On detecting signals of anthropogenic climate change in basin-scale hydrologic variables

Anthropogenic climate change is one of the most pressing global environmental challenges faced by society in recent times. Large-scale shifts are observed in precipitation and runoff patterns due to the impact of anthropogenic climate change on the regional water cycle. Understanding these impacts is critical for effectively managing and protecting water resources and for mitigating the impacts of climate change. However, the detection of the impact of anthropogenic climate change on the regional water cycle is challenging. A pattern correlation analysis using fingerprints can be carried out to evaluate the impacts of human-induced climate change. Fingerprints give the expected direction of the anthropogenic signal and help to reduce the detection problem to a univariate or low-dimensional problem. This study adopts a formal fingerprint-based detection method to analyze the trends in monsoon precipitation and streamflow in the Krishna River basin, India. In-situ observations and several climate model outputs are utilized for the analyses. Principal component analysis, statistical downscaling techniques, and an Artificial Neural Network (ANN) based rainfall-runoff model are employed. The fingerprint detection method is illustrated using three scenarios by altering the anthropogenic forcings: aerosols alone, land use alone, and a combination of greenhouse gases with aerosols. The hydrologic variables considered are the gridded monthly monsoon precipitation data for 1951-2005 at 1° latitude by 1° longitude and monthly monsoon streamflow at a downstream gauging station. Leading Empirical Orthogonal Functions (EOFs) and signal strength are used to compare the response pattern of observed hydrologic variables with the response pattern simulated by climate models, including various forcings. The hypothesis that the observed trend in hydrologic variables lies within the range expected from natural internal variability alone is validated at a 95% statistical confidence level for most of the climate models considered. This excludes the possibility of other causal factors, including solar irradiance, volcanic eruption, and other anthropogenic impacts. It is found that the signal of human influence is less distinct from that of natural variability. Hence, it is concluded that applying a formal fingerprint-based method is not fully successful in detecting anthropogenic trends in hydrologic variables at the basin scale. The results emphasize the need for robust observational data and advanced analytical techniques considering a detailed process understanding.