Evaluation and comparison of light use efficiency models for their sensitivity to the diffuse PAR fraction and aerosol loading in China

Abstract

The diffuse photosynthetically active radiation (PARdif) is highly related to the gross primary productivity (GPP) of vegetation and can enhance the photosynthetic rate of shaded leaves. The PARdif is mainly influenced by aerosols on clear sky days. To explore the impacts of the PARdif fraction and aerosol loading on the accuracy of GPP estimations, we evaluated and compared the sensitivity of one-leaf and two-leaf light use efficiency (LUE) models (MOD17 and TL-LUE models) to the PARdif fraction and aerosol optical depth (AOD) using satellite data and measurements at ChinaFLUX sites in China from 2003 to 2010. The estimated GPP from two models (GPP_MOD and GPP_TL) showed systematic underestimations when compared with the measured GPP at ChinaFLUX sites (GPP_FLUX). In addition, the underestimation of the GPP estimated by the two models was larger in months with high PARdif fractions than in months with low PARdif fractions at most sites. The TL-LUE model exhibited better performance than the MOD17 model with reduced underestimation and root-mean-square error (RMSE) values when the PARdif fraction was greater than 0.7 or 0.8. When the PARdif fraction was low, the TL-LUE model reduced the overestimation or increased the underestimations of GPP compared with the MOD17 model depending on the site. Both MOD17 and TL-LUE exhibited higher underestimations of GPP at high AOD values than at low AOD values at most sites. The TL-LUE model exhibited an overall better performance than MOD17 under high AOD. The overestimation also occurred for low AOD, which was consistent with that for low PARdif fractions. Therefore, the two LUE models showed dynamic performance with variations in the PARdif fraction and AOD, which should be considered in the application and further improvement of GPP models.