Upscaling In-Situ Leaf Area Index Measurements to Obtain the Representative Ground-Truth of the Heterogeneous Land Surface

Leaf area index (LAI) plays a significant role in the structure and the functioning of vegetation. Moderate-resolution remote sensed LAI products can be applied to monitor the crop growth, yield and global change. However, the products must be validated against in-situ observations before being used, which requires the upscaling of in-situ LAI to footprint scale of remote sensing. In this study, a two-step upscaling algorithm is developed to obtain the representative ground-truth LAI of heterogeneous surface based on continuous in-situ LAI measurements. The ASTER images are used to estimate LAI maps. First, the continuous upscaled vegetation LAI values can be obtained with a weighted combination of in-situ LAI values and the weight can be derived by Bayesian linear regression (BLR) algorithm. Secondly, the upscaled heterogeneous surface LAI can be calculated using the upscaled vegetation LAI multiplied by heterogeneous index. Our study area is located in Yingke irrigation districts of Heihe River basin, Gansu province of China and the corn LAI is measured at each point from June 25 to August 24, 2012 by newly designed LAINet. The result showed that RMSE of upscaled LAI declines with the increasing number of sampling points. Compared to the case that averaged LAI value of six points is used to represent this heterogeneous surface LAI (RMSE 0.9711), the RMSE of upscaled heterogeneous surface LAI values using the same six points falls to 0.035. It indicated that our upscaling algorithm is effective to upscale in-situ LAI values to obtain the representative LAI ground-truth of heterogeneous land surface with certain number of sampling points. Furthermore, this upscling method is used for upscaling in-situ LAI values to obtain the surface LAI in 1-2km. It suggests that upscaling method can make full use of the continuous in-situ LAI measurements.