Cross-jurisdictional monitoring for nonnative plant invasions using ndvi change detection indices in walnut canyon national monument, Arizona, USA

Nonnative invasive plants pose a significant threat to native biodiversity on National Park Service (NPS) and other public lands, which is a fact that prompts active federal prevention and control programs. Within the NPS, park units are assisted by the inventory and monitoring networks whose role is to provide long-term monitoring and limited research support to individual park management efforts. The Southern Colorado Plateau Inventory and Monitoring Network (SCPN) is currently testing methods for selecting monitoring locations to a sist in the early detection and rapid control of priority nonnative plant species including areas where there is a cross-jurisdictional threat to park resources. In response to SCPN research needs, we sought to find out if the normalized difference vegetation index (NDVI) is capable of detecting locations invaded by nonnative plants across a landscape with diverse ownership and management regimes. We compared nonnative species richness and nonnative community composition to NDVI values and disturbance types with the expectation that high numbers of nonnative plants would indicate disturbed locations that are a high monitoring priority. We hypothesized that sites with negative NDVI values would correspond to areas with increased nonnative species richness and that disturbances resulting in decreased plant cover would be greatest in those locations. Using difference maps (DMs) derived from Landsat 5 TM images, we classified five NDVI change value categories in GIS (Geographic Information System). Using NDVI classes to stratify plots and delineate categories for statistical comparison, we collected nonnative invasive plant community data and disturbance types in 150 Ã- 150 m plots. We then used plot data collected in 2008 and 2009 to test the application of NDVI for the detection of gradients of disturbance and nonnative invasive species richness at Walnut Canyon National Monument, a SCPN unit in Northern Arizona. In addition we compared three different analysis techniques (image subtraction, difference from average, and single-year image dates temporally proximal to field data collection) and seasonally different image dates (August vs. October) to find which method and season had the best modeling potential. Multiresponse permutation procedures and analysis of variance showed that NDVI DM change values were sensitive to differences in nonnative species richness, nonnative species community composition, disturbance type, and land ownership. Ordinary least quares regression analysis demonstrated that the October, 2008, NDVI image dates had the best fit to field data among the three analytical methods and the two different months. Though our expectation that negative NDVI values would correspond to disturbed areas with high invasive species richness was incorrect, the correlation between NDVI change values, management jurisdiction, disturbance, and nonnative species richness indicates that satellite-derived vegetation indices have potential applications for locating sites at high risk to nonnative invasion and for improving control efforts. Our research also demonstrated that collaboration between stakeholders is necessary to effectively meet the objectives of NPS nonnative monitoring and control efforts.