The deep learning method for natural-image object detection tasks has made tremendous progress in recent decades. However, due to multiscale targets, complex backgrounds, and high-scale small targets, methods from the field of natural images frequently fail to produce satisfactory results when applied to aerial images. To address these problems, we proposed the DET-YOLO enhancement based on YOLOv4. Initially, we employed a vision transformer to acquire highly effective global information extraction capabilities. In the transformer, we proposed deformable embedding instead of linear embedding and a full convolution feedforward network (FCFN) instead of a feedforward network in order to reduce the feature loss caused by cutting in the embedding process and improve the spatial feature extraction capability. Second, for improved multiscale feature fusion in the neck, we employed a depth direction separable deformable pyramid module (DSDP) rather than a feature pyramid network. Experiments on the DOTA, RSOD, and UCAS-AOD datasets demonstrated that our method’s average accuracy (mAP) values reached 0.728, 0.952, and 0.945, respectively, which were comparable to the existing state-of-the-art methods.

The forests in China’s temperate semi-arid region play a significant role in water conservation, carbon storage, and biodiversity protection. An accurate estimation of their aboveground biomass (AGB) is crucial for assessing key ecological characteristics, such as forest carbon storage capacity, biodiversity, and ecological productivity. This provides a scientific basis for forest resource management and ecological conservation in this region. In this study, we extract 17 features related to the dominant species (Larix gmelinii and Betula platyphylla), including 7 vegetation indices derived from remote sensing data, 14 indices from 7 satellite bands, and 3 forest site characteristics. We then analyze the correlations between the AGB and these features. We compare the performance of AGB estimation models using linear regression (LR), polynomial regression (PR), ridge regression (RR), Support Vector Regression (SVR), Extreme Gradient Boosting (XGBoost), and random forest regression (RFR). The results show that for Larix gmelinii, the Landsat 8 bands TM4 and TM7 have a greater degree of correlation with the AGB than the other features, while for Betula platyphylla, bands TM3 and TM4 show a greater degree of correlation with the AGB, and elevation has a weaker correlation with the AGB. Although the linear regression (LR) demonstrates certain advantages for AGB estimation, particularly when the AGB values range from 40 to 70 t/ha, the RFR outperforms in overall performance, with estimation accuracies reaching 85% for Betula platyphylla and 89% for Larix gmelinii. This study reveals that both the species and environmental characteristics may significantly influence the selection of the remote sensing features for AGB estimation, and the choice of algorithm for model optimization is critical. This study innovatively extracts the features related to the dominant species in temperate forests, analyses their relationships with environmental factors, and optimizes the AGB estimation model using advanced regression techniques, offering a method that can be applied to other forest regions as well.

Hemerocallis fulva, essential to urban ecosystems and landscape design, faces challenges in disease detection due to limited data and reduced accuracy in complex backgrounds. To address these issues, the Hemerocallis fulva leaf disease dataset (HFLD-Dataset) is introduced, alongside the Hemerocallis fulva Multi-Scale and Enhanced Network (HF-MSENet), an efficient model designed to improve multi-scale disease detection accuracy and reduce misdetections. The Channel–Spatial Multi-Scale Module (CSMSM) enhances the localization and capture of critical features, overcoming limitations in multi-scale feature extraction caused by inadequate attention to disease characteristics. The C3_EMSCP module improves multi-scale feature fusion by combining multi-scale convolutional kernels and group convolution, increasing fusion adaptability and interaction across scales. To address interpolation errors and boundary blurring in upsampling, the DySample module adapts sampling positions using a dynamic offset learning mechanism. This, combined with pixel reordering and grid sampling techniques, reduces interpolation errors and preserves edge details. Experimental results show that HF-MSENet achieves mAP@50 and mAP%50–95 scores of 94.9% and 80.3%, respectively, outperforming the baseline model by 1.8% and 6.5%. Compared to other models, HF-MSENet demonstrates significant advantages in efficiency and robustness, offering reliable support for precise disease detection in Hemerocallis fulva.

In object detection, boundary blurring caused by occlusion and background interference often hinders effective feature extraction. To address this challenge, we propose Edge Aware-YOLO, a novel framework designed to enhance edge awareness and efficient feature fusion. Our method integrates three key contributions. First, the Variable Sobel Compact Inverted Block (VSCIB) employs convolution kernels with adjustable orientation and size, enabling robust multi-scale edge adaptation. Second, the Spatial Pyramid Shared Convolution (SPSC) replaces standard pooling with shared dilated convolutions, minimizing detail loss during feature reconstruction. Finally, the Efficient Downsampling Convolution (EDC) utilizes a dual-branch architecture to balance channel compression with semantic preservation. Extensive evaluations on public datasets demonstrate that Edge Aware-YOLO significantly outperforms state-of-the-art models. On MS COCO, it achieves 56.3% mAP₅₀ and 40.5% mAP_50–95 (gains of 1.5% and 1.0%) with only 2.4M parameters and 5.8 GFLOPs, surpassing advanced models like YOLOv11.

Guaranteeing the ecological security of the Dongting Lake Basin is of paramount importance for national-scale programs, such as the Yangtze River Economic Belt and aquatic conservation projects. Within this framework, carbon storage and its determining drivers act as essential indicators of regional ecological stability. However, the historical trajectory of carbon pools and their response to future multi-scenario land-use transitions remain insufficiently understood. Therefore, this study aims to quantify the spatiotemporal evolution of carbon storage in the Dongting Lake Basin from 2000 to 2020 and project its future dynamics under diverse development pathways. This study, utilizing land use data from 2000 to 2020 and the carbon density database of the Dongting Lake Basin, assessed land use changes over two decades and determined the spatiotemporal distribution of carbon storage. Additionally, using 17 driving factors and various spatial policies, the study projected the land use and land cover changes (LUCC) for 2030 under four scenarios: natural development, ecological protection, economic development, and planned development. The spatiotemporal distribution of carbon storage and its response mechanisms were analyzed for each scenario. The results showed that carbon storage was directly impacted by LUCC, with an overall “decrease-increase-decrease” trend from 2000 to 2020, resulting in a net increase of 3.685 × 10⁶ t. By 2030, the changes in carbon storage under the natural development, ecological protection scenario, economic development, and planned development scenarios were projected to be −1.008 × 10⁷ t, 1.276 × 10⁷ t, 3.292 × 10⁸ t, and −1.200 × 10⁵ t, respectively. Notably, the ecological protection scenario showed a significant positive growth in carbon storage, primarily driven by an increase in forest and wetland areas. Additionally, the spatial distribution of carbon storage exhibited a pattern of “high in the west and low in the east”. These results imply that to achieve the “Dual Carbon Strategy”, future land use planning in the Dongting Lake Basin should prioritize ecological protection and planned development models, including strict control of construction land expansion, increasing ecological land area, and enhancing carbon storage.

The carbon sink capacity of forest ecosystem and its function of mitigating climate change have been confirmed. As two common coniferous species, Cunninghamia lanceolata (Lamb.) Hook. (C. lanceolata) and Pinus elliottii Engelmann (P. elliottii) are widely planted in southern China, and their carbon sink capacity has always been concerning. According to their diameter class, we randomly harvested 42 C. lanceolata trees and 38 P. elliottii trees from our entire study area, measured their carbon concentration, and constructed biomass models with DBH and tree height as variables. The biomass of the tree layer was estimated by measuring the DBH of all trees in the plots, and the biomass and carbon concentration of shrubs, herbs, dead wood and litter in the plot were measured by harvesting them. The results showed that the total biomass in C. lanceolata and P. elliottii plantations were 117.1 and 151.8 t·ha⁻¹; the biomass in the tree layer was 94.7 and 122.9 t·ha⁻¹; and in the other parts was 22.4 and 28.9 t·ha⁻¹, respectively. In addition, the total carbon densities in the C. lanceolata and P. elliottii plantation ecosystems were 166.3 and 198.6 t·ha⁻¹; the carbon densities in the soil were 108.1 and 124.6 t·ha⁻¹; and in the other parts, they were 58.2 and 74.0 t·ha⁻¹, respectively. These results indicate that there are significant differences in total biomass or total carbon storage between the two coniferous forest ecosystems, and net productivity and carbon sink capacity are higher in the P. elliottii plantation ecosystem. This study lays the foundation for the biomass estimation and carbon trading of these two coniferous forests in northern Hunan.