The purpose of this study was to use Moderate Resolution Imaging Spectroradiometer (MODIS) chlorophyll data to identify the predominant spatial and temporal patterns in chlorophyll variability in Lake Erie. Three algorithms were tested against in situ chlorophyll measurements: O’Reilly’s OC2 and OC3 algorithms (1998) and Cannizzaro’s shallow water algorithm (2005). These algorithms can be calculated from atmospherically corrected reflectance data distributed by NASA; algorithms that required data without atmospheric corrections were not considered. The initial regression results showed no correlation for any of the algorithms tested. However, outliers for all three algorithm regressions were consistently missing reflectance values at one or more wavelengths. Removing pixels that had missing reflectances at any wavelength, even those not required to compute the algorithm, greatly improved algorithm performance. This result supports the hypothesis that a more advanced correction procedure for atmospheric scattering that produced more reliable reflectance values for turbid inland waters would improve the performance of chlorophyll retrieval algorithms for Lake Erie.
While none of the algorithms were valid in Lake Erie’s turbid western basin, OC3 performed best in the central and eastern basins, with an R2 of 0.56 and RMSE of 0.73. OC3 chlorophyll concentrations were calculated for all the available 2002-2007 non-winter MODIS images of Lake Erie. Principal component analysis (PCA) was applied to the resulting time series, which extracted patterns of seasonal variability from the data set. The central basin showed more seasonal variability than the eastern basin, with elevated chlorophyll concentrations in spring and fall.