Improved gross primary production estimation in rice fields through integrated multi-scale methodologies.

Understanding productivity in agricultural ecosystems is important, as it plays a significant role in modifying regional carbon balances and capturing carbon in the form of agricultural yield. This study in particular combines information from flux determinations using the eddy covariance (EC) methodology, process-based modeling of carbon gain, remotely (satellite) sensed vegetation indices (VIs), and field surveys to assess the gross primary production (GPP) of rice, which is a primary food crop worldwide. This study relates two major variables determining GPP. The first is leaf area index (LAI) and carboxylation capacity of the rice canopy (Vc), and the second being MODIS remotely sensed vegetation indices (VIs). Success in applying such derived relationships has allowed GPP to be remotely determined over the seasonal course of rice development. The relationship to VIs of both LAI and Vc was analyzed first by using the regression approaches commonly applied in remote sensing studies. However, the resultant GPP estimations derived from these generic models were not consistently accurate and led to a large proportion of underestimations. The new, alternative approach developed to estimate LAI and Vc uses consistent development curves for rice (i.e., relies on consistent biological regulations of plant development). The modeled GPP based on this consistent development curve for both LAI and Vc agreed with from 0.76 to 0.92 (within the 95% confidence interval). The results of this study demonstrate that improved linkages between ground-based survey data, eddy flux measurements, process-based models, and remote sensing data can be constructed to estimate GPP in rice paddies. This study suggests further that the conceptual application of the consistent development curve, such as the combining of different scale measurements, has the potential to predict GPP better than the common practice of utilizing simple linear models, when seeking to estimate the critical parameters that influence carbon gain and agricultural yields.