A multi linear regression approach to satellite based chlorophyll-a estimation for trophic status mapping of Lake Victoria

Abstract

Lake Victoria is both the largest and most important water resource in East Africa. It provides livelihood to more than 20 million people in Uganda, Tanzania and Kenya. However, the water in this lake is affected by nutrient loads (mainly phosphorus and nitrogen) from adjacent based activities such as industrialization and urbanization. This causes a health threat on both humans and aquatic animals consuming this water. As a result, there is a need for continuous monitoring of the water quality on this Lake. Traditionally, this is done through taking In Situ Chlorophyll a samples and further analyzing then in the laboratory. Considering the size of the lake, it is tiresome and impractical for such an approach given the limitations of time and cost. This research thus explored the use of Landsat 8 multi linear regression models in estimating chlorophyll a concentration on Lake Victoria. Samples of Log transformed In situ data collected in 2016 were used together with five Landsat 8 Bands (Coastal Aerosol, Blue, Green, Red and Near Infrared) to develop and study three multi linear regression models. Root mean Square, R 2 , Relative Percentage Difference and adjusted R 2 values for each model were computed and analyzed. The best multi-linear regression model (Log Chl-a = -129.899B1 + 17.364B2+56.355B3+29.501B4 +8.604) had RMSE, R 2 , RPD and AdjR 2 values of 0.8111, 0.7551, 3.15265 and 0.5102 respectively. Using this model, Chlorophyll a values were computed, and classified according to Carlson’s trophic status. The results showed that that Lake Victoria was facing water quality issues with most of the water belonging to the Eutrophic state.