Riparian TMS 1996-1997 Metadata
DATASET NAME: SP_RIP97
Identification_Information:
Citation:
Citation_Information:
Originator: Bob Czaja,
USDA-ARS-SWRC
Address: 2000 East Allen Road
City: Tucson
State_or_Province: Arizona
Postal_Code: 85719-1596
Phone_Number:
Email_Address:
Publication_Date:
Title:
Description:
Abstract:
Purpose:
Time_Period_of_Content:
Time_Period_Information:
Single_Date/Time:
Calendar_Date:
Currentness_Reference:
Supplemental_Information:
Data_Type: GRID
Software: Arc/INFO 7.2
Operating_System: Unix
Path_Name: \SPDB\Coverages\sp_rip97
Reviews_Applied_to_Date:
Related_Spatial_Data_Files:
References_Cited:
Notes:
Cloud_Cover:
Status:
Progress:
Maintenance_and_Update_Frequency: As needed
Spatial_Domain:
Bounding_Coordinates:
West_Bounding_Coordinate: 569728.0959031
East_Bounding_Coordinate: 585859.0959031
North_Bounding_Coordinate: 3533380.171697
South_Bounding_Coordinate: 3471553.171697
Keywords:
Theme:
Theme_Keyword_Thesaurus:
Theme_Keyword:
Access_Constraints:
Use_Constraints:
Data_Provider_Information:
Contact_Information:
Contact_Organization_Primary:
Contact_Organization: EPA NERL Las Vegas
Contact_Person: Bill Kepner
Contact_Address:
Address_Type:
Address: P.O. Box 93478
City: Las Vegas
State_or_Province: Nevada
Postal_Code: 89193-3478
Contact_Voice_Telephone: 702 798 2193
Contact_Email_Address: kepner.william@epamail.epa.gov
Data_Processor_Information:
Contact_Information:
Contact_Organization_Primary:
Contact_Organization: USDA-ARS-SWRC
Contact_Person: Bob Czaja
Contact_Address:
Address_Type:
Address: 2000 East Allen Road
City: Tucson
State_or_Province: Arizona
Postal_Code: 85719-1596
Contact_Voice_Telephone: 520 670 6381 x162
Contact_Email_Address: czaja@tucson.ars.ag.gov
Metadata_Reference_Information:
Metadata_Date:
Metadata_Contact:
Contact_Information:
Contact_Organization: USDA-ARS-SWRC
Contact_Person: Soren Scott
Contact_Address:
Address_Type:
Address: 2000 East Allen Road
City: Tucson
State_or_Province: Arizona
Postal_Code: 85719-1596
Contact_Voice_Telephone:
Contact_Email_Address: sscott@tucson.ars.ag.gov
Metadata_Standard_Version:
Data_Quality_Information:
Attribute_Accuracy:
Attribute_Accuracy_Report: Not available
Attribute_Accuracy_Value:
Attribute_Accuracy_Explanation:
Logical_Consistency_Report:
Completeness_Report:
Positional_Accuracy:
Horizontal_Positional_Accuracy:
Vertical_Positional_Accuracy:
Lineage:
Process_Step:
Process_Description:
San Pedro Vegetation Classification
The following is a description of how the vegetation classification of the San
Pedro riparian corridor was accomplished.
The classification was accomplished using a combination of Thematic Mapper
Simulator (TMS) imagery from JPL, and high resolution color infrared photography
(CIR)from USDA ARS Weslaco Tx, supported by extensive ground truth. The TMS
imagery was from two dates. The first date was 29 May 96 with 5 meter
resolution , and the second date was 12 Aug 97 with 3 meter resolution. The CIR
was flown on 13 Aug 97 and had resolution of approximately 3 inches. Two TMS
and one CIR collections were used because all three were required to acquire the
entire area, The CIR was also used to fine tune the classification of riparian
woodland and mesquite.
The first step in the classification of the images was to correct for the TMS
wide scan angle distortion which caused a pixel size difference and to
georectify the images to UTM, Clark 1866, NAD 27. The image was projected into
UTM, Clark 1866, NAD 83 for the San Pedro Spatial Data Archive.
This classification extends from the U.S. Mexican border to a point two miles
north of St. David and includes the area on either side of the river that was
identified as the riparian strand. This identification was done by using the
USGS 1 arc second digital elevation models (DEM) of the area and by visual
interpretation of high resolution color infrared photography. The combination
of the two techniques resulted in a corridor definition that was superior to
attempts that used only one or the other data set. The approach was to select
the corridor based on 6 to 9 meter contour change from the river bottom based on
the DEM. This area was then checked against high resolution CIR photography.
The area was then adjusted to more closely fit the riparian corridor based on
photo interpretation of vegetation and terrain features. The resultant corridor
boundary was then saved and used as a mask to subset the desired area from the
classification. This technique is consistent with the methods recommended by
the U.S. Fish and Wildlife Service for defining riparian corridors.
TMS data has 10 reflected bands and 2 thermal bands. The two thermal bands have
the same band width but band 11 is low gain and band 12 is high gain. The two
data sets used here had good separation in band 11, but band 12 was somewhat
saturated, the decision was therefore made to eliminate band 12 from the data
sets. An NDVI (bands 7, 5) was done on each of the TMS images as well as an
NDSVI (bands 9, 5)and a soil enhancement (bands 10, 11). These indices were
then stacked with the other 11 TMS bands to provide a data set of 14 bands for
classification.
The vegetation classification was accomplished using established techniques
utilizing both spatial and spectral analysis. For the spectral analysis,
training sets were identified by using ground truth sites, photo truth sites,
and feature space placement for the red and NIR bands. It was necessary to
select over 100 training sites to achieve an accurate classification due to
different soil backgrounds, vegetation health, cultural activity, and cloud
shadow. Care was taken to select sites of known vegetation that exhibited
slightly different spectral properties due to shadow soil background or cultural
activity (such as farming) even though they represented the same vegetation
class. This was done in order to get the most accurate spectral classification
possible. Due to the variations in imagery (especially cloud shadow) the images
were subset into smaller areas before classification. Some of these subsets
were done to separate areas of bright sunlight from areas of cloud shadow which
required all new training sets. Other subsets were the result of spatial
analysis and were done to eliminate signatures from areas where they did not
belong. For example the agriculture signature was removed from areas on the
river where there was no agriculture but lush vegetation was classified as
agriculture. Other subsets were extracted to eliminate riparian signatures from
homogeneous mesquite stands, and to eliminate mesquite signatures from the
cottonwood willow gallery of the riparian woodland where they often accrued due
to shadowing caused by the tree canopy. The classifications were run as
supervised classifications using feature space and maximum likelihood. Once the
classifications were complete they were mosaiced together using the last overlay
method.
A raster file of all mesquite pixels was then broken down into either low or
high density based on the following procedure. The raster layer was imported
into a grid file for ArcInfo, then a window of 3X3 pixels was passed over the
file. If less than 75% of the pixels were mesquite then the center pixel, if
mesquite, was classified as mesquite low density, if 75% or more of the pixels
were mesquite then the center pixel, if mesquite, was classified as mesquite
high density. If the center pixel was not mesquite then was not changed. This
layer was then mosaiced with its parent classification. The result is a
classification where mesquite now is either high or low density based on spatial
properties.
The final step in the classification process was to extensively compare the
classification with the CIR and make corrections where necessary. At this point
the riparian wood, and mesquite were fine tuned using photo interpretation
techniques. This combination of spectral and spatial analysis took advantage of
the best of two worlds to produce a classification whose accuracy is outlined
below.
Classification Information
1. The classification resolution is 3 meters.
2. The classes are Riparian Wood, Mesquite low density, Mesquite high density,
Sacaton, Scrub, Bare Soil, Agriculture, and Water.
3. The accuracy of the classification was assessed using photo truth of all the
riparian wood, areas, water areas and mesquite areas, as well as, many sacaton,
scrub, and agricultural areas. Ground truth was carried out from Fairbanks to
Hereford at 15 sites that each contained two or more of the classification
signatures. The results of this ground truth were then used to correct errors
in the classification.
The error of the classification is estimated to be no more than 1% to 2% for
riparian wood, 2% to 5% for mesquite, and less than 10% for sacaton. The error
of the other classes (agriculture, water, scrub, and bare soil) is estimated to
be 10%. Agriculture refers only to irrigated crops that are growing. Fallow
fields classified either as scrub or bare soil depending on their condition.
These estimates are based on the aforementioned ground truth and extensive photo
truth. The higher error associated with the mesquite is due to the transition
from Chihuahua scrub with shrub mesquite to mesquite wood. The higher error
value associated with sacaton is the result of the transition between sacaton
and scrub and the difficulty is visually identifying this boundary with
photography.
4. Classification areas rounded to the nearest hectare/acre:
From Palominos to a point two miles north of St. David:
Riparian Vegetation Classes Hectares Acres
Riparian Wood 615 1521
Mesquite low 826 2042
Mesquite high 1474 3642
Sacaton 390 965
From the U.S. Mexico border to Palominos (this was done with visual
interpretation as only the CIR was available for this area):
Riparian Vegetation Hectacres Acres
Riparian Wood 20 50
Mesquite 19 47
Sacaton 10 26
Total 49 123
The figures for the total area are:
Hectacres Acres
Riparian Wood 636 1572
Mesquite 2319 5731
Sacaton 401 991
Agriculture* 28.3 70
Bare Soil 70.3 173.9
Scrub 2090.8 5166.6
Water 4.7 11.5
Unclassified** 554 1366
Total Area 6104 15082
*This does not include fallow fields. Plowed fallow fields classified as bare
soil, and old fallow fields classified as scrub. Most of the irrigated crop
area falls outside of the riparian strand. **This is the area outside the TMS
coverage that does not fall into one of the three riparian classes.
5. Location of files:
The TMS images and classification from 96 are on CD numbers 33 and 34. The TMS
images, classification, DEM, riparian corridor vector, and classification are on
CD numbers 68 and 72. The CIR photo are stored in the Remote Sensing Group.
All of these images and data sets are at USDA ARS, 2000 E. Allen St, Tucson, AZ.
6. POC for this project is Dr. Dave Goodrich, USDA ARS, (520) 670-6381 ext.
175.
Process_Date:
Process_Time:
SPATIAL REFERENCE INFORMATION
Grid Coordinate System: UTM
Horizontal Datum: NAD83
Zone: 12
Units: METERS
Scale:
Cell size (x,y): 3, 3
SPATIAL DATA ORGANIZATION INFORMATION
Number of Columns: 5377
Number of Rows: 20609
Grid Type: 1
Number of Classes: 9
Minimum Value: 0
Maximum Value: 11
ENTITY AND ATTRIBUTE INFORMATION
ATTRIBUTE LISTING FOR: SP_RIP97.BND
| COLUMN |
ITEM NAME |
WIDTH |
OUTPUT |
TYPE |
N.DEC |
ALTERNATE NAME |
INDEXED? |
| 1 |
XMIN |
8 |
18 |
F |
5 |
|
- |
| 9 |
YMIN |
8 |
18 |
F |
5 |
|
- |
| 17 |
XMAX |
8 |
18 |
F |
5 |
|
- |
| 25 |
YMAX |
8 |
18 |
F |
5 |
|
- |
ATTRIBUTE LISTING FOR: SP_RIP97.STA
| COLUMN |
ITEM NAME |
WIDTH |
OUTPUT |
TYPE |
N.DEC |
ALTERNATE NAME |
INDEXED? |
| 1 |
MIN |
8 |
15 |
F |
|
| |
|
- |
| 9 |
MAX |
8 |
15 |
F |
|
| |
|
- |
| 17 |
MEAN |
8 |
15 |
F |
|
| |
|
- |
| 25 |
STDV |
8 |
15 |
F |
|
| |
|
- |
ATTRIBUTE LISTING FOR: SP_RIP97.VAT
| COLUMN |
ITEM NAME |
WIDTH |
OUTPUT |
TYPE |
N.DEC |
ALTERNATE NAME |
INDEXED? |
| 1 |
VALUE |
4 |
10 |
B |
- |
|
Indexed |
| 5 |
COUNT |
4 |
10 |
B |
- |
|
- |
| 9 |
RED |
8 |
12 |
F |
|
| |
|
- |
| 17 |
GREEN |
8 |
12 |
F |
|
| |
|
- |
| 25 |
BLUE |
8 |
12 |
F |
|
| |
|
- |
| 33 |
CLASS_NAMES |
128 |
128 |
C |
0 |
|
- |
| 161 |
OPACITY |
8 |
12 |
F |
|
| |
|
- |
| 169 |
AREA_HECTARES |
8 |
12 |
F |
|
| |
|
- |
| 177 |
AREA_ACRES |
8 |
12 |
F |
|
| |
|
- |
DOWNLOAD sp_riparian97.avl (SP_RIP97 ArcView Legend)
![[logo] US EPA](../gif/logo_epaseal.gif){kind=logo}