Analysis of Influence of Land Use Land Cover and Climate Changes on Streamflow of Netravati Basin, India

Abstract

Massive Land Use/Land Cover (LULC) change is a result of human activities. These changes have, in turn, affected the stationarity of climate, i.e., climate change is beyond the past variability. Studies indicate the effect of LULC change and climate change on the hydrological regime and mark the necessity of its timely detection at watershed/basin scales for efficient water resource management. This study aims to analyse and predict the influence of climate change and LULC change on streamflow of Netravati basin, a tropical river basin on the south-west coast of India. For future climate data, researchers depend on general circulation models (GCMs) outputs. However, significant biases exist in GCM outputs when considered at a regional scale. Hence, six bias correction (BC) methods were used to correct the biases of high-resolution daily maximum and minimum temperature simulations. Considerable reduction in the bias was observed for all the BC methods employed except for the Linear Scaling method. While there are several BC methods, a BC considering frequency, intensity and distribution of rainfall are few. This study used an effective bias correction method which considers these characteristics of rainfall. This study also assessed and ranked the performance of 21 GCMs from the National Aeronautics Space Administration (NASA) Earth Exchange Global Daily Downscaled Projections (NEX-GDDP) dataset and bias-corrected outputs of 13 Coupled Model Inter-comparison Project, Phase 6 (CMIP6) GCMs in reproducing precipitation and temperature in the basin. Four multiple-criteria decision-making (MCDM) methods were used to identify the best GCMs for precipitation and temperature projections. For the CMIP6 dataset, BCC-CSM2-MR was seen as the best GCM for precipitation, while INM-CM5-0 and MPIESM1-2-HR were found to be the best for minimum and maximum temperature in the basin by group ranking procedure. However, the best GCMs for precipitation and temperature projections of the NEX-GDDP dataset were found to be MIROCESM-CHEM and IPSL-CM5A-LR, respectively. Multi-Model Ensembles (MMEs) are used to improve the performance of GCM simulations. This study also evaluates the performance of MMEs of precipitation and temperature developed by six methods, including mean and Machine Learning (ML) techniques.The results of the study reveal that the application of an LSTM model for ensembling performs significantly better than models. In general, all ML approaches performed better than the mean ensemble approach. Analysis and mapping of LULC is essential to improve our understanding of the human-nature interactions and their effects on land-use changes. The effects of topographic information and spectral indices on the accuracy of LULC classification were investigated in this study. Further, a comparison of the performance of Support Vector Machine (SVM) and Random Forest (RF) classifiers was evaluated. The RF classifier outperformed SVM in terms of accuracy. Finally, the classified maps by RF classifier using reflectance values, topographic factors and spectral indices, along with other driving factors are used for making the future projections of LULC in the Land Change Modeler (LCM) module of TerrSet software. The results reveal that the area of built-up is expected to increase in the future. In contrast, a drop in forest and barren land is expected. The SWAT model is used to study the impacts of LULC and climate change on streamflow. The results indicate a reduction in annual streamflow by 2100 due to climate change. While an increase in streamflow of 13.4 % is expected due to LULC change by the year 2100 when compared to the year 2020. The effect of climate change on streamflow is more compared to LULC change. A reduction in change is seen in the streamflow from near to far future.