2. Thesis and Dissertations
Permanent URI for this communityhttps://idr.nitk.ac.in/handle/1/10
Browse
Search Results
Item Assessment of Future Transition in Climate Extremes Over Western Ghats ff india Using Machine Learning Based Multi-Model Ensemble Techniques(2024) SHETTY, SWATHI; U, PRUTHVIRAJBetter hazard management in the future requires a ramifications of climate change on water resources, with particular emphasis on the regional scale. The rapid urbanization and industrialization, combined with the drastic changes in the land use and land cover of the region over the Western Ghats (WG), have driven regional heterogeneity in the climate. Key factors such as rainfall, temperature, topography, and vegetation are crucial in unraveling the intricate interactions between ecosystems and climate systems. The comprehension of forthcoming variability in these factors holds significant importance for the region owing to its global significance. The future climate projections rely on the Global Circulation Models (GCMs) ; thus, it is crucial to make sure those GCMs are accurate representations of the current climate in the region. Therefore, the study aimed to a) understand the role of topographic structure on rainfall distribution and its association with topo-climatic variables, and the vegetation, b) rank the GCM models and examine the efficacy of advanced machine learning based ensemble techniques to capture the inter-seasonal temporal variability over diverse geo-climatic basins of ghat, c) examine the uncertainties in multi-model ensembles of GCMs to capture the extreme climate indices and their trend d) model the potential occurrence of severe minimum and maximum temperatures, rainfall events, potential evapotranspiration, and historical and projected trends and e) understand the impact of climate change on the future variability in the streamflow. The dependability of rainfall on the topography and climate of the region is evaluated using the Geographically Weighted Regression method. It is observed that the effect of the terrain is amplified in the broad, gradually sloping intermediate rough mountain located in close proximity to the coast. The maximum amount of rainfall is contingent upon the steepness of a mountain’s windward side and the topographic structure resulting in the difference in the elevation of maximum rainfall occurrence. Based on this, the six river basins located in the i diverse-geoclimate of theWestern Ghat are used to evaluate the performance of the GCMs and to understand the future variability in climate and the extremes. The top-performing GCMs obtained from Technique for Order Preference by Similarity to an Ideal Solution were ensembled using simple Arithmetic Mean (AM) and seven machine learning-based ensemble methods. Further, its ability to imitate extreme climatic events is analyzed using the indices formulated by the Expert Team on Climate Change Detection and Indices. Then the frequency and trend in the projected extremes of precipitation, minimum and maximum temperature, are obtained for the Shared Socioeconomic Pathways SSP245 and SSP585 for the Near Future (2021–2050) and Far Future (2051–2100) horizon. The streamflow of the river basins is simulated using Long Short Term Memory (LSTM), a deep learning technique to assess the potential impact of climate change on streamflow. The performance of individual GCM models varies in all the basins; also, the ability to imitate the observation varies with the climatic variables, with notable disparities in the simulation of climate patterns. The ensemble of top-performing models has been proven beneficial in river basin scale by overcoming the constraints of bias correction methods. The Multi-model ensemble (MME) of Extreme Gradient Boosting (XGBR) and Random Forest Regression stand out for their superior performance across all river basins, with exceptional performance over the per-humid basins, while Adaptive Boosting, Support Vector Regression, and the AM underperform. Despite excellent accuracy in predicting daily/monthly rainfall, still, a great deal of variability in calculating climatic indices is noted, with higher relative bias in precipitation indices. Except for the duration-based precipitation indices, the XGBR calculated indices have been shown to be more accurate across all basins. The anticipated fluctuations in temperature emphasize the onset of increased warming in November, which extends up to June, resulting in a notably warmer winter and an extended summer season. In future decades, warm days and nights increase by 45–65% and 45–70% in Aghanshini and northern river basins, and 45-85%, 60-80% in southern and Netravati river basins receptively, with two fold ii warming in the winter season. After the mid-21st century, the warming trends start to slow down with decreasing trends in the pre-monsoon maximum temperature in southern and central river basins and a decrease in the monsoon minimum temperature in the northern river basins. The June and July rainfall will be highly inconsistent in the future decades, with a substantial increase in very wet to extremely wet days and medium to heavy rainfall in northern river basins. The streamflow in the monsoon season decreases substantially, with a decrease in annual streamflow in Chaliyar and Netravati and converse in other river basins. The southern river basins and the Netravati river basins are extremely vulnerable to water scarcity risk in May and June months, which extends to April and July in the high emission scenarios. These findings serve as an indication of the range of anticipated changes in the magnitude of extreme maximum and minimum temperature, rainfall, and geographical pattern over the Western Ghats.