Vegetation growth conditions strongly indicate coffee flowering anomalies

Vegetation growth conditions strongly indicate coffee flowering anomalies | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Vegetation growth conditions strongly indicate coffee flowering anomalies Thi Thu Thuy Nguyen, Jarrod Kath, Louis Reymondin, Thong Nguyen-Huy, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5018229/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 11 You are reading this latest preprint version Abstract Quantifying the timing of vegetation growth, particularly coffee plant flowering, is vital for estimating yield in advance. While satellite-based vegetation indices are effective in mapping crop growth and have a strong correlation with coffee yield, the potential contribution of plant conditions alongside climate factors in predicting coffee flowering anomalies remains underexplored. Here, our investigation aimed to determine whether satellite-based vegetation indices, in addition to climate variables, could enhance the model's predictive power for flowering anomalies of coffee trees. Utilizing a dataset on flowering dates over ten years of 558 coffee farms spread over four provinces (Dak Lak, Gia Lai, Dak Nong, and Lam Dong) in the Central Highlands of Vietnam, we analysed climate variables (rainfall and temperature) and the Normalized Difference Vegetation Index (NDVI) at various intervals prior to flowering dates. Using a Generalized Additive Model (GAM) and model selection based on Akaike’s Information Criteria (AIC), we identified the most influential predictors. Then, we performed Structural Equation Modelling (SEM) to further investigate the complex causal relationships among flowering anomalies, climate, vegetation, and management factors. Our results show that the NDVI prior to flowering dates held the most explanatory power, outperforming climate variables. Lower NDVI during the dormancy period indicated the ripe-to-flower condition of the coffee tree, informing earlier onset of the flowering stage, while higher NDVI during bud initiation and development stage suggested a delayed flowering. The best model incorporating both climate and NDVI predictors achieved good explanatory performance with an adjusted R 2 of 0.87. The analysis highlighted the advantages of vegetation indices over climate predictors in capturing plant conditions through its growing cycle, with the accumulated effects of environmental factors and agricultural management activities, especially during critical phenological stages. Our findings suggest further studies utilising vegetation indices from remote sensing data sources at multiple scales to thoroughly understand plant conditions at different crop growth phases, especially at early stages, for site-specific, timely and strategic management interventions. Biological sciences/Ecology/Ecological modelling Earth and environmental sciences/Ecology/Agri ecology Earth and environmental sciences/Ecology/Climate change ecology/Phenology Coffee Remote Sensing Normalized Difference Vegetation Index (NDVI) Climate Generalized Additive Model (GAM) Structural Equation Modelling (SEM) Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction Phenology, the study of the timing of biological events, plays a crucial role in estimating yield and production of crops 1 . The inclusion of phenology metrics, such as germination, flowering, and fruiting, helps improve crop yield estimation models 2,3 . Coffee is an important cash crop globally, with 168.2 million 60kg bags in coffee year 2022/2023 and a forecasted growth rate of 5.8% in coffee year 2023/2024 4 . It is well recognized for its complicated biennial phenological cycle 5–8 . Thus, the comprehension of key growth phases in coffee plants, such as flowering, fruit development, and cherry ripening, which are directly linked to the eventual crop yield, plays a pivotal role in estimating yield and production outcomes. Flowering is a key phenological phase that substantially influences crop yields. In coffee trees, this physiological process is especially critical as it determines the subsequent fruit set, providing essential insights into potential yield outcomes 9–12 . Aspects of coffee flowering phenology, including timing/onset, density, and frequency, have shown potential consequences on coffee production in several studies, as well as for other crops (Anwar et al., 2015; Schiessl et al., 2015; Tombesi et al., 2017; Chen et al., 2019). For instance, flowering density can have impact on production contributors such as plant reproductive success, pollinators (i.e. bee abundance), and pollen flow distances 13,14 . In terms of timing and frequency, uneven flowering would lead to unsynchronized fruit ripening, which would increase the cost of coffee production and poor beverage quality 10,14 . Numerous studies have found that climate factors significantly impact coffee flowering, especially temperature and precipitation 15,10,16–18 . For example, during the pre-flowering period, warm temperatures under 30 o C are important in flower bud initiation 19 . The rate of bud growth increased under high temperatures, and then a temperature drop following rain was decisive to break flower bud dormancy 15,9 . The requirement of rainfall in triggering flowering was well agreed upon in a number of studies 9,15,19 . Therefore, a controlled water deficit by irrigation was recommended to achieve uniform flowering 20,10,14 . Other climate variables, such as potential evapotranspiration and growing degree days, have also been shown to be related to coffee flowering 21,22 . More specifically, regarding the timing of flowering, the study by Kath et al. (2023) found that warm night (i.e., high minimum temperature) was a key climatic variable related to early flowering. Besides temperature and rainfall, the authors also found distinct effects of biological (i.e., tree age), and management (i.e., fertilizer) factors on yield under early and late flowering conditions. Monitoring plant growth conditions is also crucial in understanding crop phenological stages, including flowering 24–26 . Derived from satellite images, vegetation indices provide an effective means to assess crop growth conditions with global coverage. In addition to offering global coverage and allowing for large-scale analysis alongside gridded climate data, one of the key advantages of satellite-based vegetation indices is high spatial resolution. This capability provides critical and localized information for plentiful applications requiring precise and site-specific data. Moreover, high spatial resolution data with the ability to capture fine-scale variations can help improve the accuracy, avoiding potential over/under-estimation and misinterpretation of model outcomes 27,28 . Additionally, studies have shown that vegetation indices are effective at capturing vegetation responses to climate factors like temperature and precipitation 29,30 , as well as agricultural practices, particularly irrigation 31,32 . Hence, vegetation indices, representing plant conditions, can provide insights into how environmental and management variables influence plant health and growth. Vegetation indices have been successfully used to identify flowering phenology in various crops. For example, they have been applied to annual cereals such as rice and maize 33,34 , as well as to oilseed, fruit and nut crops such as rapeseed, canola, litchi, and almonds 35–39 . Particularly in almonds and canola with their distinctive white/cream and yellow flowers, the flowering periods were detected accurately in both temporal and spatial distribution 36,38 . These studies have illustrated the effectiveness of vegetation indices in determining peak flowering of oilseed rape at a temporal accuracy of 1 to 4 days 35 , detecting flowering transitions at an overall accuracy of 85% 38 . More importantly, it was found that the satellite-based model provided consistent results when compared to a field-based model at different scales 34 . Specifically in coffee, vegetation indices have been demonstrated in various studies to estimate coffee growth stages and yield. Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) have shown consistent correlations with coffee yield with varied R 2 from 0.58 to 0.9, highlighting the biennial phenological cycle of coffee and the stages of dormancy and flowering as the best phenophases to determine productivity 6,7,40 . Remotely sensed vegetation indices from satellite and aerial images have also been utilized to characterize the two-year phenological cycle of coffee in Brazil 41 and to monitor the ripening stage 42 . However, the potential of vegetation indices in predicting anomalies in the timing of flowering for coffee remains largely unexplored. Here, we investigate whether satellite-based vegetation indices can capture additional effects overlooked by climate and management factors in predicting flowering anomalies of coffee trees. The crop growth conditions during key phenological phases (i.e., flowering) provided by vegetation indices could inform management decisions and intervention activities towards season production. The findings of this study could also illustrate the effectiveness of employing high-resolution vegetation indices for agricultural research and applications that requires site-specific information. Our study examined ten years of data, focusing on the flowering dates of 558 coffee farms in the Central Highlands of Vietnam. The analysis consisted of two main analyses. Firstly, we employed the Generalized Additive Model (GAM) to identify the most effective explanatory variables, including climate variables and satellite-based vegetation indices, for predicting flowering timing anomalies in Robusta coffee trees. And secondly, we performed structural equation modelling (SEM) to further investigate the complex causal relationships among flowering anomalies, climate, vegetation, and management factors. 2. Method 2.1. Study area The study area covers four provinces: Gia Lai, Dak Lak, Dak Nong, and Lam Dong in the Central Highlands, Vietnam (Fig. 1 ). This region is the world’s leading producer of Robusta coffee, with approximately 648,400 hectares producing 26.6 million 60kg bags in the 2022/2023 coffee year. It represents 91.2% of the country's coffee-growing area and 95.2% of its total coffee production, contributing around 36% to global Robusta coffee output 43 . The Robusta coffee calendar in the study area can be divided into four key growth stages (Fig. 2 ): (i) start of the season/flower-bud initiation and development: November-December, (ii) flowering: January-March, (iii) fruit setting and development: April-August, and (iv) Maturation / Ripening / Harvest: September-November 44–47 . This study focused on the flowering date, marking the start of the flowering period from January to February. 2.2. Data 2.2.1. Coffee flowering data We analyzed a ten-year dataset, from 2008–2017, on the flowering dates of 558 coffee farms (total N = 5580) in the Central Highlands, Vietnam (Fig. 1 ). The dataset was collected within the Sustainable Management Services (SMS) program implemented by ECOM Agroindustrial Corporation in Vietnam, with records maintained for coffee certification. In the dataset, the flowering date was defined as the first flowering date. Over the study area, flowering is highly synchronized that 30–40% of plants start flowering on the same day, and all plants flower within a day or two. Management factors were also recorded in the dataset, including total volume of irrigation (liters tree − 1 ), and fertilizer (kg ha − 1 ), together with tree age, which was calculated based on planting year. For further details on the dataset and its collection, please see Byrareddy et al. (2019); Kath et al. (2020; 2023). To explore the explanatory power of climate variables and vegetation growth conditions on the occurrence of flowering, we calculated flowering anomalies over the study period for each year and each farm. 2.2.2. Vegetation growth conditions To explore vegetation growth conditions, we used the Normalized Difference Vegetation Index (NDVI), a widely-used metric for quantifying vegetation greenness using satellite data, from MODIS Terra Daily NDVI MOD09GA product in Google Earth Engine Data Catalog 49 . Daily NDVI values at 500m x 500m spatial resolution were extracted at farm locations with longitude and latitude coordinates, covering the period from 01/01/2007 to 31/12/2018. The sites selected were covered with maximum coffee plants and therefore the NDVI data represents coffee parameters and its growth cycle. Figure 2 presents the NDVI temporal profile with mean NDVI values for each day of year (DOY) over the study period, aligned with the corresponding coffee phenological stages in the study area. The NDVI dataset was joined with the coffee flowering dataset to get NDVI values for each farm at each year’s flowering start date. In cases where there was no NDVI value in a specific flowering date, the NDVI value available in the previous closest date was used. To explore the impact of NDVI values at different points in time prior to the flowering date, the rolling mean of NDVI values of 14, 30, 90, 120 and 365 days before the flowering dates were computed. These periods correspond to key phenological stages of coffee plants, especially bud initiation and development, and dormancy, which have accumulated impacts on flowering 10,50 . 2.2.3. Climate data To explore the explanatory power of climate variables in predicting coffee flowering anomalies, we used the global historical climate dataset ERA5 – Fifth generation of ECMWF atmospheric reanalyses of the global climate from Copernicus Climate Change Service 51 , accessed through Google Earth Engine Data Catalog. From the daily dataset with 27.8km spatial resolution, we extracted daily rainfall, minimum, and maximum temperature from 01/01/2008 to 31/12/2017 for each farm location with longitude and latitude coordinates. The climate dataset was joined with the coffee flowering dataset to select rainfall, minimum, and maximum temperature at the flowering start date over the study period. The accumulated rainfall and rolling mean of minimum and maximum temperature of 30 days prior to flowering dates were also computed. A summary of variables included in the study is presented in Table 1 below. For a detail statistical description of the variables, please see Table A1 . Statistical description of variables in the global model. Table 1 Summary of variables included in the study. Variables Spatial resolution Temporal resolution Data period Description Flowering anomalies N/A N/A 2008–2017 Flowering anomalies were calculated using flowering start dates for each farm each year. Vegetation variables 500m Daily 2007–2018 Daily NDVI values from MOD09GA dataset in Google Earth Engine Data Catalog. NDVI at flowering dates were extracted by farm locations. Rolling mean NDVI of 14, 30, 90, 120, and 365 days prior to flowering dates were computed. Climate variables 27.8km Daily 2007–2017 Global historical climate dataset ERA5 in Google Earth Engine Data Catalog. Rainfall, minimum temperature, and maximum temperature at flowering dates were extracted by farm locations. Accumulated rainfall and rolling means of minimum temperature, and maximum temperature of 30 days prior to flowering dates were computed. 2.3. Data analysis The overall aim of the analysis was to identify the key explanatory variables in predicting flowering anomalies. The analysis contained two major parts. Firstly, to investigate the explanatory power of climate and vegetation growth condition variables in predicting flowering anomalies, a Generalized Additive Model (GAM) in mgcv 52 and MuMIn 53 R packages, was used to fit the flowering anomalies model using a spline with consideration of spatial and temporal clusters. Secondly, to test the hypothesis that vegetation indices can add an explanation, besides climate and management, to predict flowering anomalies, we quantified the direct and indirect effects of NDVI and climate variables using Structural Equation Model (SEM) using piecewiseSEM package in R 54 . 2.3.1. Flowering anomalies modelling To identify the explanatory variables in predicting flowering anomalies, we used the Generalized Additive Model (GAM). GAM models have become increasingly popular in studying crop phenology, including coffee, due to their flexibility in capturing non-linear relationships and interactions between predictor variables. They were proven to be effective in explaining the effects of different variables in predicting agricultural crop yield, especially when working with spatio-temporal large data 55–57 . Several studies have utilized GAMs to investigate coffee phenology and its responses to climatic variability. For example, Craparo et al. (2021) found that warming nocturnal temperatures have a superseding effect on coffee ripening. Based on that, the authors have developed a Warm Night Index (WNI) that accurately predicts the start of the harvest season. Here, we used GAM to fit the flowering anomalies model using a spline with consideration of spatial and temporal clusters. All analyses were carried out in R 58 with mgcv 52 and MuMIn 53 packages. The global model was built using the following equation: y ~ f(x ij ) + (YR) + (ST) + (prov) Where flowering anomalies ( y ) were modelled as a non-linear function ( f ) of predictor variables ( x ) for each farm ( i ) and year ( j ) using a Gaussian distribution. There are two groups of predictors: climate variables (rainfall, minimum, and maximum temperature at flowering dates, accumulated rainfall and rolling mean of minimum, and maximum temperature of 30 days prior to flowering dates), and vegetation growth conditions (NDVI values at flowering dates, and rolling mean of NDVI of 14, 30, 90, 120 and 365 days prior to flowering dates). See Table A1 for details for all of the NDVI and climate variables used in the analysis. A random effect for each farm location ( ST ) in each province ( prov ) was included to account for the repeated measurements for each year ( YR ) at the farm-level. Restricted maximum likelihood (fREML) was selected as the smoothing parameter estimation method. Strongly correlated variables with a Pearson coefficient r > |0.7| were removed in model selection to avoid multi-collinearity 59 . Multi-model selection using the Akaike information criterion (AIC) was used to rank all possible combinations of predictors 53 . The one with the lowest AIC is the best model, representing the combination of the best explanatory predictors. 2.3.2. Structural Equation Model To further explore the explanatory powers of vegetation indices, in addition to climate and management factors, in predicting flowering anomalies, we utilized the Structural Equation Model (SEM) as piecewise estimation of local relationships. SEM modelling has two major characteristics over traditional regression approaches: (i) Paths among variables represent the hypothesized causal relationships, and (ii) variables can appear as both predictors and responses 54 . These features allowed us to develop models to capture the causal pathways from predictors to response variables, including both the direct and indirect effects of climate, vegetation, and management predictors on flowering anomalies. The dataset with climate and vegetation predictors of the best model from the previous analysis, together with management factors, including tree age, irrigation, and fertilizer, was fit in SEM models using the piecewiseSEM package in R 54 . To account for spatial autocorrelation, we fitted the SEMs using linear mixed models with a random effect for each site nested within each province. To account for temporal autocorrelation, we fitted SEM models using an autoregressive process of order 1. To test the causal relationships of the predictors under early- and late-flowering scenarios, we split the dataset into two separate sets of early flowering (i.e., a flowering day anomaly 0), to build two separate SEM models besides the first model with all available data. To compare the effect sizes among predictors and between models, coefficients were standardized [mean(x)/1.SD(x)] 60 . Finally, the direct, indirect and total effects of each predictor on flowering anomalies were calculated. Direct effects are the standardized regression coefficients of each predictor on the response variable. Indirect effects are the product of the direct effects operating along causal pathways in SEM models. The total effects are the sum of the predictor’s direct effect and all its indirect effects through intermediary variables. 3. Results 3.1. Vegetation growth conditions strongly indicate coffee flowering anomalies with greater explanatory power than climate predictors. Vegetation growth conditions strongly indicate coffee flowering anomalies. The best model included the climate predictors (rainfall at flowering dates, accumulated rainfall 30 days prior to flowering dates, and rolling mean of minimum, and maximum temperature of 30 days prior to flowering dates) and vegetation predictors (NDVI at flowering dates, rolling mean NDVI of 30, 90, and 365 days prior to flowering dates). For a detailed statistical description of all variables in the best model, please see Table A1 . Statistical description of variables in the global model.. The marginal effects of each predictor in the best model on flowering anomalies are presented in Fig. 3 . The result showed that NDVI in a month and three months prior to flowering dates have the most explanatory power for flowering anomalies. Specifically, 30-day pre-flowering NDVI (Fig. 3 f) of approximately less than 0.3 had a strong positive linear relationship with flowering anomalies, particularly early flowering. This indicates coffee trees that are under less green conditions at the end of the dormancy stage (i.e., a month prior to flowering dates) expect early flowering. In contrast, 90-day pre-flowering mean NDVI (Fig. 3 g) of approximately less than 0.3 had a strong negative linear relationship with flowering anomalies, particularly late flowering. In this case, coffee trees with less greenness at the start of the season, when flower buds initiate and develop (i.e., three months prior to flowering dates), are likely to flower late. On the other hand, NDVI at the flowering date (Fig. 3 e) had less effect on flowering anomalies, while NDVI in a year prior to the flowering date (Fig. 3 h) demonstrated a slight non-linear relationship with flowering anomalies. The best model, consisting of both climate and vegetation predictors, has an adjusted R 2 of 0.865 and can explain 86.7% of deviance (Table 2 ). The explanatory power of vegetation predictors (84.7% of deviance) was outstanding compared to that of climate variables (77.7% of deviance). The AIC values also show a higher rank for the model with vegetation predictors than the one with climate predictors. Table 2 Performance of the best model and models with climate and vegetation variables only Model Adj. R2 Deviance explained AIC Best model * 0.865 86.7% 24082.48 Best model with climate predictors only 0.775 77.7% 26914.98 Best model with vegetation predictors only 0.846 84.7% 24807.86 *: The best model based on AIC ranking includes climate predictors (rainfall at flowering dates, accumulated rainfall and rolling mean minimum temperature, and maximum temperature of 30 days prior to flowering date), and vegetation predictors (NDVI at flowering date, and mean NDVI of 30, 90, and 365 days prior to flowering date). The marginal effects of each predictor in the best model on flowering anomalies were presented in Fig. 3 . A detailed statistical description of all variables in the best model is presented in Table A1 in the Annex. 3.2. Vegetation index captures additional aspects besides climate and management factors The SEM results quantified the casual relationships, and the indirect and direct effects of NDVI, climate and management factors on flowering anomalies (Fig. 4 , Table 3 , and Table 4 ). Specifically, 30-day pre-flowering NDVI has the strongest direct positive effect on flowering anomalies in both models with all data (standardized regression coefficient = 0.787) and early flowering scenario (standardized regression coefficient = 1.623) (Fig. 4 a, b). The 90-day pre-flowering NDVI value also shows a strong direct negative effect on flowering anomalies in all three SEM models with all data (standardized regression coefficient = -0.624), early- (standardized regression coefficient = -0.925), and late-flowering scenarios (standardized regression coefficient = -0.346). Mean NDVI of a year prior to flowering dates also showed a significant negative effect on flowering anomalies in two SEM models with all data (standardized regression coefficient = -0.409) and early flowering scenario (standardized regression coefficient = -0.491). These results are consistent with GAM outputs from the previous analysis. During the bud initiation and development phase (i.e., 90 days prior to flowering date), higher NDVI values suggest an earlier onset of flowering stage, while lower NDVI values indicate a delay in the flowering. This phase corresponds to a period of active growth of coffee trees to prepare for the flowering (Fig. 2 , Table 3 ). Therefore, a high NDVI value associated with a vigorous condition suggests coffee plants are progressing well toward flowering, and an earlier onset is likely to occur. After two to three months of bud development, coffee plants enter the dormancy stage (i.e., 30 days prior to the flowering date), characterized by reduced metabolic activity, reflected in a slower rise of NDVI values. In this case, a lower 30-day pre-flowering NDVI showing a more ready, ripe-to-flower condition indicates that flowering start dates will arrive earlier (Fig. 2 , Table 3 ). SEM model results also highlight the significant effects of NDVI a year prior to the flowering date (Fig. 4 , Table 4 ). A lower NDVI a year before indicates a later flowering date the following year. Table 3 Key vegetation index indicators to coffee flowering anomaly Phenology stage and coffee plant ecophysiology Vegetation index indicator and flowering anomaly Reference Bud initiation and development 90-day pre-flowering NDVI Coffee plants are in an active growing stage, characterized by increasing chlorophyll production and photosynthetic efficiency. During this stage, the NDVI values are in an increasing trend, indicating vigorous and healthy plants. Therefore, higher NDVI values indicate that coffee plants are progressing well toward flowering, and an earlier onset is likely to happen. Wintgens, 2004, page 12 Lower NDVI values indicate that coffee plants need more time for development and are at the earlier stage of bud initiation. Therefore, the condition suggests a delay in flowering. Dormancy 30-day pre-flowering NDVI Coffee plants are quiescence, characterized by reduced metabolic activity. During the dormancy stage, the rise of NDVI values is substantially reduced. Hence, a lower 30-day pre-flowering NDVI value, showing a more ready, ripe-to-flower condition, indicates an earlier onset of the flowering stage. Wintgens, 2004, page 12 Regarding the total effects of the predictors (Fig. 4 and Table 4 ), in all three SEM models, the 30-day pre-flowering NDVI is within the top three predictors with the highest total effects on flowering anomalies. Similarly, the 90-day pre-flowering NDVI is within the top three predictors with the highest total effects on flowering anomalies in SEM models with all data and late flowering scenarios. With the additional indirect effects captured through NDVI, the 30-day pre-flowering maximum temperature has the highest total effect on flowering anomalies in the SEM model with all data. In contrast, 30-day pre-flowering minimum and maximum temperature, together with 30-day pre-flowering NDVI, are the top three predictors with the highest total effects on flowering anomalies in SEM early flowering model. Besides the direct effects of climate predictors on flowering anomalies, their indirect effects are captured by vegetation predictors in SEM models (Fig. 4 and Table 4 ). Specifically in the early flowering model, the indirect effects of 30-day pre-flowering rainfall and maximum temperature captured through 30-day pre-flowering NDVI, are two to four times larger than their direct effects. Additionally, the indirect effect of 30-day pre-flowering minimum temperature, captured through NDVI during the same period, is almost equal to the direct effect. In the late flowering model, the direct effects of 30-day pre-flowering rainfall and maximum temperature are not statistically significant. However, their total effects on flowering anomalies are statistically significant captured through the impact pathway of NDVI of the same period. Table A1 Statistical description of variables in the global model. Variables Min 1st Qu. Mean 3rd Qu. Max Flowering anomalies (day) -15.2 0.0 0.0 3.6 7.9 Vegetation index at different periods prior to flowering date Flowering NDVI 0.0003 0.2091 0.4518 0.6590 0.9711 14 days pre-flowering mean NDVI 0.0294 0.3075 0.4020 0.4975 0.7640 30 days pre-flowering mean NDVI 0.0440 0.2927 0.3814 0.4738 0.7180 90 days pre-flowering mean NDVI 0.1190 0.2995 0.3791 0.4573 0.6300 120 days pre-flowering mean NDVI 0.1283 0.2845 0.3526 0.4156 0.5991 365 days pre-flowering mean NDVI 0.1935 0.3079 0.3422 0.3756 0.5032 Climate variables Flowering rainfall (mm) 3.731 32.373 89.477 124.709 442.541 Flowering minimum temperature ( o C) 13.76 16.53 17.65 18.65 22.02 Flowering maximum temperature ( o C) 24.77 27.19 28.36 29.35 33.01 30 days pre-flowering rainfall (mm) 0.004 3.135 23.135 23.827 206.814 30 days pre-flowering minimum temperature ( o C) 13.58 17.00 18.04 19.01 22.16 30 days pre-flowering maximum temperature ( o C) 23.57 27.20 28.58 29.96 34.60 Table 4 Direct, indirect, and total effects of climate, vegetation, and management factors on flowering anomalies. “n.s.” indicates the effect is not statistically significant. The total effects in bold are the top three predictors with the highest total effects on flowering anomalies. Direct effects are the standardized regression coefficients of each predictor on the response variable. Indirect effects are the product of the direct effects operating along causal pathways in SEM models. The total effects are the sum of the predictor’s direct effect and all its indirect effects through intermediary variables. SEM models Predictor Direct effect Indirect effect Total effect ALL DATA Flowering rainfall 0.328 0.003 0.331 30-day pre-flowering rainfall -0.170 -0.071 -0.241 30-day pre-flowering tmin -0.465 -0.154 -0.619 30-day pre-flowering tmax 0.497 0.332 0.829 Total irrigation 0.235 0.033 0.268 Total fertilizer -0.002 -0.139 -0.141 Tree age -0.049 n.s. -0.049 Flowering NDVI 0.046 - 0.046 30-day pre-flowering NDVI 0.787 - 0.787 90-day pre-flowering NDVI -0.624 - -0.624 365-day pre-flowering NDVI -0.409 - -0.409 EARLY FLOWERING Flowering rainfall n.s. n.s. n.s. 30-day pre-flowering rainfall -0.094 -0.399 -0.493 30-day pre-flowering tmin -0.902 -0.790 -1.692 30-day pre-flowering tmax 0.599 1.461 2.060 Total irrigation -0.837 -0.661 -1.498 Total fertilizer n.s. n.s. n.s. Tree age -0.127 -0.078 -0.127 Flowering NDVI n.s. - n.s. 30-day pre-flowering NDVI 1.623 - 1.623 90-day pre-flowering NDVI -0.925 - -0.925 365-day pre-flowering NDVI -0.491 - -0.491 LATE FLOWERING Flowering rainfall -0.069 n.s. -0.069 30-day pre-flowering rainfall n.s. 0.040 0.040 30-day pre-flowering tmin 0.096 -0.153 -0.057 30-day pre-flowering tmax n.s. 0.161 0.161 Total irrigation 0.203 n.s. 0.203 Total fertilizer n.s. n.s. n.s. Tree age n.s. n.s. n.s. Flowering NDVI n.s. - n.s. 30-day pre-flowering NDVI 0.238 - 0.238 90-day pre-flowering NDVI -0.346 - -0.346 365-day pre-flowering NDVI 0.100 - 0.100 SEM model results also show a strong effect of irrigation, especially in the early flowering SEM model. In the early flowering model, the indirect effects of irrigation on flowering anomalies, captured through NDVI during flowering and pre-flowering stage, are closely equivalent to its direct effect. As a result, the total effect of irrigation on flowering anomalies is almost double the direct effect alone. In contrast, total fertilizer and tree age show much less effects on flowering anomalies, most of which are not statistically significant (Fig. 4 , Table 3 ). 4. Discussion Flowering phenology plays a vital role in coffee production. Deviations from the expected flowering patterns can have significant implications for crop yield and quality 9–12 . However, while the impacts of climate variables on coffee flowering have been extensively studied, how vegetation indices retrieved from satellite images can contribute to predicting coffee flowering anomalies remain underexplored. Using an extensive 10-year dataset (N = 5580) of Robusta coffee flowering dates, climate predictors (rainfall and temperature), and NDVI values at different periods prior to flowering dates, we found that vegetation growth conditions strongly indicate coffee flowering anomalies. Specifically, coffee plant conditions, represented by NDVI values, of a month and three months prior to flowering dates, corresponding to dormancy and bud growth stages, were identified to have the most explanatory power to predict the flowering timing anomalies, outperforming climate variables. The 30-day pre-flowering NDVI values had a strong positive relationship with the flowering anomalies, particularly early flowering. On the other hand, the 90-day pre-flowering NDVI values, reflecting plant conditions during bud initiation and development stage, indicated flowering anomalies in both early and late scenarios. 4.1. Vegetation growth conditions strongly indicate coffee flowering anomalies. Most published studies on coffee flowering phenology have been focused on the influences of the external environment on the plant, such as temperature, rainfall, soil moisture, etc., as triggers to flowering 23,61 . However, coffee flowering is a complex sequence of flower bud initiation, dormancy development, dormancy breakage, stimulation of regrowth, and regrowth to anthesis, which is the flower blooming 9,10 . Throughout the production sequence, crop conditions during the period of quiescence did not receive much attention. In this study, the plant’s internal conditions necessary for flowering, or in other words, the ripe-to-flower conditions, were detected by a vegetation index, which serves as an indicator of plant conditions. The analysis result showed that a lower 30-day pre-flowering NDVI value is a strong indicator for early flowering of the upcoming period. The finding is consistent with current literature on conditions to induce flowering of coffee plants that they need a pre-flowering dry period, characterized by low rainfall and high temperature, to stimulate flowering 9,10,62,18 . During the dormancy stage of approximately 30 days prior to flowering dates, coffee plants are in quiescent conditions, which are characterized by reduced metabolic activity, represented by slower rise of NDVI values comparing to the bud development phase 9,10,50 . At the end of dormancy stage, watering, either through rainfall or irrigation, will stimulate flowering, coffee plants regrow with vigorous conditions, represented by the peak of NDVI values in the temporal profile 10,50 . Furthermore, the conditions of coffee plants during the flower bud initiation and development stage, approximately 90 days prior to flowering dates, were also captured by NDVI. This active growing period, characterized by increasing chlorophyll production and photosynthetic efficiency, was presented in an increasing trend of NDVI values 9,10 . Therefore, higher NDVI values indicate that coffee plants are progressing well toward flowering, and an earlier onset is likely to happen. While lower NDVI values in this period indicate that coffee plants are at the earlier stage of bud initiation, suggesting a delay in flowering. Thus, our findings suggested that NDVI is an effective indicator for detecting shifts in the flowering period to either earlier or later than normal, or in other words, predict the anomaly of the upcoming flowering of coffee trees. 4.2. Vegetation growth conditions indicate stronger explanatory power, capturing additional aspects besides climate and management. The impact of climate factors on predicting coffee growth has been extensively studied. Temperature and rainfall are widely recognized as key determinants affecting the various stages of coffee plant development 15,10,16,18 . For flowering in particular, optimal temperatures and consistent rainfall patterns are crucial for initiating and sustaining the flowering process. However, we found that satellite-based vegetation index can capture additional aspects besides climate and management, which improves the model ability to predict flowering anomalies. Besides, the model using only vegetation predictors outperforms the one relying solely on climate variables. Even though the particular role of vegetation indices in predicting coffee flowering anomalies is still underexplored, several studies have demonstrated the effectiveness of vegetation indices retrieved from satellite images in predicting coffee yield 63 . Models using vegetation indices, such as NDVI and LAI, achieved higher accuracy than those relying solely on climate data 47 . Moreover, a study by Bolaños et al. (2023) focusing on flowering, the early stages of the coffee production cycle to predict yield, found that NDVI was among the top three predictors having the highest correlation with yield. Still, further research is needed to thoroughly investigate the specific role of satellite-based vegetation indices in predicting flowering phenology, and other key growth phases of coffee in general, ultimately leading to more accurate yield estimation. A notable aspect is the ability of vegetation indices to detect the influence of crop conditions from early stages at the start of the season on flowering anomalies. The healthy and vigorous conditions of the plants are captured by the vegetation index, indicating the position at different progressing steps within the phenological stages. This finding can be attributed to their capability to reflect the cumulative effects of environmental conditions, emphasizing the longer-term impacts of climate on plant conditions and productivity 65,66 . On the other hand, climate factors, such as temperature and solar radiation, have relatively short cumulative effects on vegetation growth of various tree-type vegetation 67 . Moreover, the influence of plant conditions in a year prior to the flowering date on the flowering anomalies of the following year is consistent with the biennial nature of coffee production. This suggests that resources, such as nutrients, of the plant are accumulated during the vegetative stage in the previous year towards the production stage in the following year 68,69 . The effectiveness of nutrient accumulation through various agricultural management practices, such as irrigation, fertilization, and pest control, is illustrated through changes in plant growth and canopy density. These changes can be measured using vegetation indices 70,71 , providing substantial information in predicting flowering anomalies. However, further study might be needed to separate the accumulated impacts of environmental conditions and agricultural management activities on the coffee production cycle, providing site-specific information for decision-making. Indirect effects of climate predictors, as captured through vegetation indices are strong, especially in the early flowering SEM model. Additionally, vegetation indices also pick up the indirect effects of management factors, particularly irrigation in early flowering SEM model, and fertilizer in all data SEM model. Notably, in the late flowering SEM model, the impacts of 30-day rainfall and maximum temperature can only be captured through the pathway of vegetation indices with their direct effects statistically insignificant. This highlights the additional insights that vegetation indices can provide beyond climate and management variables in predicting flowering anomalies of coffee trees. These results are also aligned with the findings of other published studies that vegetation indices can capture vegetation responses to climate factors (i.e., temperature and precipitation) 29,30 , and agricultural management activities (i.e., irrigation) 31,32 . Additionally, the direct effects of vegetation indices, which represent their total impacts on flowering anomalies, are consistently among the top three predictors in all three SEM models. This further explains the superior performance of vegetation-only model in GAM, where it outperformed the one relying solely on climate predictors. 4.3. Spatial variations and the effectiveness of remote sensing data This study showed a promising result when using the vegetation index retrieved from satellite data with the most popular and widely used index, such as NDVI. Additionally, other vegetation indices, such as Enhanced Vegetation Index (EVI), Leaf Area Index (LAI), Vegetation Health Index (VHI), and Vegetation Condition Index (VCI), have also shown their effectiveness in mapping crop phenology 72,73,35,74 . There are other indices developed specifically for flowering phenology, such as the Enhanced Bloom Index (EBI) 36 . However, further studies are still needed to explore more data sources and information to contribute to exploring other key growth stages within coffee complex phenology. Additionally, satellite-based vegetation indices offer several key advantages over climate data. Firstly, the detailed and site-specific information provided by high-resolution satellite images allows for more accurate modelling, as it captures fine-scale variations in vegetation that coarse resolution data may overlook and lead to potential overestimations or inaccuracies in model outcomes 27,28 . Secondly, satellite data can provide comprehensive coverage, which is particularly beneficial for multi-scale analyses. Even though gridded reanalysis climate data can also offer global coverage, the downscaled dataset often carry uncertainties, especially in remote areas where ground-based station data is sparse or unavailable due to low-density networks 75,76 . And thirdly, many satellite datasets, such as MODIS, Landsat, or Sentinel, are freely available, making them a cost-effective alternative to purchasing station-based data, which can be expensive and limited in coverage. These factors make satellite-based vegetation indices a promising data sources agricultural analyses and applications. Even though a medium resolution (i.e., 500m) of satellite images offers substantially better spatial resolution compared to gridded climate data, using MODIS dataset is still a limitation of this study for the trade-off of better temporal resolution (i.e., daily) to capture crop phenology. Moreover, with medium spatial resolution, NDVI values in each pixel can be influenced by external factors such as soil background, and shadows. With recent advances in remote sensing techniques the fusion of data from multiple satellite sensors with high spatial resolution, such as freely available Landsat and Sentinel or other commercial satellites, further studies are needed to illustrate the effectiveness of high-resolution satellite images to overcome the limitations of medium spatial resolution. Furthermore, it could be beneficial to explore other remote sensing data than satellite-based, such as drones and UAVs, which have also illustrated the effectiveness of employing very high-resolution datasets to monitor crop phenology at different scales and provide more insights into the spatial discrepancy at local levels 42,77,78 . Despite being the most popular vegetation index utilized in agricultural research and applications, NDVI primarily captures the greenness of vegetation, making it less sensitive to non-photosynthetic elements (Liu et al., 2022), such as cherries or flowers, which are important indicators of coffee phenology. Studies have found other vegetation indices that demonstrated potential relations with coffee phenology and yield, such as Leaf Area Index (LAI) 80 , EVI 6 , Soil-Adjusted Vegetation Index (SAVI), and Normalized Difference Water Index (NDWI) 40 . Thus, further studies are needed to investigate those alternatives, or even to develop a vegetation index specifically tailored for coffee as the existing vegetation indices are not produced to capture identical parameters of the crop 42 . 5. Conclusion Our analysis indicates that vegetation growth conditions provide robust indicators of coffee flowering anomalies, highlighting its advantage over climate variables. Specifically, we found that the 30-day pre-flowering NDVI values had a strong positive relationship with flowering anomalies, particularly early flowering. Additionally, the 90-day pre-flowering NDVI values, reflecting plant conditions during bud initiation and development stage, indicated flowering anomalies in both early and late scenarios. The analysis also highlighted the advantages of vegetation indices over climate predictors to capture plant conditions through its growing cycle with accumulated effects of environmental factors and agricultural management activities, especially during the vegetative stage in biennial phenology. The findings also highlight the ability of vegetation indices in capturing plant growth conditions during crucial phenological stages at the start of the crop season, providing valuable insights for management decisions and interventions to optimize seasonal production. The results of this study further demonstrate the effectiveness of employing high-resolution vegetation indices for agricultural research and applications the requires site-specific information, especially in monitoring crop growth conditions in complex landscapes with local-level spatial variations. Declarations Author Contribution T.N. conceived the initial study based on conversations with J.K. and L.R.; T.N. and V.M.B. collected the original data; T.N. processed the data and conducted the data analysis; T.N. prepared the initial manuscript; J.K., L.R., T.N.-H., V.M.B., and S.M. reviewed and edited the manuscript. All authors reviewed the manuscript and gave their approval for publication. Acknowledgement We acknowledge the funding from the German Federal Ministry for Economic Affairs and Climate Action (BMWK) through the International Climate Initiative (IKI). References 1. Wielgolaski, F.-E. Phenology in Agriculture. in Phenology and Seasonality Modeling (ed. Lieth, H.) vol. 8 369–381 (Springer Berlin Heidelberg, Berlin, Heidelberg, 1974). 2. Bolton, D. K. & Friedl, M. A. Forecasting crop yield using remotely sensed vegetation indices and crop phenology metrics. Agricultural and Forest Meteorology 173 , 74–84 (2013). 3. Sakamoto, T., Gitelson, A. A. & Arkebauer, T. J. MODIS-based corn grain yield estimation model incorporating crop phenology information. Remote Sensing of Environment 131 , 215–231 (2013). 4. International Coffee Organization. Coffee Report and Outlook . 43 https://icocoffee.org/documents/cy2023-24/Coffee_Report_and_Outlook_December_2023_ICO.pdf (2023). 5. Camargo, Â. P. D. & Camargo, M. B. P. D. Definição e esquematização das fases fenológicas do cafeeiro arábica nas condições tropicais do Brasil. Bragantia 60 , 65–68 (2001). 6. Bernardes, T., Moreira, M. A., Adami, M., Giarolla, A. & Rudorff, B. F. T. Monitoring Biennial Bearing Effect on Coffee Yield Using MODIS Remote Sensing Imagery. Remote Sensing 4 , 2492–2509 (2012). 7. Brunsell, N. A., Pontes, P. P. B. & Lamparelli, R. A. C. Remotely Sensed Phenology of Coffee and Its Relationship to Yield. GIScience & Remote Sensing 46 , 289–304 (2009). 8. De Oliveira Aparecido, L. E., De Souza Rolim, G., Camargo Lamparelli, R. A., De Souza, P. S. & Dos Santos, E. R. Agrometeorological Models for Forecasting Coffee Yield. Agronomy Journal 109 , 249–258 (2017). 9. Cannell, M. G. R. Physiology of the Coffee Crop. in Coffee (eds. Clifford, M. N. & Willson, K. C.) 108–134 (Springer US, Boston, MA, 1985). doi:10.1007/978-1-4615-6657-1_5. 10. DaMatta, F. M., Ronchi, C. P., Maestri, M. & Barros, R. S. Ecophysiology of coffee growth and production. Braz. J. Plant Physiol. 19 , 485–510 (2007). 11. DaMatta, F. M., Avila, R. T., Cardoso, A. A., Martins, S. C. V. & Ramalho, J. C. Physiological and Agronomic Performance of the Coffee Crop in the Context of Climate Change and Global Warming: A Review. J. Agric. Food Chem. 66 , 5264–5274 (2018). 12. Dos Santos Soares, L., Teruel Rezende, T., Beijo, L. A. & Silva Franco Júnior, K. Interaction between climate, flowering and production of dry coffee (Coffea arabica L.) in Minas Gerais. CS 16 , 1–10 (2021). 13. Peters, V. E. & Carroll, C. R. Temporal variation in coffee flowering may influence the effects of bee species richness and abundance on coffee production. Agroforest Syst 85 , 95–103 (2012). 14. Boreux, V., Kushalappa, C. G., Vaast, P. & Ghazoul, J. Interactive effects among ecosystem services and management practices on crop production: Pollination in coffee agroforestry systems. Proc. Natl. Acad. Sci. U.S.A. 110 , 8387–8392 (2013). 15. De T. Alvim, P. Factors Affecting Flowering of Coffee. in Genes, Enzymes, and Populations (ed. Srb, A. M.) 193–202 (Springer US, Boston, MA, 1973). doi:10.1007/978-1-4684-2880-3_13. 16. Craparo, A. C. W., Van Asten, P. J. A., Läderach, P., Jassogne, L. T. P. & Grab, S. W. Coffea arabica yields decline in Tanzania due to climate change: Global implications. Agricultural and Forest Meteorology 207 , 1–10 (2015). 17. Craparo, A. C. W., Van Asten, P. J. A., Läderach, P., Jassogne, L. T. P. & Grab, S. W. Warm nights drive Coffea arabica ripening in Tanzania. Int J Biometeorol 65 , 181–192 (2021). 18. Kath, J. et al. Not so robust: Robusta coffee production is highly sensitive to temperature. Global Change Biology 26 , 3677–3688 (2020). 19. DaMatta, F. M. & Ramalho, J. D. C. Impacts of drought and temperature stress on coffee physiology and production: a review. Braz. J. Plant Physiol. 18 , 55–81 (2006). 20. Carr, M. K. V. THE WATER RELATIONS AND IRRIGATION REQUIREMENTS OF COFFEE. Ex. Agric. 37 , 1–36 (2001). 21. Zacharias, A. O., Camargo, M. B. P. D. & Fazuoli, L. C. Modelo agrometeorológico de estimativa do início da florada plena do cafeeiro. Bragantia 67 , 249–256 (2008). 22. Pezzopane, J. R. M., Salva, T. D. J. G., De Lima, V. B. & Fazuoli, L. C. Agrometeorological parameters for prediction of the maturation period of Arabica coffee cultivars. Int J Biometeorol 56 , 843–851 (2012). 23. Kath, J., Byrareddy, V. M., Reardon-Smith, K. & Mushtaq, S. Early flowering changes robusta coffee yield responses to climate stress and management. Science of The Total Environment 856 , 158836 (2023). 24. De Beurs, K. M. & Henebry, G. M. Land surface phenology, climatic variation, and institutional change: Analyzing agricultural land cover change in Kazakhstan. Remote Sensing of Environment 89 , 497–509 (2004). 25. Reed, B. C., Schwartz, M. D. & Xiao, X. Remote Sensing Phenology. in Phenology of Ecosystem Processes (ed. Noormets, A.) 231–246 (Springer New York, New York, NY, 2009). doi:10.1007/978-1-4419-0026-5_10. 26. Hatfield, J. L. et al. Applications of Vegetative Indices from Remote Sensing to Agriculture: Past and Future. Inventions 4 , 71 (2019). 27. Dey, A. & Remesan, R. Assessing the Impact of Spatial Resolution on Land Surface Model Based on Hydrologic Simulations. in Climate Change Impacts on Water Resources (eds. Jha, R., Singh, V. P., Singh, V., Roy, L. B. & Thendiyath, R.) vol. 98 493–501 (Springer International Publishing, Cham, 2021). 28. Tian, J., Zhu, X., Wu, J., Shen, M. & Chen, J. Coarse-Resolution Satellite Images Overestimate Urbanization Effects on Vegetation Spring Phenology. Remote Sensing 12 , 117 (2020). 29. Wang, H. et al. Spatiotemporal crop NDVI responses to climatic factors in mainland China. International Journal of Remote Sensing 40 , 89–103 (2019). 30. Zhang, H. et al. NDVI dynamic changes and their relationship with meteorological factors and soil moisture. Environ Earth Sci 77 , 582 (2018). 31. Grados, D., Reynarfaje, X. & Schrevens, E. A methodological approach to assess canopy NDVI–based tomato dynamics under irrigation treatments. Agricultural Water Management 240 , 106208 (2020). 32. Maselli, F. et al. An improved NDVI-based method to predict actual evapotranspiration of irrigated grasses and crops. Agricultural Water Management 233 , 106077 (2020). 33. Wu, G., Miller, N. D., De Leon, N., Kaeppler, S. M. & Spalding, E. P. Predicting Zea mays Flowering Time, Yield, and Kernel Dimensions by Analyzing Aerial Images. Front. Plant Sci. 10 , 1251 (2019). 34. Zhang, Z. et al. Dynamic variability of the heading–flowering stages of single rice in China based on field observations and NDVI estimations. Int J Biometeorol 59 , 643–655 (2015). 35. d’Andrimont, R. et al. Detecting flowering phenology in oil seed rape parcels with Sentinel-1 and − 2 time series. Remote Sensing of Environment 239 , 111660 (2020). 36. Chen, B., Jin, Y. & Brown, P. An enhanced bloom index for quantifying floral phenology using multi-scale remote sensing observations. ISPRS Journal of Photogrammetry and Remote Sensing 156 , 108–120 (2019). 37. Lin, P. et al. A novel approach for estimating the flowering rate of litchi based on deep learning and UAV images. Front. Plant Sci. 13 , 966639 (2022). 38. Sulik, J. J. & Long, D. S. Automated detection of phenological transitions for yellow flowering plants such as Brassica oilseeds. Agrosystems Geosci & Env 3 , e20125 (2020). 39. Zang, Y. et al. Remote Sensing Index for Mapping Canola Flowers Using MODIS Data. Remote Sensing 12 , 3912 (2020). 40. Nogueira, S. M. C., Moreira, M. A. & Volpato, M. M. L. RELATIONSHIP BETWEEN COFFEE CROP PRODUCTIVITY AND VEGETATION INDEXES DERIVED FROM OLI / LANDSAT-8 SENSOR DATA WITH AND WITHOUT TOPOGRAPHIC CORRECTION. Eng. Agríc. 38 , 387–394 (2018). 41. Júnior, A. F. C., Júnior, O. A. D. C., Martins, É. D. S. & Guerra, A. F. PHENOLOGICAL CHARACTERIZATION OF COFFEE CROP (Coffea arabica L.) FROM MODIS TIME SERIES. Rev. Bras. Geof. 31 , 569 (2013). 42. Nogueira Martins, R., De Carvalho Pinto, F. D. A., Marçal De Queiroz, D., Magalhães Valente, D. S. & Fim Rosas, J. T. A Novel Vegetation Index for Coffee Ripeness Monitoring Using Aerial Imagery. Remote Sensing 13 , 263 (2021). 43. USDA. USDA Production, Supply and Distribution. https://apps.fas.usda.gov/psdonline/app/index.html#/app/downloads. 44. Amarasinghe, U. A., Hoanh, C. T., D’haeze, D. & Hung, T. Q. Toward sustainable coffee production in Vietnam: More coffee with less water. Agricultural Systems 136 , 96–105 (2015). 45. Dinh, T. L. A., Aires, F. & Rahn, E. Statistical Analysis of the Weather Impact on Robusta Coffee Yield in Vietnam. Front. Environ. Sci. 10 , 820916 (2022). 46. Kouadio, L. et al. Performance of a process-based model for predicting robusta coffee yield at the regional scale in Vietnam. Ecological Modelling 443 , 109469 (2021). 47. Thao, N. T. T. et al. Early Prediction of Coffee Yield in the Central Highlands of Vietnam Using a Statistical Approach and Satellite Remote Sensing Vegetation Biophysical Variables. Remote Sensing 14 , 2975 (2022). 48. Byrareddy, V., Kouadio, L., Mushtaq, S. & Stone, R. Sustainable Production of Robusta Coffee under a Changing Climate: A 10-Year Monitoring of Fertilizer Management in Coffee Farms in Vietnam and Indonesia. Agronomy 9 , 499 (2019). 49. MODIS Terra Daily NDVI | Earth Engine Data Catalog. Google for Developers https://developers.google.com/earth-engine/datasets/catalog/MODIS_MOD09GA_006_NDVI. 50. Coffee: Growing, Processing, Sustainable Production: A Guidebook for Growers, Processors, Traders, and Researchers . (Wiley, 2004). doi:10.1002/9783527619627. 51. Copernicus Climate Change Service (C3S). ERA5: Fifth generation of ECMWF atmospheric reanalyses of the global climate. https://cds.climate.copernicus.eu/cdsapp#!/home (2017). 52. Wood, S. N. Fast Stable Restricted Maximum Likelihood and Marginal Likelihood Estimation of Semiparametric Generalized Linear Models. Journal of the Royal Statistical Society Series B: Statistical Methodology 73 , 3–36 (2011). 53. BARTON, K. MuMIn : multi-model inference. http://r-forge.r-project.org/projects/mumin/ (2009). 54. Lefcheck, J. S. piecewiseSEM : Piecewise structural equation modelling in r for ecology, evolution, and systematics. Methods Ecol Evol 7 , 573–579 (2016). 55. Ravindra, K., Rattan, P., Mor, S. & Aggarwal, A. N. Generalized additive models: Building evidence of air pollution, climate change and human health. Environment International 132 , 104987 (2019). 56. Wellington, M. J., Lawes, R. & Kuhnert, P. A framework for modelling spatio-temporal trends in crop production using generalised additive models. Computers and Electronics in Agriculture 212 , 108111 (2023). 57. Wikle, C. K., Zammit-Mangion, A. & Cressie, N. Spatio-Temporal Statistics with R . (Chapman and Hall/CRC, Boca Raton, Florida : CRC Press, [2019], 2019). doi:10.1201/9781351769723. 58. R: The R Project for Statistical Computing. https://www.r-project.org/. 59. Dormann, C. F. et al. Collinearity: a review of methods to deal with it and a simulation study evaluating their performance. Ecography 36 , 27–46 (2013). 60. Gelman, A. & Hill, J. Data Analysis Using Regression and Multilevel/Hierarchical Models . (Cambridge University Press, 2006). doi:10.1017/CBO9780511790942. 61. Gomez, C. et al. Shift in precipitation regime promotes interspecific hybridization of introduced Coffea species. Ecology and Evolution 6 , 3240–3255 (2016). 62. Byrareddy, V. et al. Win-win: Improved irrigation management saves water and increases yield for robusta coffee farms in Vietnam. Agricultural Water Management 241 , 106350 (2020). 63. Abreu Júnior, C. A. M. D. et al. Estimating Coffee Plant Yield Based on Multispectral Images and Machine Learning Models. Agronomy 12 , 3195 (2022). 64. Bolaños, J., Corrales, J. C. & Campo, L. V. Feasibility of Early Yield Prediction per Coffee Tree Based on Multispectral Aerial Imagery: Case of Arabica Coffee Crops in Cauca-Colombia. Remote Sensing 15 , 282 (2023). 65. Pettorelli, N. et al. Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends in Ecology & Evolution 20 , 503–510 (2005). 66. Feng, J. et al. Temporal and Spatial Variation Characteristicsof NDVI and Its Relationshipwith Environmental Factors in Huangshui RiverBasin from 2000 to 2018. Pol. J. Environ. Stud. 30 , 3043–3063 (2021). 67. Du, G., Yan, S., Chen, H., Yang, J. & Wen, Y. Intra-Annual Cumulative Effects and Mechanisms of Climatic Factors on Global Vegetation Biomes’ Growth. Remote Sensing 16 , 779 (2024). 68. Salamanca-Jimenez, A., Doane, T. A. & Horwath, W. R. Nitrogen Use Efficiency of Coffee at the Vegetative Stage as Influenced by Fertilizer Application Method. Front. Plant Sci. 8 , (2017). 69. Vilela, M. S. et al. Nitrogen, phosphorus, and potassium fertilization on the incidence of brown eye spot in coffee crop in vegetative stage. Trop. plant pathol. 47 , 672–684 (2022). 70. Ennouri, K., Triki, M. A. & Kallel, A. Applications of Remote Sensing in Pest Monitoring and Crop Management. in Bioeconomy for Sustainable Development (ed. Keswani, C.) 65–77 (Springer Singapore, Singapore, 2020). doi:10.1007/978-981-13-9431-7_5. 71. Pinter, Jr., P. J. et al. Remote Sensing for Crop Management. photogramm eng remote sensing 69 , 647–664 (2003). 72. Jin, H. & Eklundh, L. A physically based vegetation index for improved monitoring of plant phenology. Remote Sensing of Environment 152 , 512–525 (2014). 73. Araya, S., Ostendorf, B., Lyle, G. & Lewis, M. CropPhenology: An R package for extracting crop phenology from time series remotely sensed vegetation index imagery. Ecological Informatics 46 , 45–56 (2018). 74. Liu, L. et al. Detecting crop phenology from vegetation index time-series data by improved shape model fitting in each phenological stage. Remote Sensing of Environment 277 , 113060 (2022). 75. Behnke, R. et al. Evaluation of downscaled, gridded climate data for the conterminous United States. Ecological Applications 26 , 1338–1351 (2016). 76. Tarek, M., Brissette, F. & Arsenault, R. Uncertainty of gridded precipitation and temperature reference datasets in climate change impact studies. Hydrol. Earth Syst. Sci. 25 , 3331–3350 (2021). 77. Fawcett, D., Bennie, J. & Anderson, K. Monitoring spring phenology of individual tree crowns using drone-acquired NDVI data. Remote Sens Ecol Conserv 7 , 227–244 (2021). 78. Ge, W., Li, X., Jing, L., Han, J. & Wang, F. Monitoring canopy-scale autumn leaf phenology at fine-scale using unmanned aerial vehicle (UAV) photography. Agricultural and Forest Meteorology 332 , 109372 (2023). 79. Liu, J., Fan, J., Yang, C., Xu, F. & Zhang, X. Novel vegetation indices for estimating photosynthetic and non-photosynthetic fractional vegetation cover from Sentinel data. International Journal of Applied Earth Observation and Geoinformation 109 , 102793 (2022). 80. Taugourdeau, S. et al. Leaf area index as an indicator of ecosystem services and management practices: An application for coffee agroforestry. Agriculture, Ecosystems & Environment 192 , 19–37 (2014). Annex Table Details Table 1. Summary of variables included in the study. Table 2. Performance of the best model and models with climate and vegetation variables only *: The best model based on AIC ranking includes climate predictors (rainfall at flowering dates, accumulated rainfall and rolling mean minimum temperature, and maximum temperature of 30 days prior to flowering date), and vegetation predictors (NDVI at flowering date, and mean NDVI of 30, 90, and 365 days prior to flowering date). The marginal effects of each predictor in the best model on flowering anomalies were presented in Fig. 3. A detailed statistical description of all variables in the best model is presented in Table A1 in the Annex. Table 3. Key vegetation index indicators to coffee flowering anomaly Table 4. Direct, indirect, and total effects of climate, vegetation, and management factors on flowering anomalies. “n.s.” indicates the effect is not statistically significant. The total effects in bold are the top three predictors with the highest total effects on flowering anomalies. Direct effects are the standardized regression coefficients of each predictor on the response variable. Indirect effects are the product of the direct effects operating along causal pathways in SEM models. The total effects are the sum of the predictor’s direct effect and all its indirect effects through intermediary variables. Annex Table A1. Statistical description of variables in the global model. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 05 Jan, 2025 Reviews received at journal 02 Jan, 2025 Reviewers agreed at journal 03 Dec, 2024 Reviews received at journal 22 Nov, 2024 Reviewers agreed at journal 22 Nov, 2024 Reviewers agreed at journal 21 Nov, 2024 Reviewers agreed at journal 16 Oct, 2024 Reviewers invited by journal 24 Sep, 2024 Editor assigned by journal 24 Sep, 2024 Submission checks completed at journal 22 Sep, 2024 First submitted to journal 02 Sep, 2024 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5018229","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":385618745,"identity":"d23cef8e-f766-483a-8abc-d0877a8af843","order_by":0,"name":"Thi Thu Thuy Nguyen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABBElEQVRIiWNgGAWjYDACCQST8QGMK4FdLQTwIMkyG5CshQ3OxqvFXrrHTIJxR23ihuO9x6p5d1jY8x1gPnibh8EusQGXLTJngFrOHE/ccOZc2m3eMxKJMw+wJVvzMCTj1iKRA9TSdixxw40cs9u8bRIJBgd4zKR5GJiJ0HL/jVkxUIu9wQH+b0At9YS01ABt4TFjBmph3HCAhw2o5TBuLTfSii0S2w4YzzyTYyw5tw3ol8NsxpZzDI4b49LCPiN5442PbXWyfcfPGH5421Znz3e8+eGNNxXVsri0AAGLRALDYQaFAzD+YRBhgFs9EDB/YGCoY5CHG3oAt9JRMApGwSgYmQAAathUYaqD9vQAAAAASUVORK5CYII=","orcid":"","institution":"University of Southern Queensland","correspondingAuthor":true,"prefix":"","firstName":"Thi","middleName":"Thu Thuy","lastName":"Nguyen","suffix":""},{"id":385618746,"identity":"36b35f1e-e097-4b16-ae00-8aae966ba586","order_by":1,"name":"Jarrod Kath","email":"","orcid":"","institution":"University of Southern Queensland","correspondingAuthor":false,"prefix":"","firstName":"Jarrod","middleName":"","lastName":"Kath","suffix":""},{"id":385618748,"identity":"dad8d6c7-af0d-45b3-a9eb-c52b3e006bc8","order_by":2,"name":"Louis Reymondin","email":"","orcid":"","institution":"Bioversity International, Parc scientifique Agropolis II","correspondingAuthor":false,"prefix":"","firstName":"Louis","middleName":"","lastName":"Reymondin","suffix":""},{"id":385618749,"identity":"78668775-db85-4dc8-93a0-c30d6bd37a87","order_by":3,"name":"Thong Nguyen-Huy","email":"","orcid":"","institution":"University of Southern Queensland","correspondingAuthor":false,"prefix":"","firstName":"Thong","middleName":"","lastName":"Nguyen-Huy","suffix":""},{"id":385618751,"identity":"398e4706-8287-4fef-b3c1-21f8a1e5db48","order_by":4,"name":"Vivekkananda Mattahalli Byrareddy","email":"","orcid":"","institution":"University of Southern Queensland","correspondingAuthor":false,"prefix":"","firstName":"Vivekkananda","middleName":"Mattahalli","lastName":"Byrareddy","suffix":""},{"id":385618753,"identity":"4b445ece-140f-4f62-8641-ba2464e7f2a4","order_by":5,"name":"Shahbaz Mushtaq","email":"","orcid":"","institution":"University of Southern Queensland","correspondingAuthor":false,"prefix":"","firstName":"Shahbaz","middleName":"","lastName":"Mushtaq","suffix":""}],"badges":[],"createdAt":"2024-09-02 12:14:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5018229/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5018229/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":70620257,"identity":"d880ebe5-18de-4ab1-8efe-77e4ff3f510a","added_by":"auto","created_at":"2024-12-05 03:02:57","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1038216,"visible":true,"origin":"","legend":"\u003cp\u003eStudy area in four provinces in the Central Highlands, Vietnam, a major Robusta coffee producing area. The dataset covers 558 coffee farms (in blue) over the study area. For detailed information on the dataset, please see \u003csup\u003e18\u003c/sup\u003e.\u003c/p\u003e","description":"","filename":"Figure01StudyArea.jpg","url":"https://assets-eu.researchsquare.com/files/rs-5018229/v1/ac7c3d6719a48754e44e422e.jpg"},{"id":70620256,"identity":"4c617f40-76ca-495d-8bca-d3454dddb8fe","added_by":"auto","created_at":"2024-12-05 03:02:57","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":314094,"visible":true,"origin":"","legend":"\u003cp\u003eNDVI temporal profile and corresponding coffee phenological stages. Black dots are the mean NDVI values for each day of year (DOY) from 2008 to 2017, based on data from 558 farms in the Robusta coffee growing areas in the Central Highlands, Vietnam. The black line is a smoothed trend of the mean NDVI values.\u003c/p\u003e","description":"","filename":"Figure02NDVIProfilePhenology.png","url":"https://assets-eu.researchsquare.com/files/rs-5018229/v1/bfeb11215d586ea737e8b597.png"},{"id":70620259,"identity":"3e4224ec-e1a4-411e-aadc-6da259f46b7b","added_by":"auto","created_at":"2024-12-05 03:02:57","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":364460,"visible":true,"origin":"","legend":"\u003cp\u003eMarginal effects of key predictors of coffee flowering anomalies when the effects of all other variables are held constant from the best GAM model identified from model selection. The solid black line is the mean effect, and grey shaded area is the 95% confidence interval. Data are from the Robusta coffee growing areas in the Central Highlands, Vietnam.\u003c/p\u003e","description":"","filename":"Figure03GAMoutputs.png","url":"https://assets-eu.researchsquare.com/files/rs-5018229/v1/5f42abd30047d3b09957bb36.png"},{"id":70620883,"identity":"bead4e3b-a06e-4f50-9101-a50154d3fef3","added_by":"auto","created_at":"2024-12-05 03:10:57","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":949218,"visible":true,"origin":"","legend":"\u003cp\u003eStructural equation models (SEM) with (a) all data (N = 5580), (b) early- (N=1382), and (c) late- (N=4181) flowering scenarios exploring the causal relationship between flowering anomaly, climate, vegetation, and management factors. Blue arrows present positive relationships, and red arrows are negative ones. Arrows for non-significant paths (P ≥ 0.05) are shown by black dashed lines. The thickness of the significant paths has been scaled based on the magnitude of the standardised regression coefficient given in the associated box.\u003c/p\u003e","description":"","filename":"Figure04SEMoutputs.png","url":"https://assets-eu.researchsquare.com/files/rs-5018229/v1/6e1d84ec5d06a3612df08f7e.png"},{"id":70620959,"identity":"0e04b83b-a0d5-4767-b2d6-22939882d0bb","added_by":"auto","created_at":"2024-12-05 03:19:00","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3869051,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5018229/v1/429223f6-a50f-44f0-b703-ac50dbe86229.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Vegetation growth conditions strongly indicate coffee flowering anomalies","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003ePhenology, the study of the timing of biological events, plays a crucial role in estimating yield and production of crops \u003csup\u003e1\u003c/sup\u003e. The inclusion of phenology metrics, such as germination, flowering, and fruiting, helps improve crop yield estimation models \u003csup\u003e2,3\u003c/sup\u003e. Coffee is an important cash crop globally, with 168.2\u0026nbsp;million 60kg bags in coffee year 2022/2023 and a forecasted growth rate of 5.8% in coffee year 2023/2024 \u003csup\u003e4\u003c/sup\u003e. It is well recognized for its complicated biennial phenological cycle \u003csup\u003e5\u0026ndash;8\u003c/sup\u003e. Thus, the comprehension of key growth phases in coffee plants, such as flowering, fruit development, and cherry ripening, which are directly linked to the eventual crop yield, plays a pivotal role in estimating yield and production outcomes.\u003c/p\u003e \u003cp\u003eFlowering is a key phenological phase that substantially influences crop yields. In coffee trees, this physiological process is especially critical as it determines the subsequent fruit set, providing essential insights into potential yield outcomes \u003csup\u003e9\u0026ndash;12\u003c/sup\u003e. Aspects of coffee flowering phenology, including timing/onset, density, and frequency, have shown potential consequences on coffee production in several studies, as well as for other crops (Anwar et al., 2015; Schiessl et al., 2015; Tombesi et al., 2017; Chen et al., 2019). For instance, flowering density can have impact on production contributors such as plant reproductive success, pollinators (i.e. bee abundance), and pollen flow distances \u003csup\u003e13,14\u003c/sup\u003e. In terms of timing and frequency, uneven flowering would lead to unsynchronized fruit ripening, which would increase the cost of coffee production and poor beverage quality \u003csup\u003e10,14\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eNumerous studies have found that climate factors significantly impact coffee flowering, especially temperature and precipitation \u003csup\u003e15,10,16\u0026ndash;18\u003c/sup\u003e. For example, during the pre-flowering period, warm temperatures under 30\u003csup\u003eo\u003c/sup\u003eC are important in flower bud initiation \u003csup\u003e19\u003c/sup\u003e. The rate of bud growth increased under high temperatures, and then a temperature drop following rain was decisive to break flower bud dormancy \u003csup\u003e15,9\u003c/sup\u003e. The requirement of rainfall in triggering flowering was well agreed upon in a number of studies \u003csup\u003e9,15,19\u003c/sup\u003e. Therefore, a controlled water deficit by irrigation was recommended to achieve uniform flowering \u003csup\u003e20,10,14\u003c/sup\u003e. Other climate variables, such as potential evapotranspiration and growing degree days, have also been shown to be related to coffee flowering \u003csup\u003e21,22\u003c/sup\u003e. More specifically, regarding the timing of flowering, the study by Kath et al. (2023) found that warm night (i.e., high minimum temperature) was a key climatic variable related to early flowering. Besides temperature and rainfall, the authors also found distinct effects of biological (i.e., tree age), and management (i.e., fertilizer) factors on yield under early and late flowering conditions.\u003c/p\u003e \u003cp\u003eMonitoring plant growth conditions is also crucial in understanding crop phenological stages, including flowering \u003csup\u003e24\u0026ndash;26\u003c/sup\u003e. Derived from satellite images, vegetation indices provide an effective means to assess crop growth conditions with global coverage. In addition to offering global coverage and allowing for large-scale analysis alongside gridded climate data, one of the key advantages of satellite-based vegetation indices is high spatial resolution. This capability provides critical and localized information for plentiful applications requiring precise and site-specific data. Moreover, high spatial resolution data with the ability to capture fine-scale variations can help improve the accuracy, avoiding potential over/under-estimation and misinterpretation of model outcomes \u003csup\u003e27,28\u003c/sup\u003e. Additionally, studies have shown that vegetation indices are effective at capturing vegetation responses to climate factors like temperature and precipitation \u003csup\u003e29,30\u003c/sup\u003e, as well as agricultural practices, particularly irrigation \u003csup\u003e31,32\u003c/sup\u003e. Hence, vegetation indices, representing plant conditions, can provide insights into how environmental and management variables influence plant health and growth.\u003c/p\u003e \u003cp\u003eVegetation indices have been successfully used to identify flowering phenology in various crops. For example, they have been applied to annual cereals such as rice and maize \u003csup\u003e33,34\u003c/sup\u003e, as well as to oilseed, fruit and nut crops such as rapeseed, canola, litchi, and almonds \u003csup\u003e35\u0026ndash;39\u003c/sup\u003e. Particularly in almonds and canola with their distinctive white/cream and yellow flowers, the flowering periods were detected accurately in both temporal and spatial distribution \u003csup\u003e36,38\u003c/sup\u003e. These studies have illustrated the effectiveness of vegetation indices in determining peak flowering of oilseed rape at a temporal accuracy of 1 to 4 days \u003csup\u003e35\u003c/sup\u003e, detecting flowering transitions at an overall accuracy of 85% \u003csup\u003e38\u003c/sup\u003e. More importantly, it was found that the satellite-based model provided consistent results when compared to a field-based model at different scales \u003csup\u003e34\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eSpecifically in coffee, vegetation indices have been demonstrated in various studies to estimate coffee growth stages and yield. Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) have shown consistent correlations with coffee yield with varied R\u003csup\u003e2\u003c/sup\u003e from 0.58 to 0.9, highlighting the biennial phenological cycle of coffee and the stages of dormancy and flowering as the best phenophases to determine productivity \u003csup\u003e6,7,40\u003c/sup\u003e. Remotely sensed vegetation indices from satellite and aerial images have also been utilized to characterize the two-year phenological cycle of coffee in Brazil \u003csup\u003e41\u003c/sup\u003e and to monitor the ripening stage \u003csup\u003e42\u003c/sup\u003e. However, the potential of vegetation indices in predicting anomalies in the timing of flowering for coffee remains largely unexplored.\u003c/p\u003e \u003cp\u003eHere, we investigate whether satellite-based vegetation indices can capture additional effects overlooked by climate and management factors in predicting flowering anomalies of coffee trees. The crop growth conditions during key phenological phases (i.e., flowering) provided by vegetation indices could inform management decisions and intervention activities towards season production. The findings of this study could also illustrate the effectiveness of employing high-resolution vegetation indices for agricultural research and applications that requires site-specific information.\u003c/p\u003e \u003cp\u003eOur study examined ten years of data, focusing on the flowering dates of 558 coffee farms in the Central Highlands of Vietnam. The analysis consisted of two main analyses. Firstly, we employed the Generalized Additive Model (GAM) to identify the most effective explanatory variables, including climate variables and satellite-based vegetation indices, for predicting flowering timing anomalies in Robusta coffee trees. And secondly, we performed structural equation modelling (SEM) to further investigate the complex causal relationships among flowering anomalies, climate, vegetation, and management factors.\u003c/p\u003e"},{"header":"2. Method","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Study area\u003c/h2\u003e \u003cp\u003eThe study area covers four provinces: Gia Lai, Dak Lak, Dak Nong, and Lam Dong in the Central Highlands, Vietnam (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). This region is the world\u0026rsquo;s leading producer of Robusta coffee, with approximately 648,400 hectares producing 26.6\u0026nbsp;million 60kg bags in the 2022/2023 coffee year. It represents 91.2% of the country's coffee-growing area and 95.2% of its total coffee production, contributing around 36% to global Robusta coffee output \u003csup\u003e43\u003c/sup\u003e. The Robusta coffee calendar in the study area can be divided into four key growth stages (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e): (i) start of the season/flower-bud initiation and development: November-December, (ii) flowering: January-March, (iii) fruit setting and development: April-August, and (iv) Maturation / Ripening / Harvest: September-November \u003csup\u003e44\u0026ndash;47\u003c/sup\u003e. This study focused on the flowering date, marking the start of the flowering period from January to February.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. Data\u003c/h2\u003e \u003cdiv id=\"Sec5\" class=\"Section3\"\u003e \u003ch2\u003e2.2.1. Coffee flowering data\u003c/h2\u003e \u003cp\u003eWe analyzed a ten-year dataset, from 2008\u0026ndash;2017, on the flowering dates of 558 coffee farms (total N\u0026thinsp;=\u0026thinsp;5580) in the Central Highlands, Vietnam (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe dataset was collected within the Sustainable Management Services (SMS) program implemented by ECOM Agroindustrial Corporation in Vietnam, with records maintained for coffee certification. In the dataset, the flowering date was defined as the first flowering date. Over the study area, flowering is highly synchronized that 30\u0026ndash;40% of plants start flowering on the same day, and all plants flower within a day or two. Management factors were also recorded in the dataset, including total volume of irrigation (liters tree\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e), and fertilizer (kg ha\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e), together with tree age, which was calculated based on planting year. For further details on the dataset and its collection, please see Byrareddy et al. (2019); Kath et al. (2020; 2023). To explore the explanatory power of climate variables and vegetation growth conditions on the occurrence of flowering, we calculated flowering anomalies over the study period for each year and each farm.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section3\"\u003e \u003ch2\u003e2.2.2. Vegetation growth conditions\u003c/h2\u003e \u003cp\u003eTo explore vegetation growth conditions, we used the Normalized Difference Vegetation Index (NDVI), a widely-used metric for quantifying vegetation greenness using satellite data, from MODIS Terra Daily NDVI MOD09GA product in Google Earth Engine Data Catalog \u003csup\u003e49\u003c/sup\u003e. Daily NDVI values at 500m x 500m spatial resolution were extracted at farm locations with longitude and latitude coordinates, covering the period from 01/01/2007 to 31/12/2018. The sites selected were covered with maximum coffee plants and therefore the NDVI data represents coffee parameters and its growth cycle. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents the NDVI temporal profile with mean NDVI values for each day of year (DOY) over the study period, aligned with the corresponding coffee phenological stages in the study area.\u003c/p\u003e \u003cp\u003eThe NDVI dataset was joined with the coffee flowering dataset to get NDVI values for each farm at each year\u0026rsquo;s flowering start date. In cases where there was no NDVI value in a specific flowering date, the NDVI value available in the previous closest date was used. To explore the impact of NDVI values at different points in time prior to the flowering date, the rolling mean of NDVI values of 14, 30, 90, 120 and 365 days before the flowering dates were computed. These periods correspond to key phenological stages of coffee plants, especially bud initiation and development, and dormancy, which have accumulated impacts on flowering \u003csup\u003e10,50\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003e2.2.3. Climate data\u003c/h2\u003e \u003cp\u003eTo explore the explanatory power of climate variables in predicting coffee flowering anomalies, we used the global historical climate dataset ERA5 \u0026ndash; Fifth generation of ECMWF atmospheric reanalyses of the global climate from Copernicus Climate Change Service \u003csup\u003e51\u003c/sup\u003e, accessed through Google Earth Engine Data Catalog. From the daily dataset with 27.8km spatial resolution, we extracted daily rainfall, minimum, and maximum temperature from 01/01/2008 to 31/12/2017 for each farm location with longitude and latitude coordinates. The climate dataset was joined with the coffee flowering dataset to select rainfall, minimum, and maximum temperature at the flowering start date over the study period. The accumulated rainfall and rolling mean of minimum and maximum temperature of 30 days prior to flowering dates were also computed.\u003c/p\u003e \u003cp\u003eA summary of variables included in the study is presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e below. For a detail statistical description of the variables, please see Table \u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003eA1\u003c/span\u003e. Statistical description of variables in the global model.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSummary of variables included in the study.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSpatial resolution\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTemporal resolution\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eData period\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDescription\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eFlowering anomalies\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2008\u0026ndash;2017\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eFlowering anomalies were calculated using flowering start dates for each farm each year.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eVegetation variables\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e500m\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDaily\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2007\u0026ndash;2018\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDaily NDVI values from MOD09GA dataset in Google Earth Engine Data Catalog. NDVI at flowering dates were extracted by farm locations. Rolling mean NDVI of 14, 30, 90, 120, and 365 days prior to flowering dates were computed.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eClimate variables\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e27.8km\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDaily\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2007\u0026ndash;2017\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eGlobal historical climate dataset ERA5 in Google Earth Engine Data Catalog. Rainfall, minimum temperature, and maximum temperature at flowering dates were extracted by farm locations. Accumulated rainfall and rolling means of minimum temperature, and maximum temperature of 30 days prior to flowering dates were computed.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Data analysis\u003c/h2\u003e \u003cp\u003eThe overall aim of the analysis was to identify the key explanatory variables in predicting flowering anomalies. The analysis contained two major parts. Firstly, to investigate the explanatory power of climate and vegetation growth condition variables in predicting flowering anomalies, a Generalized Additive Model (GAM) in mgcv \u003csup\u003e52\u003c/sup\u003e and MuMIn \u003csup\u003e53\u003c/sup\u003e R packages, was used to fit the flowering anomalies model using a spline with consideration of spatial and temporal clusters. Secondly, to test the hypothesis that vegetation indices can add an explanation, besides climate and management, to predict flowering anomalies, we quantified the direct and indirect effects of NDVI and climate variables using Structural Equation Model (SEM) using piecewiseSEM package in R \u003csup\u003e54\u003c/sup\u003e.\u003c/p\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003e2.3.1. Flowering anomalies modelling\u003c/h2\u003e \u003cp\u003eTo identify the explanatory variables in predicting flowering anomalies, we used the Generalized Additive Model (GAM). GAM models have become increasingly popular in studying crop phenology, including coffee, due to their flexibility in capturing non-linear relationships and interactions between predictor variables. They were proven to be effective in explaining the effects of different variables in predicting agricultural crop yield, especially when working with spatio-temporal large data \u003csup\u003e55\u0026ndash;57\u003c/sup\u003e. Several studies have utilized GAMs to investigate coffee phenology and its responses to climatic variability. For example, Craparo et al. (2021) found that warming nocturnal temperatures have a superseding effect on coffee ripening. Based on that, the authors have developed a Warm Night Index (WNI) that accurately predicts the start of the harvest season.\u003c/p\u003e \u003cp\u003eHere, we used GAM to fit the flowering anomalies model using a spline with consideration of spatial and temporal clusters. All analyses were carried out in R \u003csup\u003e58\u003c/sup\u003e with mgcv \u003csup\u003e52\u003c/sup\u003e and MuMIn \u003csup\u003e53\u003c/sup\u003e packages.\u003c/p\u003e \u003cp\u003eThe global model was built using the following equation: \u003cem\u003ey\u0026thinsp;~\u0026thinsp;f(x\u003c/em\u003e\u003csub\u003e\u003cem\u003eij\u003c/em\u003e\u003c/sub\u003e\u003cem\u003e) + (YR) + (ST) + (prov)\u003c/em\u003e\u003c/p\u003e \u003cp\u003eWhere flowering anomalies (\u003cem\u003ey\u003c/em\u003e) were modelled as a non-linear function (\u003cem\u003ef\u003c/em\u003e) of predictor variables (\u003cem\u003ex\u003c/em\u003e) for each farm (\u003cem\u003ei\u003c/em\u003e) and year (\u003cem\u003ej\u003c/em\u003e) using a Gaussian distribution. There are two groups of predictors: climate variables (rainfall, minimum, and maximum temperature at flowering dates, accumulated rainfall and rolling mean of minimum, and maximum temperature of 30 days prior to flowering dates), and vegetation growth conditions (NDVI values at flowering dates, and rolling mean of NDVI of 14, 30, 90, 120 and 365 days prior to flowering dates). See Table \u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003eA1\u003c/span\u003e for details for all of the NDVI and climate variables used in the analysis.\u003c/p\u003e \u003cp\u003eA random effect for each farm location (\u003cem\u003eST\u003c/em\u003e) in each province (\u003cem\u003eprov\u003c/em\u003e) was included to account for the repeated measurements for each year (\u003cem\u003eYR\u003c/em\u003e) at the farm-level. Restricted maximum likelihood (fREML) was selected as the smoothing parameter estimation method. Strongly correlated variables with a Pearson coefficient r \u0026gt; |0.7| were removed in model selection to avoid multi-collinearity \u003csup\u003e59\u003c/sup\u003e. Multi-model selection using the Akaike information criterion (AIC) was used to rank all possible combinations of predictors \u003csup\u003e53\u003c/sup\u003e. The one with the lowest AIC is the best model, representing the combination of the best explanatory predictors.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e \u003ch2\u003e2.3.2. Structural Equation Model\u003c/h2\u003e \u003cp\u003eTo further explore the explanatory powers of vegetation indices, in addition to climate and management factors, in predicting flowering anomalies, we utilized the Structural Equation Model (SEM) as piecewise estimation of local relationships. SEM modelling has two major characteristics over traditional regression approaches: (i) Paths among variables represent the hypothesized causal relationships, and (ii) variables can appear as both predictors and responses \u003csup\u003e54\u003c/sup\u003e. These features allowed us to develop models to capture the causal pathways from predictors to response variables, including both the direct and indirect effects of climate, vegetation, and management predictors on flowering anomalies.\u003c/p\u003e \u003cp\u003eThe dataset with climate and vegetation predictors of the best model from the previous analysis, together with management factors, including tree age, irrigation, and fertilizer, was fit in SEM models using the piecewiseSEM package in R \u003csup\u003e54\u003c/sup\u003e. To account for spatial autocorrelation, we fitted the SEMs using linear mixed models with a random effect for each site nested within each province. To account for temporal autocorrelation, we fitted SEM models using an autoregressive process of order 1.\u003c/p\u003e \u003cp\u003eTo test the causal relationships of the predictors under early- and late-flowering scenarios, we split the dataset into two separate sets of early flowering (i.e., a flowering day anomaly\u0026thinsp;\u0026lt;\u0026thinsp;0), and late flowering (i.e., a flowering day anomaly\u0026thinsp;\u0026gt;\u0026thinsp;0), to build two separate SEM models besides the first model with all available data. To compare the effect sizes among predictors and between models, coefficients were standardized [mean(x)/1.SD(x)] \u003csup\u003e60\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eFinally, the direct, indirect and total effects of each predictor on flowering anomalies were calculated. Direct effects are the standardized regression coefficients of each predictor on the response variable. Indirect effects are the product of the direct effects operating along causal pathways in SEM models. The total effects are the sum of the predictor\u0026rsquo;s direct effect and all its indirect effects through intermediary variables.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.1. Vegetation growth conditions strongly indicate coffee flowering anomalies with greater explanatory power than climate predictors.\u003c/h2\u003e \u003cp\u003eVegetation growth conditions strongly indicate coffee flowering anomalies. The best model included the climate predictors (rainfall at flowering dates, accumulated rainfall 30 days prior to flowering dates, and rolling mean of minimum, and maximum temperature of 30 days prior to flowering dates) and vegetation predictors (NDVI at flowering dates, rolling mean NDVI of 30, 90, and 365 days prior to flowering dates). For a detailed statistical description of all variables in the best model, please see Table \u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003eA1\u003c/span\u003e. Statistical description of variables in the global model..\u003c/p\u003e \u003cp\u003eThe marginal effects of each predictor in the best model on flowering anomalies are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. The result showed that NDVI in a month and three months prior to flowering dates have the most explanatory power for flowering anomalies. Specifically, 30-day pre-flowering NDVI (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ef) of approximately less than 0.3 had a strong positive linear relationship with flowering anomalies, particularly early flowering. This indicates coffee trees that are under less green conditions at the end of the dormancy stage (i.e., a month prior to flowering dates) expect early flowering.\u003c/p\u003e \u003cp\u003eIn contrast, 90-day pre-flowering mean NDVI (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eg) of approximately less than 0.3 had a strong negative linear relationship with flowering anomalies, particularly late flowering. In this case, coffee trees with less greenness at the start of the season, when flower buds initiate and develop (i.e., three months prior to flowering dates), are likely to flower late.\u003c/p\u003e \u003cp\u003eOn the other hand, NDVI at the flowering date (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ee) had less effect on flowering anomalies, while NDVI in a year prior to the flowering date (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eh) demonstrated a slight non-linear relationship with flowering anomalies.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe best model, consisting of both climate and vegetation predictors, has an adjusted R\u003csup\u003e2\u003c/sup\u003e of 0.865 and can explain 86.7% of deviance (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The explanatory power of vegetation predictors (84.7% of deviance) was outstanding compared to that of climate variables (77.7% of deviance). The AIC values also show a higher rank for the model with vegetation predictors than the one with climate predictors.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePerformance of the best model and models with climate and vegetation variables only\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAdj. R2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDeviance explained\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAIC\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBest model *\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.865\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e86.7%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e24082.48\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBest model with climate predictors only\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.775\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e77.7%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e26914.98\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBest model with vegetation predictors only\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.846\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e84.7%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e24807.86\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cem\u003e*: The best model based on AIC ranking includes climate predictors (rainfall at flowering dates, accumulated rainfall and rolling mean minimum temperature, and maximum temperature of 30 days prior to flowering date), and vegetation predictors (NDVI at flowering date, and mean NDVI of 30, 90, and 365 days prior to flowering date). The marginal effects of each predictor in the best model on flowering anomalies were presented in\u003c/em\u003e Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. \u003cem\u003eA detailed statistical description of all variables in the best model is presented in\u003c/em\u003e Table \u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003eA1\u003c/span\u003e \u003cem\u003ein the Annex.\u003c/em\u003e\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Vegetation index captures additional aspects besides climate and management factors\u003c/h2\u003e \u003cp\u003eThe SEM results quantified the casual relationships, and the indirect and direct effects of NDVI, climate and management factors on flowering anomalies (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, and Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Specifically, 30-day pre-flowering NDVI has the strongest direct positive effect on flowering anomalies in both models with all data (standardized regression coefficient\u0026thinsp;=\u0026thinsp;0.787) and early flowering scenario (standardized regression coefficient\u0026thinsp;=\u0026thinsp;1.623) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea, b). The 90-day pre-flowering NDVI value also shows a strong direct negative effect on flowering anomalies in all three SEM models with all data (standardized regression coefficient = -0.624), early- (standardized regression coefficient = -0.925), and late-flowering scenarios (standardized regression coefficient = -0.346). Mean NDVI of a year prior to flowering dates also showed a significant negative effect on flowering anomalies in two SEM models with all data (standardized regression coefficient = -0.409) and early flowering scenario (standardized regression coefficient = -0.491).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThese results are consistent with GAM outputs from the previous analysis. During the bud initiation and development phase (i.e., 90 days prior to flowering date), higher NDVI values suggest an earlier onset of flowering stage, while lower NDVI values indicate a delay in the flowering. This phase corresponds to a period of active growth of coffee trees to prepare for the flowering (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Therefore, a high NDVI value associated with a vigorous condition suggests coffee plants are progressing well toward flowering, and an earlier onset is likely to occur. After two to three months of bud development, coffee plants enter the dormancy stage (i.e., 30 days prior to the flowering date), characterized by reduced metabolic activity, reflected in a slower rise of NDVI values. In this case, a lower 30-day pre-flowering NDVI showing a more ready, ripe-to-flower condition indicates that flowering start dates will arrive earlier (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eSEM model results also highlight the significant effects of NDVI a year prior to the flowering date (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). A lower NDVI a year before indicates a later flowering date the following year.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eKey vegetation index indicators to coffee flowering anomaly\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePhenology stage and coffee plant ecophysiology\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eVegetation index indicator and flowering anomaly\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eBud initiation and development\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003e90-day pre-flowering NDVI\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eCoffee plants are in an active growing stage, characterized by increasing chlorophyll production and photosynthetic efficiency.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDuring this stage, the NDVI values are in an increasing trend, indicating vigorous and healthy plants. Therefore, higher NDVI values indicate that coffee plants are progressing well toward flowering, and an earlier onset is likely to happen.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eWintgens, 2004, page 12\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLower NDVI values indicate that coffee plants need more time for development and are at the earlier stage of bud initiation. Therefore, the condition suggests a delay in flowering.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDormancy\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e30-day pre-flowering NDVI\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCoffee plants are quiescence, characterized by reduced metabolic activity.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDuring the dormancy stage, the rise of NDVI values is substantially reduced. Hence, a lower 30-day pre-flowering NDVI value, showing a more ready, ripe-to-flower condition, indicates an earlier onset of the flowering stage.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eWintgens, 2004, page 12\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eRegarding the total effects of the predictors (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e and Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e), in all three SEM models, the 30-day pre-flowering NDVI is within the top three predictors with the highest total effects on flowering anomalies. Similarly, the 90-day pre-flowering NDVI is within the top three predictors with the highest total effects on flowering anomalies in SEM models with all data and late flowering scenarios. With the additional indirect effects captured through NDVI, the 30-day pre-flowering maximum temperature has the highest total effect on flowering anomalies in the SEM model with all data. In contrast, 30-day pre-flowering minimum and maximum temperature, together with 30-day pre-flowering NDVI, are the top three predictors with the highest total effects on flowering anomalies in SEM early flowering model.\u003c/p\u003e \u003cp\u003eBesides the direct effects of climate predictors on flowering anomalies, their indirect effects are captured by vegetation predictors in SEM models (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e and Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Specifically in the early flowering model, the indirect effects of 30-day pre-flowering rainfall and maximum temperature captured through 30-day pre-flowering NDVI, are two to four times larger than their direct effects. Additionally, the indirect effect of 30-day pre-flowering minimum temperature, captured through NDVI during the same period, is almost equal to the direct effect. In the late flowering model, the direct effects of 30-day pre-flowering rainfall and maximum temperature are not statistically significant. However, their total effects on flowering anomalies are statistically significant captured through the impact pathway of NDVI of the same period.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable A1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eStatistical description of variables in the global model.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMin\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1st Qu.\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMean\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3rd Qu.\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eMax\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eFlowering anomalies\u003c/b\u003e (day)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-15.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e7.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eVegetation index at different periods prior to flowering date\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFlowering NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0003\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.2091\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.4518\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.6590\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.9711\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e14 days pre-flowering mean NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0294\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.3075\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.4020\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.4975\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.7640\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e30 days pre-flowering mean NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0440\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.2927\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.3814\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.4738\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.7180\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e90 days pre-flowering mean NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.1190\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.2995\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.3791\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.4573\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.6300\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e120 days pre-flowering mean NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.1283\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.2845\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.3526\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.4156\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.5991\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e365 days pre-flowering mean NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.1935\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.3079\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.3422\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.3756\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.5032\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eClimate variables\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFlowering rainfall (mm)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.731\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e32.373\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e89.477\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e124.709\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e442.541\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFlowering minimum temperature (\u003csup\u003eo\u003c/sup\u003eC)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e13.76\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e16.53\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e17.65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e18.65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e22.02\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFlowering maximum temperature (\u003csup\u003eo\u003c/sup\u003eC)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e24.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e27.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e28.36\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e29.35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e33.01\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e30 days pre-flowering rainfall (mm)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.004\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.135\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e23.135\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e23.827\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e206.814\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e30 days pre-flowering minimum temperature (\u003csup\u003eo\u003c/sup\u003eC)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e13.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e17.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e18.04\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e19.01\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e22.16\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e30 days pre-flowering maximum temperature (\u003csup\u003eo\u003c/sup\u003eC)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e23.57\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e27.20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e28.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e29.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e34.60\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDirect, indirect, and total effects of climate, vegetation, and management factors on flowering anomalies. \u0026ldquo;n.s.\u0026rdquo; indicates the effect is not statistically significant. The total effects in bold are the top three predictors with the highest total effects on flowering anomalies. Direct effects are the standardized regression coefficients of each predictor on the response variable. Indirect effects are the product of the direct effects operating along causal pathways in SEM models. The total effects are the sum of the predictor\u0026rsquo;s direct effect and all its indirect effects through intermediary variables.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSEM models\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePredictor\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDirect effect\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eIndirect effect\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTotal effect\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eALL DATA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFlowering rainfall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.328\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.003\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.331\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30-day pre-flowering rainfall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.170\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.071\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.241\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30-day pre-flowering tmin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.465\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.154\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.619\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30-day pre-flowering tmax\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.497\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.332\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.829\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal irrigation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.235\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.033\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.268\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal fertilizer\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.002\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.139\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.141\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTree age\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.049\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.049\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFlowering NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.046\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.046\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30-day pre-flowering NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.787\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.787\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e90-day pre-flowering NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.624\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e-0.624\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e365-day pre-flowering NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.409\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.409\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eEARLY FLOWERING\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFlowering rainfall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30-day pre-flowering rainfall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.094\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.399\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.493\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30-day pre-flowering tmin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.902\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.790\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e-1.692\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30-day pre-flowering tmax\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.599\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.461\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e2.060\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal irrigation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.837\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.661\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-1.498\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal fertilizer\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTree age\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.127\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.078\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.127\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFlowering NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30-day pre-flowering NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.623\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e1.623\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e90-day pre-flowering NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.925\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.925\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e365-day pre-flowering NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.491\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.491\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eLATE FLOWERING\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFlowering rainfall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.069\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.069\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30-day pre-flowering rainfall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.040\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.040\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30-day pre-flowering tmin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.096\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.153\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.057\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30-day pre-flowering tmax\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.161\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.161\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal irrigation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.203\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.203\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal fertilizer\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTree age\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFlowering NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003en.s.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30-day pre-flowering NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.238\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.238\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e90-day pre-flowering NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.346\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e-0.346\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e365-day pre-flowering NDVI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.100\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eSEM model results also show a strong effect of irrigation, especially in the early flowering SEM model. In the early flowering model, the indirect effects of irrigation on flowering anomalies, captured through NDVI during flowering and pre-flowering stage, are closely equivalent to its direct effect. As a result, the total effect of irrigation on flowering anomalies is almost double the direct effect alone. In contrast, total fertilizer and tree age show much less effects on flowering anomalies, most of which are not statistically significant (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eFlowering phenology plays a vital role in coffee production. Deviations from the expected flowering patterns can have significant implications for crop yield and quality \u003csup\u003e9\u0026ndash;12\u003c/sup\u003e. However, while the impacts of climate variables on coffee flowering have been extensively studied, how vegetation indices retrieved from satellite images can contribute to predicting coffee flowering anomalies remain underexplored. Using an extensive 10-year dataset (N\u0026thinsp;=\u0026thinsp;5580) of Robusta coffee flowering dates, climate predictors (rainfall and temperature), and NDVI values at different periods prior to flowering dates, we found that vegetation growth conditions strongly indicate coffee flowering anomalies. Specifically, coffee plant conditions, represented by NDVI values, of a month and three months prior to flowering dates, corresponding to dormancy and bud growth stages, were identified to have the most explanatory power to predict the flowering timing anomalies, outperforming climate variables. The 30-day pre-flowering NDVI values had a strong positive relationship with the flowering anomalies, particularly early flowering. On the other hand, the 90-day pre-flowering NDVI values, reflecting plant conditions during bud initiation and development stage, indicated flowering anomalies in both early and late scenarios.\u003c/p\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e4.1. Vegetation growth conditions strongly indicate coffee flowering anomalies.\u003c/h2\u003e \u003cp\u003eMost published studies on coffee flowering phenology have been focused on the influences of the external environment on the plant, such as temperature, rainfall, soil moisture, etc., as triggers to flowering \u003csup\u003e23,61\u003c/sup\u003e. However, coffee flowering is a complex sequence of flower bud initiation, dormancy development, dormancy breakage, stimulation of regrowth, and regrowth to anthesis, which is the flower blooming \u003csup\u003e9,10\u003c/sup\u003e. Throughout the production sequence, crop conditions during the period of quiescence did not receive much attention.\u003c/p\u003e \u003cp\u003eIn this study, the plant\u0026rsquo;s internal conditions necessary for flowering, or in other words, the ripe-to-flower conditions, were detected by a vegetation index, which serves as an indicator of plant conditions. The analysis result showed that a lower 30-day pre-flowering NDVI value is a strong indicator for early flowering of the upcoming period. The finding is consistent with current literature on conditions to induce flowering of coffee plants that they need a pre-flowering dry period, characterized by low rainfall and high temperature, to stimulate flowering \u003csup\u003e9,10,62,18\u003c/sup\u003e. During the dormancy stage of approximately 30 days prior to flowering dates, coffee plants are in quiescent conditions, which are characterized by reduced metabolic activity, represented by slower rise of NDVI values comparing to the bud development phase \u003csup\u003e9,10,50\u003c/sup\u003e. At the end of dormancy stage, watering, either through rainfall or irrigation, will stimulate flowering, coffee plants regrow with vigorous conditions, represented by the peak of NDVI values in the temporal profile \u003csup\u003e10,50\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eFurthermore, the conditions of coffee plants during the flower bud initiation and development stage, approximately 90 days prior to flowering dates, were also captured by NDVI. This active growing period, characterized by increasing chlorophyll production and photosynthetic efficiency, was presented in an increasing trend of NDVI values \u003csup\u003e9,10\u003c/sup\u003e. Therefore, higher NDVI values indicate that coffee plants are progressing well toward flowering, and an earlier onset is likely to happen. While lower NDVI values in this period indicate that coffee plants are at the earlier stage of bud initiation, suggesting a delay in flowering. Thus, our findings suggested that NDVI is an effective indicator for detecting shifts in the flowering period to either earlier or later than normal, or in other words, predict the anomaly of the upcoming flowering of coffee trees.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e4.2. Vegetation growth conditions indicate stronger explanatory power, capturing additional aspects besides climate and management.\u003c/h2\u003e \u003cp\u003eThe impact of climate factors on predicting coffee growth has been extensively studied. Temperature and rainfall are widely recognized as key determinants affecting the various stages of coffee plant development \u003csup\u003e15,10,16,18\u003c/sup\u003e. For flowering in particular, optimal temperatures and consistent rainfall patterns are crucial for initiating and sustaining the flowering process. However, we found that satellite-based vegetation index can capture additional aspects besides climate and management, which improves the model ability to predict flowering anomalies. Besides, the model using only vegetation predictors outperforms the one relying solely on climate variables.\u003c/p\u003e \u003cp\u003eEven though the particular role of vegetation indices in predicting coffee flowering anomalies is still underexplored, several studies have demonstrated the effectiveness of vegetation indices retrieved from satellite images in predicting coffee yield \u003csup\u003e63\u003c/sup\u003e. Models using vegetation indices, such as NDVI and LAI, achieved higher accuracy than those relying solely on climate data \u003csup\u003e47\u003c/sup\u003e. Moreover, a study by Bola\u0026ntilde;os et al. (2023) focusing on flowering, the early stages of the coffee production cycle to predict yield, found that NDVI was among the top three predictors having the highest correlation with yield. Still, further research is needed to thoroughly investigate the specific role of satellite-based vegetation indices in predicting flowering phenology, and other key growth phases of coffee in general, ultimately leading to more accurate yield estimation.\u003c/p\u003e \u003cp\u003eA notable aspect is the ability of vegetation indices to detect the influence of crop conditions from early stages at the start of the season on flowering anomalies. The healthy and vigorous conditions of the plants are captured by the vegetation index, indicating the position at different progressing steps within the phenological stages. This finding can be attributed to their capability to reflect the cumulative effects of environmental conditions, emphasizing the longer-term impacts of climate on plant conditions and productivity \u003csup\u003e65,66\u003c/sup\u003e. On the other hand, climate factors, such as temperature and solar radiation, have relatively short cumulative effects on vegetation growth of various tree-type vegetation \u003csup\u003e67\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eMoreover, the influence of plant conditions in a year prior to the flowering date on the flowering anomalies of the following year is consistent with the biennial nature of coffee production. This suggests that resources, such as nutrients, of the plant are accumulated during the vegetative stage in the previous year towards the production stage in the following year \u003csup\u003e68,69\u003c/sup\u003e. The effectiveness of nutrient accumulation through various agricultural management practices, such as irrigation, fertilization, and pest control, is illustrated through changes in plant growth and canopy density. These changes can be measured using vegetation indices \u003csup\u003e70,71\u003c/sup\u003e, providing substantial information in predicting flowering anomalies. However, further study might be needed to separate the accumulated impacts of environmental conditions and agricultural management activities on the coffee production cycle, providing site-specific information for decision-making.\u003c/p\u003e \u003cp\u003eIndirect effects of climate predictors, as captured through vegetation indices are strong, especially in the early flowering SEM model. Additionally, vegetation indices also pick up the indirect effects of management factors, particularly irrigation in early flowering SEM model, and fertilizer in all data SEM model. Notably, in the late flowering SEM model, the impacts of 30-day rainfall and maximum temperature can only be captured through the pathway of vegetation indices with their direct effects statistically insignificant. This highlights the additional insights that vegetation indices can provide beyond climate and management variables in predicting flowering anomalies of coffee trees. These results are also aligned with the findings of other published studies that vegetation indices can capture vegetation responses to climate factors (i.e., temperature and precipitation) \u003csup\u003e29,30\u003c/sup\u003e, and agricultural management activities (i.e., irrigation) \u003csup\u003e31,32\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eAdditionally, the direct effects of vegetation indices, which represent their total impacts on flowering anomalies, are consistently among the top three predictors in all three SEM models. This further explains the superior performance of vegetation-only model in GAM, where it outperformed the one relying solely on climate predictors.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e4.3. Spatial variations and the effectiveness of remote sensing data\u003c/h2\u003e \u003cp\u003eThis study showed a promising result when using the vegetation index retrieved from satellite data with the most popular and widely used index, such as NDVI. Additionally, other vegetation indices, such as Enhanced Vegetation Index (EVI), Leaf Area Index (LAI), Vegetation Health Index (VHI), and Vegetation Condition Index (VCI), have also shown their effectiveness in mapping crop phenology \u003csup\u003e72,73,35,74\u003c/sup\u003e. There are other indices developed specifically for flowering phenology, such as the Enhanced Bloom Index (EBI) \u003csup\u003e36\u003c/sup\u003e. However, further studies are still needed to explore more data sources and information to contribute to exploring other key growth stages within coffee complex phenology.\u003c/p\u003e \u003cp\u003eAdditionally, satellite-based vegetation indices offer several key advantages over climate data. Firstly, the detailed and site-specific information provided by high-resolution satellite images allows for more accurate modelling, as it captures fine-scale variations in vegetation that coarse resolution data may overlook and lead to potential overestimations or inaccuracies in model outcomes \u003csup\u003e27,28\u003c/sup\u003e. Secondly, satellite data can provide comprehensive coverage, which is particularly beneficial for multi-scale analyses. Even though gridded reanalysis climate data can also offer global coverage, the downscaled dataset often carry uncertainties, especially in remote areas where ground-based station data is sparse or unavailable due to low-density networks \u003csup\u003e75,76\u003c/sup\u003e. And thirdly, many satellite datasets, such as MODIS, Landsat, or Sentinel, are freely available, making them a cost-effective alternative to purchasing station-based data, which can be expensive and limited in coverage. These factors make satellite-based vegetation indices a promising data sources agricultural analyses and applications.\u003c/p\u003e \u003cp\u003eEven though a medium resolution (i.e., 500m) of satellite images offers substantially better spatial resolution compared to gridded climate data, using MODIS dataset is still a limitation of this study for the trade-off of better temporal resolution (i.e., daily) to capture crop phenology. Moreover, with medium spatial resolution, NDVI values in each pixel can be influenced by external factors such as soil background, and shadows. With recent advances in remote sensing techniques the fusion of data from multiple satellite sensors with high spatial resolution, such as freely available Landsat and Sentinel or other commercial satellites, further studies are needed to illustrate the effectiveness of high-resolution satellite images to overcome the limitations of medium spatial resolution. Furthermore, it could be beneficial to explore other remote sensing data than satellite-based, such as drones and UAVs, which have also illustrated the effectiveness of employing very high-resolution datasets to monitor crop phenology at different scales and provide more insights into the spatial discrepancy at local levels \u003csup\u003e42,77,78\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eDespite being the most popular vegetation index utilized in agricultural research and applications, NDVI primarily captures the greenness of vegetation, making it less sensitive to non-photosynthetic elements (Liu et al., 2022), such as cherries or flowers, which are important indicators of coffee phenology. Studies have found other vegetation indices that demonstrated potential relations with coffee phenology and yield, such as Leaf Area Index (LAI) \u003csup\u003e80\u003c/sup\u003e, EVI \u003csup\u003e6\u003c/sup\u003e, Soil-Adjusted Vegetation Index (SAVI), and Normalized Difference Water Index (NDWI) \u003csup\u003e40\u003c/sup\u003e. Thus, further studies are needed to investigate those alternatives, or even to develop a vegetation index specifically tailored for coffee as the existing vegetation indices are not produced to capture identical parameters of the crop \u003csup\u003e42\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eOur analysis indicates that vegetation growth conditions provide robust indicators of coffee flowering anomalies, highlighting its advantage over climate variables. Specifically, we found that the 30-day pre-flowering NDVI values had a strong positive relationship with flowering anomalies, particularly early flowering. Additionally, the 90-day pre-flowering NDVI values, reflecting plant conditions during bud initiation and development stage, indicated flowering anomalies in both early and late scenarios. The analysis also highlighted the advantages of vegetation indices over climate predictors to capture plant conditions through its growing cycle with accumulated effects of environmental factors and agricultural management activities, especially during the vegetative stage in biennial phenology.\u003c/p\u003e \u003cp\u003eThe findings also highlight the ability of vegetation indices in capturing plant growth conditions during crucial phenological stages at the start of the crop season, providing valuable insights for management decisions and interventions to optimize seasonal production. The results of this study further demonstrate the effectiveness of employing high-resolution vegetation indices for agricultural research and applications the requires site-specific information, especially in monitoring crop growth conditions in complex landscapes with local-level spatial variations.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eT.N. conceived the initial study based on conversations with J.K. and L.R.; T.N. and V.M.B. collected the original data; T.N. processed the data and conducted the data analysis; T.N. prepared the initial manuscript; J.K., L.R., T.N.-H., V.M.B., and S.M. reviewed and edited the manuscript. All authors reviewed the manuscript and gave their approval for publication.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eWe acknowledge the funding from the German Federal Ministry for Economic Affairs and Climate Action (BMWK) through the International Climate Initiative (IKI).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003e1. Wielgolaski, F.-E. Phenology in Agriculture. in \u003cem\u003ePhenology and Seasonality Modeling\u003c/em\u003e (ed. Lieth, H.) vol. 8 369\u0026ndash;381 (Springer Berlin Heidelberg, Berlin, Heidelberg, 1974).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e2. Bolton, D. K. \u0026amp; Friedl, M. A. Forecasting crop yield using remotely sensed vegetation indices and crop phenology metrics. \u003cem\u003eAgricultural and Forest Meteorology\u003c/em\u003e \u003cb\u003e173\u003c/b\u003e, 74\u0026ndash;84 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e3. Sakamoto, T., Gitelson, A. A. \u0026amp; Arkebauer, T. J. MODIS-based corn grain yield estimation model incorporating crop phenology information. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e \u003cb\u003e131\u003c/b\u003e, 215\u0026ndash;231 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e4. International Coffee Organization. \u003cem\u003eCoffee Report and Outlook\u003c/em\u003e. 43 https://icocoffee.org/documents/cy2023-24/Coffee_Report_and_Outlook_December_2023_ICO.pdf (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e5. Camargo, \u0026Acirc;. P. D. \u0026amp; Camargo, M. B. P. D. Defini\u0026ccedil;\u0026atilde;o e esquematiza\u0026ccedil;\u0026atilde;o das fases fenol\u0026oacute;gicas do cafeeiro ar\u0026aacute;bica nas condi\u0026ccedil;\u0026otilde;es tropicais do Brasil. \u003cem\u003eBragantia\u003c/em\u003e \u003cb\u003e60\u003c/b\u003e, 65\u0026ndash;68 (2001).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e6. Bernardes, T., Moreira, M. A., Adami, M., Giarolla, A. \u0026amp; Rudorff, B. F. T. Monitoring Biennial Bearing Effect on Coffee Yield Using MODIS Remote Sensing Imagery. \u003cem\u003eRemote Sensing\u003c/em\u003e \u003cb\u003e4\u003c/b\u003e, 2492\u0026ndash;2509 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e7. Brunsell, N. A., Pontes, P. P. B. \u0026amp; Lamparelli, R. A. C. Remotely Sensed Phenology of Coffee and Its Relationship to Yield. \u003cem\u003eGIScience \u0026amp; Remote Sensing\u003c/em\u003e \u003cb\u003e46\u003c/b\u003e, 289\u0026ndash;304 (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e8. De Oliveira Aparecido, L. E., De Souza Rolim, G., Camargo Lamparelli, R. A., De Souza, P. S. \u0026amp; Dos Santos, E. R. Agrometeorological Models for Forecasting Coffee Yield. \u003cem\u003eAgronomy Journal\u003c/em\u003e \u003cb\u003e109\u003c/b\u003e, 249\u0026ndash;258 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e9. Cannell, M. G. R. Physiology of the Coffee Crop. in \u003cem\u003eCoffee\u003c/em\u003e (eds. Clifford, M. N. \u0026amp; Willson, K. C.) 108\u0026ndash;134 (Springer US, Boston, MA, 1985). doi:10.1007/978-1-4615-6657-1_5.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e10. DaMatta, F. M., Ronchi, C. P., Maestri, M. \u0026amp; Barros, R. S. Ecophysiology of coffee growth and production. \u003cem\u003eBraz. J. Plant Physiol.\u003c/em\u003e \u003cb\u003e19\u003c/b\u003e, 485\u0026ndash;510 (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e11. DaMatta, F. M., Avila, R. T., Cardoso, A. A., Martins, S. C. V. \u0026amp; Ramalho, J. C. Physiological and Agronomic Performance of the Coffee Crop in the Context of Climate Change and Global Warming: A Review. \u003cem\u003eJ. Agric. Food Chem.\u003c/em\u003e \u003cb\u003e66\u003c/b\u003e, 5264\u0026ndash;5274 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e12. Dos Santos Soares, L., Teruel Rezende, T., Beijo, L. A. \u0026amp; Silva Franco J\u0026uacute;nior, K. Interaction between climate, flowering and production of dry coffee (Coffea arabica L.) in Minas Gerais. \u003cem\u003eCS\u003c/em\u003e \u003cb\u003e16\u003c/b\u003e, 1\u0026ndash;10 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e13. Peters, V. E. \u0026amp; Carroll, C. R. Temporal variation in coffee flowering may influence the effects of bee species richness and abundance on coffee production. \u003cem\u003eAgroforest Syst\u003c/em\u003e \u003cb\u003e85\u003c/b\u003e, 95\u0026ndash;103 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e14. Boreux, V., Kushalappa, C. G., Vaast, P. \u0026amp; Ghazoul, J. Interactive effects among ecosystem services and management practices on crop production: Pollination in coffee agroforestry systems. \u003cem\u003eProc. Natl. Acad. Sci. U.S.A.\u003c/em\u003e \u003cb\u003e110\u003c/b\u003e, 8387\u0026ndash;8392 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e15. De T. Alvim, P. Factors Affecting Flowering of Coffee. in \u003cem\u003eGenes, Enzymes, and Populations\u003c/em\u003e (ed. Srb, A. M.) 193\u0026ndash;202 (Springer US, Boston, MA, 1973). doi:10.1007/978-1-4684-2880-3_13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e16. Craparo, A. C. W., Van Asten, P. J. A., L\u0026auml;derach, P., Jassogne, L. T. P. \u0026amp; Grab, S. W. Coffea arabica yields decline in Tanzania due to climate change: Global implications. \u003cem\u003eAgricultural and Forest Meteorology\u003c/em\u003e \u003cb\u003e207\u003c/b\u003e, 1\u0026ndash;10 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e17. Craparo, A. C. W., Van Asten, P. J. A., L\u0026auml;derach, P., Jassogne, L. T. P. \u0026amp; Grab, S. W. Warm nights drive Coffea arabica ripening in Tanzania. \u003cem\u003eInt J Biometeorol\u003c/em\u003e \u003cb\u003e65\u003c/b\u003e, 181\u0026ndash;192 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e18. Kath, J. \u003cem\u003eet al.\u003c/em\u003e Not so robust: Robusta coffee production is highly sensitive to temperature. \u003cem\u003eGlobal Change Biology\u003c/em\u003e \u003cb\u003e26\u003c/b\u003e, 3677\u0026ndash;3688 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e19. DaMatta, F. M. \u0026amp; Ramalho, J. D. C. Impacts of drought and temperature stress on coffee physiology and production: a review. \u003cem\u003eBraz. J. Plant Physiol.\u003c/em\u003e \u003cb\u003e18\u003c/b\u003e, 55\u0026ndash;81 (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e20. Carr, M. K. V. THE WATER RELATIONS AND IRRIGATION REQUIREMENTS OF COFFEE. \u003cem\u003eEx. Agric.\u003c/em\u003e \u003cb\u003e37\u003c/b\u003e, 1\u0026ndash;36 (2001).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e21. Zacharias, A. O., Camargo, M. B. P. D. \u0026amp; Fazuoli, L. C. Modelo agrometeorol\u0026oacute;gico de estimativa do in\u0026iacute;cio da florada plena do cafeeiro. \u003cem\u003eBragantia\u003c/em\u003e \u003cb\u003e67\u003c/b\u003e, 249\u0026ndash;256 (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e22. Pezzopane, J. R. M., Salva, T. D. J. G., De Lima, V. B. \u0026amp; Fazuoli, L. C. Agrometeorological parameters for prediction of the maturation period of Arabica coffee cultivars. \u003cem\u003eInt J Biometeorol\u003c/em\u003e \u003cb\u003e56\u003c/b\u003e, 843\u0026ndash;851 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e23. Kath, J., Byrareddy, V. M., Reardon-Smith, K. \u0026amp; Mushtaq, S. Early flowering changes robusta coffee yield responses to climate stress and management. \u003cem\u003eScience of The Total Environment\u003c/em\u003e \u003cb\u003e856\u003c/b\u003e, 158836 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e24. De Beurs, K. M. \u0026amp; Henebry, G. M. Land surface phenology, climatic variation, and institutional change: Analyzing agricultural land cover change in Kazakhstan. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e \u003cb\u003e89\u003c/b\u003e, 497\u0026ndash;509 (2004).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e25. Reed, B. C., Schwartz, M. D. \u0026amp; Xiao, X. Remote Sensing Phenology. in \u003cem\u003ePhenology of Ecosystem Processes\u003c/em\u003e (ed. Noormets, A.) 231\u0026ndash;246 (Springer New York, New York, NY, 2009). doi:10.1007/978-1-4419-0026-5_10.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e26. Hatfield, J. L. \u003cem\u003eet al.\u003c/em\u003e Applications of Vegetative Indices from Remote Sensing to Agriculture: Past and Future. \u003cem\u003eInventions\u003c/em\u003e \u003cb\u003e4\u003c/b\u003e, 71 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e27. Dey, A. \u0026amp; Remesan, R. Assessing the Impact of Spatial Resolution on Land Surface Model Based on Hydrologic Simulations. in \u003cem\u003eClimate Change Impacts on Water Resources\u003c/em\u003e (eds. Jha, R., Singh, V. P., Singh, V., Roy, L. B. \u0026amp; Thendiyath, R.) vol. 98 493\u0026ndash;501 (Springer International Publishing, Cham, 2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e28. Tian, J., Zhu, X., Wu, J., Shen, M. \u0026amp; Chen, J. Coarse-Resolution Satellite Images Overestimate Urbanization Effects on Vegetation Spring Phenology. \u003cem\u003eRemote Sensing\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e, 117 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e29. Wang, H. \u003cem\u003eet al.\u003c/em\u003e Spatiotemporal crop NDVI responses to climatic factors in mainland China. \u003cem\u003eInternational Journal of Remote Sensing\u003c/em\u003e \u003cb\u003e40\u003c/b\u003e, 89\u0026ndash;103 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e30. Zhang, H. \u003cem\u003eet al.\u003c/em\u003e NDVI dynamic changes and their relationship with meteorological factors and soil moisture. \u003cem\u003eEnviron Earth Sci\u003c/em\u003e \u003cb\u003e77\u003c/b\u003e, 582 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e31. Grados, D., Reynarfaje, X. \u0026amp; Schrevens, E. A methodological approach to assess canopy NDVI\u0026ndash;based tomato dynamics under irrigation treatments. \u003cem\u003eAgricultural Water Management\u003c/em\u003e \u003cb\u003e240\u003c/b\u003e, 106208 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e32. Maselli, F. \u003cem\u003eet al.\u003c/em\u003e An improved NDVI-based method to predict actual evapotranspiration of irrigated grasses and crops. \u003cem\u003eAgricultural Water Management\u003c/em\u003e \u003cb\u003e233\u003c/b\u003e, 106077 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e33. Wu, G., Miller, N. D., De Leon, N., Kaeppler, S. M. \u0026amp; Spalding, E. P. Predicting Zea mays Flowering Time, Yield, and Kernel Dimensions by Analyzing Aerial Images. \u003cem\u003eFront. Plant Sci.\u003c/em\u003e \u003cb\u003e10\u003c/b\u003e, 1251 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e34. Zhang, Z. \u003cem\u003eet al.\u003c/em\u003e Dynamic variability of the heading\u0026ndash;flowering stages of single rice in China based on field observations and NDVI estimations. \u003cem\u003eInt J Biometeorol\u003c/em\u003e \u003cb\u003e59\u003c/b\u003e, 643\u0026ndash;655 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e35. d\u0026rsquo;Andrimont, R. \u003cem\u003eet al.\u003c/em\u003e Detecting flowering phenology in oil seed rape parcels with Sentinel-1 and \u0026minus;\u0026thinsp;2 time series. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e \u003cb\u003e239\u003c/b\u003e, 111660 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e36. Chen, B., Jin, Y. \u0026amp; Brown, P. An enhanced bloom index for quantifying floral phenology using multi-scale remote sensing observations. \u003cem\u003eISPRS Journal of Photogrammetry and Remote Sensing\u003c/em\u003e \u003cb\u003e156\u003c/b\u003e, 108\u0026ndash;120 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e37. Lin, P. \u003cem\u003eet al.\u003c/em\u003e A novel approach for estimating the flowering rate of litchi based on deep learning and UAV images. \u003cem\u003eFront. Plant Sci.\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e, 966639 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e38. Sulik, J. J. \u0026amp; Long, D. S. Automated detection of phenological transitions for yellow flowering plants such as Brassica oilseeds. \u003cem\u003eAgrosystems Geosci \u0026amp; Env\u003c/em\u003e \u003cb\u003e3\u003c/b\u003e, e20125 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e39. Zang, Y. \u003cem\u003eet al.\u003c/em\u003e Remote Sensing Index for Mapping Canola Flowers Using MODIS Data. \u003cem\u003eRemote Sensing\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e, 3912 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e40. Nogueira, S. M. C., Moreira, M. A. \u0026amp; Volpato, M. M. L. RELATIONSHIP BETWEEN COFFEE CROP PRODUCTIVITY AND VEGETATION INDEXES DERIVED FROM OLI / LANDSAT-8 SENSOR DATA WITH AND WITHOUT TOPOGRAPHIC CORRECTION. \u003cem\u003eEng. Agr\u0026iacute;c.\u003c/em\u003e \u003cb\u003e38\u003c/b\u003e, 387\u0026ndash;394 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e41. J\u0026uacute;nior, A. F. C., J\u0026uacute;nior, O. A. D. C., Martins, \u0026Eacute;. D. S. \u0026amp; Guerra, A. F. PHENOLOGICAL CHARACTERIZATION OF COFFEE CROP (Coffea arabica L.) FROM MODIS TIME SERIES. \u003cem\u003eRev. Bras. Geof.\u003c/em\u003e \u003cb\u003e31\u003c/b\u003e, 569 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e42. Nogueira Martins, R., De Carvalho Pinto, F. D. A., Mar\u0026ccedil;al De Queiroz, D., Magalh\u0026atilde;es Valente, D. S. \u0026amp; Fim Rosas, J. T. A Novel Vegetation Index for Coffee Ripeness Monitoring Using Aerial Imagery. \u003cem\u003eRemote Sensing\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e, 263 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e43. USDA. USDA Production, Supply and Distribution. https://apps.fas.usda.gov/psdonline/app/index.html#/app/downloads.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e44. Amarasinghe, U. A., Hoanh, C. T., D\u0026rsquo;haeze, D. \u0026amp; Hung, T. Q. Toward sustainable coffee production in Vietnam: More coffee with less water. \u003cem\u003eAgricultural Systems\u003c/em\u003e \u003cb\u003e136\u003c/b\u003e, 96\u0026ndash;105 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e45. Dinh, T. L. A., Aires, F. \u0026amp; Rahn, E. Statistical Analysis of the Weather Impact on Robusta Coffee Yield in Vietnam. \u003cem\u003eFront. Environ. Sci.\u003c/em\u003e \u003cb\u003e10\u003c/b\u003e, 820916 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e46. Kouadio, L. \u003cem\u003eet al.\u003c/em\u003e Performance of a process-based model for predicting robusta coffee yield at the regional scale in Vietnam. \u003cem\u003eEcological Modelling\u003c/em\u003e \u003cb\u003e443\u003c/b\u003e, 109469 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e47. Thao, N. T. T. \u003cem\u003eet al.\u003c/em\u003e Early Prediction of Coffee Yield in the Central Highlands of Vietnam Using a Statistical Approach and Satellite Remote Sensing Vegetation Biophysical Variables. \u003cem\u003eRemote Sensing\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e, 2975 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e48. Byrareddy, V., Kouadio, L., Mushtaq, S. \u0026amp; Stone, R. Sustainable Production of Robusta Coffee under a Changing Climate: A 10-Year Monitoring of Fertilizer Management in Coffee Farms in Vietnam and Indonesia. \u003cem\u003eAgronomy\u003c/em\u003e \u003cb\u003e9\u003c/b\u003e, 499 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e49. MODIS Terra Daily NDVI | Earth Engine Data Catalog. \u003cem\u003eGoogle for Developers\u003c/em\u003e https://developers.google.com/earth-engine/datasets/catalog/MODIS_MOD09GA_006_NDVI.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e50. \u003cem\u003eCoffee: Growing, Processing, Sustainable Production: A Guidebook for Growers, Processors, Traders, and Researchers\u003c/em\u003e. (Wiley, 2004). doi:10.1002/9783527619627.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e51. Copernicus Climate Change Service (C3S). ERA5: Fifth generation of ECMWF atmospheric reanalyses of the global climate. https://cds.climate.copernicus.eu/cdsapp#!/home (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e52. Wood, S. N. Fast Stable Restricted Maximum Likelihood and Marginal Likelihood Estimation of Semiparametric Generalized Linear Models. \u003cem\u003eJournal of the Royal Statistical Society Series B: Statistical Methodology\u003c/em\u003e \u003cb\u003e73\u003c/b\u003e, 3\u0026ndash;36 (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e53. BARTON, K. MuMIn : multi-model inference. \u003cem\u003ehttp://r-forge.r-project.org/projects/mumin/\u003c/em\u003e (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e54. Lefcheck, J. S. \u003cspan type=\"SmallCaps\" class=\"SmallCaps\" name=\"Emphasis\"\u003epiecewiseSEM\u003c/span\u003e : Piecewise structural equation modelling in \u003cspan type=\"SmallCaps\" class=\"SmallCaps\" name=\"Emphasis\"\u003er\u003c/span\u003e for ecology, evolution, and systematics. \u003cem\u003eMethods Ecol Evol\u003c/em\u003e \u003cb\u003e7\u003c/b\u003e, 573\u0026ndash;579 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e55. Ravindra, K., Rattan, P., Mor, S. \u0026amp; Aggarwal, A. N. Generalized additive models: Building evidence of air pollution, climate change and human health. \u003cem\u003eEnvironment International\u003c/em\u003e \u003cb\u003e132\u003c/b\u003e, 104987 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e56. Wellington, M. J., Lawes, R. \u0026amp; Kuhnert, P. A framework for modelling spatio-temporal trends in crop production using generalised additive models. \u003cem\u003eComputers and Electronics in Agriculture\u003c/em\u003e \u003cb\u003e212\u003c/b\u003e, 108111 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e57. Wikle, C. K., Zammit-Mangion, A. \u0026amp; Cressie, N. \u003cem\u003eSpatio-Temporal Statistics with R\u003c/em\u003e. (Chapman and Hall/CRC, Boca Raton, Florida : CRC Press, [2019], 2019). doi:10.1201/9781351769723.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e58. R: The R Project for Statistical Computing. https://www.r-project.org/.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e59. Dormann, C. F. \u003cem\u003eet al.\u003c/em\u003e Collinearity: a review of methods to deal with it and a simulation study evaluating their performance. \u003cem\u003eEcography\u003c/em\u003e \u003cb\u003e36\u003c/b\u003e, 27\u0026ndash;46 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e60. Gelman, A. \u0026amp; Hill, J. \u003cem\u003eData Analysis Using Regression and Multilevel/Hierarchical Models\u003c/em\u003e. (Cambridge University Press, 2006). doi:10.1017/CBO9780511790942.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e61. Gomez, C. \u003cem\u003eet al.\u003c/em\u003e Shift in precipitation regime promotes interspecific hybridization of introduced \u003cem\u003eCoffea\u003c/em\u003e species. \u003cem\u003eEcology and Evolution\u003c/em\u003e \u003cb\u003e6\u003c/b\u003e, 3240\u0026ndash;3255 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e62. Byrareddy, V. \u003cem\u003eet al.\u003c/em\u003e Win-win: Improved irrigation management saves water and increases yield for robusta coffee farms in Vietnam. \u003cem\u003eAgricultural Water Management\u003c/em\u003e \u003cb\u003e241\u003c/b\u003e, 106350 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e63. Abreu J\u0026uacute;nior, C. A. M. D. \u003cem\u003eet al.\u003c/em\u003e Estimating Coffee Plant Yield Based on Multispectral Images and Machine Learning Models. \u003cem\u003eAgronomy\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e, 3195 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e64. Bola\u0026ntilde;os, J., Corrales, J. C. \u0026amp; Campo, L. V. Feasibility of Early Yield Prediction per Coffee Tree Based on Multispectral Aerial Imagery: Case of Arabica Coffee Crops in Cauca-Colombia. \u003cem\u003eRemote Sensing\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e, 282 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e65. Pettorelli, N. \u003cem\u003eet al.\u003c/em\u003e Using the satellite-derived NDVI to assess ecological responses to environmental change. \u003cem\u003eTrends in Ecology \u0026amp; Evolution\u003c/em\u003e \u003cb\u003e20\u003c/b\u003e, 503\u0026ndash;510 (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e66. Feng, J. \u003cem\u003eet al.\u003c/em\u003e Temporal and Spatial Variation Characteristicsof NDVI and Its Relationshipwith Environmental Factors in Huangshui RiverBasin from 2000 to 2018. \u003cem\u003ePol. J. Environ. Stud.\u003c/em\u003e \u003cb\u003e30\u003c/b\u003e, 3043\u0026ndash;3063 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e67. Du, G., Yan, S., Chen, H., Yang, J. \u0026amp; Wen, Y. Intra-Annual Cumulative Effects and Mechanisms of Climatic Factors on Global Vegetation Biomes\u0026rsquo; Growth. \u003cem\u003eRemote Sensing\u003c/em\u003e \u003cb\u003e16\u003c/b\u003e, 779 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e68. Salamanca-Jimenez, A., Doane, T. A. \u0026amp; Horwath, W. R. Nitrogen Use Efficiency of Coffee at the Vegetative Stage as Influenced by Fertilizer Application Method. \u003cem\u003eFront. Plant Sci.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e69. Vilela, M. S. \u003cem\u003eet al.\u003c/em\u003e Nitrogen, phosphorus, and potassium fertilization on the incidence of brown eye spot in coffee crop in vegetative stage. \u003cem\u003eTrop. plant pathol.\u003c/em\u003e \u003cb\u003e47\u003c/b\u003e, 672\u0026ndash;684 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e70. Ennouri, K., Triki, M. A. \u0026amp; Kallel, A. Applications of Remote Sensing in Pest Monitoring and Crop Management. in \u003cem\u003eBioeconomy for Sustainable Development\u003c/em\u003e (ed. Keswani, C.) 65\u0026ndash;77 (Springer Singapore, Singapore, 2020). doi:10.1007/978-981-13-9431-7_5.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e71. Pinter, Jr., P. J. \u003cem\u003eet al.\u003c/em\u003e Remote Sensing for Crop Management. \u003cem\u003ephotogramm eng remote sensing\u003c/em\u003e \u003cb\u003e69\u003c/b\u003e, 647\u0026ndash;664 (2003).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e72. Jin, H. \u0026amp; Eklundh, L. A physically based vegetation index for improved monitoring of plant phenology. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e \u003cb\u003e152\u003c/b\u003e, 512\u0026ndash;525 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e73. Araya, S., Ostendorf, B., Lyle, G. \u0026amp; Lewis, M. CropPhenology: An R package for extracting crop phenology from time series remotely sensed vegetation index imagery. \u003cem\u003eEcological Informatics\u003c/em\u003e \u003cb\u003e46\u003c/b\u003e, 45\u0026ndash;56 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e74. Liu, L. \u003cem\u003eet al.\u003c/em\u003e Detecting crop phenology from vegetation index time-series data by improved shape model fitting in each phenological stage. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e \u003cb\u003e277\u003c/b\u003e, 113060 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e75. Behnke, R. \u003cem\u003eet al.\u003c/em\u003e Evaluation of downscaled, gridded climate data for the conterminous United States. \u003cem\u003eEcological Applications\u003c/em\u003e \u003cb\u003e26\u003c/b\u003e, 1338\u0026ndash;1351 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e76. Tarek, M., Brissette, F. \u0026amp; Arsenault, R. Uncertainty of gridded precipitation and temperature reference datasets in climate change impact studies. \u003cem\u003eHydrol. Earth Syst. Sci.\u003c/em\u003e \u003cb\u003e25\u003c/b\u003e, 3331\u0026ndash;3350 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e77. Fawcett, D., Bennie, J. \u0026amp; Anderson, K. Monitoring spring phenology of individual tree crowns using drone-acquired NDVI data. \u003cem\u003eRemote Sens Ecol Conserv\u003c/em\u003e \u003cb\u003e7\u003c/b\u003e, 227\u0026ndash;244 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e78. Ge, W., Li, X., Jing, L., Han, J. \u0026amp; Wang, F. Monitoring canopy-scale autumn leaf phenology at fine-scale using unmanned aerial vehicle (UAV) photography. \u003cem\u003eAgricultural and Forest Meteorology\u003c/em\u003e \u003cb\u003e332\u003c/b\u003e, 109372 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e79. Liu, J., Fan, J., Yang, C., Xu, F. \u0026amp; Zhang, X. Novel vegetation indices for estimating photosynthetic and non-photosynthetic fractional vegetation cover from Sentinel data. \u003cem\u003eInternational Journal of Applied Earth Observation and Geoinformation\u003c/em\u003e \u003cb\u003e109\u003c/b\u003e, 102793 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e80. Taugourdeau, S. \u003cem\u003eet al.\u003c/em\u003e Leaf area index as an indicator of ecosystem services and management practices: An application for coffee agroforestry. \u003cem\u003eAgriculture, Ecosystems \u0026amp; Environment\u003c/em\u003e \u003cb\u003e192\u003c/b\u003e, 19\u0026ndash;37 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnnex\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Table Details","content":"\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003eTable\u0026nbsp;1. Summary of variables included in the study.\u003c/em\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003eTable\u0026nbsp;2. Performance of the best model and models with climate and vegetation variables only\u003c/em\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003e*: The best model based on AIC ranking includes climate predictors (rainfall at flowering dates, accumulated rainfall and rolling mean minimum temperature, and maximum temperature of 30 days prior to flowering date), and vegetation predictors (NDVI at flowering date, and mean NDVI of 30, 90, and 365 days prior to flowering date). The marginal effects of each predictor in the best model on flowering anomalies were presented in Fig.\u0026nbsp;3. A detailed statistical description of all variables in the best model is presented in Table A1 in the Annex.\u003c/em\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003eTable\u0026nbsp;3. Key vegetation index indicators to coffee flowering anomaly\u003c/em\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003eTable\u0026nbsp;4. Direct, indirect, and total effects of climate, vegetation, and management factors on flowering anomalies. \u0026ldquo;n.s.\u0026rdquo; indicates the effect is not statistically significant. The total effects in bold are the top three predictors with the highest total effects on flowering anomalies. Direct effects are the standardized regression coefficients of each predictor on the response variable. Indirect effects are the product of the direct effects operating along causal pathways in SEM models. The total effects are the sum of the predictor\u0026rsquo;s direct effect and all its indirect effects through intermediary variables.\u003c/em\u003e\u003c/p\u003e\u003cp\u003eAnnex\u003c/p\u003e\u003cp\u003eTable A1. Statistical description of variables in the global model.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-sustainable-agriculture","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [npj Sustainable Agriculture](https://www.nature.com/npjsustainagric/)","snPcode":"44264","submissionUrl":"https://submission.springernature.com/new-submission/44264/3","title":"npj Sustainable Agriculture","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Coffee, Remote Sensing, Normalized Difference Vegetation Index (NDVI), Climate, Generalized Additive Model (GAM), Structural Equation Modelling (SEM)","lastPublishedDoi":"10.21203/rs.3.rs-5018229/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5018229/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eQuantifying the timing of vegetation growth, particularly coffee plant flowering, is vital for estimating yield in advance. While satellite-based vegetation indices are effective in mapping crop growth and have a strong correlation with coffee yield, the potential contribution of plant conditions alongside climate factors in predicting coffee flowering anomalies remains underexplored.\u003c/p\u003e \u003cp\u003eHere, our investigation aimed to determine whether satellite-based vegetation indices, in addition to climate variables, could enhance the model's predictive power for flowering anomalies of coffee trees. Utilizing a dataset on flowering dates over ten years of 558 coffee farms spread over four provinces (Dak Lak, Gia Lai, Dak Nong, and Lam Dong) in the Central Highlands of Vietnam, we analysed climate variables (rainfall and temperature) and the Normalized Difference Vegetation Index (NDVI) at various intervals prior to flowering dates. Using a Generalized Additive Model (GAM) and model selection based on Akaike\u0026rsquo;s Information Criteria (AIC), we identified the most influential predictors. Then, we performed Structural Equation Modelling (SEM) to further investigate the complex causal relationships among flowering anomalies, climate, vegetation, and management factors.\u003c/p\u003e \u003cp\u003eOur results show that the NDVI prior to flowering dates held the most explanatory power, outperforming climate variables. Lower NDVI during the dormancy period indicated the ripe-to-flower condition of the coffee tree, informing earlier onset of the flowering stage, while higher NDVI during bud initiation and development stage suggested a delayed flowering. The best model incorporating both climate and NDVI predictors achieved good explanatory performance with an adjusted R\u003csup\u003e2\u003c/sup\u003e of 0.87.\u003c/p\u003e \u003cp\u003eThe analysis highlighted the advantages of vegetation indices over climate predictors in capturing plant conditions through its growing cycle, with the accumulated effects of environmental factors and agricultural management activities, especially during critical phenological stages. Our findings suggest further studies utilising vegetation indices from remote sensing data sources at multiple scales to thoroughly understand plant conditions at different crop growth phases, especially at early stages, for site-specific, timely and strategic management interventions.\u003c/p\u003e","manuscriptTitle":"Vegetation growth conditions strongly indicate coffee flowering anomalies","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-12-05 03:02:52","doi":"10.21203/rs.3.rs-5018229/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-01-05T09:37:28+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-01-02T10:53:28+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"245207180905494463545650091371059181408","date":"2024-12-03T10:14:17+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-11-22T10:19:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"258581681300924501777266419546859794627","date":"2024-11-22T05:03:15+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"57126428394692511250917096257969907332","date":"2024-11-22T03:20:58+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"261252434721296127940586027358391313691","date":"2024-10-16T07:36:38+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-09-25T02:14:25+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-09-25T01:51:51+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-09-22T13:18:41+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Sustainable Agriculture","date":"2024-09-02T12:12:39+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-sustainable-agriculture","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [npj Sustainable Agriculture](https://www.nature.com/npjsustainagric/)","snPcode":"44264","submissionUrl":"https://submission.springernature.com/new-submission/44264/3","title":"npj Sustainable Agriculture","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"d96b5cc2-bda3-4f7a-a155-42c2e48987a3","owner":[],"postedDate":"December 5th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[{"id":41081323,"name":"Biological sciences/Ecology/Ecological modelling"},{"id":41081324,"name":"Earth and environmental sciences/Ecology/Agri ecology"},{"id":41081325,"name":"Earth and environmental sciences/Ecology/Climate change ecology/Phenology"}],"tags":[],"updatedAt":"2025-01-05T09:53:05+00:00","versionOfRecord":[],"versionCreatedAt":"2024-12-05 03:02:52","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5018229","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5018229","identity":"rs-5018229","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}