WSP12518 - Laserfiche WebLink

Home Browse Search

5Z F. ZAMORA-ARROYO ET AL .-.-' .-.-.- MEXICO ~U'C "'4J"co - " . 10km Figure 2. Location of study sites in the Colorado River delta, Mexico. The gray area is the stretch of riparian corridor between the levees which supports native trees, the main focus of this study. The stripped area below the native tree zone is a mixture of habitat types, including fresh water and intertidal marshes, mud and salt flats, and vast thickets of Tamarix ramosissima. Triangles denote sites where ground transects were established to quantify vegetation; place names are the settlements (access points) nearest each transect. Closed circles denote sites where well points were established to monitor the depth and salinity of the water table under the riverbed. Numbered line segments show where strings of digital DyCam images were acquired during a low-level fight over the delta in May 1999. visually scored at each intersection to determine % cover of each class. Results were then ground-trUthed at nine locations (see Nagler et al., 2001, for details). Native trees > 6 m height could be distinguished from other vegetation based on the length of shadows they cast in the photos. Shrubs (mainly T. ramosissima) were defined as plants that had definite size and shape but were < 6 m based on shadow length. Groundcovers were green areas on photographs in which individual plants could not be distinguished. Bare soil and water were identified by color (soils were light-colored whereas water appeared nearly black in multi-band images). REGENERATION OF TREES IN RESPONSE TO FLOOD RELEASES 53 Normalized difference vegetation index (NDVI) values could also be calcuated for each image using the ratio of Red and IR bands (Nagler et aI., 2001). A wider set of images (n = 84)(Fig. 2) established a high coefficient of determination (r2 = 0.83) between percent vegetation cover and reflectance-'based NDVI values cal- culated for the aerial photographs. This relationship was used to calibrate satellite images of the delta to determine percent vegetation cover over past years. c' el W t'v Satellite imagery (' We acquired six Thematic Mapper 5 (TM) images showing summer vegetation pat- terns covering before- and after-periods of release events from 1992~ 1999. Images for Path 38, Row 38 were selected for cloudless day in May 1992, July 1994, June 1996, July 1997, June 1998, and May 1998. An additional image was obtained for 23 Febmary 1997, to delineate areas inundated by flood flows during a release event. The Febmary 1997 and May 1999 images were taken within 3 weeks of aerial surveys over the delta. Images were preprocessed and georectified by EarthSat, Inc., Rockville, MD, U.S.A. Digital numbers were converted to exoatmospheric reflectance values (0-1) using archived radiance data for each scene and sun angle functions calculated from solar azimuth and angle based on date, time of day and attitude and longitude, using ENVI software (Christopher Jengo, EarthSat, Inc., pers. corom.). The scenes were masked to include only the area of interest depicted in Fig. I, and NDVI values were calculated using ERDAS software. o...J Change analyses There was a near 1 : 1 correspondence between NDVI values calculated for water, soil and vegetation on a May 1999 satellite image of the delta and on DyCam images obtained by low-level overflight the same month (Nagler et aI., 2001) and NDVI values for different land cover classes were nearly identical over the different TM images (coefficient of variation < 10%) from 1992 to 1999 (Fig. 3). Hence, we used the formula relating % cover to NDVI, determined for DyCam images, to estimate % cover 'on TM images (y = 180x + 7'95, r2 = 0'837) (Nagler et aI., 2001). ~o -1 Min. Water Mean Soil Max, Figure 3. Means and standard deviations (error bars) of ND~ values for similar. landscape features on 1992-1999 TM images of the delta. Max and Mm refer to the maxunum and minimum NDVI values on each image; Mean refers to the mean NDVI values of all pixels on each image; and water and soil refer to NDVI values f~r five randomly-selected water and soil areas on each image. Differences among years for SOil, water and mean NDVI values were not significant at p < 0,05 (max. and min. values could not be compared among years as there was only one value available per image).