Plant invasions in dynamic desert landscapes. A field and remote sensing assessment of predictive and change modeling

Robust predictive models of invasive species inform long-term resource management. We used a scaled-down modeling approach, based on field data and high-spatial resolution imagery, to assess the predictive skill of combined genetic algorithm rule set-production (GARP) and change vector analysis (CVA) models. We hypothesized that highly dynamic desert environments are unstable, thus more vulnerable to invasion by exotic plant species than stable landscapes. Initial model results confirm this hypothesis. The GARP–CVA models identified areas vulnerable to invasion by Brassica tournefortii and Schismus arabicus over dynamic landscapes in the eastern portion of ‘El Pinacate’ Biosphere Reserve (ePBR), a natural area under potential increasing human pressure. The GARP–CVA models showed low accuracy when tested against confirmed locations of invasives due to the large modeling scale. Land cover characterization showed B. tournefortii association with microphyllous desert scrub, grassland, and sarcocaulescent desert scrub. S. arabicus was found associated with microphyllous and crassicaulescent desert scrub. The GARP–CVA models representing the most dynamic landscapes with high probability to invasion showed a good spatial agreement with the distribution of invasives per the land cover type. This relationship needs, however, to be investigated further because the ecology of these invasives is likely more complex than we can model.