Laserfiche WebLink
MEMO 89.2 <br />can handle a larger range of numbers). Since floating-point imagery occupies considerable amounts of <br />disk space and requires longer CPU time for processing, the imagery was resealed in ERDAS to facilitate <br />image-processing speed and reduce storage space in subsequent procedures (e.g., crop classification). <br />Converting the data from floating-point to 8-bit effectively reduces the numbers of decimal places <br />produced to numbers sensible for the NDVI index. <br />The NDVI data was transformed from floating-point to unsigned 8-bit format using the following <br />equation: <br />NDVI 8-bit = (NDVI float + 1) * 100 <br />Where, <br />NDVI 8-bit = Rescale NDVI imagery in unsigned 8-bit format <br />NDVI float =Original NDVI image in floating-point format <br />The result of this transformation is unsigned 8-bit NDVI imagery scaled between 0 and 200. This method <br />was chosen over the alternative minimum and maximum values rescale operation in order to ensure a <br />consistent range of NDVI values across all imagery dates (a prerequisite for the multitemporal analysis). <br />An example of the NDVI rescale model developed in ERDAS Imagine is shown in Appendix A. Figure <br />4 shows mean NDVI trends of selected irrigated crop types in the SPDSS area. <br /> Te mpor al M ean NDVI Trend <br />190 <br />180 - <br /> <br />170 ~ ' <br /> <br /> <br />160 <br />o <br />~ '~ <br />~ <br />\ tAlfalfa <br />(Small Grains <br /> <br />~ 150 <br /> <br />,~. ` ~ Corn <br />Sugar_Beets <br /> <br />~ <br />~ <br />~ 140 ~I~ Dry_Beans <br /> Grass_Pasture <br /> <br />130 Fallow (Bare Soil) <br /> <br /> <br />120 ' <br /> , <br />110 <br /> <br />PQ~ <br />~ ~ <br />~~ ~~~. <br />~ <br />~ ~~~. ~~~~ ~~~~ P~~ P~~ <br />' ~~Q ~~Q o~~ ~~~. <br />'1' <br />' <br />` <br />^o <br />ti ~ <br />h <br />~ ~ ,~ ~, q ; ~ ~ L <br />1 <br />~ <br />~L~' <br />Figure 4. Mean NDVI Trends of Selected Irrigated Crop Types in the SPDSS Area <br />An assessment of multi-temporal Tasseled Cap transformation (Grist and Cicone 1984) was conducted as <br />an additional source of information for discriminating irrigated vs. non-irrigated cover types. The <br />Tasseled Cap was later dismissed because results were similar in most respects to the results obtained <br />with NDVI, and did not add value to the final classification. <br />Figure 5 shows examples of NDVI images generated from the 2001 Landsat data using the methods <br />described above. The NDVI images are displayed in sequence from the first date, May 12 to the last one, <br />Page 6 of 45 ~RFversfde TecAnotogy, fnc. <br />4'JaYCr Resources Errgi~ecr:np an~i CansaFlrnp <br />