Monitoring aquatic plants proliferation in Lake Victoria using satellite data

Abstract

This work reports the monitoring of aquatic plants proliferation in Lake Victoria using satellite data over the period 2003 to 2010. The lake, which is the second largest freshwater body in the world and an important economic resource, is facing serious environmental challenges including growth of invasive plants. The role of some selected water quality parameters as well as meteorological conditions in aquatic plants proliferation is also investigated. Multispectral MERIS (Medium Resolution Imaging Spectrometer) imagery, were obtained from ESA in the framework of TIGER Initiative. Images were selected on the basis of spatial and temporal coverage, spatial and spectral resolution and the severity of cloud cover. Atmospheric correction was carried out as pre-processing procedure to improve on the image interpretability. The images were processed using BEAM 4.8 and ENV1 4.2 image processing and analysis software. Image derived endmembers were used to classify the images using linear spectral unmixing classification technique. Temporal variation of vegetation was obtained, and the spatial distribution was presented by cover maps. Mapping was done using ArcGIS 9.3. TSM and Chl-a values were retrieved from the images using MERIS Eutrophic Lakes Processor 1.4.1 while rainfall data was obtained from Kenya Meteorological Department (KMD). Spectral unmixing technique performed well with a mean classification accuracy of 99.48%, based on RMSE. Most images showed a high concentration of Chl-a and TSM along the shores of the lake, especially the Winam Gulf, and most of the aquatic vegetation was observed in the same regions. Vegetation cover in the Winam Gulf which was kept below 100 km2 during the years 2003 to 2006 increased to a peak of about 200 km2 in 2007, before decreasing again to below 100 km2 during the years 2008 to 2010. Regression results for Winam Gulf show that vegetation abundance has a weak linear correlation with rainfall, TSM and Chl-a of R = 0.67 (R2 = 0.44), R = 0.46 (R2 =0.21) and R - 0.57 (R = 0.32), respectively after a response period of two to three months. From these relations, vegetation abundance prediction models are proposed. At no time delay, however, vegetation abundance showed no significant relationship with these parameters, while TSM and Chl-a are significantly dependent on each other with R = -0.77 (R2 = 0.6). While traditional methods of monitoring vegetation and water quality parameters is both expensive and time consuming, remotely sensed satellite data provides reliable, consistent and repeatable information that is suitable for this study.