Abstract Climate change effects can have significant impacts worldwide. Extreme events can modify water availability and agricultural production, making climate change planning an essential task. The National Institute for Space Research (INPE in Portuguese) in Brazil has made a large dataset of regional climate model outputs (simulations and projections) available, which opens up many possibilities of carrying out high-resolution climate change studies. However, there is still no performance evaluation of the model-derived rainfall output against high-resolution ground-based observation data considering the Brazilian biomes. This paper attempts to fill this gap and evaluates the simulated precipitation throughout Brazil. We used gridded observed precipitation data and historical climate simulations from the Model for Interdisciplinary Research on Climate, version 5 (MIROC5) and from the Hadley Center Global Environment Model, version 2 (HadGEM2-ES), which were downscaled by the Eta RCM (Regional Climate Model). For the overlapping period (1980–2005), there is good agreement (PBIAS up to 10%) of downscaled annual simulations for the Amazon and Cerrado biomes and large biases (reaching 40%) in the Pampa biome, compared to the observations. Our results showed that HadGEM2-ES is capable of representing long-term mean monthly precipitation for large areas well, such as the Amazon and Cerrado. Furthermore, the Eta RCM has considerably improved the driving GCM MIROC5 simulations. In conclusion, we recommend using the HadGEM2-ES simulations for the Amazon, Eta/HadGEM2-ES for the Atlantic Forest, Cerrado, and Pampa, and Eta/MIROC5 for the Caatinga and Pantanal. Our study provides an overview of two downscaled simulation datasets in Brazil that may help verify the models' suitability for further climate change assessments.

中文翻译：

---

巴西 Eta/HadGEM2-ES 和 Eta/MIROC5 降水模拟的性能评估

摘要 气候变化的影响可以在世界范围内产生重大影响。极端事件可以改变可用水量和农业生产，使气候变化规划成为一项必不可少的任务。巴西国家空间研究所（葡萄牙语为 INPE）提供了大量区域气候模型输出（模拟和预测）数据集，这为开展高分辨率气候变化研究开辟了许多可能性。然而，考虑到巴西生物群落，仍然没有针对高分辨率地面观测数据对模型得出的降雨输出进行性能评估。本文试图填补这一空白并评估整个巴西的模拟降水。我们使用了来自跨学科气候研究模型第 5 版 (MIROC5) 和哈德利中心全球环境模型第 2 版 (HadGEM2-ES) 的网格观测降水数据和历史气候模拟，这些数据由 Eta RCM（区域气候模型）。在重叠时期（1980-2005 年），与观测结果相比，亚马逊和塞拉多生物群落的缩小年度模拟和潘帕生物群落的大偏差（达到 40%）具有良好的一致性（PBIAS 高达 10%）。我们的结果表明，HadGEM2-ES 能够很好地代表亚马逊和塞拉多等大面积地区的长期平均月降水量。此外，Eta RCM 大大改进了驱动 GCM MIROC5 模拟。总之，我们建议对亚马逊使用 HadGEM2-ES 模拟，Eta/HadGEM2-ES 用于大西洋森林、塞拉多和潘帕，Eta/MIROC5 用于卡廷加和潘塔纳尔湿地。我们的研究概述了巴西的两个缩小规模的模拟数据集，这些数据集可能有助于验证模型对进一步气候变化评估的适用性。