Assessing Vegetation Responses to Dam-Induced Hydrological Change: A 40-Year Landsat Time-Series Analysis in the Downstream Serrahis River Watershed, Cyprus

Assessing Vegetation Responses to Dam-Induced Hydrological Change: A 40-Year Landsat Time-Series Analysis in the Downstream Serrahis River Watershed, Cyprus | 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 Research Article Assessing Vegetation Responses to Dam-Induced Hydrological Change: A 40-Year Landsat Time-Series Analysis in the Downstream Serrahis River Watershed, Cyprus Marzia Gabriele This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6761511/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This study presents a methodological framework to assess the impact of dam construction on vegetation dynamics, using the Klirou–Malounta–Akaki Dam in Cyprus as a case study. Unlike many dams that incorporate controlled flood pulses, this structure lacks a designed flood regime, raising the question of whether observed changes in riparian vegetation are primarily driven by localized microclimatic or geomorphological factors rather than direct hydrological alterations. A temporal analysis covering 20 years before and after dam construction (1984–2024) was implemented to distinguish baseline vegetation conditions from post-construction shifts. The analysis focused on an upstream area within the Serrahis River watershed. Enhanced Vegetation Index (EVI) time series, derived from Landsat 5, 7, and 8 and processed via Google Earth Engine and R, were harmonized using a Random Forest cross-sensor calibration approach. Pixel-wise trend analysis at 30 m resolution revealed cluster areas of vegetation increase, suggesting localized greening effects. However, ground-truthing surveys revealed that such increases were often associated with the spread of non-native Eucalyptus spp., a drought-tolerant species adapted to stable or intermittent surface flows. While the EVI trends may imply ecosystem recovery, they also indicate a shift in species composition, potentially favouring resilient non-native taxa at the expense of native species reliant on seasonal flooding. These findings underscore the importance of interpreting vegetation indices in ecological context. Although stabilized hydrology may enhance water storage, it can simultaneously drive biodiversity loss, highlighting the need for balanced water management strategies in water scarce Mediterranean environments. Landsat GEE Riverine Ecosystem Drylands Ecology Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Introduction The Klirou–Malounta–Akaki Dam is in central Cyprus (Nicosia District) on the Akaki River, a tributary in the Serrachis River basin. Geographically, it sits about 25–26 km southwest of Nicosia, in the northern foothills of the Troodos Mountains​ (Fig. 1 ). The dam lies between the villages of Klirou and Malounta (which are approximately 2 km apart), at an elevation of roughly 300–350 m above sea level. The surrounding area is a transitional zone between the Troodos uplands and the Mesaoria plain. Upstream (southward) from the dam, the terrain rises toward the Troodos range (including the Papoutsa and Farmakas peaks), which provides the catchment for the Akaki River. Downstream (northward), the river flows through a narrow valley that opens into the broad Nicosia/Mesaoria plain on its way to Morphou Bay on the northwestern coast. In its middle reaches near Klirou and Malounta, the Akaki River has incised a small gorge into the alluvial plain, around which fertile agricultural lands and village settlements are established ( WDD, 2019). The region’s climate is Mediterranean semi-arid. Mean annual rainfall in the Klirou/Malounta area is on the order of 400–500 mm, most of it occurring in winter. Summers are typically dry with zero flow in the river. The Akaki is an ephemeral river – it experiences seasonal flow during winter rains and remains dry for much of the rest of the year. Prior to dam construction, the Akaki River’s flow regime was characterized by flashy winter floods (in wet years the river could carry substantial runoff from the 84 km² upstream catchment) followed by rapid recession. Now, the Klirou Dam intercepts and regulates these flows. The dam’s reservoir (often referred to as Klirou Reservoir or Kalo Chorio Reservoir) has a full storage volume of ~ 2.0 million m³ and a surface area of ~ 18 ha when full (Fig. 2 ; Fig. 3 ). For context, the mean annual flow of the Akaki River at the dam site is estimated at 12 million m³, which indicates that in an average year not all runoffs can be stored and some will spill. The dam is an earthfill structure with a central clay core, approximately 38 m high and 265 m in crest length. It became operational in 2007 (WDD, 2007)​. By design, this is a multi-purpose small dam serving mainly irrigation supply and groundwater recharge. There are no hydroelectric facilities, as Cyprus has no hydroelectric dams due to its limited hydrology. Instead, water stored in Klirou Reservoir is intended for controlled release to irrigation networks and to replenish the downstream aquifer in the Akaki valley. The Water Development Department (WDD) manages the dam’s operation and monitors groundwater levels to gauge recharge benefits. Importantly, the dam was not built with a dedicated spillway for continuous ecological flow releases – like most Cypriot dams of its era, it can release water through valves manually, but it does not simulate natural flow patterns on its own (WDD, 2007). This means the downstream river stretch often remains dry until either deliberate releases are made for irrigation/recharge, or the reservoir overtops in a heavy rain event. Several overflow events have been recorded in exceptionally wet winters (e.g. the dam spilled in winter 2018–2019 after heavy rains​), briefly restoring flow to the downstream channel. Study Area The land use in the downstream area is a mixture of agricultural and semi-natural riparian landscapes. The villages of Klirou, Malounta, and Akaki (further downstream ~ 10 km north) are traditional farming communities. Surrounding these settlements are cultivated fields, including olive groves, citrus orchards, grain fields, and some vegetable cultivation​. Agriculture in this region relies partly on groundwater and partly on surface water where available. The presence of the river (even if intermittent) has historically supported agriculture by providing alluvial soils and shallow groundwater; farmers have dug wells in the plains where the water table is accessible​. Livestock farming (e.g. sheep, goats) is also common in the area, taking advantage of the grazing in valley bottoms. The riparian corridor of the Akaki River area of interest of this study, (Fig. 4 ), though seasonally dry, hosts distinctive Mediterranean streamside vegetation. This includes shrubs and trees adapted to infrequent water pulses, such as tamarisk (Tamarix spp.), oleander (Nerium oleander), reeds (Phragmites spp.). Scattered riverbank stands of plane trees (Platanus orientalis) and eucalyptus (Eucalyptus spp.) occur in moister stream segments, with eucalyptus being an introduced, fast-growing taxon that tolerates periods of low surface flow by tapping deeper soil moisture. Because of its robust root systems and moderate drought tolerance, eucalyptus has been widely used across the Mediterranean to stabilize riparian embankments, reduce erosion, and provide timber or windbreaks. Much of the year, the riverbed may appear as a dry channel with patches of green where deep-rooted vegetation taps residual moisture. After winter rains, ephemeral herbs and grasses briefly flourish. Upstream of the dam, the landscape is more rugged and natural: the catchment area extends into forested slopes and shrubland of the Troodos foothills, which are part of a protected forest zone. If the dam completely cuts off ordinary flows in average years, water stress on downstream flora could increase, causing drought-tolerant species to dominate and any water-dependent species to decline. Moreover, Fig. 5 exemplifies panoramically the morphology of the Area of Interest (AOI) near the E929 road crossing. The drone perspective highlights seasonally dry channels, alluvial soils, and riparian vegetation in the corridor downstream of the dam. Much of the year, the riverbed remains barren, with deep-rooted vegetation persisting on residual moisture and ephemeral herbs flourishing briefly after winter rains. Analysing vegetation indices (which correlates with plant biomass and vigour) over time, can capture these effects quantitatively. Any observable trend downstream of the dam – relative to upstream areas – might indicate the positive influence of sustained moisture. Conversely, a decline in greenness could signal deteriorating riparian health due to reduced surface flows. Understanding these outcomes is crucial for water management: it will inform whether environmental flow provisions might be necessary at small dams and how current dam operations could be adjusted to balance human use with ecosystem needs, which after dam construction might have further skew toward drought-tolerant ecological matrixes. As for that, the study area’s mix of Mediterranean climate, ephemeral river hydrology, dam intervention, and dependent agriculture offers a microcosm to study vegetation–hydrology interactions under water stress conditions. Insights gained here will contribute to the broader knowledge of how dams in semi-arid regions affect downstream ecology, which is increasingly important as water resources scarcity is driving more infrastructure projects in similar environments (Hadjinicolaou et al., 2011 ). Problem Statement Dams worldwide significantly alter natural flow regimes and sediment transport, triggering downstream ecological changes such as wetland formation or shifts in riparian plant communities (Nilsson et al., 2005 ; Bunn & Arthington, 2002 ). Programmed flood releases are often employed to mitigate such effects by mimicking natural disturbance regimes (Poff et al., 2007 ). However, in the case of the Klirou–Malounta–Akaki Dam, no such flood pulses have been implemented. Meanwhile, precipitation trends have remained relatively stable (Hadjinicolaou et al., 2011 ), suggesting that observed vegetation changes may stem more from local geomorphological or microclimatic drivers than from major hydrological reconfiguration (Tockner & Stanford, 2002 ; Petts & Gurnell, 2005 ). This aligns with emerging perspectives on intermittent river systems where ecological dynamics are often shaped by complex interactions beyond flow alone (Larned et al., 2010 ). Objective: Investigate vegetation changes (via EVI time-series) pre- and post-dam construction. Identify pixels with positive and negative trends, to assess whether these trends might correlate with the presence and operation of the dam. Determine the most positive trends to point to identify specific areas where to plan on-field survey and investigate the in-place ecological processes and species composition. Data and Materials Satellite Data (Landsat) The lsat_export_ts() function utilizes the Landsat Collection 2, Level-2 Surface Reflectance (SR) datasets to extract long-term time series of surface reflectance from multiple Landsat missions. Specifically, it accesses six Earth Engine ImageCollections corresponding to Tier 1 and Tier 2 datasets from Landsat 5 (Thematic Mapper, TM), Landsat 7 (Enhanced Thematic Mapper Plus, ETM+), and Landsat 8 (Operational Land Imager, OLI). These are identified as: LANDSAT/LT05/C02/T1_L2, LANDSAT/LT05/C02/T2_L2 LANDSAT/LE07/C02/T1_L2, LANDSAT/LE07/C02/T2_L2 LANDSAT/LC08/C02/T1_L2, LANDSAT/LC08/C02/T2_L2 These datasets provide atmospherically corrected surface reflectance values at 30-meter spatial resolution. For Landsat 8, the Land Surface Reflectance Code (LaSRC) is applied (Vermote et al., 2016), while Landsat 5 and 7 employ the LEDAPS (Landsat Ecosystem Disturbance Adaptive Processing System) algorithm (Masek et al., 2006). Each image contains calibrated spectral reflectance for bands 1 through 7 (visible to shortwave infrared), alongside quality assurance bands such as QA_PIXEL and QA_RADSAT, which enable cloud, shadow, and saturation filtering. Metadata attributes including acquisition date, solar elevation, spacecraft ID, and processing level are also retained during export to support contextual and temporal analysis. The function overlays the max_extent band from the Global Surface Water (GSW) dataset developed by the Joint Research Centre, accessible via Earth Engine as JRC/GSW1_4/GlobalSurfaceWater. This dataset offers high-resolution information on the historical maximum extent of surface water derived from over 3 million Landsat scenes spanning the period 1984 to the present (Pekel et al., 2016). Its integration enables enhanced interpretation of water-related land cover dynamics in conjunction with vegetation indices. To ensure consistent data structure across the time series, the function pads missing bands in older images with a placeholder band (e.g., null-valued SR_B6) and coerces all outputs to floating-point format. This guarantees compatibility during pixel-wise statistical analysis or trend detection workflows. To evaluate data availability across sites and years, a summary table is generated, and a vegetation index (EVI) is computed to characterize vegetation status over time (Gurung et al., 2009; Shi et al, 2017; Zhao et al., 2009; Nagler et al., 2005; Coelho et al., 2024). Vegetation Index: Enhanced Vegetation Index (EVI) computed as: 𝐸𝑉𝐼=2.5×𝜌𝑛𝑖𝑟−𝜌𝑟𝜌𝑛𝑖𝑟+6.0𝜌𝑟−7.5𝜌𝑏+1 (1) Where 𝜌𝑛𝑖𝑟, 𝜌𝑟, and 𝜌𝑏 stand for the atmospherically corrected surface reflectance of the near-infrared, red (visible) and blue (visible) bands of Landsat data. Cross-sensor reflectance calibration is conducted to harmonize data from Landsat 5, 7, and 8 using a Random Forest modelling approach, focusing on observations collected between days of the year 151 and 242. This calibration, anchored by Landsat 7 as the temporal reference, ensures comparability among multiple Landsat missions. For the subsequent phenological and trend analyses, only cloud-free observations (below a defined 5% threshold) from clear-sky conditions are retained. Methodology The methodological workflow for Landsat Time-Series Extraction and Phenological Trend Analysis is developed by loading a suite of R packages, namely sf, dplyr, purrr, data.table, stringr, rgee (Aybar et al., 2020), and LandsatTS (Berner et al., 2023), o enable spatial data handling, data manipulation, and interaction with Google Earth Engine (GEE). After initializing the GEE (Gorelick et al., 2017) environment, Landsat 8 grid pixel canters are identified within the target polygons by employing a buffer-based function designed to avoid excessively large polygons, which can lead to computational inefficiencies. Pixel centers are assigned unique identifiers in a column such as “site_id” or “sample.id,” ensuring that subsequent functions recognize and distinguish among different locations. Following determination of point coordinates, the lsat_export_ts() function is used to export the corresponding Landsat surface reflectance data from GEE. This function packages the coordinates into sf point features, submits export tasks, and saves CSV files in a designated Google Drive folder. Once these tasks complete, the resulting CSV files are downloaded locally and imported into R for further processing, using data.table::fread() to handle large files efficiently. After loading the data, lsat_format_data() or equivalent steps are used to standardize column names, confirm the presence of a unique “sample.id,” and apply any necessary scaling to reflectance bands. The dataset “lsat.dt” was cleaned by the “lsat_clean_data” function. In this step, a maximum geometric registration error of 15 meters, a maximum allowable cloud cover of 10%, and a maximum solar zenith angle of 60° were specified. Additionally, filtering was enabled to exclude CFMask-snow, CFMask-water, and water pixels identified by the JRC Global Surface Water dataset, ensuring that only high-quality surface reflectance observations remained in the updated “lsat.dt” dataset. Data filtering removes suboptimal observations containing excessive cloud cover, snowfall, or water contamination. This produces a qualitative dataset of surface reflectance measurement, suitable for riverine vegetation monitoring. The lsat_summarize_data() function provides an overview of data availability, including temporal coverage, annual measurement frequencies, and plots of observation distributions, enabling to verify the completeness and reliability of the dataset. After extraneous observations removed, the vegetation index Enhanced Vegetation Index (EVI) is computed through lsat_calc_spectral_index(). This step parses the required reflectance bands, applies sensor-specific scaling factors, and outputs consistently named columns for further analysis. Data are acquired from multiple Landsat sensors (Landsat 5, 7, and 8), whereas the cross-calibration becomes a crucial methodological step. Cross-calibration of spectral reflectance or indices is performed using the “lsat_calibrate_rf” function, and by specifing a day-of-year range that targets the mid-growing season (in this case: days 151 to 242, corresponding to late May through late August, depending on leap years). This range captures a period of relatively stable phenological conditions and is used to determine typical reflectance values. Landsat 7 serves as a temporal benchmark given its overlap with Landsat 5 and 8, and cross-calibration is conducted on a single band or spectral index at a time using data from thousands sample sites. A Random Forest model is then trained to predict Landsat 7 reflectance from Landsat 5 or 8 reflectance recorded in the same years. Parameters such as the minimum number of observations required per site (in this case: 5), whether to include high-latitude data, the fraction of data (in this case: 0.75) to use for model training, and whether to overwrite existing columns can be tuned based on the resulting root mean square error (RMSE) to optimize calibration accuracy (in this case: TRUE). Table 1 Calibration Coefficients Satellite Band/SI B0 B1 B2 B3 B0.se B1.se B2.se B3.se LANDSAT_5 evi 0,0088 1,0495 -0.086 NA 0,0016 0,0102 0,0152 NA LANDSAT_8 evi -0,0093 0,9109 0.322 -0,319 0,003 0,0276 0,0784 0,0681 Table 2 Model Evaluation Satellite Band/SI Train Sites Eval Sites R² Uncal. RMSE Uncal. Bias Uncal. Bias (%) Xcal. RMSE Xcal. Bias Xcal. Bias (%) LANDSAT_5 evi 5291 1764 0,958 0,030 -0,014 -4,5 0,026 0,001 0,4 LANDSAT_8 evi 5964 1988 0,954 0,032 0,016 5,1 0,028 0,001 0,4 The two tables (Table 1, Table 2) present both the coefficients of the cross-calibration model for Enhanced Vegetation Index (EVI) and the corresponding validation statistics after harmonizing Landsat 5 and Landsat 8 to a Landsat 7 reference baseline. In the upper table, columns labeled “B0,” “B1,” “B2,” and “B3” represent the fitted polynomial coefficients that map EVI values from Landsat 5 or Landsat 8 to those of Landsat 7. These coefficients and their standard errors (“B0.se,” “B1.se,” “B2.se,” “B3.se”) indicate the magnitude and uncertainty of each term in the calibration function. For instance, Landsat 5 uses B0 = 0.0088, B1 = 1.0495, and B2 = − 0.086 for its calibration polynomial. Meanwhile, Landsat 8 employs a fourth-degree polynomial, evidenced by the presence of B3 = − 0.319 (with the other fitted terms given by B0, B1, and B2). The difference in polynomial order arises because the Random Forest–based model selected these coefficients to minimize calibration error, indicating complex inter-sensor relationships for Landsat 8. The lower table summarizes how well each calibration performs. “train.n.sites” and “eval.n.sites” list the number of point locations used, respectively, for training and for out-of-sample evaluation. The “r2” column shows the coefficient of determination, reflecting the proportion of variance explained by the model. The “uncal.rmse” and “uncal.bias” columns provide baseline error metrics for raw (uncalibrated) EVI, while “xcal.rmse” and “xcal.bias” report those metrics after calibration. Negative uncalibrated bias (as for Landsat 5, − 0.014) signifies that EVI estimates tended to be lower than Landsat 7, whereas a positive bias (for Landsat 8, 0.016) indicates an overestimation. The percentage columns (e.g., “uncal.bias.pcnt” and “xcal.bias.pcnt”) present those biases relative to Landsat 7. For Landsat 5, cross-calibration reduced the RMSE from 0.030 to 0.026 and bias from − 0.014 to 0.001. Landsat 8 experienced a similar improvement, with RMSE dropping from 0.032 to 0.028 and bias shrinking from 0.016 to 0.001. The final bias percentages for both sensors approximate 0.4%, indicating that the Random Forest–based approach successfully aligned EVI values with the Landsat 7 reference to within minimal error margins. Following cross-calibration of the Enhanced Vegetation Index (EVI) using lsat_calibrate_rf(), a CSV file was generated containing time series data from the 36 distinct pixels, each corresponding to a specific geographic location. The dataset incorporated crucial variables, including a unique sample ID (“sample.id”), the calendar year (“year”), the day of the year (“doy”), the EVI value (“evi”), the originating satellite (“satellite”), the percentage of cloud cover (“cloud.cover”), and a binary indicator of whether the observation was cloud-free (“clear”). To ensure data integrity in the time series construction, only records meeting strict quality thresholds were retained (specifically, cloud.cover < 5 and clear = = 1), and any observations with missing or invalid fields were removed prior to further analysis. Vegetation Time-series Analysis To identify representative vegetation dynamics across the study area, a custom procedure was implemented to select key pixels based on long-term EVI trends and temporal variability. For each pixel, a linear regression model was fitted to the EVI time series using acquisition date as the predictor variable. The slope of this model was extracted to quantify the direction and intensity of a long-term vegetation change (1984–2024): positive slopes indicated greening trends, while negative slopes pointed to degradation (Forkel et al., 2015; DeVries et al., 2020). The temporal variability of each pixel was assessed by calculating the standard deviation of EVI values, serving as an indicator of phenological fluctuation (Seddon et al., 2016). To visualize the slope value trend direction and intensity, a spatial representation of pixel-wise vegetation trends was developed using a GIS environment (QGIS 3.28). A graduated symbology was applied to the slope values extracted from the EVI time series, using a diverging colour ramp to represent trends from strong negative, to strong positive, and the size to determine the magnitude of the trend. Five trend classes were defined and visualized: Strong Negative, Mild Negative, Stable, Mild Positive, and Strong Positive. The resulting map provides a clear overview of spatial variability in long-term vegetation dynamics across the study area. To enable targeted time-series analysis, the top five pixels exhibiting strong negative and strong positive vegetation trends were extracted based on slope values. From each group, two representative pixels were selected, and their Enhanced Vegetation Index (EVI) time series was aggregated into 15-day (semi-monthly) intervals to capture seasonal dynamics and long-term patterns. Specifically, all valid EVI observations were grouped by pixel and year into 24 fixed bins corresponding to 15-day periods (DOY 15 to 360) and averaged within each bin to reduce high-frequency noise caused by cloud interference, sensor variation, or rapid short-term disturbances. This method preserves the seasonal signature while enabling robust interannual comparisons across decades (Seddon et al., 2016). The aggregated EVI curves were then smoothed using Generalized Additive Models (GAMs) with cubic regression splines (s(x, bs = "cs")), which provide flexible, data-adaptive fits while avoiding overfitting. This modelling strategy is widely adopted for land surface phenology analysis using moderate-resolution optical imagery (Forkel et al., 2015), especially in cases of non-linear vegetation response or in dryland ecosystems where phenological patterns are fragmented and noisy (DeVries et al., 2020). The combination of 15-day temporal resolution and spline-based GAM smoothing allows quantifying both gradual greening trends and abrupt degradation events, such as those induced by dam construction or vegetation suppression, and is well-suited for time-series ecological monitoring in heterogeneous landscapes (Shumway & Stoffer, 2000). Moreover, to visualize refined positive and negative spatial-temporal patterns for each pixel of interest, a linear regression model was fitted yielding a pixel-wise decadal slope, representing the direction and magnitude of vegetation change. The resulting trend statistics were joined back to the full dataset and represented using faceted violin plots. These plots illustrate the distributional shift of EVI across decades, fine-tuning the analysis. Results The spatially explicit analysis of Enhanced Vegetation Index (EVI) trends (1984–2024) allowed for a comprehensive evaluation of vegetation dynamics at the pixel scale within the study area. This analysis leveraged long-term Landsat time-series data, integrating statistical trend analysis methods and spatial visualization through Geographic Information Systems (GIS). A pixel-wise linear regression model was employed to quantify long-term vegetation trends, utilizing EVI as the dependent variable and acquisition date as the predictor. This resulted in slope coefficients that represent the annual rate of vegetation change, subsequently classified into five discrete trend categories: Strong Positive, Mild Positive, Stable, Mild Negative, and Strong Negative. Classification thresholds were determined by applying natural breaks (Jenks method), enhancing the interpretability of observed ecological shifts and facilitating spatial pattern recognition (Fig. 6 , EVI Trends Map 1984–2024). The GIS-based map, using a combination of graduated colour and symbol sizing, supported the visualization of magnitude and direction of vegetation change. Blue shades symbolized positive vegetation dynamics, whereas red hues denoted negative trends. Pixel sizes directly reflected the slope magnitude, further highlighting areas experiencing pronounced ecological transitions. Spatial analysis revealed a clear patterning of trend categories, where pixels characterized by strong positive trends appear to be predominantly clustered within the central and southern watershed riparian corridor zones. The analysis of variability, or standard deviation of EVI, (Table 3 ) among pixels exhibiting positive trends (SD ranging from approximately 0.06 to 0.09 for pixels such as pixel_45, pixel_20, pixel_52) revealed moderate seasonal variability. This might be indicative of vegetation types that exhibit stable greenness punctuated by brief seasonal fluctuations possibly tied to climatic factors (White et al., 2009 ). Conversely, pixels indicating strong negative trends (pixel_118, pixel_137, and pixel_109) were spatially scattered and mostly peripheral. These as areas are experiencing persistent stress or degradation, with minimal vegetative recovery post-disturbance. Lower EVI values and comparatively lower standard deviation (SD of about 0.03 to 0.06) supported interpretations of continuous vegetation loss or predominant and stagnant condition of intense agricultural land use, deriving from direct anthropogenic impacts, or no adequate hydrological regimes. Table 3 The analysis of variability (standard deviation of EVI). Pixel ID N Observations EVI Trend (Slope) EVI SD Trend Category 1984–2024 pixel_45 425 + 1.241 × 10⁻⁵ 0.0818 Strong Positive pixel_15 429 + 1.095 × 10⁻⁵ 0.0860 Strong Positive pixel_44 425 + 1.082 × 10⁻⁵ 0.0666 Strong Positive pixel_52 426 + 1.028 × 10⁻⁵ 0.0725 Strong Positive pixel_20 428 + 1.001 × 10⁻⁵ 0.0864 Strong Positive pixel_118 425 −3.871 × 10⁻⁶ 0.0942 Strong Negative pixel_137 426 −3.559 × 10⁻⁶ 0.0542 Strong Negative pixel_56 426 −3.426 × 10⁻⁶ 0.0560 Strong Negative pixel_100 426 −3.287 × 10⁻⁶ 0.0674 Strong Negative pixel_109 425 −3.087 × 10⁻⁶ 0.0958 Strong Negative For targeted and detailed temporal analyses, a subset of pixels representative of the strongest negative and positive categories was selected. This strategic sampling facilitated focused semi-monthly aggregated EVI analysis, further clarifying phenological patterns, ecological stability, and potential stress signals within representative locations. The 15-days semi-monthly aggregated EVI curves smoothed using Generalized Additive Models (GAMs), balanced temporal resolution and signal clarity, enhancing the capacity to detect seasonal shifts, vegetation response to climate variations, and changes associated with hydrological disturbances (Fig. 7 ). The temporal graphs for random selected pixels (pixel 20, 45, 56, and 137) supports to visualize long-term vegetation trajectories. Pixel 45 and Pixel 20 exhibit generally increasing EVI trends, categorized as strong positive pixels. There is a substantial recovery and subsequent improvement of vegetation conditions following dam construction (2004–2007), as in fact the post-dam period shows an accelerated increase in vegetation vigour. Pixel 56 presents a relatively stable to slightly declining pattern, with intermittent fluctuations across decades. Despite fluctuations, the EVI values remain relatively consistent. Pixel 137 is distinctly characterized by a strong negative trend. The declining vegetation vigour throughout the observed period indicates ongoing stress. To refine the trend analysis, the violin plots decadal variability of EVI for selected pixels, were grouped by their long-term trend classification (strong positive vs. strong negative) (Fig. 8 ). Pixel 20 shows a progressive increase in EVI over the decades, particularly pronounced during the 1990s and 2000s. The standard deviation (SD) remains relatively consistent across decades (0.03–0.07), reflecting stable ecological improvement likely due to favourable environmental conditions. Pixel 45 experiences an ecological transition, shifting from negative trends in the 1980s, to significantly positive trends during the 2000s, comprising an ecological post-dam re-stabilization in the 2010s. Pixel 137 consistently shows negative trends from the 1980s through the 2010s, with a slight positive shift in the 2020s. Variability remains relatively constant (SD around 0.04–0.06), indicating prolonged ecological stress or ongoing degradation factors. Pixel 56 displays fluctuating patterns, initially stable or slightly positive in the 1980s and 1990s, shifting to strongly negative trends during the 2000s and 2010s, again reflecting ecological responses possibly exacerbated by hydrological changes from the dam construction period. However, a recent stabilization observed in the 2020s suggests potential adaptation. Overall, the negative pixels will have to be furthermore explored by planning specific field surveys in that designated area of interest. To document species composition and examine the ecological context for interpreting EVI-based positive trends, a field survey was conducted along the riparian corridor corresponding to the areal of Strong Positive trends comprising pixel_69, pixel_68, pixel_60, pixel_59, pixel_52, pixel_51, pixel_46, pixel_45, pixel_38. By aligning in situ observations with remotely sensed data, the survey provided insights into how plant taxa and species composition contribute to observed Positive Vegetation Trends in the Akaki River basin. As for that, ground-truthing surveys confirmed that these areas are dominated by non-native eucalyptus trees ( Eucalyptus spp.). These species typically thrive under stabilized hydrological regimes, resulting from altered water flow and sediment transport caused by dam construction (Fig. 9 ). This observation suggests a potential correlation between hydrological alterations and vegetation responses to water stress, aligning closely with Mediterranean ecological succession theory (Castellano et al., 2022 ). Such dynamics are also reflected in the EVI variability observed in pixel 45. While remotely sensed positive greening could suggest overall ecological health, as observed in the time-series graphs, field validation indicated that such expansion by eucalyptus may, in fact, mask underlying ecological degradation due to the competitive exclusion of native riparian species. This raises an urgent question of whether stable but reduced flows can preserve a naturally diverse riparian zone, or if additional environmental flow provisions are necessary to sustain native species. Understanding the implications of this eucalyptus-dominated re-greening is therefore essential for future water management decisions, as it highlights the complex interplay between dam operations, groundwater recharge, species adaptability, and overall ecosystem health in water-scarce Mediterranean settings. Conclusions The Klirou-Malounta-Akaki Dam and its downstream environment provide an ideal field setting to examine how water management impacts vegetation in a water-limited ecosystem. The situation here is representative of many small dams in Mediterranean climates where intermittent streams are impounded. Because the dam markedly reduces flood peaks and lengthens the interval between flow events, changes take place in the riparian zone. For instance, without annual floods to reset the system, vegetation can encroach further into the river channel, potentially leading to denser growth (a phenomenon observed in other regulated ephemeral rivers). Additionally, groundwater recharge from controlled releases could improve sub-surface moisture availability, aiding perennial plants during the dry season. In this case, ground-truth observations, however, suggest that eucalyptus stands contribute significantly to the EVI signal in certain segments of the corridor. This non-native species can exhibit year-round foliage and relatively consistent canopy vigour, even under moderate water stress. By virtue of their deep-rooted physiology, eucalyptus individuals may benefit from enhanced groundwater levels, whether from natural seepage or incidental dam releases, leading to sustained greenness detectable by satellite indices. In contrast, some native riparian species exhibit more pronounced leaf-shedding or dormancy when surface flows are scarce. Consequently, the constant to increasing greenness trends identified in the time-series may, in part, reflect the proliferation or improved survivorship of eucalyptus in downstream areas. This underscores a subtle ecological shift: rather than native riparian flora dominating a naturally flood-driven corridor, the post-dam environment may favour introduced or drought-resistant species capable of persisting in a stabilized hydrologic regime. Should the dam entirely curtail ordinary seasonal flows, the ecological composition could further skew toward drought-tolerant matrices, including eucalyptus stands, oleander thickets, and other hardy shrubs. From a management perspective, these findings highlight that even in the absence of major flood pulses, downstream ecosystems can undergo notable vegetative changes. A re-greening trend dominated by eucalyptus may indicate sufficient subsurface water availability, potentially from groundwater recharge facilitated by the dam, yet also point to reduced habitat heterogeneity if other native species are outcompeted. Monitoring EVI provides an integrative measure of vegetative health, yet species-level identification (e.g., via ground surveys or higher-resolution imagery) becomes essential to clarify whether the observed greening denotes resilient native recovery or the spread of introduced taxa. These considerations are important for deciding whether future environmental flow provisions are necessary to maintain ecological balance in riparian zones, particularly if dam operation inadvertently supports non-native species at the expense of local biodiversity. As water scarcity and climate variability continue to escalate in Cyprus and the broader Mediterranean, integrative approaches, encompassing vegetation indices, and species-level observations, will be increasingly necessary to guide sustainable water-resource management and preserve riparian ecosystem integrity. Future developments Future developments will focus on integrating additional data sources and advanced modelling techniques to enhance the precision and robustness of vegetation response analyses. Specifically, incorporating hydrological data from the dam reservoir and climatic variables such as precipitation and temperature will allow accurate estimation of potential evapotranspiration (PET). The use of harmonized datasets combining Sentinel-2 and Landsat imagery will provide improved spatial resolution and temporal consistency, enabling more detailed recalculations of EVI trends and robust ecological interpretations. Additionally, conducting targeted ground-truthing surveys for plots exhibiting negative trends will help identify potential vegetation stress factors and inform appropriate management interventions. Declarations Author Contribution The author, M.G., wrote the work. Funding Declaration This work was not supported by any grant or fund. References Aybar, C., Wu, Q., Bautista, L., Yali, R., & Barja, A. (2020). rgee: An R package for interacting with Google Earth Engine . Journal of Open Source Software, 5 (51), 2272. https://doi.org/10.21105/joss.02272 Berner, L. T., Assmann, J. J., Normand, S., & Goetz, S. J. (2023). LandsatTS: An R package to facilitate retrieval, cleaning, cross‐calibration, and phenological modeling of Landsat time series data. Ecography, 2023 (9), e06768. Bunn, S. E., & Arthington, A. H. (2002). Basic principles and ecological consequences of altered flow regimes for aquatic biodiversity. Environmental Management, 30 (4), 492–507. Castellano, C., Bruno, D., Comin, F. A., Rey Benayas, J. M., Masip, A., & Jimenez, J. J. (2022). Environmental drivers for riparian restoration success and ecosystem services supply in Mediterranean agricultural landscapes. Coelho, C. D., da Silva, D. D., Amorim, R. S. S., Vasconcelos, B. N. F., Possato, E. L., Filho, E. I. F., ... & Silva, L. V. (2024). Development and Application of an Environmental Vulnerability Index (EVI) for Identifying Priority Restoration Areas in the São Francisco River Basin, Brazil. Land, 13(9), 1475. Cyprus Water Development Department. (2007). Technical report on the Klirou–Malounta Dam construction and operation . Nicosia, Cyprus: Ministry of Agriculture, Rural Development and Environment. Cyprus Water Development Department. (2019). Annual report 2018–2019: Hydrological data and reservoir monitoring . Nicosia: Government of Cyprus. https://www.moa.gov.cy/wdd de Jong, R., Verbesselt, J., Zeileis, A., & Schaepman, M. E. (2011). Shifts in global vegetation activity trends. Remote Sensing, 3 (3), 595–614. https://doi.org/10.3390/rs5031117 DeVries, B., Huang, C., Armston, J., Huang, W., Jones, J. W., & Lang, M. W. (2020). Rapid and robust monitoring of flood events using Sentinel-1 and Landsat data on the Google Earth Engine. Remote sensing of Environment, 240, 111664. Forkel, M., Carvalhais, N., Rödenbeck, C., Keeling, R., Heimann, M., Thonicke, K., ... & Reichstein, M. (2015). Enhanced seasonal CO₂ exchange caused by amplified plant productivity in northern ecosystems. Science, 351 (6274), 696–699. https://doi.org/10.1126/science.aac4971 Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., & Moore, R. (2017). Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sensing of Environment, 202 , 18–27. https://doi.org/10.1016/j.rse.2017.06.031 Gurung, R. B., Breidt, F. J., Dutin, A., & Ogle, S. M. (2009). Predicting Enhanced Vegetation Index (EVI) curves for ecosystem modeling applications. Remote Sensing of Environment, 113(10), 2186-2193. Hadjinicolaou, P., Giannakopoulos, C., Zerefos, C., Lange, M. A., Pashiardis, S., & Lelieveld, J. (2011). Mid-21st century climate and weather extremes in Cyprus as projected by six regional climate models. Regional Environmental Change, 11 , 441–457. Larned, S. T., Datry, T., Arscott, D. B., & Tockner, K. (2010). Emerging concepts in temporary-river ecology. Freshwater Biology, 55 (4), 717–738. https://doi.org/10.1111/j.1365-2427.2009.02322.x Masek, J. G., Vermote, E. F., Saleous, N. E., Wolfe, R., Hall, F. G., Huemmrich, K. F., ... & Gao, F. (2006). A Landsat surface reflectance dataset for North America, 1990–2000. IEEE Geoscience and Remote Sensing Letters, 3 (1), 68–72. https://doi.org/10.1109/LGRS.2005.857030 Nagler, P. L., Scott, R. L., Westenburg, C., Cleverly, J. R., Glenn, E. P., & Huete, A. R. (2005). Evapotranspiration on western US rivers estimated using the Enhanced Vegetation Index from MODIS and data from eddy covariance and Bowen ratio flux towers. Remote sensing of environment, 97(3), 337-351. Nilsson, C., Reidy, C. A., Dynesius, M., & Revenga, C. (2005). Fragmentation and flow regulation of the world’s large river systems. Science, 308 (5720), 405–408. https://doi.org/10.1126/science.1107887 Pekel, J.-F., Cottam, A., Gorelick, N., & Belward, A. S. (2016). High-resolution mapping of global surface water and its long-term changes. Nature, 540 (7633), 418–422. https://doi.org/10.1038/nature20584 Petts, G. E., & Gurnell, A. M. (2005). Dams and geomorphology: Research progress and future directions. Geomorphology, 71 (1–2), 27–47. https://doi.org/10.1016/j.geomorph.2004.02.015 Poff, N. L., Olden, J. D., Merritt, D. M., & Pepin, D. M. (2007). Homogenization of regional river dynamics by dams and global biodiversity implications. Proceedings of the National Academy of Sciences, 104 (14), 5732–5737. https://doi.org/10.1073/pnas.0609812104 Seddon, A. W. R., Macias-Fauria, M., Long, P. R., Benz, D., & Willis, K. J. (2016). Sensitivity of global terrestrial ecosystems to climate variability. Nature, 531 (7593), 229–232. https://doi.org/10.1038/nature16986 Shi, H., Li, L., Eamus, D., Huete, A., Cleverly, J., Tian, X., ... & Carrara, A. (2017). Assessing the ability of MODIS EVI to estimate terrestrial ecosystem gross primary production of multiple land cover types. Ecological Indicators, 72, 153-164. Shumway, R. H., Stoffer, D. S., & Stoffer, D. S. (2000). Time series analysis and its applications (Vol. 3, p. 4). New York: springer. Tockner, K., & Stanford, J. A. (2002). Riverine flood plains: Present state and future trends. Environmental Conservation, 29 (3), 308–330. https://doi.org/10.1017/S037689290200022X U.S. Geological Survey (USGS). (2020). Landsat Collection 2 Level-2 Science Products . https://www.usgs.gov/core-science-systems/nli/landsat/landsat-collection-2 Vermote, E., Justice, C., Claverie, M., & Franch, B. (2016). Preliminary analysis of the performance of the Landsat 8/OLI land surface reflectance product. Remote Sensing of Environment, 185 , 46–56. https://doi.org/10.1016/j.rse.2016.04.008 White, M. A., de Beurs, K. M., Didan, K., Inouye, D. W., Richardson, A. D., Jensen, O. P., ... & Lauenroth, W. K. (2009). Intercomparison, interpretation, and assessment of spring phenology in North America estimated from remote sensing for 1982–2006. Global Change Biology, 15(10), 2335-2359. Zhao, B., Yan, Y., Guo, H., He, M., Gu, Y., & Li, B. (2009). Monitoring rapid vegetation succession in estuarine wetland using time series MODIS-based indicators: an application in the Yangtze River Delta area. Ecological indicators, 9(2), 346-356. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted 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-6761511","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":470498570,"identity":"0710a343-3462-4d01-a10c-1c05113c0000","order_by":0,"name":"Marzia Gabriele","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABB0lEQVRIiWNgGAWjYLCCBBAhwcBwgKGAQY6BgbnhAIMBAw+RWgwYjBkYGCFa8OuBamEAqkxsAGoBC+DSott+9vGHBzUMdv2zex8e+GBgk76dvbHx0I0CBhl7HFrMzqQbGCQcY0iecee4wcEZBmm5O3sONhzOweMwswNpDAkJbAzJDDfSGA7zGBzO3XAjkYCW888YDiT8Y0iWB2n5Y/A/3YCglhtpjA2JbQx2BiAtDAYHEojQ8oyZIbFPIsHwzjGGgz0GyYZQv0jw8BzA5bA05o8/vtnYy91uY/7wo8JO3py9+fDnnD829uwNOKyBAIlEuLwBVASvehBARIIBQbWjYBSMglEw0gAAbtBeXhXqGYQAAAAASUVORK5CYII=","orcid":"","institution":"Built environment and Construction engineering (ABC), Politecnico di Milano","correspondingAuthor":true,"prefix":"","firstName":"Marzia","middleName":"","lastName":"Gabriele","suffix":""}],"badges":[],"createdAt":"2025-05-27 16:53:21","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6761511/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6761511/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":84779145,"identity":"ae10fa84-55a2-419b-bf2a-1afad01499dc","added_by":"auto","created_at":"2025-06-17 09:20:35","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":137061,"visible":true,"origin":"","legend":"\u003cp\u003eTechnical overview of the Akaki–Malounta Dam (also known as Klirou–Malounta–Akaki), located on the Akaki (Serrachis) River in Nicosia District, Cyprus. Source: Cyprus Water Development Department.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6761511/v1/74df3260dec9d7534dade2d0.png"},{"id":84780849,"identity":"6e95c972-18f3-4fa0-b2da-f1bc06339d26","added_by":"auto","created_at":"2025-06-17 09:28:36","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":999544,"visible":true,"origin":"","legend":"\u003cp\u003eLeft, wide view of the Klirou-Malounta dam reservoir. Right, close-up of the dam spillway designed for controlled water release and manage flow regulation. Source: Cyprus Water Development Department.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6761511/v1/a4850155fca5935a27870e7d.png"},{"id":84779153,"identity":"c0d230bb-b111-42e9-958f-e396a9e54516","added_by":"auto","created_at":"2025-06-17 09:20:36","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":832183,"visible":true,"origin":"","legend":"\u003cp\u003eLeft, downstream perspective showing the areas along riverbanks. Right, wetland area downstream of the dam. Source: Cyprus Water Development Department.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6761511/v1/c1becf23fbf9a80ba31a746a.png"},{"id":84779149,"identity":"558ad4c5-7dfd-4cbf-bbc5-43ae2ec94516","added_by":"auto","created_at":"2025-06-17 09:20:36","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":832811,"visible":true,"origin":"","legend":"\u003cp\u003eStudy area overview of the Klirou-Malounta-Akaki Dam site in Cyprus. Administrative boundaries of Cyprus (background imagery: Google Satellite); Land Productivity Dynamics (LPD) of Lefkosia-Cyprus for the year 2024, highlighting areas of declining, stable, and increasing productivity; LPD specifically within the Serrahis River watershed upstream of Klirou-Malounta-Akaki Dam; the sampled area with the georeferenced pixels used for vegetation monitoring and ground-truthing within the dam watershed.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-6761511/v1/2d5cae2d6e2fadda5d6c8c93.png"},{"id":84779147,"identity":"36416fe8-8ee2-4f1e-8994-294254809e91","added_by":"auto","created_at":"2025-06-17 09:20:36","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":3827268,"visible":true,"origin":"","legend":"\u003cp\u003eThe Klirou-Malounta dam exemplifies the complexities associated with human-altered landscapes, serving as a critical reference for studying how dams affect hydrological processes and vegetation dynamics. [Drone pictures of the study area acquired from the author M.G., 27/09/2023].\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-6761511/v1/cb0aaad2f19ab1d5163c9f03.png"},{"id":84779154,"identity":"7daf00b7-9415-4a53-be2e-ea9d368b55b0","added_by":"auto","created_at":"2025-06-17 09:20:36","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":1955097,"visible":true,"origin":"","legend":"\u003cp\u003eSpatial distribution of long-term EVI trends (1984–2024) in the area upstream of the Klirou–Malounta–Akaki Dam, located within the Serrahis River watershed. Each 30 × 30 m pixel represents a classified trend category derived from Landsat-based EVI time-series analysis.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-6761511/v1/241e7ca05bb20fb4fe07f02d.png"},{"id":84780851,"identity":"c214233b-11d9-45a1-b805-751282ab3e9d","added_by":"auto","created_at":"2025-06-17 09:28:36","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":789239,"visible":true,"origin":"","legend":"\u003cp\u003eTime-series graphs of the Enhanced Vegetation Index (EVI) trends over 15-day intervals from 1984 to 2024 for specific pixels (a) pixel 45, (b) 20, (c) 56, (d) 137. illustrating changes in vegetation over decades and highlight the period of dam construction (2004–2007).\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-6761511/v1/c4363d632fc2035efdbef441.png"},{"id":84779158,"identity":"cc5501c5-f4e8-43cc-ba0d-0bc24dada90c","added_by":"auto","created_at":"2025-06-17 09:20:36","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":546183,"visible":true,"origin":"","legend":"\u003cp\u003eDecadal distribution of EVI values for selected pixels categorized as (a) strong positive and (b) strong negative trend classes. Each violin plot represents the spread and central tendency of EVI values within a decade, calculated from 15-day composite time series.\u003c/p\u003e","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-6761511/v1/12b26fe158dac56f0176ddf3.png"},{"id":84779165,"identity":"70c5fdd7-b7e2-49c5-baa2-55751a65aa5b","added_by":"auto","created_at":"2025-06-17 09:20:36","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":1063779,"visible":true,"origin":"","legend":"\u003cp\u003eEcological states of the riparian corridor. Mature eucalyptus stands dominate, highlighting the encroachment of non-native species into these zones. The lack of native vegetation is indicative of reduced water availability and ecological stress. Opportunistic shrubs and ground vegetation dominate this section (pixel_69, pixel_68, pixel_60, pixel_59, pixel_52, pixel_51, pixel_46, pixel_45, pixel_38), emphasizing soil erosion and reduced hydrological input.\u003c/p\u003e","description":"","filename":"floatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-6761511/v1/818a6ddd3cad7552560bac31.png"},{"id":99789839,"identity":"1110315a-e1bd-44fc-a5b2-cf3bc1d0737c","added_by":"auto","created_at":"2026-01-08 12:50:45","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":11569583,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6761511/v1/29727054-f94d-4979-b034-d86e20f5afd2.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eAssessing Vegetation Responses to Dam-Induced Hydrological Change: A 40-Year Landsat Time-Series Analysis in the Downstream Serrahis River Watershed, Cyprus\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe Klirou–Malounta–Akaki Dam is in central Cyprus (Nicosia District) on the Akaki River, a tributary in the Serrachis River basin. Geographically, it sits about 25–26 km southwest of Nicosia, in the northern foothills of the Troodos Mountains​ (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The dam lies between the villages of Klirou and Malounta (which are approximately 2 km apart), at an elevation of roughly 300–350 m above sea level. The surrounding area is a transitional zone between the Troodos uplands and the Mesaoria plain. Upstream (southward) from the dam, the terrain rises toward the Troodos range (including the Papoutsa and Farmakas peaks), which provides the catchment for the Akaki River. Downstream (northward), the river flows through a narrow valley that opens into the broad Nicosia/Mesaoria plain on its way to Morphou Bay on the northwestern coast. In its middle reaches near Klirou and Malounta, the Akaki River has incised a small gorge into the alluvial plain, around which fertile agricultural lands and village settlements are established \u003cb\u003e(\u003c/b\u003eWDD, 2019).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe region’s climate is Mediterranean semi-arid. Mean annual rainfall in the Klirou/Malounta area is on the order of 400–500 mm, most of it occurring in winter. Summers are typically dry with zero flow in the river. The Akaki is an ephemeral river – it experiences seasonal flow during winter rains and remains dry for much of the rest of the year. Prior to dam construction, the Akaki River’s flow regime was characterized by flashy winter floods (in wet years the river could carry substantial runoff from the 84 km² upstream catchment) followed by rapid recession. Now, the Klirou Dam intercepts and regulates these flows. The dam’s reservoir (often referred to as Klirou Reservoir or Kalo Chorio Reservoir) has a full storage volume of ~ 2.0\u0026nbsp;million m³ and a surface area of ~ 18 ha when full (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). For context, the mean annual flow of the Akaki River at the dam site is estimated at 12\u0026nbsp;million m³, which indicates that in an average year not all runoffs can be stored and some will spill. The dam is an earthfill structure with a central clay core, approximately 38 m high and 265 m in crest length. It became operational in 2007 (WDD, 2007)​. By design, this is a multi-purpose small dam serving mainly irrigation supply and groundwater recharge. There are no hydroelectric facilities, as Cyprus has no hydroelectric dams due to its limited hydrology. Instead, water stored in Klirou Reservoir is intended for controlled release to irrigation networks and to replenish the downstream aquifer in the Akaki valley. The Water Development Department (WDD) manages the dam’s operation and monitors groundwater levels to gauge recharge benefits. Importantly, the dam was not built with a dedicated spillway for continuous ecological flow releases – like most Cypriot dams of its era, it can release water through valves manually, but it does not simulate natural flow patterns on its own (WDD, 2007). This means the downstream river stretch often remains dry until either deliberate releases are made for irrigation/recharge, or the reservoir overtops in a heavy rain event. Several overflow events have been recorded in exceptionally wet winters (e.g. the dam spilled in winter 2018–2019 after heavy rains​), briefly restoring flow to the downstream channel.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eStudy Area\u003c/h3\u003e\n\u003cp\u003eThe land use in the downstream area is a mixture of agricultural and semi-natural riparian landscapes. The villages of Klirou, Malounta, and Akaki (further downstream ~ 10 km north) are traditional farming communities. Surrounding these settlements are cultivated fields, including olive groves, citrus orchards, grain fields, and some vegetable cultivation​. Agriculture in this region relies partly on groundwater and partly on surface water where available. The presence of the river (even if intermittent) has historically supported agriculture by providing alluvial soils and shallow groundwater; farmers have dug wells in the plains where the water table is accessible​. Livestock farming (e.g. sheep, goats) is also common in the area, taking advantage of the grazing in valley bottoms. The riparian corridor of the Akaki River area of interest of this study, (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e), though seasonally dry, hosts distinctive Mediterranean streamside vegetation. This includes shrubs and trees adapted to infrequent water pulses, such as tamarisk (Tamarix spp.), oleander (Nerium oleander), reeds (Phragmites spp.). Scattered riverbank stands of plane trees (Platanus orientalis) and eucalyptus (Eucalyptus spp.) occur in moister stream segments, with eucalyptus being an introduced, fast-growing taxon that tolerates periods of low surface flow by tapping deeper soil moisture. Because of its robust root systems and moderate drought tolerance, eucalyptus has been widely used across the Mediterranean to stabilize riparian embankments, reduce erosion, and provide timber or windbreaks. Much of the year, the riverbed may appear as a dry channel with patches of green where deep-rooted vegetation taps residual moisture. After winter rains, ephemeral herbs and grasses briefly flourish. Upstream of the dam, the landscape is more rugged and natural: the catchment area extends into forested slopes and shrubland of the Troodos foothills, which are part of a protected forest zone. If the dam completely cuts off ordinary flows in average years, water stress on downstream flora could increase, causing drought-tolerant species to dominate and any water-dependent species to decline.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eMoreover, Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e exemplifies panoramically the morphology of the Area of Interest (AOI) near the E929 road crossing. The drone perspective highlights seasonally dry channels, alluvial soils, and riparian vegetation in the corridor downstream of the dam. Much of the year, the riverbed remains barren, with deep-rooted vegetation persisting on residual moisture and ephemeral herbs flourishing briefly after winter rains. Analysing vegetation indices (which correlates with plant biomass and vigour) over time, can capture these effects quantitatively. Any observable trend downstream of the dam – relative to upstream areas – might indicate the positive influence of sustained moisture. Conversely, a decline in greenness could signal deteriorating riparian health due to reduced surface flows. Understanding these outcomes is crucial for water management: it will inform whether environmental flow provisions might be necessary at small dams and how current dam operations could be adjusted to balance human use with ecosystem needs, which after dam construction might have further skew toward drought-tolerant ecological matrixes. As for that, the study area’s mix of Mediterranean climate, ephemeral river hydrology, dam intervention, and dependent agriculture offers a microcosm to study vegetation–hydrology interactions under water stress conditions. Insights gained here will contribute to the broader knowledge of how dams in semi-arid regions affect downstream ecology, which is increasingly important as water resources scarcity is driving more infrastructure projects in similar environments (Hadjinicolaou et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2011\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eProblem Statement\u003c/p\u003e \u003cp\u003eDams worldwide significantly alter natural flow regimes and sediment transport, triggering downstream ecological changes such as wetland formation or shifts in riparian plant communities (Nilsson et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Bunn \u0026amp; Arthington, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2002\u003c/span\u003e). Programmed flood releases are often employed to mitigate such effects by mimicking natural disturbance regimes (Poff et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). However, in the case of the Klirou–Malounta–Akaki Dam, no such flood pulses have been implemented. Meanwhile, precipitation trends have remained relatively stable (Hadjinicolaou et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2011\u003c/span\u003e), suggesting that observed vegetation changes may stem more from local geomorphological or microclimatic drivers than from major hydrological reconfiguration (Tockner \u0026amp; Stanford, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2002\u003c/span\u003e; Petts \u0026amp; Gurnell, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). This aligns with emerging perspectives on intermittent river systems where ecological dynamics are often shaped by complex interactions beyond flow alone (Larned et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2010\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eObjective:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eInvestigate vegetation changes (via EVI time-series) pre- and post-dam construction.\u003c/li\u003e\n \u003cli\u003eIdentify pixels with positive and negative trends, to assess whether these trends might correlate with the presence and operation of the dam.\u003c/li\u003e\n \u003cli\u003eDetermine the most positive trends to point to identify specific areas where to plan on-field survey and investigate the in-place ecological processes and species composition.\u003c/li\u003e\n\u003c/ul\u003e\n"},{"header":"Data and Materials","content":"\u003cp\u003eSatellite Data (Landsat)\u003c/p\u003e\n\u003cp\u003eThe lsat_export_ts() function utilizes the Landsat Collection 2, Level-2 Surface Reflectance (SR) datasets to extract long-term time series of surface reflectance from multiple Landsat missions. Specifically, it accesses six Earth Engine ImageCollections corresponding to Tier 1 and Tier 2 datasets from Landsat 5 (Thematic Mapper, TM), Landsat 7 (Enhanced Thematic Mapper Plus, ETM+), and Landsat 8 (Operational Land Imager, OLI). These are identified as:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eLANDSAT/LT05/C02/T1_L2, LANDSAT/LT05/C02/T2_L2\u003c/li\u003e\n\u003c/ul\u003e\n\u003cdiv id=\"Sec5\"\u003e\n \u003cul\u003e\n \u003cli\u003eLANDSAT/LE07/C02/T1_L2, LANDSAT/LE07/C02/T2_L2\u003c/li\u003e\n \u003c/ul\u003e\n \u003cdiv id=\"Sec6\"\u003e\n \u003cul\u003e\n \u003cli\u003eLANDSAT/LC08/C02/T1_L2, LANDSAT/LC08/C02/T2_L2\u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003eThese datasets provide atmospherically corrected surface reflectance values at 30-meter spatial resolution. For Landsat 8, the Land Surface Reflectance Code (LaSRC) is applied (Vermote et al., 2016), while Landsat 5 and 7 employ the LEDAPS (Landsat Ecosystem Disturbance Adaptive Processing System) algorithm (Masek et al., 2006). Each image contains calibrated spectral reflectance for bands 1 through 7 (visible to shortwave infrared), alongside quality assurance bands such as QA_PIXEL and QA_RADSAT, which enable cloud, shadow, and saturation filtering. Metadata attributes including acquisition date, solar elevation, spacecraft ID, and processing level are also retained during export to support contextual and temporal analysis. The function overlays the max_extent band from the Global Surface Water (GSW) dataset developed by the Joint Research Centre, accessible via Earth Engine as JRC/GSW1_4/GlobalSurfaceWater. This dataset offers high-resolution information on the historical maximum extent of surface water derived from over 3\u0026nbsp;million Landsat scenes spanning the period 1984 to the present (Pekel et al., 2016). Its integration enables enhanced interpretation of water-related land cover dynamics in conjunction with vegetation indices. To ensure consistent data structure across the time series, the function pads missing bands in older images with a placeholder band (e.g., null-valued SR_B6) and coerces all outputs to floating-point format. This guarantees compatibility during pixel-wise statistical analysis or trend detection workflows. To evaluate data availability across sites and years, a summary table is generated, and a vegetation index (EVI) is computed to characterize vegetation status over time (Gurung et al., 2009; Shi et al, 2017; Zhao et al., 2009; Nagler et al., 2005; Coelho et al., 2024).\u003c/p\u003e\n \u003cp\u003eVegetation Index: Enhanced Vegetation Index (EVI) computed as:\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003ch3\u003e𝐸𝑉𝐼=2.5×𝜌𝑛𝑖𝑟−𝜌𝑟𝜌𝑛𝑖𝑟+6.0𝜌𝑟−7.5𝜌𝑏+1 (1)\u003c/h3\u003e\n\u003cp\u003eWhere 𝜌𝑛𝑖𝑟, 𝜌𝑟, and 𝜌𝑏 stand for the atmospherically corrected surface reflectance of the near-infrared, red (visible) and blue (visible) bands of Landsat data. Cross-sensor reflectance calibration is conducted to harmonize data from Landsat 5, 7, and 8 using a Random Forest modelling approach, focusing on observations collected between days of the year 151 and 242. This calibration, anchored by Landsat 7 as the temporal reference, ensures comparability among multiple Landsat missions. For the subsequent phenological and trend analyses, only cloud-free observations (below a defined 5% threshold) from clear-sky conditions are retained.\u003c/p\u003e\n\u003cp\u003eMethodology\u003c/p\u003e\n\u003cp\u003eThe methodological workflow for Landsat Time-Series Extraction and Phenological Trend Analysis is developed by loading a suite of R packages, namely sf, dplyr, purrr, data.table, stringr, rgee (Aybar et al., 2020), and LandsatTS (Berner et al., 2023), o enable spatial data handling, data manipulation, and interaction with Google Earth Engine (GEE). After initializing the GEE (Gorelick et al., 2017) environment, Landsat 8 grid pixel canters are identified within the target polygons by employing a buffer-based function designed to avoid excessively large polygons, which can lead to computational inefficiencies. Pixel centers are assigned unique identifiers in a column such as “site_id” or “sample.id,” ensuring that subsequent functions recognize and distinguish among different locations. Following determination of point coordinates, the lsat_export_ts() function is used to export the corresponding Landsat surface reflectance data from GEE. This function packages the coordinates into sf point features, submits export tasks, and saves CSV files in a designated Google Drive folder. Once these tasks complete, the resulting CSV files are downloaded locally and imported into R for further processing, using data.table::fread() to handle large files efficiently. After loading the data, lsat_format_data() or equivalent steps are used to standardize column names, confirm the presence of a unique “sample.id,” and apply any necessary scaling to reflectance bands. The dataset “lsat.dt” was cleaned by the “lsat_clean_data” function. In this step, a maximum geometric registration error of 15 meters, a maximum allowable cloud cover of 10%, and a maximum solar zenith angle of 60° were specified. Additionally, filtering was enabled to exclude CFMask-snow, CFMask-water, and water pixels identified by the JRC Global Surface Water dataset, ensuring that only high-quality surface reflectance observations remained in the updated “lsat.dt” dataset. Data filtering removes suboptimal observations containing excessive cloud cover, snowfall, or water contamination. This produces a qualitative dataset of surface reflectance measurement, suitable for riverine vegetation monitoring. The lsat_summarize_data() function provides an overview of data availability, including temporal coverage, annual measurement frequencies, and plots of observation distributions, enabling to verify the completeness and reliability of the dataset.\u003c/p\u003e\n\u003cp\u003eAfter extraneous observations removed, the vegetation index Enhanced Vegetation Index (EVI) is computed through lsat_calc_spectral_index(). This step parses the required reflectance bands, applies sensor-specific scaling factors, and outputs consistently named columns for further analysis. Data are acquired from multiple Landsat sensors (Landsat 5, 7, and 8), whereas the cross-calibration becomes a crucial methodological step. Cross-calibration of spectral reflectance or indices is performed using the “lsat_calibrate_rf” function, and by specifing a day-of-year range that targets the mid-growing season (in this case: days 151 to 242, corresponding to late May through late August, depending on leap years). This range captures a period of relatively stable phenological conditions and is used to determine typical reflectance values. Landsat 7 serves as a temporal benchmark given its overlap with Landsat 5 and 8, and cross-calibration is conducted on a single band or spectral index at a time using data from thousands sample sites. A Random Forest model is then trained to predict Landsat 7 reflectance from Landsat 5 or 8 reflectance recorded in the same years. Parameters such as the minimum number of observations required per site (in this case: 5), whether to include high-latitude data, the fraction of data (in this case: 0.75) to use for model training, and whether to overwrite existing columns can be tuned based on the resulting root mean square error (RMSE) to optimize calibration accuracy (in this case: TRUE).\u003c/p\u003e \u0026nbsp;\u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 1\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eCalibration Coefficients\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSatellite\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eBand/SI\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eB0\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eB1\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eB2\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eB3\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eB0.se\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eB1.se\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eB2.se\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eB3.se\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLANDSAT_5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eevi\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0,0088\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1,0495\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e-0.086\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,0016\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,0102\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,0152\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNA\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLANDSAT_8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eevi\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-0,0093\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,9109\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.322\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-0,319\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,003\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,0276\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,0784\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0,0681\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cdiv\u003e\n \u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 2\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eModel Evaluation\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"11\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSatellite\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eBand/SI\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTrain Sites\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eEval Sites\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eR²\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eUncal. RMSE\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eUncal. Bias\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eUncal. Bias (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eXcal. RMSE\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eXcal. Bias\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eXcal. Bias (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLANDSAT_5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eevi\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5291\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1764\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,958\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,030\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-0,014\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-4,5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,026\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLANDSAT_8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eevi\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5964\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1988\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,954\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,032\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0,016\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5,1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,028\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0,4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eThe two tables (Table\u0026nbsp;1, Table\u0026nbsp;2) present both the coefficients of the cross-calibration model for Enhanced Vegetation Index (EVI) and the corresponding validation statistics after harmonizing Landsat 5 and Landsat 8 to a Landsat 7 reference baseline. In the upper table, columns labeled “B0,” “B1,” “B2,” and “B3” represent the fitted polynomial coefficients that map EVI values from Landsat 5 or Landsat 8 to those of Landsat 7. These coefficients and their standard errors (“B0.se,” “B1.se,” “B2.se,” “B3.se”) indicate the magnitude and uncertainty of each term in the calibration function. For instance, Landsat 5 uses B0 = 0.0088, B1 = 1.0495, and B2 = − 0.086 for its calibration polynomial. Meanwhile, Landsat 8 employs a fourth-degree polynomial, evidenced by the presence of B3 = − 0.319 (with the other fitted terms given by B0, B1, and B2). The difference in polynomial order arises because the Random Forest–based model selected these coefficients to minimize calibration error, indicating complex inter-sensor relationships for Landsat 8. The lower table summarizes how well each calibration performs. “train.n.sites” and “eval.n.sites” list the number of point locations used, respectively, for training and for out-of-sample evaluation. The “r2” column shows the coefficient of determination, reflecting the proportion of variance explained by the model. The “uncal.rmse” and “uncal.bias” columns provide baseline error metrics for raw (uncalibrated) EVI, while “xcal.rmse” and “xcal.bias” report those metrics after calibration. Negative uncalibrated bias (as for Landsat 5, − 0.014) signifies that EVI estimates tended to be lower than Landsat 7, whereas a positive bias (for Landsat 8, 0.016) indicates an overestimation. The percentage columns (e.g., “uncal.bias.pcnt” and “xcal.bias.pcnt”) present those biases relative to Landsat 7. For Landsat 5, cross-calibration reduced the RMSE from 0.030 to 0.026 and bias from − 0.014 to 0.001. Landsat 8 experienced a similar improvement, with RMSE dropping from 0.032 to 0.028 and bias shrinking from 0.016 to 0.001. The final bias percentages for both sensors approximate 0.4%, indicating that the Random Forest–based approach successfully aligned EVI values with the Landsat 7 reference to within minimal error margins.\u003c/p\u003e\n\u003cp\u003eFollowing cross-calibration of the Enhanced Vegetation Index (EVI) using lsat_calibrate_rf(), a CSV file was generated containing time series data from the 36 distinct pixels, each corresponding to a specific geographic location. The dataset incorporated crucial variables, including a unique sample ID (“sample.id”), the calendar year (“year”), the day of the year (“doy”), the EVI value (“evi”), the originating satellite (“satellite”), the percentage of cloud cover (“cloud.cover”), and a binary indicator of whether the observation was cloud-free (“clear”). To ensure data integrity in the time series construction, only records meeting strict quality thresholds were retained (specifically, cloud.cover \u0026lt; 5 and clear = = 1), and any observations with missing or invalid fields were removed prior to further analysis.\u003c/p\u003e\n\u003cp\u003eVegetation Time-series Analysis\u003c/p\u003e\n\u003cp\u003eTo identify representative vegetation dynamics across the study area, a custom procedure was implemented to select key pixels based on long-term EVI trends and temporal variability. For each pixel, a linear regression model was fitted to the EVI time series using acquisition date as the predictor variable. The slope of this model was extracted to quantify the direction and intensity of a long-term vegetation change (1984–2024): positive slopes indicated greening trends, while negative slopes pointed to degradation (Forkel et al., 2015; DeVries et al., 2020). The temporal variability of each pixel was assessed by calculating the standard deviation of EVI values, serving as an indicator of phenological fluctuation (Seddon et al., 2016). To visualize the slope value trend direction and intensity, a spatial representation of pixel-wise vegetation trends was developed using a GIS environment (QGIS 3.28). A graduated symbology was applied to the slope values extracted from the EVI time series, using a diverging colour ramp to represent trends from strong negative, to strong positive, and the size to determine the magnitude of the trend. Five trend classes were defined and visualized: Strong Negative, Mild Negative, Stable, Mild Positive, and Strong Positive. The resulting map provides a clear overview of spatial variability in long-term vegetation dynamics across the study area.\u003c/p\u003e\n\u003cp\u003eTo enable targeted time-series analysis, the top five pixels exhibiting strong negative and strong positive vegetation trends were extracted based on slope values. From each group, two representative pixels were selected, and their Enhanced Vegetation Index (EVI) time series was aggregated into 15-day (semi-monthly) intervals to capture seasonal dynamics and long-term patterns.\u003c/p\u003e\n\u003cp\u003eSpecifically, all valid EVI observations were grouped by pixel and year into 24 fixed bins corresponding to 15-day periods (DOY 15 to 360) and averaged within each bin to reduce high-frequency noise caused by cloud interference, sensor variation, or rapid short-term disturbances. This method preserves the seasonal signature while enabling robust interannual comparisons across decades (Seddon et al., 2016). The aggregated EVI curves were then smoothed using Generalized Additive Models (GAMs) with cubic regression splines (s(x, bs = \"cs\")), which provide flexible, data-adaptive fits while avoiding overfitting. This modelling strategy is widely adopted for land surface phenology analysis using moderate-resolution optical imagery (Forkel et al., 2015), especially in cases of non-linear vegetation response or in dryland ecosystems where phenological patterns are fragmented and noisy (DeVries et al., 2020). The combination of 15-day temporal resolution and spline-based GAM smoothing allows quantifying both gradual greening trends and abrupt degradation events, such as those induced by dam construction or vegetation suppression, and is well-suited for time-series ecological monitoring in heterogeneous landscapes (Shumway \u0026amp; Stoffer, 2000). Moreover, to visualize refined positive and negative spatial-temporal patterns for each pixel of interest, a linear regression model was fitted yielding a pixel-wise decadal slope, representing the direction and magnitude of vegetation change. The resulting trend statistics were joined back to the full dataset and represented using faceted violin plots. These plots illustrate the distributional shift of EVI across decades, fine-tuning the analysis.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eThe spatially explicit analysis of Enhanced Vegetation Index (EVI) trends (1984\u0026ndash;2024) allowed for a comprehensive evaluation of vegetation dynamics at the pixel scale within the study area. This analysis leveraged long-term Landsat time-series data, integrating statistical trend analysis methods and spatial visualization through Geographic Information Systems (GIS).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eA pixel-wise linear regression model was employed to quantify long-term vegetation trends, utilizing EVI as the dependent variable and acquisition date as the predictor. This resulted in slope coefficients that represent the annual rate of vegetation change, subsequently classified into five discrete trend categories: Strong Positive, Mild Positive, Stable, Mild Negative, and Strong Negative. Classification thresholds were determined by applying natural breaks (Jenks method), enhancing the interpretability of observed ecological shifts and facilitating spatial pattern recognition (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e, EVI Trends Map 1984\u0026ndash;2024). The GIS-based map, using a combination of graduated colour and symbol sizing, supported the visualization of magnitude and direction of vegetation change. Blue shades symbolized positive vegetation dynamics, whereas red hues denoted negative trends. Pixel sizes directly reflected the slope magnitude, further highlighting areas experiencing pronounced ecological transitions. Spatial analysis revealed a clear patterning of trend categories, where pixels characterized by strong positive trends appear to be predominantly clustered within the central and southern watershed riparian corridor zones.\u003c/p\u003e \u003cp\u003eThe analysis of variability, or standard deviation of EVI, (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) among pixels exhibiting positive trends (SD ranging from approximately 0.06 to 0.09 for pixels such as pixel_45, pixel_20, pixel_52) revealed moderate seasonal variability. This might be indicative of vegetation types that exhibit stable greenness punctuated by brief seasonal fluctuations possibly tied to climatic factors (White et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). Conversely, pixels indicating strong negative trends (pixel_118, pixel_137, and pixel_109) were spatially scattered and mostly peripheral. These as areas are experiencing persistent stress or degradation, with minimal vegetative recovery post-disturbance. Lower EVI values and comparatively lower standard deviation (SD of about 0.03 to 0.06) supported interpretations of continuous vegetation loss or predominant and stagnant condition of intense agricultural land use, deriving from direct anthropogenic impacts, or no adequate hydrological regimes.\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\u003eThe analysis of variability (standard deviation of EVI).\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=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026times;\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" 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\u003ePixel ID\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN Observations\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEVI Trend (Slope)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eEVI SD\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTrend Category\u003c/p\u003e \u003cp\u003e1984\u0026ndash;2024\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003epixel_45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e425\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e+\u0026thinsp;1.241 \u0026times; 10⁻⁵\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.0818\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStrong Positive\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003epixel_15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e429\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e+\u0026thinsp;1.095 \u0026times; 10⁻⁵\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.0860\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStrong Positive\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003epixel_44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e425\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e+\u0026thinsp;1.082 \u0026times; 10⁻⁵\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.0666\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStrong Positive\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003epixel_52\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e426\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e+\u0026thinsp;1.028 \u0026times; 10⁻⁵\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.0725\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStrong Positive\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003epixel_20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e428\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e+\u0026thinsp;1.001 \u0026times; 10⁻⁵\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.0864\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStrong Positive\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003epixel_118\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e425\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;3.871 \u0026times; 10⁻⁶\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.0942\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStrong Negative\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003epixel_137\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e426\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;3.559 \u0026times; 10⁻⁶\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.0542\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStrong Negative\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003epixel_56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e426\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;3.426 \u0026times; 10⁻⁶\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.0560\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStrong Negative\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003epixel_100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e426\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;3.287 \u0026times; 10⁻⁶\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.0674\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStrong Negative\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003epixel_109\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e425\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;3.087 \u0026times; 10⁻⁶\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.0958\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStrong Negative\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\u003eFor targeted and detailed temporal analyses, a subset of pixels representative of the strongest negative and positive categories was selected. This strategic sampling facilitated focused semi-monthly aggregated EVI analysis, further clarifying phenological patterns, ecological stability, and potential stress signals within representative locations. The 15-days semi-monthly aggregated EVI curves smoothed using Generalized Additive Models (GAMs), balanced temporal resolution and signal clarity, enhancing the capacity to detect seasonal shifts, vegetation response to climate variations, and changes associated with hydrological disturbances (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe temporal graphs for random selected pixels (pixel 20, 45, 56, and 137) supports to visualize long-term vegetation trajectories. Pixel 45 and Pixel 20 exhibit generally increasing EVI trends, categorized as strong positive pixels. There is a substantial recovery and subsequent improvement of vegetation conditions following dam construction (2004\u0026ndash;2007), as in fact the post-dam period shows an accelerated increase in vegetation vigour. Pixel 56 presents a relatively stable to slightly declining pattern, with intermittent fluctuations across decades. Despite fluctuations, the EVI values remain relatively consistent.\u003c/p\u003e \u003cp\u003ePixel 137 is distinctly characterized by a strong negative trend. The declining vegetation vigour throughout the observed period indicates ongoing stress.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo refine the trend analysis, the violin plots decadal variability of EVI for selected pixels, were grouped by their long-term trend classification (strong positive vs. strong negative) (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e). Pixel 20 shows a progressive increase in EVI over the decades, particularly pronounced during the 1990s and 2000s. The standard deviation (SD) remains relatively consistent across decades (0.03\u0026ndash;0.07), reflecting stable ecological improvement likely due to favourable environmental conditions. Pixel 45 experiences an ecological transition, shifting from negative trends in the 1980s, to significantly positive trends during the 2000s, comprising an ecological post-dam re-stabilization in the 2010s. Pixel 137 consistently shows negative trends from the 1980s through the 2010s, with a slight positive shift in the 2020s. Variability remains relatively constant (SD around 0.04\u0026ndash;0.06), indicating prolonged ecological stress or ongoing degradation factors. Pixel 56 displays fluctuating patterns, initially stable or slightly positive in the 1980s and 1990s, shifting to strongly negative trends during the 2000s and 2010s, again reflecting ecological responses possibly exacerbated by hydrological changes from the dam construction period. However, a recent stabilization observed in the 2020s suggests potential adaptation. Overall, the negative pixels will have to be furthermore explored by planning specific field surveys in that designated area of interest.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo document species composition and examine the ecological context for interpreting EVI-based positive trends, a field survey was conducted along the riparian corridor corresponding to the areal of Strong Positive trends comprising pixel_69, pixel_68, pixel_60, pixel_59, pixel_52, pixel_51, pixel_46, pixel_45, pixel_38. By aligning in situ observations with remotely sensed data, the survey provided insights into how plant taxa and species composition contribute to observed Positive Vegetation Trends in the Akaki River basin.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAs for that, ground-truthing surveys confirmed that these areas are dominated by non-native eucalyptus trees (\u003cem\u003eEucalyptus\u003c/em\u003e spp.). These species typically thrive under stabilized hydrological regimes, resulting from altered water flow and sediment transport caused by dam construction (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003e). This observation suggests a potential correlation between hydrological alterations and vegetation responses to water stress, aligning closely with Mediterranean ecological succession theory (Castellano et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Such dynamics are also reflected in the EVI variability observed in pixel 45. While remotely sensed positive greening could suggest overall ecological health, as observed in the time-series graphs, field validation indicated that such expansion by eucalyptus may, in fact, mask underlying ecological degradation due to the competitive exclusion of native riparian species. This raises an urgent question of whether stable but reduced flows can preserve a naturally diverse riparian zone, or if additional environmental flow provisions are necessary to sustain native species. Understanding the implications of this eucalyptus-dominated re-greening is therefore essential for future water management decisions, as it highlights the complex interplay between dam operations, groundwater recharge, species adaptability, and overall ecosystem health in water-scarce Mediterranean settings.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThe Klirou-Malounta-Akaki Dam and its downstream environment provide an ideal field setting to examine how water management impacts vegetation in a water-limited ecosystem. The situation here is representative of many small dams in Mediterranean climates where intermittent streams are impounded. Because the dam markedly reduces flood peaks and lengthens the interval between flow events, changes take place in the riparian zone. For instance, without annual floods to reset the system, vegetation can encroach further into the river channel, potentially leading to denser growth (a phenomenon observed in other regulated ephemeral rivers). Additionally, groundwater recharge from controlled releases could improve sub-surface moisture availability, aiding perennial plants during the dry season.\u003c/p\u003e \u003cp\u003eIn this case, ground-truth observations, however, suggest that eucalyptus stands contribute significantly to the EVI signal in certain segments of the corridor. This non-native species can exhibit year-round foliage and relatively consistent canopy vigour, even under moderate water stress. By virtue of their deep-rooted physiology, eucalyptus individuals may benefit from enhanced groundwater levels, whether from natural seepage or incidental dam releases, leading to sustained greenness detectable by satellite indices. In contrast, some native riparian species exhibit more pronounced leaf-shedding or dormancy when surface flows are scarce. Consequently, the constant to increasing greenness trends identified in the time-series may, in part, reflect the proliferation or improved survivorship of eucalyptus in downstream areas. This underscores a subtle ecological shift: rather than native riparian flora dominating a naturally flood-driven corridor, the post-dam environment may favour introduced or drought-resistant species capable of persisting in a stabilized hydrologic regime. Should the dam entirely curtail ordinary seasonal flows, the ecological composition could further skew toward drought-tolerant matrices, including eucalyptus stands, oleander thickets, and other hardy shrubs.\u003c/p\u003e \u003cp\u003eFrom a management perspective, these findings highlight that even in the absence of major flood pulses, downstream ecosystems can undergo notable vegetative changes. A re-greening trend dominated by eucalyptus may indicate sufficient subsurface water availability, potentially from groundwater recharge facilitated by the dam, yet also point to reduced habitat heterogeneity if other native species are outcompeted. Monitoring EVI provides an integrative measure of vegetative health, yet species-level identification (e.g., via ground surveys or higher-resolution imagery) becomes essential to clarify whether the observed greening denotes resilient native recovery or the spread of introduced taxa. These considerations are important for deciding whether future environmental flow provisions are necessary to maintain ecological balance in riparian zones, particularly if dam operation inadvertently supports non-native species at the expense of local biodiversity. As water scarcity and climate variability continue to escalate in Cyprus and the broader Mediterranean, integrative approaches, encompassing vegetation indices, and species-level observations, will be increasingly necessary to guide sustainable water-resource management and preserve riparian ecosystem integrity.\u003c/p\u003e"},{"header":"Future developments","content":"\u003cp\u003eFuture developments will focus on integrating additional data sources and advanced modelling techniques to enhance the precision and robustness of vegetation response analyses. Specifically, incorporating hydrological data from the dam reservoir and climatic variables such as precipitation and temperature will allow accurate estimation of potential evapotranspiration (PET). The use of harmonized datasets combining Sentinel-2 and Landsat imagery will provide improved spatial resolution and temporal consistency, enabling more detailed recalculations of EVI trends and robust ecological interpretations. Additionally, conducting targeted ground-truthing surveys for plots exhibiting negative trends will help identify potential vegetation stress factors and inform appropriate management interventions.\u003c/p\u003e "},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eThe author, M.G., wrote the work.\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eFunding Declaration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was not supported by any grant or fund.\u003c/p\u003e\n"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAybar, C., Wu, Q., Bautista, L., Yali, R., \u0026amp; Barja, A. (2020). \u003cem\u003ergee: An R package for interacting with Google Earth Engine\u003c/em\u003e. \u003cem\u003eJournal of Open Source Software, 5\u003c/em\u003e(51), 2272. https://doi.org/10.21105/joss.02272\u003c/li\u003e\n\u003cli\u003eBerner, L. T., Assmann, J. J., Normand, S., \u0026amp; Goetz, S. J. (2023). LandsatTS: An R package to facilitate retrieval, cleaning, cross‐calibration, and phenological modeling of Landsat time series data. \u003cem\u003eEcography, 2023\u003c/em\u003e(9), e06768.\u003c/li\u003e\n\u003cli\u003eBunn, S. E., \u0026amp; Arthington, A. H. (2002). Basic principles and ecological consequences of altered flow regimes for aquatic biodiversity. \u003cem\u003eEnvironmental Management, 30\u003c/em\u003e(4), 492\u0026ndash;507.\u003c/li\u003e\n\u003cli\u003eCastellano, C., Bruno, D., Comin, F. A., Rey Benayas, J. M., Masip, A., \u0026amp; Jimenez, J. J. (2022). Environmental drivers for riparian restoration success and ecosystem services supply in Mediterranean agricultural landscapes.\u003c/li\u003e\n\u003cli\u003eCoelho, C. D., da Silva, D. D., Amorim, R. S. S., Vasconcelos, B. N. F., Possato, E. L., Filho, E. I. F., ... \u0026amp; Silva, L. V. (2024). Development and Application of an Environmental Vulnerability Index (EVI) for Identifying Priority Restoration Areas in the S\u0026atilde;o Francisco River Basin, Brazil. Land, 13(9), 1475.\u003c/li\u003e\n\u003cli\u003eCyprus Water Development Department. (2007). \u003cem\u003eTechnical report on the Klirou\u0026ndash;Malounta Dam construction and operation\u003c/em\u003e. Nicosia, Cyprus: Ministry of Agriculture, Rural Development and Environment.\u003c/li\u003e\n\u003cli\u003eCyprus Water Development Department. (2019). \u003cem\u003eAnnual report 2018\u0026ndash;2019: Hydrological data and reservoir monitoring\u003c/em\u003e. Nicosia: Government of Cyprus. https://www.moa.gov.cy/wdd\u003c/li\u003e\n\u003cli\u003ede Jong, R., Verbesselt, J., Zeileis, A., \u0026amp; Schaepman, M. E. (2011). Shifts in global vegetation activity trends. \u003cem\u003eRemote Sensing, 3\u003c/em\u003e(3), 595\u0026ndash;614. https://doi.org/10.3390/rs5031117\u003c/li\u003e\n\u003cli\u003eDeVries, B., Huang, C., Armston, J., Huang, W., Jones, J. W., \u0026amp; Lang, M. W. (2020). Rapid and robust monitoring of flood events using Sentinel-1 and Landsat data on the Google Earth Engine. Remote sensing of Environment, 240, 111664.\u003c/li\u003e\n\u003cli\u003eForkel, M., Carvalhais, N., R\u0026ouml;denbeck, C., Keeling, R., Heimann, M., Thonicke, K., ... \u0026amp; Reichstein, M. (2015). Enhanced seasonal CO₂ exchange caused by amplified plant productivity in northern ecosystems. \u003cem\u003eScience, 351\u003c/em\u003e(6274), 696\u0026ndash;699. https://doi.org/10.1126/science.aac4971\u003c/li\u003e\n\u003cli\u003eGorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., \u0026amp; Moore, R. (2017). Google Earth Engine: Planetary-scale geospatial analysis for everyone. \u003cem\u003eRemote Sensing of Environment, 202\u003c/em\u003e, 18\u0026ndash;27. https://doi.org/10.1016/j.rse.2017.06.031\u003c/li\u003e\n\u003cli\u003eGurung, R. B., Breidt, F. J., Dutin, A., \u0026amp; Ogle, S. M. (2009). Predicting Enhanced Vegetation Index (EVI) curves for ecosystem modeling applications. Remote Sensing of Environment, 113(10), 2186-2193.\u003c/li\u003e\n\u003cli\u003eHadjinicolaou, P., Giannakopoulos, C., Zerefos, C., Lange, M. A., Pashiardis, S., \u0026amp; Lelieveld, J. (2011). Mid-21st century climate and weather extremes in Cyprus as projected by six regional climate models. \u003cem\u003eRegional Environmental Change, 11\u003c/em\u003e, 441\u0026ndash;457.\u003c/li\u003e\n\u003cli\u003eLarned, S. T., Datry, T., Arscott, D. B., \u0026amp; Tockner, K. (2010). Emerging concepts in temporary-river ecology. \u003cem\u003eFreshwater Biology, 55\u003c/em\u003e(4), 717\u0026ndash;738. https://doi.org/10.1111/j.1365-2427.2009.02322.x\u003c/li\u003e\n\u003cli\u003eMasek, J. G., Vermote, E. F., Saleous, N. E., Wolfe, R., Hall, F. G., Huemmrich, K. F., ... \u0026amp; Gao, F. (2006). A Landsat surface reflectance dataset for North America, 1990\u0026ndash;2000. \u003cem\u003eIEEE Geoscience and Remote Sensing Letters, 3\u003c/em\u003e(1), 68\u0026ndash;72. https://doi.org/10.1109/LGRS.2005.857030\u003c/li\u003e\n\u003cli\u003eNagler, P. L., Scott, R. L., Westenburg, C., Cleverly, J. R., Glenn, E. P., \u0026amp; Huete, A. R. (2005). Evapotranspiration on western US rivers estimated using the Enhanced Vegetation Index from MODIS and data from eddy covariance and Bowen ratio flux towers. Remote sensing of environment, 97(3), 337-351.\u003c/li\u003e\n\u003cli\u003eNilsson, C., Reidy, C. A., Dynesius, M., \u0026amp; Revenga, C. (2005). Fragmentation and flow regulation of the world\u0026rsquo;s large river systems. \u003cem\u003eScience, 308\u003c/em\u003e(5720), 405\u0026ndash;408. https://doi.org/10.1126/science.1107887\u003c/li\u003e\n\u003cli\u003ePekel, J.-F., Cottam, A., Gorelick, N., \u0026amp; Belward, A. S. (2016). High-resolution mapping of global surface water and its long-term changes. \u003cem\u003eNature, 540\u003c/em\u003e(7633), 418\u0026ndash;422. https://doi.org/10.1038/nature20584\u003c/li\u003e\n\u003cli\u003ePetts, G. E., \u0026amp; Gurnell, A. M. (2005). Dams and geomorphology: Research progress and future directions. \u003cem\u003eGeomorphology, 71\u003c/em\u003e(1\u0026ndash;2), 27\u0026ndash;47. https://doi.org/10.1016/j.geomorph.2004.02.015\u003c/li\u003e\n\u003cli\u003ePoff, N. L., Olden, J. D., Merritt, D. M., \u0026amp; Pepin, D. M. (2007). Homogenization of regional river dynamics by dams and global biodiversity implications. \u003cem\u003eProceedings of the National Academy of Sciences, 104\u003c/em\u003e(14), 5732\u0026ndash;5737. https://doi.org/10.1073/pnas.0609812104\u003c/li\u003e\n\u003cli\u003eSeddon, A. W. R., Macias-Fauria, M., Long, P. R., Benz, D., \u0026amp; Willis, K. J. (2016). Sensitivity of global terrestrial ecosystems to climate variability. \u003cem\u003eNature, 531\u003c/em\u003e(7593), 229\u0026ndash;232. https://doi.org/10.1038/nature16986\u003c/li\u003e\n\u003cli\u003eShi, H., Li, L., Eamus, D., Huete, A., Cleverly, J., Tian, X., ... \u0026amp; Carrara, A. (2017). Assessing the ability of MODIS EVI to estimate terrestrial ecosystem gross primary production of multiple land cover types. Ecological Indicators, 72, 153-164.\u003c/li\u003e\n\u003cli\u003eShumway, R. H., Stoffer, D. S., \u0026amp; Stoffer, D. S. (2000). Time series analysis and its applications (Vol. 3, p. 4). New York: springer.\u003c/li\u003e\n\u003cli\u003eTockner, K., \u0026amp; Stanford, J. A. (2002). Riverine flood plains: Present state and future trends. \u003cem\u003eEnvironmental Conservation, 29\u003c/em\u003e(3), 308\u0026ndash;330. https://doi.org/10.1017/S037689290200022X\u003c/li\u003e\n\u003cli\u003eU.S. Geological Survey (USGS). (2020). \u003cem\u003eLandsat Collection 2 Level-2 Science Products\u003c/em\u003e. https://www.usgs.gov/core-science-systems/nli/landsat/landsat-collection-2\u003c/li\u003e\n\u003cli\u003eVermote, E., Justice, C., Claverie, M., \u0026amp; Franch, B. (2016). Preliminary analysis of the performance of the Landsat 8/OLI land surface reflectance product. \u003cem\u003eRemote Sensing of Environment, 185\u003c/em\u003e, 46\u0026ndash;56. https://doi.org/10.1016/j.rse.2016.04.008\u003c/li\u003e\n\u003cli\u003eWhite, M. A., de Beurs, K. M., Didan, K., Inouye, D. W., Richardson, A. D., Jensen, O. P., ... \u0026amp; Lauenroth, W. K. (2009). Intercomparison, interpretation, and assessment of spring phenology in North America estimated from remote sensing for 1982\u0026ndash;2006. Global Change Biology, 15(10), 2335-2359. \u003c/li\u003e\n\u003cli\u003eZhao, B., Yan, Y., Guo, H., He, M., Gu, Y., \u0026amp; Li, B. (2009). Monitoring rapid vegetation succession in estuarine wetland using time series MODIS-based indicators: an application in the Yangtze River Delta area. Ecological indicators, 9(2), 346-356.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"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":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Landsat, GEE, Riverine Ecosystem, Drylands, Ecology","lastPublishedDoi":"10.21203/rs.3.rs-6761511/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6761511/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study presents a methodological framework to assess the impact of dam construction on vegetation dynamics, using the Klirou\u0026ndash;Malounta\u0026ndash;Akaki Dam in Cyprus as a case study. Unlike many dams that incorporate controlled flood pulses, this structure lacks a designed flood regime, raising the question of whether observed changes in riparian vegetation are primarily driven by localized microclimatic or geomorphological factors rather than direct hydrological alterations. A temporal analysis covering 20 years before and after dam construction (1984\u0026ndash;2024) was implemented to distinguish baseline vegetation conditions from post-construction shifts. The analysis focused on an upstream area within the Serrahis River watershed. Enhanced Vegetation Index (EVI) time series, derived from Landsat 5, 7, and 8 and processed via Google Earth Engine and R, were harmonized using a Random Forest cross-sensor calibration approach. Pixel-wise trend analysis at 30 m resolution revealed cluster areas of vegetation increase, suggesting localized greening effects. However, ground-truthing surveys revealed that such increases were often associated with the spread of non-native Eucalyptus spp., a drought-tolerant species adapted to stable or intermittent surface flows. While the EVI trends may imply ecosystem recovery, they also indicate a shift in species composition, potentially favouring resilient non-native taxa at the expense of native species reliant on seasonal flooding. These findings underscore the importance of interpreting vegetation indices in ecological context. Although stabilized hydrology may enhance water storage, it can simultaneously drive biodiversity loss, highlighting the need for balanced water management strategies in water scarce Mediterranean environments.\u003c/p\u003e","manuscriptTitle":"Assessing Vegetation Responses to Dam-Induced Hydrological Change: A 40-Year Landsat Time-Series Analysis in the Downstream Serrahis River Watershed, Cyprus","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-06-17 09:20:31","doi":"10.21203/rs.3.rs-6761511/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c94685c4-63c8-4cd9-8d0d-682ee1daddc1","owner":[],"postedDate":"June 17th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-01-05T00:08:31+00:00","versionOfRecord":[],"versionCreatedAt":"2025-06-17 09:20:31","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6761511","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6761511","identity":"rs-6761511","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}