This session highlights innovative applications of satellite, aerial, and lidar remote sensing to monitor ecosystems, agriculture, planetary surfaces, and conservation landscapes. Presentations demonstrate how multi-sensor imagery, machine learning, and geospatial analytics are being used to map invasive species, improve crop classification, support planetary mapping, and establish environmental baselines for long-term monitoring and management.
10:30 AM – 10:45 AM
Mapping the Spread of the Goldspotted Oak Borer Using Remote Sensing
Presented by Zhuba Goldenlamb, California State University Long Beach
This project explores the spatial relationship between canopy stress and field observations of invasive beetle activity in Marshall Canyon, Southern California. The project integrates 2024 lidar point cloud data from LARIAC and 2024 NAIP multispectral imagery to identify potential forest decline associated with Goldspotted Oak Borer (GSOB) infestation. By combining canopy height and vegetation health measurements, the map helps visualize patterns of canopy stress that may not be immediately visible through field surveys alone.
The analysis compares snag clusters—areas where canopy height exceeds 5 meters and NDVI values fall below 0.30—with observed GSOB indicators such as D-shaped exit holes. Results show that GSOB observations occur significantly closer to canopy stress areas, with a mean distance of 8.86 meters compared to 24.44 meters for randomly generated points. This suggests a strong spatial association between remotely sensed indicators of forest decline and the presence of invasive beetles.
The project was created using ArcGIS Pro through a workflow that combined lidar-derived Canopy Height Models, NDVI raster analysis, aerial imagery, and spatial proximity analysis. Key layers included canopy height, vegetation stress, snag clusters, and mapped GSOB field observations. The map demonstrates how remote sensing and GIS can support scalable environmental monitoring and invasive species detection across forested landscapes.
10:45 AM – 11:00 AM
An Integrated Sentinel-2 and NASA Environmental Data Framework for Crop Classification in the Navajo Nation
Presented by Varatharajaperumal Thangavel, Florida Atlantic University
Crop classification is important for monitoring agricultural land, evaluating seasonal crop conditions, and supporting water and resource planning. In the Navajo Nation, crop mapping is challenging because agricultural fields are spatially dispersed, field reference data are limited, and crop development is strongly affected by drought and variations in the availability and quality of irrigation water. Sentinel-2 imagery provides detailed information on crop reflectance and seasonal vegetation development; however, optical imagery alone does not directly represent the precipitation, soil-moisture, evapotranspiration, and temperature conditions that influence crop growth. The objective of this study is to develop an integrated crop-classification framework using multi-temporal Sentinel-2 imagery and NASA environmental datasets to improve crop mapping and agricultural monitoring in the Navajo Nation.
The study uses satellite observations from the 2017–2024 growing seasons. Sentinel-2 surface-reflectance imagery acquired from April through November is processed into monthly median composites. Visible, red-edge, near-infrared, and shortwave-infrared bands are used together with the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Normalized Difference Red Edge Index (NDRE), Green Chlorophyll Vegetation Index (GCVI), and Land Surface Water Index (LSWI). Environmental variables include precipitation from the NASA Global Precipitation Measurement mission, root-zone soil moisture from the Soil Moisture Active Passive mission, evapotranspiration from the MOD16A2GF product, and daytime and nighttime land-surface temperature from the MOD11A2 product. These environmental variables will be summarized for the same monthly periods as the Sentinel-2 observations to represent growing-season water and temperature conditions.
Annual United States Department of Agriculture (USDA) Cropland Data Layer (CDL) maps are used to prepare crop-reference samples for the corresponding years. The presentation will discuss the Random Forest models developed using two main predictor configurations: (i) Sentinel-2 spectral and seasonal variables alone, and (ii) combined Sentinel-2 and NASA environmental variables. The same crop classes, reference samples, model parameters, and validation locations will be maintained for both models. The presentation will discuss spatial block cross-validation, which is used to reduce the influence of nearby training and validation samples. Data from 2017-2023 are used for model development, with 2024 as an independent temporal test year.
The presentation will compare model performance using overall accuracy, balanced accuracy, macro-F1 score, class-specific precision, recall, F1 score, and confusion matrices. In addition, the performance of Sentinel-2-only and integrated models will be compared, the classification performance across different growing seasons will be examined, and the crop classes that benefit from the inclusion of environmental information will be identified. The study will produce annual crop-classification maps, crop-level accuracy assessments, model-comparison results, and an evaluation of model performance across years. The presentation will provide a structured method for combining optical and environmental satellite observations for agricultural monitoring in the Navajo Nation.
11:00 AM – 11:15 AM
Integration of MOLA-Based DEMs and HiRISE Images for Large-Scale Mars Mapping
Presented by Ahmed Elaksher, NMSU
Precise surface mapping is fundamental for Martian geomorphological studies, surface change analysis, and selecting safe landing zones. However, accurately aligning high-resolution (sub-meter) HiRISE imagery with the broader MOLA geodetic coordinate system is challenging due to the significant disparity in scale and sensor technology, which precludes direct, dense registration. Furthermore, factors such as inconsistent illumination, varying surface textures (e.g., shifting dunes or slope streaks), differing radiometric properties, and specific cartographic standards (such as longitude direction and reference surfaces) necessitate a reliable and repeatable method for integrating these datasets.
In this study, we propose a workflow that utilizes the Context Camera (CTX) as an intermediary to bridge HiRISE imagery to the MOLA frame, validated across 29 image pairs representing five distinct geological environments. By employing SIFT features and RANSAC for robust matching, we connect HiRISE control points to MOLA data via a nearest-neighbor search. We evaluated six different 2D and 2D–3D transformation models to georeference the images. Our results indicate that the Direct Linear Transformation (DLT) provides the highest accuracy, achieving a Root Mean Square Error (RMSE) of approximately 2–3 meters, with optimal performance reached using roughly 10 well-distributed ground control points. Simpler models, such as parallel and extended parallel projections, yielded RMSEs in the 7–10 meter range.
It is important to note that these error metrics are calculated based on MOLA ground-track consistency within the Mars 2000 reference frame, rather than an independent validation. While performance is influenced by the terrain—with tectonic and mineralogical landscapes yielding the best results compared to more difficult aeolian or mass-wasting areas—we conclude that the DLT model combined with at least 10 well-distributed control points is the most effective approach for practical applications.
11:15 AM – 11:30 AM
Multi-Sensor Satellite Remote Sensing for Holistic Biome Characterization: Supporting Conservation and Eradication Efforts on Aldabra Atoll
Presented by Kyle Goodrich, TCarta
Satellite remote sensing offers unparalleled capabilities for monitoring logistically challenging, ecologically vital ecosystems. The Aldabra Atoll, Seychelles requires comprehensive baseline datasets to support active conservation and management amid climate variability. This study presents a dual-framework assessment by TCarta that integrates Synthetic Aperture Radar (SAR) and high-resolution multispectral optical satellite imagery. By capturing both terrestrial and marine biomes simultaneously, this multi-sensor approach provides a holistic understanding of Aldabra’s interconnected environments. Specifically, this work addresses two urgent conservation needs: mapping dynamic inland pools to support invasive rat eradication efforts, and delivering an updated benthic habitat map for a protected lagoon that has gone unmapped for decades despite its designation as a UNESCO heritage site.
To assess the terrestrial biome, high-resolution X-band Capella SAR Spotlight imagery was utilized across the 2025 wet and dry seasons. Because radar penetrates cloud cover and is highly sensitive to surface texture, it allows for the precise extraction of small, isolated inland pools using an AI-based extraction method. This mapping provides critical spatial intelligence for terrestrial eradication campaigns where targeted monitoring relies on identifying seasonal water access.
Concurrently, the marine biome was evaluated using Maxar WorldView-2 multispectral optical imagery. Pre-processing involved glint correction and Gram-Schmidt pansharpening to achieve a 0.5-meter resolution while preserving the true radiometric quality through the water. Advanced feature extraction—including natural-log ratio band permutations, colorspace conversions to decouple depth from illumination, and Satellite-Derived Bathymetry (SDB)—informed supervised machine learning models (Histogram Gradient Boosting) to classify complex seafloor habitats.
The integration of multi-sensor satellite data successfully bridged the gap between Aldabra’s land and sea habitats. The SAR analysis revealed severe seasonal hydrological contractions across the atoll, mapping a dramatic reduction in both the abundance and surface area of pools during the dry season. Documenting these drying trends provides crucial insights into wildlife concentration points and freshwater availability vital for successful eradication logistics.
In the marine environment, the machine learning pipeline successfully categorized the expansive lagoon into seven distinct benthic zones with high overall accuracy. Macroalgae and seagrass assemblages on sand, dense macroalgal mats, and bare carbonate sand were identified as the dominant lagoon features, providing the first highly accurate structural breakdown of this protected marine environment in over fifteen years.
This project underscores the transformative impact of combining radar and optical satellite technologies for island conservation. Moving beyond legacy methodologies, the use of advanced machine learning and multi-temporal sensor pipelines overcomes the historical challenges of cloud cover, water depth, and structural complexity. The resulting datasets establish a standardized, repeatable baseline designed for direct change detection. By empowering managers with actionable insights into both terrestrial hydrology and marine habitats, satellite remote sensing serves as a cornerstone for safeguarding Aldabra’s unique biomes.
