藏北典型高寒草甸地上生物量的遥感估算模型

摘要

Normalized difference vegetation indices (NDVI), enhanced vegetation indices (EVI) and land surface water indices (LSWI) were calculated based on MYD09A1 during 2010 - 2011 to estimate the aboveground biomass (ANPP) in the Northern Tibet Plateau using regression analysis. Air temperature, air moisture, soil temperature and soil moisture were compared in 2010 and 2011 when there was a great difference in aboveground biomass during the two years. Correlation analysis between aboveground biomass and vegetation indices showed that single factor regression model would be sufficient to reflect the relationship between aboveground biomass and vegetation indices. Regression analysis showed that both NDVI and EVI could be used to estimate aboveground biomass, however LSWI performed relatively poor. The simple linear regression model best reflected the relationship between vegetation indices and aboveground biomass while the power fitting was not quite as good as linear regression. The relationships between NDVI, EVI and aboveground biomass can be expressed by the linear regression equation: ANPP=115.478 NDVI (R2 = 0. 956, P<0. 001) and ANPP=174. 225 EVI (R2=0. 975, P<0. 001). Results of verification showed that the power model had a better estimating capability than the linear model. When the power model was used to estimate ANPP, the relative error was 9. 76% compared with 10. 8% for the linear model. This difference demonstrated that the power model may be more robust in estimating the ANPP.%本研究利用2010和2011年6-9月高寒嵩草草甸群落地上生物量数据和同期的中分辨率成像光谱仪(moderate resolution imaging spectroradiometer,MODIS)影像数据,建立了西藏自治区拉萨当雄高寒草甸的地上生物量遥感估算模型.在探讨群落地上生物量与归一化植被指数(normalized difference vegetation indices,NDVI)、增强型植被指数(enhanced vegetation indices,EVI)以及地表水分指数(land surface water indices,LSWI)相关关系的基础上,构建了群落地上生物量与各指数的线性或非线性(包括对数函数、二次多项式、三次多项式、幂函数、增长曲线、指数函数)估算模型,并进行了模型验证.结果表明,1)3个指数中EVI的模拟效果最好,NDVI次之,LSWI最差；2)所有模型中线性模型的模拟效果最好,EVI的线性模型为:y=174.225x,R2=0.975,P＜0.001,NDVI的线性模型为:y=115.478x,R2=0.956,P＜0.001;3)在预测效果方面,幂函数模型最好,线性模型次之,其平均误差分别仅为9.76％和10.8％,因此,两者都可以用于高寒草甸群落地上生物量的模拟.