干旱区生态环境知识资源中心

The sampling protocol adopted during a field campaign at an Alpine meadow site (Shandan site), during July 2002 is based on the so-called "Valeri" protocol (VALERI). The field campaign LAI measurements in Shandan are scaled up to 30x30 m(2) raster maps based on Landsat ETM+ imagery. Regression analysis is applied to construct empirical transfer functions for the determination of Leaf Area Index (LAI) raster imagery from ETM+ Normalized Difference Vegetation Index (NDVI) and Simple Ratio (SR) data. Subsequently, the scaling up of the LAI raster maps is performed by the aggregation of the 30x30 m(2) data into 1 x 1 km(2) pixels by calculating the average LAI values for the low resolution pixels. The up-scaled data are used to validate the MODIS LAI product at the Shandan site. A power regression model (LAI=2.3758*NDVI3.5216, R-2=0.66, P<0.01), established between field measured LAI and ETM+ NDVI, elicits a high statistical significance. A linear regression model (LA1=0.1798*SR-0.3574, R-2=0.55, P<0.01) is established between field measured LAI and ETM+ SR. The MODIS LAI product correlates best with the ETM+ LAI transfer function obtained with NDVI data. Its R-2 reaches 0.46, its slope 0.97, but the intercept is 0.7, which suggests that MODIS LAI is systematically underestimated. The results illustrate that LAI measured with a LAI-2000 instrument at the VALERI Shandan site leads to an underestimation of the MODIS LAI product. A plausible cause for the systematic underestimation related with the LAI field measurements is discussed.