基于多模式集合的陆面蒸散发研究

As the exchange of water and energy between land surface and atmosphere, terrestrial evapotranspiration plays a key role in many fields such as hydrology, climate, agriculture and ecology. Compared with in-situ observation and remote sensing retrieval, land surface modeling is more skillful at characterizing the spatial and temporal variability of terrestrial evapotranspiration. However, there are large differences among models' simulations due to their diverse parameterizations. Bayesian Model Averaging (BMA) method can fully consider the performance of each model, and is able to dramatically increase the accuracy of the estimate. In this study, based on both offline land surface models and coupled ocean-land-atmosphere models, and with the Bayesian Model Averaging (BMA) method, multimodel ensemble scheme of terrestrial evapotranspiration will be built. More reliable terrestrial evapotranspiration estimate in China during the past 60 years will be derived through offline land surface modelings, while climate forecast of terrestrial evapotranspiration in China will be available via multiple coupled models in CMIP5. The above multimodel ensemble scheme is capable of providing a reasonable estimate of terrestrial evapotranspiration so as to profoundly characterize its temporal and spatial variability in China during the past 60 years and its response to different climate scenarios.

陆表蒸散发是陆-气间水量和能量交换的纽带，对水文、气候、农业、生态等众多领域具有重要影响。与站点观测和遥感反演相比，陆面过程模型在刻画蒸散发的时空变率方面具有明显优势。然而不同的物理参数化方案导致了陆面模拟结果差异较大。相关研究表明，基于贝叶斯模型平均(BMA)的多模式集合方法能够充分考虑各模式的性能优劣，从而显著提高估算的准确性。因此本项目拟以离线的陆面模型以及耦合的海-陆-气模型为模式框架，以贝叶斯模型平均(BMA)为方法基础，建立基于多模式集合的陆面蒸散发估算方案。一方面结合离线的陆面模型结果获得中国区域过去近60年的高质量蒸散发估算数据，另一方面结合CMIP5的多个耦合模式结果获得中国区域未来近100年较为可靠的蒸散发预测数据，进而科学地揭示中国区域过去近60年蒸散发的时空变化强度和规律，并且较为合理地刻画未来高、中、低排放情景下蒸散发的可能响应特征。

陆面蒸散是连接地表能量循环与水量循环的纽带。然而目前陆面蒸散的地表观测数据较少，同时缺乏足够的空间代表性。数值模拟能够提供时空连续的陆表蒸散，但是模式间不同的参数化方案导致模拟结果具有较大的差异。多模式集合能够有效地降低模拟结果的不确定性。本研究中我们开展了中国区域陆面蒸散发的数值模拟与评估工作，分析了模拟结果之间的差异。另外，分析了影响陆面蒸散的关键因子太阳短波辐射和大气向下长波辐射的不确定性，在此基础上结合常规的气象和卫星数据发展了一个适用于全球的陆面蒸散模型。