Satellite Soil Moisture Data-Driven to Predict Paddy Production Using Machine Learning Approach

Abstract: Measuring and mapping how wet the soil is accurately are very important for farming in a sustainable way, managing water resources, and studying the climate. On top of that, it also important to find out that, based on soil moisture of a land, what could be the potential paddy production. This study aims to explore the mapping of soil moisture of Rajshahi district in Bangladesh using multi-spectral remote-sensing satellite images and based on historical paddy production data, what could be the possible paddy production in the year 2025. To find out the possible production in the year of 2025, I have used the Machine Learning techniques to predict the paddy production based on the data generated from the soil moisture map and integrated the data with the paddy production statistics data of Bangladesh Bureau of Statistics (BBS). A total of 14 Landsat scenes covers the entire Rajshahi district. Thus, a set of Landsat imagery (a total of 14 scenes) for the year 2017, 2020 and 2023 was used in this study to map the soil moisture of Bangladesh through the application of Geographical Information System (GIS) and Remote Sensing. Satellite Image preprocessing, correction, and analysis were done with the ArcGIS software (version 10.8, developed by Environmental Systems Research Institute, USA) and prediction was found out through Google Colab. The map in this study indicates the level of soil wetness in Rajshahi, ranging from Very Dry, Dry, Wet to Very Wet soils. Based upon the study, it is found out that, it is possible to increase paddy production rate by 47.95% in the year 2025 due to improvement of soil moisture conditions.