< Back  ||  Home > Reports > Elk Habitat

Modeling Elk Habitat with Remote Sensing

This portion of the study was performed to explore the relationship between the temporal dynamics of vegetated landscapes and the elk population in Arizona. The analysis is based on four main data sources: 

(1). Two elk population estimates were used based on AGFD estimates and estimates calculated using the method of Bender and Spencer (1999). Elk population estimates were provided for the following sets of GMUs covering the years indicated.

(2). We used multitemporal NOAA-AVHRR data to characterize the temporal dynamics of the GMUs for which elk population estimates were available. Seasonal and interannual changes in landscapes reflect vegetation phenology and condition. These landscape dynamics were captured using data collected by the NOAA-AVHRR sensor (EDC, 1994). The relatively coarse spatial resolution of the sensor (1km2) is offset by its high temporal resolution: images are acquired daily, as opposed to the bi-weekly revisit cycle of many other sensors. AVHRR data are often used in research on land-cover change because the frequency of data availability permits the detection of short-lived vegetation and landscape changes that may be missed by other sensors. 

A derived image called the Normalized Difference Vegetation Index (NDVI) is calculated from the AVHRR reflectance values in the visible and near-infrared regions of the electromagnetic spectrum (see Figure 1). The NDVI represents a measure of vegetation photosynthetic activity, and is sensitive to various biophysical vegetation characteristics, such as biomass and percent cover (Huete and Jackson 1987). To produce a cloud-free image representing actual ground conditions, the maximum value of the daily NDVI images is composited over each 2-week period. These composite data sets are produced by the U.S. Geological Survey (USGS) (EDC, 1994) and are available for the state of Arizona through the ARIA website. The methods we develop are directly applicable to the Moderate Resolution Imaging Spectroradiometer (MODIS) data. MODIS data is also being collected daily, but it has a much finer spatial resolution (250 meters sq.). The MODIS deliverables will include an Enhanced Vegetation Index (EVI) that is analogous to the NDVI but incorporates corrections for view-angle, soil background and atmosphere. 

(3). High spatial resolution data from the Landsat-7 Enhanced Thematic Mapper (ETM+) were used to map vegetation association. The size and pattern of vegetation are landscape characteristics known to influence wildlife movement (Krausman, 1996). For example, canopy closure affects the amount of annual vegetation available for elk forage and may influence visibility of elk to predation (Risenhoover and Bailey, 1980). Such structural characteristics are captured in the Landsat-7 15-meter panchromatic and 30-meter multispectral data and, therefore, are reflected in the derived vegetation classification. Ikonos data, with a spatial resolution of 1-meter pixels, are available for the V Bar V ranch. This fine scale allows for the discrimination of individual tree canopies and permits interpretation of precise details of vegetation community structure in the ranch area where we have information on grazing activities. 

(4). We also utilized detailed vegetation mapping, fine-scale topography, and terrestrial ecosystem units available for the region surrounding the Walker Basin area to refine the classification of the Landsat-7 data. The U.S. Forest Service Terrestrial Ecosystem Survey (TES) of the Coconino National Forest recognizes 134 mapping units. Initial classifications were based upon delineation of topography, geology, and vegetation using stereo pairs of 1:24000-scale aerial photographs. Fieldwork was conducted to assess the accuracy of these initial units and to collect additional terrain data. The vegetation classification employed in developing the TES units is hierarchical, with broadest categories applicable over larger spatial scales. The classification hierarchy, in descending order, considers vegetation structure, regional phenology, and physiology of dominant species (Miller et al., 1995). The TES units were subsequently merged, also on a hierarchical basis into broad cover classes. This vegetation cover map covered a portion of the study area and allowed us to extrapolate the available vegetation mapping into the entire study area using the remote sensing data and to interpret the spectral response patterns of the various classified vegetation units. 

More...

Creating a vegetation cover map

Characterizing landscape temporal dynamics

Linking the temporal and spatial data sets

Relating population estimates to landscape characteristics

Identifying preferred landscapes based on vegetation greenness

Mapping landscapes accessed by wildlife using sightings data

Relating rangeland condition to climate indices (ENSO)