Gross Primary Production Estimation In Maize Using MODIS And Flux Data

Gross primary productivity (GPP) is the amount of carbon fixed by vegetationthrough photosynthesis at unit time and unit area. GPP includes not only the productionof the plant, but also the plant communities’ respiration which maintains their ownsurvival at the same period. GPP determines the initial substance and energy into theterrestrial ecosystem, and is an important part of measuring the ecosystem carbon fluxand the carbon balance between the biosphere and atmosphere. Accurate estimation ofGPP is important significance to understanding of the global carbon cycle andevaluating effects of climate variation.In this paper, the light use efficiency model REG-PEM combined with remotesensing data and eddy covariance flux data was studied to estimate the GPP of maizeagro-ecosystem at the three flux stations named Yucheng station, Fermi station andYingke station. The feasibility and applicability of REG-PEM model in estimatingmaize agro-ecosystem was analyzed.The main methods and results are:(1) The daily and8-day MODIS surface reflectance products (MOD09GA andMOD09A1) were used to calculate the vegetation index by band operation. The timeseries of vegetation index curve was built. The time series vegetation index curve wasreconstructed by Harmonic Analysis of Time Series (HANTS), which could remove theoutliers caused by cloud and sensor noise. The contrast between original vegetationindex curve and reconstructed vegetation index curve by HANTS demonstrated that thereconstructed vegetation index curve was smoother and the seasonal variation of thevegetation index curve was more apparent than the original vegetation index curve. Ingeneral, HANTS realized the purpose of removal cloud and reconstruction of timeseries remote sensing image.(2) According to the actual situation of the study area, the station REG-PEM modelwas established. The daily and8-day GPP of maize agro-ecosystem was calculated. Bycomparing the model estimated GPP and eddy covariance flux observed GPP, theestimated GPP by the REG-PEM model agreed well with observed GPP from the fluxdata, and the seasonal dynamics of estimated GPP matched reasonably well with those of observed GPP. The correlation coefficient was R~2=0.64(Yucheng daily), R~2=0.82(Fermi daily), R~2=0.70(Yingke daily), R~2=0.77(Yucheng8-day), R~2=0.87(Fermi8-day), R~2=0.83(Yingke8-day) respectively. The results demonstrated that theREG-PEM model had high feasibility and applicability in estimating maizeagro-ecosystem.(3) The model parameters including the plant potential light use efficiency, theminimum temperature, the optimum temperature and the maximum temperature wereset to be fixed constant values in REG-PEM model. These constants were empiricalvalues which might not apply to this paper, and was one of the causes generating theestimation error. So in this paper Gauss-Marquardt-Levenberg algorithm was used tooptimize the four model parameters. The overall estimation error reduced from30.1%(Yucheng station),19.7%(Fermi station),5.5%(Yingke station) to8.4%(Yuchengstation),3.9%(Fermi station),1.2%(Yingke station).