Abstract
Accurate and continuous monitoring of forest production is critical for quantifying the dynamics of regional-to-global carbon cycles. MOD17A2 provides high frequency observations of terrestrial gross primary productivity (GPP) and is widely used to evaluate the spatiotemporal variability and responses to changing climate. However, the effectiveness of the Moderate Resolution Imaging Spectroradiometer (MODIS) in measuring GPP is directly constrained by the large uncertainties in the modeling process, specifically for complicated and extensive forest ecosystems. Although there have been plenty of studies to verify the MODIS GPP product with groundbased measurements covering a range of biome types, few have comprehensively validated the performance of MODIS estimates (C5.5) for diverse forests. Thus, this study examined the degree of correspondence between the MODIS-derived GPP and the EC-measured GPP at seasonal and interannual time scales for the main forest ecosystems, encompassing evergreen broadleaf forest (EBF), evergreen needleleaf forest (ENF), deciduous broadleaf forest (DBF), and mixed forest (MF) relying on 16 flux towers with a total dataset of 68 site-years. Overall, the site-specific evaluation of multi-year mean annual GPP estimates indicates that the current MODIS product works more significantly for DBF and MF, less for ENF, and least for EBF. Except for the tropical forest, MODIS estimates could capture the broad trends of GPP at an 8-day time scale for the other sites. At the seasonal time scale, the highest performance was observed in ENF, followed by MF and DBF, and the least performance was observed in EBF. Trend analyses also revealed the weak performance in EBF and DBF. This study suggested that current MODIS GPP estimates still need to improve the quality of different upstream inputs in addition to the algorithm for accurately quantifying forest production.
Tang, Xuguang