In the current study, the uncertainties of wheat modeling using gridded soil and weather datasets were analyzed under dry farming and irrigated conditions. In this regard, the performance of the CSM-CERES-Wheat model forced with different weather-soil data combinations was studied in some dryland regions in Iran based on normalized Root Mean Square Error (nRMSE), Kling-Gupta Efficiency (KGE), and Percent Bias (PBIAS). The data combination scenarios were WS-SO: soil observations and gridded weather datasets including ERA5-Land (WE-SO) and the combinations of non-precipitation ERA5-Land forcings with CHIRPS (WCE-SO) and PERSIANN-CDR (WPE-SO), SoilGrids250m gridded soil data and weather observations (WO-SS), and soil and weather observations (WO-SO). Although the CHIRPS-ERA5L improved simulations relative to ERA5-Land and PERSIANN-CDR-ERA5-Land, there was still an nRMSE greater than 30 %, a KGE below 0.50, and an absolute PBIAS exceeding 25 % for dry farming yield in most drylands under WS-SS and WS-SO, indicating significant input uncertainties. The high uncertainty in dry farming wheat yield under WS-SS and WS-SO can be attributed to the uncertainties in simulating the water stress index in CSM-CERES-Wheat. The dry farming wheat yield was, however, simulated satisfactorily with SoilGrids250m products for WO-SS. The dry farming wheat yield showed the largest sensitivity to the uncertainty in precipitation forcing. The notable uncertainty in water stress simulation, and therefore in dry farming yield, appears to stem from the high uncertainty in precipitation products. These findings demonstrate that dry farming modeling is subject to notable input uncertainty when reliable meteorological records are lacking in our study area. SoilGrids250m can be reliably used to model wheat yield under dry farming conditions in the study area when weather observations are available. However, the applicability of SoilGrids250m largely depends on the availability of regional soil observations. Irrigated wheat yield was successfully simulated due to the reduced uncertainty in water stress. Therefore, using alternate weather-soil data provides a robust solution to data unavailability when wheat water requirements are sufficiently met. Nonetheless, caution is needed when using gridded weather datasets to force the CSM-CERES-Wheat model for dry farming.

中文翻译：

---

CSM-CERES-wheat 建模中的输入不确定性：使用替代天气和土壤数据的旱作农业和灌溉条件


在目前的研究中，分析了在旱作和灌溉条件下使用网格化土壤和天气数据集进行小麦建模的不确定性。在这方面，基于归一化均方根误差 （nRMSE）、Kling-Gupta 效率 （KGE） 和百分比偏差 （PBIAS），在伊朗的一些旱地地区研究了使用不同天气-土壤数据组合强制的 CSM-CERES-Wheat 模型的性能。数据组合情景为 WS-SO：土壤观测和网格化天气数据集，包括 ERA5-Land （WE-SO） 和非降水 ERA5-Land 强迫与 CHIRPS （WCE-SO） 和 PERSIANN-CDR （WPE-SO）、SoilGrids250m 网格化土壤数据和天气观测 （WO-SS） 以及土壤和天气观测 （WO-SO）。尽管 CHIRPS-ERA5L 相对于 ERA5-Land 和 PERSIANN-CDR-ERA5-Land 改进了模拟，但在 WS-SS 和 WS-SO 下，大多数旱地的旱作产量仍存在大于 30% 的 nRMSE、低于 0.50 的 KGE 和绝对 PBIAS 超过 25%，表明存在重大的输入不确定性。WS-SS 和 WS-SO 下旱作小麦产量的高不确定性可归因于模拟 CSM-CERES-小麦水分胁迫指数的不确定性。然而，使用 WO-SS 的 SoilGrids250m 产品对旱作小麦产量进行了令人满意的模拟。旱作小麦产量对降水强迫的不确定性表现出最大的敏感性。水分胁迫模拟的显著不确定性，以及旱作农业产量的显著不确定性，似乎源于降水产物的高度不确定性。这些发现表明，当我们的研究区域缺乏可靠的气象记录时，旱作农业建模会受到显着输入不确定性的影响。 当天气观测可用时，SoilGrids250m 可以可靠地用于模拟研究区域旱作条件下的小麦产量。然而，SoilGrids250m 的适用性在很大程度上取决于区域土壤观测的可用性。由于水分胁迫的不确定性降低，成功模拟了灌溉小麦产量。因此，当小麦水分需求得到充分满足时，使用替代天气-土壤数据为数据不可用提供了一个强大的解决方案。尽管如此，在使用网格化天气数据集强制 CSM-CERES-Wheat 模型进行旱作农业时需要谨慎。