Assessing the impact of drought conditions on groundwater potential in Godavari Middle Sub-Basin, India using analytical hierarchy process and random forest machine learning algorithm

Severe drought conditions have affected ever increasing demand of water in Godavari Middle Sub-Basin of India. Thus, assessment of drought and its impact on groundwater potential zones is essential for effective management of water resource in the Sub-basin. We utilized site-specific factors for ascertaining the groundwater potential zones. Analytical hierarchy process (AHP) was used for assigning weights to the factors and preparing map of groundwater potential zones. One-month (meteorological), three and six months (agricultural) and twelve months (hydrological) droughts were determined using standardized precipitation index (SPI) during 1979-2013. Vulnerable drought zones were identified using weighted sum overlay analysis. Random forest (RF) algorithm was utilized for assessing the impact of drought conditions on groundwater potential zones. Validation of random forest was carried out by dividing groundwater potential and drought maps into training (80%) and testing (20%) datasets in Jupyter notebook Python IDE (Integrated Development environment). Findings revealed that nearly half of the area of the Sub-basin experienced low and very low groundwater potential. RF analysis revealed that the groundwater potential in the northern and southern parts of the Sub-basin was severely affected by drought during the study period. Random forest regression (R-2 being 0.858) indicated high accuracy of the model. Thus, RF model has proved to be an effective tool for analyzing relationship between drought and groundwater potential zones. The other drought prone areas interested to attempt drought and its impact assessment on groundwater potential may find this approach useful for policy measures.