Abstract This study develops a new method for selecting the most representative climate model for use in the Soil and Water Assessment Tool (SWAT) and evaluates this method in the Maumee River Basin. Downscaled temperature and precipitation data from 32 Coupled Model Intercomparison Project Phase 5 (CMIP5) models were used to drive the SWAT. Our model selection approach uses a sensitivity analysis to evaluate the impacts of temperature and precipitation on streamflow and to determine the optimal weights for both variables. This weighting scheme is then used to rank the performance of all climate models and identify which ones are most representative of climate forcing required by the hydrological model. The results of our sensitivity analysis indicate that the response of streamflow to precipitation is consistent throughout the year, while the response to temperature is more complex. During the warm season, higher temperatures cause evapotranspiration to increase and streamflow to decrease; during the cool season, higher temperatures cause more snowmelt and this causes streamflow to increase. Based on our weighting scheme, the GISS-E2-R was identified as the most representative climate model in the Maumee River Basin. We then compared the performance of SWAT-simulated streamflow using all CMIP5 models. The most representative climate model ranked first based on correlation and mean absolute error, and second based on Nash-Sutcliffe efficiency, while the model selected using a simple average method was ranked fifth, fourth and twelfth, respectively. Based on the evaluation of other key variables, such as sediment transport, nitrogen and phosphorus, the optimal climate model also has a higher correlation and a more representative standard deviation than the other CMIP5 models. The selected model also performs better than the 32-model ensemble mean based on correlation and Taylor’s skill score. When applying this method at different time scales, the responding time of streamflow to climate factors needs to be considered. The model selection method that was developed in this paper can be applied by hydrological modelers to select the most representative climate models for evaluating the impacts of climate change in other watersheds around the world.

中文翻译：

---

为水文建模优化气候模型选择：使用 SWAT 在 Maumee 河流域进行的案例研究

摘要 本研究开发了一种选择最具代表性的气候模型以用于土壤和水评估工具 (SWAT) 的新方法，并在 Maumee 河流域评估了该方法。来自 32 个耦合模型比对项目第 5 阶段 (CMIP5) 模型的缩减温度和降水数据用于驱动 SWAT。我们的模型选择方法使用敏感性分析来评估温度和降水对流量的影响，并确定这两个变量的最佳权重。然后使用该加权方案对所有气候模型的性能进行排序，并确定哪些模型最能代表水文模型所需的气候强迫。我们的敏感性分析结果表明，河流流量对降水的响应全年是一致的，而对温度的反应则更为复杂。在温暖的季节，较高的温度导致蒸发量增加和流量减少；在凉爽的季节，较高的温度会导致更多的融雪，从而导致流量增加。根据我们的加权方案，GISS-E2-R 被确定为莫米河流域最具代表性的气候模型。然后，我们使用所有 CMIP5 模型比较了 SWAT 模拟水流的性能。最具代表性的气候模型根据相关性和平均绝对误差排名第一，根据 Nash-Sutcliffe 效率排名第二，而使用简单平均法选择的模型分别排名第五、第四和第十二。基于对其他关键变量的评价，如泥沙输移、氮和磷，最佳气候模型也比其他 CMIP5 模型具有更高的相关性和更具代表性的标准偏差。根据相关性和泰勒技能得分，所选模型的性能也优于 32 模型集成平均值。在不同时间尺度应用该方法时，需要考虑径流对气候因素的响应时间。水文建模者可以应用本文开发的模型选择方法来选择最具代表性的气候模型，以评估气候变化对世界其他流域的影响。在不同时间尺度应用该方法时，需要考虑径流对气候因素的响应时间。水文建模者可以应用本文开发的模型选择方法来选择最具代表性的气候模型，以评估气候变化对世界其他流域的影响。在不同时间尺度应用该方法时，需要考虑径流对气候因素的响应时间。水文建模者可以应用本文开发的模型选择方法来选择最具代表性的气候模型，以评估气候变化对世界其他流域的影响。