Convolutional neural networks (CNNs) have achieved success in HSI classification, but the performance is constrained by the limited reception field. In this regard, vision transformer is introduced recently, which is of powerful capabilities in long-range feature extraction for HSI classification. However, transformers are computation intensive and poor for local feature extraction. The motivation for this study is to build a lightweight hybrid model, which ensembles the respective inductive bias from CNNs and global receptive field from transformers. In this work, we propose a concise and efficient framework—the spatial-spectral feature extraction network with patch attention module (PASSNet), to simultaneously extract both local and global features. Specifically, we design an innovative plugin called patch attention module (PAM), which can be easily integrated into both CNNs and transformers blocks to extract spatial-spectral features from multiple spatial perspectives. Besides, a novel partial convolution operation is introduced, with a reduced computational cost than vanilla convolution operation. Through coupling the local attention from the CNNs with the global receptive fields in the transformers, the proposed PASSNet exhibits a superior classification performance on three well-known datasets with a small training sample size.

Grassland ecosystems are a significant part of the global ecosystem and support the livelihoods of millions of people. The Inner Mongolia grassland is the largest temperate grassland in the world, and an important ecological barrier for China, but due to human activities and climate change it has been faced with an ecological crisis in recent years. In this study, a modified Carnegie-Ames-Stanford approach (CASA) model based on the Google Earth Engine platform was used to determine the net primary production (NPP) in the Inner Mongolia grassland from 2000 to 2020. The results show that the average annual NPP of the Inner Mongolia grassland is 278.63 g C/m², and 83.22% of the total area has shown an increasing trend during the study period. We also analyzed the impact of land-use/cover change (LUCC) and climatic factors on NPP. We found that: (1) the total area of grassland increased from 2000 to 2010 and then decreased from 2010 to 2020. During the whole study period, although the grassland area increased slightly by 4728.69 km² because of LUCC, the overall effect of LUCC on grassland NPP was negative, with a loss of 17.63 Tg C compared to an increase of 16.38 Tg C. (2) The main meteorological factor affecting the NPP of the Inner Mongolia grassland is precipitation, followed by sunshine duration and temperature. About 97.06% of the grassland shows a positive correlation between NPP and precipitation. (3) The results for NPP and its changing trends are not completely consistent in the long- and short-term study periods. Considering the instability of grassland growth, it is necessary to take the periodic variation of precipitation into account when studying NPP. These results could provide basic information for policy formulation and scientific research into the ecological environment management of grassland areas in the future.

This article describes a new teaching method for a university geography course based on the core idea that everyone should be “makers” rather than passive learners. After enough teaching and learning preparation, freshman students in a Chinese university were asked to select keywords on the topic “What is human geography?” Nearly half of the selected keywords were different from what they had learned previously. They were linked to the frontier of human geography, to some extent, demonstrating that students were “makers.” The result shows the method of keyword selection and the idea that “we are all makers” can help the teacher and students reconstruct the process of teaching and learning, even for newcomers to the field. Thus, students could learn more autonomously and efficiently. At the same time, they would try to make new progress. Key Words: China, human geography, keyword selection, learner, maker.

Hyperspectral image feature extraction plays a crucial role in reducing the redundancy and correlation among spectral bands while preserving the essential information. Knowledge-driven feature extraction methods, such as spectral indices (SIs), leverage the interaction mechanisms between electromagnetic waves and materials to enhance the characteristic attributes of ground objects through band operations. These methods offer key advantages, including strong physical interpretability, simple construction, and robust scene reusability. However, most of the existing SIs still rely on expert knowledge tailored to specific scenarios, leading to inherent limitations, such as subjectivity, high time consumption, and implementation complexity. In this article, to address these challenges, we propose a hyperspectral image multi-agent deep reinforcement learning feature extraction (HI-MAFE) algorithm, aiming to alleviate the burden of manual SIs design by human experts. HI-MAFE employs a heuristic “generation-selection” strategy to simulate the decision-making process of domain experts, with specifically designed deep reinforcement learning (DRL) models for both the generation and selection steps. To accelerate exploration in a high-dimensional action space, the model incorporates a multi-agent deep reinforcement learning (MADRL) framework. The experimental results demonstrate the effectiveness and superiority of the proposed algorithm for hyperspectral image classification. The proposed HI-MAFE framework leverages DRL to autonomously generate meaningful environmental interpretation from spectral data, thereby reducing the reliance on manually designed SIs. This research can inspire future work in SIs construction and complement the limitations of data-driven approaches.