Laserfiche WebLink
3. Method Used to Map Estimated Groundwater ET <br />ET calculations for the South Platte drainage were made as annual totals. For these <br />estimates, the vegetation index, NDVI*, from Landsat TM data, with a distribution from <br />zero to one, was used in the manner of a crop coefficient. The decimal fraction for <br />NDVI* was set equivalent to ET from groundwater (ETg) that was also transformed from <br />zero to one. ETg is simply ET minus the component provided by precipitation. <br />The techniques used here for phreatophyte ET estimation were adapted from scientific <br />analysis by Groeneveld and Baugh and Groeneveld, et al. (both currently in review, <br />Journal of Hydrology). The technique uses peak-growing-season satellite data as the basis <br />for estimation and mapping of water use. These techniques and their adaptation are <br />described here for background. <br />Landsat TM data were used for this project. Pixels of Landsat TM have an area of 0.2 <br />acres, and so, every 0.2-acre pixel received an estimate of annual ET in this analysis. The <br />technique uses NDVI, normalized difference vegetation index, and an adaptation of <br />NDVI, called NDVI* that is processed to remove non-systematic data scatter. Baugh and <br />Groeneveld (2006) evaluated NDVI* against all published vegetation indices, including <br />simple NDVI, and found that it was superior to all other indices for predicting a linear <br />relationship between vegetation growth and antecedent precipitation, thus reflecting <br />removal of the scatter inherent in satellite-derived vegetation indices. <br />3.1 Satellite Data and NDVI* <br />NDVI was calculated from the red (Band 3) and the near infrared (Band 4) channels of <br />Landsat TM according to: <br />NDVI = (NIR - R) / (NIR + R) (1) <br />Where R =red <br />NIR =near infrared <br />The calculation of NDVI* stretches NDVI from 0 to 1 for the lowest and highest values <br />of NDVI across the complete range of vegetation within each image. These lowest and <br />highest expected values are predicted using cumulative distribution functions as <br />described below. To enhance these estimates, pixel values extracted from each image <br />were chosen from the immediate regions surrounding the riparian corridor areas <br />containing phreatophytes (rather than, for example, performing this operation on the <br />entire image). This geographic pairing is important since this procedure corrects for <br />differential scatter or absorption in Bands 3 and 4 due to atmospheric pollutants that vary <br />spatially, especially due to the altitude of the target surface. Thus, these analyses avoided <br />higher-elevation mountainous areas that would lack the path length and atmospheric <br />particulate/aerosol content more appropriate for the Denver region and Front Range <br />urban corridor through which the South Platte River flows. <br />