Remote sensing of the water quality of shallow lakes: A mixture modelling approach to quantifying phytoplankton in water characterized by high-suspended sediment

Remote sensing has the potential to provide truly synoptic views of water quality, in particular, the spatial distributions of phytoplankton. Whilst the spectral capabilities of satellites used in ocean colour work have improved significantly over recent years, the application of satellite remote sensing to lake water is constrained by the need for high spatial resolution image data and thus remains limited by spectral resolution capabilities. This becomes a significant problem when attempting to quantify chlorophyll a (Chl a) in waters characterized by high and heterogeneous suspended sediment concentrations (SSC). The SSC dominates the spectral reflectance, masking the spectral influence from other components in broad spectral band systems, making Chl a determination from remote sensing imagery difficult. This paper presents a linear mixture modelling approach to derive accurate estimates of Chl a from Landsat Thematic Mapper (TM) imagery. This approach was tested in Lake Balaton, Europe's largest shallow lake characterized by high suspended sediment and, until recently, frequent eutrophic and hypereutrophic episodes. The last significant bloom occurred in September of 2000 and a Landsat TM image was acquired for 11th September, during which ground reference data of water quality was collected. The modelled image-derived results of Chl a demonstrate an excellent correspondence (r(2) = 50.95) between the ground-based measurements of Chl a, and yield considerable detail of lake phytoplankton distributions. The September 2000 calibration was then successfully applied to a July 1994 Landsat TM image and validated with Chl a data collected coincidently within two days of the image. The comparability between water sample data and image results demonstrates that there is temporal stability and robustness in the approach and calibration described.