Development of a Satellite-based Monitoring Approach for Augmenting Open-source Indian Lake Water Quality Datasets with Seasonal Information

Lake water quality monitoring in India faces a critical paradox—one where sub-daily or daily data needs are only met with sparse, annually available information. The largest publicly available water quality dataset in India is hosted by the Central Pollution Control Board (CPCB), which only provides annual maxima and minima for a few monitored quality parameters, providing no details on their intra-annual variability. To bridge this critical data gap, this analysis attempts to build a Satellite-based Monitoring approach, demonstrated for two India lakes- Lake Nainital, a source water body in Uttarakhand (0.438 km² area) and Lake Sukhna, a wastewater receiving water body in Chandigarh (1.38 km² area). Using Sentinel-2 imagery from 2016-2023, we reconstructed water quality values for 11 parameters of interest, including optically active (chlorophyll-a, turbidity, TSS, etc.) and optically inactive (electrical conductivity, fecal bacteria, BOD, etc.), derived on 3 separate grid sizes: 10m*10m, 20m*20m and 30m*30m. For Lake Nainital, lake quality was analysed for a 100m buffer zone around the water intake point and the following analyses were performed

(i) Seasonal Random Forest models were trained with CPCB ground-truth data, achieving promising predictive accuracy. In that, for Lake Nainital, winter served as the optimal period for nutritional monitoring with R² values exceeding 0.9, whereas temperature prediction was most accurate in the monsoon (R²=0.93).  Fecal coliform demonstrated remarkable accuracy in summer (R²=0.90), in stark contrast to its diminished performance during the monsoon (R²=0.80). Sukhna exhibited contrasting seasonal dependencies: temperatures peaked in summer (R²=0.81), while electrical conductivity spiked in winter (R²=0.90). Also, BOD prediction enhanced significantly from summer (R²=0.66) to winter (R²=0.91).

(ii) Using Modified Robust Principal Component Analysis (MRPCA) Lake Naintial successfully diagnosed a single-factor dominance to multi-stressor complexity, e.g., during the COVID-19 pandemic in 2020 anthropogenic pressures temporarily eased then resurged with altered patterns. Further, the chronic nutrient impairment of Lake Sukhna was also diagnosed using this approach.

 

The advantages of the proposed satellite-based lake monitoring approach are significant- allowing water treatment plant operators to seasonally forecast coagulant demand fluctuations. This novel satellite-to-tap approach demonstrates an alternative future for water quality monitoring-one which need not rely on extensive grab sampling or sensor-based data as inputs. It also allows regulatory monitoring of chronically impaired lakes of the country and monitoring the upkeep of restored and rejuvenated lakes.