25 Sep 2025

Expansion of a Landform Reference Dataset in the Chihuahuan Desert for Dust Source Characterization Applications

Abstract: This report details the development of an extensive landform reference dataset for the Chihuahuan Desert region to support validation of a machine-learning-based landform classification model. Building upon previous work by Cook et al. (2022), we expanded both the quantity and spatial coverage of reference points to better represent the study domain’s geomorphic diversity. Analysts integrated information from published literature, government databases, and satellite imagery interpretation to create a dataset of 236,582 points across 12 landform classes, aligned to a 500 m resolution grid. The bedrock/pediment/plateau class was the dominant class (58%), followed by alluvial fans (21%), aeolian sands (11%), and aeolian dunes (5%). Approximately 85% of the reference points received high analyst confidence ratings, and ratings were especially high for classes with distinctive signatures, such as bedrock features, fine-grained lake deposits, urban/developed areas, water, and agricultural lands. Classification challenges consistently emerged in transitional zones between land-forms, areas with anthropogenic modifications, and complex landform assemblages where mapping resolution proved insufficient. The resulting dataset is a valuable resource for model validation and offers insights into arid region geomorphology. Additionally, it has the potential to support multiple applications, including dust hazard forecasting, terrain mobility assessment, soil property inference, and rangeland management.

24 Sep 2025

Simulating Environmental Conditions for a Severe Dust Storm in Southwest Asia Using the Weather Research and Forecasting Model: A Model Configuration Sensitivity Study

Abstract: Dust aerosols create hazardous air quality conditions that affect human health, visibility, and military operations. Numerical weather prediction models are important tools for predicting atmospheric dust by simulating dust emission, transport, and chemical evolution. We assessed the Weather Research and Forecasting (WRF) model’s ability to simulate the atmospheric conditions that drove a major dust event in Southwest Asia during July–August 2018. We evaluated five WRF configurations against satellite observations and Reanalysis Version 5 (ERA5) reanalysis data, focusing on the event’s synoptic evolution, storm progression, vertical structure, and surface wind fields. Results revealed substantial differences between configurations using Noah and Noah Multiparameterization (Noah-MP) land surface models (LSMs), with Noah providing a superior representation of meteorological conditions despite theoretical expectations of similar performance in arid environments. The best-performing configuration (Noah LSM, Mellor–Yamada–Nakanishi–Niino planetary boundary layer scheme, and spectral nudging) of the five considered accurately simulated the progression of a low-level jet streak and the associated surface winds responsible for dust mobilization throughout the event. This study supports the US Army Engineer Research and Development Center’s efforts to improve dust forecasting and establishes a foundation for evaluating dust emission parameterizations by isolating meteorological forcing errors from dust model physics.

23 Sep 2025

Using the Robot Operating System for Uncrewed Surface Vehicle Navigation to Avoid Beaching

Abstract: Our research explores the use of the Robotic Operating System (ROS) to autonomously navigate an uncrewed surface vehicle (USV). As a proof of concept, we set up a simulated world and spawned a virtual Wave Adaptive Modular Vehicle (WAM-V). We used the robot_localization package to localize the WAM-V in the virtual world and used move_base for the navigation of waypoints. The move_base package used both costmaps and path planners to reach its intended goal while simultaneously avoiding sub-merged shallow-water obstacles. Shallow-water obstacles are obstacles at a depth that is less than a user-defined value (1 meter in this case). Finally, we investigated using vizanti as a mission planner. This report provides a detailed explanation of the parameters that were modified to demonstrate a successful proof of concept.

17 Sep 2025

Validating Predicted Soil Boundaries with In Situ Collections

Abstract: This US Army Engineer Research and Development Center (ERDC) technical note describes the process used by the Intelligent Environmental Battlefield Awareness (IEBA) team to validate the spatial distribution and texture class attribution of soil boundary predictions. The predicted global soil boundary polygons will serve as a primary base layer for populating other environmental variables; thus, it is essential to assess their robustness prior to the attribution stage.

3 Sep 2025

Exploring Burnt Area Delineation with Cross-Resolution Mapping: A Case Study of Very High and Medium-Resolution Data

Abstract: Remote sensing is essential for mapping and monitoring burnt areas. Integrating Very High-Resolution (VHR) data with medium-resolution datasets like Landsat and deep learning algorithms can enhance mapping accuracy. This study employs two deep learning algorithms, UNET and Gated Recurrent Unit (GRU), to classify burnt areas in the Bandipur Forest, Karnataka, India. We explore using VHR imagery with limited samples to train models on Landsat imagery for burnt area delineation. Four models were analyzed:(a) custom UNET with Landsat labels, (b) custom UNET with PlanetScope-labeled data on Landsat, (c) custom UNET-GRU with Landsat labels, and (d) custom UNET-GRU with PlanetScope-labeled data on Landsat. Custom UNET with Landsat labels achieved the best performance, excelling in precision (0.89), accuracy (0.98), and segmentation quality (Mean IOU: 0.65, Dice Coefficient: 0.78). Using PlanetScope labels resulted in slightly lower performance, but its high recall (0.87 for UNET-GRU) demonstrating its potential for identifying positive instances. In the study, we highlight the potential and limitations of integrating VHR with medium-resolution satellite data for burnt area delineation using deep learning.

18 Aug 2025

Bare Ground Classification Using a Spectral Index Ensemble and Machine Learning Models Optimized Across 12 International Study Sites

Abstract: This research investigates a global approach to map bare ground across diverse geographies with an ensemble of spectral indices using optimal thresholds identified in testing to train and evaluate machine learning models to extract bare ground pixels from Sentinel-2 imagery. Twelve locations in four Köppen climate zones with data from two seasons were evaluated. Accuracy assessment showed a mean F1 score of 80% and a mean Overall Accuracy (OA) of 81% for random forest and an F1 score of 78% and OA of 79% for support vector machine. Higher accuracies were observed in climate region-based models with mean F1 = 84% in three of four climate zones. Low accuracies occurred in winter imagery with leaf-off tree cover or building materials similar to bare ground. This framework provides a global approach to map bare ground without need for high-density time-series or deep learning models and moves beyond locally effective methods.

14 Aug 2025

Creating an Augmented Soil Texture Master List Using the Gridded Soil Survey Geographic Database (gSSURGO)

Purpose: This US Army Engineer Research and Development Center (ERDC) technical note (TN) describes the workflow for creating an augmented soil texture master list that describes the surface-most (i.e., uppermost) USDA soil texture class and coarse fragment modifier. In conjunction with a soil similarity search algorithm, the soil texture master list fulfills a need identified by the Intelligent Environmental Battlefield Awareness (IEBA) project to generate detailed global soil boundary polygons. These polygons will serve as the base layer for populating other environmental variables, like soil temperature, soil moisture, depth to permafrost, and vegetation type, in the battlespace. This TN describes the purpose of the augmented soil texture master list, provides an overview of the gridded Soil Survey Geographic Database (gSSURGO), and describes the methodology used to create the soil texture master list.

13 Aug 2025

A Revised Landform Map for Areas Prone to Dust Emission in the Southwestern United States

Abstract: An area’s landform composition can provide insight into its dust emission potential. In 2017, geomorphologists from the Desert Research Institute provided the US Army Engineer Research and Development Center with a 32-class landform map for portions of the Mojave and Sonoran Deserts in the southwest United States (SWUS) to support air quality and dust hazard modeling applications. We collaborated with the University of California to independently assess the map. Our review identified opportunities to improve the dataset, such as using a simpler landform classification system and revising individual geomorphic unit assignments to ensure consistent labeling across the study area. This report describes our approaches for refining the SWUS map and documents the updated 15-class landform map that resulted from our efforts.

9 Jun 2025

KANICE: Kolmogorov-Arnold Networks with Interactive Convolutional Elements

Abstract: We introduce KANICE, a novel neural architecture that com-bines Convolutional Neural Networks (CNNs) with Kolmogorov-Arnold Network (KAN) principles. KANICE integrates Interactive Convolutional Blocks (ICBs) and KAN linear layers into a CNN framework. This leverages KANs’ universal approximation capabilities and ICBs’ adaptive feature learning. KANICE captures complex, non-linear data relationships while enabling dynamic, context-dependent feature extraction based on the Kolmogorov-Arnold representation theorem. We evaluated KANICE on four datasets: MNIST, Fashion-MNIST, EMNIST, and SVHN, comparing it against standard CNNs, CNN-KAN hybrids, and ICB variants. KANICE consistently outperformed baseline models, achieving 99.35% accuracy on MNIST and 90.05% on the SVHN dataset. Furthermore, we introduce KANICE-mini, a compact variant designed for efficiency. A comprehensive ablation study demonstrates that KANICE-mini achieves comparable performance to KANICE with significantly fewer parameters. KANICE-mini reached 90.00% accuracy on SVHN with 2,337,828 parameters, compared to KAN-ICE’s 25,432,000. This study highlights the potential of KAN-based architectures in balancing performance and computational efficiency in image classification tasks. Our work contributes to research in adaptive neural networks, integrates mathematical theorems into deep learning architectures, and explores the trade-offs between model complexity and performance, advancing computer vision and pattern recognition. The source code for this paper is publicly accessible through our GitHub repository (https://github.com/m-ferdaus/kanice).

14 May 2025

Autonomous Robotics Development in Robot Operating System (ROS) 2 Humble

Abstract: This report presents a novel Robot Operating System (ROS) 2–based simulation framework designed to facilitate the development and testing of an autonomous navigation stack. Elements of the navigation stack, including lidar odometry, simultaneous localization and mapping (SLAM), and frontier exploration, are discussed in detail. The key features of the navigation stack include real-time performance and scalable architecture. The simulation results were applied to a physical robot. As a result, the physical robot was able to autonomously map the interior of a building and to generate 2D occupancy and 3D point clouds of the environment.