Separating climate change and human activities' effects on flow regime with hydrological model error correction

Abstract

Climate change and human activities have markedly altered flow regimes, leading to deleterious consequences for riverine ecosystems. The widely employed hydrological modeling restoration method for conducting flow regime attribution often tends to underemphasize the impact of simulation errors on attribution outcomes. Moreover, the issue of ambiguous attribution targets in flow regime attribution remains a critical concern. This study presented a flow regime attribution framework integrated the hydrological model error correction. Indicators of hydrologic alteration (IHA) simulated by the XAJ and GR4J hydrological models and corrected using the quantile-mapping (QM) and convolutional neural network-long short term memory (CNN-LSTM) error correction methods, were employed to create 6 scenarios for assessing the impacts of climate change and human activities on flow regime changes. A case study was conducted in the Xitiaoxi basin, located in the upper reaches of Lake Tai in China. The results demonstrated that CNN-LSTM's superior performance over QM in correcting IHA metrics simulated by both XAJ and GR4J. The mean flow regime attribution results obtained from XAJ and GR4J simulations following the application of CNN-LSTM correction, mitigated the underestimation of contributions from the primary driving factors, in contrast to uncorrected or QM-corrected IHA metrics attribution results. Additionally, the regime attribution findings revealed all 33 IHA metrics experienced changes, with 73 % of them attributing human activities as the predominant driving forces. Substantial variations in IHA metrics exerted a certain degree of pressure on riverine habitats, particularly impacting the spawning and reproductive activities of piscine species.

Keywords

• Flow regime
• IHA
• Climate change
• Human activities
• Hydrological model error correction
• CNN-LSTM