Speaker
Description
Introduction
Cropping system models (CSMs) are vital tools in agricultural research, typically designed for plot level assessments but increasingly applied across different scales, aiding in understanding how crops respond to various factors like climate change (Asseng et al., 2013). However, when these models are scaled up, they face challenges due to limited data availability for parameterization (Angulo et al., 2013). Researchers use preset coefficients from model developers or previous studies, which can introduce uncertainties into the model-based assessments and respective outputs. It's crucial to accurately parameterize these models to tackle this issue, especially when exploring new locations or crop varieties. One promising approach is utilizing data from crop variety trials to fine-tune the models, ensuring they accurately reflect real-world conditions (Liang et al., 2021). In addition, one needs to consider that even with careful calibration, there is still inherent uncertainty related to model structure (Wallach et al., 2017). By employing a multi-model approach, researchers can minimize these uncertainties and improve the robustness of their predictions (Röll et al., 2021). Our study addresses these challenges by training three wheat models with a comprehensive dataset under German growth conditions, aiming to enhance yield predictions and optimize agricultural practices amidst changing environmental conditions.
Materials and Methods
In our study, we employed three wheat models from DSSAT: CSM-CERES, CSM-CROPSIM, and CSM-Nwheat. Ensuring consistent parameterization involved gathering cultivar-specific data from Germany’s Plant Variety Trials, supported by detailed growth data from a rain-out shelter trial in Braunschweig (52.296° N, 10.436° E; 75m a.s.l. (Schittenhelm et al., 2014). Soil data originated from the German Federal Institute for Geosciences and Natural Resources (BGR), and daily weather data from the German Weather Service (DWD), at a 1×1 km grid resolution. Model application encompassed four pedoclimatically contrasting sites across Germany: Feldkirchen (southeast, with good soil and high rainfall), Bad Lauchstädt (central, with good soil but limited rainfall), Thyrow (northeast, with poor soil and limited rainfall), and Gudendorf (north, with poor soil but high rainfall). We used the three models to simulate Management × Environment interactions in wheat production covering a span of 30 years (1991-2020). We evaluated the effects of six nitrogen treatments (0, 60, 120, 150, 180, and 200 kg Nitrogen per hectare) under rainfed vs. irrigated conditions considering productivity, efficiency and environmental impact of wheat production.
**
Results and discussion
**
Our study highlights the effective parameterization of DSSAT-CSM-CERES, DSSAT-CSM-CROPSIM, and DSSAT-CSM-Nwheat using plant variety trial data from diverse locations and cropping seasons nationwide. Using the same dataset, we achieved reasonable accuracy in predicting phenology, while for grain yield prediction, CSM-CERES exhibited an RMSE of 1405 kg ha-1, CSM-CROPSIM 2126 kg ha-1, and CSM-Nwheat 1886 kg ha-1.The distribution of simulated yield compared to observed yield (Figure 1) indicates close alignment, confirming model accuracy. Irrigation consistently increased yields in Bad Lauchstädt and Thyrow across all nitrogen levels, while no significant effects were observed in Feldkirchen and Gudendorf, suggesting sufficient rainfall for optimal plant growth under rainfed conditions. The CSM-CERES model exhibited higher yields at higher nitrogen levels (150N, 180N, 200N) across all locations, while CSM-CROPSIM and CSM-Nwheat showed similar effects in fewer locations. In conclusion, our study underscores the importance of robust model parameterization using diverse datasets, informing agricultural policy and climate change adaptation strategies nationally.
**
References
**
- Asseng, Senthold, et al. "Uncertainty in simulating wheat yields
under climate change." Nature climate change 3.9 (2013): 827-832. - Angulo, Carlos, et al. "Characteristic ‘fingerprints’ of crop model
responses to weather input data at different spatial resolutions."
European Journal of Agronomy 49 (2013): 104-114. - Liang, Xi, et al. "Deriving genetic coefficients from variety trials
to determine sorghum hybrid performance using the CSM–CERES–Sorghum
model." Agronomy Journal 113.3 (2021): 2591-2606. - Wallach, Daniel, et al. "Estimating uncertainty in crop model
predictions: Current situation and future prospects." European
Journal of Agronomy 88 (2017): A1-A7. - Röll, Georg, et al. "Implementation of an automatic time‐series
calibration method for the DSSAT wheat models to enhance multi‐model
approaches." Agronomy Journal 112.5 (2020): 3891-3912. - Schittenhelm, Siegfried et al. "Performance of winter cereals grown
on field-stored soil moisture only." European Journal of Agronomy 52
(2014): 247-258.
| Keywords | DSSAT wheat model; GxExM; Parameterization; Plant variety trials |
|---|
