Due to climate change and human activities, the assumption of the stationarity of hydrologic features will no longer hold. Moreover, uncertainties, such as the apparent randomness of hydrologic elements, and complexities, such as the existence of various water users with different characteristics, also introduce huge challenges for water managers. To address nonstationarity, uncertainty, and complexity, a new approach is proposed for the optimal allocation of regional water resources. This objective was achieved via two steps: First, the generalized additive model was chosen to analyze the nonstationary probability distribution of the hydrologic dataset; then, an interval two-stage classified-allocation model is formulated by incorporating two-stage stochastic programming, interval parameter programming and classification thought. The model can not only address uncertainties, which were expressed as interval parameters and probability distributions, but can also handle complexities by classifying the water users into agricultural and nonagricultural users. The approach was applied to the Zhanghe Irrigation District to optimize available water allocation for municipality, industry, hydropower, and agriculture in two planning years (namely 2010 and 2015). The annual inflow of the Zhanghe Reservoir is found to be nonstationary and can be well fitted by Gamma distribution with one location parameter based on a nonlinear function of time. Moreover, the difference in output between the two years with different inflow probability distributions indicates the need for nonstationary analysis. Comparison to the inexact two-stage water management model that did not consider the variation of agricultural water requirement shows the meaning of classification. From the results, municipality and industry are more competitive than agriculture and then hydropower. For agriculture, winter rape and cotton have higher priority than rice. These solutions of the optimal targets and optimal water allocation for different water users can help managers to accurately develop allocation plans under uncertain and nonstationary conditions.

中文翻译：

---

非平稳条件下区域水资源管理的间隔两阶段分类分配模型

由于气候变化和人类活动的影响，水文特征平稳的假设将不再成立。此外，不确定性（例如水文要素的明显随机性）和复杂性（例如存在具有不同特征的各种用水者）也给水资源管理者带来了巨大挑战。针对非平稳性、不确定性和复杂性，提出了一种新的区域水资源优化配置方法。该目标通过两个步骤实现：首先，选择广义可加模型来分析水文数据集的非平稳概率分布；然后，结合两阶段随机规划、区间参数规划和分类思想，建立了区间两阶段分类分配模型。该模型不仅可以解决以区间参数和概率分布表示的不确定性，还可以通过将用水用户分为农业用户和非农业用户来处理复杂性。该方法应用于漳河灌区，在两个规划年（即2010年和2015年）优化市政、工业、水电和农业可用水资源配置。发现漳河水库的年流入量是非平稳的，可以很好地拟合 Gamma 分布，其中一个位置参数基于时间的非线性函数。此外，流入概率分布不同的两年之间的产出差异表明需要进行非平稳分析。与不考虑农业需水量变化的不精确的两阶段水管理模型相比，显示了分类的意义。从结果来看，市政和工业比农业和水电更具竞争力。在农业方面，冬油菜和棉花的优先级高于水稻。这些针对不同用水户的最优目标和最优配水的解决方案可以帮助管理者在不确定和非平稳条件下准确制定配水计划。