The principal objective of the proposed research was to integrate empirically based models of non-native plant invasion, fire, and native species habitat in a spatially explicit decision-support package that informs sustainable resource management and recovery in the face of ongoing climate change. The project team modeled distribution, biomass, invasion risk, and fire risk associated with the following non-native invasive species (NIS): African buffelgrass (Pennisetum ciliare), red brome (Bromus rubens), Sahara mustard (Brassica tournefortii), Mediterranean grass (Schismus spp.), and arugula (Eruca vesicaria sativa) on the U.S. Army Yuma Proving Ground, Barry M. Goldwater Air Force Range, Kofa and Cabeza Prieta National Wildlife Refuges, and Organ Pipe Cactus National Monument. 

Technical Approach

Research involved extensive field sampling efforts to train and test regional- and landscape-scale models of non-native invasive plant distribution and biomass. Species-specific models incorporated novel remote sensing techniques that identified NIS based on both phenological and spectral differences using satellite platforms of differing spatial, temporal, and spectral resolutions. Species distribution maps at landscape and regional scales were used to assess biogeographical relationships of invasive plants to land use and climate and to model changing invasion risk with global change. Biomass maps were used to model fuel loads and to predict areas of high fire risk, hazard, and behavior. Invasion and fire risk predictions, taking into account potential management and mitigation responses, were integrated with models of resource use and habitat connectivity for wildlife species. The research culminated with the embedding of the above results into a spatial decision-support package to guide management on Department of Defense (DoD) and surrounding lands.


During this project, detections of B. tournefortii and Schismus spp. were relatively common across the study area, whereas P. ciliare, E. vesicaria sativa, and B. rubens were relatively uncommon. B. tournefortii demonstrated relatively specific habitat conditions under which it becomes dominant, whereas Schismus spp. exhibited more generalist habitat requirements and were present in most sampled ecosystems. By contrast, P. ciliare and B. rubens exhibited greater invasion potential in upland ecosystems. E. vesicaria sativa appeared likely to spread beyond its current distribution. Modeling results confirmed that the advanced remote sensing and modeling techniques developed by the project enabled identification of NIS habitat. In particular, models derived from MODIS and Landsat TM were appropriate for describing the likelihood of finding B. tournefortii, due to the unique phenology of the species and a strong contrast with native vegetation green-up. Schismus spp. were less distinct, both spectrally and temporally. The use of a spatially weighted ensemble approach to mapping improved our B. tournefortii models, likely because spatial heterogeneity in precipitation drove phenological variability across the study area for this species. For both species, models were challenged by low abundances of the target species as a result of unusually low precipitation in both 2011 and 2012. In all, the project (1) generated more than 400 post-processed time series (2000 to 2014) MODIS, Landsat, WorldView-2 (WV2), and SPOT5 images and associated derived phenometrics/indices; (2) generated probabilistic models of habitat suitability across the study area for all target species; (3) developed regional models of presence, cover, and biomass for all target species (based on MODIS, Landsat, WV2, SPOT5); (4) refined existing spatial databases of Schismus spp., B. tournefortii, B. rubens, P. ciliare, and E. vesicaria sativa occurrence for the study area and region (CA, AZ, NV, UT, NM); (5) updated regional models of current and future risk of invasion by B. rubens and B. tournefortii; and (6) derived regional models of significant fire risk under different fuel load/climate scenarios, as well as a novel regional model of fire connectivity. The project’s sampling design was deliberately iterative and targeted. Results from this design, along with the habitat suitability and occurrence models, enables mapping of key species of ongoing management concern over a large geographic area encompassing multiple DoD installations and other jurisdictions. Results also highlighted areas where ongoing or new fuels monitoring activities might be focused.


This research effort developed new techniques, models, and maps related to fundamental ecological changes driven by NIS, fire, and global change in the Southwest. The research products are practical and relevant to management on DoD lands and across the surrounding Sonoran Desert ecoregion. Importantly, the approaches advanced by the project were designed to leverage multiple new and freely-available information sources, so that the methods would be easily transferrable and repeatable. The project’s assessment of landscape- and regional-scale ecological risk in a spatial management framework enables DoD to integrate environmental objectives with training needs, and to provide leadership on feasible management responses to global change.