With the influence of climate change and human activities, the frequency and scale of forest fires have been increasing continuously, posing a significant threat to the environment and human safety. Therefore, rapid and accurate forest fire detection has become essential for effective control of forest fires. This study proposes a Forest Fire Detection and Segmentation Model (FFDSM) based on unmanned aerial vehicle (UAV) infrared images to address the problems of forest fire occlusion and the poor adaptability of traditional forest fire detection methods. The FFDSM integrates the YOLO (You Only Look Once) v5s-seg, Efficient Channel Attention (ECA), and Spatial Pyramid Pooling Fast Cross-Stage Partial Channel (SPPFCSPC) to improve the detection accuracy of forest fires of different sizes. The FFDSM enhances the detection and extraction capabilities of forest fire features, enabling the accurate segmentation of forest fires of different sizes and shapes. Furthermore, we conducted ablation and controlled experiments on different attention mechanisms, spatial pyramid pooling (SPP) modules, and fire sizes to verify the effectiveness of the added modules and the adaptability of the FFDSM model. The results of the ablation experiment show that, compared to the original YOLOv5s-seg model, the models fused with the ECA and SPPFCSPC achieve an improved accuracy, with FFDSM showing the greatest improvement. FFDSM achieves a 2.1% increase in precision, a 2.7% increase in recall, a 2.3% increase in mAP@0.5, and a 4.2% increase in mAP@0.5:0.95. The results of the controlled experiments on different attention mechanisms and SPP modules demonstrate that the ECA+SPPFCSPC model (FFDSM) performs the best, with a precision, recall, mAP@0.5, and mAP@0.5:0.95 reaching 0.959, 0.870, 0.907, and 0.711, respectively. The results of the controlled experiment on different fire sizes show that FFDSM outperforms YOLOv5s-seg for all three fire sizes, and it performs the best for small fires, with a precision, recall, mAP@0.5, and mAP@0.5:0.95 reaching 0.989, 0.938, 0.964, and 0.769, respectively, indicating its good adaptability for early forest fire detection. The results indicate that the forest fire detection model based on UAV infrared images (FFDSM) proposed in this study exhibits a high detection accuracy. It is proficient in identifying obscured fires in optical images and demonstrates good adaptability in various fire scenarios. The model effectively enables real-time detection and provides early warning of forest fires, providing valuable support for forest fire prevention and scientific decision making.

How to distinguish the relative role of climate change and human activities in vegetation dynamics has attracted increasing attention. However, most of the current studies concentrate on arid and semiarid regions, while the relative contributions of climate change and human activities to vegetation changes remain unclear in warm-humid regions. Based on the normalized difference vegetation index (NDVI) and climatic variables (temperature, precipitation, radiation) during 2001–2020, this study used the Theil–Sen median trend analysis, partial correlation analysis, and residual trend analysis to analyze the spatiotemporal pattern of vegetation trends, the response of vegetation to climate variations, and the climatic and anthropogenic contributions to vegetation dynamics in the warm and humid Guangdong Province of south China. Results showed that the NDVI in most areas exhibited an increasing trend. Changes in climatic variables displayed different spatial variations which, however, were not significant in most areas. Vegetation responded diversely to climate change with temperature as the most important climatic factor for vegetation improvement in most areas, while precipitation was the dominant climatic factor in the southern edge region and radiation was the dominant climatic factor in the central and western regions. Vegetation in most areas was influenced by both climate change and human activities, but the contribution rate of human activities was commonly much higher than climate change. The findings of this study are expected to enhance our understanding of the relative climatic and anthropogenic contributions to vegetation changes in warm-humid regions and provide a scientific basis for future ecological policies and ecosystem management in highly urbanized regions.

Accurate and detailed cropland maps are essential for food security, yet existing products for China exhibit substantial discrepancies. This study presents CropLayer, a 2 m resolution cropland map of China for 2020, developed from Mapbox and Google satellite imagery. The framework comprises three key stages: (1) image quality assessment (IQA) using a ResNet model to compensate for missing acquisition metadata; (2) cropland extraction via an active learning strategy guided by a Mask2Former segmentation model and XGBoost-based semantic correctness evaluation; and (3) integration of Mapbox and Google results through an XGBoost model informed by four feature groups: Geography, IQA, Regional Property, and Consistency. A three-level validation scheme (pixel, block, and region) ensures robust and interpretable accuracy across spatial scales. CropLayer achieves a pixel-level accuracy of 88.73 %, a block-level semantic correctness of 96.5 %, and provincial-level consistency, with 30 out of 32 provinces showing area estimates within ±10 % of official statistics. In comparison, only 1–9 provinces meet this criterion across eight existing datasets. CropLayer provides a reliable, high-resolution baseline for agricultural structure analysis, yield estimation, and land use planning in China. The CropLayer dataset is available at https://doi.org/10.5281/zenodo.14726428 (Jiang et al., 2025).

The canopy of perennial evergreen fruit trees in southern China has a unique Bidirectional Reflectance Factor (BRF) due to its complex multi-branch structure and density changes. This study aimed to address the lack of clarity regarding the changes in BRF of evergreen fruit trees in southern China. Litchi, a typical fruit tree in this region, was chosen as the subject for establishing a three-dimensional (3D) real structure model. The canopy BRF of litchi was simulated under different leaf components, illumination geometry, observed geometry, and leaf area index (LAI) using a 3D radiation transfer model. The corresponding changes in characteristics were subsequently analyzed. The findings indicate that the chlorophyll content and equivalent water thickness of leaves exert significant influences on canopy BRF, whereas the protein content exhibit relatively weak effects. Variation in illumination and observation geometry results in the displacement of hotspots, with the solar zenith angle and view zenith angle exerting significant influence on the BRF. As the LAI of the litchi orchard increases, the distribution of hotspots becomes more concentrated, and the differences in angle information are relatively smaller when observed from multiple angles. With the increase in LAI in litchi orchards, the BRF on the principal plane would be saturated, but observation at hotspots could alleviate this phenomenon. The above analysis provides a reference for quantitative inversion of vegetation parameters using remote sensing monitoring information of typical perennial evergreen fruit trees.

An efficient monitoring of forest fire risk and its dynamic changes is an important way to prevent and reduce forest fire hazards. This study integrated the ignition effect of fire sources and the impact of landform into the calculation of forest fire risk using digital-terrain-slope factor, and developed an optimized forest fire risk model (fire-potential-index slope, FPIS). Combined with Landsat 8 satellite images, the study retrieved and analyzed the variations of forest fire risk in Zhaoqing City, Guangdong province, for four consecutive periods in the dry season, 2019. It was found that the high forest fire risk area was mainly distributed in the valley plains of Huaiji district, Fengkai district and Guangning district, the depressions of the Sihui district, and mountain-edge areas of Dinghu district and Gaoyao district, and accounted for 8.9% on 20 October but expanded to 19.89% on 7 December 2019. However, the further trend analysis indicated that the forest fire risk with significant increasing trend only accounted for 6.42% in Zhaoqing. Compared to the single high forest fire risk results, the changing trend results effectively narrowed the key areas for forest fire prevention (2.48%–12.47%) given the actual forest fires in the city. For the four forest fire events (Lingshan mountain, Hukeng industrial area, Xiangang county and Huangniuling ridge forest fires), it was found that the forest fire risk with significant increasing trend in these regions accounted for 26.63%, 35.84%, 54.6% and 73.47%, respectively, which further proved that the forest fire risk changing trend had a better indicated significance for real forest fire events than the high forest fire risk results itself (1.89%–71.69%). This study suggested that the forest fire risk increasing trend could be well used to reduce the probability of misjudgment and improve the accuracy of the early-warning areas when predicting forest fires.

Accurate and detailed cropland maps are essential for agricultural planning, resource management, and food security, particularly in countries like China, where agricultural productivity is high but resources are limited. Despite the availability of several medium-to-high-resolution satellite-based cropland maps, significant discrepancies in area estimates and spatial distribution persist, limiting their utility. This study proposes a data-driven framework for cropland mapping that leverages 2 m High Resolution (HR) imagery from Mapbox and Google. The framework consists of three main stages: First, national imagery is partitioned into 0.05°×0.05° blocks for efficient parallel computation. An Image Quality Assessment (IQA) using ResNet models is performed on both sources to address the challenge of missing image acquisition metadata. Second, a robust cropland identification model integrates Mask2Former for precise segmentation and XGBoost for error evaluation, facilitating iterative improvements through active learning. Finally, a novel integration strategy combines four feature groups—Geography, IQA, Region Property, and Consistency—using XGBoost to merge the datasets into a unified cropland layer, named Croplayer. The Croplayer dataset achieves an overall mapping accuracy of 88.73 %, with 30 out of 32 provincial units reporting area estimates within ±10 % of official statistics. In contrast, only 1 to 9 provinces from seven other existing datasets meet the same accuracy standard. The results highlight Croplayer's potential for applications such as crop yield estimation and agricultural structure analysis, offering a reliable tool for addressing agricultural and food security challenges.