eCognition Method for Classifying Sagebrush and Land Cover from Landsat TM Scene P37/R29

Metadata:


Identification_Information:
Citation:
Citation_Information:
Originator: Wildlife Spatial Analysis Lab, The University of Montana
Publication_Date: 20021202
Title:
eCognition Method for Classifying Sagebrush and Land Cover from Landsat TM Scene P37/R29
Publication_Information:
Publication_Place: Missoula, Montana
Publisher: Wildlife Spatial Analysis Lab, The University of Montana
Description:
Abstract:
BLM3729ECOG is an ArcInfo 15 meter region grouped grid (raster file) of ground feature polygons derived from the 09/08/1999 P37/R29 Landsat 7 Thematic Mapper (TM) image using the eCognition method. An image segmentation and supervised classification of land cover types with sagebrush / xeric shrub canopy cover classes was performed as part of a sagebrush mapping comparison for the Bureau of Land Management.

Two image segmentation techniques and a pixel classification were compared for mapping sagebrush in southern Montana. The SILC3 method used an image segmentation based on a 30m unsupervised classification of a July 2000 Landsat image. The eCognition method used an image segmentation based on a 15m bottom up region merging algorithm of a September 1999 Landsat image. The modified USFS Region Four principal components analysis (PCA) method used a pixel classification based on a 15m isodata analysis of the first 3 principle components of a September 1999 Landsat image. A study area was defined for the project based on the intersection of the July 2000 and September 1999 Landsat imagery. To reduce spectral and ecological differences, the study area was additionally restricted to the intersection of the two Landsat images within Montana and 24 miles south of the Montana border in Wyoming. Below is a detailed description of how the BLM3729ECOG grid was created using the eCognition image segmentation method.

BLM3729ECOG Land Cover Type Classification using Sagebrush Canopy Classes
Covertype  Acres   Description
------------------------------------------------
1100      3937.1  Urban or Developed Lands
2010     82499.4  Agricultural - Dry
2020    171705.7  Agricultural - Irrigated
3130    542839.6  Very Low Cover Grasslands
3150    368958.0  Low / Moderate Cover Grasslands
3170     43581.6  Moderate / High Cover Grasslands
3370    416683.7  Sagebrush / Xeric Shrubs 05-14% Cover
3380    437902.4  Sagebrush / Xeric Shrubs 15-24% Cover
3390    130989.9  Sagebrush / Xeric Shrubs 25-34% Cover
3395     63706.3  Sagebrush / Xeric Shrubs >= 35% Cover
3610    134172.8  Mesic Shrublands / Willow
4101     18069.6  Aspen
4150     32303.1  Mixed Broadleaf / Cottonwood Forest
4203     29811.1  Lodgepole Pine
4204    108338.3  Whitebark Pine
4205     81662.1  Limber Pine
4206    38666.5  Ponderosa Pine
4212    143962.2  Douglas-fir
4216    175274.3  Utah Juniper
4223      5185.4  Douglas-fir/Lodgepole Pine
4237    115601.7  Subalpine Fir/Spruce
4241     74013.2  Mixed Upper Subalpine Conifer Forest
4242     31861.2  Mixed Lower Subalpine Conifer Forest
4244     25927.0  Mixed Xeric Conifer Forest
4400    342704.2  Burns
5000     30281.1  Water
7300    325728.8  Rock/Barren
7500       484.8  Mines/Quarries
9100     35524.5  Snow

BLM3729ECOG Sagebrush Species Classes (for regions classified as a sagebrush type above) Covertype Acres Description ------------------------------------------------ 3311 35802.5 Greasewood 3351 28977.4 Mountain Big Sagebrush 3352 861631.5 Wyoming Big Sagebrush 3353 9605.0 Basin Big Sagebrush 3354 113265.9 Black / Low Sagebrush

First, the September 1999 Landsat TM data was clipped to the study area and then resampled to 15m using the panchromatic band. Next an NDVI layer was created using TM bands 3 and 4:

NDVI = (TM4 - TM3) / (TM4 + TM3)

An image segmentation was run in eCognition using a bottom up region growing algorithm. First the TM band and NDVI layers were loaded into the eCognition program. Segmentation settings for scale and color/shape contrast were selected and image segmentations were created. Ground feature delineations were analyzed using DOQ's and the original imagery. Additional segmentation settings for shape features are also used. Lower scale factors resulted in smaller minimum map units. Setting the color/shape contrast to favor pixel color separation verses region shape resulted in more linear feature delineation. This was particularly important in delineating draws with high canopy cover sagebrush. A scale factor of 10 and a color/shape contrast of 0.8 to 0.2 were selected for the final image segmentation. The image segmentation was exported from eCognition and imported into an ARC/INFO region grid. Each region was assigned a set of attributes for TM channels 1-7 and panchromatic from the imagery, along with elevation, aspect, and slope from the 30 meter digital elevation model (DEM); these attribute values were calculated based on the mean values of the 30 meter pixels within each region.

Vegetation overstory (land cover type) was labeled using a supervised classification. The land cover type supervised classification was performed using a combination of manual and training data classifications. The manual classification was performed by on screen analysis of the imagery and directly assigning cover type labels. The labels were filled into the MANLABEL attribute. The following classes were manually labeled:

COVERTYPE  DESCRIPTION
------------------------------------------------
1100       Urban or Developed Lands
2010       Agricultural - Dry
2020       Agricultural - Irrigated
4400       Burns
5000       Water
7500       Mines/Quarries

Twenty One land cover types were labeled using a training data classification. Canopy cover classes were used for both grassland and sagebrush/xeric shrub types because canopy classes provided significantly better separation between the two lifeforms than species classes did. A euclidian distance classifier based on TM band data and topographic variables with a nearest neighborhood of 15 was used in conjunction with a mean inverse spatial classifier to assign labels to the regions. The labels were filled into the COV_CODE_1 ATTRIBUTE. The following classes were used in the training data classification:

COVERTYPE  DESCRIPTION
------------------------------------------------
3130       Very Low Cover Grasslands
3150       Low / Moderate Cover Grasslands
3170       Moderate / High Cover Grasslands
3370       Sagebrush / Xeric Shrubs 05-14% Cover
3380       Sagebrush / Xeric Shrubs 15-24% Cover
3390       Sagebrush / Xeric Shrubs 25-34% Cover
3395       Sagebrush / Xeric Shrubs >= 35% Cover
3610       Mesic Shrublands / Willow
4101       Aspen
4150       Mixed Broadleaf / Cottonwood Forest
4203       Lodgepole Pine
4204       Whitebark Pine
4205       Limber Pine
4206       Ponderosa Pine
4212       Douglas-fir
4216       Utah Juniper
4223       Douglas-fir/Lodgepole Pine
4237       Subalpine Fir/Spruce
4240       Mixed Conifer Forest
7300       Rock/Barren
9100       Snow

After the training data classification, regions labeled as Mixed Conifer Forest (4240) were assigned to one of three mixed conifer types using a ruleset based on elevation and PVT type for the region (see mixcon_recode.aml).

4241       Mixed Upper Subalpine Conifer Forest
4242       Mixed Lower Subalpine Conifer Forest
4244       Mixed Xeric Conifer Forest

A second training data classification was used to assign sagebrush species to regions classified as sagebrush/xeric shrub types in the first train data classification. A euclidian distance classifier based on TM band data and topographic variables with a nearest neighborhood of 15 was used in conjunction with a mean inverse spatial classifier to assign labels to the regions. The labels were filled into the SAGESPP1 ATTRIBUTE. The following classes were used in the sagebrush species classification:

SAGESPECIES  DESCRIPTION
------------------------------------------------
3311         Greasewood
3351         Mountain Big Sagebrush
3352         Wyoming Big Sagebrush
3353         Basin Big Sagebrush
3354         Black / Low Sagebrush

An accuracy assessment was performed on the land cover type and sagebrush species labels assigned from the training data classifications. A cross validation (leave-one-out) accuracy assessment was performed using the entire training data set.

To help track similarities and differences in map legends between this classification and previous ones (SILC1 and MT-GAP) cover type names and codes numbers were not changed if the actual types were mapped using similar training data.

Purpose:
This land cover classification of P37/R29 was prepared for a sagebrush mapping comparison project under contract for the Montan State office of the Bureau of Land Management. These data were produced to delineate ground features and map existing land cover in a standardized, consistent manner across the project area. This land cover grid is suited for analysis at the regional, sub-regional, and landscape levels; it can also provide support for many management disciplines, including timber, wildlife, fisheries, and recreation.
Time_Period_of_Content:
Time_Period_Information:
Single_Date/Time:
Calendar_Date: 20021202
Currentness_Reference: calendar date
Status:
Progress: Complete
Maintenance_and_Update_Frequency: None Scheduled
Spatial_Domain:
Bounding_Coordinates:
West_Bounding_Coordinate: -110.20289889
East_Bounding_Coordinate: -107.57261354
North_Bounding_Coordinate: 45.54039749
South_Bounding_Coordinate: 44.62271160
Keywords:
Theme:
Theme_Keyword_Thesaurus: none
Theme_Keyword: vegetation
Theme_Keyword: land cover
Theme_Keyword: classification
Theme_Keyword: Landsat Thematic Mapper scenes
Theme_Keyword: remote-sensing image
Place:
Place_Keyword_Thesaurus: Geographic Names Information System
Place_Keyword: Montana
Access_Constraints:
This data set is in the public domain, and the recipient may not assert any proprietary rights thereto nor represent it to anyone as other than a data set produced by the Bureau of Land Management and the Wildlife Spatial Analysis Lab at the University of Montana. COPYRIGHT (c) Copyright The University of Montana, 2002. This Arc Info dataset is copyrighted by the University of Montana.
Use_Constraints:
This data set is provided "as-is" without warranty of any kind, including, but not limited to, the implied warranties of merchantability and fitness for a particular purpose. The user assumes all responsibility for the accuracy and suitability of this data set for a specific application. In no event will the creators, the University of Montana, or the Bureau of Land Management, be liable for any damages, including lost profits, lost savings, or other incidental or consequential damages arising from the use of or inability to use this data set. Use of these data requires the ability to read ArcInfo Grid data sets. Users must assume responsibility for determining the suitability of these data for their purposes. Not for use at scales greater than 1:100000.
Point_of_Contact:
Contact_Information:
Contact_Organization_Primary:
Contact_Organization: Wildlife Spatial Analysis Lab, The University of Montana
Contact_Person: Roland L. Redmond
Contact_Position: Principal Investigator
Contact_Address:
Address_Type: mailing and physical address
Address: Wildlife Spatial Analysis Lab, The University of Montana
City: Missoula
State_or_Province: MT
Postal_Code: 59812-1063
Country: USA
Contact_Voice_Telephone: 406 243 5208 (email preferred)
Contact_Facsimile_Telephone: 406 243 6064
Contact_Electronic_Mail_Address: red@wru.umt.,edu
Hours_of_Service: Monday-Friday, 8-5, Mountain Time
Data_Set_Credit:
The Wildlife Spatial Analysis Lab for creation of the geospatial data set.
Native_Data_Set_Environment:
The Wildlife Spatial Analysis Lab uses IBM RS/6000 Workstations running AIX 4.3.3 with ArcInfo software versions 8.2.0, and Imagine version 8.5.


Data_Quality_Information:
Attribute_Accuracy:
Attribute_Accuracy_Report:
Thematic accuracy of the land cover type and sagebrush species cover labels were assessed using a leave-one-out cross-validation. The cross-validation involved removing the first training observation from the training set and constructing a new classification rule from the reduced training set. This new rule was then used to classify the polygon containing the left-out training observation. The process was repeated until all observations were held out once and used as a singleton test set. The estimated accuracy rate was the percentage of (held-out) observations correctly classified. The manually classified land cover types did not have an accuracy assessment performed. The Mixed Conifer Forest types were classified as one group (4240 Mixed Conifer), and were then recoded into 3 types; 4241 Mixed Upper Subalpine Conifer Forest, 4242 Mixed Lower Subalpine Conifer Forest, 4244 Mixed Xeric Conifer Forest. The accuracy of 4240 Mixed Conifer type was used to report accuracies for the 3 recoded Mixed Conifer types.

Users Accuracies of Land Cover Types were:
Acc%  Covertype  Definition
-------------------------------------------
82.5  3130       Very Low Cover Grasslands
81.7  3150       Low / Moderate Cover Grasslands
80.2  3170       Moderate / High Cover Grasslands
56.2  3370       Sagebrush / Xeric Shrubs 05-14% Cover
52.7  3380       Sagebrush / Xeric Shrubs 15-24% Cover
50.5  3390       Sagebrush / Xeric Shrubs 25-34% Cover
84.0  3395       Sagebrush / Xeric Shrubs >= 35% Cover
74.4  3610       Mesic Shrublands / Willow
68.0  4101       Aspen
74.3  4150       Mixed Broadleaf / Cottonwood Forest
70.2  4203       Lodgepole Pine
75.8  4204       Whitebark Pine
67.3  4205       Limber Pine
75.8  4206       Ponderosa Pine
78.1  4212       Douglas-fir
75.7  4216       Utah Juniper
38.4  4223       Douglas-fir/Lodgepole Pine
56.1  4237       Subalpine Fir/Spruce
65.5  4240       Mixed Conifer Forest
99.0  7300       Rock/Barren
100.0 9100       Snow

Users Accuracies of Sagebrush Species were: Acc% SAGESPECIES DESCRIPTION ------------------------------------------------ 76.4 3311 Greasewood 89.8 3351 Mountain Big Sagebrush 97.5 3352 Wyoming Big Sagebrush 33.3 3353 Basin Big Sagebrush 67.8 3354 Black / Low Sagebrush

Logical_Consistency_Report:
All grid attributes were checked for consistency of appropriate range values.
Completeness_Report:
All areas within the BLM3729ECOG grid boundary have attributed regions.
Positional_Accuracy:
Horizontal_Positional_Accuracy:
Horizontal_Positional_Accuracy_Report:
Claimed root-mean square error for horizontal position of the terrain-corrected Landsat TM images is 18 meters in the x direction (WNW-ESE) and 30 meters in the y direction (NNE-SSW).
Vertical_Positional_Accuracy:
Vertical_Positional_Accuracy_Report:
Each region has a mean elevation in meters determined from Digital Elevation Models whose Root Mean Square accuracy varies from 7 meters to approximately 30 meters.
Lineage:
Source_Information:
Source_Citation:
Citation_Information:
Originator: USGS EROS Data Center
Publication_Date: 19990908
Title: Terrain-corrected Landsat 7 Thematic Mapper Images P37/R29
Geospatial_Data_Presentation_Form: remote-sensing image
Publication_Information:
Publication_Place: Sioux Falls, South Dakota
Publisher: USGS
Source_Scale_Denominator: 60000
Type_of_Source_Media: Online
Source_Time_Period_of_Content:
Time_Period_Information:
Single_Date/Time:
Calendar_Date: 19990908
Source_Currentness_Reference: calendar date
Source_Citation_Abbreviation: EROS
Source_Contribution:
Provided base imagery for classification of Eastside Assessment Area.
Source_Information:
Source_Citation:
Citation_Information:
Originator: USDA Forest Service, Region 1
Publication_Date: 199504
Title: Potential natural vegetation (PVT) classification
Geospatial_Data_Presentation_Form: map
Publication_Information:
Publication_Place: Montana
Publisher: USDA Forest Service
Source_Scale_Denominator: 100000
Type_of_Source_Media: digital
Source_Time_Period_of_Content:
Time_Period_Information:
Multiple_Dates/Times:
Calendar_Date: 199808
Calendar_Date: 199904
Source_Currentness_Reference: calendar date
Source_Citation_Abbreviation: PVT
Source_Contribution:
Provided PVT seed for unsupervised classification of image pixels.
Source_Information:
Source_Citation:
Citation_Information:
Originator: USDA Forest Service, Region 1
Publication_Date: 199504
Title: Existing Ground-Truth Databases and TSMRS Databases
Geospatial_Data_Presentation_Form: map
Publication_Information:
Publication_Place: Montana
Publisher: USDA Forest Service
Source_Scale_Denominator: 24000
Type_of_Source_Media: digital
Source_Time_Period_of_Content:
Time_Period_Information:
Multiple_Dates/Times:
Calendar_Date: 199504
Calendar_Date: 199510
Calendar_Date: 199912
Source_Currentness_Reference: calendar date
Source_Citation_Abbreviation: GRTR
Source_Contribution:
Provided training data and TSMRS data for assigning cover type labels.
Source_Information:
Source_Citation:
Citation_Information:
Originator: USDA Forest Service, Region 4
Publication_Date: 199504
Title: Existing Ground-Truth Databases
Geospatial_Data_Presentation_Form: map
Publication_Information:
Publication_Place: Montana
Publisher: USDA Forest Service
Source_Scale_Denominator: 24000
Type_of_Source_Media: digital
Source_Time_Period_of_Content:
Time_Period_Information:
Multiple_Dates/Times:
Calendar_Date: 199609
Calendar_Date: 199705
Source_Currentness_Reference: calendar date
Source_Citation_Abbreviation: GRTR
Source_Contribution: Provided training data for assigning cover type labels.
Source_Information:
Source_Citation:
Citation_Information:
Originator: U.S. Bureau of Land Management
Publication_Date: 199704
Title: Existing Ground-Truth Databases
Geospatial_Data_Presentation_Form: map
Publication_Information:
Publication_Place: Montana
Publisher: U.S. Bureau of Land Management
Source_Scale_Denominator: 60000
Type_of_Source_Media: digital
Source_Time_Period_of_Content:
Time_Period_Information:
Single_Date/Time:
Calendar_Date: 199704
Source_Currentness_Reference: calendar date
Source_Citation_Abbreviation: GRTR
Source_Contribution: Provided training data for assigning cover type labels.
Source_Information:
Source_Citation:
Citation_Information:
Originator: Natural Resource Conservation Service
Publication_Date: 199706
Title: Existing Ground-Truth Soil Survey Databases
Geospatial_Data_Presentation_Form: map
Publication_Information:
Publication_Place: Montana
Publisher: Natural Resource Conservation Service
Source_Scale_Denominator: 24000
Type_of_Source_Media: digital
Source_Time_Period_of_Content:
Time_Period_Information:
Single_Date/Time:
Calendar_Date: 199706
Source_Currentness_Reference: calendar date
Source_Citation_Abbreviation: GRTR
Source_Contribution:
Provided training data for assigning cover type labels. To protect private landowners, NRCS overlaid WSAL polygons with their points and identified polygons containing ground truth plots; source scale is 1:24,000 or finer.
Source_Information:
Source_Citation:
Citation_Information:
Originator: U.S. Bureau of Indian Affairs
Publication_Date: 199706
Title: Existing Ground-Truth Databases
Geospatial_Data_Presentation_Form: map
Publication_Information:
Publication_Place: Montana
Publisher: U.S. Bureau of Indian Affairs
Source_Scale_Denominator: 24000
Type_of_Source_Media: digital
Source_Time_Period_of_Content:
Time_Period_Information:
Single_Date/Time:
Calendar_Date: 199706
Source_Currentness_Reference: calendar date
Source_Citation_Abbreviation: GRTR
Source_Contribution: Provided training data for assigning cover type labels.
Source_Information:
Source_Citation:
Citation_Information:
Originator: U.S. Geological Survey
Publication_Date: 199603
Title: 1:100,000-scale Digital Line Graphs
Geospatial_Data_Presentation_Form: map
Publication_Information:
Publication_Place: Reston, VA
Publisher: U.S. Geological Survey
Source_Scale_Denominator: 100000
Type_of_Source_Media: digital
Source_Time_Period_of_Content:
Time_Period_Information:
Multiple_Dates/Times:
Calendar_Date: 199603
Calendar_Date: 199704
Source_Currentness_Reference: calendar date
Source_Citation_Abbreviation: HYDR
Source_Contribution: Provided hydrography layer.
Source_Information:
Source_Citation:
Citation_Information:
Originator: U.S. Geological Survey
Publication_Date: 199507
Title: 7.5-minute Digital Elevation Models
Geospatial_Data_Presentation_Form: map
Publication_Information:
Publication_Place: Reston, VA
Publisher: U.S. Geological Survey
Source_Scale_Denominator: 24000
Type_of_Source_Media: digital
Source_Time_Period_of_Content:
Time_Period_Information:
Multiple_Dates/Times:
Calendar_Date: 199507
Calendar_Date: 199609
Source_Currentness_Reference: calendar date
Source_Citation_Abbreviation: DEM
Source_Contribution: Provided majority of topography layer.
Source_Information:
Source_Citation:
Citation_Information:
Originator: Defense Mapping Agency
Publication_Date: 199507
Title: 1-degree Digital Elevation Models
Geospatial_Data_Presentation_Form: map
Publication_Information:
Publication_Place: Bethesda, MD
Publisher: Defense Mapping Agency
Source_Scale_Denominator: 250000
Type_of_Source_Media: digital
Source_Time_Period_of_Content:
Time_Period_Information:
Multiple_Dates/Times:
Calendar_Date: 199507
Calendar_Date: 199609
Source_Currentness_Reference: calendar date
Source_Citation_Abbreviation: DEM
Source_Contribution:
Provided topography layer where 7.5-minute DEM data were unavailable.
Process_Step:
Process_Description:
IMAGESEG: unsupervised classification of input image pixels into regions using eCognition software program. First, the September 1999 Landsat TM data was clipped to the study area and then resampled to 15m using the panchromatic band. Next an NDVI layer was created using TM bands 3 and 4:

NDVI = (TM4 - TM3) / (TM4 + TM3)

An image segmentation was run in eCognition using a bottom up region growing algorithm. First the TM band and NDVI layers were loaded into the eCognition program. Segmentation settings for scale and color/shape contrast were selected and image segmentations were created. Ground feature delineations were analyzed using DOQ's and the original imagery. Different weights were assigned to the TM and NDVI bands which had and effect on the color segmentation limits. Additional region shape limits for region smoothness and compactness shape were also used. The eCognition region growing alogrithm attempts to build the most homogenous region possible based in the segmentation settings entered. Lower scale factors resulted in smaller minimum map units. Setting the color/shape contrast to favor pixel color separation verses region shape resulted in more linear feature delineation. This was particularly important in delineating draws with high canopy cover sagebrush. The final image segmentation used the following settings:

Band Weights                  Scale  Color  Shape  Shape-Additional
TM1 TM2 TM3 TM4 TM5 TM7 NDVI                       (smooth) (compact)
---------------------------------------------------------------------
0.5 0.5 1.0 1.0 1.0 1.0 2.0    10     0.8    0.2     0.9       0.1

The final image segmentation was exported from eCognition and imported into an ARC/INFO region grid. The final region grid had 549,246 regions with an average size of 9.0 acres.

Source_Used_Citation_Abbreviation: MRLC, EROS, PVT
Process_Date: 20020615
Source_Produced_Citation_Abbreviation: IMSEG
Process_Step:
Process_Description:
MAKE-DEM: extracted a digital elevation model (DEM) from the WSAL database for use during labeling.

MAKE-HYDROGRAPHY: extracted a hydrography coverage from the WSAL hydrography database for use during labeling.

ATTRIBUTE-GRID: Filled in VALUE, COUNT, HECTARES, PERIMETER, TM1, TM2, TM3, TM4, TM5, TM7, PAN, MNDVI, ELE, SLP, ASP, and SLPASP fields.

Source_Used_Citation_Abbreviation: IMSEG, DEM, HYDR
Process_Date: 20020620
Source_Produced_Citation_Abbreviation: EXTR
Process_Step:
Process_Description:
LABEL_MANUALLY: manually labeled urban, agriculture, water, recent burns, mines, and cloud areas. Manual labeling was done by manually identifying the above covertypes in the imagery. The manually classified labels were filled MANLABEL field.
Source_Used_Citation_Abbreviation: EXTR, CLAS
Process_Date: 20020701
Source_Produced_Citation_Abbreviation: EXTR
Process_Step:
Process_Description:
ASSEMBLE-TRAINDATA: training data was originally assembled for the SILC3 classification in 2001. Additional training data were collected during the summer of 2002 for sagebrush canopy cover and species. The new training data were added to the SILC3 train data set.

PROCESS-GT: filled in spectral, topographic, and hydrologic information in ground-truth training data for use in supervised classification.

Source_Used_Citation_Abbreviation: GRTR, EXTR
Process_Date: 20020715
Source_Produced_Citation_Abbreviation: GRT
Process_Step:
Process_Description:
LABEL_COVERTYPE: performed train data classification, assigning each region a land cover type label base on a train data set. The training data were analyzed for data attribute accuracy, positional accuracy, and agreement with the imagery. The grass train data points were assigned to a grass canopy cover group based on their MNDVI values:

CODE  DESCRIPTION                         MNDVI
-----------------------------------------------
3130  very low cover grasslands           < -10
3150  low / moderate cover grasslands    -10 to 34.9
3170  moderate / high cover grasslands    >= 35

Note: Snow was classified using training data. Normally snow is manually classified based on spectral classes associated with snow. However, ecognition does not produce spectral classes, thus snow was classified using training data.

The training data were run through a cross-validation test where each train data was removed and then classified using the remaining train data set. A Dudani distance-weighted euclidian distance classifier using variables: TM1, TM2, TM3, TM4, TM5, TM7, MNDVI, ELEV (scaled elevation), SLP, and SLPASP with a nearest neighborhood of 15 was used in conjunction with a mean inverse spatial classifier. Outlier train data were identified based on the cross-validation test and were removed from the train data set. After completing the train data evaluation, the final train data set was used to run a supervised classification of all regions in the BLM3729ECOG grid. The euclidian distance and spatial classifiers described above were used to classify the regions. The top three most probable land cover types were filled into the COV_CODE_1, COV_CODE_2, and COV_CODE_3 fields.

The final land cover type label was filled into the COVERTYPE field from the MANLABEL field if it was > -1 else from the COV_CODE_1 field.

Regions classified as Mixed Conifer Forest (4240) were recoded to one of three types based on a ruleset using the pvt type and elevation of the region.

4241       Mixed Upper Subalpine Conifer Forest
4242       Mixed Lower Subalpine Conifer Forest
4244       Mixed Xeric Conifer Forest

The aml with the ruleset is available through the data contact.

Source_Used_Citation_Abbreviation: EXTR, GRT
Process_Date: 20021115
Source_Produced_Citation_Abbreviation: EXTR
Process_Step:
Process_Description:
LABEL-SAGESPP: performed train data classification, assigning each region a sagebrush species label base on a train data set. The training data were analyzed for data attribute accuracy, positional accuracy, and agreement with the imagery. The training data were run through a cross-validation test where each train data was removed and then classified using the remaining train data set. A Dudani distance-weighted euclidian distance classifier using variables: TM1, TM2, TM3, TM4, TM5, TM7, MNDVI, ELEV (scaled elevation), SLP, and SLPASP with a nearest neighborhood of 15 was used in conjunction with a mean inverse spatial classifier. Outlier train data were identified based on the cross- validation test and were removed from the train data set. After completing the train data evaluation, the final train data set was used to run a supervised classification of all regions in the BLM3729ECOG grid. The euclidian distance and spatial classifiers described above were used to classify the regions. The most probable sagebrush species from the classification was filled into the SAGESPP1 field.
Source_Used_Citation_Abbreviation: EXTR, GRT
Process_Date: 20021115
Source_Produced_Citation_Abbreviation: EXTR
Process_Step:
Process_Description:
ACCURACY_ASSESS: performed accuracy assessment of the land cover type and sagebrush species using a leave-one-out cross-validation. The cross-validation involved removing the first training observation from the training set and constructing a new classification rule from the reduced training set. This new rule was then used to classify the polygon containing the left-out training observation. The process was repeated until all observations were held out once and used as a singleton test set.
Source_Used_Citation_Abbreviation: EXTR, GRT
Process_Date: 20021118
Source_Produced_Citation_Abbreviation: ACC
Process_Step:
Process_Description: VALIDATE: validation was performed for all items in the VAT.
Source_Used_Citation_Abbreviation: EXTR
Process_Date: 20021120
Source_Produced_Citation_Abbreviation: EXTR
Process_Step:
Process_Description:
LAND-COVER: created ECOGCOVDEN, 30m ArcInfo grid of overstory land cover types with sagebrush canopy cover types from the BLM3729ECOG grid COVERTYPE attribute.
Source_Used_Citation_Abbreviation: EXTR
Process_Date: 20021121
Source_Produced_Citation_Abbreviation: COVR
Process_Step:
Process_Description:
SAGE-SPECIES: created ECOGCOVSPP, 30m ArcInfo grid of overstory land cover types with sagebrush species types. First, all regions with a sagebrush canopy cover type in the COVERTYPE field were selected. Then the COVERTYPE field was filled in from the SAGESPP1 field. The ECOGCOVSPP grid was created from the BLM3729ECOG grid COVERTYPE attribute. Afterwards the COVERTYPE field was reset to the sagebrush canopy cover labels originally assigned.
Source_Used_Citation_Abbreviation: EXTR
Process_Date: 20021122
Source_Produced_Citation_Abbreviation:


Spatial_Data_Organization_Information:
Direct_Spatial_Reference_Method: Raster
Raster_Object_Information:
Raster_Object_Type: Grid cell
Row_Count: 6903
Column_Count: 13758


Spatial_Reference_Information:
Horizontal_Coordinate_System_Definition:
Planar:
Map_Projection:
Map_Projection_Name: Albers Conical Equal Area
Albers_Conical_Equal_Area:
Standard_Parallel: 45 37 30
Standard_Parallel: 48 30 00
Longitude_of_Central_Meridian: -110 00 00
Latitude_of_Projection_Origin: 44 30 00
False_Easting: 0
False_Northing: 0
Planar_Coordinate_Information:
Planar_Coordinate_Encoding_Method: row and column
Coordinate_Representation:
Abscissa_Resolution: 15
Ordinate_Resolution: 15
Planar_Distance_Units: meters
Geodetic_Model:
Horizontal_Datum_Name: North American Datum of 1927
Ellipsoid_Name: Clarke 1866
Semi-major_Axis: 6378206.4
Denominator_of_Flattening_Ratio: 294.98


Entity_and_Attribute_Information:
Detailed_Description:
Entity_Type:
Entity_Type_Label: BLM3729ECOG.VAT
Entity_Type_Definition: Value attribute table for grid
Entity_Type_Definition_Source: None
Attribute:
Attribute_Label: VALUE
Attribute_Definition:
Unique region identification number, assigned by ArcInfo software when the BUILDVAT process is run.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 1
Range_Domain_Maximum: 549246
Attribute:
Attribute_Label: COUNT
Attribute_Definition: Number of 15m x 15m pixels for each region in the zone grid.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 2
Range_Domain_Maximum: 3140
Attribute:
Attribute_Label: COVERTYPE
Attribute_Definition:
Final land cover type class assigned after training data and manual classifications, if MANLABEL > -1 then COVERTYPE = MANLABEL, else COVERTYPE = COV_CODE_1.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Unrepresentable_Domain:
COVERTYPE  DESCRIPTION
------------------------------------------------
1100       Urban or Developed Lands
2010       Agricultural - Dry
2020       Agricultural - Irrigated
3130       Very Low Cover Grasslands
3150       Low / Moderate Cover Grasslands
3170       Moderate / High Cover Grasslands
3370       Sagebrush / Xeric Shrubs 05-14% Cover
3380       Sagebrush / Xeric Shrubs 15-24% Cover
3390       Sagebrush / Xeric Shrubs 25-34% Cover
3395       Sagebrush / Xeric Shrubs >= 35% Cover
3610       Mesic Shrublands / Willow
4101       Aspen
4150       Mixed Broadleaf / Cottonwood Forest
4203       Lodgepole Pine
4204       Whitebark Pine
4205       Limber Pine
4206       Ponderosa Pine
4212       Douglas-fir
4216       Utah Juniper
4223       Douglas-fir/Lodgepole Pine
4237       Subalpine Fir/Spruce
4241       Mixed Upper Subalpine Forest
4242       Mixed Lower Subalpine Forest
4244       Mixed Xeric Conifer Forest
4400       Burns
5000       Water
7300       Rock/Barren
9100       Snow
Attribute:
Attribute_Label: COV_CODE_1
Attribute_Definition:
Most probable land cover type assigned in train data classification.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Unrepresentable_Domain:
COVERTYPE  DESCRIPTION
------------------------------------------------
3130       Very Low Cover Grasslands
3150       Low / Moderate Cover Grasslands
3170       Moderate / High Cover Grasslands
3370       Sagebrush / Xeric Shrubs 05-14% Cover
3380       Sagebrush / Xeric Shrubs 15-24% Cover
3390       Sagebrush / Xeric Shrubs 25-34% Cover
3395       Sagebrush / Xeric Shrubs >= 35% Cover
3610       Mesic Shrublands / Willow
4101       Aspen
4150       Mixed Broadleaf / Cottonwood Forest
4203       Lodgepole Pine
4204       Whitebark Pine
4205       Limber Pine
4206       Ponderosa Pine
4212       Douglas-fir
4216       Utah Juniper
4223       Douglas-fir/Lodgepole Pine
4237       Subalpine Fir/Spruce
4240       Mixed Conifer Forest
7300       Rock/Barren
9100       Snow
Attribute:
Attribute_Label: COV_PROB_1
Attribute_Definition:
Posterior probability of COV_CODE_1 prediction from train data classification.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 0.180000
Range_Domain_Maximum: 1.0
Attribute:
Attribute_Label: COV_CODE_2
Attribute_Definition:
Second most probable land cover type assigned in train data classification.
Attribute_Definition_Source: None
Attribute_Domain_Values:
Unrepresentable_Domain: See COV_CODE_1.
Attribute:
Attribute_Label: COV_PROB_2
Attribute_Definition:
Posterior probability of COV_CODE_2 prediction from train data classification.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 0.162000
Range_Domain_Maximum: 1.0
Attribute:
Attribute_Label: COV_CODE_3
Attribute_Definition:
Third most probable land cover type assigned in train data classification.
Attribute_Definition_Source: None
Attribute_Domain_Values:
Unrepresentable_Domain: See COV_CODE_1.
Attribute:
Attribute_Label: COV_PROB_3
Attribute_Definition:
Posterior probability of COV_CODE_3 prediction from train data classification.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 0.163000
Range_Domain_Maximum: 1.0
Attribute:
Attribute_Label: MANLABEL
Attribute_Definition: Land cover type class assigned from manual classifications.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Unrepresentable_Domain:
Code Definition
----------------
-1   Not Manually Classified
1100 Urban or Developed Lands
2010 Agricultural - Dry
2020 Agricultural - Irrigated
4400 Burns
5000 Water
7500 Mines / Quarries
Attribute:
Attribute_Label: TM1
Attribute_Definition: Mean spectral value of region in Thematic Mapper channel 1.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 0
Range_Domain_Maximum: 241.885910
Attribute_Units_of_Measure: Units of radiance
Attribute_Measurement_Resolution: 0.000099
Attribute:
Attribute_Label: TM2
Attribute_Definition: Mean spectral value of region in Thematic Mapper channel 2.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 0.000000
Range_Domain_Maximum: 240.491547
Attribute_Units_of_Measure: Units of radiance
Attribute_Measurement_Resolution: 0.000099
Attribute:
Attribute_Label: TM3
Attribute_Definition: Mean spectral value of region in Thematic Mapper channel 3.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 0.000000
Range_Domain_Maximum: 244.209854
Attribute_Units_of_Measure: Units of radiance
Attribute_Measurement_Resolution: 0.000099
Attribute:
Attribute_Label: TM4
Attribute_Definition: Mean spectral value of region in Thematic Mapper channel 4.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 0.000000
Range_Domain_Maximum: 236.594360
Attribute_Units_of_Measure: Units of radiance
Attribute_Measurement_Resolution: 0.000099
Attribute:
Attribute_Label: TM5
Attribute_Definition: Mean spectral value of region in Thematic Mapper channel 5.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 0
Range_Domain_Maximum: 228.500000
Attribute_Units_of_Measure: Units of radiance
Attribute_Measurement_Resolution: 0.000099
Attribute:
Attribute_Label: TM7
Attribute_Definition: Mean spectral value of region in Thematic Mapper channel 7.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 0
Range_Domain_Maximum: 244.075470
Attribute_Units_of_Measure: Units of radiance
Attribute_Measurement_Resolution: 0.000099
Attribute:
Attribute_Label: PAN
Attribute_Definition: Mean spectral value of region in panchromatic band.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 0.000000
Range_Domain_Maximum: 255.000000
Attribute_Units_of_Measure: Units of radiance
Attribute_Measurement_Resolution: 0.000099
Attribute:
Attribute_Label: MNDVI
Attribute_Definition:
Modified normalized difference vegetation index; calculated according to the following equation: (TM4 - TM3) / (TM4 + TM3 + 1) * ( 256 / (TM5 + 1)) * 100.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: -144.855972
Range_Domain_Maximum: 125.313774
Attribute:
Attribute_Label: ELE
Attribute_Definition: Mean elevation of region calculated from USGS DEM.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 1059.796875
Range_Domain_Maximum: 3876.258057
Attribute_Units_of_Measure: meters
Attribute_Measurement_Resolution: 0.000099
Attribute:
Attribute_Label: SLP
Attribute_Definition: Mean slope of region calculated from USGS DEM.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 0.000000
Range_Domain_Maximum: 74.952377
Attribute_Units_of_Measure: decimal degrees
Attribute_Measurement_Resolution: 0.000099
Attribute:
Attribute_Label: ASP
Attribute_Definition: Majority aspect of the region calculated from USGS DEM.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Unrepresentable_Domain:
Code  Description
-----------------
0     Flat
1     North
2     Northeast
3     East
4     Southeast
5     South
6     Southwest
7     West
8     Northwest
Attribute:
Attribute_Label: SLPASP
Attribute_Definition:
Mean slope/aspect combination of region; calculated according to the following equation: cos ( ( ( ( (ASP - 1) * 45) + 135) / 360) * 6.2832).
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: -69.066666
Range_Domain_Maximum: 66.900002
Attribute:
Attribute_Label: XCOORD
Attribute_Definition: X-coordinate from center of region in meters.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 545902.500000
Range_Domain_Maximum: 750487.500000
Attribute:
Attribute_Label: YCOORD
Attribute_Definition: Y-coordinate from center of region in meters.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 41664.351562
Range_Domain_Maximum: 145133.421875
Attribute:
Attribute_Label: PERIMETER
Attribute_Definition: Perimeter of the region.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 90
Range_Domain_Maximum: 14160
Attribute_Units_of_Measure: meters
Attribute_Measurement_Resolution: 1
Attribute:
Attribute_Label: HECTARES
Attribute_Definition: Area for the region in hectares (COUNT * 0.09).
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 0.045000
Range_Domain_Maximum: 70.650002
Attribute:
Attribute_Label: SAGESPP1
Attribute_Definition:
The most probable sagebrush species from the sagebrush species training data classification.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Unrepresentable_Domain:
Code  Description
-----------------
3311  Greasewood
3351  Mountain Big Sagebrush
3352  Wyoming Big Sagebrush
3353  Basin Big Sagebrush
3354  Black / Low Sagebrush
Attribute:
Attribute_Label: SAGESPP1_PROB
Attribute_Definition:
Posterior probability of SAGESPP1 prediction from train data classification.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Range_Domain:
Range_Domain_Minimum: 0.218000
Range_Domain_Maximum: 1.0
Attribute:
Attribute_Label: PVTCODE
Attribute_Definition:
Region One Potential Vegetation Type (PVT) based on zonal majority. Please contact the Geospatial Services Group at Region One of the USDA Forest Service to get more information on the definitions of the PVTCODES and how they were developed.
Attribute_Definition_Source: none
Attribute_Domain_Values:
Unrepresentable_Domain:
Code Definition
----------------
0    no pvt code assigned
1    abla1 (lower elevation subalpine fir type)
2    abla2 (high elevation subalpine fir type)
6    picea (spruce type)
7    pincon (lodgepole pine)
8    pifl (limber pine type)
9    pinpon (ponderosa pine)
11   psemen (Douglas-fir type)
12   agric (agriculture type)
13   agrsmi
14   alpgrass (alpine grasslands type)
15   drygrass (xeric, valley bottom or foothills grassland type)
16   fesida (Idaho fescue type)
18   ripgrass (riparian grass type)
19   rock (rock dominated type)
20   bigsage (big sagebrush type)
21   dryshrub (non-sagebrush xeric shrub type)
22   messhrub
23   potfru (shrubby cinquefoil type)
25   ripshrub (riparian shrub type)
27   urban
33   pial/laly (white bark pine/alpine larch)


Distribution_Information:
Distributor:
Contact_Information:
Contact_Organization_Primary:
Contact_Organization: Montana State Office of the Bureau of Land Management
Contact_Address:
Address_Type: mailing and physical address
Address: PO Box 36800, 5001 Southgate Drive
City: Billings
State_or_Province: MT
Postal_Code: 59107
Country: USA
Contact_Voice_Telephone: 406 896 5293
Contact_Facsimile_Telephone: 406 896 5293
Contact_Electronic_Mail_Address: Roxanne_Falise@blm.gov
Hours_of_Service: Monday-Friday, 8-5, Pacific Time
Contact_Instructions: email or call for data requests
Resource_Description: BLM3729ECOG ArcInfo grid
Distribution_Liability:
This data set is in the private domain, and the recipient may not assert any proprietary rights thereto nor represent it to anyone as other than a data set produced by The University of Montana under contract to the the Bureau of Land Management; it is provided "as-is" without warranty of any kind, including, but not limited to, the implied warranties of merchantability and fitness for a particular purpose. The user assumes all responsibility for the accuracy and suitability of this data set for a specific application. In no event will the data set producers at The University of Montana or the Bureau of Land Management be liable for any damages, including lost profits, lost savings, or other incidental or consequential damages arising from the use of or the inability to use this data set.
Standard_Order_Process:
Digital_Form:
Digital_Transfer_Information:
Format_Name: ArcInfo Grid data sets
Digital_Transfer_Option:
Offline_Option:
Offline_Media: CD-ROM
Recording_Format: ISO 9660
Fees:
None.


Metadata_Reference_Information:
Metadata_Date: 20021122
Metadata_Contact:
Contact_Information:
Contact_Organization_Primary:
Contact_Organization: Wildlife Spatial Analysis Lab
Contact_Person: Chip Fisher
Contact_Position: Image Analyst
Contact_Address:
Address_Type: mailing and physical address
Address: Wildlife Spatial Analysis Lab, The University of Montana
City: Missoula
State_or_Province: Montana
Postal_Code: 59812-1063
Country: USA
Contact_Voice_Telephone: 406 243 5208 (email preferred)
Contact_Facsimile_Telephone: 406 243 6064
Contact_Electronic_Mail_Address: cfisher@wru.umt.edu
Hours_of_Service: Monday-Friday, 8-5, Mountain Time
Metadata_Standard_Name: FGDC Content Standards For Digital Geospatial Metadata
Metadata_Standard_Version: 19940608
Metadata_Time_Convention: local time

Generated by mp on Mon Dec 2 13:07:52 2002