Pedological properties, including soil structure, and particle size distribution have been accepted as important factors controlling soil moisture. In the context of the increased climate change, local weather monitoring is crucial from the economic and agricultural point of view, as the classical meteo prediction schemes do not apply very precisely - this was our very first challenge. To find out which parameters affect the soil moisture levels the most. As the soil moisture levels are a seasonal dataset, the SARIMA model was deployed to test the seasonality in the data. for the same - We melted the soil moisture levels while taking the pm3 variable and its values. to improve the model's predictive performance we also calculated the first, second, and third differences in the moisture levels and used them for a comparative study.

Training an RFR pipeline was a tough and exhausting task, The slow learning rate due to the non-imputed null values was the 2nd challenge that we faced. For the same, we made an XGboost Imputed RFR Pipeline which resulted in a comparatively faster learning rate and even better accuracy - 99.7% with a Mean Squared error - 32.23%