Section I

Vegetation index values derived from AVHRR channel 1 and 2 (visible and near-IR) data, and surface radiant temperature (AVHRR channel 4) will be examined on a single date, and for a composite that includes several AVHRR images for the DFW region. This region has been selected for this module due to the sustained urbanization that it has experienced over the last 15 years. A general map of the area with the location of lakes/reservoirs that will assist in locating DFW in the satellite images is shown in Figure 1.1.

Figure 1.1. Map of the Dallas-Ft. Worth (DFW) region acquired from U.S. Bureau of the Census Tiger Map Service (http://tiger.census.gov/)

AVHRR Channels 1, 2 and 4 data.

The data acquired from channels 1, 2 and 4 of the AVHRR on 10 July 1991 for the Dallas-Ft. Worth region are contained in three data files respectively: dfwch1.dat, dfwch2.dat and dfwch4.dat. Each of these binary files has 135 lines (rows) and 155 samples (columns) of single-band data with zero header bytes. The data acquired in AVHRR channel 1 is in the visible part of the electromagnetic spectrum (0.58- 0.68 micrometers) while channel 2 is located in the near-infrared (0.72 - 1.1 micrometers) part of the spectrum. Channel 4 acquires data in the thermal-infrared portion of the spectrum (10.3 - 11.3 micrometers). The data of Channels 1 and 2 are provided as reflectance values while that of Channel 4 is provided as radiant temperature of the measured surface.

The channel 1 and 2 reflectance data have been converted to a byte range of 0 to 255 by scaling values from 0 to 63.5 to 0 to 254 (.25% per byte value). All reflectance values greater than 63.5 were assigned a byte value of 255 as these values were representative of clouds, snow, or other bright non-vegetated surfaces. Channel 1 or 2 (Chan1 or Chan2) data can be rescaled to their actual reflectance values with Equation 1.1. The Channel 4 surface temperature values (T4sfc) have also been scaled and can be rescaled to their actual temperature values with Equation 1.2. Further information on the processing of the AVHRR data set used in this module can be found HERE (http://s un1.cr.usgs.gov/glis/hyper/guide/usavhrr).

Chan1 (or Chan2) reflectance value = (Chan1byte / 4)      
(Equation 1.1)



T4sfc (Degrees Celcius) = (T4sfcbyte)/2 - 83.0                         
(Equation 1.2)

1.P1. Use the display system of your software to view the three single-channel AVHRR images of the DFW region or view Figures 1.2a, 1.2b, and 1.2c that correspond to the data files above. The data displayed in the figures have been "stretched" (i.e., re-assignment of the greyscale of the display) to enhance the contrast between low and high reflectance values (See implementation note 1.P.1).

Figure 1.2. Images that display the Chan1 (2a), Chan2 (2b) and T4sfc (2c) AVHRR data acquired for the DFW region on 10 July 1991.

1.Q1. Compare the three channels of AVHRR imagery (Figures 1.2a, b, and c)with the map of the DFW region in Figure 1.1. In which image does the DFW urban area appear to be different from the surrounding region? What other features appear in the images?

In analyzing the single-channel AVHRR images, it is important to consider that the data is integrated over a band of the electromagnetic spectrum, and that the reflectances of terrestrial surfaces can change within a given channel and between channels. Figure 1.3 shows the spectral behavior of water, vegetation and bare soil between 0.4 and 1.2 micrometers.

Figure 1.3. Generalized spectral behavior of water, vegetation and bare soil between 0.4 and 1.2 micrometers. The spectral ranges of measurement for channels 1 and 2 of the AVHRR are also shown (after Richards, 1993).

1.Q2. Compare the images for AVHRR channels 1 and 2. In which image are water bodies better defined? Might you expect that bare soil could be distinguished from vegetation using only the channel 2 image? Explain how changes in reflectance of terrestrial surface types influence changes in contrast between channels 1 and 2.

The distinction between channels 1 and 2 of the AVHRR can be more clearly seen by combining the two bands into a false color image. Such a composite has been created based on the assignment of the data from channel 2 to the red image plane and channel 1 to the blue and green. The false color image includes the weather station ID numbers that will be utilized as a part of this module.

1.P2. Use your software system to create a false color image of the DFW region as described above with the AVHRR channel 1 (dfwch1.dat) and channel 2 data (dfwch2.dat) or examine Figure 1.4 (See implementation note 1.P.2).

Figure 1.4. False color image of the DFW region with weather station locations.


The AVHRR images for a single date (Figures 1.2 and 1.4) serve as a good example of why composites (produced from several single dates of satellite acquired data) of the data are often utilized. Observations of the earth on any single date are often contaminated by the presence of clouds, which absorb, scatter, or reflect radiation in the AVHRR sensor wavelengths. Note the clouds in the southeast corner of the DFW false color composite (bright white patches).

1.Q3. Look at the variation in color in the false color composite image. There are two predominant colors in the image. What are they? What areas do they appear to correspond to (based on Figure 1.1)? Are the urban areas uniquely defined by one of these colors? Why or why not? (HINT: Think about land cover reflectances shown in Figure 1.3).

Computation of the vegetation index

The normalized difference vegetation index (http://eros.usgs.gov/greenness/whatndvi.html) is utilized in this module. This index is computed by taking the difference between the near-IR and visible reflectance values divided by their sum:

NDVI = (Chan2 - Chan1) /  (Chan2 + Chan1)                  
     (Equation 1.3).

The NDVI, like most other vegetation indices, is based on the greater reflectance by vegetation in the near-infrared wavelengths (Chan2 of AVHRR) compared to visible wavelengths (Chan1 of AVHRR), as shown in Figure 1.3. Features that have similar responses in both visible and near-IR wavelengths would have negative or low values of the NDVI (range of -1.0 to 1.0). The NDVI data for the 10 July 1991 image are available as dfwnd.dat (135 lines, 155 samples) and displayed in figure 1.5.

The conversion algorithms for the NDVI data to the actual values is:

NDVI = (NDVIbyte - 100)/100                    (Equation 1.4).

Figure 1.5. The NDVI image of the DFW region computed from the Chan1 and Chan2 data acquired on 10 July 1991.

1.P3. Use your software display system to view the NDVI data of the DFW region: dfwnd.dat or examine Figure 1.5 (See implementation note 1.P.3).

1.Q4. Based on Figure 1.3, approximately what NDVI values would you expect for i.) lakes, ii.) bare soil locations and iii.) densely vegetated locations? How do your calculations of these values compare to the brightness intensity shown in the NDVI image in Figure 1.5 (where dark areas correspond to low NDVI and bright area correspond to high NDVI)?

1.Q5. Are urban areas (based on Figure 1.1) better defined in the NDVI image than in the false color composite image? Explain.

Data Composites

In addition to the data from 10 July 1991, a composite of data acquired from 5 July through 1August 1991 for the DFW region will also be evaluated. Each image acquired during the 5 Julythrough 1 August 1991 interval was compared temporally pixel to pixel. The maximum observed value of NDVI observed during the interval was retained for each pixel in the composite image. The composite process minimizes the likelihood that clouds (with low NDVI) remain in the composite image. The composite NDVI data area available at dfwndc.dat (135 lines, 155 samples) and Channel 4 data are available at dfwch4c.dat (135 lines, 155 samples).

1.P4. Use your software display system to view the temporally composited AVHRR images for NDVI dfwndc.dat and channel 4: dfwch4c.dat or examine Figures 1.6a and 1.6b (See implementation note 1.P.4).

Figure 1.6. The composite NDVI (1.6a) and Tsfc4 (1.6b) images for the DFW region.

1.Q6. Visually compare the two composite images to their respective counterparts from 10 July 1991. Should your software permit, compute the mean and standard deviation of the data values that comprise the entire images. What are the differences in each pair? Which composite appears to provide the greater improvement over the 10 July 1991 image? Why? (See implementation note 1.Q.6)

Data Extraction and Comparison.

Initially the NDVI and thermal-IR based land surface temperature (AVHRR Channel 4) data will be extracted in the vicinity of the 7 surface observation weather stations around the Dallas-Ft. Worth metropolitan region. The data will be extracted and compared for both the single date and composite data sets. The location of the pixels associated with the weather stations that are displayed in Figure 1.4 are listed in Table 1.1. A 3 by 3 pixel window was utilized when possible for each station. Station 280 was located near a water body and thus only is one sample (column) in width was extracted.

Table 1.1. Station location and AVHRR data extraction information.

Once the data was extracted they needed to be rescaled to their actual (real) values from the byte values of the images (Equations 1.2 and 1.4).

1.P5. Use your software to extract the single date and composite NDVI and T4sfc values for the pixels associated with the stations identified in Table 1. 1. The values of NDVI and T4sfc should be averaged over the pixels of each station. Convert these values into the actual values using Eq. 1.2 and 1.4. and fill out Table 1.2 (See implementation note 1.P.5).

Table 1.2. NDVI, T4sfc, and the ratio of T4sfc/NDVI computed for the pixels associated with the weather observation stations from the 10 July 1991 and 5 July - 1 August 1991 images.

Note that the NDVI values for the stations are all larger in the composite data compared to the single date. This should be expected since the largest NDVI value observed during the interval was retained. The T4sfc values are also greater in the composite data set, except for station 285. The T4sfc data that are retained in the composite product are the values observed on the date when the maximum NDVI was observed.

The relationship between T4sfc and NDVI is shown in Figure 1.7 for the 10 July 1991 image and Figure 1.8 for the temporally composited (5 July - 1 August 1991) image. Along the ordinate, values of T4sfc/NDVI are used instead of T4sfc. These values act to emphasize the significance of NDVI in influencing values of T4sfc.

Figure 1.7. T4sfc/NDVI displayed as a function of NDVI for the 10 July 1991 image.

Figure 1.8. T4sfc/NDVI displayed as a function of NDVI for the 5 July through 1 August 1991 image.

1.Q7. What type of relationship between T4sfc/NDVI and NDVI exists in Figure 1.7? In Figure 1.8? Explain the apparent difference.

1.Q8. What type of surface features would be expected at and around the surface observation stations with low values of T4sfc and high values of NDVI? With high values of T4sfc and low values of NDVI?

1.Q9. Based on the T4sfc and NDVI data, which meteorological stations are likely surrounded by urban, compared to rural, land surface features?