This paper aims to reveal the effects of multi-generational succession of eucalyptus on soil fertility, organic structure and biological properties. Soil samples were collected from eucalyptus plantations of different stand ages (5, 11, 17 and 21 years old) in a typical area in south Asia, soil organic fraction structure and content characteristics were investigated using Fourier transform infrared (FTIR), and structural equation modelling (SEM) was used to explore influences of soil fertility, enzyme activity and organic fraction on stand biomass. FTIR analysis showed that 11 infrared absorption peaks existed in the soils of this study area, attributed to silicates, aromatics, carbonate ions, sugars, esters, polysaccharides, aliphatic hydrocarbons and phenolic alcohols. Combined with the results of peak area integration, the content of esters, aromatics and phenolic alcohols was significantly higher in 17- and 21-year-old stand soils than in control soils. The results of SEM showed that organic components were negatively related ( p < 0.01) to enzyme activity and biomass, with standardized coefficients of 0.53 and 0.49, respectively. In summary, multi-generation succession of eucalyptus trees can change the structure of soil organic functional group composition and promote the enrichment of aromatic and phenolic alcohol functional groups. Such changes can directly inhibit the increase in eucalyptus biomass and indirectly negatively affect biomass by inhibiting enzyme activity.

Soil Organic Matter (SOM) is crucial for soil fertility, and effective detection methods are of great significance for the development of agriculture and forestry. This study uses 206 hyperspectral soil samples from the state-owned Yachang and Huangmian Forest Farms in Guangxi, using the SPXY algorithm to partition the dataset in a 4:1 ratio, to provide an effective spectral data preprocessing method and a novel SOM content prediction model for the study area and similar regions. Three denoising methods (no denoising, Savitzky–Golay filter denoising, and discrete wavelet transform denoising) were combined with nine mathematical transformations (original spectral reflectance (R), first-order differential (1DR), second-order differential (2DR), MSC, SNV, logR, (logR)′, 1/R, ((1/R)′) to form 27 combinations. Through Pearson heatmap analysis and modeling accuracy comparison, the SG-1DR preprocessing combination was found to effectively highlight spectral data features. A CNN-SVM model based on the Black Kite Algorithm (BKA) is proposed. This model leverages the powerful parameter tuning capabilities of BKA, uses CNN for feature extraction, and uses SVM for classification and regression, further improving the accuracy of SOM prediction. The model results are RMSE = 3.042, R² = 0.93, MAE = 4.601, MARE = 0.1, MBE = 0.89, and PRIQ = 1.436.

Pest control in economic forests is a crucial aspect of sustainable forest resource management, yet it faces bottlenecks such as low efficiency and high miss rates for small objects. Based on the RT-DETR model, this paper proposes LightFAD-DETR, a lightweight architecture integrated with feature aggregation diffusion, designed for complex economic forest scenarios. Firstly, by employing the YOLOv9 lightweight backbone network to compress the computational base load, we introduce the RepNCSPELAN4-CAA module, which integrates re-parameterization techniques and one-dimensional strip convolution. This enhances the model’s ability for cross-regional modeling of slender insect morphologies. Secondly, a feature aggregation diffusion network is designed, incorporating a dimension-aware selective integration mechanism. This dynamically fuses shallow detail features with deep semantic features, effectively mitigating information loss for small objects occluded by foliage. Finally, a re-parameterized batch normalization technique is introduced to reconstruct the AIFI module. Combined with a progressive training strategy, this eliminates redundant parameters, thereby enhancing inference efficiency on edge devices. Experimental validation demonstrates that compared to the baseline RT-DETR model, LightFAD-DETR achieves a 1.4% improvement in ${mAP}_{0.5:0.95}$, while reducing parameters by 41.7% and computational load by 35.0%. With an inference speed reaching 106.3 FPS, the method achieves balanced improvements in both accuracy and lightweight design.

High-resolution remote sensing imagery comprises spatial structure features of multispectral bands varying in scale, color, and shape. These heterogeneous geographical features introduce grave challenges to the fine segmentation required for classification applications in remote sensing imagery, where direct application of traditional image classification models fails to deliver optimal results. To overcome these challenges, a multispectral, multi-label model, MMDL-Net, has been developed. This model is integrated with the multi-label BigEarthNet dataset, primarily employed for land cover classification research in remote sensing imagery, with each image composed of 13 spectral bands and spatial resolutions of 10 m, 20 m, and 60 m. To effectively utilize the information across these bands, a multispectral stacking module has been introduced to concatenate this spectral information. To proficiently process three distinct large-scale remote sensing image datasets, a multi-label classification module has been incorporated for training and inference. To better learn and represent the intricate features within the images, a twin-number residual structure has been proposed. The results demonstrate that the MMDL-Net model achieves a top accuracy of 83.52% and an F1 score of 77.97%, surpassing other deep learning models and conventional methods, thereby exhibiting exceptional performance in the task of multispectral multi-label classification of remote sensing imagery.

In forest monitoring networks, the computational capabilities of sensors cannot meet the latency requirements for complex tasks, and the limited battery capacity of these sensors hinders the long-term execution of monitoring tasks. Mobile edge computing (MEC) acts as an effective solution for this issue by offloading tasks to edge servers, significantly reducing both task latency and energy consumption. However, the computational capacity of MEC servers and the bandwidth in the system are limited, and the communication environment in forested areas is complex. To simulate the complexity of the forest communication environment, we incorporate empirical path loss and multipath fading into the calculation of signal transmission rates. The computational offloading problem is then converted into a minimum-cost optimization problem with multiple constraints related to energy consumption and latency, which we formulate as an NP-hard problem. We propose a dung beetle optimization (DBO) strategy for computational offloading, enhancing it with an improved circle chaotic mapping, a dimension decomposition strategy, and Cauchy disturbance. This algorithm has the beauty of symmetry in the search range, and the symmetrical features can comprehensively search for existing solutions. Experimental results demonstrate that the improved dung beetle optimization algorithm (IDBO) achieves better convergence, lower complexity, and superior optimization outcomes compared to local offloading strategies and other metaheuristic algorithms, confirming the effectiveness of the proposed algorithm and ensuring the service quality of the forest monitoring network.

Accurate and rapid acquisition of soil texture information is crucial to evaluating soil quality, formulating soil and water conservation strategies, and guiding agricultural resource management. Compared with traditional machine learning methods, convolutional neural networks (CNNs) demonstrate superior accuracy in soil texture prediction. To overcome the limitations of existing lightweight models in spectral modeling, such as insufficient single-scale feature representation, limited channel utilization, and branch redundancy, and to meet the demand for lightweight architectures, we propose a novel dynamic feature modeling approach: Multi-scale Routing Attention Network (MSRA-Net). MSRA-Net integrates grouped multi-scale convolutions with an intra-group Efficient Channel Attention (gECA) mechanism, combined with a multi-scale weighting strategy based on a Branch Routing Attention (BRA) mechanism, thereby enhancing inter-channel feature interaction and improving the model’s ability to capture complex spectral patterns. Furthermore, we introduce a multi-task learning variant, MSRA-MT, which employs uncertainty dynamic weighting to balance gradients magnitude across tasks, thereby improving both stability and predictive accuracy. Experimental results on the LUCAS and ICRAF datasets demonstrate that the MSRA-MT model consistently outperforms baseline models in terms of performance and robustness ($RMS{E}_{mean}$ = 9.190 and $RMS{E}_{mean}$ = 8.189 for ICRAF and LUCAS, respectively). Prior knowledge-based soft constraints may hinder optimization by amplifying intrinsic noise, rather than improving learning effectiveness.