Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → États-Unis → 2009 → Efficacy of using remotely sensed data for early detection of invasive plants in the Chihuahuan Desert


Efficacy of using remotely sensed data for early detection of invasive plants in the Chihuahuan Desert

Young, Kendal Eugene

Titre : Efficacy of using remotely sensed data for early detection of invasive plants in the Chihuahuan Desert

Auteur : Young, Kendal Eugene

Université de soutenance : NEW MEXICO STATE UNIVERSITY

Grade : Doctor of Philosophy (PhD) 2009

Invasive plants have the ability to displace native species, disrupt nutrient and fire cycles, and alter the character of the plant community by enhancing additional invasions. Early detection of invasive plants can circumvent negative ecological impacts to communities and ecosystems. This study evaluated the utility of using remotely sensed and presence-only datasets as a tool for the early detection of invasive plants in the Chihuahuan Desert. Thirteen invasive plant species were selected in Big Bend National Park (BIBE) and Holloman Air Force Base (HAFB). Locations of selected species were obtained from BIBE and HAFB and during field surveys conducted in 2006–2007. Predictive habitat models were created using a presence-only data modeling approach. Seven Landsat Enhanced Thematic Mapper Plus (Landsat 7 ETM+) datasets representing three seasons (fall, summer, spring) and five years (1999–2003) of spectral data were used for analyses on BIBE. Quickbird imagery was used on HAFB, and corresponded to the season (spring/summer 2007) of field data collection. Spectral bands, vegetation indices, digital elevation model, and land cover map represented environmental variables. Risk assessments were conducted by incorporating locations of known plants, suitable habitats, and vectors and pathways. Model performance was evaluated using the area under curve (AUC) metric from receiver operating characteristic curves (threshold independent), and the testing omission rate (threshold dependent). All plant predictive habitat models had high discriminate ability (AUC > 0.90) to differentiate suitable invasive plant habitat from random background locations. Predictive models yielded omission rates ≤ 19%, but were typically 5–15%. Quickbird imagery may be more desirable for invasive plant habitat modeling as habitat models derived from Quickbird imagery yielded fewer errors of omission than Landsat 7 ETM+ derived models. Habitat models and risk analysis were an effective approach to assess landscapes for the potential of invasive plant incursion. Models can inform managers regarding the potential introduction of invasive species and allocation of resources for invasive plant control. In addition to prioritizing areas to monitor or control invasive plants, spatial models can be used to evaluate the effects of invasive species occurrences on Threatened and Endangered species, species of concern, recreational activities, and management actions.

Search Oxford Libraries Online

Proquest Dissertations & Theses

Page publiée le 4 novembre 2011, mise à jour le 11 novembre 2018