Development and application of a remote sensing-based salinity prediction model for a large estuarine lake in the US Gulf of Mexico coast

SummarySalinity in estuaries is highly variable due to river discharge, tidal motion, and winds. Information on the spatial and temporal changes in salinity can provide important ecological indications, but accurate monitoring of the space–time variability for a large estuary is often costly and time-consuming. This study applied remote sensing techniques to develop a salinity prediction model for Lake Pontchartrain, a large estuarine lake located in the Northern Gulf of Mexico, USA. “Ground truth” salinity was measured along two transects across the lake and near the shoreline. Water-leaving reflectance from the measurement locations was extracted from Landsat Thematic Mapper (TM) images pre-processed through “banding” noise reduction and radiometrical correction approaches. Ordinary least square and ridge regression methods were performed to identify model parameters and to determine relationships between salinity and reflectance. Salinity in the lake on eight dates was predicted with the developed model. Difference in salinity level and patterns, and impacts of Hurricanes Katrina and Rita on salinity were assessed with ANOVA and Fuzzy Similarity methods. The results showed that the model achieved a high power in prediction of the lake salinity (R2 = 0.89 and RMSE of validation = 0.27). Reflectance from TM bands 1, 2, and 4 was positively correlated to salinity levels and explained 1.9%, 20.3%, and 10.2% variance in salinity levels. Reflectance from bands 3 and 5 was negatively correlated to salinity and explained 34.1% and 31.2% variance. Under normal circumstances without the impacts of hurricanes, the lake salinity presented two patterns with average salinity level of 5.5 ppt. After Katrina’s landfall, the average was significantly increased by 1.1 ppt and the spatial patterns were altered. The pattern on 30 August 2005 was the most dissimilar one as compared to the two normal patterns, and then followed by the patterns on 9 and 25 October, and 7 September 2005. This study demonstrated that satellite remote sensing techniques can be applied to monitor salinity in coastal environments, and that freshwater discharge not only affects salinity levels and patterns under normal conditions but also is crucial for the return of salinity patterns to normal conditions after hurricane disturbance.