Abstract Crop models may potentially explore alternative ways to improve agroecosystem resilience in arid regions of Middle East and North Africa. Mapping the outputs behavior as a function of the inputs and quantifying the uncertainty contribution of inputs to the variability of outputs are crucial for understanding and applying complex mathematical models to a new environment. Objectives of present research are (i) to calibrate and evaluate the Decision Support System for Agrotechnology Transfer (DSSAT) cropping system model using detailed experimental datasets on maize production in arid sandy soils (Entisol) and (ii) to determine the model’s sensitivity to soil, genotype and crop management inputs under the currently explored conditions (low fertility and water holding capacity) based on multivariate analysis and variance decomposition methods. The goodness-of-fit statistics between observed and simulated data indicated that the calibrated model reasonably well simulates maize phenology, growth and yield, evapotranspiration, soil water content, grain N concentration, and postharvest soil NO3-N in eight year site field experiments. A global sensitivity analysis using the co-inertia method was carried out to link 14 output variables and 25 soil and genotype input parameters. Maize growth and yield variables were strongly correlated with soil hydrological and fertility input parameters such as soil water upper limit (SDUL) and soil organic carbon (SOC), whereas simulation of maize phenology was largely determined by phenological genotype-specific cultivar input parameters. A strong association was also observed between the output variables of yield and soil fertility. The effect of carbon (C) related soil input parameters of initial SOC and stable SOC and crop management factors of maize residue retention and compost application under no-till system on the long-term (10 years) simulation of yield and SOC dynamics was further explored using Sobolʹ method. Simulated grain yield, water productivity, active SOC, and cumulative soil CO2 efflux were most sensitive to initial stable SOC and compost application. Maize residue retention significantly affected the simulation of cumulative N mineralization, SOC % in 0.2 m depth, and cumulative soil CO2 efflux through interactions effect, i.e. total-order sensitivity index (STi) > 0.05, with other inputs. Compost application increased grain yield by 13 %, SOC stock by 5%, and cumulative soil CO2 efflux by 95 % compared with no application. However, compost application with maize residue retained significantly reduced cumulative soil CO2 efflux by 12 % compared with compost application with maize residue removed. Therefore, the application of compost with maize residue retained under no-till system is a plausible crop management option for agronomically improved and environmentally sound maize production in arid sandy soils.

中文翻译：

---

DSSAT 模型模拟干旱地中海气候下玉米产量和土壤碳动态的敏感性：土壤、基因型和作物管理的影响

摘要 作物模型可能会探索提高中东和北非干旱地区农业生态系统恢复力的替代方法。将输出行为映射为输入的函数并量化输入对输出可变性的不确定性贡献对于理解复杂的数学模型并将其应用于新环境至关重要。本研究的目标是 (i) 使用干旱沙质土壤 (Entisol) 玉米生产的详细实验数据集校准和评估农业技术转让决策支持系统 (DSSAT) 种植系统模型和 (ii) 确定模型对土壤的敏感性基于多元分析和方差分解方法的当前探索条件（低肥力和持水能力）下的基因型和作物管理投入。观测数据和模拟数据之间的拟合优度统计表明，校准模型在 8 年的现场田间试验中较好地模拟了玉米物候、生长和产量、蒸发蒸腾、土壤含水量、谷物 N 浓度和收获后土壤 NO3-N。使用共惯性方法进行了全局敏感性分析，将 14 个输出变量和 25 个土壤和基因型输入参数联系起来。玉米生长和产量变量与土壤水分上限 (SDUL) 和土壤有机碳 (SOC) 等土壤水文和肥力输入参数密切相关，而玉米物候模拟很大程度上取决于物候基因型特定品种输入参数。在产量和土壤肥力的输出变量之间也观察到了很强的关联。进一步研究了免耕系统下初始 SOC 和稳定 SOC 的碳 (C) 相关土壤输入参数以及免耕系统下玉米残留物保持和堆肥施用的作物管理因素对产量和 SOC 动态的长期（10 年）模拟的影响使用 Sobolʹ 方法进行探索。模拟谷物产量、水分生产力、活性 SOC 和累积土壤 CO2 流出对初始稳定 SOC 和堆肥应用最敏感。玉米残留物保留显着影响了累积氮矿化、0.2 m 深度的 SOC % 和累积土壤 CO2 流出的模拟，通过相互作用效应，即全阶敏感性指数 (STi) > 0.05，与其他输入。与不施用相比，堆肥施用使谷物产量提高了 13%，SOC 储量提高了 5%，土壤 CO2 累积流出量提高了 95%。然而，与去除玉米残留物的堆肥应用相比，保留玉米残留物的堆肥应用显着减少了 12% 的累积土壤 CO2 流出。因此，在免耕系统下使用保留玉米残留物的堆肥是一种合理的作物管理选择，可在干旱沙质土壤中进行农艺改良和环境无害的玉米生产。