In recent years, different models to simulate urban climate variables have been applied to various cities. As the model outputs are usually validated individually, this raises the question about which urban climate model to choose for what specific purpose. The present study aims to find answers to this by intercomparing air temperature outputs of four different urban climate models that have been applied in the city of Bern, Switzerland. This includes a geostatistical land use regression model and the numerical models MUKLIMO_3, PALM, and FITNAH 3D. Using data from 70 stations of an urban air temperature measurement network, we intercompare the four models by focusing on the modeled urban air temperature variability. MUKLIMO_3 outputs show a weak urban air temperature variability, while strong small-scale temperature gradients are modeled by FITNAH 3D. PALM outputs are the only ones that reproduce the impact of a large-scale ventilation pattern, but have a large positive bias. The most accurate estimates of the urban air temperature variability are obtained from the land use regression model. For future applications of urban climate models, we reinforce the need of validation with in-situ measurements, since the outputs (and subsequent policies) depend substantially on the selection of the model.

中文翻译：

---

多种选择 - 四种不同城市气候模型的相互比较


近年来，模拟城市气候变量的不同模型已应用于各种城市。由于模型输出通常是单独验证的，这就提出了一个问题，即为什么特定目的选择哪个城市气候模型。本研究旨在通过比较瑞士伯尔尼市应用的四种不同城市气候模型的气温输出来找到答案。这包括地统计土地利用回归模型以及数值模型 MUKLIMO_3、PALM 和 FITNAH 3D。使用来自城市空气温度测量网络 70 个站点的数据，我们通过关注建模的城市空气温度变化来相互比较四个模型。MUKLIMO_3 输出显示较弱的城市空气温度变化，而较强的小尺度温度梯度由 FITNAH 3D 建模。PALM 输出是唯一再现大规模通风模式影响的输出，但具有较大的正偏差。城市气温变化的最准确估计值是从土地利用回归模型中获得的。对于城市气候模型的未来应用，我们强调了通过原位测量进行验证的必要性，因为输出（和后续政策）在很大程度上取决于模型的选择。