Sensitivity of Crop Gross Primary Production Simulations to In-situ and Reanalysis Meteorological Data

Friday, 19 December 2014
Cui Jin1, Xiangming Xiao2 and Pradeep Wagle1, (1)University of Oklahoma Norman Campus, Norman, OK, United States, (2)University of Oklahoma, Norman, OK, United States
Accurate estimation of crop Gross Primary Production (GPP) is important for food securityand terrestrial carbon cycle. Numerous publications have reported the potential of the satellite-based Production Efficiency Models (PEMs) to estimate GPP driven by in-situ climate data. Simulations of the PEMs often require surface reanalysis climate data as inputs, for example, the North America Regional Reanalysis datasets (NARR). These reanalysis datasets showed certain biases from the in-situ climate datasets. Thus, sensitivity analysis of the PEMs to the climate inputs is needed before their application at the regional scale. This study used the satellite-based Vegetation Photosynthesis Model (VPM), which is driven by solar radiation (R), air temperature (T), and the satellite-based vegetation indices, to quantify the causes and degree of uncertainties in crop GPP estimates due to different meteorological inputs at the 8-day interval (in-situ AmeriFlux data and NARR surface reanalysis data). The NARR radiation (RNARR) explained over 95% of the variability in in-situ RAF and TAF measured from AmeriFlux. The bais of TNARR was relatively small. However, RNARR had a systematical positive bias of ~3.5 MJ m-2day-1 from RAF. A simple adjustment based on the spatial statistic between RNARR and RAF produced relatively accurate radiation data for all crop site-years by reducing RMSE from 4 to 1.7 MJ m-2day-1. The VPM-based GPP estimates with three climate datasets (i.e., in-situ, and NARR before and after adjustment, GPPVPM,AF, GPPVPM,NARR, and GPPVPM,adjNARR) showed good agreements with the seasonal dynamics of crop GPP derived from the flux towers (GPPAF). The GPPVPM,AF differed from GPPAF by 2% for maize, and -8% to -12% for soybean on the 8-day interval. The positive bias of RNARR resulted in an overestimation of GPPVPM,NARR at both maize and soybean systems. However, GPPVPM,adjNARR significantly reduced the uncertainties of the maize GPP from 25% to 2%. The results from this study revealed that the errors of the NARR surface reanalysis data introduced significant uncertainties of the PEMs-based GPP estimates. Therefore, it is important to develop more accurate radiation datasets at the regional and global scales to estimate gross and net primary production of terrestrial ecosystems at the regional and global scales.