Quantifying Burdur Lake Shrinkage (2018–2025): Trend Analysis and Uncertainty Quantification with Sentinel-2 Imagery and Monte Carlo

Quantifying Burdur Lake Shrinkage (2018–2025): Trend Analysis and Uncertainty Quantification with Sentinel-2 Imagery and Monte Carlo | 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 Quantifying Burdur Lake Shrinkage (2018–2025): Trend Analysis and Uncertainty Quantification with Sentinel-2 Imagery and Monte Carlo Tarik Emre Yorulmaz This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8269709/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 Burdur Lake, a Ramsar site in Turkey, exemplifies the global crisis of shrinking endorheic lakes under anthropogenic and climatic pressures. This study quantifies its surface area decline from 2018 to 2025 using Sentinel-2 imagery (10 m resolution), revealing a median shrinkage rate of -1.64 km²/year (95% CI: -1.83 to -1.47 km²/year), totaling 11.59 km². A Monte Carlo Simulation (MCS) framework, integrated with the non-parametric Sen’s Slope estimator, propagates classification uncertainty (± 1%, validated at 98–99% accuracy), addressing a common gap in remote sensing time-series analysis. Adaptive Otsu thresholding outperformed conventional zero-thresholding (McNemar test, p < 0.05), ensuring reliable water-land delineation in this dynamic, saline basin. The decline, driven by dams and groundwater abstraction, mirrors trends in lakes like Urmia and Aral Sea, affecting biodiversity, including the endangered White-headed Duck. These findings provide a baseline for policy interventions, such as revised reservoir management and irrigation optimization to restore hydrological balance. This methodology offers an approach for monitoring lake dynamics, supporting water management and ecological conservation. Geographic Information Systems lake shrinkage Sentinel Otsu trend uncertainty Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1 - INTRODUCTION Surface water bodies, such as lakes and reservoirs, are fundamental to the Earth’s hydrologic cycle, global climate regulation, and biodiversity support ( Du et al., 2016 ). These ecosystems are increasingly threatened by anthropogenic pressures—including irrigation demands, dam construction, and groundwater extraction—and climate change-induced evaporation, necessitating accurate, scalable monitoring for sustainable water resource management ( Tesfaye and Breuer, 2024 ) . Remote sensing, with its cost-effective, large-scale capabilities, surpasses traditional in-situ methods, providing dynamic data essential for tracking environmental change ( Tesfaye and Breuer, 2024 ; Che et al., 2025 ). Among vulnerable regions, Mediterranean endorheic lakes are particularly at risk, with Burdur Lake in southwest Turkey—a Ramsar site of international ecological significance—exemplifying this trend. Historical analyses have documented severe shrinkage, with a 37% surface area loss from 210 km² (1975) to 131 km² (2016) ( Davraz et al., 2019 ), and recent studies report a 1.47% decline between 2000 and 2024 ( Ghorbani and Pamucar, 2025 ) and a one-fifth reduction from 1987 to 2000 ( Sarp and Ozcelik, 2017 ) . These changes are predominantly attributed to upstream water extraction for agriculture via dams and groundwater pumping ( Dervisoglu et al., 2022 ), trends likely persisting into the 2018–2025 period. Despite these insights, a critical methodological challenge persists in remote sensing of lake dynamics. The transition from continuous spectral indices, such as the Normalized Difference Water Index (NDWI), to binary water/land maps introduces classification uncertainty. This process relies on thresholding techniques (e.g., Otsu or zero-threshold) and spatial post-processing filters (e.g., 3×3 or 5×5 majority filters) to mitigate noise from shadows or mixed pixels ( Chen and Zhao, 2022 ; Sekertekin, 2021 ) . However, existing studies often adopt a single method without assessing sensitivity, leaving the cumulative impact of these choices on error propagation—and thus the reliability of long-term trends—unquantified ( Kirby et al., 2024 ). This gap undermines the statistical robustness of reported shrinkage rates, limiting their utility for policy and ecological management. To address this, this study applies a methodology using Sentinel-2 imagery (2018–2025) to quantify classification uncertainty and propagate it into a statistically robust trend analysis for Burdur Lake. Four combinations of two thresholding techniques (Otsu, zero-threshold) and two filtering sizes (3×3, 5×5) were systematically evaluated, validated against 300 reference points (98–99% accuracy). A Monte Carlo Simulation (MCS) framework, integrated with the non-parametric Sen’s Slope estimator, models this uncertainty (± 1% based on validation) across 1,000 iterations. Objectives are to: (1) assess classification robustness using McNemar tests and accuracy metrics; (2) quantify uncertainty propagation; (3) derive a statistically robust annual shrinkage rate; and (4) provide an evidence base linking trends to anthropogenic drivers ( Dervisoglu et al., 2022 ) for informed water policy and ecological restoration in this stressed basin. 2 - MATERIAL AND METHODS 2.1- STUDY AREA Lake Burdur, situated in southwest Turkey across the provinces of Burdur and Isparta, is a pivotal component of the Turkish Lake District. As one of the country’s largest and deepest endorheic lakes, it occupies a tectonic depression formed along a major fault system (NW–SE to E–W trending) between the Söğüt Mountains and adjacent geological structures, reaching a surface area of approximately 127 km² and a maximum depth of 110 m ( Ghorbani and Pamucar, 2025 ; Erdem et al., 2025; Atalay et al., 2020 ). Registered as a Ramsar site and Wildlife Development Area, Lake Burdur holds significant international ecological importance. Historically, it serves as a critical wintering ground for a significant portion of the globally endangered White-headed Duck ( Oxyura leucocephala ) and hosts endemic fish adapted to its brackish waters. While reed beds in the northern and southern alluvial zones enhance its biodiversity, recent decades have seen population declines driven by habitat loss ( Ataol, 2010 ) . Situated in a transition zone between Mediterranean and Continental climates, Burdur Lake experienced stable water levels until 1987, after which consistent shrinkage began, driven by increased evaporation and reduced inflows ( Davraz et al., 2019 ). The Turkish Ministry of Agriculture and Forestry’s River Basin Management Plan (RBMP) identifies six key hydraulic structures—Karaçal, Karamanlı, and Bademli dams; Beylerli, Belenli, and Tefenni ponds—as primary sources of hydromorphological pressure, impounding water on tributaries such as Bozçay and reducing inflows to Lake Burdur by an estimated 25–35% since the 1980s ( Davraz et al., 2019 ; Şener et al., 2020; T.C. TOB SYGM, 2020 ). These, along with additional smaller reservoirs (up to 16 total as of 2022), collectively retain approximately 126 hm³ annually, with substantial diversions of approximately 72.4 hm³ for irrigation exacerbating shrinkage ( Çolak et al., 2022 ; Dervisoglu et al., 2022 ; Sargın et al., 2011 ). This anthropogenic structural alteration, combined with reduced rainfall, has severely restricted the basin's hydrology, a situation further exacerbated during the 2018–2025 period under study. The synergistic effects of this structural pressure and climatic stressors underscore the urgent need for rigorous monitoring to inform effective restoration strategies for this ecologically important basin. Figure 1 illustrates the geographical context of the study area and highlighting key features. 2.2- DATA COLLECTION AND PROCESSING This study utilized the Google Earth Engine (GEE) cloud computing platform to process Sentinel-2 Level-2A (L2A) imagery, providing Bottom-of-Atmosphere (BOA) reflectance—atmospherically corrected data—at a 10 m spatial resolution. Data were sourced from the COPERNICUS/S2_SR_HARMONIZED ImageCollection, spanning January 1, 2018, to December 31, 2025, aligning with the study period. To capture Burdur Lake’s annual minimum water extent, imagery was filtered to August–September, reflecting the dry-season peak, ensuring temporal consistency across 2018–2025. A cloud cover threshold of < 10%, based on the CLOUDY_PIXEL_PERCENTAGE metadata, was applied to minimize atmospheric interference, with additional manual inspection to exclude residual haze. The Area of Interest (AOI) was defined as a rectangular bounding box (approximately 30 km × 20 km, encompassing 600 km²) manually delineated around the lake, validated against high-resolution Google Earth Pro imagery. Annual median composites were generated for each year by calculating pixel-wise median reflectance values across all qualifying images, reducing noise from clouds, shadows, and temporal variability—a necessary step for subsequent thresholding. This approach, implemented using GEE’s JavaScript API, leveraged the platform’s geometry clipping function to extract AOI-specific data, ensuring precise spatial alignment. These composites serve as the foundation for uncertainty quantification, mitigating mixed-pixel effects that impact classification accuracy, thus supporting the trend analysis with Monte Carlo simulation. The complete workflow, demonstrating the integration of preprocessing, optimal classification selection, trend estimation, and the uncertainty quantification framework, is illustrated in Fig. 2 . 2.3- NORMALIZED DIFFERENCE WATER INDEX The Normalized Difference Water Index (NDWI), originally proposed by McFeeters ( 1996 ) , was employed to delineate Burdur Lake’s water surface area from Sentinel-2 imagery, enhancing water features while suppressing soil and terrestrial vegetation signals. NDWI leverages the contrast between visible green light and reflected near-infrared (NIR) radiation, calculated as: NDWI = (Green - NIR) / (Green + NIR) (1) where Green (Band 3) and NIR (Band 8) correspond to Sentinel-2’s 10 m resolution bands ( Jumaah et al., 2023 ). This study computed NDWI annually (2018–2025) on the Google Earth Engine (GEE) platform using median composite images derived from August–September data, capturing the dry-season minimum extent. The resulting NDWI images are continuous grayscale representations with pixel values ranging from − 1 to 1, where water typically exhibits positive values, while soil and vegetation yield zero or negative values due to higher NIR reflectance (McFeeters, 1996 ). This differentiation, however, is challenged by mixed pixels and shadow effects, particularly in shallow, turbid waters like Burdur Lake, necessitating robust thresholding. The NDWI output serves as the foundation for subsequent classification, where median compositing reduces noise, enhancing reliability for uncertainty quantification via Monte Carlo simulation. This step is important for propagating classification errors into the trend analysis. 2.4- THRESHOLDING METHODS A critical step in extracting water pixels from water index images is the selection of an appropriate threshold, which significantly reduces omission and commission errors in spectral indices such as the Normalized Difference Water Index (NDWI) ( Yang et al., 2018 ; Rad et al., 2021 ). Conventional approaches often employ a zero-threshold to delineate water bodies ( Buma et al., 2018 ), exploiting this spectral distinction. However, this method may introduce inaccuracies in dynamic environments like Burdur Lake, where spatiotemporal variations in image brightness and contrast—due to salinity, turbidity, or seasonal changes—can misclassify pixels ( Kirby et al., 2024 ). To address these challenges, the Otsu algorithm provides an automated, data-driven thresholding technique, widely utilized in remote sensing ( Kulkarni et al., 2022 ; Otsu, 1975 ) . Otsu determines the optimal threshold by minimizing intra-class variance and maximizing inter-class variance within the NDWI histogram, enhancing separation between water and non-water pixels. This process involves preprocessing to ensure a bimodal histogram, followed by iterative variance analysis to identify the threshold that best partitions pixel intensities. In this study, both zero-thresholding and Otsu methods were applied to NDWI-derived median composites (2018–2025) on the Google Earth Engine (GEE) platform. The resulting binary maps classify pixels above the threshold as water (1) and below as land (0), serving as inputs for subsequent classification and uncertainty quantification. This dual approach assesses thresholding efficacy for water area estimation. 2.5- MAJORITY FILTERING An essential post-processing step in binary map generation is majority filtering, which addresses omission and commission errors arising from the salt-and-pepper effect—isolated pixel noise caused by low backscattering intensity from vegetation, shadows, and mixed pixels (Chen and Zhao, 2022 ) . This technique applies a moving window to each pixel, reassigning its class based on the most frequent value within the window, thereby smoothing spatial inconsistencies and enhancing the delineation of water bodies. Widely adopted in remote sensing, majority filtering mitigates the impact of misclassified pixels, particularly in dynamic environments like Burdur Lake, where shoreline variability complicates classification ( Sekertekin, 2021 ) . In this study, majority filtering was applied to binary maps derived from zero-threshold and Otsu thresholding of NDWI composites (2018–2025) to refine water area representations. Using the Rasterio library in Python, a single-pass moving window approach was implemented, testing 3×3 and 5×5 window sizes. The 3×3 window targets localized noise, preserving finer shoreline details, while the 5×5 window offers broader smoothing, potentially at the cost of edge accuracy. This step reduces classification noise for subsequent water area estimation. The filtered outputs, classifying pixels as water (1) or land (0) based on majority vote, provide a more reliable foundation for downstream analysis, contributing to the robust mapping of Burdur Lake’s dynamic extent. 2.6- SURFACE AREA CALCULATION Accurate quantification of Burdur Lake’s surface area is fundamental to assessing its shrinkage over 2018–2025, requiring the conversion of binary raster maps—generated from Otsu and zero-thresholding with 3×3 and 5×5 majority filters (six maps per year)—into vector polygons. This process was automated using Python geospatial libraries. To enable precise area calculations, the coordinate system was transformed from the original WGS84 to the Projected Coordinate System, Universal Transverse Mercator (UTM) Zone 36N (EPSG:32636), minimizing distortion over the study area. The resulting vector maps were exported to QGIS for further processing. Polygons representing the lake were visually identified and manually extracted based on shoreline alignment with high-resolution Google Earth Pro imagery, with erroneous polygons removed to enhance accuracy. The $ area function—a QGIS Field Calculator tool—was applied to compute surface areas in square meters, subsequently converted to square kilometers for consistency with hydrological reporting standards. These area estimates form an essential input for downstream analysis, supporting the study’s goal of rigorous quantification of lake dynamics. 2.7- ACCURACY ASSESSMENT Assessing the accuracy of binary maps generated for Burdur Lake (2018–2025) validates the thresholding and filtering processes. This study employed a stratified sampling approach, generating 275 random points using the “Random Points Inside Polygons” tool in QGIS, with a minimum distance constraint of 100 m to ensure spatial independence and mitigate autocorrelation. To address potential biases from omission and commission errors—such as misclassified pixels near shorelines—25 additional points were manually selected based on visual inspection of error patterns, validated against high-resolution Google Earth Pro images temporally matched to the Sentinel-2 acquisition dates (August–September). Pixel values were automatically extracted from binary raster maps using a Python script with the rasterio library, storing predicted and actual classifications in the point shapefile’s attribute table. Reference data were collected from Google Earth Pro imagery, cross-checked for cloud-free conditions, ensuring alignment with the study period. Accuracy was evaluated through a suite of metrics, including the Confusion Matrix, Overall Accuracy, Kappa Coefficient, F1-Score, Intersection over Union (IoU), Producer’s Accuracy, and User’s Accuracy. The sample size of 300 points was determined to cover approximately 5% of the lake’s variable shoreline, providing a representative assessment. This process quantifies classification errors, feeding into the study’s uncertainty quantification framework, and supports the reliable estimation of Burdur Lake’s shrinkage dynamics. 2.8- MCNEMAR TEST Evaluating the statistical significance of differences in classification accuracy between thresholding methods validates Burdur Lake’s water-land binary maps (2018–2025). The McNemar test, introduced by McNemar (1947) ( Rudke et al., 2021 ), was employed to compare the performance of Otsu and zero-thresholding, as well as their filtered variants (3×3 and 5×5), using the same set of 300 independent reference points. This non-parametric test facilitates 15 pairwise comparisons (Otsu/Zero, Filtered/Unfiltered) annually, addressing the underexplored cumulative impact of thresholding and filtering on classification reliability. The test constructs 2×2 confusion matrices from reference points, comparing pixel classifications between two methods. The chi-square (χ²) statistic, calculated as: $$\:{X}^{2}\:=\:\frac{{({f}_{12}\:-\:{f}_{21})}^{2}}{{f}_{12}+{f}_{21}}$$ 2 where \(\:{f}_{12}\) represents pixels correctly classified by Otsu ( \(\:{C}_{1}\) ) but incorrectly by zero-threshold ( \(\:{C}_{2}\) ), and \(\:{f}_{21}\) denotes the reverse, quantifies the difference ( Yan et al., 2006 ; Dermosinoglou and Petropoulos, 2024 ; Foody, 2004 ) . The null hypothesis ( \(\:{H}_{0}\) ) posits equal performance, rejected at a significance level of p < 0.05. Implementation utilized the statsmodels library in Python, with a custom script to automate matrix generation and χ² computation across the time series. This analysis supports the reliability of thresholding outcomes, providing a foundation for downstream mapping accuracy. By isolating method-specific differences, the McNemar test supports the study’s objective of accurate classification validation for Burdur Lake’s dynamic extent. 2.9- TREND ANALYSIS Quantifying the temporal dynamics of Burdur Lake’s surface area requires robust trend detection, addressed through nonparametric statistical methods. The Mann-Kendall test, introduced by Mann ( 1945 ) and refined by Kendall ( 1948 ) , is a widely used nonparametric approach to identify statistically significant monotonic trends—consistent upward or downward changes—without assuming data normality ( de Brito Neto et al., 2016 ; Arra et al., 2024 ). Its resilience to sudden breaks makes it suitable for heterogeneous time series, such as those affected by seasonal variability ( Arra et al., 2025 ). Complementing this, the Sen’s Slope estimator, developed independently by Theil ( 1950 ) and Sen ( 1968 ) , quantifies the magnitude of the trend by calculating the median slope of all pairwise differences between data points, effectively handling non-normal distributions and outliers. In this study, these methods were applied to annual median composites (2018–2025) derived from Otsu-thresholded NDWI images of Burdur Lake. The Mann-Kendall test assessed trend significance at a p < 0.05 threshold, preprocessing data to remove extreme outliers via a 95% quantile filter to ensure stability. Sen’s Slope estimated the median annual change rate by computing slopes across all pairwise area differences, implemented with a sliding window to capture seasonal effects. Both techniques were executed using the pymannkendall library in Python, with a custom script to automate calculations across the eight-year series. This approach enhances trend reliability, providing a foundation for subsequent uncertainty quantification via Monte Carlo simulation. 2.10- UNCERTAINTY QUANTIFICATION While the methodology thus far yields a robust point estimate for Burdur Lake’s annual shrinkage rate, it does not account for the systematic propagation of classification uncertainty into the final trend, a common challenge in remote sensing time series. These uncertainties arise primarily from difficulties in defining the water-land boundary through thresholding and spatial filtering, impacting the reliability of area estimates ( Sekertekin, 2021 ) . Addressing this gap is essential to enhance the credibility of models reliant on satellite data ( Cockx et al., 2014 ; Feizizadeh et al., 2014 ). To this end, a Monte Carlo Simulation (MCS) framework was integrated with the Sen’s Slope trend analysis, effectively addressing the non-linear complexities of polygon area calculations on binary maps without assuming error distributions ( Papadopoulos and Yeung, 2001 ; Biljecki et al., 2014 ; Feizizadeh et al., 2014 ). The MCS process began by stochastically perturbing annual surface area values from the optimal classification map (Otsu with majority filtering) over 1,000 iterations. The perturbation range of ± 1% was empirically derived from classification variability in challenging boundary pixels, validated against 300 reference points (98–99% accuracy from Google Earth Pro), ensuring robustness across plausible scenarios through dynamic probability density functions in Monte Carlo simulations ( Healey et al., 2014 ). This generated an ensemble of 1,000 synthetic time series, each reflecting a potential realization of the lake’s area history. The Sen’s Slope estimator was then applied to each series, propagating classification uncertainty into trend quantification. The final shrinkage rate was determined as the median of the 1,000 slope values, with the 95% Confidence Interval (CI) defined by the 2.5th and 97.5th percentiles of the simulated distribution, providing a quantitative measure of reliability ( McMurray et al., 2017 ; Lee et al., 2024 ). Implemented using Python with the numpy library and a fixed random seed (42) for consistency, this approach accounted for computational efficiency on the Google Earth Engine platform. This integration ensures a statistically defended trend estimate, supporting the study’s goal of rigorous shrinkage assessment. 3 - RESULTS 3.1- SURFACE AREA ESTIMATION AND VALIDATION The estimation of Burdur Lake’s annual surface area (2018–2025) was conducted using Sentinel-2 imagery from August–September, capturing the dry-season minimum extent. Otsu and zero thresholding, combined with 3×3 and 5×5 majority filtering, produced six maps per year. Otsu-thresholded areas declined from 126,757,096 m² (126.76 km²) in 2018 to 115,167,474 m² (115.17 km²) in 2025, with filtered variants showing slight variations: Otsu 3×3 from 126,757,571 m² to 115,168,184 m², and Otsu 5×5 from 126,754,255 m² to 115,166,605 m². Zero-thresholded areas ranged from 127,454,300 m² (127.45 km²) to 115,788,929 m² (115.79 km²), with similar filtered adjustments. This 11.59 km² net loss (Otsu) indicates ongoing shrinkage, influenced by filtering’s edge-smoothing effects. Validation against 300 reference points, randomly generated in QGIS and manually adjusted for omission/commission errors, used high-resolution Google Earth Pro images temporally aligned with August–September acquisitions. Otsu achieved accuracies of 98% (Kappa 0.92) in 2018, improving to 99% (Kappa 0.96) in 2025, with producer’s accuracy rising from 0.93 to 0.95 and user’s accuracy from 0.91 to 0.98, indicating consistent performance. Zero thresholding started at 93% accuracy (Kappa 0.76) in 2018, reaching 96% (Kappa 0.85) in 2025, suggesting adaptation to boundary variability. Confusion matrices for 2025 showed 257 true positives, 1 false positive, and 42 false negatives, highlighting Otsu’s precision. McNemar tests confirmed no significant differences among Otsu variants (p ≥ 0.25) or Zero variants (p ≥ 0.06), but significant differences between Otsu and Zero (p < 0.05) annually ( Van Tricht et al., 2023 ), underscoring Otsu’s reliability. The comprehensive accuracy assessment, including F1-Scores and Producer's and User's Accuracy metrics, is fully detailed in Table 1 , providing the quantitative evidence for the optimal selection of the Otsu 3x3 classification variant. Figure 3 visually substantiates these findings by charting the superior and stable performance of the Otsu-thresholding variants, particularly emphasizing the critical distinction in Overall Accuracy and Kappa Coefficient when compared against the conventional zero-thresholding method across the entire time series. Table 1 Summary of the classification accuracy metrics and annual McNemar test results for 2025. The comprehensive accuracy metrics consistently demonstrate the superiority of the Otsu variants (OA \(\:\ge\:\) 0.99, \(\:K\:\ge\:\) 0.96). The McNemar test confirms that Otsu-based methods are statistically different from Zero-thresholding (p < 0.004 vs. No Filter), justifying the optimal selection of the Otsu 3x3 variant and validating the use of a low \(\:\pm\:1\%\:\) error margin in the subsequent Monte Carlo Simulation. Metric / Method Otsu (No Filter) Otsu 3x3 (Optimal) Otsu 5x5 Zero (No Filter) Zero 3x3 Zero 5x5 Overall Accuracy (OA) 0.99 0.99 0.99 0.96 0.98 0.98 Kappa Coefficient (K) 0.96 0.96 0.96 0.85 0.91 0.91 F1-Score 0.96 0.96 0.96 0.87 0.92 0.92 Producer's Accuracy 0.95 0.95 0.95 0.95 0.95 0.95 User's Accuracy 0.98 0.98 0.98 0.80 0.89 0.89 McNemar p-value vs. Otsu 3x3 \(\:\ge\:\) 1.00 - \(\:\ge\:\) 1.00 < 0.004 0.125 0.125 3.2- TREND ANALYSIS Assessing the temporal dynamics of Burdur Lake’s surface area (2018–2025) is essential to quantify its shrinkage, utilizing the Otsu 3×3 filtered time series derived from Sentinel-2 imagery. The Mann-Kendall (MK) test, a nonparametric method, was initially applied to detect statistically significant monotonic trends—consistent decreases over time—without assuming data normality. The test yielded a Z value of -3.45 (p = 0.0008, α = 0.05), confirming a significant downward trend. Subsequently, the Sen’s Slope estimator quantified the trend magnitude by calculating the median slope of all pairwise area differences, resulting in a rate of -1,638,994.40 m²/year (-1.64 km²/year), with an intercept of 125,926,857.90 m². This corresponds to a total reduction of 11,591,622 m² (11.59 km²) over the study period, reflecting a 9.2% decline from the 2018 baseline. The analysis leveraged the pymannkendall library in Python, with preprocessing to remove outliers via a 95% quantile filter, ensuring trend stability. This point estimate, while robust, captures only the central tendency, with variability across thresholding variants (e.g., Otsu 5×5 at -1.63 km²/year) suggesting method sensitivity. The decline aligns with historical trends ( Davraz et al., 2019 ), likely driven by anthropogenic pressures, underscoring the need for ecological intervention. However, the final, statistically defended confidence limits, accounting for classification uncertainty, are detailed in the Uncertainty Quantification section (3.3), enhancing the reliability of this trend for Burdur Lake’s management. To document the precise input for the trend estimation, the annual surface area values derived from the optimal Otsu 3x3 classification are itemized in Table 2 . Table 2 Annual Surface Area Time Series for Burdur Lake (2018–2025). The annual surface area, derived from the Otsu 3x3 optimal classification, serves as the foundational data for the Mann-Kendall and Sen's Slope trend analysis. The documented area decline from 126.76 km² in 2018 to 115.17 km² in 2025 demonstrates a net loss of 11.59 km² over the eight-year period. Year Area (m²) (Otsu 3x3) Area (km²) (Otsu 3x3) 2018 126.757,571 126.76 2019 124.903,447 124.90 2020 123.265,115 123.27 2021 120.958,300 120.96 2022 119.422,455 119.42 2023 118.263,205 118.26 2024 116.708,843 116.71 2025 115.168,184 115.17 3.3- UNCERTAINTY QUANTIFICATION The Sen’s Slope-derived point estimate of -1.64 km²/year for Burdur Lake’s shrinkage (2018–2025) provides a robust trend measure, yet it overlooks the systematic propagation of classification uncertainty inherent in remote sensing time series. These uncertainties, primarily arising from challenges in defining the water-land boundary through thresholding and spatial filtering ( Sekertekin, 2021 ) , undermine trend reliability, necessitating rigorous quantification to enhance model credibility ( Cockx et al., 2014 ; Feizizadeh et al., 2014 ). The MCS framework (detailed in Section 2.10 ) was applied to propagate classification uncertainty, yielding a median slope of -1.64 km²/year with a 95% CI of -1.83 to -1.47 km²/year. The negative CI indicates a decline, validated by the methodology’s robustness. These findings provide a basis for water management strategies. Table 3 provides a consolidated, statistically defended summary of the key trend metrics, encompassing the Mann-Kendall test results and the final Confidence Interval derived from the 1,000-iteration Monte Carlo Simulation. Figure 4 graphically validates the robustness of this approach, illustrating the median annual lake area and the derived 95% Confidence Interval (CI) that rigorously quantifies the propagated classification uncertainty across the entire 2018–2025 time series. Furthermore, the core output of the Monte Carlo Simulation is visualized in Fig. 5 , which presents the density distribution of the 1,000 resulting Sen's Slope values, quantifying the uncertainty of the annual shrinkage rate itself. Table 3 Summary of Trend Analysis and 95% Uncertainty Quantification (Otsu 3x3, 2018–2025). The results confirm a statistically highly significant monotonic downward trend (p = 0.0008). The 95% Confidence Interval (CI), derived from the Monte Carlo Simulation, is entirely negative and tightly constrained, providing high confidence that the annual shrinkage rate lies between − 1.83 and − 1.47 km²/year. Metric Value Units Median Sen’s Slope -1.64 km²/year Total Area Reduction 11.59 km² Mann-Kendall Z Statistic -3.45 Dimensionless Mann-Kendall p-value 0.0008 Dimensionless 95% CI (Lower Bound) -1.83 km²/year 95% CI (Upper Bound) -1.47 km²/year 4 - DISCUSSION 4.1- Interpretation of Trends and Methodological Insights The quantified shrinkage indicates a total loss of 11.59 km² during the study period and a persistent deficit in the lake's water budget. When placed in historical context, (37% reduction from 1975 to 2016; Davraz et al., 2019 ), this finding is consistent. The continuing high-magnitude shrinkage strongly suggests that anthropogenic factors remain the dominant drivers, overshadowing the contribution of climatic variability ( Dervisoglu et al., 2022 ). The persistence of this decline aligns with prior research attributing the water deficit primarily to streamflow diversion and groundwater pumping ( Dervisoglu et al., 2022 ). The efficacy of human intervention is quantitatively supported by numerical models, such as MODFLOW. These models demonstrate that water levels can be engineered to rebound significantly by prioritizing the release of surface water flows, even despite the negative impacts of climate change (e.g., increased evaporation). This confirms that policy and management decisions hold the controlling leverage over the lake’s fate ( Kılıç Germeç, 2023 ) . Methodologically, this study offers insights for remote sensing time series. The systematic comparison between classification methods demonstrated the superior performance and stability of the Otsu thresholding algorithm over the common zero-thresholding approach, a difference that the McNemar test confirmed was statistically significant. The Otsu method's adaptive, data-driven approach minimized intra-class variance supported consistent delineation (98–99% accuracy). 4.2- Uncertainty and Validation Considerations The transition from continuous spectral index imagery to binary water/land maps is the primary source of epistemic uncertainty in remotely sensed water resource monitoring, particularly in complex, shallow endorheic basins like Burdur Lake. The validation protocol in this study systematically quantified this uncertainty, demonstrating that the choice of classification methodology critically impacts the resulting area estimates. The comparison against 300 reference points confirmed the superior performance of the Otsu algorithm, which achieved high overall accuracies (ranging from 98% in 2018 to 99% in 2025) and strong agreement (Kappa coefficients between 0.92 and 0.96). Crucially, the McNemar test provided the necessary statistical rigor, showing a significant annual difference between the Otsu method and the conventional zero-thresholding approach (p < 0.05). While the 300-point validation dataset, adjusted for omission and commission errors, achieved high accuracy (98–99%), it may not fully capture the spatial heterogeneity of Burdur Lake’s dynamic shoreline, particularly in shallow, turbid regions. Additionally, despite applying a < 10% cloud cover threshold, residual atmospheric interference (e.g., haze) in some Sentinel-2 images could introduce minor classification errors, especially for mixed pixels near the water-land boundary. These pixels, influenced by salinity and seasonal variability, may affect area estimates, though median compositing mitigates this to an extent. Temporal variability in image acquisition dates within the August–September window could also contribute to subtle inconsistencies, as water levels fluctuate slightly within the dry season. Future studies could address these by incorporating multi-sensor data (e.g., Sentinel-1 for cloud-penetrating capabilities) or machine learning-based classification to enhance boundary delineation. These uncertainties, while minor, underscore the need for cautious interpretation of the reported shrinkage rate and highlight opportunities for methodological refinement. Despite these minor uncertainties, the Monte Carlo Simulation framework, integrated with the Sen’s Slope estimator, provides an estimate of Burdur Lake’s shrinkage rate (-1.64 km²/year, 95% CI: -1.83 to -1.47 km²/year), accounting for classification errors. This trend, validated at 98–99% accuracy, offers a baseline for understanding the lake’s hydrological decline. By quantifying the area loss, this study offers a foundation for informing water resource management strategies, particularly in addressing the anthropogenic drivers exacerbating the crisis. 4.3- Implications for Water Resource Management The shrinkage rate of -1.64 km²/year indicates a water resource issue in the Burdur Closed Basin (BCB). This decline confirms that anthropogenic activities, such as irrigation diversions and groundwater abstraction, dominate over climatic stressors ( Dervisoglu et al., 2022 ). Consequently, the lake's ecological fate hinges on fundamental, basin-wide policy restructuring, not on mitigating external climate change alone. Firstly, this finding serves as a timely metric for policy intervention. Previous MODFLOW groundwater model studies established that, while excessive pumping and climate change forecast a 7 m decline over 46 years, strategically releasing surface water flows could force a rebound of up to 3 m ( Kılıç Germeç, 2023 ) . This confirms that human management holds the controlling leverage. The observed persistent shrinkage in this study indicates that this leverage has not been utilized, necessitating immediate action to reverse reservoir operational policies that divert nearly all natural surface flows. Policy adjustment is supported by studies estimating that mandatory conversion to efficient irrigation methods (e.g., pressurized systems) could conserve 62.6 hm³ of water annually ( Sargın et al., 2011 ). This potential saving matches the estimated annual water loss rate for the lake (approximately 40 hm³) ( Ataol, 2010 ) , demonstrating that the crisis is solvable through policy-driven water use optimization. Secondly, the management imperative extends to groundwater abstraction. The drilling of numerous boreholes has severely depleted the regional aquifer, eliminating the essential baseflow contribution to the lake ( Ataol, 2010 ) . The combination of this abstraction, regional temperature projections (a rise of at least 2°C by 2100) ( Çolak et al., 2022 ), and increased irrigation demand is projected to exacerbate the drying process unless groundwater extraction rates are effectively regulated. Resource management must implement strict regulatory enforcement (MoEU, 2020) to allow the aquifer system to recover and re-establish a hydraulic connection with the lake. Finally, the decline affects the lake's status as an internationally significant Ramsar site, which is the most important wintering site for the critically endangered White-headed Duck ( Oxyura leucocephala ) (Doğa Derneği, n.d.) . The substantial area loss confirmed here indicates a continuing collapse of critical shallow littoral habitats and accelerates salinization ( Çolak et al., 2022 ). Failure to adopt a "dynamic lake management plan" prioritizing ecosystem needs will contribute to local climate changes—increasing continentality and public health hazards (AECOM, 2019) from exposed, dusty lakebed sediments. The shrinkage rate suggests a shift in the current resource paradigm from exploitation to restoration. 4.4- Future Research Directions This shrinkage trend provides a metric for immediate policy action and a foundation for critical, subsequent research. Future studies must build upon the methodological advancements presented here to tackle the outstanding uncertainties in hydrological attribution, policy effectiveness, and socio-economic dynamics. Methodologically, the robust Monte Carlo Simulation (MCS) uncertainty quantification framework successfully applied here to surface area should be extended to the third dimension. Future research could integrate high-resolution, multi-source altimetry data ( Hou et al., 2024 ) (e.g., Sentinel-3, ICESat-2) and historical Digital Elevation Models (DEMs) to derive annual rate of volume loss and water level change. This 3D approach, incorporating elevation error propagation, supports understanding of the lake's desiccation that current 2D area studies lack. Hydrologically, while this research confirms the dominance of anthropogenic factors, the quantitative attribution of future climate stress requires refinement. Studies utilizing the MODFLOW numerical model have established the potential for management interventions to generate a multi-meter rebound (+ 3 m) even under projected climate decline. Future work should prioritize refining the long-term water budget by integrating high-resolution Regional Climate Models (CORDEX RCP scenarios) with empirical data on regional thermal stress (e.g., the documented 2.2°C Lake Surface Water Temperature increase) to accurately partition the contributions of increasing evaporative output and decreased natural inflow under long-term drought conditions ( Kılıç Germeç, 2023 ) . Finally, future research must transition from documenting decline to evaluating the efficacy of proposed mitigation strategies. Given the potential for optimized irrigation to save water exceeding the lake's annual deficit ( Ataol, 2010 ) , a crucial research direction is the development of socio-economic models. These models must assess the economic costs, policy feasibility, and financial incentives necessary for a mandatory basin-wide conversion to pressurized irrigation systems, dynamically linking Land Use/Land Cover (LULC) change, water abstraction, and local policy implementation. Furthermore, high-temporal-resolution ecological monitoring remains necessary (MoEF, 2007) to quantify the rate of biodiversity loss—especially the collapse of essential shallow habitats—and inform the designation of newly critical preservation zones. 5 - CONCLUSION This study developed and implemented a robust methodology for quantifying Burdur Lake’s surface area decline using a high-resolution Sentinel-2 time series (2018–2025). Sen’s Slope estimator was integrated with an MCS framework to account for classification uncertainty, yielding a robust shrinkage rate of -1.64 km²/year, totaling 11.59 km², consistent with the historical trajectory (37% reduction, 1975–2016). The 95% Confidence Interval (CI) of -1.83 to -1.47 km²/year, derived from 1,000 MCS iterations, validates sustained area loss, driven by anthropogenic factors. This decline indicates a water budget deficit, affecting the lake’s Ramsar status and its endangered White-headed Duck population. It provides a basis for policy intervention, such as revised reservoir management and stringent groundwater controls to restore hydrological balance. Future research should extend this MCS framework by integrating multi-source altimetry (e.g., Sentinel-3, ICESat-2) with Digital Elevation Models to develop a 3D Uncertainty Quantification model, quantifying volume loss and water level changes. Additionally, coupling these physical insights with socio-economic analyses of irrigation efficiency will enhance policy feasibility, supporting sustainable restoration and ecological recovery. Declarations ORCID 0009-0009-4974-0206 Author Contribution The author confirms sole responsibility for the following: conceptualization and design of the study; data acquisition, preprocessing, analysis; interpretation of results; drafting and critical revision of the manuscript; and final submission approval of the published version. Funding Declaration The author declares that no external funding was received from any external source or organization for the research, authorship, and/or publication of this article. Competing Interests The author declares no competing interests. Data Availability Data will be made available on request. References AECOM. (2019), February 15 Consequences of drying lake systems around the world (Summary of the February 15, 2019 report prepared for the Great Salt Lake Advisory Council). Great Salt Lake Advisory Council Al Garni HZ, Awasthi A (2020) A Monte Carlo approach applied to sensitivity analysis of criteria impacts on solar PV site selection. Handbook of Probabilistic Models. Butterworth-Heinemann, pp 489–504 Arra AA, Alashan S, Şişman E (2024) Trends of meteorological and hydrological droughts and associated parameters using innovative approaches. J Hydrol 640:131661 Arra AA, Keskin MZ, Şişman E (2025) Trend Analysis of Hydro-Meteorological Variables Using Mann-Kendall and Sen's Slope with Standardization (SSS): Case Study of the Kızılırmak Catchment, Türkiye. Phys Chem Earth Parts A/B/C, 104115 Atalay İ, Altunbaş S, Siler M (2020) The Formation and Evaluation of the Faulted Topography in the Burdur Basin, Lakes Region, SW Anatolia. J Geogr, (41), 41 Ataol M (2010) Burdur Gölü’nde seviye değişimleri. Coğrafi Bilimler Dergisi 8(1):77–92 Biljecki F, Ledoux H, Stoter J (2014) Error propagation in the computation of volumes in 3D city models with the Monte Carlo method. ISPRS Annals Photogrammetry Remote Sens Spat Inform Sci 2:31–39 Buma WG, Lee SI, Seo JY (2018) Recent surface water extent of lake Chad from multispectral sensors and GRACE. Sensors , 18 (7), 2082 Che L, Li S, Liu X (2025) Improved surface water mapping using satellite remote sensing imagery based on optimization of the Otsu threshold and effective selection of remote-sensing water index. J Hydrol 654:132771 Chen Z, Zhao S (2022) Automatic monitoring of surface water dynamics using Sentinel-1 and Sentinel-2 data with Google Earth Engine. Int J Appl Earth Obs Geoinf 113:103010 Cockx K, Van de Voorde T, Canters F (2014) Quantifying uncertainty in remote sensing-based urban land-use mapping. Int J Appl Earth Obs Geoinf 31:154–166 Crosetto M, Tarantola S, Saltelli A (2000) Sensitivity and uncertainty analysis in spatial modelling based on GIS. Agric Ecosyst Environ 81(1):71–79 Çolak MA, Öztaş B, Özgencil İK, Soyluer M, Korkmaz M, Ramírez-García A, Akyürek Z (2022) Increased water abstraction and climate change have substantial effect on morphometry, salinity, and biotic communities in lakes: examples from the semi-arid burdur Basin (Turkey). Water 14(8):1241 Davraz A, Sener E, Sener S (2019) Evaluation of climate and human effects on the hydrology and water quality of Burdur Lake, Turkey. J Afr Earth Sc 158:103569 de Brito Neto RT, Santos CA, Mulligan K, Barbato L (2016) Spatial and temporal water-level variations in the Texas portion of the Ogallala Aquifer. Nat Hazards 80(1):351–365 Dermosinoglou A, Petropoulos GP (2024) Exploring long term Impervious Surface Areas (ISA) dynamics using Landsat imagery, Μachine Learning and GEE: The case of Attica, Greece. Remote Sens Applications: Soc Environ 36:101338 Dervisoglu A, Yağmur N, Fıratlı E, Musaoğlu N, Tanık A (2022) Spatio-temporal assessment of the shrinking Lake Burdur, Turkey. Int J Environ Geoinformatics 9(2):169–176 Doğa Derneği (n.d.). Burdur lake . https://dogadernegi.org/en/burdur-lake/ Du Y, Zhang Y, Ling F, Wang Q, Li W, Li X (2016) Water bodies’ mapping from Sentinel-2 imagery with modified normalized difference water index at 10-m spatial resolution produced by sharpening the SWIR band. Remote Sens 8(4):354 Erdem KC, Bakırman T, Bayram B Temporal Dynamics of Lake Burdur's Water Surface Area: A Two-Decade Remote Sensing Analysis and Future Forecasts. Mersin Photogrammetry J, 7 (1), 22–28 Feizizadeh B, Jankowski P, Blaschke T (2014) A GIS based spatially-explicit sensitivity and uncertainty analysis approach for multi-criteria decision analysis. Comput Geosci 64:81–95 Foody GM (2004) Thematic map comparison. Photogrammetric Eng Remote Sens 70(5):627–633 Ghorbani S, Pamucar D (2025) Remote Sensing-Based Evaluation of Lake Area Dynamics: A Quantitative Assessment for Environmental Management in Turkey. Spectr Oper Res 3(1):352–358 Healey SP, Urbanski SP, Patterson PL, Garrard C (2014) A framework for simulating map error in ecosystem models. Remote Sens Environ 150:207–217 Hong S, Heo J, Vonderohe AP (2013) Simulation-based approach for uncertainty assessment: Integrating GPS and GIS. Transp Res Part C: Emerg Technol 36:125–137 Hou J, Van Dijk AI, Renzullo LJ, Larraondo PR (2024) GloLakes: water storage dynamics for 27 000 lakes globally from 1984 to present derived from satellite altimetry and optical imaging. Earth Syst Sci Data 16(1):201–218 Jumaah HJ, Ameen MH, Kalantar B (2023) Surface water changes and water depletion of Lake Hamrin, Eastern Iraq, using Sentinel-2 images and geographic information systems. Adv Environ Eng Res 4(1):1–11 Kendall MG (1948) Rank correlation methods Kılıç Germeç H (2023) Assessment of the impacts of future climatic variations and anthropogenic activities on Burdur Lake levels Kirby K, Ferguson S, Rennie CD, Cousineau J, Nistor I (2024) Identification of the best method for detecting surface water in Sentinel-2 multispectral satellite imagery. Remote Sens Applications: Soc Environ 36:101367 Kulkarni R, Khare K, Khanum H (2022) Detecting, extracting, and mapping of inland surface water using Landsat 8 Operational Land Imager: A case study of Pune district, India. F1000Research , 11 , 774 Lee S, Moon S, Kim K, Sung S, Hong Y, Lim W, Park SK (2024) A comparison of green, delta, and Monte Carlo methods to select an optimal approach for calculating the 95% confidence interval of the Population-attributable fraction: guidance for epidemiological research. J Prev Med Public Health 57(5):499 Mann HB (1945) Nonparametric tests against trend. Econometrica: J econometric Soc, 245–259 McFeeters SK (1996) The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features. Int J Remote Sens 17(7):1425–1432 McMurray A, Pearson T, Casarim F (2017) Guidance on applying the Monte Carlo approach to uncertainty analyses in forestry and greenhouse gas accounting. Winrock International: Arlington, VA, USA , 26 Otsu N (1975) A threshold selection method from gray-level histograms. Automatica 11(285–296):23–27 Papadopoulos CE, Yeung H (2001) Uncertainty estimation and Monte Carlo simulation method. Flow Meas Instrum 12(4):291–298 Rad AM, Kreitler J, Sadegh M (2021) Augmented Normalized Difference Water Index for improved surface water monitoring. Environ Model Softw 140:105030 Republic of Turkey, Ministry of Environment and Forestry, General Directorate of Nature Conservation and National Parks, Department of Nature Conservation (2007) The national biological diversity strategy and action plan Republic of Turkey Ministry of Environment and Urbanization, General Directorate of Environmental Impact Assessment, Permit and Inspection. (2020). 6th state of environment report for Republic of Turkey (Publication No. 48/2) Rudke AP, Xavier ACF, Fujita T, Abou Rafee SA, Martins LD, Morais MVB, Martins JA (2021) Mapping past landscapes using landsat data: Upper Paraná River Basin in 1985. Remote Sens Applications: Soc Environ 21:100436 Sargın A, Taşkıran F, Yılmaz E, Sönmez Y, Yeniyurt C (2011) No lake, no Burdur! Doğa Derneği. Sarp G, Ozcelik M (2017) Water body extraction and change detection using time series: A case study of Lake Burdur, Turkey. J Taibah Univ Sci 11(3):381–391 Sekertekin A (2021) A Survey on Global Thresholding Methods for Mapping Open Water Body Using Sentinel-2 Satellite Imagery and Normalized Difference Water Index. Arch Comput Methods Eng, 28 (3) Sen PK (1968) Estimates of the regression coefficient based on Kendall's tau. J Am Stat Assoc 63(324):1379–1389 Şehnaz Ş, Şener E, Davraz A, Varol S (2020) Hydrogeological and hydrochemical investigation in the Burdur Saline Lake Basin, southwest Turkey. Geochemistry 80(4):125592 T.C. Tarım ve Orman Bakanlığı, Su Yönetimi Genel Müdürlüğü. (2020, April). Burdur Basin river basin management plan final report Annex 1–6 Tesfaye M, Breuer L (2024) Performance of water indices for large-scale water resources monitoring using Sentinel-2 data in Ethiopia. Environ Monit Assess 196(5):467 Theil H (1950) A rank-invariant method of linear and polynomial regression analysis. Indagationes Math 12(85):173 Van Tricht K, Degerickx J, Gilliams S, Zanaga D, Battude M, Grosu A, Szantoi Z (2023) WorldCereal: a dynamic open-source system for global-scale, seasonal, and reproducible crop and irrigation mapping. Earth Syst Sci Data 15(12):5491–5515 Yan G, Mas JF, Maathuis BHP, Xiangmin Z, Van Dijk PM (2006) Comparison of pixel-based and object‐oriented image classification approaches—a case study in a coal fire area, Wuda, Inner Mongolia, China. Int J Remote Sens 27(18):4039–4055 Yang X, Qin Q, Grussenmeyer P, Koehl M (2018) Urban surface water body detection with suppressed built-up noise based on water indices from Sentinel-2 MSI imagery. Remote Sens Environ 219:259–270 Additional Declarations The authors declare no competing interests. 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-8269709","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":554641360,"identity":"b013c1ef-5e45-4187-b111-42ca2db4f6b3","order_by":0,"name":"Tarik Emre Yorulmaz","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3UlEQVRIie3RMQuCQBTA8SeCLaLruVQf4SKIhsivchI0OYfQUCK4uee3cHJ+EXjLVatjETTnN+islpbLseH+8IbH4wcHB6DT/WsXOa6bIL427EKYHG9XsbdBMLadCMWQQidCeXYnQTj3AUVzaFLoOzXr3SIVEXxCgnJhGklW4D6FsVczIxYqUi8tSUzLhOOLBEVLVC/7kI1tQXhpyaYrORAbQmgJo7+IJypzGpScElJRFCcyysU1zlXE4alRN+Xa98/J7RGtZgOHL7BRkSF+70TOj58cqM86nU6nkz0BcG1YrcfFx9UAAAAASUVORK5CYII=","orcid":"https://orcid.org/0009-0009-4974-0206","institution":"Istanbul Technical University","correspondingAuthor":true,"prefix":"","firstName":"Tarik","middleName":"Emre","lastName":"Yorulmaz","suffix":""}],"badges":[],"createdAt":"2025-12-03 11:33:58","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8269709/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8269709/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":97667389,"identity":"a13da167-8eb5-4097-842e-7b5f3e3ea399","added_by":"auto","created_at":"2025-12-08 09:23:22","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3799745,"visible":true,"origin":"","legend":"","description":"","filename":"PAPERDRAFT.docx","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/fa56c72ca10cfaec0758b288.docx"},{"id":97668072,"identity":"57d5451c-80a3-4015-b6ae-673f7c7dc144","added_by":"auto","created_at":"2025-12-08 09:24:46","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":342,"visible":true,"origin":"","legend":"","description":"","filename":"rs8269709.json","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/88f6448f98cef5bcaec4a0b8.json"},{"id":97422924,"identity":"e82de9db-dac4-4b44-b051-0b1a251462cc","added_by":"auto","created_at":"2025-12-04 08:47:10","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":130116,"visible":true,"origin":"","legend":"","description":"","filename":"rs82697090enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/0b95dabfa90203a142be9840.xml"},{"id":97666842,"identity":"cbc9ce1d-226e-408e-bf6b-cd40ae7b1cef","added_by":"auto","created_at":"2025-12-08 09:22:13","extension":"jpeg","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":94020,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/79546d8abcc1ec0e64751982.jpeg"},{"id":97667096,"identity":"7f732c23-58fc-4127-9f9e-43a7a24d2fb4","added_by":"auto","created_at":"2025-12-08 09:22:43","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":116459,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/7cfc3a8ff39570bda2628030.png"},{"id":97422929,"identity":"59a08582-19e1-437d-a420-eaacdabd1b42","added_by":"auto","created_at":"2025-12-04 08:47:10","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":83296,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/9fa2f6be689956a1095a45a3.jpeg"},{"id":97667778,"identity":"75cde084-7fec-4461-9176-b00ebe19a31b","added_by":"auto","created_at":"2025-12-08 09:24:15","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":47002,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/1cb516ee251392922dcf23c1.jpeg"},{"id":97667790,"identity":"f536329a-b04e-450b-8bde-91005c04056e","added_by":"auto","created_at":"2025-12-08 09:24:16","extension":"jpeg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":41822,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/08c30a533c6e90cba5508b44.jpeg"},{"id":97666812,"identity":"c75b1543-9468-460f-b9bc-491cac771491","added_by":"auto","created_at":"2025-12-08 09:22:11","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82727,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/e14e24a9c9bbc837c667e816.png"},{"id":97667121,"identity":"887be2b4-2278-45ef-b994-e9c299cb1b47","added_by":"auto","created_at":"2025-12-08 09:22:51","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":31154,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/70ee7b440349558243ca279a.png"},{"id":97666411,"identity":"8a826c89-c10d-47f3-a06c-a97d8c28720c","added_by":"auto","created_at":"2025-12-08 09:21:09","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39652,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/c1eb9cb79490c701dd7bc497.png"},{"id":97422934,"identity":"8859d54a-2484-47ed-a9b9-ec7624b6716c","added_by":"auto","created_at":"2025-12-04 08:47:11","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":28268,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/65ebba1c2c109c3d076917d7.png"},{"id":97422937,"identity":"7970e197-6bc7-485a-a36d-1493814bbb05","added_by":"auto","created_at":"2025-12-04 08:47:11","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":25237,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/98f090cfccd20a112445536e.png"},{"id":97667663,"identity":"93b17878-2bad-4503-9a34-cda4be150211","added_by":"auto","created_at":"2025-12-08 09:24:03","extension":"xml","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":128345,"visible":true,"origin":"","legend":"","description":"","filename":"rs82697090structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/e4ab3fdfe21ca77fdf1c551c.xml"},{"id":97422939,"identity":"e3deb0e4-9e73-4ac3-b6d3-d31e5544bb6d","added_by":"auto","created_at":"2025-12-04 08:47:11","extension":"html","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":136064,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/313f9927b33d2e538378582c.html"},{"id":97422921,"identity":"0b86c5a3-f766-437b-b166-e9b83488a6cc","added_by":"auto","created_at":"2025-12-04 08:47:10","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":142739,"visible":true,"origin":"","legend":"\u003cp\u003eStudy area map of Lake Burdur, southwest Turkey, depicting its location within the Burdur Closed Basin, and six primary hydraulic structures\u003c/p\u003e","description":"","filename":"figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/6d4140be4cfae99073b00487.jpg"},{"id":97666342,"identity":"d941eac0-df93-480d-a4e0-81700e79a791","added_by":"auto","created_at":"2025-12-08 09:21:02","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":69010,"visible":true,"origin":"","legend":"\u003cp\u003eIntegrated Methodology for Trend Analysis and Uncertainty Quantification. The schematic illustrates the two-stage analytical approach.\u003c/p\u003e","description":"","filename":"figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/3bc8959e6a8accf8687cbe1c.png"},{"id":97422928,"identity":"2a8f2228-1927-40e5-a362-a068ebd12bdc","added_by":"auto","created_at":"2025-12-04 08:47:10","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":137365,"visible":true,"origin":"","legend":"\u003cp\u003eComparative Classification Accuracy Metrics for Burdur Lake (2018–2025). (A) Overall Accuracy (OA) and (B) Kappa Coefficient (\u003cimg height=\"20\" src=\"data:image/png;base64,R0lGODlhDQAZAHcAMSH+GlNvZnR3YXJlOiBNaWNyb3NvZnQgT2ZmaWNlACH5BAEAAAAALAAABgANAAwAhAAAAAAAAAAAOgAAZgA6kABmtjoAADpmtjqQ22YAAGY6Oma222a2/5A6AJC225Db/7ZmALZmOrb//9uQOtu2Ztv///+2Zv/bkP/btv//tv//2wECAwECAwECAwECAwECAwVGIABYQcmMZXCKU7CIWiM4IjtILFHVQFz0kcOOl0kgMIoXT0Qq3ZY2CUUFBUB0Rd2yiGAJHsuLYcVdQqgiyJPlxIkDBBwgBAA7\" width=\"10\"/\u003e) are shown for four core classification methods.\u003c/p\u003e","description":"","filename":"figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/993b1a509e6caa35b5681d6a.png"},{"id":97666450,"identity":"4cd187d8-bbe0-4e39-9ba8-93c01dd909a5","added_by":"auto","created_at":"2025-12-08 09:21:15","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":69158,"visible":true,"origin":"","legend":"\u003cp\u003eAnnual Surface Area Time Series of Burdur Lake (2018–2025) with 95% Confidence Intervals (CI).\u003c/p\u003e","description":"","filename":"figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/c1e6a8a5417f3e1e6c8f3dd1.png"},{"id":97422926,"identity":"9aed8dbc-b2a7-40bb-8759-f86077dc468d","added_by":"auto","created_at":"2025-12-04 08:47:10","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":62628,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of Annual Shrinkage Rates from Monte Carlo Simulation (2018–2025).\u003c/p\u003e","description":"","filename":"figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/e7f19462804dae787e0f2824.png"},{"id":97677523,"identity":"7bdb0ece-2f07-465d-bc96-b7d260da2d03","added_by":"auto","created_at":"2025-12-08 09:53:23","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1241221,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8269709/v1/0b937f95-0549-42e6-9c61-25e8f96c3767.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eQuantifying Burdur Lake Shrinkage (2018–2025): Trend Analysis and Uncertainty Quantification with Sentinel-2 Imagery and Monte Carlo\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"1 - INTRODUCTION","content":"\u003cp\u003eSurface water bodies, such as lakes and reservoirs, are fundamental to the Earth\u0026rsquo;s hydrologic cycle, global climate regulation, and biodiversity support \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eDu et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). These ecosystems are increasingly threatened by anthropogenic pressures\u0026mdash;including irrigation demands, dam construction, and groundwater extraction\u0026mdash;and climate change-induced evaporation, necessitating accurate, scalable monitoring for sustainable water resource management \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eTesfaye and Breuer, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2024\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e. Remote sensing, with its cost-effective, large-scale capabilities, surpasses traditional in-situ methods, providing dynamic data essential for tracking environmental change \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eTesfaye and Breuer, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Che et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Among vulnerable regions, Mediterranean endorheic lakes are particularly at risk, with Burdur Lake in southwest Turkey\u0026mdash;a Ramsar site of international ecological significance\u0026mdash;exemplifying this trend. Historical analyses have documented severe shrinkage, with a 37% surface area loss from 210 km\u0026sup2; (1975) to 131 km\u0026sup2; (2016) \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eDavraz et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), and recent studies report a 1.47% decline between 2000 and 2024 \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eGhorbani and Pamucar, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2025\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e and a one-fifth reduction from 1987 to 2000 \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eSarp and Ozcelik, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2017\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e. These changes are predominantly attributed to upstream water extraction for agriculture via dams and groundwater pumping \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eDervisoglu et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), trends likely persisting into the 2018\u0026ndash;2025 period.\u003c/p\u003e\u003cp\u003eDespite these insights, a critical methodological challenge persists in remote sensing of lake dynamics. The transition from continuous spectral indices, such as the Normalized Difference Water Index (NDWI), to binary water/land maps introduces classification uncertainty. This process relies on thresholding techniques (e.g., Otsu or zero-threshold) and spatial post-processing filters (e.g., 3\u0026times;3 or 5\u0026times;5 majority filters) to mitigate noise from shadows or mixed pixels \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eChen and Zhao, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Sekertekin, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2021\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e. However, existing studies often adopt a single method without assessing sensitivity, leaving the cumulative impact of these choices on error propagation\u0026mdash;and thus the reliability of long-term trends\u0026mdash;unquantified \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eKirby et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). This gap undermines the statistical robustness of reported shrinkage rates, limiting their utility for policy and ecological management.\u003c/p\u003e\u003cp\u003eTo address this, this study applies a methodology using Sentinel-2 imagery (2018\u0026ndash;2025) to quantify classification uncertainty and propagate it into a statistically robust trend analysis for Burdur Lake. Four combinations of two thresholding techniques (Otsu, zero-threshold) and two filtering sizes (3\u0026times;3, 5\u0026times;5) were systematically evaluated, validated against 300 reference points (98\u0026ndash;99% accuracy). A Monte Carlo Simulation (MCS) framework, integrated with the non-parametric Sen\u0026rsquo;s Slope estimator, models this uncertainty (\u0026plusmn;\u0026thinsp;1% based on validation) across 1,000 iterations. Objectives are to: (1) assess classification robustness using McNemar tests and accuracy metrics; (2) quantify uncertainty propagation; (3) derive a statistically robust annual shrinkage rate; and (4) provide an evidence base linking trends to anthropogenic drivers \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eDervisoglu et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) for informed water policy and ecological restoration in this stressed basin.\u003c/p\u003e"},{"header":"2 - MATERIAL AND METHODS","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1- STUDY AREA\u003c/h2\u003e\u003cp\u003eLake Burdur, situated in southwest Turkey across the provinces of Burdur and Isparta, is a pivotal component of the Turkish Lake District. As one of the country\u0026rsquo;s largest and deepest endorheic lakes, it occupies a tectonic depression formed along a major fault system (NW\u0026ndash;SE to E\u0026ndash;W trending) between the S\u0026ouml;ğ\u0026uuml;t Mountains and adjacent geological structures, reaching a surface area of approximately 127 km\u0026sup2; and a maximum depth of 110 m \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eGhorbani and Pamucar, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eErdem et al., 2025;\u003c/span\u003e Atalay et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Registered as a Ramsar site and Wildlife Development Area, Lake Burdur holds significant international ecological importance. Historically, it serves as a critical wintering ground for a significant portion of the globally endangered White-headed Duck (\u003cem\u003eOxyura leucocephala\u003c/em\u003e) and hosts endemic fish adapted to its brackish waters. While reed beds in the northern and southern alluvial zones enhance its biodiversity, recent decades have seen population declines driven by habitat loss \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eAtaol, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2010\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eSituated in a transition zone between Mediterranean and Continental climates, Burdur Lake experienced stable water levels until 1987, after which consistent shrinkage began, driven by increased evaporation and reduced inflows \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eDavraz et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). The Turkish Ministry of Agriculture and Forestry\u0026rsquo;s River Basin Management Plan (RBMP) identifies six key hydraulic structures\u0026mdash;Kara\u0026ccedil;al, Karamanlı, and Bademli dams; Beylerli, Belenli, and Tefenni ponds\u0026mdash;as primary sources of hydromorphological pressure, impounding water on tributaries such as Boz\u0026ccedil;ay and reducing inflows to Lake Burdur by an estimated 25\u0026ndash;35% since the 1980s \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eDavraz et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eŞener et al., 2020; T.C. TOB SYGM, 2020\u003c/span\u003e). These, along with additional smaller reservoirs (up to 16 total as of 2022), collectively retain approximately 126 hm\u0026sup3; annually, with substantial diversions of approximately 72.4 hm\u0026sup3; for irrigation exacerbating shrinkage \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003e\u0026Ccedil;olak et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Dervisoglu et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Sargın et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). This anthropogenic structural alteration, combined with reduced rainfall, has severely restricted the basin's hydrology, a situation further exacerbated during the 2018\u0026ndash;2025 period under study. The synergistic effects of this structural pressure and climatic stressors underscore the urgent need for rigorous monitoring to inform effective restoration strategies for this ecologically important basin. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates the geographical context of the study area and highlighting key features.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2- DATA COLLECTION AND PROCESSING\u003c/h2\u003e\u003cp\u003eThis study utilized the Google Earth Engine (GEE) cloud computing platform to process Sentinel-2 Level-2A (L2A) imagery, providing Bottom-of-Atmosphere (BOA) reflectance\u0026mdash;atmospherically corrected data\u0026mdash;at a 10 m spatial resolution. Data were sourced from the COPERNICUS/S2_SR_HARMONIZED ImageCollection, spanning January 1, 2018, to December 31, 2025, aligning with the study period. To capture Burdur Lake\u0026rsquo;s annual minimum water extent, imagery was filtered to August\u0026ndash;September, reflecting the dry-season peak, ensuring temporal consistency across 2018\u0026ndash;2025. A cloud cover threshold of \u0026lt;\u0026thinsp;10%, based on the CLOUDY_PIXEL_PERCENTAGE metadata, was applied to minimize atmospheric interference, with additional manual inspection to exclude residual haze.\u003c/p\u003e\u003cp\u003eThe Area of Interest (AOI) was defined as a rectangular bounding box (approximately 30 km \u0026times; 20 km, encompassing 600 km\u0026sup2;) manually delineated around the lake, validated against high-resolution Google Earth Pro imagery. Annual median composites were generated for each year by calculating pixel-wise median reflectance values across all qualifying images, reducing noise from clouds, shadows, and temporal variability\u0026mdash;a necessary step for subsequent thresholding. This approach, implemented using GEE\u0026rsquo;s JavaScript API, leveraged the platform\u0026rsquo;s geometry clipping function to extract AOI-specific data, ensuring precise spatial alignment. These composites serve as the foundation for uncertainty quantification, mitigating mixed-pixel effects that impact classification accuracy, thus supporting the trend analysis with Monte Carlo simulation. The complete workflow, demonstrating the integration of preprocessing, optimal classification selection, trend estimation, and the uncertainty quantification framework, is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3- NORMALIZED DIFFERENCE WATER INDEX\u003c/h2\u003e\u003cp\u003eThe Normalized Difference Water Index (NDWI), originally proposed by McFeeters (\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e1996\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e, was employed to delineate Burdur Lake\u0026rsquo;s water surface area from Sentinel-2 imagery, enhancing water features while suppressing soil and terrestrial vegetation signals. NDWI leverages the contrast between visible green light and reflected near-infrared (NIR) radiation, calculated as:\u003c/p\u003e\u003cp\u003eNDWI = (Green - NIR) / (Green\u0026thinsp;+\u0026thinsp;NIR) (1)\u003c/p\u003e\u003cp\u003ewhere Green (Band 3) and NIR (Band 8) correspond to Sentinel-2\u0026rsquo;s 10 m resolution bands \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eJumaah et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). This study computed NDWI annually (2018\u0026ndash;2025) on the Google Earth Engine (GEE) platform using median composite images derived from August\u0026ndash;September data, capturing the dry-season minimum extent.\u003c/p\u003e\u003cp\u003eThe resulting NDWI images are continuous grayscale representations with pixel values ranging from \u0026minus;\u0026thinsp;1 to 1, where water typically exhibits positive values, while soil and vegetation yield zero or negative values due to higher NIR reflectance (McFeeters, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e1996\u003c/span\u003e). This differentiation, however, is challenged by mixed pixels and shadow effects, particularly in shallow, turbid waters like Burdur Lake, necessitating robust thresholding. The NDWI output serves as the foundation for subsequent classification, where median compositing reduces noise, enhancing reliability for uncertainty quantification via Monte Carlo simulation. This step is important for propagating classification errors into the trend analysis.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.4- THRESHOLDING METHODS\u003c/h2\u003e\u003cp\u003eA critical step in extracting water pixels from water index images is the selection of an appropriate threshold, which significantly reduces omission and commission errors in spectral indices such as the Normalized Difference Water Index (NDWI) \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eYang et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Rad et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Conventional approaches often employ a zero-threshold to delineate water bodies \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eBuma et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), exploiting this spectral distinction. However, this method may introduce inaccuracies in dynamic environments like Burdur Lake, where spatiotemporal variations in image brightness and contrast\u0026mdash;due to salinity, turbidity, or seasonal changes\u0026mdash;can misclassify pixels \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eKirby et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eTo address these challenges, the Otsu algorithm provides an automated, data-driven thresholding technique, widely utilized in remote sensing \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eKulkarni et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Otsu, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e1975\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e. Otsu determines the optimal threshold by minimizing intra-class variance and maximizing inter-class variance within the NDWI histogram, enhancing separation between water and non-water pixels. This process involves preprocessing to ensure a bimodal histogram, followed by iterative variance analysis to identify the threshold that best partitions pixel intensities. In this study, both zero-thresholding and Otsu methods were applied to NDWI-derived median composites (2018\u0026ndash;2025) on the Google Earth Engine (GEE) platform. The resulting binary maps classify pixels above the threshold as water (1) and below as land (0), serving as inputs for subsequent classification and uncertainty quantification. This dual approach assesses thresholding efficacy for water area estimation.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.5- MAJORITY FILTERING\u003c/h2\u003e\u003cp\u003eAn essential post-processing step in binary map generation is majority filtering, which addresses omission and commission errors arising from the salt-and-pepper effect\u0026mdash;isolated pixel noise caused by low backscattering intensity from vegetation, shadows, and mixed pixels (Chen and Zhao, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e. This technique applies a moving window to each pixel, reassigning its class based on the most frequent value within the window, thereby smoothing spatial inconsistencies and enhancing the delineation of water bodies. Widely adopted in remote sensing, majority filtering mitigates the impact of misclassified pixels, particularly in dynamic environments like Burdur Lake, where shoreline variability complicates classification \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eSekertekin, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2021\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eIn this study, majority filtering was applied to binary maps derived from zero-threshold and Otsu thresholding of NDWI composites (2018\u0026ndash;2025) to refine water area representations. Using the Rasterio library in Python, a single-pass moving window approach was implemented, testing 3\u0026times;3 and 5\u0026times;5 window sizes. The 3\u0026times;3 window targets localized noise, preserving finer shoreline details, while the 5\u0026times;5 window offers broader smoothing, potentially at the cost of edge accuracy. This step reduces classification noise for subsequent water area estimation. The filtered outputs, classifying pixels as water (1) or land (0) based on majority vote, provide a more reliable foundation for downstream analysis, contributing to the robust mapping of Burdur Lake\u0026rsquo;s dynamic extent.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e2.6- SURFACE AREA CALCULATION\u003c/h2\u003e\u003cp\u003eAccurate quantification of Burdur Lake\u0026rsquo;s surface area is fundamental to assessing its shrinkage over 2018\u0026ndash;2025, requiring the conversion of binary raster maps\u0026mdash;generated from Otsu and zero-thresholding with 3\u0026times;3 and 5\u0026times;5 majority filters (six maps per year)\u0026mdash;into vector polygons. This process was automated using Python geospatial libraries. To enable precise area calculations, the coordinate system was transformed from the original WGS84 to the Projected Coordinate System, Universal Transverse Mercator (UTM) Zone 36N (EPSG:32636), minimizing distortion over the study area.\u003c/p\u003e\u003cp\u003eThe resulting vector maps were exported to QGIS for further processing. Polygons representing the lake were visually identified and manually extracted based on shoreline alignment with high-resolution Google Earth Pro imagery, with erroneous polygons removed to enhance accuracy. The \u003cspan\u003e$\u003c/span\u003earea function\u0026mdash;a QGIS Field Calculator tool\u0026mdash;was applied to compute surface areas in square meters, subsequently converted to square kilometers for consistency with hydrological reporting standards. These area estimates form an essential input for downstream analysis, supporting the study\u0026rsquo;s goal of rigorous quantification of lake dynamics.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e2.7- ACCURACY ASSESSMENT\u003c/h2\u003e\u003cp\u003eAssessing the accuracy of binary maps generated for Burdur Lake (2018\u0026ndash;2025) validates the thresholding and filtering processes. This study employed a stratified sampling approach, generating 275 random points using the \u0026ldquo;Random Points Inside Polygons\u0026rdquo; tool in QGIS, with a minimum distance constraint of 100 m to ensure spatial independence and mitigate autocorrelation. To address potential biases from omission and commission errors\u0026mdash;such as misclassified pixels near shorelines\u0026mdash;25 additional points were manually selected based on visual inspection of error patterns, validated against high-resolution Google Earth Pro images temporally matched to the Sentinel-2 acquisition dates (August\u0026ndash;September). Pixel values were automatically extracted from binary raster maps using a Python script with the rasterio library, storing predicted and actual classifications in the point shapefile\u0026rsquo;s attribute table.\u003c/p\u003e\u003cp\u003eReference data were collected from Google Earth Pro imagery, cross-checked for cloud-free conditions, ensuring alignment with the study period. Accuracy was evaluated through a suite of metrics, including the Confusion Matrix, Overall Accuracy, Kappa Coefficient, F1-Score, Intersection over Union (IoU), Producer\u0026rsquo;s Accuracy, and User\u0026rsquo;s Accuracy. The sample size of 300 points was determined to cover approximately 5% of the lake\u0026rsquo;s variable shoreline, providing a representative assessment. This process quantifies classification errors, feeding into the study\u0026rsquo;s uncertainty quantification framework, and supports the reliable estimation of Burdur Lake\u0026rsquo;s shrinkage dynamics.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e2.8- MCNEMAR TEST\u003c/h2\u003e\u003cp\u003eEvaluating the statistical significance of differences in classification accuracy between thresholding methods validates Burdur Lake\u0026rsquo;s water-land binary maps (2018\u0026ndash;2025). The McNemar test, introduced by McNemar (1947) \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eRudke et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), was employed to compare the performance of Otsu and zero-thresholding, as well as their filtered variants (3\u0026times;3 and 5\u0026times;5), using the same set of 300 independent reference points. This non-parametric test facilitates 15 pairwise comparisons (Otsu/Zero, Filtered/Unfiltered) annually, addressing the underexplored cumulative impact of thresholding and filtering on classification reliability.\u003c/p\u003e\u003cp\u003eThe test constructs 2\u0026times;2 confusion matrices from reference points, comparing pixel classifications between two methods. The chi-square (χ\u0026sup2;) statistic, calculated as:\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$\\:{X}^{2}\\:=\\:\\frac{{({f}_{12}\\:-\\:{f}_{21})}^{2}}{{f}_{12}+{f}_{21}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{f}_{12}\\)\u003c/span\u003e\u003c/span\u003e represents pixels correctly classified by Otsu (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{C}_{1}\\)\u003c/span\u003e\u003c/span\u003e) but incorrectly by zero-threshold (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{C}_{2}\\)\u003c/span\u003e\u003c/span\u003e), and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{f}_{21}\\)\u003c/span\u003e\u003c/span\u003e denotes the reverse, quantifies the difference \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eYan et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Dermosinoglou and Petropoulos, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Foody, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2004\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e. The null hypothesis (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{H}_{0}\\)\u003c/span\u003e\u003c/span\u003e) posits equal performance, rejected at a significance level of p\u0026thinsp;\u0026lt;\u0026thinsp;0.05. Implementation utilized the statsmodels library in Python, with a custom script to automate matrix generation and χ\u0026sup2; computation across the time series.\u003c/p\u003e\u003cp\u003eThis analysis supports the reliability of thresholding outcomes, providing a foundation for downstream mapping accuracy. By isolating method-specific differences, the McNemar test supports the study\u0026rsquo;s objective of accurate classification validation for Burdur Lake\u0026rsquo;s dynamic extent.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e2.9- TREND ANALYSIS\u003c/h2\u003e\u003cp\u003eQuantifying the temporal dynamics of Burdur Lake\u0026rsquo;s surface area requires robust trend detection, addressed through nonparametric statistical methods. The Mann-Kendall test, introduced by Mann (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e1945\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e and refined by Kendall (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e1948\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e, is a widely used nonparametric approach to identify statistically significant monotonic trends\u0026mdash;consistent upward or downward changes\u0026mdash;without assuming data normality \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003ede Brito Neto et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Arra et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Its resilience to sudden breaks makes it suitable for heterogeneous time series, such as those affected by seasonal variability \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eArra et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Complementing this, the Sen\u0026rsquo;s Slope estimator, developed independently by Theil (\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e1950\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e and Sen (\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e1968\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e, quantifies the magnitude of the trend by calculating the median slope of all pairwise differences between data points, effectively handling non-normal distributions and outliers.\u003c/p\u003e\u003cp\u003eIn this study, these methods were applied to annual median composites (2018\u0026ndash;2025) derived from Otsu-thresholded NDWI images of Burdur Lake. The Mann-Kendall test assessed trend significance at a p\u0026thinsp;\u0026lt;\u0026thinsp;0.05 threshold, preprocessing data to remove extreme outliers via a 95% quantile filter to ensure stability. Sen\u0026rsquo;s Slope estimated the median annual change rate by computing slopes across all pairwise area differences, implemented with a sliding window to capture seasonal effects. Both techniques were executed using the pymannkendall library in Python, with a custom script to automate calculations across the eight-year series. This approach enhances trend reliability, providing a foundation for subsequent uncertainty quantification via Monte Carlo simulation.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e2.10- UNCERTAINTY QUANTIFICATION\u003c/h2\u003e\u003cp\u003eWhile the methodology thus far yields a robust point estimate for Burdur Lake\u0026rsquo;s annual shrinkage rate, it does not account for the systematic propagation of classification uncertainty into the final trend, a common challenge in remote sensing time series. These uncertainties arise primarily from difficulties in defining the water-land boundary through thresholding and spatial filtering, impacting the reliability of area estimates \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eSekertekin, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2021\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e. Addressing this gap is essential to enhance the credibility of models reliant on satellite data \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eCockx et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Feizizadeh et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). To this end, a Monte Carlo Simulation (MCS) framework was integrated with the Sen\u0026rsquo;s Slope trend analysis, effectively addressing the non-linear complexities of polygon area calculations on binary maps without assuming error distributions \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003ePapadopoulos and Yeung, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2001\u003c/span\u003e; Biljecki et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Feizizadeh et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe MCS process began by stochastically perturbing annual surface area values from the optimal classification map (Otsu with majority filtering) over 1,000 iterations. The perturbation range of \u0026plusmn;\u0026thinsp;1% was empirically derived from classification variability in challenging boundary pixels, validated against 300 reference points (98\u0026ndash;99% accuracy from Google Earth Pro), ensuring robustness across plausible scenarios through dynamic probability density functions in Monte Carlo simulations \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eHealey et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). This generated an ensemble of 1,000 synthetic time series, each reflecting a potential realization of the lake\u0026rsquo;s area history. The Sen\u0026rsquo;s Slope estimator was then applied to each series, propagating classification uncertainty into trend quantification. The final shrinkage rate was determined as the median of the 1,000 slope values, with the 95% Confidence Interval (CI) defined by the 2.5th and 97.5th percentiles of the simulated distribution, providing a quantitative measure of reliability \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eMcMurray et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Lee et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Implemented using Python with the numpy library and a fixed random seed (42) for consistency, this approach accounted for computational efficiency on the Google Earth Engine platform. This integration ensures a statistically defended trend estimate, supporting the study\u0026rsquo;s goal of rigorous shrinkage assessment.\u003c/p\u003e\u003c/div\u003e"},{"header":"3 - RESULTS","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e3.1- SURFACE AREA ESTIMATION AND VALIDATION\u003c/h2\u003e\u003cp\u003eThe estimation of Burdur Lake\u0026rsquo;s annual surface area (2018\u0026ndash;2025) was conducted using Sentinel-2 imagery from August\u0026ndash;September, capturing the dry-season minimum extent. Otsu and zero thresholding, combined with 3\u0026times;3 and 5\u0026times;5 majority filtering, produced six maps per year. Otsu-thresholded areas declined from 126,757,096 m\u0026sup2; (126.76 km\u0026sup2;) in 2018 to 115,167,474 m\u0026sup2; (115.17 km\u0026sup2;) in 2025, with filtered variants showing slight variations: Otsu 3\u0026times;3 from 126,757,571 m\u0026sup2; to 115,168,184 m\u0026sup2;, and Otsu 5\u0026times;5 from 126,754,255 m\u0026sup2; to 115,166,605 m\u0026sup2;. Zero-thresholded areas ranged from 127,454,300 m\u0026sup2; (127.45 km\u0026sup2;) to 115,788,929 m\u0026sup2; (115.79 km\u0026sup2;), with similar filtered adjustments. This 11.59 km\u0026sup2; net loss (Otsu) indicates ongoing shrinkage, influenced by filtering\u0026rsquo;s edge-smoothing effects.\u003c/p\u003e\u003cp\u003eValidation against 300 reference points, randomly generated in QGIS and manually adjusted for omission/commission errors, used high-resolution Google Earth Pro images temporally aligned with August\u0026ndash;September acquisitions. Otsu achieved accuracies of 98% (Kappa 0.92) in 2018, improving to 99% (Kappa 0.96) in 2025, with producer\u0026rsquo;s accuracy rising from 0.93 to 0.95 and user\u0026rsquo;s accuracy from 0.91 to 0.98, indicating consistent performance. Zero thresholding started at 93% accuracy (Kappa 0.76) in 2018, reaching 96% (Kappa 0.85) in 2025, suggesting adaptation to boundary variability. Confusion matrices for 2025 showed 257 true positives, 1 false positive, and 42 false negatives, highlighting Otsu\u0026rsquo;s precision. McNemar tests confirmed no significant differences among Otsu variants (p\u0026thinsp;\u0026ge;\u0026thinsp;0.25) or Zero variants (p\u0026thinsp;\u0026ge;\u0026thinsp;0.06), but significant differences between Otsu and Zero (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) annually \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eVan Tricht et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), underscoring Otsu\u0026rsquo;s reliability. The comprehensive accuracy assessment, including F1-Scores and Producer's and User's Accuracy metrics, is fully detailed in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, providing the quantitative evidence for the optimal selection of the Otsu 3x3 classification variant. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e visually substantiates these findings by charting the superior and stable performance of the Otsu-thresholding variants, particularly emphasizing the critical distinction in Overall Accuracy and Kappa Coefficient when compared against the conventional zero-thresholding method across the entire time series.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSummary of the classification accuracy metrics and annual McNemar test results for 2025. The comprehensive accuracy metrics consistently demonstrate the superiority of the Otsu variants (OA \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\ge\\:\\)\u003c/span\u003e\u003c/span\u003e 0.99, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:K\\:\\ge\\:\\)\u003c/span\u003e\u003c/span\u003e 0.96). The McNemar test confirms that Otsu-based methods are statistically different from Zero-thresholding (p \u0026lt; 0.004 vs. No Filter), justifying the optimal selection of the Otsu 3x3 variant and validating the use of a low \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\pm\\:1\\%\\:\\)\u003c/span\u003e\u003c/span\u003eerror margin in the subsequent Monte Carlo Simulation.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"7\"\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=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMetric / Method\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOtsu (No Filter)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eOtsu 3x3 (Optimal)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eOtsu 5x5\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eZero (No Filter)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eZero 3x3\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eZero 5x5\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eOverall Accuracy (OA)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.99\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.99\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.99\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.98\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eKappa Coefficient (K)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.91\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.91\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eF1-Score\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.87\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.92\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.92\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eProducer's Accuracy\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.95\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eUser's Accuracy\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.80\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.89\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMcNemar p-value vs. Otsu 3x3\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\ge\\:\\)\u003c/span\u003e\u003c/span\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\ge\\:\\)\u003c/span\u003e\u003c/span\u003e1.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u0026lt;\u0026thinsp;0.004\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e0.125\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.125\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=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e3.2- TREND ANALYSIS\u003c/h2\u003e\u003cp\u003eAssessing the temporal dynamics of Burdur Lake\u0026rsquo;s surface area (2018\u0026ndash;2025) is essential to quantify its shrinkage, utilizing the Otsu 3\u0026times;3 filtered time series derived from Sentinel-2 imagery. The Mann-Kendall (MK) test, a nonparametric method, was initially applied to detect statistically significant monotonic trends\u0026mdash;consistent decreases over time\u0026mdash;without assuming data normality. The test yielded a Z value of -3.45 (p\u0026thinsp;=\u0026thinsp;0.0008, α\u0026thinsp;=\u0026thinsp;0.05), confirming a significant downward trend. Subsequently, the Sen\u0026rsquo;s Slope estimator quantified the trend magnitude by calculating the median slope of all pairwise area differences, resulting in a rate of -1,638,994.40 m\u0026sup2;/year (-1.64 km\u0026sup2;/year), with an intercept of 125,926,857.90 m\u0026sup2;. This corresponds to a total reduction of 11,591,622 m\u0026sup2; (11.59 km\u0026sup2;) over the study period, reflecting a 9.2% decline from the 2018 baseline.\u003c/p\u003e\u003cp\u003eThe analysis leveraged the pymannkendall library in Python, with preprocessing to remove outliers via a 95% quantile filter, ensuring trend stability. This point estimate, while robust, captures only the central tendency, with variability across thresholding variants (e.g., Otsu 5\u0026times;5 at -1.63 km\u0026sup2;/year) suggesting method sensitivity. The decline aligns with historical trends \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eDavraz et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), likely driven by anthropogenic pressures, underscoring the need for ecological intervention. However, the final, statistically defended confidence limits, accounting for classification uncertainty, are detailed in the Uncertainty Quantification section (3.3), enhancing the reliability of this trend for Burdur Lake\u0026rsquo;s management. To document the precise input for the trend estimation, the annual surface area values derived from the optimal Otsu 3x3 classification are itemized in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\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\u003e\u003cb\u003eAnnual Surface Area Time Series for Burdur Lake (2018\u0026ndash;2025).\u003c/b\u003e The annual surface area, derived from the Otsu 3x3 optimal classification, serves as the foundational data for the Mann-Kendall and Sen's Slope trend analysis. The documented area decline from 126.76 km\u0026sup2; in 2018 to 115.17 km\u0026sup2; in 2025 demonstrates a net loss of 11.59 km\u0026sup2; over the eight-year period.\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\u003eYear\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eArea (m\u0026sup2;) (Otsu 3x3)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eArea (km\u0026sup2;) (Otsu 3x3)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2018\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e126.757,571\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e126.76\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=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e124.903,447\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e124.90\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=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e123.265,115\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e123.27\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=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e120.958,300\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e120.96\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=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e119.422,455\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e119.42\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=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e118.263,205\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e118.26\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=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e116.708,843\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e116.71\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=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e115.168,184\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e115.17\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=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003e3.3- UNCERTAINTY QUANTIFICATION\u003c/h2\u003e\u003cp\u003eThe Sen\u0026rsquo;s Slope-derived point estimate of -1.64 km\u0026sup2;/year for Burdur Lake\u0026rsquo;s shrinkage (2018\u0026ndash;2025) provides a robust trend measure, yet it overlooks the systematic propagation of classification uncertainty inherent in remote sensing time series. These uncertainties, primarily arising from challenges in defining the water-land boundary through thresholding and spatial filtering \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eSekertekin, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2021\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e, undermine trend reliability, necessitating rigorous quantification to enhance model credibility \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eCockx et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Feizizadeh et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). The MCS framework (detailed in Section \u003cspan refid=\"Sec12\" class=\"InternalRef\"\u003e2.10\u003c/span\u003e) was applied to propagate classification uncertainty, yielding a median slope of -1.64 km\u0026sup2;/year with a 95% CI of -1.83 to -1.47 km\u0026sup2;/year. The negative CI indicates a decline, validated by the methodology\u0026rsquo;s robustness. These findings provide a basis for water management strategies. Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e provides a consolidated, statistically defended summary of the key trend metrics, encompassing the Mann-Kendall test results and the final Confidence Interval derived from the 1,000-iteration Monte Carlo Simulation. Figure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e graphically validates the robustness of this approach, illustrating the median annual lake area and the derived 95% Confidence Interval (CI) that rigorously quantifies the propagated classification uncertainty across the entire 2018\u0026ndash;2025 time series.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFurthermore, the core output of the Monte Carlo Simulation is visualized in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e, which presents the density distribution of the 1,000 resulting Sen's Slope values, quantifying the uncertainty of the annual shrinkage rate itself.\u003c/p\u003e\u003cp\u003e\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\u003e\u003cb\u003eSummary of Trend Analysis and 95% Uncertainty Quantification (Otsu 3x3, 2018\u0026ndash;2025).\u003c/b\u003e The results confirm a statistically highly significant monotonic downward trend (p\u0026thinsp;=\u0026thinsp;0.0008). The 95% Confidence Interval (CI), derived from the Monte Carlo Simulation, is entirely negative and tightly constrained, providing high confidence that the annual shrinkage rate lies between \u0026minus;\u0026thinsp;1.83 and \u0026minus;\u0026thinsp;1.47 km\u0026sup2;/year.\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=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMetric\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eValue\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eUnits\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMedian Sen\u0026rsquo;s Slope\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-1.64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ekm\u0026sup2;/year\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTotal Area Reduction\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e11.59\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ekm\u0026sup2;\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMann-Kendall Z Statistic\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-3.45\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eDimensionless\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMann-Kendall p-value\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.0008\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eDimensionless\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e95% CI (Lower Bound)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-1.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ekm\u0026sup2;/year\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e95% CI (Upper Bound)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-1.47\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ekm\u0026sup2;/year\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"},{"header":"4 - DISCUSSION","content":"\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003e4.1- Interpretation of Trends and Methodological Insights\u003c/h2\u003e\u003cp\u003eThe quantified shrinkage indicates a total loss of 11.59 km\u0026sup2; during the study period and a persistent deficit in the lake's water budget. When placed in historical context, (37% reduction from 1975 to 2016; Davraz et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), this finding is consistent. The continuing high-magnitude shrinkage strongly suggests that anthropogenic factors remain the dominant drivers, overshadowing the contribution of climatic variability \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eDervisoglu et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe persistence of this decline aligns with prior research attributing the water deficit primarily to streamflow diversion and groundwater pumping \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eDervisoglu et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The efficacy of human intervention is quantitatively supported by numerical models, such as MODFLOW. These models demonstrate that water levels can be engineered to rebound significantly by prioritizing the release of surface water flows, even despite the negative impacts of climate change (e.g., increased evaporation). This confirms that policy and management decisions hold the controlling leverage over the lake\u0026rsquo;s fate \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eKılı\u0026ccedil; Germe\u0026ccedil;, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2023\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eMethodologically, this study offers insights for remote sensing time series. The systematic comparison between classification methods demonstrated the superior performance and stability of the Otsu thresholding algorithm over the common zero-thresholding approach, a difference that the McNemar test confirmed was statistically significant. The Otsu method's adaptive, data-driven approach minimized intra-class variance supported consistent delineation (98\u0026ndash;99% accuracy).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\u003ch2\u003e4.2- Uncertainty and Validation Considerations\u003c/h2\u003e\u003cp\u003eThe transition from continuous spectral index imagery to binary water/land maps is the primary source of epistemic uncertainty in remotely sensed water resource monitoring, particularly in complex, shallow endorheic basins like Burdur Lake. The validation protocol in this study systematically quantified this uncertainty, demonstrating that the choice of classification methodology critically impacts the resulting area estimates. The comparison against 300 reference points confirmed the superior performance of the Otsu algorithm, which achieved high overall accuracies (ranging from 98% in 2018 to 99% in 2025) and strong agreement (Kappa coefficients between 0.92 and 0.96). Crucially, the McNemar test provided the necessary statistical rigor, showing a significant annual difference between the Otsu method and the conventional zero-thresholding approach (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e\u003cp\u003eWhile the 300-point validation dataset, adjusted for omission and commission errors, achieved high accuracy (98\u0026ndash;99%), it may not fully capture the spatial heterogeneity of Burdur Lake\u0026rsquo;s dynamic shoreline, particularly in shallow, turbid regions. Additionally, despite applying a\u0026thinsp;\u0026lt;\u0026thinsp;10% cloud cover threshold, residual atmospheric interference (e.g., haze) in some Sentinel-2 images could introduce minor classification errors, especially for mixed pixels near the water-land boundary. These pixels, influenced by salinity and seasonal variability, may affect area estimates, though median compositing mitigates this to an extent. Temporal variability in image acquisition dates within the August\u0026ndash;September window could also contribute to subtle inconsistencies, as water levels fluctuate slightly within the dry season. Future studies could address these by incorporating multi-sensor data (e.g., Sentinel-1 for cloud-penetrating capabilities) or machine learning-based classification to enhance boundary delineation. These uncertainties, while minor, underscore the need for cautious interpretation of the reported shrinkage rate and highlight opportunities for methodological refinement.\u003c/p\u003e\u003cp\u003eDespite these minor uncertainties, the Monte Carlo Simulation framework, integrated with the Sen\u0026rsquo;s Slope estimator, provides an estimate of Burdur Lake\u0026rsquo;s shrinkage rate (-1.64 km\u0026sup2;/year, 95% CI: -1.83 to -1.47 km\u0026sup2;/year), accounting for classification errors. This trend, validated at 98\u0026ndash;99% accuracy, offers a baseline for understanding the lake\u0026rsquo;s hydrological decline. By quantifying the area loss, this study offers a foundation for informing water resource management strategies, particularly in addressing the anthropogenic drivers exacerbating the crisis.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003e4.3- Implications for Water Resource Management\u003c/h2\u003e\u003cp\u003eThe shrinkage rate of -1.64 km\u0026sup2;/year indicates a water resource issue in the Burdur Closed Basin (BCB). This decline confirms that anthropogenic activities, such as irrigation diversions and groundwater abstraction, dominate over climatic stressors \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eDervisoglu et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Consequently, the lake's ecological fate hinges on fundamental, basin-wide policy restructuring, not on mitigating external climate change alone.\u003c/p\u003e\u003cp\u003eFirstly, this finding serves as a timely metric for policy intervention. Previous MODFLOW groundwater model studies established that, while excessive pumping and climate change forecast a 7 m decline over 46 years, strategically releasing surface water flows could force a rebound of up to 3 m \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eKılı\u0026ccedil; Germe\u0026ccedil;, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2023\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e. This confirms that human management holds the controlling leverage. The observed persistent shrinkage in this study indicates that this leverage has not been utilized, necessitating immediate action to reverse reservoir operational policies that divert nearly all natural surface flows. Policy adjustment is supported by studies estimating that mandatory conversion to efficient irrigation methods (e.g., pressurized systems) could conserve 62.6 hm\u0026sup3; of water annually \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eSargın et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). This potential saving matches the estimated annual water loss rate for the lake (approximately 40 hm\u0026sup3;) \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eAtaol, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2010\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e, demonstrating that the crisis is solvable through policy-driven water use optimization.\u003c/p\u003e\u003cp\u003eSecondly, the management imperative extends to groundwater abstraction. The drilling of numerous boreholes has severely depleted the regional aquifer, eliminating the essential baseflow contribution to the lake \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eAtaol, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2010\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e. The combination of this abstraction, regional temperature projections (a rise of at least 2\u0026deg;C by 2100) \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003e\u0026Ccedil;olak et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), and increased irrigation demand is projected to exacerbate the drying process unless groundwater extraction rates are effectively regulated. Resource management must implement strict regulatory enforcement \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(MoEU, 2020)\u003c/span\u003e to allow the aquifer system to recover and re-establish a hydraulic connection with the lake.\u003c/p\u003e\u003cp\u003eFinally, the decline affects the lake's status as an internationally significant Ramsar site, which is the most important wintering site for the critically endangered White-headed Duck (\u003cem\u003eOxyura leucocephala\u003c/em\u003e) \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(Doğa Derneği, n.d.)\u003c/span\u003e. The substantial area loss confirmed here indicates a continuing collapse of critical shallow littoral habitats and accelerates salinization \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003e\u0026Ccedil;olak et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Failure to adopt a \"dynamic lake management plan\" prioritizing ecosystem needs will contribute to local climate changes\u0026mdash;increasing continentality and public health hazards \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(AECOM, 2019)\u003c/span\u003e from exposed, dusty lakebed sediments. The shrinkage rate suggests a shift in the current resource paradigm from exploitation to restoration.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e\u003ch2\u003e4.4- Future Research Directions\u003c/h2\u003e\u003cp\u003eThis shrinkage trend provides a metric for immediate policy action and a foundation for critical, subsequent research. Future studies must build upon the methodological advancements presented here to tackle the outstanding uncertainties in hydrological attribution, policy effectiveness, and socio-economic dynamics.\u003c/p\u003e\u003cp\u003eMethodologically, the robust Monte Carlo Simulation (MCS) uncertainty quantification framework successfully applied here to surface area should be extended to the third dimension. Future research could integrate high-resolution, multi-source altimetry data \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eHou et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) (e.g., Sentinel-3, ICESat-2) and historical Digital Elevation Models (DEMs) to derive annual rate of volume loss and water level change. This 3D approach, incorporating elevation error propagation, supports understanding of the lake's desiccation that current 2D area studies lack.\u003c/p\u003e\u003cp\u003eHydrologically, while this research confirms the dominance of anthropogenic factors, the quantitative attribution of future climate stress requires refinement. Studies utilizing the MODFLOW numerical model have established the potential for management interventions to generate a multi-meter rebound (+\u0026thinsp;3 m) even under projected climate decline. Future work should prioritize refining the long-term water budget by integrating high-resolution Regional Climate Models (CORDEX RCP scenarios) with empirical data on regional thermal stress (e.g., the documented 2.2\u0026deg;C Lake Surface Water Temperature increase) to accurately partition the contributions of increasing evaporative output and decreased natural inflow under long-term drought conditions \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eKılı\u0026ccedil; Germe\u0026ccedil;, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2023\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eFinally, future research must transition from documenting decline to evaluating the efficacy of proposed mitigation strategies. Given the potential for optimized irrigation to save water exceeding the lake's annual deficit \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(\u003c/span\u003eAtaol, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2010\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e, a crucial research direction is the development of socio-economic models. These models must assess the economic costs, policy feasibility, and financial incentives necessary for a mandatory basin-wide conversion to pressurized irrigation systems, dynamically linking Land Use/Land Cover (LULC) change, water abstraction, and local policy implementation. Furthermore, high-temporal-resolution ecological monitoring remains necessary \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e(MoEF, 2007)\u003c/span\u003e to quantify the rate of biodiversity loss\u0026mdash;especially the collapse of essential shallow habitats\u0026mdash;and inform the designation of newly critical preservation zones.\u003c/p\u003e\u003c/div\u003e"},{"header":"5 - CONCLUSION","content":"\u003cp\u003eThis study developed and implemented a robust methodology for quantifying Burdur Lake\u0026rsquo;s surface area decline using a high-resolution Sentinel-2 time series (2018\u0026ndash;2025). Sen\u0026rsquo;s Slope estimator was integrated with an MCS framework to account for classification uncertainty, yielding a robust shrinkage rate of -1.64 km\u0026sup2;/year, totaling 11.59 km\u0026sup2;, consistent with the historical trajectory (37% reduction, 1975\u0026ndash;2016). The 95% Confidence Interval (CI) of -1.83 to -1.47 km\u0026sup2;/year, derived from 1,000 MCS iterations, validates sustained area loss, driven by anthropogenic factors.\u003c/p\u003e\u003cp\u003eThis decline indicates a water budget deficit, affecting the lake\u0026rsquo;s Ramsar status and its endangered White-headed Duck population. It provides a basis for policy intervention, such as revised reservoir management and stringent groundwater controls to restore hydrological balance. Future research should extend this MCS framework by integrating multi-source altimetry (e.g., Sentinel-3, ICESat-2) with Digital Elevation Models to develop a 3D Uncertainty Quantification model, quantifying volume loss and water level changes. Additionally, coupling these physical insights with socio-economic analyses of irrigation efficiency will enhance policy feasibility, supporting sustainable restoration and ecological recovery.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eORCID\u003c/h2\u003e\u003cp\u003e0009-0009-4974-0206\u003c/p\u003e\u003cp\u003eAuthor Contribution\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe author confirms sole responsibility for the following: conceptualization and design of the study; data acquisition, preprocessing, analysis; interpretation of results; drafting and critical revision of the manuscript; and final submission approval of the published version.\u003c/p\u003e\n\u003ch3\u003eFunding Declaration\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eThe author declares that no external funding was received from any external source or organization for the research, authorship, and/or publication of this article.\u003c/p\u003e\n\u003cp\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe author declares no competing interests.\u003c/p\u003e\n\u003cp\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData will be made available on request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAECOM. (2019), February 15 Consequences of drying lake systems around the world (Summary of the February 15, 2019 report prepared for the Great Salt Lake Advisory Council). Great Salt Lake Advisory Council\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAl Garni HZ, Awasthi A (2020) A Monte Carlo approach applied to sensitivity analysis of criteria impacts on solar PV site selection. Handbook of Probabilistic Models. Butterworth-Heinemann, pp 489\u0026ndash;504\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eArra AA, Alashan S, Şişman E (2024) Trends of meteorological and hydrological droughts and associated parameters using innovative approaches. J Hydrol 640:131661\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eArra AA, Keskin MZ, Şişman E (2025) Trend Analysis of Hydro-Meteorological Variables Using Mann-Kendall and Sen's Slope with Standardization (SSS): Case Study of the Kızılırmak Catchment, T\u0026uuml;rkiye. Phys Chem Earth Parts A/B/C, 104115\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAtalay İ, Altunbaş S, Siler M (2020) The Formation and Evaluation of the Faulted Topography in the Burdur Basin, Lakes Region, SW Anatolia. J Geogr, (41), 41\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAtaol M (2010) Burdur G\u0026ouml;l\u0026uuml;\u0026rsquo;nde seviye değişimleri. Coğrafi Bilimler Dergisi 8(1):77\u0026ndash;92\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBiljecki F, Ledoux H, Stoter J (2014) Error propagation in the computation of volumes in 3D city models with the Monte Carlo method. ISPRS Annals Photogrammetry Remote Sens Spat Inform Sci 2:31\u0026ndash;39\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBuma WG, Lee SI, Seo JY (2018) Recent surface water extent of lake Chad from multispectral sensors and GRACE. \u003cem\u003eSensors\u003c/em\u003e, \u003cem\u003e18\u003c/em\u003e(7), 2082\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChe L, Li S, Liu X (2025) Improved surface water mapping using satellite remote sensing imagery based on optimization of the Otsu threshold and effective selection of remote-sensing water index. J Hydrol 654:132771\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen Z, Zhao S (2022) Automatic monitoring of surface water dynamics using Sentinel-1 and Sentinel-2 data with Google Earth Engine. Int J Appl Earth Obs Geoinf 113:103010\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCockx K, Van de Voorde T, Canters F (2014) Quantifying uncertainty in remote sensing-based urban land-use mapping. Int J Appl Earth Obs Geoinf 31:154\u0026ndash;166\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCrosetto M, Tarantola S, Saltelli A (2000) Sensitivity and uncertainty analysis in spatial modelling based on GIS. Agric Ecosyst Environ 81(1):71\u0026ndash;79\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e\u0026Ccedil;olak MA, \u0026Ouml;ztaş B, \u0026Ouml;zgencil İK, Soyluer M, Korkmaz M, Ram\u0026iacute;rez-Garc\u0026iacute;a A, Aky\u0026uuml;rek Z (2022) Increased water abstraction and climate change have substantial effect on morphometry, salinity, and biotic communities in lakes: examples from the semi-arid burdur Basin (Turkey). Water 14(8):1241\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDavraz A, Sener E, Sener S (2019) Evaluation of climate and human effects on the hydrology and water quality of Burdur Lake, Turkey. J Afr Earth Sc 158:103569\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ede Brito Neto RT, Santos CA, Mulligan K, Barbato L (2016) Spatial and temporal water-level variations in the Texas portion of the Ogallala Aquifer. Nat Hazards 80(1):351\u0026ndash;365\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDermosinoglou A, Petropoulos GP (2024) Exploring long term Impervious Surface Areas (ISA) dynamics using Landsat imagery, Μachine Learning and GEE: The case of Attica, Greece. Remote Sens Applications: Soc Environ 36:101338\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDervisoglu A, Yağmur N, Fıratlı E, Musaoğlu N, Tanık A (2022) Spatio-temporal assessment of the shrinking Lake Burdur, Turkey. Int J Environ Geoinformatics 9(2):169\u0026ndash;176\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDoğa Derneği (n.d.). \u003cem\u003eBurdur lake\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://dogadernegi.org/en/burdur-lake/\u003c/span\u003e\u003cspan address=\"https://dogadernegi.org/en/burdur-lake/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDu Y, Zhang Y, Ling F, Wang Q, Li W, Li X (2016) Water bodies\u0026rsquo; mapping from Sentinel-2 imagery with modified normalized difference water index at 10-m spatial resolution produced by sharpening the SWIR band. Remote Sens 8(4):354\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eErdem KC, Bakırman T, Bayram B Temporal Dynamics of Lake Burdur's Water Surface Area: A Two-Decade Remote Sensing Analysis and Future Forecasts. Mersin Photogrammetry J, \u003cem\u003e7\u003c/em\u003e(1), 22\u0026ndash;28\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFeizizadeh B, Jankowski P, Blaschke T (2014) A GIS based spatially-explicit sensitivity and uncertainty analysis approach for multi-criteria decision analysis. Comput Geosci 64:81\u0026ndash;95\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFoody GM (2004) Thematic map comparison. Photogrammetric Eng Remote Sens 70(5):627\u0026ndash;633\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGhorbani S, Pamucar D (2025) Remote Sensing-Based Evaluation of Lake Area Dynamics: A Quantitative Assessment for Environmental Management in Turkey. Spectr Oper Res 3(1):352\u0026ndash;358\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHealey SP, Urbanski SP, Patterson PL, Garrard C (2014) A framework for simulating map error in ecosystem models. Remote Sens Environ 150:207\u0026ndash;217\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHong S, Heo J, Vonderohe AP (2013) Simulation-based approach for uncertainty assessment: Integrating GPS and GIS. Transp Res Part C: Emerg Technol 36:125\u0026ndash;137\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHou J, Van Dijk AI, Renzullo LJ, Larraondo PR (2024) GloLakes: water storage dynamics for 27 000 lakes globally from 1984 to present derived from satellite altimetry and optical imaging. Earth Syst Sci Data 16(1):201\u0026ndash;218\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJumaah HJ, Ameen MH, Kalantar B (2023) Surface water changes and water depletion of Lake Hamrin, Eastern Iraq, using Sentinel-2 images and geographic information systems. Adv Environ Eng Res 4(1):1\u0026ndash;11\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKendall MG (1948) Rank correlation methods\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKılı\u0026ccedil; Germe\u0026ccedil; H (2023) Assessment of the impacts of future climatic variations and anthropogenic activities on Burdur Lake levels\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKirby K, Ferguson S, Rennie CD, Cousineau J, Nistor I (2024) Identification of the best method for detecting surface water in Sentinel-2 multispectral satellite imagery. Remote Sens Applications: Soc Environ 36:101367\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKulkarni R, Khare K, Khanum H (2022) Detecting, extracting, and mapping of inland surface water using Landsat 8 Operational Land Imager: A case study of Pune district, India. \u003cem\u003eF1000Research\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e, 774\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLee S, Moon S, Kim K, Sung S, Hong Y, Lim W, Park SK (2024) A comparison of green, delta, and Monte Carlo methods to select an optimal approach for calculating the 95% confidence interval of the Population-attributable fraction: guidance for epidemiological research. J Prev Med Public Health 57(5):499\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMann HB (1945) Nonparametric tests against trend. Econometrica: J econometric Soc, 245\u0026ndash;259\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMcFeeters SK (1996) The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features. Int J Remote Sens 17(7):1425\u0026ndash;1432\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMcMurray A, Pearson T, Casarim F (2017) Guidance on applying the Monte Carlo approach to uncertainty analyses in forestry and greenhouse gas accounting. \u003cem\u003eWinrock International: Arlington, VA, USA\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOtsu N (1975) A threshold selection method from gray-level histograms. Automatica 11(285\u0026ndash;296):23\u0026ndash;27\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePapadopoulos CE, Yeung H (2001) Uncertainty estimation and Monte Carlo simulation method. Flow Meas Instrum 12(4):291\u0026ndash;298\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRad AM, Kreitler J, Sadegh M (2021) Augmented Normalized Difference Water Index for improved surface water monitoring. Environ Model Softw 140:105030\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRepublic of Turkey, Ministry of Environment and Forestry, General Directorate of Nature Conservation and National Parks, Department of Nature Conservation (2007) The national biological diversity strategy and action plan\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRepublic of Turkey Ministry of Environment and Urbanization, General Directorate of Environmental Impact Assessment, Permit and Inspection. (2020). 6th state of environment report for Republic of Turkey (Publication No. 48/2)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRudke AP, Xavier ACF, Fujita T, Abou Rafee SA, Martins LD, Morais MVB, Martins JA (2021) Mapping past landscapes using landsat data: Upper Paran\u0026aacute; River Basin in 1985. Remote Sens Applications: Soc Environ 21:100436\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSargın A, Taşkıran F, Yılmaz E, S\u0026ouml;nmez Y, Yeniyurt C (2011) No lake, no Burdur! Doğa Derneği.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSarp G, Ozcelik M (2017) Water body extraction and change detection using time series: A case study of Lake Burdur, Turkey. J Taibah Univ Sci 11(3):381\u0026ndash;391\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSekertekin A (2021) A Survey on Global Thresholding Methods for Mapping Open Water Body Using Sentinel-2 Satellite Imagery and Normalized Difference Water Index. Arch Comput Methods Eng, \u003cem\u003e28\u003c/em\u003e(3)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSen PK (1968) Estimates of the regression coefficient based on Kendall's tau. J Am Stat Assoc 63(324):1379\u0026ndash;1389\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eŞehnaz Ş, Şener E, Davraz A, Varol S (2020) Hydrogeological and hydrochemical investigation in the Burdur Saline Lake Basin, southwest Turkey. Geochemistry 80(4):125592\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eT.C. Tarım ve Orman Bakanlığı, Su Y\u0026ouml;netimi Genel M\u0026uuml;d\u0026uuml;rl\u0026uuml;ğ\u0026uuml;. (2020, April). Burdur Basin river basin management plan final report Annex 1\u0026ndash;6\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTesfaye M, Breuer L (2024) Performance of water indices for large-scale water resources monitoring using Sentinel-2 data in Ethiopia. Environ Monit Assess 196(5):467\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTheil H (1950) A rank-invariant method of linear and polynomial regression analysis. Indagationes Math 12(85):173\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVan Tricht K, Degerickx J, Gilliams S, Zanaga D, Battude M, Grosu A, Szantoi Z (2023) WorldCereal: a dynamic open-source system for global-scale, seasonal, and reproducible crop and irrigation mapping. Earth Syst Sci Data 15(12):5491\u0026ndash;5515\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYan G, Mas JF, Maathuis BHP, Xiangmin Z, Van Dijk PM (2006) Comparison of pixel-based and object‐oriented image classification approaches\u0026mdash;a case study in a coal fire area, Wuda, Inner Mongolia, China. Int J Remote Sens 27(18):4039\u0026ndash;4055\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang X, Qin Q, Grussenmeyer P, Koehl M (2018) Urban surface water body detection with suppressed built-up noise based on water indices from Sentinel-2 MSI imagery. Remote Sens Environ 219:259\u0026ndash;270\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Istanbul Technical University","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":"lake, shrinkage, Sentinel, Otsu, trend, uncertainty","lastPublishedDoi":"10.21203/rs.3.rs-8269709/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8269709/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eBurdur Lake, a Ramsar site in Turkey, exemplifies the global crisis of shrinking endorheic lakes under anthropogenic and climatic pressures. This study quantifies its surface area decline from 2018 to 2025 using Sentinel-2 imagery (10 m resolution), revealing a median shrinkage rate of -1.64 km\u0026sup2;/year (95% CI: -1.83 to -1.47 km\u0026sup2;/year), totaling 11.59 km\u0026sup2;. A Monte Carlo Simulation (MCS) framework, integrated with the non-parametric Sen\u0026rsquo;s Slope estimator, propagates classification uncertainty (\u0026plusmn;\u0026thinsp;1%, validated at 98\u0026ndash;99% accuracy), addressing a common gap in remote sensing time-series analysis. Adaptive Otsu thresholding outperformed conventional zero-thresholding (McNemar test, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05), ensuring reliable water-land delineation in this dynamic, saline basin. The decline, driven by dams and groundwater abstraction, mirrors trends in lakes like Urmia and Aral Sea, affecting biodiversity, including the endangered White-headed Duck. These findings provide a baseline for policy interventions, such as revised reservoir management and irrigation optimization to restore hydrological balance. This methodology offers an approach for monitoring lake dynamics, supporting water management and ecological conservation.\u003c/p\u003e","manuscriptTitle":"Quantifying Burdur Lake Shrinkage (2018–2025): Trend Analysis and Uncertainty Quantification with Sentinel-2 Imagery and Monte Carlo","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-04 08:47:06","doi":"10.21203/rs.3.rs-8269709/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":"7974093e-eca5-4c99-9fca-d229ae19aac9","owner":[],"postedDate":"December 4th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":59025184,"name":"Geographic Information Systems"}],"tags":[],"updatedAt":"2025-12-04T08:47:06+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-04 08:47:06","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8269709","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8269709","identity":"rs-8269709","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}