Application of the land surface temperature from MODIS in the estimation of gross primary productivity for a subtropical pinus plantation in southern China

Junxia Yan Junxia Yan,Hongjian Li Hongjian Li,Jinhua Tao Jinhua Tao,Liangfu Chen,Yanhua Gao Yanhua Gao

doi:10.1109/rsete.2011.5964230

Abstract

In this study, the land surface temperature (LST) from Moderate Resolution Imaging Spectroradiometer (MODIS) sensor which was chosen as one scalar temperature directly modifying light use efficiency (LUE), together with the 8-day composited MODIS data and the site measurements of photosynthetically active radiation (PAR) of a subtropical evergreen coniferous forest ecosystem in southern China during 2005-2006 are used to estimate the seasonal variations of gross primary production (GPP) with the Production Efficiency Model (PEM). To investigate the possibility of LST instead of atmospheric temperature (T a ) in the PEM model to estimate the GPP values, the LST predicted GPP (GPP LST ) and T a predicted GPP (GPP air ) are used to compare the GPP values observed from eddy covariance measurements (GPP obs ). It was found that the coefficient of determination of the fitted exponential equations between the GPP obs and the T a (r 2 = 0.89) was slightly higher than that between the GPP obs and the LST(r 2 = 0.80), and both variables could explain above 80 % variations of the GPP obs in this area. Both the GPP LST and GPP air agreed well with the observed GPP obs .There was a very strong linear relation between GPP LST and GPP air (r 2 = 0.99), and they both were highly correlated with the GPP obs (r 2 = 0.91 and 0.91, respectively). The annual total GPP LST and GPP air values (1630.6 and 1598.6 g C m -2 for 2005; 1740.2 and 1660.8 g C m -2 for 2006, respectively) were slightly lower than the GPP obs (1646.8 and 1867.4 g C m -2 , respectively), and the error of the predicted GPP from LST (0.98% and 6.8% for 2005 and 2006, respectively) was smaller than the predicted GPP from T a (2.9% and 11.1% for 2005 and 2006, respectively) for the whole year data. These results demonstrate that the LST can be potentially used to estimate GPP instead of T a .

Full Text