Water Supply Forecasting Model in the Chi River from Satellite Images

In the present, water resource management is important for the livelihoods of people living in river basin areas. One of the guidelines for water resource management is a water supply forecast which is a prediction of streamflow volume expected to pass a particular location in a stream during a specified season. The objective of developing the water supply forecasting model is to predict the water volume in the Chi River from satellite images. Only the clear view images of the Chi River basin from Sentinel-2A and Sentinel-2B are processed. Then, k-means clustering is used to identify between water and non-water surface areas on satellite images. After that, data cleaning is provided by image segmentation for separating only Chi River areas from other areas. Blue pixels (water surface areas) are also counted to estimate water volumes. Next, normalization is applied to the data to make data to be the same scale. Last, the data is processed by Gaussian Process Regression (GPR) to develop the water supply forecasting model as a graph that predicts trends in water volume levels. The results of this study indicate that the amount of water trend predicted graph generated by GPR can forecast the near-future water volume trends and analyze potential water situations that could occur.