Myles B. Traphagen | Geology Reports

USGS2026

High-resolution transboundary vegetation community maps of the Sonoran and Mojave Desert ecoregion to support critical landscape conservation planning and habitat management needs

We produced a 30-m resolution binational land cover map of Bird Conservation Region 33 (BCR 33) for the U.S. North American Bird Conservation Initiative. The region covers large portions of the Sonoran and Mojave Deserts. The map can support the U.S. Fish and Wildlife Service (FWS) Migratory Bird Program’s recovery planning efforts and constitutes the first known binational land cover dataset spanning sections of the United States–Mexico border and using a consistent classification system for both countries. The mapped region includes 152 distinct land cover classes, covering a total area of 38,421,453 ha (148,345 mi 2 ), of which 13,148,345 ha (52,706 mi 2 ) are located in Mexico and 24,770,640 ha (95,639 mi 2 ) in the United States. We primarily used Landsat 8 (OLI) imagery, supplemented by limited ground surveys from two field campaigns, drone-based aerial data, and existing vegetation classification frameworks from both countries. The classification applied a data-fusion approach integrating 30-m Landsat 8 imagery, decadal phenology metrics from vegetation indices, and a random forest model trained mainly with datasets from a comprehensive national mapping project from the U.S. Geological Survey (USGS) GAP Analysis Project (GAP) and federal wildland fire agencies’ Landscape Fire and Resource Management Planning Tools (LANDFIRE) (GAP/LANDFIRE) [United States side] and the National Institute of Statistics and Geography (INEGI) [Mexico side] as well as land cover maps and opportunistic open-access and field observations. Mapping of the full BCR 33 region was carried out in two phases: 1) Phase I, the prototype map, covered a smaller portion of the transboundary area and identified 31 land cover classes, and 2) Phase II, the full BCR 33 map (refer to Figure 1), which resulted in 152 land cover classes. Using a Random Forest classifier, we achieved an overall prediction accuracy of 92% for the Phase I map and 87% for the Phase II full region map. This slight decrease can be attributed to working on a larger, more complex area with a greater number of land cover classes. No formal validation was conducted, aside from using a subset of the collected field observations and training data to assess model performance during and after training. The training sites were further verified using Google Earth (Google, 2026) imagery. Two undergraduate students who worked for over a year visually inspected imagery and open access public images to confirm each training site during model training using in-house developed, online, visual tools. A portion of this field training data was reserved for model validation, and the corresponding results are to be presented in later sections. The project developed an end-to-end, medium- and fine-resolution remote sensing–based data fusion mapping approach. This effort produced a map (Nagler et al., 2025) and the online tools to support a dynamic, live, online map for visualizing the transboundary vegetation communities in BCR 33. The toolset is currently hosted by the University of Arizona (UofA) Vegetation Index and Phenology (VIP) Lab to support FWS partners (https://vip.arizona.edu/viplab_data_explorer?LCM_BCR33). The online map is designed to allow rapid updates using new training, validation, or correction data, making it dynamic and maintainable. The approach we took established a framework for rapid updating and correction of land cover maps, as the model can be quickly retrained with new field observations, updated training data, or other sources. This enables dynamic mapping and change detection of the region’s vegetation. This framework is an advance in data fusion and crowdsourced mapping of complex, vulnerable regions, providing support to regional stakeholders and the wider user community. This transboundary map can inform the protection, conservation, and restoration of vegetation, habitat, and ecosystems, particularly for threatened and endangered species across the two nations using consistent and harmonized binational mapping systems. Beyond supporting land management decisions and stakeholders in the transboundary desert ecoregions, this BCR 33 mapping effort establishes a foundation for future rapid, low-cost, cross-border land cover mapping that can benefit and advance ecosystem management.

Arizona, Baja California, California, Nevada, Sina↗

USGS2023

Random forest classification of multitemporal Landsat 8 spectral data and phenology metrics for land cover mapping in the Sonoran and Mojave Deserts

Geospatial data and tools evolve as new technologies are developed and landscape change occurs over time. As a result, these data may become outdated and inadequate for supporting critical habitat-related work across the international boundary in the Sonoran and Mojave Deserts Bird Conservation Region (BCR 33) due to the area’s complex vegetation communities and the discontinuity in data availability across the United States (US) and Mexico (MX) border. This research aimed to produce the first 30 m continuous land cover map of BCR 33 by prototyping new methods for desert vegetation classification using the Random Forest (RF) machine learning (ML) method. The developed RF classification model utilized multitemporal Landsat 8 Operational Land Imager spectral and vegetation index data from the period of 2013–2020, and phenology metrics tailored to capture the unique growing seasons of desert vegetation. Our RF model achieved an overall classification F-score of 0.80 and an overall accuracy of 91.68%. Our results portrayed the vegetation cover at a much finer resolution than existing land cover maps from the US and MX portions of the study area, allowing for the separation and identification of smaller habitat pockets, including riparian communities, which are critically important for desert wildlife and are often misclassified or nonexistent in current maps. This early prototyping effort serves as a proof of concept for the ML and data fusion methods that will be used to generate the final high-resolution land cover map of the entire BCR 33 region.

Remote Sensing↗

USGS2022

Spatial models of jaguar energy expenditure in response to border wall construction and remediation

The construction of a wall at the United States-Mexico border is known to impede and deter movement of terrestrial wildlife between the two countries. One such species is the jaguar, in its northernmost range in the borderlands of Arizona and Sonora. We developed an anisotropic cost distance model for jaguar in a binational crossing area of the Madrean Sky Islands at the United States-Mexico border in Southern Arizona as a case study by using previously collected GPS tracking data for jaguars, bioenergetic calculations for pumas, and a digital elevation model. This model describes projected energy expenditure for jaguar to reach key water sources north of the international border. These desert springs and the broader study region provide vital habitat for jaguar conservation and reintroduction efforts in the United States. An emerging impediment to jaguar conservation and reintroduction is border infrastructure including border wall. By comparing walled and un-walled border sections, and three remediation scenarios, we demonstrate that existing border infrastructure significantly increases energy expenditure by jaguars and that some partial remediation scenarios are more beneficial than others. Our results demonstrate opportunities for remediation. Improved understanding of how border infrastructure impacts physiological requirements and resulting impacts to jaguar and other terrestrial wildlife in the United States-Mexico borderlands may inform conservation management.

Arizona↗