Unmanned aerial vehicle (UAV) remote sensing has become an important tool for high-resolution tree species identification in orchards and forests. However, irregular spatial distribution, overlapping canopies, and small crown sizes still limit detection accuracy. To overcome these challenges, we propose YOLOv11-OAM, an enhanced one-stage object detection model based on YOLOv11. The model incorporates three key modules: omni-dimensional dynamic convolution (ODConv), adaptive spatial feature fusion (ASFF), and a multi-point distance IoU (MPDIoU) loss. A class-balanced augmentation strategy is also applied to mitigate category imbalance. We evaluated YOLOv11-OAM on UAV imagery of six fruit tree species—walnut, prune, apricot, pomegranate, saxaul, and cherry. The model achieved a mean Average Precision (mAP@0.5) of 93.1%, an 11.4% improvement over the YOLOv11 baseline. These results demonstrate that YOLOv11-OAM can accurately detect small and overlapping tree crowns in complex orchard environments, offering a reliable solution for precision agriculture and smart forestry applications.

Xinjiang is an important forest and fruit production area in China, and Cytospora canker, caused by the genus Cytospora Ehrenb., has caused serious losses to forestry production in Xinjiang. In this study, we constructed ensemble models based on Biomod2 to assess the potential geographical distribution of Cytospora chrysosperma, C. nivea, and C. mali in Xinjiang, China and their changes under different climate change scenarios, using species occurrence data and four types of environmental variables: bioclimatic, topographic, NDVI, and soil. The model performance assessment metrics (AUC and TSS) indicated that the ensemble models are highly reliable. The results showed that NDVI had the most important effect on the distribution of all three species, but there were differences in the response patterns, and bioclimatic factors such as temperature and precipitation also significantly affected the distribution of the three species. C. chrysosperma showed the broadest ecological adaptation and the greatest potential for expansion. C. nivea and C. mali also showed expansion trends, but to a lesser extent. The overlapping geographical distribution areas of the three species increased over time and with an intensification of the climate scenarios, especially under the high-emission SSP585 scenario. The centroids of the geographical distribution for all three species generally shifted towards higher latitude regions in the northeast, reflecting their response to climate warming. C. chrysosperma may become a more prevalent forest health threat in the future, and an increase in the overlapping geographical distribution areas of the three species may lead to an increased risk of multiple infections. These findings provide an important basis for understanding and predicting the distribution and spread of the genus Cytospora in Xinjiang and are important for the development of effective forest disease prevention and control strategies.

With global climate change, urbanization, and agricultural resource limitations, precision agriculture and crop monitoring are crucial worldwide. Integrating multi-source remote sensing data with deep learning enables accurate crop mapping, but selecting optimal network architectures remains challenging. To improve remote sensing-based fruit planting classification and support orchard management and rural revitalization, this study explored feature selection and network optimization. We proposed an improved CF-EfficientNet model (incorporating FGMF and CGAR modules) for fruit planting classification. Multi-source remote sensing data (Sentinel-1, Sentinel-2, and SRTM) were used to extract spectral, vegetation, polarization, terrain, and texture features, thereby constructing a high-dimensional feature space. Feature selection identified 13 highly discriminative bands, forming an optimal dataset, namely the preferred bands (PBs). At the same time, two classification datasets—multi-spectral bands (MS) and preferred bands (PBs)—were constructed, and five typical deep learning models were introduced to compare performance: (1) EfficientNetB0, (2) AlexNet, (3) VGG16, (4) ResNet18, (5) RepVGG. The experimental results showed that the EfficientNetB0 model based on the preferred band performed best in terms of overall accuracy (87.1%) and Kappa coefficient (0.677). Furthermore, a Fine-Grained Multi-scale Fusion (FGMF) and a Condition-Guided Attention Refinement (CGAR) were incorporated into EfficientNetB0, and the traditional SGD optimizer was replaced with Adam to construct the CF-EfficientNet architecture. The results indicated that the improved CF-EfficientNet model achieved high performance in crop classification, with an overall accuracy of 92.6% and a Kappa coefficient of 0.830. These represent improvements of 5.5 percentage points and 0.153, compared with the baseline model, demonstrating superiority in both classification accuracy and stability.

Jujube is an important economic fruit tree in China, with Xinjiang being the largest production region, where flowering-stage heat stress threatens yield stability. Assessing heat risk in jujube orchards requires consideration of crop-specific phenological sensitivity and heat injury thresholds. Therefore, this study developed a Heat Injury Accumulating Index for jujube (HIS_j) to assess flowering-stage heat injury by integrating temperature–humidity stress and phenological sensitivity. A Hazard–Vulnerability–Exposure framework combined with Landsat image was applied for spatial heat risk assessment, atmospheric circulation analysis and interpretable machine learning were integrated to identify the drivers of hazard and vulnerability. Results show that HIS_j exhibited pronounced spatial heterogeneity and a distinct spatial dipole-like pattern, with higher values mainly concentrated in the oasis regions of eastern and southern Xinjiang, and an approximately six-year oscillation across most regions. High risk areas were mainly concentrated in Turpan, Yizhou District, and southern Bazhou, whereas Aksu, Kashgar, and most of Hotan were predominantly characterized by moderate-risk conditions. Climatic heat hazard and integrated orchard-scale risk were not spatially equivalent, with some intensively managed orchard areas exhibiting comparatively lower risk despite relatively strong meteorological heat stress. Variations in HIS_j were associated with the Western Pacific Subtropical High, Eurasian zonal circulation, and the India–Burma trough. NDVI was used as an observational proxy for orchard vegetation condition to characterize vegetation-based vulnerability. NDVI showed nonlinear associations with hydrothermal conditions and elevation, with threshold-like responses around 0.10–0.15 cm³/cm³ for soil moisture and around 800 m for elevation. The relatively weak and non-monotonic SHAP contribution of HIS_j to NDVI suggests that the association between meteorological heat stress and orchard vegetation conditions may vary with local environmental conditions. These findings provide a spatially explicit basis for understanding flowering-stage heat risk and supporting targeted adaptation in arid oasis jujube orchards.

Accurate vineyard mapping is important for agricultural resource monitoring, land-use management, and inventory updating in arid regions. However, vineyard identification in Xinjiang, China, is challenged by fragmented parcels, exposed soil backgrounds, irrigation-driven heterogeneity, and strong accumulated-temperature gradients. This study developed a multi-source remote-sensing framework for vineyard mapping and inventory updating by integrating Sentinel-1/2 data, terrain variables, growing-degree-day-derived agrothermal zones (ATZs), and seasonal-difference features. RF, LightGBM, XGBoost, 1D-CNN, and a Stacking ensemble were evaluated using polygon-level in-distribution testing and Leave-One-ATZ-Out cross-zone validation. The in-distribution test was used to assess vineyard separability under similar sample distributions, whereas cross-ATZ validation was used to evaluate model transferability across heterogeneous thermal domains. Tree-based models and Stacking achieved near-ceiling performance under the in-distribution setting, but cross-ATZ validation revealed substantial performance degradation, indicating that conventional local validation can overestimate operational transferability. RF achieved the highest mean cross-zone F1-score, while Stacking achieved the highest cross-zone AP and Recall and provided a flexible probability surface for thresholding, mosaicking, vector post-processing, and patch-level mapping. In Gaochang District, the Stacking-derived result identified 22,998.84 ha of potential vineyard area, achieved an Area Recall of 0.7908 against the historical inventory, and covered 88.03% of existing parcels at ≥30% spatial overlap. The workflow also identified 1356 candidate vineyard patches for inventory updating covering 2502.32 ha for subsequent inventory verification. These results demonstrate that multi-source feature integration and probability-based ensemble mapping can support vineyard mapping and inventory updating in arid regions, while cross-zone validation is essential for assessing operational generalization.