Groundwater plays a vital role in the arid inland river basins, in which the groundwater management is critical to the sustainable development of area economy and ecology. Traditional sustainable management approaches are to analyze different scenarios subject to assumptions or to construct simulation–optimization models to obtain optimal strategy. However, groundwater system is time-varying due to exogenous inputs. In this sense, the groundwater management based on static data is relatively outdated. As part of the Heihe River Basin (HRB), which is a typical arid river basin in Northwestern China, the Daman irrigation district was selected as the study area in this paper. First, a simulation–optimization model was constructed to optimize the pumping rates of the study area according to the groundwater level constraints. Three different groundwater level constraints were assigned to explore sustainable strategies for groundwater resources. The results indicated that the simulation–optimization model was capable of identifying the optimal pumping yields and satisfy the given constraints. Second, the simulation–optimization model was integrated with wireless sensors network (WSN) technology to provide real-time features for the management. The results showed time-varying feature for the groundwater management, which was capable of updating observations, constraints, and decision variables in real time. Furthermore, a web-based platform was developed to facilitate the decision-making process. This study combined simulation and optimization model with WSN techniques and meanwhile attempted to real-time monitor and manage the scarce groundwater resource, which could be used to support the decision-making related to sustainable management.

STEM education aims to cultivate innovative talents by improving students' ability to comprehensively apply interdisciplinary knowledge in solving practical problems. This paper first develops an innovation event model through the analysis of 50 historical innovation events that can be traced back to whole human history. The model divides the realization process of those innovation events into four steps: 1) pointing out a problem, 2) proposing solutions to the problem, 3) concrete implementation of those solutions, and 4) iterative modification process. And then, the relationship between innovation event model and STEM education is established from the perspectives of subject integration and constructivism of STEM education. Based on this model, we can understand some key issues in the implementation of STEM education from a top-down view, including the nature of STEM education, the knowledge integration model, and the relationship between subject-specific education and integrated education. This will help to gradually improve our cognition and understanding of STEM education, so as to achieve its initial goal of integrated and innovative education. This article will contribute to a holistic rethinking about how to renovate STEM education in different levels of schools and colleges, equipped with such an innovation event model.

Federated Learning is an emerging application paradigm of edge computing in smart cities. On the one hand, it enables efficient, private, and secure processing of sensitive data. On the other hand, it alleviates the burden of centralized data processing for the smart city. However, in real-world scenarios, performance degradation caused by domain adaptation has become a bottleneck that limits the widespread application of federated learning. Most existing approaches tackle the issue by designing novel local learning approaches to transfer knowledge among different domains while ignoring the optimization for global model aggregation. To address this issue, we propose a novel approach that leverages the label-free adversarial learning technique to evaluate the representations learned by the different domains under the global model. With the constraints of the federated setting, we minimize the discrepancy by aligning each distribution to the global distribution. Additionally, we have developed a fast detector to enhance the quality of the generated images. Through extensive experiments on image classification tasks, we have demonstrated promising results and shown that our approach can serve as a robust plugin for other local optimizers in Federated Learning.