变网格模式LMDZ4对东亚夏季气候的模拟检验

Simulation of the East Asian summer climate with LMDZ4—a variable resolution model

对法国动力气象实验室发展的变网格全球大气环流模式LMDZ4在东亚地区的模拟性能进行评估。LMDZ4作为海气耦合模式IPSL-CM5A的大气模块,本文局地格点加密区域位于东亚上空,加密区外使用ERA-interim再分析资料进行驱动,积分时段为1979—2009年。通过对模式在东亚东部区域夏季气候的模拟能力评估,发现模式整体上能够真实地模拟出南亚高压、西风急流、西太副高、水汽输送以及夏季风等环流系统的气候平均态,地面雨带和气温的分布及极值中心都得到较好再现。同时模式仍显现一定的偏差,具体表现为,华南沿海的模拟偏冷偏湿,江淮流域偏暖偏干,而东北地区则偏暖偏湿,导致这种地面要素场偏差的原因来自于中上层环流场的系统性模拟偏差。例如,高层南亚高压、西风急流模拟偏弱,而印度季风和来自印度洋的水汽输送偏强,导致华南沿海降水偏多。西太平洋副高偏东偏弱,低纬东风带强度偏弱,西太平洋区域比湿偏小,使得东南风带来的水汽无法到达江淮流域。上述偏差的产生可能由于LMDZ4模式中的云参数化方案尚存在不足之处,使得云量的模拟偏少,并且该版本模式的垂直分辨率不足,限制了它对高层环流系统的模拟能力,增加了模式的系统误差。

LMDZ4, a variable-resolution global atmospheric general circulation model,is assessed against observed data in Southeast Asia before its further utilization in climate change downscaling studies.The model is developed in the Laboratoire de Meteorologie Dynamique in Paris, as the atmospheric component of the ocean--atmosphere coupled system IPSL-CM5A.In order to investigate the ability of LMDZ4 in simulating East Asian summer precipitation, temperature and circulation, this paper analyzes the causes of simulation deviation and some of the uncertainties in this model.The study is expected to lay a foundation for further forecasting of future climate changes.In this study, LMDZ4 has its domain centered over Southeast Asia[ domain center at （30°N, 110°E） ;cover-age：（5-55°N, 85-135°E）], and is driven by ERA-Interim reanalysis data from 1979 to 2009. It is actually used as a traditional limited-area model with the whole globe as the buffer zone that receives the ERA-Interim data.The nudged variables include zonal wind,meridional wind,temperature and specific humidity,with a time interval of 6 h.There are 121×120（latitudexlongitude）grids and 19 layers in the vertical direction.The spatial resolution inside the domain is approximately 0.6°×0.6°.The driving forcing is added through a relaxation procedure. We impose a long（ 10 days,but not infinite,to be comparable with other simulations using this historical set-up） relaxation time scale inside the domain,while the relaxation time scale outside the domain is 1.5 h.The use of two distinct relaxation time scales inside and outside the domain enables the model to run almost freely inside the domain, whereas it completely follows the forcing outside the domain. For the purpose of model assessment, daily temperature and rainfall at a few meteorological stations in China are used.These two datasets are available with a resolution of 0.5°×0.5°. Results show that the mean state of the major atmospheric general circulation systems can be simulated well, including the South Asian High, westerly jet, western Pacific subtropical high, and moisture transport. The main characteristics of rainfall and surface air temperature are also successfully reproduced.Nevertheless ,the simulation shows some biases.Wet and cold biases are found over southern coastal areas,warm and dry biases over the Huaihe river basin, and warm and wet biases over northeast regions,which are related to the simulation deviations of mid-upper level circulations.For example, the simulated strengths of the South Asian high and westerly jet stream are too weak, while the water vapor transport from the Indian Ocean is stronger, leading to more rainfall over southern coastal areas.The missed rainfall belt along the Yangtze River Valley is due to a lack of moisture brought by southeast winds that are too weak, and is directly related to the weak western Pacific subtropical high, which leads to less moisture over western Pacific regions.The deviations mentioned above may largely be attributable to the cloud parameterization scheme in LMDZ4, which yields insufficient cloud cover. In addition, the coarse vertical resolution in this model version is also thought to be a factor contributing to some of the imperfections in upper-level circulation,and to model biases.As we use only one regional climate model driven by one set of reanalysis data, we cannot fully explore the uncertainties of this model in terms of climate changes for precipitation and temperature.It is clear that more research is needed to assess these uncertainties.We need to include as many different types of downscaling model（ e.g. ,dynamical and statistical downscaling）as possible when generating climate-change information at the local scale.