Central Asia, which is one of the most complex regions in the world, is crisscrossed with transboundary rivers shared by several countries. This paper analyzes the effects of climate change and human activities on hydrological processes and water resources in these transboundary rivers over the past half century. The results show that the average temperature in the Tienshan Mountains experienced “sharp” increases in 1998, with the average temperature after 1998 being 1.0°C higher than that during 1960–1998. This rapid warming has accelerated glacier shrinkage and decreases in snow cover. These changes influenced the hydrological processes, causing an earlier runoff peak and aggravated extreme hydrological events. Changes in mountainous hydrological processes affected regional water availability and intensified future water crisis in Central Asia. The mismatched spatial distributions of water and land resources, along with the intense human activities (e.g., overexploitation of water resources), have ultimately led to the present water crisis in Central Asia's river basins. This is the main reason for ongoing water conflicts in the region's transboundary rivers and the ecological disaster of the Aral Sea.

Object detection has attracted increasing attention in the field of remote sensing image analysis. Complex backgrounds, vertical views, and variations in target kind and size in remote sensing images make object detection a challenging task. In this work, considering that the types of objects are often closely related to the scene in which they are located, we propose a convolutional neural network (CNN) by combining scene-contextual information for object detection. Specifically, we put forward the scene-contextual feature pyramid network (SCFPN), which aims to strengthen the relationship between the target and the scene and solve problems resulting from variations in target size. Additionally, to improve the capability of feature extraction, the network is constructed by repeating a building aggregated residual block. This block increases the receptive field, which can extract richer information for targets and achieve excellent performance with respect to small object detection. Moreover, to improve the proposed model performance, we use group normalization, which divides the channels into groups and computes the mean and variance for normalization within each group, to solve the limitation of the batch normalization. The proposed method is validated on a public and challenging dataset. The experimental results demonstrate that our proposed method outperforms other state-of-the-art object detection models.

The Tarim River Basin in Xinjiang, China, has a typical desert riparian forest ecosystem. Analysis of the resilience of this type of ecosystem under extreme drought conditions and ecological rehabilitation projects could provide a theoretical basis for understanding ecosystem stability and resistance, and provide new ecological rehabilitation measures to improve ecosystem resilience. We employed a quantitative framework to assess net primary productivity (NPP) resilience, emphasizing four aspects of NPP dynamics: NPP, NPP stability, NPP resistance, and maximum NPP potential. We compared ecosystem resilience across four time periods: before the implementation of ecological rehabilitation projects (1990–2000), during construction and partial implementation of ecological rehabilitation projects (2001–2012), during the initial project stage of ecological rehabilitation (2013–2015), and during the late project stage of ecological rehabilitation (2016–2018). There are three main finding of this research. (1) Mean NPP was increased significantly from 2013 and was decreased from 2016, especially in the main stream of the Tarim River and in the basins of eight of its nine tributary rivers. (2) Ecosystem resilience in 2013–2018 was greater than in 1990–2012, with the greatest NPP stability, mean NPP and NPP resistance, especially in part one of the river basin (the Aksu River, the Weigan-Kuche River, the Dina River, the Kaidu-Konqi River, and the main stream of the Tarim River). Ecosystem resilience in 2001–2012 was lowest when compared to 1990–2000 and 2013–2018, with lowest mean NPP, NPP stability, NPP resistance and maximum NPP potential, particularly in part two of the river basin (the Kashigr River, the Yarkand River and the Hotan River basins). Therefore, part one was most affected by ecological restoration projects. When 2013–2018 was divided into two distinct stages, 2013–2015 and 2016–2018, resilience in the latter stage was the lowest, with lowest mean NPP, NPP resistance and maximum NPP potential, especially in the main stream of the Tarim River. This may be due to unreasonable water conveyance in 2014–2015. (3) Ecological resilience has increased significantly in 2013–2015 after the implementation of ecological water transfer projects, river regulation, and natural vegetation enclosure projects. Ecosystem resilience could continue to increase even more in the future with the continued implementation of reasonable ecological water transfer projects.

With exploiting contextual information over large image regions in an efficient way, the deep convolutional neural network has shown an impressive performance for single-image super-resolution (SR). In this paper, we propose a new deep convolutional network by cascading multiple well-designed inception-residual blocks within the deep Laplacian pyramid framework to progressively restore the missing high-frequency details in the low-resolution images. By optimizing our network structure, the trainable depth of our proposed network gains a significant improvement, which in turn improves super-resolving accuracy. However, the saturation and degradation of training accuracy remains a critical problem. With regard to this, we propose an effective two-stage training strategy, in which we first use the images downsampled from the ground-truth high-resolution (HR) images to pretrain the inception-residual blocks on each pyramid level with an extremely high learning rate enabled by gradient clipping, and then the original ground-truth HR images are used to fine-tune all the pretrained inception-residual blocks for obtaining our final SR models. Furthermore, we present a new loss function operating in both image space and local rank space to optimize our network for exploiting the contextual information among different output components. Extensive experiments on benchmark data sets validate that the proposed method outperforms the existing state-of-the-art SR methods in terms of the objective evaluation as well as the visual quality.