Touching briefly on the methods and science behind this project

For those of you wondering what the plots might look like, here is one from the fall sampling. Click on the photo to get a larger version in which you can see the plot boundary (large square) and subplot boundaries (small squares). Point sampling is performed where the yellow lines (transects) intersect subplot boundaries or other transects. Thus, there are 25 point intercepts per subplot. The subplots are 30m by 30m (the size of a Landsat TM pixel). The plots are 250m by 250m (the size of a MODIS pixel). These satellite sensors are considered moderate resolution and you can see from the image that there is a lot of spatial heterogeneity at a scale smaller than what either a Landsat TM or MODIS TM pixel can see. Here is a photo taken at the furthest right subplot in the image above to give you a sense of the community structure and diversity:

For comparison, I've included a crude map showing what the same place on the ground looks like in a single Landsat TM image below. The outline of a single MODIS pixel is shown in black.

There are two hypotheses being tested. The first hypothesis is that invaded pixels have a different spectral signature. Landsat TM has 6 reflective bands, so each pixel can be described by a set of 6 values. One of the more famous derivatives of Landsat data (and many other satellite imagery) is the Normalized Difference Vegetation Index, or NDVI. NDVI utilizes the fact that vegetation has strong absorption in the red wavelength due to chlorophyll activity while and highly reflective in the near-infrared wavelengths due to cell structures. These characteristics are unique to vegetation, so NDVI is used to quantify the amount or health of vegetation. NDVI utilizes two of the six Landsat TM bands, so ostensibly, more information can be obtained from the full set of Landsat bands. Of course, the spectral "signature" of vegetation and other materials change over the course of season due to factors such as green-up and senescence, the wetting and drying of soils, and the presence of snow. Therefore, we seek to find an optimal timing in which invaded pixels are distinguishable from uninvaded pixels.

The second hypothesis is that the phenology of invasive species in the desert differ from native species. This has been demonstrated already for many species, including cheatgrass (Bromus tectorum), and one of our study species, Sahara mustard (Brassica tournefortii). The premise is that these species are more aggressive in their water usage, green up earlier, and senesce earlier. Additional evidence suggests that invasive grasses mute bimodal growing seasons.

What makes our approach viable is the incorporation of long time series of these satellite images. Shown below is a time series of MODIS NDVI over the course of 10 years. The goal is to match unique phenological signals in the time series with invasive species and biomass estimates gained in the field this spring and develop models that predict invasive species occurrence, abundance, and biomass at other locations.