| 摘 要： |
Large-scale land surface phenology (LSP) information has been developed from remote sensing-based vegetation indices (VIs) data. However, there are considerable discrepancies and uncertainties in the LSP data products for the start and end of growing seasons (SOS; EOS) as different vegetation indices and algorithms are used. Here, we used the TROPOspheric Monitoring Instrument (TROPOMI) solar-induced chlorophyll fluorescence (SIF) data to estimate SOS and EOS in the Tibetan Plateau, a global hotspot of vegetation response to climate change. We compared SIF-based phenological metrics to those derived from VIs (e.g., normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), and near-infrared reflectance of vegetation (NIRv)), and gross primary production (GPP) simulated by the vegetation photosynthesis model (VPM). We found relatively small dis-crepancies in SOS between SIF and VIs, but large differences in EOS. Thus, the length of the growing season derived from VIs was as much as two months longer than that estimated by SIF. These results were consistent across three products, bidirectional reflectance distribution function (BRDF) adjusted (MCD43), standard MODIS (MOD09), and TROPOMI products. The EOS discrepancy remained after excluding two mismatches (solar illu-mination and viewing angle) between satellite sensors. We also found that VIs-based EOS occurred below the freezing point, while SIF-based EOS occurred above the freezing point, suggesting that SIF-based EOS is more physiologically meaningful. Our study proposed that the late VIs-based EOS was caused in part by the effect of changes in soil background on VIs and VIs-based EOS. Our results highlight the need to re-evaluate current LSP data products derived from reflectance-based VIs and to develop new vegetation phenology data products using emitted energy such as SIF. |
在线留言
联系我们
所长信箱