A Deep Convolutional Neural Network Model to Predict Rainfall and Flood Using Sentinel Images

Abstract

With the shift in the Earth's axis of rotation, there is drastic change in the climate which causes significant movement of the tectonic plates. One such change is the pattern of rainfall, which, among other things, has a direct bearing on the frequency of floods. Different models and algorithms assist in estimating the pattern of rainfall today, which helps prevent floods. We present a model to investigate the pattern of rainfall and flood occurrences using real-time meteorological variables and satellite photos. We can forecast the areas that are likely to have flash floods by examining real-time satellite images. Our model estimates the likelihood of a flood based on the water body's path, depth, and the layout of the city it runs through. Using Random Forest and convolutional neural networks, it is also possible to anticipate the pattern of rainfall and floods based on meteorological variables such as temperature, humidity, pressure, and others. We have created a model that can reliably forecast the likelihood that a flood will occur in a specific city at a specific time. This forecast would assist the government in handling the problem prophylactically and warning those who might be in danger.