Estimating wheat yield by integrating the WheatGrow and PROSAIL models

• Published in: Field crops research Vol. 192; pp. 55 - 66
• Main Authors: Zhang, L., Guo, C.L., Zhao, L.Y., Zhu, Y., Cao, W.X., Tian, Y.C., Cheng, T., Wang, X.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.06.2016
• Subjects: algorithms; crop models; crops; Data fusion; ecoregions; Grain yield; heading; leaf area index; LUT; monitoring; nitrogen; prediction; PROSAIL model; radiative transfer; remote sensing; sowing date; spatial data; Triticum aestivum; uncertainty; Wheat; WheatGrow model; winter wheat; Data fusion; LUT; Wheat; PROSAIL model; Grain yield; WheatGrow model
• ISSN: 0378-4290; 1872-6852
• DOI: 10.1016/j.fcr.2016.04.014

•PROSAIL model was integrated with the WheatGrow model, integrating RS and growth•We developed a look-up table between VIs and measured wheat parameters.•Time-series VIs obtained by fusing high spatial- and temporal-resolution images.•Three-band Vi RNir/(RRed+RGreen) and two-band VI SAVI(RNir,RRed) were optimal.•Accuracy is best using multiple RS data from late jointing to initial filling. Coupling remote sensing data with crop models is an important way to predict crop yield on a regional scale. Here, we developed a method for wheat yield forecasting based on integrating a wheat growth model (WheatGrow) with a radiative transfer model (PROSAIL). Based on the assimilation algorithm, the PROSAIL model was joined to the coupling model, integrating the remote sensing (RS) and growth models. Wheat VIs (vegetation indices) from different growth periods, which were obtained by fusing high spatial resolution and high temporal resolution images, were used for coupling parameters. Moreover, a LUT (look-up table) was developed between VIs and the measured management parameters (sowing date, sowing rate and nitrogen rate) based on the coupling model. Management parameters including sowing date, sowing rate and nitrogen rate, which are difficult to accurately obtain, were then inverted at the regional scale. Reducing the uncertainty from the input parameters of the WheatGrow model, such as management parameters, improved the accuracy of regional-scale yield prediction. We also used measured data from different years and different ecological sites to determine the optimum coupling VI and coupling frequency based on the integration of RS and crop models. The results show that (1) the three-band VI RNir/(RRed+RGreen) and the two-band VI SAVI (RNir,RRed) are the best coupling VIs; (2) the heading or anthesis stage is the best coupling stage if only one RS image is available; (3) good prediction accuracy can be achieved if three to four high spatial and temporal resolution images from the late jointing to initial filling stages are assimilated. In the study areas, the spatial and temporal distribution of winter wheat growth and productivity parameters were well simulated. The RRMSE of leaf area index and leaf nitrogen accumulation between predicted and measured values were 17.8% and 20.3%, respectively, and the RRMSE of grain yield were less than 10%. The results of this study provide an important theoretical and technical foundation for monitoring and estimating winter wheat growth status on a regional scale and can be extended to other types of crops from varied ecological regions.