This dissertation consists of three self-contained, yet closely related, papers summarizing the results of a comprehensive study aimed at the development of a decision support system for sustainable water resources management for the Zhangye Basin in northwestern China. The first paper presents a 3D groundwater flow model to represent groundwater dynamics of the basin from 1999 through 2010 using MODFLOW-2005. The regional 3D groundwater model provides reliable information of the flow field and produces detailed water budgets for management purposes. It defines the intensive groundwater-surface water interaction zones, and reveals the increasing flux exchange due to both climate change and human activities. The second paper presents an integrated 3D groundwater-surface water flow model using GSFLOW. The model calibration was done by first running PRMS and MODFLOW-2005 models separately, and then followed by the calibration of the integrated GSFLOW model. The model shows a detailed trend of water storage changes and their relationship with each inflow and outflow item. More importantly, this study demonstrates the applicability of integrated basin-scale models in characterizing the groundwater-surface water(GWSW) interaction, reproducing the flow system, and supporting sustainable water resources management while accounting for the effects of climate change in arid inland river basins. The third paper presents an efficient decision support tool, taking into consideration the relevant complexities and interactions in different water resource components, to inform decision making for water resources management for the Zhangye basin. On the basis of data collection and data mining, incorporated with integrated hydrological conceptual models and numerical models, a Bayesian network (BN) has been developed and calibrated by K-fold cross validation. The trained BN model captures the important hydrological cycle characteristics and uncertainty of related factors to provide the optimal management solutions under consideration. While this study is based on the Zhangye basin, the concepts and approaches developed in this study are of general applicability. The integrated GWSW modeling, coupled with a BN construct, provides an innovative tool to inform decision making in water resources management.