Forest fires pose a severe threat to ecological environments and the safety of human lives and property, making real-time forest fire monitoring crucial. This study addresses challenges in forest fire image object detection, including small fire targets, sparse smoke, and difficulties in feature extraction, by proposing TFNet, a Transformer-based multi-scale feature fusion detection network. TFNet integrates several components: SRModule, CG-MSFF Encoder, Decoder and Head, and WIOU Loss. The SRModule employs a multi-branch structure to learn diverse feature representations of forest fire images, utilizing 1 × 1 convolutions to generate redundant feature maps and enhance feature diversity. The CG-MSFF Encoder introduces a context-guided attention mechanism combined with adaptive feature fusion (AFF), enabling effective multi-scale feature fusion by reweighting features across layers and extracting both local and global representations. The Decoder and Head refine the output by iteratively optimizing target queries using self- and cross-attention, improving detection accuracy. Additionally, the WIOU Loss assigns varying weights to the IoU metric for predicted versus ground truth boxes, thereby balancing positive and negative samples and improving localization accuracy. Experimental results on two publicly available datasets, D-Fire and M4SFWD, demonstrate that TFNet outperforms comparative models in terms of precision, recall, F1-Score, mAP50, and mAP50–95. Specifically, on the D-Fire dataset, TFNet achieved metrics of 81.6% precision, 74.8% recall, an F1-Score of 78.1%, mAP50 of 81.2%, and mAP50–95 of 46.8%. On the M4SFWD dataset, these metrics improved to 86.6% precision, 83.3% recall, an F1-Score of 84.9%, mAP50 of 89.2%, and mAP50–95 of 52.2%. The proposed TFNet offers technical support for developing efficient and practical forest fire monitoring systems.

The cation exchange capacity (CEC) of the clay fraction (<2 μm), denoted as CEC_clay, serves as a crucial indicator for identifying low-activity clay (LAC) soils and is an essential criterion in soil classification. Traditional methods of estimating CEC_clay, such as dividing the whole-soil CEC (CEC_soil) by the clay content, can be problematic due to biases introduced by soil organic matter and different types of clay minerals. To address this issue, we introduced a soil pedotransfer functions (PTFs) approach to predict CEC_clay from CEC_soil using experimental soil data. We conducted a study on 122 pedons in South China, focusing on highly weathered and strongly leached soils. Samples from the B horizon were used, and eight models and PTFs (four machine learning methods, multiple linear regression (MLR) and three PTFs from publication) were evaluated for their predictive performance. Four covariate datasets were combined based on available soil data and environmental variables and various parameters for machine learning techniques including an artificial neural network, a deep belief network, support vector regression and random forest were optimized. The results, based on 10-fold cross-validation, showed that the simple division of CEC_soil by clay content led to significant overestimation of CEC_clay, with a mean error of 14.42 cmol(+) kg⁻¹. MLR produced the most accurate predictions, with an R² of 0.63–0.71 and root mean squared errors (RMSE) of 3.21–3.64 cmol(+) kg⁻¹. The incorporation of environmental variables improved the accuracy by 2–10%. A linear model was fitted to enhance the current calculation method, resulting in the equation: CEC_clay = 15.31 + 15.90 × (CEC_soil/Clay), with an R² of 0.41 and RMSE of 4.48 cmol(+) kg⁻¹. Therefore, given limited soil data, the MLR PTFs with explicit equations were recommended for predicting the CEC_clay of B horizons in humid subtropical regions.

Accurate runoff simulation is essential for water resources planning and development projects. At present, the commonly employed runoff simulation approaches are categorized into two types: process- and data-driven models. Process-driven models pertain to the enhancement of the structural framework in conceptual rainfall–runoff models using hydrological principles to estimate runoff but have low accuracy at the monthly scale. Unlike the process-driven models, data-driven models (DDMs) can simulate the relationship between input factors and output runoff data without regard to complex and unknown runoff production and acquire satisfactory simulation results. Here, we comparatively investigate the applicability of DDMs, including traditional shallow DDMs, deep learning-based (DL) models for monthly runoff simulation, and select the Autoregressive (AR) model as the baseline model for comparison. Moreover, four evaluation indicators, including mean absolute percentage error (MAPE), root mean squared error (RMSE), Nash–Sutcliffe efficiency (NSE), and coefficient of determination (R²), are employed to evaluate the runoff simulation effects of the above methods. We systematically compare the AR model, and shallow and deep learning-based DDMs for runoff simulation at four hydrological stations in the Yalong River basin (YRB), respectively. The finding results reveal that the DDMs demand limited data and can offer satisfactory prediction effects. Also, the DL models outperform other shallow DDMs and the AR model in terms of the above evaluation criteria.

Recent advances in underwater image enhancement (UIE) have achieved notable progress using deep learning techniques; however, existing methods often struggle with limited receptive fields, inadequate frequency modeling, and poor structural perception, leading to sub-optimal visual quality and weak generalization in complex underwater environments. To tackle these issues, we propose FACMamba, a Mamba-based framework augmented with frequency-aware mechanisms, enabling efficient modeling of long-range spatial relations for underwater image restoration. Specifically, FACMamba incorporates three key components: a Multi-Directional Vision State-Space Module (MVSM) to model directional spatial context via the proposed 8-direction selective scan block (SS8D), a Frequency-Aware Guidance Module (FAGM) for learning informative frequency representations with low overhead, and a Structure-Aware Fusion Module (SAFM) to preserve fine-grained structural cues through adaptive multi-scale integration. Recognizing the importance of spatial-frequency interaction, our model fuses these representations via lightweight architecture to enhance both texture and color fidelity. Experiments on standard UIE benchmarks demonstrate that FACMamba achieves a favorable balance between enhancement quality and computational efficiency, outperforming many existing UIE methods.

Point-of-interest (POI) recommendation is highly sensitive to temporal factors, including fluctuations in user preferences, variation in user similarity, and decay in the attraction of locations. However, current studies overlook the temporal dynamics of user similarity and the timeliness of POIs, resulting in a disconnect between recommendations and users’ recent preferences. This paper proposes a new framework for dynamic and timely POI recommendation by integrating spatio-temporal influences and social relationships. Dynamic prediction is achieved through an enhanced user-based collaborative filtering approach. A time slot clustering technique was designed based on the statistical check-in features in each time slot. Ratings within the same cluster were shared to address data sparsity. To reflect user similarity drift, we took the time variable as a crucial parameter to dynamically calculate user similarity. Moreover, timely prediction was achieved by integrating the timeliness, popularity, and spatial features of POIs. We introduce a novel method to evaluate the timeliness of POI recommendation, aimed at assessing whether the recommendations align with users’ recent preferences. Comprehensive experiments are performed on Brightkite and Gowalla datasets, with the data divided into workdays and weekends. The experimental results reveal that our algorithm outperforms seven state-of-the-art recommenders in terms of prediction accuracy and system timeliness.

Object detection in UAV aerial imagery plays a pivotal role across a wide spectrum of applications. However, existing detection models continue to face significant challenges stemming from small object scales, dense spatial distributions, and highly complex backgrounds. To address these challenges, this paper proposes a novel dual-backbone network model named WCDB-YOLO. The core innovation of this work lies in introducing a “target-context decoupled perception” paradigm, which utilizes two structurally complementary backbone networks to separately process local object features and global background information: one backbone focuses on extracting fine-grained local features of objects, while the other innovatively incorporates a wavelet convolution module to efficiently model the global contextual semantics of complex scenes with minimal computational cost by constructing a large receptive field. To further enhance the scale adaptability for small objects, a Dilation-wise Residual (DWR) module is designed, which employs parallel convolutional branches with different dilation rates to achieve dynamic adaptation to multi-scale small object features. Additionally, the model optimizes the feature pyramid structure by integrating high-resolution P2/4 features into the detection head, significantly improving the localization accuracy of tiny objects. Experimental results on the VisDrone dataset show that the proposed method achieves an 8.4% improvement in mAP50 over the baseline YOLOv11s model and outperforms current state-of-the-art (SOTA) approaches. This work presents a highly accurate and robust solution for small object detection from UAV platforms in complex environments.