Land Cover Classification in the SGP97 Region

The Data Procedures Specifications Accuracy The Files Filenames and Formats Data Access and Contacts FTP Site Points of Contact

Landcover classification Data

Landcover Map

Vegetation page

 

The Data

Procedure for Creating Oklahoma TM Classification

1. Ground truth in and around the ARM site, El Reno and Little Washita Basin were collected.
2. Ground truth was converted to AOI's and 1/3 were set aside for verification purposes.
3. Landsat TM data for 4 dates: April 20, June 7, July 9^th and July 25 were collected and imported to ERDAS for Path 28/Row 34-36.
4. Landsat TM images were masked for clouds. For April 20, June 7^th and July 9^th the pixels with values in band 1 (.45 - .52 um ) greater than 100 were removed. On July 25^th, band 1 greater than 100 and band 5 ( 1.55 - 1.75 um) less than 175 were removed.
5. Shadowed areas; Band 4 ( .76 -.90um ) less than 75, band 5 values between 25 and 50 and values less than 40 in band 3 were removed.
6. Bands 3,4,5,7 were extracted from April 20, June 7^th and July 9^th and merged to create a 12 band image.
7. Training classifications within AOI's used ISODATA reiterations. Means were initialized along diagonal axes and standard deviations were set to one Maximum, iterations were set to 24, convergence thresholds to .995. For wheat and pasture, convergence thresholds were set to .95.
8. Signatures based on AOI's were combined into one large signature file and used within supervised classification using maximum likelihood
9. Using a 5 band (2,3,4,5,7) image for July 25^th (the least cloudy image) a second classification was created using the same methods as described in 7-8.
10. Areas that were masked because of clouds or shadows in the original classified image were replaced by the July 25^th derived classified image creating the merged classification.
11. An 8 band image (3,4,5,7) for July 9^th and July 25^th was created and classified using the same methods as described in 7-8 to create a summer classification.
12. Wheat areas in the merged classification were replaced if they coincided with legumes or corn in the summer classification. Otherwise the crops were maintained. This was to give a better representation of what was on the ground during the study time period.
13. Accuracies were calculated using the AOI's that had not been used in the classification process.
14. After the initial classification, while the overall accuracies were good, it was noted that in certain areas of the study sites, the separation of tree categories were misclassified and a reclassification of the urban or soil categories was conducted. The final classification has an acceptable overall classification and avoids the problems of misclassifications in the study areas.

Specifications

Specifications of the classified image: Format Rows: 17720 Columns: 9840 Classifications Row Class_Names 0 Unclassified 1 Alfalfa 2 Bare (wheat ) 3 Corn 4 Forage 5 Legume 6 Pasture/Rangeland 7 Tree 8 Urban 9 Water 10 Wheat 11 summer Corn 12 summer Legume 13 Shrub Extrema xmin: 464385 xmax: 759585 ymin: 3722115 ymax: 4253715 Projection UTM, Zone 14 Units meters

 

Accuracy

Hectares and percentage correct   Alfalfa Bare Corn Forage Legume Pasture Tree Urban Water Wheat Accuracy Alfalfa 23 0 0 10 0 0 0 0 0 2 66 Bare 1 1 0 0 0 0 0 1 0 11 79 Corn 0 0 14 1 8 3 0 0 0 2 50 Forage 7 0 4 32 2 24 0 2 0 4 43 Legume 0 1 0 0 77 0 0 0 0 0 99 Pasture 28 4 12 43 4 483 10 13 1 20 78 Tree 0 0 0 0 0 2 37 0 0 0 95 Urban 0 0 0 0 0 0 0 0 0 0   Water 0 0 0 0 0 0 0 0 0 0   Wheat 43 2 0 10 28 22 0 1 0 339 76 User accu 23 13 47 33 65 90 79 0 0 90 75

 

The Files

File Name and Format Information

File name File type Format Description File size Storage required okla-class.bil binary BYTEORDER M LAYOUT BIL NROWS 17720 NCOLS 9840 NBANDS 1 NBITS 8 BANDROWBYTES 9840 TOTALROWBYTES 9840 BANDGAPBYTES 0 Landuse/Landcover classification in the SGP 97 study area ~166.3MB ~166.5MB okla-class.gif Image gif gif file displaying the binary file ~85KB okla-class.hdr Text ascii Formatting information for .bil file NA oklarev.rtf Document Rich text format explanation of the procedures used to create the classification as well as class description and accuracy assesment. ~84KB oklarev.wpd Word Perfect ~45KB Note: File okla-class.hdr contains the formatting information displayed above.

 

Data Access and Contacts

FTP Site

The SGP97 Landcover Classification data is in the following GES DISC ftp site:

Landcover classification Data

Points of Contact

The principal investigator for the land classification data set is

Paul Doraiswamy

USDA/Remote Sensing and Modeling Lab

Beltsville, MD 20705

E-mail: pdoraisw@asrr.arsusda.gov

Voice: 301-504-6576

Fax: 301-504-5031

For information about or assistance in using DAAC data, contact

Hydrology Data Support Team

Goddard Earth Sciences
Data and Information Services Center (GES DISC)

Code 610.2

NASA Goddard Space Flight Center

Greenbelt, Maryland 20771

 

E-mail: hydrology-disc@listserv.gsfc.nasa.gov

Voice: 301-614-5165

Fax: 301-614-5268

---

Last update:Thu Jan 28 09:34:12 EST 1999 Page Author: Hydrology Data Support Team -- hydrology-disc@listserv.gsfc.nasa.gov Web Curator: Anthony Drake NASA official: Steve Kempler, DAAC Manager -- Steven.J.Kempler@nasa.gov