Scenario analysis is the basis of developing rational water management practices (WMPs) for watersheds. How to predict future hydrological responses on a regional-scale is still a challenge for modeling work in irrigated watersheds with shallow groundwater environments. Therefore, this paper presents an efficient realization of predicting regional agroecosystem responses and searching for appropriate WMPs, through using a water balance-based, semi-distributed hydrological model (SWAT-AG). The scenario case study is carried out in the Jiyuan Irrigation System located in the Hetao of upper Yellow River basin, based on the calibrated and validated modeling work in our previous companion paper. Eight scenarios of water-saving practices (WSPs) are proposed, with consideration for reducing irrigation depth and controlling initial groundwater depth. Then the coupled responses of agroecosystem processes to various WSPs are predicted for the case study region in 2012 and 2013, mainly related to the groundwater depth, root zone soil water and salinity, and crop yield/natural vegetation biomass. Based on the analysis for proposed scenarios, the 100% of present irrigation depth combined with increasing initial GWD by 50 cm are recommended as appropriate WSPs for dry years, and the 80% of present irrigation depth combined with increasing initial GWD by 100 cm are recommended for wet years, in order to maintain good environmental conditions for both crops and natural vegetation. In addition, results show that SWAT-AG could overcome the scale/function limitations of traditional soil/crop models and also avoid computational issues of numerical models. We further point out that the scenarios in reality will be more complicated and comprehensive in time and space, and thus the predictions should be updated accordingly. Overall, this case study fully presents the feasibility and practicality of using the SWAT-AG model to realize the scenario response analysis and water management decision-making on a region scale for irrigated watersheds with shallow groundwater environments.

情景分析是为流域制定合理的水管理规范（WMP）的基础。在具有浅层地下水环境的灌溉流域中，如何对工作进行建模仍然是一个挑战。因此，本文通过使用基于水平衡的半分布式水文模型（SWAT-AG），提出了预测区域农业生态系统响应和寻找合适WMP的有效方法。在我们之前的伴随论文中经过校准和验证的建模工作的基础上，在黄河上游河套的济源灌溉系统中进行了案例研究。提出了八种节水措施（WSP），以减少灌溉深度和控制初始地下水深度。然后，以2012年和2013年为例，预测了农业生态系统过程对各种WSPs的耦合响应，主要与地下水深度，根区土壤水分和盐分以及作物产量/天然植被生物量有关。根据对拟议方案的分析，建议在干旱年份将当前灌溉深度的100％加上初始GWD增加50 cm作为适当的WSP，并建议将当前灌溉深度的80％加上初始GWD增加100 cm在潮湿的年份，以保持作物和自然植被的良好环境条件。此外，结果表明，SWAT-AG可以克服传统土壤/作物模型的规模/功能限制，并且可以避免数值模型的计算问题。我们进一步指出，现实中的场景在时间和空间上将更加复杂和全面，因此，预测也应相应地进行更新。总体而言，本案例研究充分展示了使用SWAT-AG模型在具有浅层地下水环境的流域进行区域情景下的情景响应分析和水管理决策的可行性和实用性。