Context

Process-based crop growth models can explain soil and crop dynamics that influence the optimal N rate for crop production. Currently, there is a lack of understanding regarding the accuracy of process-based models for site-specific zones within fields, as well as the key factors that need to be considered when calibrating these models for zone-specific economic optimum N rate (EONR).

Objective

We calibrated the Agricultural Production Systems sIMulator (APSIM) model in contrasting zones within fields, quantified the model performance, and used the calibrated model to develop long-term corn yield response to N to assess the temporal variability between zones and sites to assist decision making.

Methods

We conducted four N rate experiments (2 fields × 2 zones within a field) over two years in southeast Nebraska. Experimental data were used to calibrate and test the APSIM model. APSIM simulated corn yield response to N for each zone and site was obtained by running numerous iterations of the calibrated model at different N rates. Observed and simulated corn yield response to N rate were analyzed with statistical models to estimate the EONR.

Results and conclusions

The APSIM model predicted corn yield over 11 historical years with a relative root mean square error (RRMSE) of 12% and yield at EONR in the N studies with RRMSE of 8.8%. The simulated EONR was lower than the observed EONR across sites, years, and zones with greater error than yield. The simulated yield increase with N fertilization was under-estimated in fine textured soils and over-estimated in medium textured soils. Long-term corn yield response to N showed that temporal variation in simulated EONR was greater than spatial variation. Long-term EONR and yield at EONR increased with increasing rainfall, while yield at zero N was greatest in normal years. Temporal variation was driven primarily by year-to-year variation in N loss (CV of 67% ± 9.5). Soil texture, hydrological properties, water table, and tile drainage were key variables for accurate site-specific model calibration. Improvements in simulating site-specific EONR may be realized by including in-situ or remotely sensed data for better estimation of N dynamics. We concluded that APSIM can provide valuable insights into systems dynamics in this region, but it can’t provide precise N-rate estimates. Our study contributes to understanding of the within-field variability using simulation modeling.

中文翻译：

---

利用 APSIM 模型模拟田间玉米产量对氮的时空响应

 语境

基于过程的作物生长模型可以解释影响作物生产最佳施氮量的土壤和作物动态。目前，人们对油田内特定区域基于过程的模型的准确性以及在校准这些模型以获得特定区域经济最佳氮肥率（EONR）时需要考虑的关键因素缺乏了解。 。

 客观的

我们在田间的对比区域中校准了农业生产系统模拟器 (APSIM) 模型，量化了模型性能，并使用校准后的模型开发了对氮的长期玉米产量响应，以评估区域和地点之间的时间变化，以协助决策。

 方法

我们在内布拉斯加州东南部两年内进行了四次氮速率实验（2 个田地 × 田地内的 2 个区域）。实验数据用于校准和测试APSIM模型。 APSIM 模拟每个区域和地点的玉米产量对氮的响应是通过在不同施氮量下运行校准模型的多次迭代获得的。使用统计模型分析观察和模拟的玉米产量对施氮量的响应，以估计 EONR。

 结果和结论

APSIM 模型预测 11 个历史年份的玉米产量，相对均方根误差 (RRMSE) 为 12%，N 研究中 EONR 的产量，RRMSE 为 8.8%。模拟的 EONR 低于不同地点、年份和区域的观测 EONR，且误差大于产量。细质地土壤中施氮肥的模拟产量增加被低估，而中质地土壤中的模拟产量增加被高估。玉米产量对氮的长期响应表明，模拟 EONR 的时间变化大于空间变化。长期 EONR 和 EONR 产量随着降雨量的增加而增加，而零氮产量在正常年份最高。时间变化主要是由氮素损失的逐年变化驱动的（CV为67%±9.5）。土壤质地、水文特性、地下水位和瓷砖排水是准确的特定地点模型校准的关键变量。模拟特定地点 EONR 的改进可以通过包括原位或遥感数据来实现，以便更好地估计氮动态。我们得出的结论是，APSIM 可以为该区域的系统动力学提供有价值的见解，但它无法提供精确的 N 速率估计。我们的研究有助于使用模拟建模来理解场内变异性。