Evaluation of four long time-series global leaf area index products

Abstract

As an important vegetation biophysical variable, leaf area index (LAI) is a critical input parameter in many climate and ecological models. There exist four long time-series global LAI products since the 1980s, namely GLASS AVHRR, NCEI AVHRR, GIMMS3g and GLOBMAP. Currently, no inter-comparison studies exist to evaluate these LAI products and understand their differences for effective applications. In this study, the four long time-series global LAI products were inter-compared to evaluate their temporal and spatial discrepancies. These LAI products were also compared with MODIS LAI product and LAI values derived from high-resolution reference maps at VAlidation of Land European Remote sensing Instruments (VALERI) sites. The results show that the GLASS AVHRR and GLOBMAP LAI products are spatially complete, but the NCEI AVHRR and GIMMS3g LAI products contain many missing pixels, especially in rainforest regions. These LAI products reasonably represent the global vegetation characteristics and their seasonal variability. A relatively large discrepancy among these LAI products was observed in tropical forest regions, where the GLASS AVHRR and NCEI AVHRR LAI values achieved good agreement with the MODIS LAI values, but were between 0.5 and 1.0 LAI units lower than the GLOBMAP LAI values and higher than the GIMMS3g LAI values (more than 0.5 LAI units). Over the last three decades, the GLASS AVHRR, NCEI AVHRR, GIMMS3g LAI products show increasing trends for all biome types except evergreen needleleaf forests and deciduous needleleaf forests, but the GLOBMAP LAI product shows positive trends only for the grasses/cereal crops and shrubs. A comparison of these LAI products against the LAI values derived from high-resolution reference maps demonstrated that the GLASS AVHRR LAI values provided the better performance (RMSE=0.9014 and Bias=−0.1885) than the NCEI AVHRR LAI values (RMSE=1.0459 and Bias=−0.5695), the GIMMS3g LAI values (RMSE=1.0971 and Bias=−0.3904) and the GLOBMAP LAI values (RMSE=1.6145 and Bias=−0.9414).