This paper provides a review of paleoclimate modeling activities in China. Rather than attempt to cover all topics, we have chosen a few climatic intervals and events judged to be particularly informative to the international community. In historical climate simulations, changes in solar radiation and volcanic activity explain most parts of reconstructions over the last millennium prior to the industrial era, while atmospheric greenhouse gas concentrations play the most important role in the20 th century warming over China. There is a considerable model–data mismatch in the annual and boreal winter temperature change over China during the mid-Holocene [6000 years before present（ka BP）], while coupled models with an interactive ocean generally perform better than atmospheric models. For the Last Glacial Maximum（21 ka BP）, climate models successfully reproduce the surface cooling trend over China but fail to reproduce its magnitude, with a better performance for coupled models. At that time, reconstructed vegetation and western Pacific sea surface temperatures could have significantly affected the East Asian climate, and environmental conditions on the Qinghai–Tibetan Plateau were most likely very different to the present day. During the late Marine Isotope Stage 3（30–40 ka BP）, orbital forcing and Northern Hemisphere glaciation, as well as vegetation change in China, were likely responsible for East Asian climate change. On the tectonic scale,the Qinghai–Tibetan Plateau uplift, the Tethys Sea retreat, and the South China Sea expansion played important roles in the formation of the East Asian monsoon-dominant environment pattern during the late Cenozoic.
The Regional Eta-coordinate Model（REM） has performed well in forecasting heavy rainfalls in China in recent years.A four-dimensional variational assimilation system（4DVar） is developed to improve the forecast skill of the REM.The tangent linear model and adjoint model codes are written according to the＂code to code＂rule,and the establishment of the REM adjoint modeling system is introduced in detail in this paper.The tangent linear and adjoint models of the REM are validated against the observational data,and so is the gradient of the given cost function.It is shown that for the tangent linear model and cost function,when the magnitude of perturbations is reduced,the verification results approach 1.0;when the rounding error of computer is increased,the verification results depart off 1.0.In the validation of the adjoint model,the values on the left- and right-hand sides of the algebraic formula are equal with 13-digit accuracy.These results indicate that the tangent linear model and the adjoint model system of the REM are successfully coded,and the gradient of the cost function is correctly calculated.By using the REM adjoint modeling system,two 4DVar experiments and extended forecasts are performed using observational data for two real cases in June 1998 and August 2000.The results show that forecasts of temperature,wind speed, and specify humidity using the 4DVar-assimilated initial data are all improved at the end of the forecast period.However,the performance of the 4DVar in forcasting rainfall is different in these two cases.The prediction of location and amount of the accumulated rainfall is well improved in the first case,while in the second case the prediction has no significant improvement.The problem may result from the fact that the observational data used in the 4DVar for the second case are inadequate.This case will be studied further in future work.
