A multivariate analysis and machine learning approach to assess the impact of climate change on water quality

This study proposes a systematic method to investigate the impact of climate change on water quality using multivariate analysis and machine learning. This approach is applied in the Upper Guadiana Basin (UGB), a semi-arid region in central Spain, by analyzing historical temperature and precipitation data from 1980 to 2018 alongside future climate projections obtained from the MIROC-ESM-CHEM global climate model under the CMIP5 framework. We evaluate long-term trends across four representative concentration pathways (RCPs): 2.6. 4.5. 6.0. and 8.5, covering a period from 2019 to 2100. Given that water quality is a critical concern in this region, understanding how climate variability influences water quality is crucial for sustainable resource management. To analyze changes in key water quality parameters, Hierarchical Clustering (HC) and Principal Component Analysis (PCA) were employed to explore temporal patterns and their correlation with climate variables. Moreover, machine learning techniques were employed to improve forecasting accuracy through ARIMA (Auto Regressive Integrated Moving Average) modeling and RCP-based projections covering the period 2019–2100. The ARIMA model captured trends based on historical temperature data, while the RCP-based projections incorporated simulated changes in temperature and precipitation. Results show a significant warming trend, especially under high-emission scenarios (RCP 6.0 and 8.5), along with increased inter-annual variability in precipitation. Considering these climatic shifts, we methodically explore the maximum detectable influence of temperature and precipitation on selected water quality parameters. These findings highlight the necessity of adaptive water management strategies that consider climate variability and extreme events, showing the intricate relationships between climate drivers and water quality.