Fu, G; Shen, ZX; Zhang, XZ; Shi, PL; He, YT; Zhang, YJ; Sun, W; Wu, JS; Zhou, YT; Pan, X (2012). Calibration of MODIS-based gross primary production over an alpine meadow on the Tibetan Plateau. CANADIAN JOURNAL OF REMOTE SENSING, 38(2), 157-168.
Moderate-resolution imaging spectroradiometer (MODIS) gross primary production (GPP) was compared with estimated GPP (GPP_EC) from eddy covariance measurements over an alpine meadow on the Tibetan Plateau in 2005-2007. The MODIS GPP (GPP_MOD17A2) with a bias of -0.38 g C m(-2) d(-1) (i.e., about -40.58% of the mean of the GPP_EC) strongly underestimated the GPP_EC for the alpine meadow. The MODIS GPP was recalibrated using measured surface meteorological data, including photosynthetically active radiation (PAR), daily minimum air temperature (T-amin) and daytime mean vapor pressure deficit (VPD), revised fractional photosynthetically active radiation (FPAR), and the revised maximum light use efficiency (LUEmax) of 0.81 g C MJ(-1) (compared with the default value of 0.68 g C MJ(-1) for grassland in the MODIS GPP algorithm) for the alpine meadow. The MODIS-based FPAR was about 14.70% larger than the surface-estimated FPAR using surface-measured leaf area index (LAI) data. Additionally, the temporal resolution of surface-measured LAI data was relatively low. Therefore, the linear relationship between surface-measured LAI and MODIS-based LAI was established (R-2 > 0.80, P < 0.001). Then the revised MODIS LAI datasets were used to calculate the revised FPAR. The revised LUEmax was optimized from the MOD17A2 algorithm using daily surface measurements, including LAI, PAR, VPD, T-amin and GPP_EC. The calibrated MOD17A2 algorithm could explain 88% of GPP_EC variance for the alpine meadow. The bias between GPP_MOD17A2 and calculated GPP from the MOD17A2 algorithm using surface-measured PAR, T-amin, and VPD, MODIS-based FPAR, and the default LUEmax of 0.68 g C MJ(-1) was -0.17 g C m(-2) d(-1) (i.e., about -17.60% of the mean of the GPP_EC). The underestimation of LUEmax caused a 13.78% underestimation of GPP. In contrast, the overestimation of FPAR resulted in a 7.17% overestimation of GPP. The net effect of meteorology data and FPAR resulted in a 13.84% underestimation of GPP. These results showed that MODIS-based meteorology data, FPAR, and LUEmax for the alpine meadow needed to be adjusted.