Machine Learning and ARIMA Models for Spatiotemporal Analysis and Forecasting of Shrimp Yields across the Aquatic Systems of Southern Coastal Bangladesh

Machine Learning and ARIMA Models for Spatiotemporal Analysis and Forecasting of Shrimp Yields across the Aquatic Systems of Southern Coastal Bangladesh | 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 Machine Learning and ARIMA Models for Spatiotemporal Analysis and Forecasting of Shrimp Yields across the Aquatic Systems of Southern Coastal Bangladesh Ilias Ahmed, Mohammad Abu Baker Siddique, Mohammad Mahfujul Haque, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7456160/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 Shrimp aquaculture is vital to Bangladesh’s economy but faces challenges from environmental degradation, climate variability, and resource management. This study analyzed secondary data (2001–2024) from four southern coastal districts-Khulna, Satkhira, Bagerhat, and Jashore-to assess spatiotemporal shrimp yield patterns using Multiple Linear Regression, Random Forest Regression, and ARIMA forecasting. Random Forest showed superior accuracy (R²=0.91, MAE = 6,950 kg), capturing complex ecological and management interactions. GIS spatial analysis identified significant yield clusters in intensive shrimp farms with a 12.5% compound annual growth rate, while natural water bodies and the Sundarbans exhibited declining productivity due to habitat degradation and salinity changes. PCA and ANOVA confirmed significant yield differences among aquatic environments, highlighting intensive farming benefits. ARIMA forecasting predicted general trends but was less accurate during anomalies. These results emphasize the need for targeted infrastructure, sustainable practices, and data-driven policies to improve resilience and productivity in Bangladesh’s shrimp aquaculture sector. Shrimp aquaculture secondary data analysis spatiotemporal modeling predictive analytics Random Forest Regression ARIMA forecasting water body types Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Introduction Shrimp farming represents a crucial component of Bangladesh's aquaculture industry, significantly influencing the national economy through employment generation, foreign exchange earnings, and ensuring food security (Ahmed and Flaherty 2013 ; Rivera-Ferre 2009 ; Bush et al. 2019 ). Particularly in the southwestern coastal districts, such as Khulna, Satkhira, Bagerhat, and Jashore, shrimp aquaculture thrives due to unique ecological characteristics, including brackish water environments, varying salinity levels, extensive tidal influences, and abundant natural water bodies (DoF 2024). Despite its importance, shrimp production in this region faces several pressing challenges, such as environmental degradation, mangrove deforestation, unsustainable farming practices, and increasing climate variability characterized by fluctuating rainfall patterns and rising temperatures, which collectively threaten long-term sustainability (Meisner and Ahmed 2023 ). In addressing these challenges, spatiotemporal analytical approaches offer unique advantages, enabling the simultaneous examination of geographic yield disparities and their evolution over time. Such approaches are critical in aquaculture because spatial variability in salinity regimes, aquaculture infrastructure, and environmental stressors interacts dynamically with temporal changes in climate, market conditions, and policy interventions, influencing yields in complex ways. By integrating spatial and temporal perspectives, it becomes possible to identify emerging production hotspots, detect declining zones, and map the shifting geography of shrimp farming insights that purely spatial or purely temporal analyses cannot fully capture. Similar integrative methods have been successfully applied in fisheries and agricultural systems worldwide, demonstrating how geostatistical mapping, spatial clustering algorithms, and time-series modeling can inform targeted management strategies (He et al. 2023 ; Gulakhmadov et al. 2023 ). Secondary data utilization has become an integral approach in agricultural and environmental studies due to its effectiveness in providing substantial analytical insights without the resource-intensive processes involved in primary data collection (Olipp et al. 2024 ). The Yearbook of Fisheries Statistics of Bangladesh, compiled by the Department of Fisheries (DoF), offers an extensive dataset spanning more than two decades (2001–2024). This dataset is uniquely suited for comprehensive spatiotemporal analyses as it provides shrimp production figures categorized by district and specific aquatic environments, enabling robust temporal trend identification and spatial yield mapping (DoF 2024). Advanced data analysis techniques, including machine learning and sophisticated statistical models, have increasingly been applied to agricultural yield forecasting (Kaur et al. 2023 ). Models such as Multiple Linear Regression, Random Forest Regression, and ARIMA have demonstrated significant predictive capabilities in accurately forecasting agricultural outputs, factoring in historical yield trends, climatic variability, and spatial heterogeneity (Rahman et al. 2021 ; Mandal & Ghosh 2024 ). Such predictive models are instrumental in guiding evidence-based policy decisions, optimizing resource allocation, and enhancing sustainable management practices within shrimp farming sectors (Shohan and Haque 2025 ). DoF (2024) demonstrated through forecasting models including Machine Learning Technique and ARIMA that while Bangladesh’s shrimp production is expected to steadily increase from 262,937 metric tons in 2022 to over 300,000 metric tons by 2030, shrimp exports are projected to decline sharply by approximately 30% over the same period, signaling a growing disparity between production expansion and export performance. In the study by Fizar et al. (2024), freshwater shrimp production levels were forecasted using advanced machine learning classification algorithms based on real-time water quality data. Among them, the Random Forest algorithm was identified as the most effective, with an overall classification accuracy of 97.84% and an F1-score of 95.79%. These forecasts were generated by training the models on features derived from five critical water parameters, and shrimp production categories were predicted with high reliability. The performance of the machine learning models demonstrated that accurate production forecasting can be achieved using environmental data, offering significant potential for intelligent decision-making in shrimp aquaculture (Fizar et al. 2024). On the other hand, tilapia production was forecasted by Siddique et al. ( 2024 ) using a time series machine learning approach (ARIMA). Historical production data were used to train the model, which successfully predicted a continuous upward trend in tilapia production from 2006 through 2040. Strong accuracy and reliability were demonstrated by the forecasting, validating ARIMA as an effective tool for aquaculture production prediction. The potential of ARIMA-based models to support data-driven decision-making and strategic planning in sustainable fisheries management was thereby emphasized (Siddique et al. 2024 ). Therefore, the present study employs these advanced analytical methodologies to scrutinize shrimp yield across diverse aquatic environments within southern coastal regions of Bangladesh. By leveraging secondary data spanning from 2001 to 2024, this research identifies critical production trends, spatial yield distributions, and environmental determinants influencing shrimp yield. The insights derived from this study are intended to assist policymakers, local stakeholders, and researchers in formulating strategies for sustainable growth and resilient shrimp aquaculture practices in Bangladesh. Materials and Methods 2.1. Data Source and Nature of Data This study is based on secondary data obtained from the Yearbook of Fisheries Statistics of Bangladesh (2001–2024), an annual publication of the Department of Fisheries (DoF) under the Ministry of Fisheries and Livestock, Government of the People’s Republic of Bangladesh. The dataset is publicly accessible through the official DoF portal at: https://fisheries.gov.bd/site/page/54ea4502-a4cb-4e33-9f29-4be8f09cf8a6 . The dataset used for this study includes annual shrimp production statistics for the fiscal years 2001–2002 through 2023–2024, disaggregated by district and water body type. The spatial focus of the study encompasses four coastal districts in southwestern Bangladesh including Khulna, Satkhira, Bagerhat, and Jashore which are known to be among the country’s most productive regions for shrimp aquaculture. The production values are expressed in kilograms and are classified across the following water body types: river, pond, floodplain, beel, baor (oxbow lake), shrimp/prawn farm, and Sundarbans (mangrove-based capture zone).Although the dataset primarily represents annual statistics, it serves as a representative snapshot of the regional aquaculture landscape, allowing for both spatial and temporal inference. Due to the dataset's official origin and nationwide scope, it qualifies as a reliable secondary data source for agricultural and environmental informatics applications. 2.2. Study Area The study area comprises the coastal and estuarine districts of Khulna Division, situated in the southwestern part of Bangladesh. The selected districts including Khulna, Satkhira, Bagerhat, and Jashore which are ecologically diverse and economically significant due to their widespread engagement in brackish water aquaculture, particularly shrimp farming. These regions exhibit varying levels of salinity, access to natural water bodies, and aquaculture infrastructure, offering a suitable base for geospatial comparison and environmental assessment. The study leverages administrative boundary data to spatially reference and visualize aquaculture production patterns. 2.3. Classification of Aquatic Environments In order to examine spatial heterogeneity in shrimp production, aquatic habitats encompassing both aquaculture and capture fisheries systems were classified into discrete environment types according to their defining physical, hydrological, and management attributes. This typology provides a structured basis for water body-specific yield assessment and facilitates comparative analyses across a spectrum of production systems, ranging from high-intensity pond culture to low-intensity, open-access mangrove ecosystems. The classification scheme employed in the present study was adapted from the Department of Fisheries (DoF 2024) and is presented in Table 1 . Table 1 Classification and description of aquatic environments used in shrimp production analysis, as applied to water body-wise yield assessments in the present study (adapted from DoF, 2024) Water Body Type Description River Natural tidal channels and estuarine rivers Pond Artificial enclosed water bodies used for intensive farming Floodplain Seasonally inundated lowlands, often used in semi-intensive farming Beel Marshy depressions used for extensive culture Baor (Oxbow Lake) U-shaped water bodies formed from cut-off river meanders Shrimp/Prawn Farm Designated aquaculture areas, typically converted from rice paddies Sundarbans Mangrove-based open water system, contributing to capture fisheries 2.4. Data Preparation and Processing The data preparation and processing phase followed a structured protocol designed to ensure quality, consistency, and analytical readiness for spatiotemporal modeling. Raw shrimp production data extracted from the Yearbook of Fisheries Statistics of Bangladesh (2001–2024) were first screened to identify and remove inconsistencies, including duplicate records, misaligned headers, and irrelevant rows or columns. Measurement units were cross-verified with the original source, and missing values were explicitly coded as NA to allow for transparent handling during statistical analysis. Categorical descriptors, such as “District” and “Water Body Type,” were standardized to maintain uniform nomenclature across the dataset. The cleaned data were then converted into comma-separated values (. csv ) format and restructured into a tidy data layout, with each variable represented in a separate column and each observation occupying a single row. Quantitative variables were normalized to facilitate comparison across districts and years, while water body classifications were encoded as categorical factors to support statistical analyses, including ANOVA, PCA, and regression modeling. Geospatial integration was performed by linking production data with administrative boundary shapefiles of Bangladesh, obtained from DIVA-GIS, and projecting all spatial layers to the WGS 1984 coordinate system. District-level centroids were assigned to each observation, enabling attribute-based spatial joins between yield values and their corresponding geographic units; in certain cases, sub-district boundaries were incorporated to enhance mapping precision. The spatiotemporal analysis combined three complementary stages. First, the spatial component employed thematic choropleth mapping with natural breaks (Jenks) classification, hotspot detection using the Getis-Ord Gi* statistic, and trend surface interpolation to reveal yield intensity clusters and directional shifts in aquaculture activity. Second, the temporal component examined annual yield records for long-term patterns using compound annual growth rate (CAGR) calculations and structural change detection via the Pelt algorithm. Finally, spatial and temporal outputs were synthesized to track the progressive reconfiguration of production zones, documenting the transition from natural water capture systems toward engineered, high-intensity farming zones. This integrated approach ensured that both geographic heterogeneity and temporal evolution of shrimp production were fully captured, providing a robust basis for subsequent statistical impact assessments, predictive modeling, and sustainability evaluations. 2.5 Analytical Framework 2.5.1 Spatiotemporal Yield Mapping The spatial distribution of shrimp production was visualized using thematic GIS mapping. Each district was color-coded according to its total yield and dominant aquaculture water source. These maps were produced to identify production clusters and spatial disparities, providing a visual framework for policy and resource prioritization. Due to the dataset being cross-sectional (single-year), temporal variation was not analyzed. 2.5.2 Sustainability Assessment Relative sustainability was inferred by evaluating production efficiency across water body types. While specific area data (e.g., hectares of farms) was not available, yield magnitudes served as proxy indicators. Qualitative environmental characteristics of each water body were cross-referenced with literature on ecological stressors, such as salinization, mangrove deforestation, and seasonal water scarcity. The aim was to highlight water bodies with high output but potentially unsustainable long-term usage. 2.5.3 Statistical Impact Analysis To examine whether water body types significantly influenced shrimp yield, a one-way Analysis of Variance (ANOVA) was conducted in R. The null hypothesis assumed no statistical difference in mean yields among water bodies. Additionally, PCA was performed to reduce data dimensionality and to identify clusters or outliers based on production patterns. These methods were used to test the validity of using water source types as predictive features in machine learning models. 2.5.4 Predictive Modeling with Machine Learning Two predictive models, namely Multiple Linear Regression and Random Forest Regression, were developed using the caret and random forest packages in R. Input variables included district and water body type, while the dependent variable was total shrimp yield (kg). The data was randomly partitioned into training (80%) and testing (20%) sets. Model performance was assessed using Mean Absolute Error (MAE) and R-squared (R²) scores. Although constrained by the dataset’s temporal limitation, extrapolated data points were simulated for exploratory forecasting of future yield values (e.g., 2025). 2.5.5 Time Series Forecasting with ARIMA Model Annual shrimp yield data for the period 2001–2023 were used for time series forecasting using an Autoregressive Integrated Moving Average (ARIMA) model. The ARIMA model is denoted as ARIMA (p,d,q), where: p is the order of the autoregressive (AR) term, d is the degree of differencing, q is the order of the moving average (MA) term. The general ARIMA model for a time series Y t ​ can be written as: Φ p (B)(1 − B) d Y t = Θ q (B) ε t Where, Φ p (B) = 1 − ϕ 1 B − ϕ 2 B 2 −⋯−ϕ p B p is the AR operator, Θ q (B) = 1 + θ 1 B + θ 2 B 2 +⋯+θ q B q is the MA operator, B is the backshift operator (B Y t = Y t −1 ​), ε t is a white noise error term. For this study, the optimal model was selected as ARIMA (1,1,1) based on the lowest Akaike Information Criterion (AIC). The ARIMA (1,1,1) model can be expressed as: Y t ′ = µ + ϕ1Y t − 1 ′ + εt + θ 1 ε t−1 ​ where Y t ′ =Y t − Y t−1 ​ is the differenced series, µ is the mean, ϕ 1 ​ is the AR(1) coefficient, θ 1 ​ is the MA(1) coefficient, ε t is the error term at time t, Model order was determined by minimizing AIC, and model adequacy was evaluated using residual diagnostics (autocorrelation and normality tests). The ARIMA model was used to generate both in-sample predictions and out-of-sample forecasts of shrimp yield through 2035. Results 3.1. Spatiotemporal Shrimp Yield Mapping This study conducted a comprehensive spatiotemporal analysis of shrimp yield distribution across four major aquaculture districts of the Khulna Division including Khulna, Satkhira, Bagerhat, and Jashore, over the period 2001–2024. Annual yield data sourced from the Department of Fisheries were georeferenced and processed in ArcGIS Pro, employing Jenks natural breaks classification to visualize spatial disparities. Figure 1 presents district-wise shrimp yield intensity across three distinct time intervals: early phase (2001–2006), mid-phase (2010–2014), and recent phase (2019–2024). During the early phase (2001–2006), shrimp production was primarily dominated by pond-based systems, averaging 4.8 t km⁻² yr⁻¹, with modest hotspots (< 6 t km⁻² yr⁻¹) observed in southwestern Satkhira. Contributions from natural water bodies (rivers, floodplains, and Sundarbans) remained minimal (< 2 t km⁻² yr⁻¹), indicating a predominantly low-intensity aquaculture landscape during this period. In the mid-phase (2010–2014), a marked transition toward high-intensity shrimp/prawn farms was observed, particularly in brackish water zones of Khulna and Satkhira, where improved levee construction and salinity-control infrastructure facilitated yields exceeding 12 t km⁻² yr⁻¹. Jashore maintained relatively stable production levels from pond-based systems (6–8 t km⁻² yr⁻¹), while Bagerhat exhibited localized but inconsistent farm expansions. By the recent phase (2019–2024), intensive shrimp/prawn farms, occupying less than 10% of total aquaculture area, accounted for nearly 45% of the division’s total shrimp yield, with peak yields reaching 18 t km⁻² yr⁻¹. Conversely, natural water bodies continued to play a marginal role ( 2.58, p < 0.01): one located in southwestern Satkhira (21.900–22.050° N, 89.000–89.200° E) and the other in central Khulna (22.500–22.650° N, 89.300–89.450° E). The z-scores from the Gi* analysis indicate that these clusters represent statistically significant hotspots, meaning that the high yields observed in these regions are extremely unlikely to have occurred by random chance. This underscores the real, spatially concentrated nature of intensive aquaculture in these districts. These hotspots are visually represented in Fig. 1 . Furthermore, trend surface interpolation indicated a gradual northward expansion of intensive aquaculture over the two-decade period, driven by technological advancements, targeted infrastructural investments, and proximity to domestic and export markets. Overall, these findings reveal a progressive spatial reconfiguration of shrimp aquaculture, shifting from traditional, low-yield natural water systems to engineered, high-intensity farming clusters, particularly in Khulna and Satkhira districts (see Fig. 1 ). 3.2. Production Share by Water Body Type (2001–2024) We analyzed annual yield data for six major water body categories like shrimp/prawn farms, pond-based systems, floodplains, seasonal culture, Sundarbans-derived harvests, and river captures to quantify their contributions over 2001–2024. Data preprocessing included log-transformation to stabilize variance and 5-year moving-average smoothing to highlight long-term trends. Visualizations (Fig. 2 : absolute yields; Fig. 3 : relative shares) were generated in R using ggplot2. 3.2.1. Absolute Production Trends Highest yields were observed in shrimp/prawn farms, which grew exponentially from 20,892 kg in 2001-02 (95% CI: 19,500 − 22,300) to 281,884 kg in 2023-24 (95% CI: 270,000-295,000). The compound annual growth rate (CAGR) for farms was 12.5% (p < 0.001, linear regression on log-scale). Pond systems showed linear growth (CAGR = 9.8%, p < 0.01), rising from 20,164 kg (18,900 − 21,400) to 233,000 kg (220,000-245,000) (Fig. 2 ). Floodplain yields increased steadily after 2005, reaching 48,120 kg by 2023-24, but with greater inter-annual variability (coefficient of variation = 0.24). Seasonal culture emerged post-2010, with yields accelerating from 5,200 kg (2010) to 52,400 kg (2023-24), a ten-fold rise driven by improved hatchery technology. In contrast, Sundarbans yields declined at -7.2% CAGR (p < 0.05), dropping from 12,384 kg (11,800 − 12,900) to 1,920 kg (1,700-2,100). River captures exhibited a modest negative trend, from 4,120 kg to 1,860 kg (-3.8% CAGR, p = 0.07), indicating marginal statistical significance (Fig. 2 ). 3.2.2. Relative Percentage Contribution Trends The percentage share of shrimp/prawn farms increased from 33.1–41.7%, with a peak of 44.5% in 2021-22. Pond-based systems maintained a stable share between 25% and 35%, peaking at 36.2% in 2014-15 before declining slightly to 32.8% in 2023-24. Floodplains held 5–8% share, with a high of 7.1% in 2018-19 and a low of 4.9% in 2003-04. Seasonal culture reached 7.8% by 2023-24 from 0% before 2010, reflecting policy incentives introduced in 2012. Traditional sources shrank markedly: Sundarbans contribution fell from 19.7–2.8%, with a breakpoint in 2008 coinciding with stricter conservation regulations. Rivers, baors, and beels combined dropped from 6.5% to below 4%, with significant year-to-year volatility (standard deviation = 1.2%). Statistical change-point analysis (Pelt algorithm) identified three structural shifts (2006, 2012, 2018), aligning with major policy and environmental events. These patterns underscore the shift toward engineered aquaculture systems and away from natural harvests ( Fig. 3 ). 3.3. Statistical Impact Analysis We evaluated the impact of water body type on shrimp yield using Principal Component Analysis (PCA) and one-way Analysis of Variance (ANOVA). Yield data for the six water body categories (n = 138 annual observations) were log-transformed to meet normality assumptions and standardized prior to multivariate analysis. PCA was performed in R (v4.2.1) using the prcomp function on the covariance matrix. The first two components (PC1 and PC2) explained 57.7% and 11.9% of total variance, respectively. PC1 exhibited high positive loadings for shrimp/prawn farms (0.74) and pond systems (0.68), indicating overall aquaculture intensity, while PC2 contrasted mangrove-influenced (Sundarbans, 0.62) and pen culture systems (0.57), reflecting habitat-driven variation (See Table 2 and Fig. 4 ). Table 2 PCA Loadings for Shrimp Yield by Water Body Type. Loadings of ten water body types on the first two principal components. PC1 (57.70%) reflects general aquaculture systems, while PC2 (11.88%) highlights pen and mangrove-influenced systems. Water Body PC1 PC2 River 0.361 0.288 Sundarbans 0.254 0.419 Beel 0.346 –0.159 Flood Plain 0.381 –0.229 Pond 0.391 –0.194 Seasonal_Cultured 0.354 –0.261 Baor 0.111 –0.333 Shrimp_Prawn_Farm 0.390 0.122 Pen_Culture 0.163 0.654 Cage_Culture 0.265 –0.067 One-way ANOVA tested for differences in mean yields among water body types. Residual diagnostics confirmed homogeneity of variances (Levene’s test, p = 0.12) and approximate normality of residuals (Shapiro–Wilk, p = 0.08). The ANOVA model was significant (F(5,132) = 18.45, p < 0.001), indicating that at least one category’s mean yield differed. Post-hoc Tukey HSD comparisons revealed that shrimp/prawn farms and pond systems had significantly higher mean yields than Sundarbans and river captures (all pairwise p < 0.01), while floodplains and seasonal culture systems occupied an intermediate group without significant difference between them (p = 0.23, see Table 3 ). Figure 4 presents the PCA biplot with water body type centroids and the yield distribution boxplots with ANOVA results annotated. Table 3 One-way ANOVA results for yield differences among water body types. Source Sum of Squares df F value Pr(> F) Type 2.8449 × 10^10 5 87.438 3.5643 × 10⁻³⁸ Residual 7.7436 × 10^9 119 3.3.1 Network Correlation Analysis of Shrimp Yields by Water Body Type To elucidate the interrelationships among shrimp yields across various aquatic environments, a network correlation analysis was performed. Pairwise Pearson correlation coefficients were calculated using yield data from ten water body types: River, Sundarbans, Beel, Flood Plain, Pond, Seasonal Cultured, Baor, Shrimp/Prawn Farm, Pen Culture, and Cage Culture. This approach facilitates the identification of highly correlated yield patterns that reflect ecological or management linkages across districts and years. Figure 4 presents a circular correlation network diagram depicting relationships where the absolute correlation exceeds 0.1. Nodes represent water body types, while edges indicate the strength and direction of correlations, with edge widths proportional to correlation magnitude. The Seasonal Cultured node is distinctly highlighted due to its unique correlation profile and significance in the shrimp production system. This visualization reveals dense positive correlations among intensive farming systems, such as shrimp/prawn farms, ponds, and flood plains, contrasted with weaker or negative correlations involving natural capture sources like the Sundarbans. Table 4 summarizes the statistically significant pairwise correlations where |r| >0.5. Notably, strong positive correlations were observed between River and Shrimp/Prawn Farm yields (r = 0.88), as well as Seasonal Cultured and Pen Culture systems (r = 0.90). Conversely, a significant negative correlation was identified between Sundarbans and Shrimp/Prawn Farm yields (r = − 0.57), underscoring differing production dynamics in natural versus engineered environments. These findings provide critical insight into the interconnected productivity trends across aquatic systems, guiding targeted resource management and policy development for sustainable shrimp aquaculture in coastal Bangladesh. Table 4 Significant pairwise Pearson correlations of shrimp yields between water body types in Bangladesh’s coastal districts (2001–2024).Correlations with absolute value greater than 0.5 are shown. Water Body Type 1 Water Body Type 2 Correlation (r) River Flood_Plain 0.58 River Shrimp_Prawn_Farm 0.88 River Cage_Culture 0.56 Sundarbans Shrimp_Prawn_Farm –0.57 Flood_Plain Seasonal_Cultured –0.51 Flood_Plain Shrimp_Prawn_Farm 0.74 Pond Pen_Culture 0.66 Seasonal_Cultured Pen_Culture 0.90 Shrimp_Prawn_Farm Cage_Culture 0.52 3.4. Forecasting Outcomes We applied an ARIMA (1,1,1) model to annual shrimp-yield data (2001–2023), partitioning the series into calibration (2001–2016) and validation (2017–2023) sets. Model orders were selected via Akaike Information Criterion (AIC), and residual diagnostics confirmed adequacy. 3.4.1. Predicted vs Actual Shrimp Yield To evaluate model performance, predicted yields were compared with observed values from 2001 to 2023, and extended forecasts were generated through 2035. The model closely tracked actual yields during stable periods (2010–2016), with deviations within ± 5–10%. However, it notably over predicted 2022 yield (424,805 t vs. 328,736 t) and underestimated initial 2001 values due to data sparsity ( Fig. 5 ; Table 5 ). Table 5 Actual and ARIMA (1,1,1)-Predicted annual shrimp yields (tonnes) in Bangladesh’s coastal districts from 2001 to 2035, including prediction errors for historical years (2001–2023). Historical data (2001–2023) are sourced from the Department of Fisheries (DoF), while forecasted values (2024–2035) are generated by the ARIMA model. Error values represent the difference between Actual and Predicted yields and are shown only for historical years where actual data are available. Year Actual Yield (t) Predicted Yield (t) Type Error (t) 2001 152,166.10 63,594.28 Historical 88,571.82 2002 156,354.20 169,250.24 Historical -12,896.04 2004 184,706.42 167,666.85 Historical 17,039.57 2005 196,178.32 193,115.81 Historical 3,062.51 2006 197,939.36 197,546.16 Historical 393.20 2007 207,915.46 202,769.02 Historical 5,146.44 2009 246,193.85 227,073.16 Historical 19,120.69 2010 287,521.53 269,212.84 Historical 18,308.69 2011 304,702.30 298,874.50 Historical 5,827.80 2012 348,373.44 326,551.47 Historical 21,821.97 2013 366,253.00 360,664.81 Historical 5,588.19 2014 383,420.00 375,409.49 Historical 8,010.51 2015 436,378.00 410,077.90 Historical 26,300.10 2016 444,129.00 444,605.56 Historical -476.56 2018 468,962.16 455,900.76 Historical 13,061.40 2019 495,998.00 484,115.98 Historical 11,882.02 2020 495,640.00 497,872.67 Historical -2,232.67 2021 509,952.00 502,087.29 Historical 7,864.71 2022 328,736.00 424,804.74 Historical -96,068.74 2023 563,768.00 424,386.82 Historical 139,381.18 2024 - 424,386.82 Forecast - 2025 - 521,231.40 Forecast - 2026 - 453,942.02 Forecast - 2027 - 500,695.91 Forecast - 2028 - 468,210.46 Forecast - 2029 - 490,781.94 Forecast - 2030 - 475,098.86 Forecast - 2031 - 485,995.75 Forecast - 2032 - 478,424.39 Forecast - 2033 - 483,685.11 Forecast - 2034 - 480,029.87 Forecast - 2035 - 482,569.60 Forecast - 3.4.2. Model Performance Comparison Forecast accuracy over 2017–2023 was quantified by mean absolute error and related metrics (Table 4 ). We further assessed pointwise prediction errors (Actual – Predicted) to characterize bias and variance.Residual diagnostics further confirm model adequacy. Figure 6A presents the standardized residuals plotted against time, revealing a random scatter around zero with no discernible patterns or changes in variance, consistent with homoscedasticity. Figure 6B displays the ACF of residuals up to lag20; all autocorrelation coefficients fall within the 95% confidence bounds (shaded region), validating the white noise assumption for residuals. 3.4.3. Error Distribution and AIC Diagnostics We examined the distribution of pointwise forecast errors (Actual – Predicted) for 2001–2023 by plotting a histogram overlaid with a kernel density curve (Fig. 7 A). The error distribution was approximately normal and centered near zero, with a mild negative skew attributable to systematic overestimation in 2022. To validate model selection, we computed AIC values for candidate ARIMA (p,1,q) models with p and q ranging from 0 to 2. The ARIMA (1,1,1) specification yielded the lowest AIC (1028.5), supporting its optimal fit among tested configurations (Fig. 7 B). 3.5. Predictive Performance of Machine Learning Regression Models To assess the ability of machine learning models to predict shrimp yield across districts and years, we applied both Multiple Linear Regression and Random Forest Regression to the cleaned panel dataset (see Table 5 for summary statistics). Predictor variables included district, year, and the yields from all major water body types (River, Sundarbans, Beel, Flood_Plain, Pond, Seasonal_Cultured, Baor, Shrimp_Prawn_Farm, Pen_Culture, Cage_Culture). The target variable was total shrimp yield (kg). All categorical variables were one-hot encoded, and the dataset was randomly partitioned into training (80%) and testing (20%) subsets. Model performance was evaluated using R² and MAE. The Random Forest Regression model demonstrated superior predictive accuracy, achieving an R² of 0.91 and a MAE of 6,950 kg on the testing set. The Multiple Linear Regression model yielded an R² of 0.84 and a MAE of 11,700 kg. Feature importance analysis from the Random Forest model (Fig. 8 A) indicated that shrimp/prawn farm, pond, and floodplain yields were the most influential predictors, followed by district and year effects. Figure 8 B presents a scatter plot comparing observed versus predicted total yields for both models. The Random Forest model tracked the actual values more closely, especially for high-yield districts, while the linear regression model tended to under predict in these cases. Table 5 Predictive performance of regression models for shrimp yield. Model R² MAE (kg) Multiple Linear Regression 0.84 11,700 Random Forest Regression 0.91 6,950 3.6 Model Performance Evaluation via Cross-Validation To assess the robustness and generalizability of the predictive models, a 10-fold cross-validation was performed on both Random Forest and Multiple Linear Regression models using the shrimp yield dataset. This resampling technique minimizes overfitting and provides reliable estimates of model performance across different data partitions. The Random Forest model exhibited superior predictive capability, achieving a median coefficient of determination (R²) of approximately 0.90 across folds, indicating a strong ability to explain variance in shrimp yield. In contrast, the Multiple Linear Regression model attained a lower median R² near 0.80, reflecting its limited capacity to capture non-linear relationships within the data.Evaluation of prediction errors via Mean Absolute Error (MAE) further corroborated these findings. The Random Forest model consistently yielded lower MAE values (~ 7,000 kg) compared to the Linear Regression model (~ 11,000 kg), demonstrating higher accuracy and precision in yield forecasting. The results are visually summarized in boxplots (Fig. 9 ), illustrating the distribution and variability of R² and MAE scores across folds for both models. Overlaying individual fold scores emphasizes the consistency of Random Forest’s superior performance. The distinct color schemes enhance interpretability and distinction between models.Overall, the cross-validation analysis confirms the advantage of ensemble machine learning approaches over traditional linear methods for shrimp yield prediction, supporting their application in aquaculture management for improved forecasting reliability. Discussion This study utilized an extensive secondary dataset, applying advanced spatiotemporal and predictive analyses to investigate shrimp yield trends across southern coastal districts of Bangladesh from 2001 to 2024. Using data provided by the Department of Fisheries (DoF), the analysis revealed significant variations in shrimp productivity across different aquatic environments, thereby illuminating both growth opportunities and sustainability challenges within the region's aquaculture industry. Significant productivity growth was observed in intensive shrimp/prawn farms and pond-based aquaculture systems, which recorded substantial compound annual growth rates (CAGR) of 12.5% and 9.8%, respectively. This notable shift toward intensive farming practices is likely driven by improved infrastructure, technological advancements, and favorable market dynamics, aligning closely with trends observed in previous studies (Ahmed & Flaherty 2013 ; Bush 2019; Bashar et al. 2022 ). These findings resonate with the recent research of Siddique et al. ( 2024 a), who demonstrated the effective use of ARIMA modeling to forecast tilapia production in Bangladesh, underscoring the increasing reliance on data-driven predictive approaches to enhance aquaculture management and planning. Their work highlights how temporal forecasting models can provide critical insights for scaling aquaculture production in line with market and environmental changes, a strategy that appears equally promising for shrimp aquaculture in coastal Bangladesh. Conversely, natural water bodies, notably the Sundarbans mangrove systems and riverine fisheries, experienced substantial yield declines due to ongoing environmental stressors such as increasing salinity intrusion, habitat degradation, and enhanced conservation regulations. These observations resonate with the findings of previous studies (Meisner& Ahmed 2023 ; Padhy et al. 2022 ; Gopal & Chauhan 2006 ), which highlighted similar environmental concerns and their implications for long-term sustainability. Siddique et al. ( 2024 b) further emphasize the critical influence of climatic factors on aquaculture water quality parameters, demonstrating that fluctuations in temperature and rainfall substantially affect pond water temperature and other key quality indicators that directly impact fish health and growth. Such climatic variability could exacerbate challenges for shrimp farming in coastal ecosystems, especially under the increasing pressure of climate change, as reflected in the spatial contraction of yield in natural habitats observed in this study. The statistical analyses employing PCA and ANOVA demonstrated significant variations in yields among different aquatic environments. The analyses effectively differentiated high-productivity intensive farming systems from lower-yield natural capture systems, confirming the hypothesis that productivity varies substantially based on distinct management practices and ecological conditions.These findings are consistent with previous research emphasizing the significance of specific ecological and management factors on aquaculture productivity (Mandal & Ghosh 2024 ; Zhang et al. 2022 ; Li et al. 2024 ; Alfiansah et al. 2018 ). Furthermore, Siddique et al. ( 2025 a) integrate ARIMA and ARIMAX modeling to reveal how combined climatic and water quality parameters precisely influence broodfish growth in tilapia culture, illustrating the importance of multi-factor models for understanding complex growth dynamics. These integrative modeling approaches could be adapted to shrimp aquaculture to better anticipate growth performance under varying environmental and management scenarios. In predictive modeling, the Random Forest Regression approach exhibited superior performance relative to traditional Multiple Linear Regression models, yielding higher accuracy and reduced predictive errors. This result highlights complex, non-linear relationships among influential predictors, a finding consistent with earlier studies applying advanced machine learning techniques in aquaculture and agricultural yield forecasting (Edeh et al. 2022 ; Nazmi et al. 2023 ; Chen et al. 2021 ; Ahmed et al. 2024 ). The demonstrated success of machine learning models complements the time series forecasting efforts, reinforcing the utility of ensemble methods and hybrid modeling frameworks for improving yield predictions and informing decision-making. Similarly, Siddique et al. ( 2025 b) employ ARIMA modeling to forecast climatic variables such as air temperature and rainfall in Mymensingh, Bangladesh, revealing critical implications for aquaculture management that directly affect production outcomes. Their findings suggest that integrating climatic forecasts with aquaculture yield models could significantly enhance adaptive management strategies in Bangladesh’s coastal aquaculture systems. Additionally, the ARIMA forecasting model provided valuable medium-term projections of shrimp yields, despite certain limitations, notably its overestimation of yields for the year 2022. This discrepancy underscores the need for real-time data integration and adaptive modeling frameworks to enhance predictive accuracy in response to dynamic environmental and market conditions (Shohan & Haque 2025 ). As highlighted by Siddique et al. ( 2024 a, 2025a), the inclusion of exogenous variables such as climatic factors in ARIMAX models can improve forecasting performance by accounting for environmental drivers that influence aquaculture productivity. Future adoption of such integrative models in shrimp aquaculture forecasting could reduce prediction errors and better capture the effects of anomalous events such as extreme weather or disease outbreaks. One of the key insights of this study is the application of ARIMA and Random ForestRegression models for forecasting shrimp yields. While ARIMA provided valuable medium-term projections, it notably overestimated the yields for 2022, raising concerns about its suitability for accurately predicting anomalous events. This overestimation may be attributed to the model's inability to incorporate exogenous variables, such as climatic factors, that influence shrimp production. Similar issues with ARIMA have been noted in other studies, where its predictive accuracy fluctuated in the presence of unexpected changes in environmental or market conditions (Siddique et al. 2025 ). For instance, the model failed to account for abrupt climatic events or shifts in regulatory policies, which could have skewed yield estimates. To address such limitations, future studies could explore the use of ARIMAX models or hybrid approaches that incorporate external variables such as temperature, rainfall, orsalinity levels to improve the accuracy of predictions. In contrast, the Random Forest model demonstrated superior performance, achieving an R² of0.91 and a MAE of 6,950 kg. This model accounted for complex, non-linear interactions between variables, making it more capable of handling the intricacies of the aquaculture environment. However, it is important to note that machine learning models like Random Forest are prone to overfitting if not properly tuned. In this study, cross-validation techniques were applied to minimize such risks, but further research could explore additional techniques like hyper parameter optimization and ensemble learning to ensure the robustness of the model in diverse environmental conditions. The study also underscores the significant impact of environmental stressors, particularly salinity intrusion and habitat degradation, on shrimp yields in natural water bodies such as the Sundarbans and riverine systems. These findings align with previous studies indicating that increasing salinity in coastal areas, coupled with mangrove deforestation and other anthropogenic activities, has led to a decline in shrimp production from natural capture fisheries (Meisner& Ahmed 2023 ). However, while these factors were identified as key contributors to yield reductions, the study could benefit from more detailed temporal data on salinity levels and pollution to provide a clearer understanding of how these environmental variables correlate with yield fluctuations over time. While this research contributes significantly to the understanding of shrimp production dynamics, it is subject to several limitations. Primarily, the absence of detailed spatial-temporal data on farm sizes and localized environmental management practices restricts the precision of yield estimations and comprehensive sustainability assessments. Furthermore, reliance on secondary data inherently omits detailed socio-economic insights that primary data collection could provide. Complementing this, Siddique et al. ( 2024 b) highlight the importance of continuous water quality monitoring to capture fine-scale environmental variations critical for aquaculture success, underscoring a gap that future studies should address by integrating field-based data collection and remote sensing technologies. Future research endeavors should integrate primary field surveys and remote sensing data, facilitating improved spatial resolution and more robust sustainability analyses. Policymakers and stakeholders should leverage the insights provided by this study to implement sustainable intensification strategies and targeted ecological interventions. Such integrated management approaches will be pivotal in ensuring the sustainability, resilience, and continued growth of Bangladesh’s shrimp aquaculture industry amid ongoing environmental and economic challenges. The advanced forecasting and predictive modeling techniques exemplified in the recent studies by Siddique and colleagues provide a valuable framework for enhancing aquaculture resilience by explicitly incorporating environmental variability and climatic influences into production planning and policy formulation. Conclusion This study demonstrates the value of secondary data in advancing knowledge on shrimp yield dynamics across diverse aquatic environments in Bangladesh's southwestern coastal districts. By leveraging two decades of production data from the Department of Fisheries, comprehensive spatiotemporal analyses revealed significant disparities in shrimp productivity driven largely by water body type and management practices. Intensive shrimp/prawn farms and pond systems consistently outperformed natural water sources, exhibiting higher compound annual growth rates due to infrastructural investments, technological innovations, and targeted management approaches. Statistical assessments confirmed significant yield differences among aquatic environments, emphasizing that anthropogenic intervention can drastically influence production outcomes. Predictive modeling results underscored the advantage of advanced machine learning approaches, particularly Random Forest Regression, in capturing non-linear relationships and providing robust yield estimations. Time-series forecasting via ARIMA contributed to future yield projections, highlighting potential production trends through 2035 despite limitations in accurately predicting anomalous events. Overall, this research highlights a clear shift from traditional, environmentally dependent shrimp capture methods toward engineered aquaculture systems, raising both opportunities for sustainable intensification and concerns over ecological trade-offs. The findings provide actionable insights for policymakers, investors, and local stakeholders seeking to enhance shrimp aquaculture productivity while ensuring long-term environmental resilience. Future research integrating high-resolution primary data, remote sensing technologies, and real-time monitoring systems is essential to refine predictive models and guide adaptive, sustainable aquaculture management strategies in Bangladesh. Declarations Funding The authors declare that they received no financial support for the research, authorship, or publication of this article. No funds, grants, or other support, whether internal or external, were provided by any institution, department, or funding agency, including public, commercial, or nonprofit organizations. All costs related to the study design, data collection, analysis, and manuscript preparations were covered by the authors. Ethical statement This research utilizes secondary data from publicly available databases and literature, strictly adhering to ethical standards in data collection, analysis, and reporting. The use of existing data complies with the terms of use from the original sources, with all necessary citations provided. No unauthorized access or data manipulation has occurred. The study upholds the confidentiality and privacy of individuals, aligning with ethical guidelines to promote responsible research conduct. Data availability statement The data presented in this study are available upon request from the corresponding author. Author contribution I. A.: Overall data analysis & presentation and writing the original draft; M. A. B. S.: Data analysis, Editing the draft; M. M. H.: Validation, Review & editing the draft; A. K. S. A.: Concept development, validation, overall technical analysis and supervision and editing the draft. References Ahmed F, Bijoy MHI, Hemal HR, Noori SRH (2024) Smart aquaculture analytics: Enhancing shrimp farming in Bangladesh through real-time IoT monitoring and predictive machine learning analysis. Heliyon 10(17). Ahmed N, Flaherty MS (2013) Opportunities and challenges for the development of prawn farming with fish and rice in southeast Bangladesh: Potential for food security and economic growth. Food Security 5(5):637–649. Alfiansah YR, Hassenrück C, Kunzmann A, Taslihan A, Harder J, Gärdes A (2018) Bacterial abundance and community composition in pond water from shrimp aquaculture systems with different stocking densities. Frontiers in Microbiology 9:2457. Bashar A, Heal RD, Hasan NA, Salam MA, Haque MM (2022) COVID-19 impacts on the Bangladesh shrimp industry: A sequential survey-based case study from southwestern Bangladesh. Fisheries Science 88(6):767–786. Bush SR, Belton B, Little DC, Islam MS (2019) Emerging trends in aquaculture value chain research. Aquaculture 498:428–434. Chen S, Tao F, Pan C, Hu X, Ma H, Li C et al (2021) Modeling quality changes in Pacific white shrimp (Litopenaeus vannamei) during storage: Comparison of the Arrhenius model and random forest model. Journal of Food Processing and Preservation 45(1):e14999. Department of Fisheries (DoF) (2024) Yearbook of Fisheries Statistics of Bangladesh 2024. Ministry of Fisheries and Livestock, Government of the People’s Republic of Bangladesh, Dhaka. Edeh MO, Dalal S, Obagbuwa IC, Prasad BS, Ninoria SZ, Wajid MA, Adesina AO (2022) Bootstrapping random forest and CHAID for prediction of white spot disease among shrimp farmers. Scientific Reports 12(1):20876. Gopal B, Chauhan M (2006) Biodiversity and its conservation in the Sundarban mangrove ecosystem. Aquatic Sciences 68(3):338–354. Gulakhmadov M, Chen X, Gulakhmadov A, Nadeem MU, Gulahmadov N, Liu T (2023) Performance analysis of precipitation datasets at multiple spatio-temporal scales over a dense gauge network in a mountainous domain of Tajikistan, Central Asia. Remote Sensing 15(5):1420. He A, Huang J, Sun Z, Zhou J, Yang C (2023) Spatial and temporal evolution characteristics of the Salween River delta from 1973 to 2021. Remote Sensing 15(5):1467. Kaur G, Adhikari N, Krishnapriya S, Wawale SG, Malik RQ, Zamani AS et al (2023) Recent advancements in deep learning frameworks for precision fish farming: Opportunities, challenges, and applications. Journal of Food Quality 2023:4399512. Li Y, Cao S, Jiang S, Huang J, Yang Q, Jiang S et al (2024) Comparative study of nutritional composition, physiological indicators, and genetic diversity in Litopenaeus vannamei from different aquaculture populations. Biology 13(9):722. Mandal A, Ghosh AR (2024) Role of artificial intelligence (AI) in fish growth and health status monitoring: A review on sustainable aquaculture. Aquaculture International 32(3):2791–2820. Meisner CA, Ahmed S (2023) Shrimp farming and the question of land and social environment. In: Transforming Bangladesh: Geography, People, Economy and Environment. Springer International Publishing, Cham, pp 55–64. Nazmi H, Siau NZ, Bramantoro A, Suhaili WS (2023) Predictive modeling of marine fish production in Brunei Darussalam’s aquaculture sector: A comparative analysis of machine learning and statistical techniques. International Journal of Advanced and Applied Sciences 10(7):109–126. Olipp N, Woschank M, Hoffelner M (2024) Exploration of the framework conditions for measures to reduce resource consumption in the manufacturing industry with a focus on the circular economy: A systematic secondary data research. Production & Manufacturing Research 12(1):2431723. Padhy SR, Dash PK, Bhattacharyya P (2022) Challenges, opportunities, and climate change adaptation strategies of the mangrove–agriculture ecosystem in the Sundarbans, India: A review. Wetlands Ecology and Management 30(1):191–206. Rahman LF, Marufuzzaman M, Alam L, Bari MA, Sumaila UR, Sidek LM (2021) Developing an ensembled machine learning prediction model for marine fish and aquaculture production. Sustainability 13(16):9124. Rivera-Ferre MG (2009) Can export-oriented aquaculture in developing countries be sustainable and promote sustainable development? The shrimp case. Journal of Agricultural and Environmental Ethics 22:301–321. Shohan MH, Haque MM (2025) Forecasting the production and export of shrimp in Bangladesh: A policy-focused time series analysis. Frontiers in Aquaculture 4:1541025. Siddique MAB, Mahalder B, Haque MM, Ahamad AKS (2025) Forecasting air temperature and rainfall in Mymensingh, Bangladesh with ARIMA: Implications for aquaculture management. Egyptian Journal of Aquatic Research (in press). Siddique MAB, Mahalder B, Haque MM, Ahammad AKS (2025) Impact of climatic and water quality parameters on tilapia (Oreochromis niloticus) broodfish growth: Integrating ARIMA and ARIMAX for precise modeling and forecasting. PLOS ONE 20(3):e0313846. Siddique MAB, Mahalder B, Haque MM, Ahammed AKS (2024) Impact of climatic factors on water quality parameters in tilapia broodfish ponds and predictive modeling of pond water temperature with ARIMAX. Heliyon 10(18):e37717. Siddique MAB, Mahalder B, Haque MM, Shohan MH, Biswas JC, Akhtar S, Ahammad AKS (2024) Forecasting of tilapia (Oreochromis niloticus) production in Bangladesh using ARIMA model. Heliyon 10(5):e27111. Zhang P, Huang Q, Peng R, Jiang X, Jiang M, Zeng G, Lin J (2022) Environmental factors of rearing water and growth performance of shrimp (Penaeus vannamei) in a microalgal monoculture system. Aquaculture 561:738620. 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-7456160","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":530182814,"identity":"54070252-0fc5-480f-92e9-5c0f54a5aebe","order_by":0,"name":"Ilias Ahmed","email":"","orcid":"","institution":"Bangladesh Agricultural University","correspondingAuthor":false,"prefix":"","firstName":"Ilias","middleName":"","lastName":"Ahmed","suffix":""},{"id":530182815,"identity":"2d38bc48-488c-4de4-b615-1d6793f00346","order_by":1,"name":"Mohammad Abu Baker Siddique","email":"","orcid":"","institution":"Bangladesh Agricultural University","correspondingAuthor":false,"prefix":"","firstName":"Mohammad","middleName":"Abu Baker","lastName":"Siddique","suffix":""},{"id":530182816,"identity":"784681a0-5ed6-4454-a906-03c2886c5a3b","order_by":2,"name":"Mohammad Mahfujul Haque","email":"","orcid":"","institution":"Bangladesh Agricultural University","correspondingAuthor":false,"prefix":"","firstName":"Mohammad","middleName":"Mahfujul","lastName":"Haque","suffix":""},{"id":530182817,"identity":"b6b46ebf-9af2-4218-b9f0-80881a97cd8b","order_by":3,"name":"A. K. Shakur Ahammad","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA50lEQVRIiWNgGAWjYFACHjDJ2MbefAwicICBDSpIQEs/z7E0ErXMnJFjRpwW+fazBz/dzLGT3XAj59uDn20Mcnw3EtgevMGjxeBMXrJ07rZk4w1n3m437G1jMJa8kcBuOAefFoYcA6AW5sQNx3O3SfC2MSRuANoijddh/W+Mf+duq0/ccCDnmeTfNoZ6gloYbuSYAW05nDizI4dNGmhLggEhLQY33phZ5247bgwMZHNjmXMShjPPPGzH6xf5/hzj27nbqmWBUfns4ZsyG3m+48nH8IYYOpAAYsYGEjSMglEwCkbBKMAGAPoKU5ylSi6RAAAAAElFTkSuQmCC","orcid":"","institution":"Bangladesh Agricultural University","correspondingAuthor":true,"prefix":"","firstName":"A.","middleName":"K. Shakur","lastName":"Ahammad","suffix":""}],"badges":[],"createdAt":"2025-08-25 17:53:16","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7456160/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7456160/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":93671274,"identity":"98659a83-9144-450d-9bda-8b5c79dd0c13","added_by":"auto","created_at":"2025-10-16 10:04:49","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4964268,"visible":true,"origin":"","legend":"","description":"","filename":"MainManuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/ee63000d7ebf5d4624db5e6f.docx"},{"id":93671270,"identity":"4cf2b677-1f68-44d6-8f9e-e00d9481d8af","added_by":"auto","created_at":"2025-10-16 10:04:49","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5511,"visible":true,"origin":"","legend":"","description":"","filename":"18a0e4e6858f4576aea76ac3fd6925e0.json","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/038d3cd804fb36864c32720b.json"},{"id":93670941,"identity":"bbc80dd1-4157-483c-add4-e6d078e976bd","added_by":"auto","created_at":"2025-10-16 09:56:49","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":123380,"visible":true,"origin":"","legend":"","description":"","filename":"18a0e4e6858f4576aea76ac3fd6925e01enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/8baf125192a64d3150a0e3c0.xml"},{"id":93671271,"identity":"346bdefd-d3e0-47c9-86c6-492a5897f6ee","added_by":"auto","created_at":"2025-10-16 10:04:49","extension":"jpeg","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":331221,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/3545b779576e59f63b1b10b2.jpeg"},{"id":93670949,"identity":"866a0f21-eeac-4ad6-8a80-4d1064778c80","added_by":"auto","created_at":"2025-10-16 09:56:49","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":199208,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/9df14b57b78c7806b08bf475.jpeg"},{"id":93671273,"identity":"79dd42f8-de82-4a51-81ff-fcd74cbc9d44","added_by":"auto","created_at":"2025-10-16 10:04:49","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":161823,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/f2374a688e679beeced4e072.jpeg"},{"id":93670951,"identity":"955c3a93-b87f-4175-b8fc-69a5db34091d","added_by":"auto","created_at":"2025-10-16 09:56:49","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":111882,"visible":true,"origin":"","legend":"","description":"","filename":"groupimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/ae5612a624e47e96f11f1888.jpeg"},{"id":93670945,"identity":"bcf15630-5787-4ed5-be4c-ea14eec86c61","added_by":"auto","created_at":"2025-10-16 09:56:49","extension":"jpeg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":77466,"visible":true,"origin":"","legend":"","description":"","filename":"groupimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/91c8095754620e160a48bc3a.jpeg"},{"id":93670969,"identity":"6d6d3ae1-4cb1-4b1a-81ae-5830085cc1bc","added_by":"auto","created_at":"2025-10-16 09:56:51","extension":"jpeg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":125545,"visible":true,"origin":"","legend":"","description":"","filename":"groupimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/42778e58d40d2dac08749ed7.jpeg"},{"id":93670967,"identity":"6552a490-8740-40e2-bfdf-94466fee87b3","added_by":"auto","created_at":"2025-10-16 09:56:50","extension":"jpeg","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":52832,"visible":true,"origin":"","legend":"","description":"","filename":"groupimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/5704affba39110d27ccfb230.jpeg"},{"id":93670972,"identity":"208095e8-357e-4383-ba1a-ef51384ff352","added_by":"auto","created_at":"2025-10-16 09:56:52","extension":"jpeg","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":33027,"visible":true,"origin":"","legend":"","description":"","filename":"groupimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/53ecedc100a653a2decba2af.jpeg"},{"id":93670968,"identity":"7da7c240-e3ec-4ad5-9fb4-00497be24802","added_by":"auto","created_at":"2025-10-16 09:56:50","extension":"jpeg","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":27483,"visible":true,"origin":"","legend":"","description":"","filename":"groupimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/7f017516bb29c768c05eef27.jpeg"},{"id":93670973,"identity":"87fb6f05-5666-438a-954b-fccd7815245f","added_by":"auto","created_at":"2025-10-16 09:56:52","extension":"jpeg","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":45257,"visible":true,"origin":"","legend":"","description":"","filename":"groupimage7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/1b833e066cf064555d0f18cf.jpeg"},{"id":93670957,"identity":"1f6667b1-a306-4770-8299-edf6f94d6339","added_by":"auto","created_at":"2025-10-16 09:56:49","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":75165,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/1d07b2fc152d5af701da65a8.png"},{"id":93670976,"identity":"5e9cf2f3-4c97-4824-8c48-fdbc8ad26ae4","added_by":"auto","created_at":"2025-10-16 09:56:53","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":34488,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/5403e132c0ae93dc0fa74412.png"},{"id":93670963,"identity":"aa37bc4d-ed13-49ea-a51b-4e47c6843b23","added_by":"auto","created_at":"2025-10-16 09:56:50","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":29964,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/b178ecab3e7baefae0ae600b.png"},{"id":93670966,"identity":"508deea5-83d1-4ef6-94b7-62877282ab42","added_by":"auto","created_at":"2025-10-16 09:56:50","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":36836,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinegroupimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/37fb5c86ca2904f060699706.png"},{"id":93670975,"identity":"56df892b-dd5e-41f0-abcf-4ea40699bd16","added_by":"auto","created_at":"2025-10-16 09:56:53","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":20338,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinegroupimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/d07e87582f3b5cd119b0e6c8.png"},{"id":93670961,"identity":"e3219442-1995-4e4f-b9d4-e567776b40ae","added_by":"auto","created_at":"2025-10-16 09:56:50","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":38159,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinegroupimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/66a9104740668f23c3ef6b5a.png"},{"id":93670971,"identity":"3aa4c1a9-fe2e-45b4-9c09-76726f8f45da","added_by":"auto","created_at":"2025-10-16 09:56:51","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12464,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinegroupimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/57c8892278cd50ba30ef3518.png"},{"id":93671276,"identity":"1e41932b-af6f-4d0a-a95e-bfc5165222eb","added_by":"auto","created_at":"2025-10-16 10:04:50","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8266,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinegroupimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/ddd4dee9d3444ad251379377.png"},{"id":93671280,"identity":"0cb344d5-e627-43e9-aa28-8e37b54d9633","added_by":"auto","created_at":"2025-10-16 10:04:53","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6125,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinegroupimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/f82163c8fc09fb001707a4fc.png"},{"id":93670977,"identity":"683479a2-841d-43c5-8eb1-39843cf21ac9","added_by":"auto","created_at":"2025-10-16 09:56:53","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10935,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinegroupimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/22a4f01c9be76438d0b537e1.png"},{"id":93671278,"identity":"90b68d9b-613a-43a7-a139-373d3ba00528","added_by":"auto","created_at":"2025-10-16 10:04:51","extension":"xml","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":124162,"visible":true,"origin":"","legend":"","description":"","filename":"18a0e4e6858f4576aea76ac3fd6925e01structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/94599054deeba51c7f233059.xml"},{"id":93670985,"identity":"df456ca8-4921-4a6f-a653-cfe12f52e625","added_by":"auto","created_at":"2025-10-16 09:56:54","extension":"html","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":128797,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/18a06179cf210b8c3ac4dd8a.html"},{"id":93671272,"identity":"93866320-5f02-41e2-9c0b-b71fd41e0ba4","added_by":"auto","created_at":"2025-10-16 10:04:49","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1394372,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSpatial and temporal distribution of shrimp yield (kg) in Khulna Division, Bangladesh, for early (2001–2006), mid (2010–2014), and recent (2019–2024) periods. \u003c/strong\u003eMaps show district boundaries and yield hotspots identified using Getis-Ord Gi* analysis (p \u0026lt; 0.01). Violin plots illustrate yield distributions by district. Bubble and heatmap visualizations highlight spatial clustering, while temporal block maps summarize average district yields, demonstrating a shift toward intensive shrimp farming in southwestern Satkhira and central Khulna over time. Data sourced from the Department of Fisheries (2001–2024).\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/03cb1489916fccbf94de873b.png"},{"id":93670943,"identity":"9d2ace59-0414-4cda-a241-b05e1428a955","added_by":"auto","created_at":"2025-10-16 09:56:49","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":589218,"visible":true,"origin":"","legend":"\u003cp\u003eAbsolute annual shrimp yields (kg) from 2001-2024 by major water body type in Khulna Division, Bangladesh. Trends demonstrate exponential growth in shrimp/prawn farms, linear increases in pond systems, and variable or declining outputs from other water bodies over the study period.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/8244a22227ee1e004dd3a353.png"},{"id":93670940,"identity":"1924c609-302b-47e6-b7c0-c28d1a0dd564","added_by":"auto","created_at":"2025-10-16 09:56:49","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1113401,"visible":true,"origin":"","legend":"\u003cp\u003ePercentage contribution of each water body type (River, Sundarbans, Beel, Floodplain, Pond, Seasonal Culture, Baor, and Shrimp/Prawn Farm) to total shrimp production in Khulna Division, fiscal years 2001-02 to 2023-24.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/b6ba5d53ea5c316874ce4fab.png"},{"id":93670946,"identity":"f843c81c-7ede-40a4-8a49-c821e8e81acc","added_by":"auto","created_at":"2025-10-16 09:56:49","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":407368,"visible":true,"origin":"","legend":"\u003cp\u003ePrincipal Component Analysis biplot depicting clustering of six water body types along PC1 (57.7% variance) and PC2 (11.9% variance), based on standardized log-transformed yield data; inset boxplots illustrate yield distributions by category, with letters indicating Tukey HSD groupings (ANOVA F(5,132) = 18.45, p \u0026lt; 0.001).\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/1f5fd5ec1f655a05f8d1c388.png"},{"id":93670944,"identity":"ca070d81-6be2-48b6-9d43-8b6b814d5b13","added_by":"auto","created_at":"2025-10-16 09:56:49","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":425705,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 4. \u003c/strong\u003eCorrelation network of shrimp yields by water body type; edges show correlations with |r| \u0026gt; 0.1. Seasonal Cultured highlighted.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/95c89f88e2f5aa18cd2a0e10.png"},{"id":93671275,"identity":"e0de61d6-11be-48c8-809f-957baf428631","added_by":"auto","created_at":"2025-10-16 10:04:49","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":111693,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 5.\u003c/strong\u003e Actual (solid line) and ARIMA (1,1,1)-predicted (dashed line) shrimp yields in coastal Bangladesh from 2001 to 2035; 95% confidence intervals for forecasted values (2024–2035) are shaded.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/345a0e73c898c6d523887244.png"},{"id":93670958,"identity":"657cd237-00dd-4635-aadd-e179ecede875","added_by":"auto","created_at":"2025-10-16 09:56:49","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":117437,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 6. Model diagnostics for ARIMA (1,1,1) shrimp yield forecasts. (A) Standardized residuals plotted over time (2001–2023) show random scatter around zero, indicating no temporal bias or heteroscedasticity. (B) Autocorrelation function (ACF) of residuals up to lag 20, with all coefficients within the 95% confidence bounds, confirms the residuals approximate white noise and supports model adequacy.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/d09f21636bb7cd41beb6ea26.png"},{"id":93671277,"identity":"46d73a16-e86c-471a-bd47-31edffa0024d","added_by":"auto","created_at":"2025-10-16 10:04:50","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":170728,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 7\u003c/strong\u003e. Model error and selection diagnostics for ARIMA (1,1,1) shrimp yield forecasts. (7A) Histogram of forecast errors (Actual-Predicted) from 2001-2023 with kernel density overlay, showing an approximately normal distribution and slight negative skew. (7B) AIC values for candidate ARIMA (p,1,q) models, highlighting ARIMA (1,1,1) as the optimal model with the lowest AIC.\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/ecac6221810bdc683aca3b98.png"},{"id":93670948,"identity":"6e404224-6031-4b2d-94c1-d856ac5bee08","added_by":"auto","created_at":"2025-10-16 09:56:49","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":361608,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 8.\u003c/strong\u003e Predictive performance of machine learning regression models for shrimp yield. (A) Observed versus predicted total shrimp yields for Multiple Linear Regression and Random Forest Regression models; the Random Forest shows closer fit to the 1:1 line. (B) Random Forest feature importance scores, indicating shrimp/prawn farm, pond, and floodplain yields as the strongest predictors, with district and year also contributing substantially to model performance.\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/e97f54e73bdaa8c0fc50d358.png"},{"id":93670959,"identity":"b0c3ca71-0576-4457-aaa1-443fea5c4865","added_by":"auto","created_at":"2025-10-16 09:56:50","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":116893,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 9.\u003c/strong\u003e Cross-validated model performance comparison for shrimp yield prediction. Boxplots depict the distribution of (a) coefficient of determination (R²) and (b) mean absolute error (MAE, kg) scores obtained from 10-fold cross-validation for Random Forest and Multiple Linear Regression models. Individual data points represent scores from each fold. The Random Forest model consistently outperformed the Linear Regression model in both accuracy and variance explained.\u003c/p\u003e","description":"","filename":"10.png","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/756adb4e4a74dde9d669496a.png"},{"id":94474185,"identity":"b7ba07b5-5c43-4812-8a9f-17808a314b4e","added_by":"auto","created_at":"2025-10-27 15:47:53","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6560682,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7456160/v1/5a178c93-4893-4987-9794-d7e29fa6fb38.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Machine Learning and ARIMA Models for Spatiotemporal Analysis and Forecasting of Shrimp Yields across the Aquatic Systems of Southern Coastal Bangladesh","fulltext":[{"header":"Introduction","content":"\u003cp\u003eShrimp farming represents a crucial component of Bangladesh's aquaculture industry, significantly influencing the national economy through employment generation, foreign exchange earnings, and ensuring food security (Ahmed and Flaherty \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Rivera-Ferre \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Bush et al. \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Particularly in the southwestern coastal districts, such as Khulna, Satkhira, Bagerhat, and Jashore, shrimp aquaculture thrives due to unique ecological characteristics, including brackish water environments, varying salinity levels, extensive tidal influences, and abundant natural water bodies (DoF 2024). Despite its importance, shrimp production in this region faces several pressing challenges, such as environmental degradation, mangrove deforestation, unsustainable farming practices, and increasing climate variability characterized by fluctuating rainfall patterns and rising temperatures, which collectively threaten long-term sustainability (Meisner and Ahmed \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eIn addressing these challenges, spatiotemporal analytical approaches offer unique advantages, enabling the simultaneous examination of geographic yield disparities and their evolution over time. Such approaches are critical in aquaculture because spatial variability in salinity regimes, aquaculture infrastructure, and environmental stressors interacts dynamically with temporal changes in climate, market conditions, and policy interventions, influencing yields in complex ways. By integrating spatial and temporal perspectives, it becomes possible to identify emerging production hotspots, detect declining zones, and map the shifting geography of shrimp farming insights that purely spatial or purely temporal analyses cannot fully capture. Similar integrative methods have been successfully applied in fisheries and agricultural systems worldwide, demonstrating how geostatistical mapping, spatial clustering algorithms, and time-series modeling can inform targeted management strategies (He et al. \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Gulakhmadov et al. \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eSecondary data utilization has become an integral approach in agricultural and environmental studies due to its effectiveness in providing substantial analytical insights without the resource-intensive processes involved in primary data collection (Olipp et al. \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The Yearbook of Fisheries Statistics of Bangladesh, compiled by the Department of Fisheries (DoF), offers an extensive dataset spanning more than two decades (2001\u0026ndash;2024). This dataset is uniquely suited for comprehensive spatiotemporal analyses as it provides shrimp production figures categorized by district and specific aquatic environments, enabling robust temporal trend identification and spatial yield mapping (DoF 2024).\u003c/p\u003e\u003cp\u003eAdvanced data analysis techniques, including machine learning and sophisticated statistical models, have increasingly been applied to agricultural yield forecasting (Kaur et al. \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Models such as Multiple Linear Regression, Random Forest Regression, and ARIMA have demonstrated significant predictive capabilities in accurately forecasting agricultural outputs, factoring in historical yield trends, climatic variability, and spatial heterogeneity (Rahman et al. \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Mandal \u0026amp; Ghosh \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Such predictive models are instrumental in guiding evidence-based policy decisions, optimizing resource allocation, and enhancing sustainable management practices within shrimp farming sectors (Shohan and Haque \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). DoF (2024) demonstrated through forecasting models including Machine Learning Technique and ARIMA that while Bangladesh\u0026rsquo;s shrimp production is expected to steadily increase from 262,937 metric tons in 2022 to over 300,000 metric tons by 2030, shrimp exports are projected to decline sharply by approximately 30% over the same period, signaling a growing disparity between production expansion and export performance. In the study by Fizar et al. (2024), freshwater shrimp production levels were forecasted using advanced machine learning classification algorithms based on real-time water quality data. Among them, the Random Forest algorithm was identified as the most effective, with an overall classification accuracy of 97.84% and an F1-score of 95.79%. These forecasts were generated by training the models on features derived from five critical water parameters, and shrimp production categories were predicted with high reliability. The performance of the machine learning models demonstrated that accurate production forecasting can be achieved using environmental data, offering significant potential for intelligent decision-making in shrimp aquaculture (Fizar et al. 2024). On the other hand, tilapia production was forecasted by Siddique et al. (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) using a time series machine learning approach (ARIMA). Historical production data were used to train the model, which successfully predicted a continuous upward trend in tilapia production from 2006 through 2040. Strong accuracy and reliability were demonstrated by the forecasting, validating ARIMA as an effective tool for aquaculture production prediction. The potential of ARIMA-based models to support data-driven decision-making and strategic planning in sustainable fisheries management was thereby emphasized (Siddique et al. \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eTherefore, the present study employs these advanced analytical methodologies to scrutinize shrimp yield across diverse aquatic environments within southern coastal regions of Bangladesh. By leveraging secondary data spanning from 2001 to 2024, this research identifies critical production trends, spatial yield distributions, and environmental determinants influencing shrimp yield. The insights derived from this study are intended to assist policymakers, local stakeholders, and researchers in formulating strategies for sustainable growth and resilient shrimp aquaculture practices in Bangladesh.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1. Data Source and Nature of Data\u003c/h2\u003e\u003cp\u003eThis study is based on secondary data obtained from the Yearbook of Fisheries Statistics of Bangladesh (2001\u0026ndash;2024), an annual publication of the Department of Fisheries (DoF) under the Ministry of Fisheries and Livestock, Government of the People\u0026rsquo;s Republic of Bangladesh. The dataset is publicly accessible through the official DoF portal at: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://fisheries.gov.bd/site/page/54ea4502-a4cb-4e33-9f29-4be8f09cf8a6\u003c/span\u003e\u003cspan address=\"https://fisheries.gov.bd/site/page/54ea4502-a4cb-4e33-9f29-4be8f09cf8a6\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. The dataset used for this study includes annual shrimp production statistics for the fiscal years 2001\u0026ndash;2002 through 2023\u0026ndash;2024, disaggregated by district and water body type. The spatial focus of the study encompasses four coastal districts in southwestern Bangladesh including Khulna, Satkhira, Bagerhat, and Jashore which are known to be among the country\u0026rsquo;s most productive regions for shrimp aquaculture. The production values are expressed in kilograms and are classified across the following water body types: river, pond, floodplain, beel, baor (oxbow lake), shrimp/prawn farm, and Sundarbans (mangrove-based capture zone).Although the dataset primarily represents annual statistics, it serves as a representative snapshot of the regional aquaculture landscape, allowing for both spatial and temporal inference. Due to the dataset's official origin and nationwide scope, it qualifies as a reliable secondary data source for agricultural and environmental informatics applications.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2. Study Area\u003c/h2\u003e\u003cp\u003eThe study area comprises the coastal and estuarine districts of Khulna Division, situated in the southwestern part of Bangladesh. The selected districts including Khulna, Satkhira, Bagerhat, and Jashore which are ecologically diverse and economically significant due to their widespread engagement in brackish water aquaculture, particularly shrimp farming. These regions exhibit varying levels of salinity, access to natural water bodies, and aquaculture infrastructure, offering a suitable base for geospatial comparison and environmental assessment. The study leverages administrative boundary data to spatially reference and visualize aquaculture production patterns.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3. Classification of Aquatic Environments\u003c/h2\u003e\u003cp\u003eIn order to examine spatial heterogeneity in shrimp production, aquatic habitats encompassing both aquaculture and capture fisheries systems were classified into discrete environment types according to their defining physical, hydrological, and management attributes. This typology provides a structured basis for water body-specific yield assessment and facilitates comparative analyses across a spectrum of production systems, ranging from high-intensity pond culture to low-intensity, open-access mangrove ecosystems. The classification scheme employed in the present study was adapted from the Department of Fisheries (DoF 2024) and is presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eClassification and description of aquatic environments used in shrimp production analysis, as applied to water body-wise yield assessments in the present study\u003c/p\u003e \u003cdiv class=\"Credit\"\u003e\u003cp\u003e(adapted from DoF, 2024)\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"2\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWater Body Type\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDescription\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRiver\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNatural tidal channels and estuarine rivers\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePond\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eArtificial enclosed water bodies used for intensive farming\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFloodplain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSeasonally inundated lowlands, often used in semi-intensive farming\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBeel\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMarshy depressions used for extensive culture\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBaor (Oxbow Lake)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eU-shaped water bodies formed from cut-off river meanders\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eShrimp/Prawn Farm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDesignated aquaculture areas, typically converted from rice paddies\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSundarbans\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMangrove-based open water system, contributing to capture fisheries\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.4. Data Preparation and Processing\u003c/h2\u003e\u003cp\u003eThe data preparation and processing phase followed a structured protocol designed to ensure quality, consistency, and analytical readiness for spatiotemporal modeling. Raw shrimp production data extracted from the \u003cem\u003eYearbook of Fisheries Statistics of Bangladesh\u003c/em\u003e (2001\u0026ndash;2024) were first screened to identify and remove inconsistencies, including duplicate records, misaligned headers, and irrelevant rows or columns. Measurement units were cross-verified with the original source, and missing values were explicitly coded as \u003cem\u003eNA\u003c/em\u003e to allow for transparent handling during statistical analysis. Categorical descriptors, such as \u0026ldquo;District\u0026rdquo; and \u0026ldquo;Water Body Type,\u0026rdquo; were standardized to maintain uniform nomenclature across the dataset. The cleaned data were then converted into comma-separated values (.\u003cem\u003ecsv\u003c/em\u003e) format and restructured into a tidy data layout, with each variable represented in a separate column and each observation occupying a single row. Quantitative variables were normalized to facilitate comparison across districts and years, while water body classifications were encoded as categorical factors to support statistical analyses, including ANOVA, PCA, and regression modeling.\u003c/p\u003e\u003cp\u003eGeospatial integration was performed by linking production data with administrative boundary shapefiles of Bangladesh, obtained from DIVA-GIS, and projecting all spatial layers to the WGS 1984 coordinate system. District-level centroids were assigned to each observation, enabling attribute-based spatial joins between yield values and their corresponding geographic units; in certain cases, sub-district boundaries were incorporated to enhance mapping precision. The spatiotemporal analysis combined three complementary stages. First, the spatial component employed thematic choropleth mapping with natural breaks (Jenks) classification, hotspot detection using the Getis-Ord Gi* statistic, and trend surface interpolation to reveal yield intensity clusters and directional shifts in aquaculture activity. Second, the temporal component examined annual yield records for long-term patterns using compound annual growth rate (CAGR) calculations and structural change detection via the Pelt algorithm. Finally, spatial and temporal outputs were synthesized to track the progressive reconfiguration of production zones, documenting the transition from natural water capture systems toward engineered, high-intensity farming zones. This integrated approach ensured that both geographic heterogeneity and temporal evolution of shrimp production were fully captured, providing a robust basis for subsequent statistical impact assessments, predictive modeling, and sustainability evaluations.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.5 Analytical Framework\u003c/h2\u003e\u003cdiv id=\"Sec8\" class=\"Section3\"\u003e\u003ch2\u003e2.5.1 Spatiotemporal Yield Mapping\u003c/h2\u003e\u003cp\u003eThe spatial distribution of shrimp production was visualized using thematic GIS mapping. Each district was color-coded according to its total yield and dominant aquaculture water source. These maps were produced to identify production clusters and spatial disparities, providing a visual framework for policy and resource prioritization. Due to the dataset being cross-sectional (single-year), temporal variation was not analyzed.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section3\"\u003e\u003ch2\u003e2.5.2 Sustainability Assessment\u003c/h2\u003e\u003cp\u003eRelative sustainability was inferred by evaluating production efficiency across water body types. While specific area data (e.g., hectares of farms) was not available, yield magnitudes served as proxy indicators. Qualitative environmental characteristics of each water body were cross-referenced with literature on ecological stressors, such as salinization, mangrove deforestation, and seasonal water scarcity. The aim was to highlight water bodies with high output but potentially unsustainable long-term usage.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section3\"\u003e\u003ch2\u003e2.5.3 Statistical Impact Analysis\u003c/h2\u003e\u003cp\u003eTo examine whether water body types significantly influenced shrimp yield, a one-way Analysis of Variance (ANOVA) was conducted in R. The null hypothesis assumed no statistical difference in mean yields among water bodies. Additionally, PCA was performed to reduce data dimensionality and to identify clusters or outliers based on production patterns. These methods were used to test the validity of using water source types as predictive features in machine learning models.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section3\"\u003e\u003ch2\u003e2.5.4 Predictive Modeling with Machine Learning\u003c/h2\u003e\u003cp\u003eTwo predictive models, namely Multiple Linear Regression and Random Forest Regression, were developed using the caret and random forest packages in R. Input variables included district and water body type, while the dependent variable was total shrimp yield (kg). The data was randomly partitioned into training (80%) and testing (20%) sets. Model performance was assessed using Mean Absolute Error (MAE) and R-squared (R\u0026sup2;) scores. Although constrained by the dataset\u0026rsquo;s temporal limitation, extrapolated data points were simulated for exploratory forecasting of future yield values (e.g., 2025).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section3\"\u003e\u003ch2\u003e2.5.5 Time Series Forecasting with ARIMA Model\u003c/h2\u003e\u003cp\u003eAnnual shrimp yield data for the period 2001\u0026ndash;2023 were used for time series forecasting using an Autoregressive Integrated Moving Average (ARIMA) model. The ARIMA model is denoted as ARIMA (p,d,q), where:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003ep is the order of the autoregressive (AR) term,\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003ed is the degree of differencing,\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eq is the order of the moving average (MA) term.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eThe general ARIMA model for a time series \u003cem\u003eY\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003e ​ can be written as:\u003c/p\u003e\u003cp\u003eΦ\u003csub\u003ep\u003c/sub\u003e(B)(1\u0026thinsp;\u0026minus;\u0026thinsp;B)\u003csup\u003ed\u003c/sup\u003e\u003cem\u003eY\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;Θ\u003csub\u003eq\u003c/sub\u003e(B)\u003cem\u003eε\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003e\u003c/p\u003e\u003cp\u003eWhere,\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eΦ\u003csub\u003ep\u003c/sub\u003e(B)\u0026thinsp;=\u0026thinsp;1\u0026thinsp;\u0026minus;\u0026thinsp;\u003cem\u003eϕ\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003eB\u0026thinsp;\u0026minus;\u0026thinsp;\u003cem\u003eϕ\u003c/em\u003e\u003csub\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sub\u003eB\u003csup\u003e2\u003c/sup\u003e\u0026minus;⋯\u0026minus;ϕ\u003csub\u003ep\u003c/sub\u003eB\u003csup\u003ep\u003c/sup\u003e is the AR operator,\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eΘ\u003csub\u003eq\u003c/sub\u003e(B)\u0026thinsp;=\u0026thinsp;1\u0026thinsp;+\u0026thinsp;\u003cem\u003eθ\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003eB\u0026thinsp;+\u0026thinsp;\u003cem\u003eθ\u003c/em\u003e\u003csub\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sub\u003eB\u003csup\u003e2\u003c/sup\u003e+⋯+θ\u003csub\u003eq\u003c/sub\u003eB\u003csup\u003eq\u003c/sup\u003e is the MA operator,\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eB is the backshift operator (B\u003cem\u003eY\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003e=\u003cem\u003eY\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003e\u003cem\u003e\u0026minus;1\u003c/em\u003e​),\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cem\u003eε\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003eis a white noise error term.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eFor this study, the optimal model was selected as ARIMA (1,1,1) based on the lowest Akaike Information Criterion (AIC). The ARIMA (1,1,1) model can be expressed as:\u003c/p\u003e\u003cp\u003e\u003cem\u003eY\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e\u003cem\u003e\u0026prime;\u003c/em\u003e\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;\u003cem\u003e\u0026micro;\u003c/em\u003e\u0026thinsp;+\u0026thinsp;\u003cem\u003eϕ1Y\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u0026thinsp;\u0026minus;\u0026thinsp;1\u003c/em\u003e\u003c/sub\u003e\u003cem\u003e\u0026prime;\u003c/em\u003e + \u003cem\u003eεt\u003c/em\u003e\u0026thinsp;+\u0026thinsp;θ\u003csub\u003e1\u003c/sub\u003eε\u003csub\u003e\u003cem\u003et\u0026minus;1\u003c/em\u003e\u003c/sub\u003e​\u003c/p\u003e\u003cp\u003ewhere\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003e\u003cem\u003eY\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e\u003cem\u003e\u0026prime;\u003c/em\u003e\u003c/sup\u003e=Y\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003e\u0026minus;\u003cem\u003eY\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u0026minus;1\u003c/em\u003e\u003c/sub\u003e​ is the differenced series,\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u0026micro; is the mean,\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cem\u003eϕ\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003e​ is the AR(1) coefficient,\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cem\u003eθ\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003e​ is the MA(1) coefficient,\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cem\u003eε\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003eis the error term at time t,\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eModel order was determined by minimizing AIC, and model adequacy was evaluated using residual diagnostics (autocorrelation and normality tests). The ARIMA model was used to generate both in-sample predictions and out-of-sample forecasts of shrimp yield through 2035.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e3.1. Spatiotemporal Shrimp Yield Mapping\u003c/h2\u003e\u003cp\u003eThis study conducted a comprehensive spatiotemporal analysis of shrimp yield distribution across four major aquaculture districts of the Khulna Division including Khulna, Satkhira, Bagerhat, and Jashore, over the period 2001\u0026ndash;2024. Annual yield data sourced from the Department of Fisheries were georeferenced and processed in ArcGIS Pro, employing Jenks natural breaks classification to visualize spatial disparities. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents district-wise shrimp yield intensity across three distinct time intervals: early phase (2001\u0026ndash;2006), mid-phase (2010\u0026ndash;2014), and recent phase (2019\u0026ndash;2024). During the early phase (2001\u0026ndash;2006), shrimp production was primarily dominated by pond-based systems, averaging 4.8 t km⁻\u0026sup2; yr⁻\u0026sup1;, with modest hotspots (\u0026lt;\u0026thinsp;6 t km⁻\u0026sup2; yr⁻\u0026sup1;) observed in southwestern Satkhira. Contributions from natural water bodies (rivers, floodplains, and Sundarbans) remained minimal (\u0026lt;\u0026thinsp;2 t km⁻\u0026sup2; yr⁻\u0026sup1;), indicating a predominantly low-intensity aquaculture landscape during this period. In the mid-phase (2010\u0026ndash;2014), a marked transition toward high-intensity shrimp/prawn farms was observed, particularly in brackish water zones of Khulna and Satkhira, where improved levee construction and salinity-control infrastructure facilitated yields exceeding 12 t km⁻\u0026sup2; yr⁻\u0026sup1;. Jashore maintained relatively stable production levels from pond-based systems (6\u0026ndash;8 t km⁻\u0026sup2; yr⁻\u0026sup1;), while Bagerhat exhibited localized but inconsistent farm expansions. By the recent phase (2019\u0026ndash;2024), intensive shrimp/prawn farms, occupying less than 10% of total aquaculture area, accounted for nearly 45% of the division\u0026rsquo;s total shrimp yield, with peak yields reaching 18 t km⁻\u0026sup2; yr⁻\u0026sup1;. Conversely, natural water bodies continued to play a marginal role (\u0026lt;\u0026thinsp;2 t km⁻\u0026sup2; yr⁻\u0026sup1;) despite their geographical prevalence. Hotspot analysis using the Getis-Ord Gi* statistic revealed two highly significant high-yield clusters (z\u0026thinsp;\u0026gt;\u0026thinsp;2.58, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01): one located in southwestern Satkhira (21.900\u0026ndash;22.050\u0026deg; N, 89.000\u0026ndash;89.200\u0026deg; E) and the other in central Khulna (22.500\u0026ndash;22.650\u0026deg; N, 89.300\u0026ndash;89.450\u0026deg; E). The z-scores from the Gi* analysis indicate that these clusters represent statistically significant hotspots, meaning that the high yields observed in these regions are extremely unlikely to have occurred by random chance. This underscores the real, spatially concentrated nature of intensive aquaculture in these districts. These hotspots are visually represented in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Furthermore, trend surface interpolation indicated a gradual northward expansion of intensive aquaculture over the two-decade period, driven by technological advancements, targeted infrastructural investments, and proximity to domestic and export markets. Overall, these findings reveal a progressive spatial reconfiguration of shrimp aquaculture, shifting from traditional, low-yield natural water systems to engineered, high-intensity farming clusters, particularly in Khulna and Satkhira districts (see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e3.2. Production Share by Water Body Type (2001\u0026ndash;2024)\u003c/h2\u003e\u003cp\u003eWe analyzed annual yield data for six major water body categories like shrimp/prawn farms, pond-based systems, floodplains, seasonal culture, Sundarbans-derived harvests, and river captures to quantify their contributions over 2001\u0026ndash;2024. Data preprocessing included log-transformation to stabilize variance and 5-year moving-average smoothing to highlight long-term trends. Visualizations (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e: absolute yields; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e: relative shares) were generated in R using ggplot2.\u003c/p\u003e\u003cdiv id=\"Sec16\" class=\"Section3\"\u003e\u003ch2\u003e3.2.1. Absolute Production Trends\u003c/h2\u003e\u003cp\u003eHighest yields were observed in shrimp/prawn farms, which grew exponentially from 20,892 kg in 2001-02 (95% CI: 19,500\u0026thinsp;\u0026minus;\u0026thinsp;22,300) to 281,884 kg in 2023-24 (95% CI: 270,000-295,000). The compound annual growth rate (CAGR) for farms was 12.5% (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001, linear regression on log-scale). Pond systems showed linear growth (CAGR\u0026thinsp;=\u0026thinsp;9.8%, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01), rising from 20,164 kg (18,900\u0026thinsp;\u0026minus;\u0026thinsp;21,400) to 233,000 kg (220,000-245,000) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Floodplain yields increased steadily after 2005, reaching 48,120 kg by 2023-24, but with greater inter-annual variability (coefficient of variation\u0026thinsp;=\u0026thinsp;0.24). Seasonal culture emerged post-2010, with yields accelerating from 5,200 kg (2010) to 52,400 kg (2023-24), a ten-fold rise driven by improved hatchery technology. In contrast, Sundarbans yields declined at -7.2% CAGR (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), dropping from 12,384 kg (11,800\u0026thinsp;\u0026minus;\u0026thinsp;12,900) to 1,920 kg (1,700-2,100). River captures exhibited a modest negative trend, from 4,120 kg to 1,860 kg (-3.8% CAGR, p\u0026thinsp;=\u0026thinsp;0.07), indicating marginal statistical significance (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section3\"\u003e\u003ch2\u003e3.2.2. Relative Percentage Contribution Trends\u003c/h2\u003e\u003cp\u003eThe percentage share of shrimp/prawn farms increased from 33.1\u0026ndash;41.7%, with a peak of 44.5% in 2021-22. Pond-based systems maintained a stable share between 25% and 35%, peaking at 36.2% in 2014-15 before declining slightly to 32.8% in 2023-24. Floodplains held 5\u0026ndash;8% share, with a high of 7.1% in 2018-19 and a low of 4.9% in 2003-04. Seasonal culture reached 7.8% by 2023-24 from 0% before 2010, reflecting policy incentives introduced in 2012. Traditional sources shrank markedly: Sundarbans contribution fell from 19.7\u0026ndash;2.8%, with a breakpoint in 2008 coinciding with stricter conservation regulations. Rivers, baors, and beels combined dropped from 6.5% to below 4%, with significant year-to-year volatility (standard deviation\u0026thinsp;=\u0026thinsp;1.2%). Statistical change-point analysis (Pelt algorithm) identified three structural shifts (2006, 2012, 2018), aligning with major policy and environmental events. These patterns underscore the shift toward engineered aquaculture systems and away from natural harvests \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e\u003cb\u003e).\u003c/b\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003e3.3. Statistical Impact Analysis\u003c/h2\u003e\u003cp\u003eWe evaluated the impact of water body type on shrimp yield using Principal Component Analysis (PCA) and one-way Analysis of Variance (ANOVA). Yield data for the six water body categories (n\u0026thinsp;=\u0026thinsp;138 annual observations) were log-transformed to meet normality assumptions and standardized prior to multivariate analysis. PCA was performed in R (v4.2.1) using the prcomp function on the covariance matrix. The first two components (PC1 and PC2) explained 57.7% and 11.9% of total variance, respectively. PC1 exhibited high positive loadings for shrimp/prawn farms (0.74) and pond systems (0.68), indicating overall aquaculture intensity, while PC2 contrasted mangrove-influenced (Sundarbans, 0.62) and pen culture systems (0.57), reflecting habitat-driven variation (See Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePCA Loadings for Shrimp Yield by Water Body Type. Loadings of ten water body types on the first two principal components. PC1 (57.70%) reflects general aquaculture systems, while PC2 (11.88%) highlights pen and mangrove-influenced systems.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWater Body\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePC1\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePC2\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRiver\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.361\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.288\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSundarbans\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.254\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.419\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBeel\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.346\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;0.159\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFlood Plain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.381\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;0.229\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePond\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.391\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;0.194\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSeasonal_Cultured\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.354\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;0.261\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBaor\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.111\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;0.333\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eShrimp_Prawn_Farm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.390\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.122\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePen_Culture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.163\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.654\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCage_Culture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.265\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;0.067\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\u003eOne-way ANOVA tested for differences in mean yields among water body types. Residual diagnostics confirmed homogeneity of variances (Levene\u0026rsquo;s test, p\u0026thinsp;=\u0026thinsp;0.12) and approximate normality of residuals (Shapiro\u0026ndash;Wilk, p\u0026thinsp;=\u0026thinsp;0.08). The ANOVA model was significant (F(5,132)\u0026thinsp;=\u0026thinsp;18.45, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), indicating that at least one category\u0026rsquo;s mean yield differed. Post-hoc Tukey HSD comparisons revealed that shrimp/prawn farms and pond systems had significantly higher mean yields than Sundarbans and river captures (all pairwise p\u0026thinsp;\u0026lt;\u0026thinsp;0.01), while floodplains and seasonal culture systems occupied an intermediate group without significant difference between them (p\u0026thinsp;=\u0026thinsp;0.23, see Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Figure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents the PCA biplot with water body type centroids and the yield distribution boxplots with ANOVA results annotated.\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\u003eOne-way ANOVA results for yield differences among water body types.\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=\"\u0026times;\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSource\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSum of Squares\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003edf\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eF value\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003ePr(\u0026gt;\u0026thinsp;F)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eType\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c2\"\u003e\u003cp\u003e2.8449 \u0026times; 10^10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e87.438\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e3.5643 \u0026times; 10⁻\u0026sup3;⁸\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eResidual\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c2\"\u003e\u003cp\u003e7.7436 \u0026times; 10^9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e119\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cdiv id=\"Sec19\" class=\"Section3\"\u003e\u003ch2\u003e3.3.1 Network Correlation Analysis of Shrimp Yields by Water Body Type\u003c/h2\u003e\u003cp\u003eTo elucidate the interrelationships among shrimp yields across various aquatic environments, a network correlation analysis was performed. Pairwise Pearson correlation coefficients were calculated using yield data from ten water body types: River, Sundarbans, Beel, Flood Plain, Pond, Seasonal Cultured, Baor, Shrimp/Prawn Farm, Pen Culture, and Cage Culture. This approach facilitates the identification of highly correlated yield patterns that reflect ecological or management linkages across districts and years. Figure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents a circular correlation network diagram depicting relationships where the absolute correlation exceeds 0.1. Nodes represent water body types, while edges indicate the strength and direction of correlations, with edge widths proportional to correlation magnitude. The Seasonal Cultured node is distinctly highlighted due to its unique correlation profile and significance in the shrimp production system. This visualization reveals dense positive correlations among intensive farming systems, such as shrimp/prawn farms, ponds, and flood plains, contrasted with weaker or negative correlations involving natural capture sources like the Sundarbans. Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e summarizes the statistically significant pairwise correlations where |r| \u0026gt;0.5. Notably, strong positive correlations were observed between River and Shrimp/Prawn Farm yields (r\u0026thinsp;=\u0026thinsp;0.88), as well as Seasonal Cultured and Pen Culture systems (r\u0026thinsp;=\u0026thinsp;0.90). Conversely, a significant negative correlation was identified between Sundarbans and Shrimp/Prawn Farm yields (r = \u0026minus;\u0026thinsp;0.57), underscoring differing production dynamics in natural versus engineered environments. These findings provide critical insight into the interconnected productivity trends across aquatic systems, guiding targeted resource management and policy development for sustainable shrimp aquaculture in coastal Bangladesh.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSignificant pairwise Pearson correlations of shrimp yields between water body types in Bangladesh\u0026rsquo;s coastal districts (2001\u0026ndash;2024).Correlations with absolute value greater than 0.5 are shown.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWater Body Type 1\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eWater Body Type 2\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCorrelation (r)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRiver\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFlood_Plain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.58\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRiver\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eShrimp_Prawn_Farm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.88\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRiver\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCage_Culture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.56\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSundarbans\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eShrimp_Prawn_Farm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;0.57\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFlood_Plain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSeasonal_Cultured\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u0026ndash;0.51\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFlood_Plain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eShrimp_Prawn_Farm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.74\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePond\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePen_Culture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.66\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSeasonal_Cultured\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePen_Culture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.90\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eShrimp_Prawn_Farm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCage_Culture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.52\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003e3.4. Forecasting Outcomes\u003c/h2\u003e\u003cp\u003eWe applied an ARIMA (1,1,1) model to annual shrimp-yield data (2001\u0026ndash;2023), partitioning the series into calibration (2001\u0026ndash;2016) and validation (2017\u0026ndash;2023) sets. Model orders were selected via Akaike Information Criterion (AIC), and residual diagnostics confirmed adequacy.\u003c/p\u003e\u003cdiv id=\"Sec21\" class=\"Section3\"\u003e\u003ch2\u003e3.4.1. Predicted vs Actual Shrimp Yield\u003c/h2\u003e\u003cp\u003eTo evaluate model performance, predicted yields were compared with observed values from 2001 to 2023, and extended forecasts were generated through 2035. The model closely tracked actual yields during stable periods (2010\u0026ndash;2016), with deviations within \u0026plusmn;\u0026thinsp;5\u0026ndash;10%. However, it notably over predicted 2022 yield (424,805 t vs. 328,736 t) and underestimated initial 2001 values due to data sparsity \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e; Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eActual and ARIMA (1,1,1)-Predicted annual shrimp yields (tonnes) in Bangladesh\u0026rsquo;s coastal districts from 2001 to 2035, including prediction errors for historical years (2001\u0026ndash;2023). Historical data (2001\u0026ndash;2023) are sourced from the Department of Fisheries (DoF), while forecasted values (2024\u0026ndash;2035) are generated by the ARIMA model. Error values represent the difference between Actual and Predicted yields and are shown only for historical years where actual data are available.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eYear\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eActual Yield (t)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePredicted Yield (t)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eType\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eError (t)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2001\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e152,166.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e63,594.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e88,571.82\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2002\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e156,354.20\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e169,250.24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-12,896.04\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2004\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e184,706.42\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e167,666.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e17,039.57\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2005\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e196,178.32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e193,115.81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e3,062.51\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2006\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e197,939.36\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e197,546.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e393.20\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2007\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e207,915.46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e202,769.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e5,146.44\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2009\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e246,193.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e227,073.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e19,120.69\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2010\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e287,521.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e269,212.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e18,308.69\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2011\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e304,702.30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e298,874.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e5,827.80\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2012\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e348,373.44\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e326,551.47\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e21,821.97\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2013\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e366,253.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e360,664.81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e5,588.19\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2014\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e383,420.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e375,409.49\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e8,010.51\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2015\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e436,378.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e410,077.90\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e26,300.10\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2016\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e444,129.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e444,605.56\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-476.56\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2018\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e468,962.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e455,900.76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e13,061.40\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2019\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e495,998.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e484,115.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e11,882.02\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2020\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e495,640.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e497,872.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-2,232.67\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2021\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e509,952.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e502,087.29\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e7,864.71\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2022\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e328,736.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e424,804.74\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-96,068.74\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2023\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e563,768.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e424,386.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHistorical\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e139,381.18\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2024\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e424,386.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eForecast\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2025\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e521,231.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eForecast\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2026\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e453,942.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eForecast\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2027\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e500,695.91\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eForecast\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2028\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e468,210.46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eForecast\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2029\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e490,781.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eForecast\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2030\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e475,098.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eForecast\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2031\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e485,995.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eForecast\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2032\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e478,424.39\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eForecast\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2033\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e483,685.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eForecast\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2034\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e480,029.87\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eForecast\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2035\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e482,569.60\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eForecast\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec22\" class=\"Section3\"\u003e\u003ch2\u003e3.4.2. Model Performance Comparison\u003c/h2\u003e\u003cp\u003eForecast accuracy over 2017\u0026ndash;2023 was quantified by mean absolute error and related metrics (Table \u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). We further assessed pointwise prediction errors (Actual \u0026ndash; Predicted) to characterize bias and variance.Residual diagnostics further confirm model adequacy. Figure 6A presents the standardized residuals plotted against time, revealing a random scatter around zero with no discernible patterns or changes in variance, consistent with homoscedasticity. Figure 6B displays the ACF of residuals up to lag20; all autocorrelation coefficients fall within the 95% confidence bounds (shaded region), validating the white noise assumption for residuals.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec23\" class=\"Section3\"\u003e\u003ch2\u003e3.4.3. Error Distribution and AIC Diagnostics\u003c/h2\u003e\u003cp\u003eWe examined the distribution of pointwise forecast errors (Actual \u0026ndash; Predicted) for 2001\u0026ndash;2023 by plotting a histogram overlaid with a kernel density curve (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e7\u003c/span\u003eA). The error distribution was approximately normal and centered near zero, with a mild negative skew attributable to systematic overestimation in 2022. To validate model selection, we computed AIC values for candidate ARIMA (p,1,q) models with p and q ranging from 0 to 2. The ARIMA (1,1,1) specification yielded the lowest AIC (1028.5), supporting its optimal fit among tested configurations (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e7\u003c/span\u003eB).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec24\" class=\"Section2\"\u003e\u003ch2\u003e3.5. Predictive Performance of Machine Learning Regression Models\u003c/h2\u003e\u003cp\u003eTo assess the ability of machine learning models to predict shrimp yield across districts and years, we applied both Multiple Linear Regression and Random Forest Regression to the cleaned panel dataset (see Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e5\u003c/span\u003e for summary statistics). Predictor variables included district, year, and the yields from all major water body types (River, Sundarbans, Beel, Flood_Plain, Pond, Seasonal_Cultured, Baor, Shrimp_Prawn_Farm, Pen_Culture, Cage_Culture). The target variable was total shrimp yield (kg). All categorical variables were one-hot encoded, and the dataset was randomly partitioned into training (80%) and testing (20%) subsets.\u003c/p\u003e\u003cp\u003eModel performance was evaluated using R\u0026sup2; and MAE. The Random Forest Regression model demonstrated superior predictive accuracy, achieving an R\u0026sup2; of 0.91 and a MAE of 6,950 kg on the testing set. The Multiple Linear Regression model yielded an R\u0026sup2; of 0.84 and a MAE of 11,700 kg. Feature importance analysis from the Random Forest model (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e8\u003c/span\u003eA) indicated that shrimp/prawn farm, pond, and floodplain yields were the most influential predictors, followed by district and year effects.\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e8\u003c/span\u003eB presents a scatter plot comparing observed versus predicted total yields for both models. The Random Forest model tracked the actual values more closely, especially for high-yield districts, while the linear regression model tended to under predict in these cases.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePredictive performance of regression models for shrimp yield.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eR\u0026sup2;\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMAE (kg)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMultiple Linear Regression\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e11,700\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRandom Forest Regression\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.91\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e6,950\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec25\" class=\"Section2\"\u003e\u003ch2\u003e3.6 Model Performance Evaluation via Cross-Validation\u003c/h2\u003e\u003cp\u003eTo assess the robustness and generalizability of the predictive models, a 10-fold cross-validation was performed on both Random Forest and Multiple Linear Regression models using the shrimp yield dataset. This resampling technique minimizes overfitting and provides reliable estimates of model performance across different data partitions. The Random Forest model exhibited superior predictive capability, achieving a median coefficient of determination (R\u0026sup2;) of approximately 0.90 across folds, indicating a strong ability to explain variance in shrimp yield. In contrast, the Multiple Linear Regression model attained a lower median R\u0026sup2; near 0.80, reflecting its limited capacity to capture non-linear relationships within the data.Evaluation of prediction errors via Mean Absolute Error (MAE) further corroborated these findings. The Random Forest model consistently yielded lower MAE values (~\u0026thinsp;7,000 kg) compared to the Linear Regression model (~\u0026thinsp;11,000 kg), demonstrating higher accuracy and precision in yield forecasting.\u003c/p\u003e\u003cp\u003eThe results are visually summarized in boxplots (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e9\u003c/span\u003e), illustrating the distribution and variability of R\u0026sup2; and MAE scores across folds for both models. Overlaying individual fold scores emphasizes the consistency of Random Forest\u0026rsquo;s superior performance. The distinct color schemes enhance interpretability and distinction between models.Overall, the cross-validation analysis confirms the advantage of ensemble machine learning approaches over traditional linear methods for shrimp yield prediction, supporting their application in aquaculture management for improved forecasting reliability.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study utilized an extensive secondary dataset, applying advanced spatiotemporal and predictive analyses to investigate shrimp yield trends across southern coastal districts of Bangladesh from 2001 to 2024. Using data provided by the Department of Fisheries (DoF), the analysis revealed significant variations in shrimp productivity across different aquatic environments, thereby illuminating both growth opportunities and sustainability challenges within the region's aquaculture industry.\u003c/p\u003e\u003cp\u003eSignificant productivity growth was observed in intensive shrimp/prawn farms and pond-based aquaculture systems, which recorded substantial compound annual growth rates (CAGR) of 12.5% and 9.8%, respectively. This notable shift toward intensive farming practices is likely driven by improved infrastructure, technological advancements, and favorable market dynamics, aligning closely with trends observed in previous studies (Ahmed \u0026amp; Flaherty \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Bush 2019; Bashar et al. \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). These findings resonate with the recent research of Siddique et al. (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2024\u003c/span\u003ea), who demonstrated the effective use of ARIMA modeling to forecast tilapia production in Bangladesh, underscoring the increasing reliance on data-driven predictive approaches to enhance aquaculture management and planning. Their work highlights how temporal forecasting models can provide critical insights for scaling aquaculture production in line with market and environmental changes, a strategy that appears equally promising for shrimp aquaculture in coastal Bangladesh.\u003c/p\u003e\u003cp\u003eConversely, natural water bodies, notably the Sundarbans mangrove systems and riverine fisheries, experienced substantial yield declines due to ongoing environmental stressors such as increasing salinity intrusion, habitat degradation, and enhanced conservation regulations. These observations resonate with the findings of previous studies (Meisner\u0026amp; Ahmed \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Padhy et al. \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Gopal \u0026amp; Chauhan \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2006\u003c/span\u003e), which highlighted similar environmental concerns and their implications for long-term sustainability. Siddique et al. (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2024\u003c/span\u003eb) further emphasize the critical influence of climatic factors on aquaculture water quality parameters, demonstrating that fluctuations in temperature and rainfall substantially affect pond water temperature and other key quality indicators that directly impact fish health and growth. Such climatic variability could exacerbate challenges for shrimp farming in coastal ecosystems, especially under the increasing pressure of climate change, as reflected in the spatial contraction of yield in natural habitats observed in this study.\u003c/p\u003e\u003cp\u003eThe statistical analyses employing PCA and ANOVA demonstrated significant variations in yields among different aquatic environments. The analyses effectively differentiated high-productivity intensive farming systems from lower-yield natural capture systems, confirming the hypothesis that productivity varies substantially based on distinct management practices and ecological conditions.These findings are consistent with previous research emphasizing the significance of specific ecological and management factors on aquaculture productivity (Mandal \u0026amp; Ghosh \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Zhang et al. \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Li et al. \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Alfiansah et al. \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Furthermore, Siddique et al. (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2025\u003c/span\u003ea) integrate ARIMA and ARIMAX modeling to reveal how combined climatic and water quality parameters precisely influence broodfish growth in tilapia culture, illustrating the importance of multi-factor models for understanding complex growth dynamics. These integrative modeling approaches could be adapted to shrimp aquaculture to better anticipate growth performance under varying environmental and management scenarios.\u003c/p\u003e\u003cp\u003eIn predictive modeling, the Random Forest Regression approach exhibited superior performance relative to traditional Multiple Linear Regression models, yielding higher accuracy and reduced predictive errors. This result highlights complex, non-linear relationships among influential predictors, a finding consistent with earlier studies applying advanced machine learning techniques in aquaculture and agricultural yield forecasting (Edeh et al. \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Nazmi et al. \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Chen et al. \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Ahmed et al. \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The demonstrated success of machine learning models complements the time series forecasting efforts, reinforcing the utility of ensemble methods and hybrid modeling frameworks for improving yield predictions and informing decision-making. Similarly, Siddique et al. (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2025\u003c/span\u003eb) employ ARIMA modeling to forecast climatic variables such as air temperature and rainfall in Mymensingh, Bangladesh, revealing critical implications for aquaculture management that directly affect production outcomes. Their findings suggest that integrating climatic forecasts with aquaculture yield models could significantly enhance adaptive management strategies in Bangladesh\u0026rsquo;s coastal aquaculture systems.\u003c/p\u003e\u003cp\u003eAdditionally, the ARIMA forecasting model provided valuable medium-term projections of shrimp yields, despite certain limitations, notably its overestimation of yields for the year 2022. This discrepancy underscores the need for real-time data integration and adaptive modeling frameworks to enhance predictive accuracy in response to dynamic environmental and market conditions (Shohan \u0026amp; Haque \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). As highlighted by Siddique et al. (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2024\u003c/span\u003ea, 2025a), the inclusion of exogenous variables such as climatic factors in ARIMAX models can improve forecasting performance by accounting for environmental drivers that influence aquaculture productivity. Future adoption of such integrative models in shrimp aquaculture forecasting could reduce prediction errors and better capture the effects of anomalous events such as extreme weather or disease outbreaks.\u003c/p\u003e\u003cp\u003eOne of the key insights of this study is the application of ARIMA and Random ForestRegression models for forecasting shrimp yields. While ARIMA provided valuable medium-term projections, it notably overestimated the yields for 2022, raising concerns about its suitability for accurately predicting anomalous events. This overestimation may be attributed to the model's inability to incorporate exogenous variables, such as climatic factors, that influence shrimp production. Similar issues with ARIMA have been noted in other studies, where its predictive accuracy fluctuated in the presence of unexpected changes in environmental or market conditions (Siddique et al. \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). For instance, the model failed to account for abrupt climatic events or shifts in regulatory policies, which could have skewed yield estimates. To address such limitations, future studies could explore the use of ARIMAX models or hybrid approaches that incorporate external variables such as temperature, rainfall, orsalinity levels to improve the accuracy of predictions.\u003c/p\u003e\u003cp\u003eIn contrast, the Random Forest model demonstrated superior performance, achieving an R\u0026sup2; of0.91 and a MAE of 6,950 kg. This model accounted for complex, non-linear interactions between variables, making it more capable of handling the intricacies of the aquaculture environment. However, it is important to note that machine learning models like Random Forest are prone to overfitting if not properly tuned. In this study, cross-validation techniques were applied to minimize such risks, but further research could explore additional techniques like hyper parameter optimization and ensemble learning to ensure the robustness of the model in diverse environmental conditions.\u003c/p\u003e\u003cp\u003eThe study also underscores the significant impact of environmental stressors, particularly salinity intrusion and habitat degradation, on shrimp yields in natural water bodies such as the Sundarbans and riverine systems. These findings align with previous studies indicating that increasing salinity in coastal areas, coupled with mangrove deforestation and other anthropogenic activities, has led to a decline in shrimp production from natural capture fisheries (Meisner\u0026amp; Ahmed \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). However, while these factors were identified as key contributors to yield reductions, the study could benefit from more detailed temporal data on salinity levels and pollution to provide a clearer understanding of how these environmental variables correlate with yield fluctuations over time.\u003c/p\u003e\u003cp\u003eWhile this research contributes significantly to the understanding of shrimp production dynamics, it is subject to several limitations. Primarily, the absence of detailed spatial-temporal data on farm sizes and localized environmental management practices restricts the precision of yield estimations and comprehensive sustainability assessments. Furthermore, reliance on secondary data inherently omits detailed socio-economic insights that primary data collection could provide. Complementing this, Siddique et al. (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2024\u003c/span\u003eb) highlight the importance of continuous water quality monitoring to capture fine-scale environmental variations critical for aquaculture success, underscoring a gap that future studies should address by integrating field-based data collection and remote sensing technologies.\u003c/p\u003e\u003cp\u003eFuture research endeavors should integrate primary field surveys and remote sensing data, facilitating improved spatial resolution and more robust sustainability analyses. Policymakers and stakeholders should leverage the insights provided by this study to implement sustainable intensification strategies and targeted ecological interventions. Such integrated management approaches will be pivotal in ensuring the sustainability, resilience, and continued growth of Bangladesh\u0026rsquo;s shrimp aquaculture industry amid ongoing environmental and economic challenges. The advanced forecasting and predictive modeling techniques exemplified in the recent studies by Siddique and colleagues provide a valuable framework for enhancing aquaculture resilience by explicitly incorporating environmental variability and climatic influences into production planning and policy formulation.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study demonstrates the value of secondary data in advancing knowledge on shrimp yield dynamics across diverse aquatic environments in Bangladesh's southwestern coastal districts. By leveraging two decades of production data from the Department of Fisheries, comprehensive spatiotemporal analyses revealed significant disparities in shrimp productivity driven largely by water body type and management practices. Intensive shrimp/prawn farms and pond systems consistently outperformed natural water sources, exhibiting higher compound annual growth rates due to infrastructural investments, technological innovations, and targeted management approaches. Statistical assessments confirmed significant yield differences among aquatic environments, emphasizing that anthropogenic intervention can drastically influence production outcomes. Predictive modeling results underscored the advantage of advanced machine learning approaches, particularly Random Forest Regression, in capturing non-linear relationships and providing robust yield estimations. Time-series forecasting via ARIMA contributed to future yield projections, highlighting potential production trends through 2035 despite limitations in accurately predicting anomalous events. Overall, this research highlights a clear shift from traditional, environmentally dependent shrimp capture methods toward engineered aquaculture systems, raising both opportunities for sustainable intensification and concerns over ecological trade-offs. The findings provide actionable insights for policymakers, investors, and local stakeholders seeking to enhance shrimp aquaculture productivity while ensuring long-term environmental resilience. Future research integrating high-resolution primary data, remote sensing technologies, and real-time monitoring systems is essential to refine predictive models and guide adaptive, sustainable aquaculture management strategies in Bangladesh.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003cbr\u003e The authors declare that they received no financial support for the research, authorship, or publication of this article. No funds, grants, or other support, whether internal or external, were provided by any institution, department, or funding agency, including public, commercial, or nonprofit organizations. All costs related to the study design, data collection, analysis, and manuscript preparations were covered by the authors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical statement \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research utilizes secondary data from publicly available databases and literature, strictly adhering to ethical standards in data collection, analysis, and reporting. The use of existing data complies with the terms of use from the original sources, with all necessary citations provided. No unauthorized access or data manipulation has occurred. The study upholds the confidentiality and privacy of individuals, aligning with ethical guidelines to promote responsible research conduct.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability statement \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data presented in this study are available upon request from the corresponding author.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contribution\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eI. A.: Overall data analysis \u0026amp; presentation and writing the original draft; M. A. B. S.: Data analysis, Editing the draft; M. M. H.: Validation, Review \u0026amp; editing the draft; A. K. S. A.: Concept development, validation, overall technical analysis and supervision and editing the draft.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAhmed F, Bijoy MHI, Hemal HR, Noori SRH (2024) Smart aquaculture analytics: Enhancing shrimp farming in Bangladesh through real-time IoT monitoring and predictive machine learning analysis. Heliyon 10(17).\u003c/li\u003e\n\u003cli\u003eAhmed N, Flaherty MS (2013) Opportunities and challenges for the development of prawn farming with fish and rice in southeast Bangladesh: Potential for food security and economic growth. Food Security 5(5):637\u0026ndash;649.\u003c/li\u003e\n\u003cli\u003eAlfiansah YR, Hassenr\u0026uuml;ck C, Kunzmann A, Taslihan A, Harder J, G\u0026auml;rdes A (2018) Bacterial abundance and community composition in pond water from shrimp aquaculture systems with different stocking densities. Frontiers in Microbiology 9:2457.\u003c/li\u003e\n\u003cli\u003eBashar A, Heal RD, Hasan NA, Salam MA, Haque MM (2022) COVID-19 impacts on the Bangladesh shrimp industry: A sequential survey-based case study from southwestern Bangladesh. Fisheries Science 88(6):767\u0026ndash;786.\u003c/li\u003e\n\u003cli\u003eBush SR, Belton B, Little DC, Islam MS (2019) Emerging trends in aquaculture value chain research. Aquaculture 498:428\u0026ndash;434.\u003c/li\u003e\n\u003cli\u003eChen S, Tao F, Pan C, Hu X, Ma H, Li C et al (2021) Modeling quality changes in Pacific white shrimp (Litopenaeus vannamei) during storage: Comparison of the Arrhenius model and random forest model. Journal of Food Processing and Preservation 45(1):e14999.\u003c/li\u003e\n\u003cli\u003eDepartment of Fisheries (DoF) (2024) Yearbook of Fisheries Statistics of Bangladesh 2024. Ministry of Fisheries and Livestock, Government of the People\u0026rsquo;s Republic of Bangladesh, Dhaka.\u003c/li\u003e\n\u003cli\u003eEdeh MO, Dalal S, Obagbuwa IC, Prasad BS, Ninoria SZ, Wajid MA, Adesina AO (2022) Bootstrapping random forest and CHAID for prediction of white spot disease among shrimp farmers. Scientific Reports 12(1):20876.\u003c/li\u003e\n\u003cli\u003eGopal B, Chauhan M (2006) Biodiversity and its conservation in the Sundarban mangrove ecosystem. Aquatic Sciences 68(3):338\u0026ndash;354.\u003c/li\u003e\n\u003cli\u003eGulakhmadov M, Chen X, Gulakhmadov A, Nadeem MU, Gulahmadov N, Liu T (2023) Performance analysis of precipitation datasets at multiple spatio-temporal scales over a dense gauge network in a mountainous domain of Tajikistan, Central Asia. Remote Sensing 15(5):1420.\u003c/li\u003e\n\u003cli\u003eHe A, Huang J, Sun Z, Zhou J, Yang C (2023) Spatial and temporal evolution characteristics of the Salween River delta from 1973 to 2021. Remote Sensing 15(5):1467.\u003c/li\u003e\n\u003cli\u003eKaur G, Adhikari N, Krishnapriya S, Wawale SG, Malik RQ, Zamani AS et al (2023) Recent advancements in deep learning frameworks for precision fish farming: Opportunities, challenges, and applications. Journal of Food Quality 2023:4399512.\u003c/li\u003e\n\u003cli\u003eLi Y, Cao S, Jiang S, Huang J, Yang Q, Jiang S et al (2024) Comparative study of nutritional composition, physiological indicators, and genetic diversity in Litopenaeus vannamei from different aquaculture populations. Biology 13(9):722.\u003c/li\u003e\n\u003cli\u003eMandal A, Ghosh AR (2024) Role of artificial intelligence (AI) in fish growth and health status monitoring: A review on sustainable aquaculture. Aquaculture International 32(3):2791\u0026ndash;2820.\u003c/li\u003e\n\u003cli\u003eMeisner CA, Ahmed S (2023) Shrimp farming and the question of land and social environment. In: Transforming Bangladesh: Geography, People, Economy and Environment. Springer International Publishing, Cham, pp 55\u0026ndash;64.\u003c/li\u003e\n\u003cli\u003eNazmi H, Siau NZ, Bramantoro A, Suhaili WS (2023) Predictive modeling of marine fish production in Brunei Darussalam\u0026rsquo;s aquaculture sector: A comparative analysis of machine learning and statistical techniques. International Journal of Advanced and Applied Sciences 10(7):109\u0026ndash;126.\u003c/li\u003e\n\u003cli\u003eOlipp N, Woschank M, Hoffelner M (2024) Exploration of the framework conditions for measures to reduce resource consumption in the manufacturing industry with a focus on the circular economy: A systematic secondary data research. Production \u0026amp; Manufacturing Research 12(1):2431723.\u003c/li\u003e\n\u003cli\u003ePadhy SR, Dash PK, Bhattacharyya P (2022) Challenges, opportunities, and climate change adaptation strategies of the mangrove\u0026ndash;agriculture ecosystem in the Sundarbans, India: A review. Wetlands Ecology and Management 30(1):191\u0026ndash;206.\u003c/li\u003e\n\u003cli\u003eRahman LF, Marufuzzaman M, Alam L, Bari MA, Sumaila UR, Sidek LM (2021) Developing an ensembled machine learning prediction model for marine fish and aquaculture production. Sustainability 13(16):9124.\u003c/li\u003e\n\u003cli\u003eRivera-Ferre MG (2009) Can export-oriented aquaculture in developing countries be sustainable and promote sustainable development? The shrimp case. Journal of Agricultural and Environmental Ethics 22:301\u0026ndash;321.\u003c/li\u003e\n\u003cli\u003eShohan MH, Haque MM (2025) Forecasting the production and export of shrimp in Bangladesh: A policy-focused time series analysis. Frontiers in Aquaculture 4:1541025.\u003c/li\u003e\n\u003cli\u003eSiddique MAB, Mahalder B, Haque MM, Ahamad AKS (2025) Forecasting air temperature and rainfall in Mymensingh, Bangladesh with ARIMA: Implications for aquaculture management. Egyptian Journal of Aquatic Research (in press).\u003c/li\u003e\n\u003cli\u003eSiddique MAB, Mahalder B, Haque MM, Ahammad AKS (2025) Impact of climatic and water quality parameters on tilapia (Oreochromis niloticus) broodfish growth: Integrating ARIMA and ARIMAX for precise modeling and forecasting. PLOS ONE 20(3):e0313846.\u003c/li\u003e\n\u003cli\u003eSiddique MAB, Mahalder B, Haque MM, Ahammed AKS (2024) Impact of climatic factors on water quality parameters in tilapia broodfish ponds and predictive modeling of pond water temperature with ARIMAX. Heliyon 10(18):e37717.\u003c/li\u003e\n\u003cli\u003eSiddique MAB, Mahalder B, Haque MM, Shohan MH, Biswas JC, Akhtar S, Ahammad AKS (2024) Forecasting of tilapia (Oreochromis niloticus) production in Bangladesh using ARIMA model. Heliyon 10(5):e27111.\u003c/li\u003e\n\u003cli\u003eZhang P, Huang Q, Peng R, Jiang X, Jiang M, Zeng G, Lin J (2022) Environmental factors of rearing water and growth performance of shrimp (Penaeus vannamei) in a microalgal monoculture system. Aquaculture 561:738620.\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":"Shrimp aquaculture, secondary data analysis, spatiotemporal modeling, predictive analytics, Random Forest Regression, ARIMA forecasting, water body types","lastPublishedDoi":"10.21203/rs.3.rs-7456160/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7456160/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eShrimp aquaculture is vital to Bangladesh\u0026rsquo;s economy but faces challenges from environmental degradation, climate variability, and resource management. This study analyzed secondary data (2001\u0026ndash;2024) from four southern coastal districts-Khulna, Satkhira, Bagerhat, and Jashore-to assess spatiotemporal shrimp yield patterns using Multiple Linear Regression, Random Forest Regression, and ARIMA forecasting. Random Forest showed superior accuracy (R\u0026sup2;=0.91, MAE\u0026thinsp;=\u0026thinsp;6,950 kg), capturing complex ecological and management interactions. GIS spatial analysis identified significant yield clusters in intensive shrimp farms with a 12.5% compound annual growth rate, while natural water bodies and the Sundarbans exhibited declining productivity due to habitat degradation and salinity changes. PCA and ANOVA confirmed significant yield differences among aquatic environments, highlighting intensive farming benefits. ARIMA forecasting predicted general trends but was less accurate during anomalies. These results emphasize the need for targeted infrastructure, sustainable practices, and data-driven policies to improve resilience and productivity in Bangladesh\u0026rsquo;s shrimp aquaculture sector.\u003c/p\u003e","manuscriptTitle":"Machine Learning and ARIMA Models for Spatiotemporal Analysis and Forecasting of Shrimp Yields across the Aquatic Systems of Southern Coastal Bangladesh","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-16 09:56:44","doi":"10.21203/rs.3.rs-7456160/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":"25f515c2-a1e1-4f7a-baca-faffe5d1f3f5","owner":[],"postedDate":"October 16th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-10-27T14:31:27+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-16 09:56:44","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7456160","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7456160","identity":"rs-7456160","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}