Development of a Composite Eutrophication Proxy Index (CEPI) Using Sentinel-2 GeoAI Time-Series for Lake Toba

Abstract

Lake Toba faces increasing environmental pressure due to eutrophication, characterized by excessive algal growth and declining water quality. Conventional monitoring based on in situ measurements has spatial and temporal limitations, making it difficult to fully capture the spatial and temporal dynamics of eutrophication in large lake environments. Therefore, an efficient and sustainable alternative approach is needed for spatial-temporal water quality assessment. This study proposes the development of a Composite Eutrophication Proxy Index (CEPI) based on Sentinel-2 imagery and GeoAI as a preliminary bio-optical proxy framework for eutrophication-related monitoring under limited field-data conditions. The analysis utilized Sentinel-2 Level-2A imagery from 2021 to 2025 within a 100-meter shoreline buffer zone. The CEPI was constructed by integrating normalized NDCI, AFAI, and TSS components. Temporal analysis, K-Means clustering, Isolation Forest anomaly detection, and statistical evaluation were conducted to assess the spatial-temporal consistency and behavior of the index. CEPI values ranged from 0.30 to 0.75, with an average value of 0.50, indicating relative mesotrophic-like conditions. Seasonal fluctuations were observed, with higher CEPI values generally occurring at the end of the rainy season and the beginning of the dry season. Spatial clustering identified relative CEPI zones concentrated around floating net cage aquaculture areas. Strong correlations were observed between CEPI and its constituent bio-optical components, particularly AFAI (r = 0.88) and NDCI (r = 0.81). Temporal stability analysis produced a coefficient of variation (CV) of 0.029, indicating relatively stable temporal behavior. Internal consistency assessment yielded R² values of 0.997 for training data and 0.930 for temporal test data, indicating temporal coherence within the CEPI framework. The results indicate that CEPI provides a consistent spatial-temporal proxy for eutrophication-related dynamics in Lake Toba and may support scalable monitoring approaches for tropical lakes under data-limited conditions.