Spatial–temporal changes in runoff and terrestrial ecosystem water retention under 1.5 and 2 °C warming scenarios across China

Abstract

Abstract. The Paris Agreement set a long-term temperature goal of holding the global average temperature increase to below 2.0 ∘C above pre-industrial levels, pursuing efforts to limit this to 1.5 ∘C; it is therefore important to understand the impacts of climate change under 1.5 and 2.0 ∘C warming scenarios for climate adaptation and mitigation. Here, climate scenarios from four global circulation models (GCMs) for the baseline (2006–2015), 1.5, and 2.0 ∘C warming scenarios (2106–2115) were used to drive the validated Variable Infiltration Capacity (VIC) hydrological model to investigate the impacts of global warming on runoff and terrestrial ecosystem water retention (TEWR) across China at a spatial resolution of 0.5∘. This study applied ensemble projections from multiple GCMs to provide more comprehensive and robust results. The trends in annual mean temperature, precipitation, runoff, and TEWR were analyzed at the grid and basin scale. Results showed that median change in runoff ranged from 3.61 to 13.86 %, 4.20 to 17.89 %, and median change in TEWR ranged from −0.45 to 6.71 and −3.48 to 4.40 % in the 10 main basins in China under 1.5 and 2.0 ∘C warming scenarios, respectively, across all four GCMs. The interannual variability of runoff increased notably in areas where it was projected to increase, and the interannual variability increased notably from the 1.5 to the 2.0 ∘C warming scenario. In contrast, TEWR would remain relatively stable, the median change in standard deviation (SD) of TEWR ranged from −10 to 10 % in about 90 % grids under 1.5 and 2.0 ∘C warming scenarios, across all four GCMs. Both low and high runoff would increase under the two warming scenarios in most areas across China, with high runoff increasing more. The risks of low and high runoff events would be higher under the 2.0 than under the 1.5 ∘C warming scenario in terms of both extent and intensity. Runoff was significantly positively correlated to precipitation, while increase in maximum temperature would generally cause runoff to decrease through increasing evapotranspiration. Likewise, precipitation also played a dominant role in affecting TEWR. Our results were supported by previous studies. However, there existed large uncertainties in climate scenarios from different GCMs, which led to large uncertainties in impact assessment. The differences among the four GCMs were larger than differences between the two warming scenarios. Our findings on the spatiotemporal patterns of climate impacts and their shifts from the 1.5 to the 2.0 ∘C warming scenario are useful for water resource management under different warming scenarios.

Sensitivity Analysis‐Based Automatic Parameter Calibration of the VIC Model for Streamflow Simulations Over China

“…According to SA results discussed above, the parameters we identified as sensitive parameters for modeling streamflow in the 10 major river basins did not match the seven default parameters (B, D 1 , D 2 , D 3 , Dm, Ws, Ds). Rather than using these default parameters in model calibration (Ran et al, 2017;Shi et al, 2008;Zhang et al, 2014), we can eliminate calibration of the unnecessary parameters (those resulting from type I error) and instead add the sensitive parameters missing from the default calibration procedure (type II error). So, how does the SA-based ASMO calibration framework avoid those two errors?…”

“…As for the Southwest River basin, it originates in the highest part of the Tibetan Plateau, which has the most varied topography in the world and which has a great amount of globalwarming-induced glacier melting augmenting the surface-runoff rate (Immerzeel et al, 2010). Despite the poor performance in arid and high-altitude basins, the current framework is still a computationally low-cost and effective method for parameter optimization as compared to manual calibration method for the seven default parameters applied in previous studies in China (e.g., Ran et al, 2017;Wang et al, 2012;Xie et al, 2007). Figure 6.…”

“…Streamflow modeling plays an important role in many practical hydrological problems (Sorooshian & Chu, 2013), such as water-related hazards forecasting , water resource management (Liu et al, 2018;Zheng et al, 2019), and management of ecosystem services (Wenger et al, 2010). Additionally, evidence from recent years has shown that the issue of global warming is becoming more serious in nature (IPCC, 2014;Niraula et al, 2017), and there is a need to use regional hydrological modeling tools to identify changes in the magnitude and timing of associated hydrological processes (Niraula et al, 2017;Ran et al, 2017;Wang et al, 2018). However, models are just approximations of natural systems, and there remains substantial discrepancy between model results and reality (Sorooshian & Chu, 2013).…”

“…VIC is able to capture transient basin discharge and has 46 or more tunable parameters (Bennett et al, 2018). Although we know that parameter sensitivities respond to the dynamic environment in a basin and automatic calibration performs better than manual calibration, the practice of manually calibrating the seven expert-recommended parameters (hereinafter referred to as the seven default parameters) is still largely used in VIC model calibration (Niraula et al, 2017;Ran et al, 2017;Wang et al, 2018;Xie et al, 2007;Zhang et al, 2014). The use of the expertrecommended parameters in the process of calibration precludes our ability to quantify the potential impact of uncertainties in parameters or climate data and consequently makes it impossible to quantify the associated uncertainties in model performance.…”

Evaluating the Impacts of Climate Change and Vegetation Restoration on the Hydrological Cycle over the Loess Plateau, China

“…Using a parameter dataset (i.e., six parameters with calibration) to characterize all grid cells of a catchment was a generalized description of the catchment features. Furthermore, parameter implementation, which was actually based on an idealized assumption for simulation in uncalibrated regions, was essential for hydrological simulation from catchment-scale up to region-scale [57,77]. In this study, though further validations were conducted in some catchments to ensure the reliability of parameter transplantation, and obtained a relatively satisfactory result, a potential uncertainty still existed, which may directly influence the model performance and call for improved parameterization.…”