Interannual variation of gross primary productivity (GPP), its annual maximum (GPPmax), and growing season length (GSL) are crucial indicators for assessing forest ecosystem responses to climate change. Coarse-resolution satellite observations (e.g., MODIS, 250–500 m) have been widely used to upscale GPP metrics measured by eddy-covariance (EC) flux towers during 2000-present. However, primarily due to data sparsity, studies that use Landsat data to upscale GPP metrics and investigate long-term carbon dynamics are rare, despite its finer resolution (30 m) and extended temporal coverage (1980s-present), which align well with most EC measurements. Here, by using a recently developed Bayesian land surface phenology (BLSP) method that addresses data sparsity of Landsat and a vegetation photosynthesis model (VPM), we explored the potential of Landsat in upscaling long-term GPP metrics. We found that Landsat had comparable performance (R2 = 0.9, RMSE = 1.32 g C m-2d-1, Bias = -0.3 g C m-2d-1) with MODIS (R2 = 0.92, RMSE = 1.13 g C m-2d-1, Bias = -0.22 g C m-2d-1) in estimating GPP validated by EC measurements. Both data sources had higher performance in deciduous (Landsat R2 = 0.89; MODIS R2 = 0.91) than in evergreen forests (Landsat R2 = 0.79; MODIS R2 = 0.88). More importantly, Landsat substantially extended the temporal coverage of GPP, especially prior the MODIS era. This extension enabled a more robust assessment of long-term GPP dynamics, as evidenced by our result that long-term trends of GPP metrics derived from Landsat aligned more closely with EC measurements than those derived from MODIS. Therefore, our study shows that Landsat, coupled with the BLSP model, offers a powerful tool to temporally extend EC-measurements and investigate long-term vegetation ecosystem carbon dynamics.
扫码关注我们
求助内容:
应助结果提醒方式:
