Estimating Gross Primary Productivity (GPP) over Rice-Wheat-Rotation Croplands by Using the Random Forest Model and Eddy Covariance Measurements: Upscaling and Comparison with the MODIS Product

Despite advances in remote sensing-based gross primary productivity (GPP) modeling, the calibration of the Moderate Resolution Imaging Spectroradiometer (MODIS) GPP product (GPP(MOD)) is less well understood over rice-wheat-rotation cropland. To improve the performance of GPP(MOD), a random forest (RF) machine learning model was constructed and employed over the rice-wheat double-cropping fields of eastern China. The RF-derived GPP (GPP(RF)) agreed well with the eddy covariance (EC)-derived GPP (GPP(EC)), with a coefficient of determination of 0.99 and a root-mean-square error of 0.42 g C m(-2) d(-1). Therefore, it was deemed reliable to upscale GPP(EC) to regional scales through the RF model. The upscaled cumulative seasonal GPP(RF) was higher for rice (924 g C m(-2)) than that for wheat (532 g C m(-2)). By comparing GPP(MOD) and GPP(EC), we found that GPP(MOD) performed well during the crop rotation periods but underestimated GPP during the rice/wheat active growth seasons. Furthermore, GPP(MOD) was calibrated by GPP(RF), and the error range of GPP(MOD) (GPP(RF) minus GPP(MOD)) was found to be 2.5-3.25 g C m(-2) d(-1) for rice and 0.75-1.25 g C m(-2) d(-1) for wheat. Our findings suggest that RF-based GPP products have the potential to be applied in accurately evaluating MODIS-based agroecosystem carbon cycles at regional or even global scales.

Automated summary

By utilizing a random forest machine learning model, the accuracy of MODIS GPP product can be improved in rice-wheat rotation areas, and the eddy covariance-derived GPP can be reliably upscaled to regional scales. The seasonal GPP for rice was higher than that for wheat, and MODIS product performed well during crop rotation periods but underestimated GPP during rice/wheat active growth seasons.