基于CSCS改进CASA模型的中国草地净初级生产力模拟

将草原综合顺序分类系统（CSCS）中的热量指标（∑θ）和湿润度指标（K）引入CASA模型。利用该模型模拟了2004-2008年中国41个草地类的净初级生产力（NPP），并分析了其时空变化和不同草地类NPP变化。结果表明：2004-2008年中国草地NPP模拟平均值与实测平均值分别为503.8 g·m^-2·a^-1和567.3 g·m^-2·a^-1，两者较为接近。各类草地的平均误差和平均相对误差均值分别为4.85 g·m^-2·a^-1和7.6%。草地NPP的实测值和模拟值相关性较好。改进CASA模型模拟值比Miami和Thornthwaite Memorial模型模拟值更接近实测值。NPP空间分布呈东高西低，南高北低，从西北向东南逐渐增加的趋势，体现了K和∑θ的水平和垂直地带性分布规律。2004-2008年中国草地NPP总体呈现增加趋势，其总量增加了23.0%。草地NPP年均值在不同植被类型中差异显著，分布规律与CSCS划分草地类的K和∑θ密切相关。总之，改进后的CASA模型模拟精度较高，实现了草地NPP模拟与草地分类的相互关联。

Estimation of Grassland Net Primary Production in China with Improved CASA Model Based on the CSCS

In this study, >0 ℃ average annual cumulative temperature (∑θ) and moisture index (K) in grasslands comprehensive and sequential classification system (CSCS) were introduced into Carnegie Ames Stanford Approach (CASA) model. The net primary production (NPP) of 41 grassland classes in China from 2004 to 2008 was estimated by the improved model. In addition, the temporal and spatial variations of NPP and NPP variations of different grassland classes were analyzed. The results showed that the average values of estimation data of the improved model and the observed data were 503.8 g·^-2·a^-1and 567.3 g·m^-2·a^-1, indicating the difference between them was small. The mean error and mean relative error of 41 grassland classes were 4.85 g·m^-2·a^-1 and 7.6%. There were significant correlations between observed data and estimation data. The estimation data of the improved model was closer to the observed data than that of Miami model and Thornthwaite Memorial model. The grassland NPP in east was higher than that in west, and that in south was higher than that in north. It increased from Northwest China toward Southeast China, which may show the characteristic of horizontal and vertical distribution of ∑θ and K. The trend of grassland NPP in China from 2004 to 2008 was increased and it increased by 23% totally. The variations of annual average NPP in different grassland classes were significant in China from 2004 to 2008, which was consistent with the ∑θ and K for determining grassland class in CSCS. In conclusion, the precision of the improved CASA model is enough for the estimation of grassland NPP. The mutual relation of the classes in CSCS with the simulation of grassland NPP has been completed to a certain extent in this study.