A Semi-Analytical Model for the Multitemporal Prediction of Chlorophyll-a in an Iowa Lake Using Hyperion Data

The aim of this study was to use an analytical approach to monitor water quality in an Iowa lake using multitemporal Hyperion satellite imagery. Cloud-free hyperspectral images were acquired from the Hyperion sensor on individual days in June, July, and August of 2006. Water samples with accurate locations using GPS were collected simultaneously with image acquisition. Water samples were analyzed for various water quality constituents. Chlorophyll-a (CHL), was estimated for each sampling date using a bio-optical model with Specific Inherent Optical Properties (SIOPs) of the lake and light field variables derived from a radiative transfer numerical model. The model was then applied to the Hyperion images to create spatially continuous CHL maps for the study area every month. These results were compared with traditional linear regression model outputs. Maps produced using the bio-optical model effectively demonstrated spatial and temporal variability of CHL for the lake. The CHL concentration from this model across the lake ranged from 28 to 121μg/L for the month of June, 18 to 111μg/L for July, and 31 to 125μg/L for August. The validation and accuracy assessment for the bio-optical model with in-situ data showed R2 values of 0.90978, 0.96794 and 0.93057 for June, July, and August, respectively, and Nash-Sutcliffe coefficient values of 0.499478, 0.733072, and 0.878757 for June, July, and August, respectively.