Abstract: GPP(Gross Primary Productivity) is a key indicator to describe terrestrial ecosystem, which provides a quantitative description of carbon cycle under global climate change. It is an important indicator of ecosystem function, and it's a key element in the carbon cycle, which reflects the results of the comprehensive influence of climate change and human activities on land vegetation. As a key parameter in remote sensing estimation model, the value of LUE(Light Use Efficiency, LUE) is affected by many factors such as environmental factors, spatial and temporal distribution differences, vegetation types and so on. In order to quantitatively evaluate the ability of remote sensing vegetation parameters in estimating ecosystem GPP, Jinzhou corn production area was selected as the research object, based on the surface flux data and MODIS data from 2013 to 2014. APAR(Absorbed Photosynthetically Active Radiation, APAR) model, PRI(Photochemical Reflectance Index, PRI) model and REG-PEM(REGion Productivity Efficiency Model, REG-PEM) model were established to estimate the GPP of sites on different time scales. With the help of correlation analysis method, the results are as follows: (1) on diurnal scale, the seasonal dynamics of estimated GPP from REG-PEM model and APAR model both matched reasonably well with those of observed GPP from eddy covariance flux. Relative error of estimated GPP from APAR model was less than that from REG-PEM model. However, the phenomenon of estimated GPP was overrated in GPP low-value area while underrated in high-value area, existed in both two models. The main reason is that LUEmax was overestimated in the low vegetation coverage area, and the influence of air temperature and moisture on LUE was underestimated. There are inevitable errors in the reconstruction of vegetation index curve EVI and LSWI. (2) On hour scale, especially at midday, the solar radiation and the temperature are increased, the phenomenon of light saturation and midday break in vegetation leaves greatly weakens the response ability of APAR to GPP and weakens the simulation effect, the ability of APAR in response to GPP had weakened. Compared with APAR model, the accuracy of GPP estimation can be improved by using PRI model, but the simulation effect needs to be improved.

Key words: MODIS, LUE Model, GPP