Estimating Algae Concentration by Using LANDSAT 8 Images-A Case Study of Techi Reservoir in Taiwan

105

Taiwan is a tropical–subtropical transition island characterized by high temperatures and ample rainfall. It has more than 40 reservoirs that provide water for drinking, other everyday activities, and industrial use. Significant seasonal variation in annual rainfall complicates reservoir operation and affects water quality, as do some reservoirs’ upstream locations. During heavy rain, nutrient-rich fertilizers often flow into reservoirs and spur the growth of algae species. A high concentration of algae species may lower the water quality. Therefore, long-term and continuous monitoring of water quality in reservoirs is necessary, because the water quality in reservoirs directly affects the quality of the water supply. New remote sensing technology provides data on water body parameters through images that reflect a wide range of spatial and temporal changes. This thesis, with the Techi reservoir as the study area, discusses estimates of reservoir algae concentration according to Landsat 8 images. Linear regression (LR) and a genetic algorithm combining operation tree (GAOT) were used to establish the predictive models by using image data and in situ algae data. The GAOT is a data mining method used to automatically discover relationships among nonlinear systems. The results show that the GAOT performs more efficiently (CCs = 0.83, 0.8, 0.82 and low RMSE = 34.5, 23.32, 21.35 for Bacillariophyta, Chlorophyta, and Dinophyta, respectively) than does LR (CCs = 0.78, 0.74, 0.62 and low RMSE = 31.68, 21.87, 17.64 for Bacillariophyta, Chlorophyta, and Dinophyta, respectively).