Convolutional Encoder-Decoder with LSTM Approach-Based Groundwater Potential Mapping Using Remote Sensing and GIS

In recent years, the mapping of ground water capacity has grown more and more dependent on Remote Sensing (RS) and Geographic Information System (GIS). Recent groundwater potential mapping methods have made use of GIS and remote sensing, but they still have issues like effectively recording complicated geology and topography. The study proposes a convolutional encoder-decoder with Long Short Term Memory (LSTM) approach to overcome these issues. Due to the complicated geology and terrain of the research location, Thiruvallur district, it was challenging to map the potential of ground water using conventional techniques. The characteristics from the input data were extracted using a convolutional encoder-decoder system, and the temporal relationships among these characteristics were modelled using an LSTM. This work uses RS, GIS, and deep learning to determine the water shortage and analyze the water index for the Thiruvallur district. The research's findings demonstrated that the convolutional encoder-decoder with LSTM technique could precisely map the study area's ground water capacity with accuracy of 99.9%. The method was capable of capturing the intricate connections between the various data layers, and the model's usage of the LSTM made it possible to accurately represent the temporal dependencies between these levels. According to this study, there are five categories for groundwater quality: very high, high, moderate, low, and very low. Ultimately, the work showed the possibility for mapping ground water capability utilizing remote sensing and GIS data employing a convolutional encoder-decoder with LSTM method. By contrasting the detection rate of deep learning and GIS, the efficiency of the suggested technique is assessed. The strategy may be used in other regions with complicated geology and terrain where conventional approaches would not be successful.