Bromus tectorum, also known as cheatgrass, is a wide ranging invasive grass impacting the ecology of many regions of North America. This invasive grass has been shown to encroach and alter ecological systems in shrub land environments by utilizing critical resources and providing fuel for fire. Through these vectors, cheatgrass inhibits regeneration of native plant species and has been shown to initiate a cascading effect on the ecosystem as a whole, impacting both flora and fauna.
The conversion of native shrub lands to cheatgrass is rapidly becoming a principal concern of land management agencies, ecologists, and environmental groups within the Great Basin region. Additionally, the concern is being compounded as the public demands a solution to successful and efficient wildland accounting. Added to these issues, the overall lack of monetary resources and an increasing time deficit at the forefront of many management projects, agencies are looking to remote sensing and geospatial technologies for solutions to administer and monitor, in real-time or near real-time, ecological changes across landscapes. To this end, the Moderate Resolution Imaging Spectroradiometer (MODIS) was employed to determine its feasibility and usability with regards to mapping the occurrence and spread of cheatgrass across the Great Basin.
While MODIS makes use of 36 spectral bands, it was concluded that the 16-day MODIS Normalized Difference Vegetation Index (NDVI) composite data product would best suit this study. The NDVI product is generated from the bidirectional reflectance factors of band 1 (620 – 670 nm) and band 2 (841 – 876 nm) as a normalized ratio of the NIR and red bands. Using this measure, it was determined that the inter-annual, or seasonal, variability would be visible in this dry region. Imagery was collected for years 2000 thru 2004 (Julian dates 081 – 305) then reprojected, mosaiced, and subset for analysis.
Southwest Regional GAP (SWReGap) analysis vegetation sample points were used to delineate location, presence, and percent cover of cheatgrass. Additional biophysical variables such as elevation, slope, aspect, and soils were included to create a regression tree model delineating current distribution of cheatgrass across a sub-region of the Great Basin. This method was met with moderate success.
Utilizing 30 percent of the points sampled and set aside for validation purposes, a retrospective accuracy assessment was completed. This assessment identified that the highly negatively skewed distribution of cheatgrass greatly impacted the overall validity of both the regression models. Due impart to the high number of low percent cover cheatgrass sample sites, the model included many areas that are otherwise unimpeded by Bromus tectorum. The modeling exercise was a successful attempt to utilize MODIS NDVI data to track and evaluate the spread of non-native species on large areas.