Effect of urbanisation & population density on groundwater in India using satellite remote sensing data
MetadataShow full item record
This work investigates the relationship of urbanization, population density and meteorological variables (temperature & precipitation) on groundwater storage for the selected study regions in India. Variations in groundwater storage are analyzed using Gravity Recovery and Climate Experiment (GRACE) derived variations of Terrestrial Water Storage (ΔTWS). In the first part of the thesis, we have examined and established the correspondence between GRACE ΔTWS and groundwater level for selected study sites across different geographic regions in India using Linear regression, Support Vector Regression and Artificial Neural Network models. It has been observed in our study that ΔTWS is a highly significant predictor of GWL and the amount of variation in GWL that could be explained with the help of ΔTWS varies from 36.48% to 74.28%. Next, we have studied changes in ΔTWS across India from January 2003 to January 2017 and have found evidence of its significant declining trend (−0.912 ± 0.455 cm/year) in the northern part of India encompassing Ganga-Brahmaputra river basin and North-West India. As ΔTWS serves as a strong indicator for groundwater storage, its declining trend implies significant depletion of groundwater in this belt during this period. Interestingly, for the same time period, this particular belt with declining ΔTWS has observed a significant positive trend in precipitation and no significant trend for temperature. Also, for the mentioned time period, we’ve observed a higher growth rate in agricultural electricity consumption and population density in this region compared to the rest of India. These observations strongly suggest that the depletion of TWS in this region could be primarily attributed to anthropogenic activities rather than to changes in meteorological variables. Motivated by this observation, we investigate further the relationship between ΔTWS & urbanization. To measure the temporal changes in urbanization, we’ve developed an index PB1BI (Powered B1 Built Up Index) to classify pixels from Landsat7 images. We have compared the performance of PB1BI with existing indices and machine learning methods (SVM & ANN). Validation with manually verified test pixels and qualitative assessments indicate that PB1BI outperforms existing indices and it’s performance matches with that of ML methods. In addition to the existing Otsu thresholding method, this study has proposed a thresholding method using bootstrapping. For all the indices compared test pixel, accuracy measures show that the bootstrapping method works better than the Otsu method. Further, the multi-temporal analysis conducted in this study has demonstrated consistent performance of PB1BI. To reduce misclassification of “river sand” pixels as “built-up” ones while classifying “built-up” pixels from Landsat7 satellite imagery, we’ve developed another index-based methodology BRSSI (Built-Up & River Sand Separation Index) to separate these two land cover types. The performance of this index which is computationally inexpensive is comparable to that of the support vector machine. Both quantitative & qualitative assessment for the effectiveness of the developed methodology confirms a significant reduction of the misclassification. Finally, to understand the relationship of the urbanization & population density with ΔTWS, panel data regression analysis was conducted for 9 selected study sites across different geographic locations in India for the period 2003-2017. The newly developed algorithms (PB1BI & BRSSI) have been applied jointly to compute the percentage of urbanization from Landsat7 imagery. Population density, precipitation and temperature along with urbanization, have been used as explanatory variables in the panel data regression for understanding the variations in ΔTWS. Results suggest that precipitation & urbanization exhibit significant positive & negative effects respectively with ΔTWS and together they could explain 66.93% of the variability in the data. Similarly, it has been observed that interaction effect of urbanization & population density exhibit a significant negative association with GRACE ΔTWS and 77.76% of the variation in ΔTWS could be explained with the help of the same along with precipitation. This indicates the significant effect of an increase in anthropogenic indicators like urbanization & population density on the groundwater storage.