How well does MODIS GPP capture alpine productivity of a Tibetan grassland?

Résumé fourni par la source

Discrepancies between satellite-derived gross primary productivity (GPP) and eddy covariance (EC) based estimates have been widely reported across ecosystems. Although spatial and scale mismatch between time-varying EC flux footprints and fixed satellite pixels is frequently cited as one of the potential causes, this assumption has rarely been tested quantitatively. In this study, we quantified the role of spatial representativeness in explaining discrepancies between moderate resolution imaging spectroradiometer (MODIS) GPP and EC-derived GPP. We explicitly accounted for vegetation distribution within individual EC footprints and the corresponding MODIS pixel using half-hourly flux footprints and high-resolution daily normalized difference vegetation index. MODIS GPP underestimated EC-derived GPP at this site, with a bias and RMSE of −8.73 and 9.19 g C m⁻² 8 d⁻¹, respectively. During periods of active vegetation, MODIS GPP showed moderate temporal agreement with EC-derived GPP (R² = 0.60) but a large relative bias (−65.30 %). Restricting EC observations to the footprints exhibiting high vegetation similarity to the MODIS pixel improved the temporal agreement (R 2 = 0.69), but did not substantially reduce the bias (−9.16 g C m⁻² 8 d⁻¹; −66.3 %). These results indicate that the spatial mismatch between EC footprints and MODIS pixels is unlikely to be the primary cause of the underestimation. Instead, algorithmic constraints in MODIS GPP and temporal scaling differences between satellite and tower measurements are likely the main contributors to the lower MODIS GPP at this site.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.