Lab exercise 6, due right before the class on October 5, 2005
EES5053: Remote Sensing. Earth and Environmental Science, UTSA
http://www.utsa.edu/LRSG/
Student Name: ___________________
Atmospheric correction, radiance, reflectance,
and NDVI from Landsat image
Purpose
In this lab, you will learn a procedure of image processing from atmospheric correction, converting the digital number to radiance, and then to reflectance by using band math. Based on reflectance, you will learn how to calculate the NDVI.
1. Preparation
Create a subdirectory Lab6 under your EES5053, and make a directory MyWork under the Lab6 for you save your works. The Landsat image you will use is the same image (ETM2001.img) you used for Labs 4 and 5. So you do not need to copy them to here.
2. Atmospheric corrections
In lectures 1-5, we talked a lot about the interactions between atmosphere and EMR, and we also talked about the radiance of the remote sensor records include two parts: one from the target area (which is what we want), the other one from the path (path radiance, which is what we do not want). The process to remove the path radiance is called atmospheric correction. There are twp types of atmospheric corrections: (1) absolute atmospheric correction: radiative transfer-based atmospheric correction and empirical line calibration and (2) relative radiometric correction: Dark Object Subtraction (DOS) and multiple-data image normalization using regression. In this class, you want to do a simple DOS correction. The principals of this DOS includes (1) find a dark object in the image, (2) assume that its spectrum should be all zero reflectance, (3) any spectrum observed for the object is assumed to be the atmosphere noise (path radiance), (4) subtract the path radiance from each pixel of the image. Usually these dark objects are water bodies (see below figure, the fresh water has very no reflectance, meaning water absorbing most of light, especially after 0.75 µm).
In ENVI, the DOC is called Dark Subtract, you can find it from Basic Tool -> Preprocessing -> General Purpose Utilities -> Dark Subtract -> select the ETM2001 image that you used before as the input -> select the Band Minimum, which means that the minimum value of each band will be automatically selected, and then this value will be subtracted from all pixels in this band. Output the new image to Memory or MyWork directory. In this image, the darkest objects should be the water bodies in lakes in the southern San Antonio. After this correction, the resulting image is ready for the following steps.
Question 1, calculate and show the basic statistics of the images before and after DOC and make simple comparison (hint, from the ENVI Main menu, Basic Tools -> Statistic -> Computer Statistics)
3. Spectra radiance calculation
Equation 1 is the basic equation for calculating spectral radiance for band l (Ll) from the Digital Number (DN) of Landsat 4, 5 and 7:
(1)
where, DN is the Digital Number of each pixel in the image, LMAX and LMIN are the calibration constants, and QCALMAX and QCALMIN are the highest and the lowest points of the range of rescaled radiance in DN.
3.1 For Landsat 4 and 5, the QCALMAX is 255 and the QCALMIN is zero
Table 1. LMAX and LMIN values for Landsat 4 and 5 TM (After Markham and Barker, 1986)
3.2. For Landsat 7, QCALMAX in the Equation 1 is 255 and the QCALMIN is 0 or 1 depending on what type of product generating system is applied by the USGS. QCALMIN is 0 in NLAPS and 1 in LPGS, where NLAPS and LPGS are two different USGS Landsat 7 Level 1 product generation systems (for more detail, See the Chap.11 and 12 of Landsat 7 Science User Data Handbook, 2002). The image that we used in Lab 4 is a subset of image downloaded from here http://www.fri.sfasu.edu/data/sensor/l7/p027/r040/y2001/d202/. When you connect to this link, you will see a directory called “nlaps”, which means that the data we used is in NLAPS format. SO the QCALMIN here should be 0.
The Table 2 shows the values of LMAX and LMIN for LANDSAT 7 ETM+, for Equation 1. Similar to the Landsat 5 TM,
Table 2. LMAX and LMIN values for Landsat 7 ETM+ (Landsat 7 Science User Data Handbook Chap.11, 2002)
The unit is the same as you saw in the lecture or book: W m-2 sr-1 um-1
The gain combination for the particular image must be clearly known when selecting the LMIN and LMAX values in Table 1. Usually for Land, band 1, 2, 3, 5, 6, 7 are high gain, 4 and 8 low gain (see more from here: http://landsat7.usgs.gov/documents/gain_setting_rules.doc). In San Antonio, we can use this combination.
3.3 For Landsat 7, However, there is a more simple way for calculating Ll (Landsat 7 Science User Data Handbook Chap.11, 2002). This is what we will use in this lab.
(2)
In Equation 2, the “gain” corresponds to the “Gain” in the header file, and the “offset” corresponds to the “Bias” in the header file. The unit is the same as you saw in the lecture or book: W m-2 sr-1 um-1
If you click http://www.fri.sfasu.edu/data/sensor/l7/p027/r040/y2001/d202/ and then click “nlaps”, then click LE7027040000120250.H1. This is the header file for Band1, 2, 3, 4, 5, 7. LE7027040000120250.H2 is the head file for band 6. band 6 has always high gain and low gain images. LE7027040000120250.H3 is the header file for band 8 image. Below is the table showing all gain and offset for each band from those header files. So if you can not open those header files (since the Hurricane Rita), You will just use the numbers in this table below.
Band | Ref DN to Radiance Default
| Detector gain offset Abs Calib?
-------------------------------------------------------
1 | 15 0.775686 -6.20000 FALSE
2 | 12 0.795686 -6.39999 FALSE
3 | 8 0.619216 -5.00000 FALSE
4 | 7 0.965490 -5.10001 FALSE
5 | 14 0.125725 -0.99999 FALSE
6 | 8 0.066823 0.000000 FALSE
7 | 10 0.043726 -0.35000 FALSE
8 | 27 0.971765 -4.70000 FALSE
9 | 8 0.037059 3.200000 FALSE
In this lab, we only do band 3 and band 4. You can use the Band Math tool to do so. from the main ENVI menu, click Basic Tools -> Band Math, you can type the equation for band 3 as the figure below, and click OK. A new window will popup for you to select the atmospheric corrected band 3 as b3. then save this image to your MyWork directory. In the similar way, you can do the band 4.
Question 2, calculate and show the basic statistics of the two radiance images, make some comparisons
4. Spectra Reflectance calculation
The reflectance for band l is computed by the following equation (Markham and Barker,1986 and Landsat 7 Science User Data Handbook Chap.11, 2002):
(3)
where Ll is at satellite spectral radiance which is the outgoing radiation energy of the band observed at the top of atmosphere by the satellite (in this lab, we use the results calculated from step 3), d is the Earth-Sun distance in astronomical units, ESUNl is mean solar exoatmospheric irradiances for the band l, and cosq is the cosine of the solar incident angle. Supposing a horizontal land surface is flat, the cosine of solar incident angle (cosq) can be calculated from the Sun Elevation cos(90-SunElevation). The Sun elevation angle for the image is 64.36º (you can get this from the head file of LE7027040000120250.H1 mentioned above)
Since the inverse of d2 (which is 1/d2) in Equation 3 is equivalent to “inverse squared relative distance Earth-Sun, dr“, the Equation 3 can be rewritten as:
(4)
The annual averaged value of dr is 1.0, and it ranges from about 0.97 to 1.03. You can find a real number for a special date (such as the 202 day: July 21 for this image used) from Table 11.4 of this link here at: http://ltpwww.gsfc.nasa.gov/IAS/handbook/handbook_htmls/chapter11/chapter11.html
The values for ESUNl in Equation 4 are given in Table 3. The value of ESUNl for band 6 is not available.
Table 3. ESUNl for Landsat 4 and 5 TM in mW/cm2/μm (Markham and Barker, 1986), and for Landsat 7 ETM+ in W/m2/μm (Landsat 7 Science User Data Handbook Chap.11, 2002)
In this lab, we only calculate the reflectance of band 3 and band 4, using the Band Math tool.
Question 3
Calculate and compare the basic statistics of reflectance Band 3 and 4.
5. Calculate NDVI
Now you can calculate the NDVI: (b4-b3)/(b4+b3) using the reflectance band 3 and band 4 as input.
Question 4
Calculate the basic statistics of NDVI. Open the NDVI image, using Density Slice (from the Main window), you will be able to see the distribution of the NDVI for the area. Please make a simple discussion about the spatial distribution of the NDVI values. Do you see any linkage between NDVI and vegetation coverage (or density)? You may link the 1 meter orthoimage with the NDVI image (30 meters), then you can see more details about the possible linkage between them. If you can add this into your discussion, you will be more than welcomed!