planting density, climatic factors, nutrient status, irrigation management, salinity, and
other conditions.
VI methods replace (or supplement) crop coefficients with a VI that reflects the
actual growth stage of the crop at the time of measurement. VI-based crop coefficients
(K c , VI ) have been developed for individual and mixed crops in agricultural regions
starting nearly 30 years ago (e.g., Neale et al., 1989). More recently, the concept has
been applied to natural ecosystems at local, regional, and global scales of measurement (Glenn et al., 2010):
ET = ET o VI*
(5.5)
where VI* is a vegetation index scaled between bare soil (VI* = 0) and full vegetation
cover transpiring at the rate of ET o (VI* = 1.0) by the formula
VI* = 1 −
VI max − VI
VI max − VI soil
(5.6)
where VI max is the VI of fully transpiring vegetation and VI soil is the VI of bare, dry
soil with assumed ET of 0. The VI max and VI soil are usually determined by sampling
areas of bare soil and dense vegetation within each image (Nagler et al., 2005a,e;
Baugh and Groeneveld, 2006; Groeneveld and Baugh, 2007; Groeneveld et al., 2007).
It would be desirable to have paired sites, with and without beetle damage, to
compare imagery taken at the same time. However, beetle damage was always
widespread at a given site, so this sampling method was not possible. In our saltcedar
beetle studies, we obtained one or two annual summer Landsat 5 images for each year
from 2000 to 2010 for beetle release sites on six western U.S. rivers, encompassing
years before and after beetle release at each site (Nagler et al., 2012). Images with
no cloud cover over the sites of interest acquired from June 15 to August 15 were
selected. Level 1T processed images referenced to fixed ground points were
obtained from the U.S. Geological Survey Earth Explorer website (http://
edcsns17.cr.usgs.gov/NewEarthExplorer/). Sites were chosen that encompassed
one or more MODIS pixel footprints with the riparian area.
Methods for processing band data, converting NDVI values to scaled values
(NDVI *
TM ), and calculating ET followed methods developed to estimate annual ET by
western U.S. phreatophye communities, including Tamarix sites, from single summer
Landsat images (Baugh and Groeneveld, 2006; Groeneveld and Baugh, 2007;
Groeneveld et al., 2007). DN values were converted to apparent at-satellite reflectance
values using data in the header files and equations and tabular information in the
Landsat 7 Science Data Users Handbook (Irish, 1999). NDVI values were then
scaled (NDVI
*
TM ) between bare soil (NDVI Soil ) and maximum vegetation response
(NDVI max ) on each image using the relationship in Equation (5.6) where NDVI Max
was the highest NDVI value on the image determined from a display of the pixel
statistics in ERDAS Imagine (ERDAS, Inc., Norcross, GA) and NDVI Soil was the
NDVI of an area (1–2 ha) of dry lakebed or rock outcrop that was apparently
unvegetated and stable year to year. This bare-soil (or rock) area was selected on
92
CHANGE DETECTION USING VEGETATION INDICES AND MULTIPLATFORM
other conditions.
VI methods replace (or supplement) crop coefficients with a VI that reflects the
actual growth stage of the crop at the time of measurement. VI-based crop coefficients
(K c , VI ) have been developed for individual and mixed crops in agricultural regions
starting nearly 30 years ago (e.g., Neale et al., 1989). More recently, the concept has
been applied to natural ecosystems at local, regional, and global scales of measurement (Glenn et al., 2010):
ET = ET o VI*
(5.5)
where VI* is a vegetation index scaled between bare soil (VI* = 0) and full vegetation
cover transpiring at the rate of ET o (VI* = 1.0) by the formula
VI* = 1 −
VI max − VI
VI max − VI soil
(5.6)
where VI max is the VI of fully transpiring vegetation and VI soil is the VI of bare, dry
soil with assumed ET of 0. The VI max and VI soil are usually determined by sampling
areas of bare soil and dense vegetation within each image (Nagler et al., 2005a,e;
Baugh and Groeneveld, 2006; Groeneveld and Baugh, 2007; Groeneveld et al., 2007).
It would be desirable to have paired sites, with and without beetle damage, to
compare imagery taken at the same time. However, beetle damage was always
widespread at a given site, so this sampling method was not possible. In our saltcedar
beetle studies, we obtained one or two annual summer Landsat 5 images for each year
from 2000 to 2010 for beetle release sites on six western U.S. rivers, encompassing
years before and after beetle release at each site (Nagler et al., 2012). Images with
no cloud cover over the sites of interest acquired from June 15 to August 15 were
selected. Level 1T processed images referenced to fixed ground points were
obtained from the U.S. Geological Survey Earth Explorer website (http://
edcsns17.cr.usgs.gov/NewEarthExplorer/). Sites were chosen that encompassed
one or more MODIS pixel footprints with the riparian area.
Methods for processing band data, converting NDVI values to scaled values
(NDVI *
TM ), and calculating ET followed methods developed to estimate annual ET by
western U.S. phreatophye communities, including Tamarix sites, from single summer
Landsat images (Baugh and Groeneveld, 2006; Groeneveld and Baugh, 2007;
Groeneveld et al., 2007). DN values were converted to apparent at-satellite reflectance
values using data in the header files and equations and tabular information in the
Landsat 7 Science Data Users Handbook (Irish, 1999). NDVI values were then
scaled (NDVI
*
TM ) between bare soil (NDVI Soil ) and maximum vegetation response
(NDVI max ) on each image using the relationship in Equation (5.6) where NDVI Max
was the highest NDVI value on the image determined from a display of the pixel
statistics in ERDAS Imagine (ERDAS, Inc., Norcross, GA) and NDVI Soil was the
NDVI of an area (1–2 ha) of dry lakebed or rock outcrop that was apparently
unvegetated and stable year to year. This bare-soil (or rock) area was selected on
92
CHANGE DETECTION USING VEGETATION INDICES AND MULTIPLATFORM