Water quality and freshwater ecosystems are affected by river discharge and temperature. Models are frequently used to estimate river temperature on large spatial and temporal scales due to limited observations of discharge and temperature. In this study, we use physically based river routing and temperature models to simulate daily discharge and river temperature for rivers in 138 basins in Alaska, including the entire Yukon River basin, from 1990–2021. The river temperature model was optimized for ice free months using a surrogate-based model optimization method, improving model performance at uncalibrated river gages. A common statistical model relating local air and water temperature was used as a benchmark. The physically based river temperature model exhibited superior performance compared to the benchmark statistical model after optimization, suggesting river temperature model optimization could become more routine. The river temperature model demonstrated high sensitivity to air temperature and model parameterization, and lower sensitivity to discharge. Validation of the models showed a Kling-Gupta Efficiency of 0.46 for daily river discharge and a root mean square error of 2.04°C for daily river temperature, improving on the non-optimized physical model and the benchmark statistical model, which had root mean square errors of 3.24 and 2.97°C, respectively. The simulation shows that rivers in northern Alaska have higher maximum summer temperatures and more variability than rivers in the Central and Southern regions. Furthermore, this framework can be readily adapted for use across models and regions.

中文翻译：

---

来自耦合和优化物理模型的阿拉斯加河流量和温度的高分辨率每日后报（1990-2021）

水质和淡水生态系统受到河流流量和温度的影响。由于对流量和温度的观测有限，模型经常用于估计大空间和时间尺度的河流温度。在本研究中，我们使用基于物理的河流走向和温度模型来模拟 1990 年至 2021 年阿拉斯加 138 个流域（包括整个育空河流域）的河流的每日流量和河流温度。使用基于代理的模型优化方法对无冰月份的河流温度模型进行了优化，提高了未校准河流水位的模型性能。使用与当地气温和水温相关的通用统计模型作为基准。与优化后的基准统计模型相比，基于物理的河流温度模型表现出优越的性能，这表明河流温度模型优化可以变得更加常规。河流温度模型对气温和模型参数化表现出较高的敏感性，而对流量的敏感性较低。模型验证显示，每日河流流量的 Kling-Gupta 效率为 0.46，每日河流温度的均方根误差为 2.04°C，改进了非优化物理模型和具有均方根的基准统计模型误差分别为 3.24 和 2.97°C。模拟显示，阿拉斯加北部的河流比中部和南部地区的河流夏季最高气温更高，且变化更大。此外，该框架可以轻松适应跨模型和区域的使用。