H11A-0847:
A Linear Mixed Effect Model to Estimate Chla Concentration in Lake Erie from Satellite Data

Monday, 15 December 2014
Kiana Zolfaghari and Claude R Duguay, University of Waterloo, Waterloo, ON, Canada
Abstract:
Water quality parameters change frequently in time and space. Therefore, synoptic measurements of these changes are required in order to investigate potential water quality problems associated in lakes. However, obtaining in situ measurements of water quality parameters at sufficient temporal and spatial resolutions is a challenge. Considering the declining network of in situ measurements and availability of datasets across the globe, alternatives to conventional water quality measurements are necessary. Satellite remote sensing provides a mean to retrieve parameters related to lake optical properties for many lakes over large areas and with frequent temporal coverage. Chlorophyll-a (chla) concentration is the photosynthetic pigment available in all kinds of phytoplankton and can be addressed to monitor algal blooms and determine the eutrophication status. In this study, a linear mixed effect model (LME) is applied to MERIS satellite imagery to estimate the concentration of chla in Lake Erie, which is considered as Case II water (i.e. turbid and productive). The increasing phosphorus concentration in Lake Erie, especially in the western basin, stimulates occasional algal blooms. Hence, overall reduction in total phosphorus inputs to the lake is necessary. MERIS spectral bands are extended further in red and NIR wavelengths. These bands are required for band ratio algorithms to derive chla concentration over turbid lakes with complex optical properties, as the shorter wavelengths are affected by other optically active water constituents besides chla. The LME model considers the correlation in the repeated measurements of in situ data collected in space and time. This method is developed between the logarithmic scale of chla and the band ratio B7:665/B9:705. Cross validation resulted in RMSE of 0.30 for log10chla. The regionally tuned model was therefore reliable for application on a time series of images acquired over Lake Erie from 2004 to 2008. Results show that the offshore areas of the lake, especially the western basin, are experiencing high concentration of chla which intensifies in summer. The concentration of chla is the highest in western basin and decreases toward the central and eastern parts of the lake. The central basin of the lake experiences the highest concentration of chla in fall.