利用MODIS增强型植被指数反演草地地上生物量

利用2002年5～9月MODIS增强型植被指数(EVI)、NDVI及同期地上生物量(ANPP)资料,分别按照5个月份、4种草地类型、2种植被指数,建立了VI-ANPP的线性模型和幂模型.研究结果显示,地面样地采样时间对模型影响较大,7月份模型相关性高,6,9月份相关性低.按照4种草地类型,草甸和典型草地模型相关性高,沙地和荒漠模型相关性低.按照植被指数,月份EVI-ANPP的相关性比NDVI-ANPP的相关性普遍提高,尤其是草地已经开始衰退的9月份;但草地类型EVI-ANPP的相关性只有在草甸和荒漠这两种草地中有明显改善,而在典型草地和沙地草地中相关性降低.按照不同模型类型,幂函数模型比线性模型的相关性普遍提高,只有当线性模型已经很好的草甸才例外.如果只使用5,7,8三个月的样地建立VI-ANPP模型,则相关性普遍提高.NDVI-ANPP线性模型的R2由原来的0.49增加为0.59,NDVI-ANPP幂模型的R2由原来的0.54增加为0.71;EVI-ANPP线性模型的R2由原来的0.52增加为0.63,NDVI-ANPP线性模型的R2由原来的0.55增加为0.73.在这种普遍增长趋势下,只有荒漠草地的EVI-ANPP线性模型的R2由原来的0.54降低为0.46.

ANPP estimate from MODIS-EVI for the grassland region of Xilingol, China

Above-ground Net Primary Productivity (ANPP) of grasses in Xilingol, Inner Mongolia, China, were correlated over May-September 2002 to Enhanced Vegetation Indices (EVI) derived from 250/500 m visible and near-infrared bands of the Moderate Resolution Imaging Spectroradiometer (MODIS). This article discussed the potentials and limits of MODIS-EVI over MODIS-NDVI by 5 months, 4 grasslands and 2 kinds of models. Temporally, July was the optimal time to estimate ANPP from vegetation indices, with a R =0.548 3 for linear EVI-ANPP relation and a R2 =0.5436 for linear NDVI-ANPP relation; June and September were less suitable for establishing the VI-ANPP model. Month by month, all 5 EVI-ANPP models had higher R2 than correspond NDVI-ANPP models. Spatially, linear NDVI-ANPP models worked best for typical steppe and meadow steppe, with R2= 5056 and _R2=0.5647 respectively. By contrast, linear EVI-ANPP models worked best for meadow steppe and desert steppe, with R2= 6334 and _R2=0.5403. With regards to the different models, power-function models had a generally closer VI-ANPP relation than its linear counterparts, the overall R2 =0.5492 for power-function EVI-ANPP, better than R2 =0.5184 for linear EVI-ANPP model. To improve the models, only site data from May, July, and August were used to re-establish the models. The resulting models were improved impressively. R for linear EVI-ANPP increased from 0.5184 to 0.6250; R for power function EVI-ANPP increased from 0.5492 to 0.7324. More research is needed to investigate the inter-year stability of these models.