Increasing planting density is one of the main management practices contributing to maize grain yield improvements across the world, but the soil water sustainability after long-term continuous high-density under rainfed farming systems is not clear. Crop simulation models may help explore suitable management systems for increasing crop productivity and economic benefits. The present study calibrated and validated the DSSAT-CERES-Maize model in a four-year field experiment, which coupled four densities from 52,500 to 97,500 plant ha⁻¹ and two cultivars under rainfed conditions. The calibrated CERES-Maize model performed fairly well in simulating the phenological dates, and the average root mean-squared error (RMSE) ranged from 0.7 to 2.8 d for anthesis and 0 to 2.8 d for maturity date. The normalized root mean squared errors (nRMSE) for biomass and grain yield were 17.5% and 12.4%, respectively. The average nRMSE for soil water dynamics in the 0–200 cm soil layers was 15.6% among the different growth stages. The calibrated model was subsequently used to evaluate the soil water regime and crop productivity in response to planting density under 38 years of meteorological data. The results showed that maize water productivity and evapotranspiration (ET) fluctuated with seasonal rainfall, and normal and wet years were significantly higher than dry years. Although ET during the growing season tended to increase with increasing density, the long-term continuous high planting density did not cause excessive soil water consumption. Grain yield and water use efficiency (WUE) tended to exhibit a parabolic relationship with planting density during the long-term simulated seasons between different experimental sites. No significant differences were detected between different cultivars in water productivity under long-term simulation. However, the simulation results suggested that the optimal planting density was often related to variability in climate conditions between sites and years. The scenario simulation results suggest that the optimal density should not exceed 67,500 plant ha⁻¹ when the annual precipitation is less than 500 mm, but it should not exceed 82,500 plant ha⁻¹ in areas where the rainfall is greater than 500 mm. Therefore, this study suggests that moderate planting density has the potential to realize sustainable maize development in dryland farming systems on the Loess Plateau and similar areas.

提高种植密度是世界范围内提高玉米产量的主要管理措施之一，但在雨养农业系统下长期连续高密度种植后土壤水分的可持续性尚不清楚。作物模拟模型可能有助于探索合适的管理系统，以提高作物生产力和经济效益。本研究在为期四年的田间试验中校准和验证了 DSSAT-CERES-玉米模型，该模型将 52,500 至 97,500 公顷植物的四种密度耦合^-1和两个雨养条件下的栽培品种。校准后的 CERES-Maize 模型在模拟物候日期方面表现相当好，平均均方根误差 (RMSE) 范围为开花期为 0.7 至 2.8 天，成熟期为 0 至 2.8 天。生物量和谷物产量的归一化均方根误差 (nRMSE) 分别为 17.5% 和 12.4%。0-200 cm 土层土壤水分动力学的平均 nRMSE 在不同生长阶段为 15.6%。校准模型随后用于评估土壤水分状况和作物生产力，以响应 38 年气象数据下的种植密度。结果表明，玉米水分生产力和蒸散量（ET）随季节降雨而波动，正常年和湿润年显着高于干旱年。虽然生长季ET随着密度的增加有增加的趋势，但长期连续高种植密度并未造成土壤耗水量过大。在不同试验地点之间的长期模拟季节中，粮食产量和水分利用效率（WUE）往往与种植密度呈抛物线关系。在长期模拟下，不同品种的水分生产力没有显着差异。然而，模拟结果表明，最佳种植密度通常与地点和年份之间气候条件的变化有关。情景模拟结果表明，最佳种植密度不应超过 67,500 公顷 长期连续高种植密度并未造成土壤耗水量过大。在不同试验地点之间的长期模拟季节中，粮食产量和水分利用效率（WUE）往往与种植密度呈抛物线关系。在长期模拟下，不同品种的水分生产力没有显着差异。然而，模拟结果表明，最佳种植密度通常与地点和年份之间气候条件的变化有关。情景模拟结果表明，最佳种植密度不应超过 67,500 公顷 长期连续高种植密度并未造成土壤耗水量过大。在不同试验地点之间的长期模拟季节中，粮食产量和水分利用效率（WUE）往往与种植密度呈抛物线关系。在长期模拟下，不同品种的水分生产力没有显着差异。然而，模拟结果表明，最佳种植密度通常与地点和年份之间气候条件的变化有关。情景模拟结果表明，最佳种植密度不应超过 67,500 公顷 在不同试验地点之间的长期模拟季节中，粮食产量和水分利用效率（WUE）往往与种植密度呈抛物线关系。在长期模拟下，不同品种的水分生产力没有显着差异。然而，模拟结果表明，最佳种植密度通常与地点和年份之间气候条件的变化有关。情景模拟结果表明，最佳种植密度不应超过 67,500 公顷 在不同试验地点之间的长期模拟季节中，粮食产量和水分利用效率（WUE）往往与种植密度呈抛物线关系。在长期模拟下，不同品种的水分生产力没有显着差异。然而，模拟结果表明，最佳种植密度通常与地点和年份之间气候条件的变化有关。情景模拟结果表明，最佳种植密度不应超过 67,500 公顷^-1当年降水量小于500 毫米时，但在降水量大于500 毫米的地区不应超过82,500 公顷^-1。因此，本研究表明，适度的种植密度有可能在黄土高原和类似地区的旱地农业系统中实现玉米的可持续发展。