Evaluation Of AOCGCMs For Precipitation In East Asia

Mainly based on ECMWF ERA40 reanalysis data from 1958 to 1999, ground-based observation data in recent 50 years, and CPC Merged Analysis of Precipitation from 1979 to 1999, and other related data from different origins, the simulating abilities for precipitation in East Asia of current Atmosphere-Ocean Coupled Global Circulation Models (AOCGCMs) are evaluated. Firstly, the primary climate elements simulated by the models in global scale are simply analyzed. And secondly, the following research are focused on evaluation of simulating ability for precipitation in East Asia, which mainly includes the climate pattern of precipitation and the rain band seasonal process, the beginning and end of rainy season, the extreme heavy rain, and temporal multi-scale characters of precipitation in China. And based on the above analysis, the observed and simulated climate change and abrupt jumping in Qinghai-Tibet Plateau are all analyzed. Finally, the projected climate changes in future 50 to 100 years based on the SRES emission scenarios are provided. The mainly results are as follows:(1) The analysis of several primary climate elements simulated by AOCGCMs show that, most of the models are able to capture the main climate patterns of the elements in global scale accompanying with some biases. Generally, most models have better results in the Northern Hemisphere than in the Southern Hemisphere, and the largest biases of the analyzed elements are often around the South Pole, except for zonal wind. The areas with larger precipitation biases rates are often over the bigger mountains or complex surface. Most models underestimate surface air temperature, intensity of circulation, geo-potential height, specific humidity in most areas of the mid-low latitude in the Northern Hemisphere. Based on the comparison of the results for each element, the follow models are found to be with much better simulating abilities. Those models are CCSM3, CGCM3.1(T47), CGCM3.1(T63), ECHAM5-MPI-OM, GFDL-CM2.0, GFDL-CM2.1, MIROC3.2(hires), MIROC3.2(medres), UKMO-HadCM3, UKMO-HadGEM1, which are generally with mid-high resolutions and also have relative better simulating ability for precipitation in East Asia. (2) The analysis about precipitation in East Asia show that, these AOCGCMs are able to reproduce the precipitation in East Asia in some extent, which are represented in the following aspects: 1) most of the models can reproduce the basic precipitation pattern that precipitation decreases from the ocean in the southeastern part of East Asia to the inland of China in the northwestern part of East Asia; 2) most of the models also can generally capture the advance of rain belt in eastern China. However, there are still considerable differences, which are 1) the climatic mean values and variability of rainfall in most models are much less than observation; 2) there is bias in the simulation of advance and retreat of rain belt, especially over the ocean, and some models even cannot capture the basic seasonal progress over the eastern ocean. Therefore, the current CGCMs have limited capability to simulate precipitation in East Asian monsoon areas, and need to be further improved. The multi-model ensemble shows that the simulated circulation intensity in summer is weaker and precipitation is lesser than observation, especially in the eastern part of China. Furthermore, moisture content is lower and thus moisture transport is weaker, which may be one reason that results in the underestimate of summer precipitation in East Asian monsoon areas in the models.(3) Most models are able to obtain the basic character of rainy season in China, which is start from the south China and then advance northward and westward, meantime the rainy season of northwest also start and advance toward southern and eastern region. However, the simulation for rainy season in the models also exist some problems, which include that : 1)the abrupt increase and decrease of precipitation are very difficult to be grasped by most models; 2) In eastern China, the precipitation values are generally be underestimated, that the start of rainy seasons are ahead of observation while the end of rainy seasons are lagged and the seasonal process of precipitation are slowed down, which result in the rainy season generally longer than observation; 3) In the western China, most models overestimated the precipitation, and the result about start of rainy season in northwestern China is earlier than observation. MIROC3.2(hires) as a model with higher resolution show some advantage in the simulation about the rainy season process in southern China, and it is able to basically grasp the three peaks in Yangtze River Basin and two peaks in south China, but the exact times still also have some biases.(4) The models overestimate the rainy days of flurry and moderate rain, but underestimate the rainy days of heavy rain and rainstorm. The trend analysis for heavy rain show that although some models can reflect the feature in some region of north China, mid-low reaches of Yangtze River basin and south China, no model can reproduce the general characters in the three regions and even in the total eastern China. The thresholds of extreme heavy rain in the models are all lower than observation, and even no one model can capture the extreme heavy rain trend pattern in eastern China. Such results imply that the future trends of extreme heavy rain projected by the models based on some SRES emission scenarios maybe have high uncertainties. Generally, the simulation abilities for extreme heavy rain of the current AOCGCMs should be greatly improved.(5) The evaluation for different temporal scale variability of precipitation show that, most models are able to capture the signals of interannual and decadal scale periods, while the intensity of the signals and especially the positive and negative phase for each scale period are all great different between simulation and observation, which maybe result in the poor phase consistency in precipitation. Further analysis show that simulation abilities of the models for sea surface temperature variability (ENSO and PDO) still need to be greatly improved, and the current models still could not reproduce the relationship between the PDO and precipitation in eastern China, which maybe finally affect the precipitation by some physical process, especially the interaction between atmosphere and ocean in the AOCGCMs. The biases of the contribution of 10-30-day and 30-60-day oscillation between simulations and observation are smaller than that of other oscillation in intra-seasonal scales. As a model with higher resolution, MIROC3.2(hires) show some advantages, especially in the result about 30-60-day oscillation in eastern China.(6) Most models underestimate the surface and tropospheric air temperature in Qinghai-Tibet Plateau. The difference of the biases between land (Qinghai-Tibet Plateau and the around region) and ocean (the east and south ocean areas) maybe affect the monsoon circulation intensity. Most models overestimate the precipitation in Qinghai-Tibet Plateau, especially in the south and southeast side of the plateau. The models with higher resolutions do better than the lower ones in the simulation of the trends of precipitation change in Qinghai-Tibet Plateau. Although some models can capture the climate abrupt jumping features in some extent, no model is able to reproduce the observed abrupt jumping character in the whole China. (7) In the future 50 to 100 years, the surface and tropospheric air temperature will continuously increase in East Asia and China, accompanying with the decrease of upper stratospheric air temperature. In SRESA1B, the mean result of Multi-model shows that the precipitation in northern China will have been increased in the first ten years in21st century, and the precipitation anomalies will be negative in southern China and positive in northern China in 2010s and 2030s, while negative in southwestern China in 2020s, and then generally positive since 2040s in the eastern China. In the future, the days with extreme heavy rain will increase and the amplitude in 2081-2100 is larger than that in 2046-2065, which has much high consistency among the models.