Estimating water quality through neural networks using Terra ASTER data, water depth, and temperature of Lake Hachiroko, Japan

In recent years, the need to reduce water pollution and improve environmental water quality has increased. In this study, we estimated the spatial distribution of suspended solids (SS) and the nitrogen-to-phosphorus (NP) ratio as water quality parameters by combining three types of information: satellite remote sensing data, water depth, and water temperature. A water quality estimation method using a neural network was also developed. The proposed method is effective and easy to apply as it does not use many parameters. The results showed that the maximum improvements in the SS and NP ratio estimates compared to the results of the fuzzy regression analysis and the conventional method were 6 mg/L and 2.25, respectively. In the SS estimation, the learning dataset based on texture dissimilarity helped improve the accuracy. The proposed method will contribute to a more detailed understanding of water quality conditions.

Automated summary

In this study, a water quality estimation method using a neural network was developed, combining satellite remote sensing data, water depth, and water temperature. The results showed significant improvements in the accuracy of estimating suspended solids and the nitrogen-to-phosphorus ratio compared to the conventional methods, providing a more detailed understanding of water quality conditions.