High-resolution Leaf Area Index estimation from synthetic Landsat data generated by a spatial and temporal data fusion model

Leaf area index (LAI) is an important input parameter for biogeochemical and ecosystem process models. Mapping LAI using remotely sensed data has been a major objective in remote sensing research to date. However, the current LAI product mapped by remote sensing is both spatially and temporally discontinuous as a result of cloud cover, seasonal snows, and instrumental constraints. This has limited the application of LAI to ground surface process simulations, climatic modeling, and global change research. To fill these gaps in LAI products, this study develops an algorithm to provide high spatial and temporal resolution LAI products with synthetic Landsat data, generated by a spatial and temporal data fusion model (STDFA). The model has been developed and validated within the Changping District of Beijing, China. Using Moderate Resolution Imaging Spectroradiometer (MODIS) reflectance data and real Landsat data, this method can generate LAI data whose spatial (temporal) resolution is the same as that of the Landsat (MODIS) data. Linear regression analysis was performed to compare the modeled data with field-measured LAI data, and indicates that this new method can provide accurate estimates of LAI, with R2 equal to 0.977 and root mean square error (RMSE) equal to 0.1585m2m−2 (P<0.005), which is superior to the standard MODIS LAI product. Further, various STDFA model application strategies were tested, with the results showing that the application strategy of the STDFA model has an important influence on the accuracy of LAI estimation: the vegetation index fusion strategy produced a better result than the reflectance fusion strategy. The applications of the STDFA model to eight commonly used vegetation indices were also compared. The results show that some vegetation indices (e.g., Enhanced Vegetation Index (EVI), Normalized difference vegetation index (NDVI), and Normalized difference infrared index (NDII)) exhibited better performance than others (e.g., Infrared simple ratio (ISR), Reduced infrared simple ratio (RISR), Reduced normalized difference vegetation Index (RNDVI), Reduced simple ratio (RSR), and Simple ratio (SR)). However, ISR, RISR, and NDII data produced lower saturation effects than other spectral vegetation indices in the estimation of LAI values higher than 2m2m−2.