Padmanava Dash

Cyanobacteria represent a major harmful algal group in fresh to brackish water environments. Lac des Allemands, a freshwater lake of 49 km2 southwest of New Orleans, Louisiana, provides a natural laboratory for remote characterization of cyanobacteria blooms because of their seasonal occurrence. This dissertation makes a contribution to research methodology pertaining to atmospheric correction of satellite data and development of remote sensing algorithms to quantify cyanobacterial pigments.

The Ocean Color Monitor (OCM) sensor provides radiance measurements similar to Seaviewing Wide Field-of-View Sensor (SeaWiFS) but with higher spatial resolution. However, OCM does not have a standard atmospheric correction procedure and the comprehensive suite of atmospheric correction procedures for ocean (or lake) is not available in the literature in one place. Atmospheric correction of satellite data over inland lakes, estuaries and coastal waters is also challenging due to difficulties in the estimation of aerosol scattering accurately over these optically complex water bodies. Thus an atmospheric correction procedure was developed to obtain more accurate spectral remote sensing reflectance ( Rrs ) over Lac des Allemands from OCM data based on NASA's extensive work for SeaWiFS. Since OCM was not well calibrated, a new vicarious calibration procedure was also developed to adjust OCM radiance values to SeaWiFS radiance as SeaWiFS is well calibrated over its entire life.

Empirical inversion algorithms were developed to convert the OCM Rrs at bands centered at 510.6 and 556.4 nm to concentrations of phycocyanin (PC), the primary cyanobacterial pigment. For the algorithms to be uniformly valid over all areas (or all bio-optical regimes) of the lake, a holistic approach was developed to minimize the influence of other optically active constituents on the PC algorithms. Similarly, empirical algorithms to estimate chlorophyll a (Chl a) concentrations were developed using OCM bands centered at 556.4 and 669 nm. The best PC algorithm (R^2=0.7450, p<0.0001, n=72) yielded a root mean square error (RMSE) of 36.92 μg/L with a relative RMSE of 10.27%, and a mean absolute error (MAE) of 21.79 μg/L with a relative MAE of 6.06% (PC from 2.75 to 363.50 μg/L, n=48). The best algorithm for Chl a ^2=0.7510, p<0.0001) produced an RMSE of 31.19 μg/L, with relative RMSE = 15.70% and a MAE of 16.56 μg/L, with relative MAE = 8.33% (Chl a from 9.46 to 212.76 μg/L, n=48). The results demonstrate the preliminary success of using the 360x 236 m resolution OCM data to map cyanobacterial blooms in a small lake. While more field data are required to further validate the long-term performance of the algorithms, at present the algorithms may be implemented to process OCM data in an automated setup to provide timely information on the lake's bloom conditions. Similarly, retrospective processing may provide a long-term time series of bloom characteristics to document potential trends. The applicability of the algorithms can be extended to other lakes after necessary testing.

Ref: PhD Dissertation, Louisiana State University, May 31, 2011

Back to Publications