Cultivar coefficient stability and effects on yield projections in the SPUDSIM model

Abstract

Crop model calibration refers to the process where values are obtained for a subset of model parameters that represent cultivar traits. Calibrated parameters, however, are just as likely to compensate for limitations in model structure as well as reflect true phenotypic characteristics. This confounding of genetics with production environment limits model accuracy particularly for climate assessments and can result in crop parameter values that are location dependent. We evaluated the calibration stability of the potato (Solanum tuberosum L.) model SPUDSIM for two cultivars grown in two sites in the United States sites using a location specific (R1) and cross-location (R2) methodology. Differences between R1 coefficient values ranged from 4 to 17% between the two sites depending on cultivar. This likely reflected over-coupling of canopy expansion rate within the model structure as well as an over-sensitivity to temperature and photoperiod. Broader site differences were observed for growth coefficients, particularly when compared with the R2 method. Validation year end-ofseason yields (17% error or less) and in-season RMSE (<36.1 g plant −1 ) were similar for both methods. The R2 calibration overpredicted yields by 5% or less as compared with R1, with a maximum discrepancy of 0.35 Mg ha −1 in climate forecasted impacts. Consistent differences between R1 and R2 calibrated models in rates of yield decline per decade were not observed for cultivar, location, or climate scenario, and averaged 0.034 Mg ha −1 yr −1 for R1 and 0.038 for R2. Thus, calibration for these cultivars and locations was relatively stable in SPUDSIM for the temperate climate evaluated.

Yield Response of an Ensemble of Potato Crop Models to Elevated CO2 in Continental Europe

“…Specific calibration methods thus varied among modeling groups and included manual and/or automated procedures for minimizing differences between predicted versus observed values via the use of 'genetic coefficients'. Genetic coefficients in these models represent crop phenotype response to weather and management factors (Fleisher et al, 2020). A list of the major genetic coefficients calibrated for each model, their resulting values, and the institute of each modeling group that performed the calibration was shown in supplemental Table S1.…”

“…Asterisks indicate locations that were under-predicted by the mean of all models. phenology parameters first (those that capture influence of, for example, temperature and/or photoperiod on tuber initiation and maturity dates) followed by growth parameters which is generally the accepted approach (Fleisher et al, 2020). Comparisons of other growth and phenology responses is outside the scope of this manuscript and limited in some locations by data availability.…”

Simulations of Water and Thermal Dynamics for Soil Surfaces With Residue Mulch and Surface Runoff

“…Models established in Sections 2.1–2.3 are implemented in MAIZSIM, a modular‐based soil‐crop simulator including multiple soil physical and chemical processes (via MAIZSIM sub‐model “2DSOIL,” Timlin et al., 1996), as well as maize growth and soil‐plant interactions (via a crop sub‐model in MAIZSIM, Kim et al., 2012; Wang et al., 2021). With proper modifications in the crop sub‐model, variations of MAIZSIM can be extended for other crops, for example, SPUDSIM for potato (Fleisher et al., 2020) or an updated GLYCIM for soybean (Sun et al., 2021). Climate factors, crop growth, agricultural management including fertilizer and irrigation, and soil water, heat and solute transfer are coded as separated modules.…”