The combination of unoccupied aerial vehicles and deep learning offers a promising approach for mapping invasive alien plant species (IAPS), though its effectiveness in early detection case remains uncertain. In this study, we evaluated the suitability of this approach based on a convolutional neural network for mapping the location of common reed (Phragmites australis subsp. australis) within Parc national des Îles-de-Boucherville located in southern Québec, Canada. We collected data on six distinct dates (July-October 2022) during the growing season, covering environments with different levels of reed invasion (Dense, Establishing and Post-treatment). Overall, model performance was high for the different dates and zones, especially for recall (mean of 0.89). The results showed an increase in performance, reaching a peak following the appearance of the inflorescence in September (highest F1-score at 0.98). Despite challenges associated with common reed mapping in a post-treatment monitoring context with poorer detection, this approach has the potential to serve as an effective tool for speeding up the work of biologists in the field and ensuring better management of IAPS. To this end, we provide a comprehensive dataset of high-resolution imagery and deep learning models enabling the detection of common reed across its phenological cycle.

Landfast ice polynyas are important features near many northern coastal communities, and their automated detection from synthetic aperture radar (SAR) imagery is positioned to support on-ice travel safety under changing Arctic sea ice conditions. This research leveraged original datasets of over 5,000 Sentinel-1 SAR observations of wintertime polynyas mapped near the Canadian communities of Sanikiluaq and Nain to investigate deep learning-based landfast ice polynya detection. The Faster-RCNN object detection network was optimized for polynya detection through several modifications to network design elements and training strategies. Resulting detection models generalized well between regions and accurately detected polynyas with local backscatter contrasts above 5 dB, e.g. achieving 90% target recall at 24% precision. Polynyas smaller than 500 meters and with local backscatter contrasts less than 3 dB, constituting approximately half of all observations, were frequently missed. Precision scores below 30% were consistently incurred in attempts to optimize recall. Results highlight challenges to the consistent performance of single-image detectors due to variable and frequently weak polynya signatures in dual-polarized Sentinel-1 SAR backscatter. Future investigations into multi-temporal, multi-frequency, and/or higher-resolution SAR imagery could further support the delivery of robust hazard detection systems relevant to community sea ice safety and monitoring.

Extracting road features from high-resolution remote sensing imagery is crucial for urban planning, navigation systems. However, this task is challenged by factors such as occlusions from buildings, interruptions in road continuity, and variations in road width and appearance. These issues often lead to segmentation discontinuities and misclassifications. This paper presents the Urban Road Extraction Network (UBR-Net), an architecture that enhances the DeepLabv3+ model to address these challenges. The key contributions lie in the specific design and integration of several modules. We introduce Cross-Residual Encoding blocks designed to preserve fine-grained details and mitigate the gradient vanishing problem, thereby improving road continuity. Additionally, UBR-Net incorporates a Multiscale Context Features Extraction (MCFE) module, enhanced with an Improved Self-Attention Block (ISAB), to capture rich, hierarchical feature representations with a focus on long-range dependencies. A Channel Spatial Attention Module (CSAM) is also integrated to refine the feature extraction process by focusing on critical channels and spatial regions. Evaluations on public datasets, including DeepGlobe and Massachusetts, show that UBR-Net reduces extraction errors from occlusions, achieving a 75.18% F1 score and a 57.22% Intersection over Union (IoU), surpassing existing methods. These results highlight UBR-Net’s effectiveness for urban analysis and its potential for more precise and efficient urban planning.

Repeatable methods capable of quantifying Arctic surface water extent at high resolutions are important, but still require development. Here, we present a study using very-high resolution (VHR) X-band Synthetic Aperture Radar (SAR) imagery from Capella Space for fine-scale semantic segmentation of Arctic surface water features. Our study proposes a modified U-Net encoder-decoder model for this task, optimized using the Nadam algorithm. Otsu thresholding was leveraged to rapidly generate 512 × 512-pixel patches for the U-Net, resulting in an efficient and automated training pipeline. Within this study, we also quantitatively compared the deep learning (DL) U-Net to a shallow machine learning (ML) algorithm, XGBoost (XGB), and evaluated the Capella Space imagery against spatially and temporally coincident Sentinel-1 C-band. Performance evaluations showed the U-Net outperforms XGB measured under several statistical metrics, reaching an Intersection over Union (IoU) of 0.955. An explainability analysis was conducted to complement this finding, using Gradient-weighted Class Activation Mapping (Grad-Cam). Visual analysis also underscored the extreme detail of small water features captured by Capella Space imagery, which are at times omitted or lack clarity in conventional Sentinel-1. This research makes several contributions to Arctic surface water mapping, demonstrating the effectiveness of combining VHR SAR imagery with DL.

Monitoring aboveground biomass (AGB) is critical for carbon reporting and quantifying ecosystem change. AGB from field data can be scaled to the region using airborne lidar. However, lidar-based AGB products emphasize upland forests, which may not represent the conditions in rapidly changing peatland complexes in the southern Taiga of western Canada. In addition, to ensure that modeled AGB changes do not incorporate systematic error due to differences between older and newer lidar technologies, model transfer tests are required. The aim of this study was to develop one bi-temporal lidar-based AGB model applicable to (1) vegetation structures at varying vertical and horizontal continuity in this region and to (2) data collected with an earlier generation lidar system for which Canada-wide aerial coverage is available. Goodness-of-fit metrics show that AGB can be modeled with moderate (R2 = 48%–58% Taiga Shield, peatlands) to high accuracies (R2 = 83%–89% Taiga Plains, upland/permafrost plateau forests including ecotones) by using the point clouds average height and 90th height percentile within a weighted approach as function of modeled AGB and calibrating the earlier lidar data. These results are important for quantifying climate change effects on forest to peatland ecotones.

Remote sensing is key for large-scale forest mapping, yet the limited integration of LiDAR sensors restricts the spatial coverage of forest attribute estimation. Our study aimed to accurately map Foliage Height Diversity (FHD) in five key North American regions, vital for ongoing research in ecosystem dynamics in western North America. We used a combination of GEDI data and LVIS data, incorporating measures of forest complexity and relative forest heights (RH25, RH50, RH75, and RH98), to predict FHD with a random forest regression in the Kluane region, southwest Yukon—a northernmost site where GEDI data are not available. This method was designed to overcome spatial coverage limitations of different sensors, enabling the production of consistent, precise, multi-temporal FHD maps across all sites. We found strong agreement between predicted and observed FHD values estimated from Airborne Laser Scanning in the Yukon (R2 = 0.72; RMSE = 0.46). Additionally, we upscaled GEDI FHD predictions in all sites by integrating Landsat imagery, ALOS PALSAR, and topographical data, resulting in high accuracy (R2 = 0.85; RMSE = 0.26). Our findings demonstrate that by harmonizing full-wave form LiDAR sensors, we can significantly expand the coverage of LiDAR data, allowing for consistent broad-scale analyses of forest attributes.

In recent years, numerous deep learning-based methods have gained increasing attention in hyperspectral classification, particularly the Graph Neural Network, which exhibits superior capabilities in structural description. However, a single graph structure is not suitable for hyperspectral feature representation. Therefore, we propose a novel Multiple-Scale graph network structure, known as the Multi-Scale Dense Graph Attention network for hyperspectral classification. Firstly, semi-supervised local Fisher discriminant analysis and superpixel segmentation were employed for dimensionality reduction and multi-scale graph construction, respectively. Secondly, Spectral-Spatial convolution is applied to extract shallow features from the image. Subsequently, an improved graph self-attention network is sequentially applied to each scale graph, and the different scale graphs are densely connected through spatial feature alignment modules, designed using twice matrix multiplication. Finally, the combined pixel-level feature map from multiple graph spaces is derived, and Spectral-Spatial convolution is employed to fuse the abundant feature maps for hyperspectral classification. Experimental results on various hyperspectral datasets demonstrate the superiority of our MSDesGATnet over many state-of-the-art methods. The code is available at https://github.com/l7170/MSDesGAT.git.

In this paper, the central Indian Ocean (60°–95°E, 0°–37°S) has been selected as the research area, and Argo salinity data are used as the measured values. The Catboost algorithm is introduced for the first time to retrieve sea surface salinity, and a comparison is made with the traditional artificial neural network (ANN) and random forest (RF) machine learning algorithm. The results show that: (1) Through linear fitting with the Argo salinity, the R 2 of the three machine learning methods are 0.9299, 0.88 and 0.83, respectively. The corresponding RMSE were 0.2360, 0.3004, and 0.3156 psu, and MAE were 0.1816, 0.2486, and 0.2641 psu, respectively. (2) The spatial distribution of salinity of Argo and SMOS was compared with the inversion results of the model. It was found that the salinity of the sea area was lower at (83°–88°E, 24°–27°S) and (68°–72°E, 17°–20°S), and higher at 30°–35° south latitude, showing consistent with Argo. (3) The stability of the model was independently verified using the data from January to March 2020, and it was found that the R2 of the RF model shows the best stability, while the R 2 of the ANN model shows the worst stability.

Deep-learning-based methods for building-edge-detection have been widely researched and applied in the field of image processing. However, these methods often emphasis the analysis of deep features, which may result in neglecting crucial shallow information representation. Furthermore, abstract features in the deep layers can potentially interfere with the accuracy of edge extraction. To address these challenges, we propose a densely connected edge-detection enhancement network (DCEFE-Net) for building-edge-detection in high-resolution remote sensing images. Firstly, by introducing spatial land channel attention modules, we effectively captured low-level spatial information and high-level semantic information from the input image. Secondly, the proposed edge-aware feature enhancement (EAFE) module emphasis the representation of informative edge features. By alliteratively generating multiple layers of edge-detection maps, it addresses the issue of edge detail loss and enhances edge-detection accuracy. Finally, the dense connectivity blocks strengthen the connections between the convolutional layers, thereby preventing the loss of edge features. Experimental results on the WHU and the Inria Aerial Image Labeling datasets validate the effectiveness of DCEFE-Net, as it consistently produces clear and reliable building-edge results.

Wetlands are considered one of the most valuable ecosystems around the world and provide numerous environmental services, including water purification, flood protection, and habitat for a variety of species. Wetlands loss is an increasing trend due to anthropogenic activities and natural processes. As such, spatial knowledge regarding the extent and dynamics of surface water is demanding for wetland conservation and protection. The Landsat program, with five decades of historical Earth observation data, has a unique advantage for monitoring wetland surface water changes and dynamics with 30 m spatial resolution. We monitored 266 Ramsar wetland sites in North America for the past 40 years using the open-access Landsat data within the Google Earth Engine cloud computing platform. Landsat Collection 2 Level-2 surface reflectance products were preprocessed and cloud-screened, and a time series of spectral bands and indices were created. The unsupervised Dynamic Surface Water Extent method classified each image into water classes with different confidence levels. An average overall agreement of 92,97% and an average F-score of 96.31% were achieved in this study. Water occurrence maps, in addition to inundation class and change maps, were created for the entire North America, and quantified spatial information was calculated for Ramsar wetland sites.