Remote Estimation of Nutrients for a Drinking Water Source Through Adaptive Modeling

Morse Reservoir, a major water supply for the Indianapolis metropolitan area, IN, USA, experiences nuisance cyanobacterial blooms due to agricultural and point source nutrient loadings. Hyperspectral remote sensing data from both in situ and airborne AISA measurements were applied to an adaptive model based on Genetic Algorithms-Partial Least Squares (GA-PLS) by relating the spectral signal with total nitrogen (TN) and phosphorus (TP) concentrations. Results indicate that GA-PLS relating in situ spectral reflectance to the nutrients yielded high coefficients of determination (TN: R ²?=?0.88; TP: R ²?=?0.91) between measured and estimated TN (RMSE?=?0.07 mg/L; Range: 0.6–1.88 mg/L), and TP (RMSE?=?0.017; Range: 0.023–0.314 mg/L). The GA-PLS model also yielded high performance with AISA imaging data, showing close correlation between measured and estimated values (TN: RMSE?=?0.11 mg/L; TP: RMSE?=?0.02 mg/L). An analysis of in situ data indicated that TN and TP were highly correlated with chlorophyll-a and suspended matter in the water column, setting a basis for remotely sensed estimates of TN and TP. Spatial correlation of TN, TP with chlorophyll-a and suspended matters further confirmed that remote quantification of nutrients for inland waters is based on the strong association of optically active constituents with nutrients. Based on these results, in situ and airborne hyperspectral remote sensors can provide both quantitative and qualitative information on the distribution and concentration of nutrients in Morse Reservoir. Our modeling approach combined with hyperspectral remote sensing is applicable to other productive waters, where algal blooms are triggered by nutrients.