Deepak KumarShomya Kumari2021-12-242021-12-242021-11https://krishikosh.egranth.ac.in/handle/1/5810179829Groundwater is an important natural freshwater reserve on which billions of habitants depend for their diverse utilization. Global water demand has far exceeded the total available water resources which in turn have put a serious concern on food security. India is one of the largest agricultural user of groundwater in the world where there has been a large scale revolutionary shift from surface water management to a widespread groundwater abstraction. Increased industrialization, rapid population growth, climate change, changes in the land use and land cover, has influenced the extensive use of groundwater which simultaneously affects the groundwater level. Groundwater drought occurs when this groundwater level falls below the critical level. In the present study, analysis of groundwater drought of the state of Bihar, India, has been carried using a drought index called Standardized Groundwater Index (SGI) and the spatial and temporal distribution of SGI has been reflected using Remote sensing and GIS approach. The rainfall and groundwater data of 38 districts of Bihar from 2002-2019 has been used and was divided seasonally into pre- monsoon, monsoon, post- monsoon and winter seasons. Further, SGI was modelled using Artificial Neural Network and Random Forest machine learning techniques with different input models. GRACE satellite water equivalent data along with rainfall and below groundwater level was used to predict SGI. Finally, the trend analysis of groundwater level data of 38 districts of Bihar for all the four seasons was studied using Mann- Kendall test statistics and Thein Sen's slope estimator. The results of SGI spatial and temporal distribution showed that districts like Aurangabad, Gaya, Buxar, Bhojpur, Kishanganj, Katihar, Kaimur, Rohtas, Nawada, Saran Chappra, Siwan, Samastipur, Supaul are prone to the critical groundwater drought condition. On comparing the performance of the two models to predict, SGI it was found that RF models performs superior than the ANN model with correlation coefficient value of (r) as 0.95. The trend analysis results showed that 45% of the districts are showing decline in the groundwater level particularly in pre-monsoon season.EnglishThesis