Run-time calibration, i.e. adjusting simulation results for field observations of model driving variables during run-time, may allow correcting for deviations between complex mechanistic simulation model results and actual field conditions. Leaf area index (LAI) and canopy nitrogen contents (LeafNWt) are the most important driving variables for these models, as they govern light interception and photosynthetic production capacity of the crop. Remote sensing may provide (spatial) data from which such information can be estimated. How, when and at what frequency such additional information is integrated in the simulation process may have various effects on the simulations. The objective of this study was to quantify the effects of different run-time calibration scenarios for final grain yield (FGY) simulations in order to optimize remote sensing image (RS) acquisition. The PlantSys model was calibrated on LAI and LeafNWt for maize in France and used to simulate maize crop growth in the Argentina and the USA, for which non-destructive estimates of LAI and leaf chlorophyll contents were acquired by optical measurement techniques. Leaf chlorophyll data were used to estimate LeafNWt. Due to its structure, the PlantSys model was more susceptible to run-time calibration with LeafNWt than with LAI. Run-time calibration with LAI showed the largest effect on FGY before and around flowering, and could mainly be related to maintenance respiration costs. Run-time calibration with LeafNWt showed the largest effect on FGY at and after flowering and could mainly be related to the change in effective radiation interception due to change in leaf life. The accuracy of LAI estimates showed a major effect on FGY for underestimations but was small in absolute sense. The accuracy of LeafNWt estimates had significant impact at all crop development stages, but was the strongest after flowering where crop growth and nitrogen uptake are less able to recuperate from changes in LeafNWt. In absolute sense, the effect on FGY was as strong as the accuracy of the LeafNWt estimates when applied in the early reproductive stages. Based on these results it was concluded that remotely sensed in-field variability of LAI and LeafNWt is valuable information that can be used to spatially differentiate model simulations. Run-time calibration at sub-field level may lead to more accurate simulation results for whole fields.
- leaf-area index
- hyperspectral vegetation indexes
- chlorophyll density
- light reflectance