ArcGIS REST Services Directory Login | Get Token
JSON | SOAP | WMS

Mosaics/WMS_DBO_IBL_ENR_FMD_MVI_Landcover_VegetationTypeDensity (ImageServer)

View In:   ArcGIS JavaScript   ArcGIS Online Map Viewer   ArcGIS Earth   ArcMap

View Footprint In:   ArcGIS Online Map Viewer

Service Description:

An update to the land cover over Phase 1 of the Northwest Territories study area was undertaken using Landsat TM imagery collected from 2006 to 2008 (c.2007); the Phase 2 and Phase 3 area consisted of Landsat TM imagery collected from 2007 to 2013 (c.2010). The image classification procedure followed the methods of Earth Observation of Sustainable Development of Forests (EOSD), which employs unsupervised classifications and cluster labeling techniques to classify twenty-four land cover classes. The main objectives for updating the land cover were: i) to replace the original circa 2000 EOSD land cover with an EOSD product more current to the natural disturbances of the landscape (ex., fire and forest regeneration), and ii) to improve upon some of the issues in creating the original land cover data (e.g., NoData, seamlines, misclassifications).

A large collection of aerial survey data (including field and air plots provided by Ducks Unlimited and the NWT Ecological Land Classification photo library), forest inventory datasets, and high resolution satellite imagery was referenced for labeling clusters. In Phase 2 a total of 2,692 calibration locations were used, and classification accuracy was assessed for 17 classes using 954 independent validation locations (stratified random sampling). In Phase 3 a total of 1,340 calibration locations were used, and classification accuracy was assessed for 17 classes using 535 independent validation locations (stratified random sampling). Calibration and validation data locations were captured within the last ten years prior to the Landsat scenes and were removed if located in areas that were subsequently burned (> 2003). Any changes in relative abundance of vegetative species between the calibration/validation data and the image data were expected to be minor because of the slow growth and successional rates in the NWT.

Data gaps due to cloud cover were filled with additional Landsat TM and OLI scenes using the same EOSD classification process. An iterative filtering technique was applied to the final product to reduce granularity. Hydrology and road layers (CANVEC 1:50K, NRCAN National Road Network) were used to further update and delineate linear features to ensure compatibility with other datasets. Mountains were manually delineated.

The results of this image classification project were a more consistent and accurate land cover map of the vegetation composition and density relative to c.2000 EOSD product. Based on the independent validation in Phase 2, overall accuracy improved from 29% (c. 2000 EOSD) to 50% for the original EOSD classification schema (95% confidence interval: 47 % to 53%). The same independent data (n = 954 samples across 17 classes) were used to assess the c. 2000 EOSD classification. Based on the independent validation, in Phase 3, overall accuracy improved from 29% (c. 2000 EOSD) to 52% for the original EOSD classification schema (95% confidence interval: 47% to 56%).

Classification schema:

Phase 1 (c.2007)

Field data collected in 2007 for independent assessment of the remote sensing inventory products was used to assess the accuracy of the updated circa 2007 land cover. A classification error matrix of the EOSD forested classes was generated below:

EOSD class (#)

N

EOSD c.2000

EOSD c.2007

Prod %

User %

Prod %

User %

Wetland Treed (81)

3

0

0

67

100

Conifer Dense (211)

8

38

16

63

56

Conifer Open (212)

15

27

22

73

61

Conifer Sparse (213)

5

20

20

40

100

Deciduous Dense (221)

6

33

100

50

75

Deciduous Open (222)

3

67

100

100

75

Mixedwood Dense (231)

11

0

0

82

60

Mixedwood Open (232)

5

0

0

20

100

Overall Accuracy: c.2000 = 21%; c.2007 = 64%

By combining the density classes of the table above and assessing the broad-level classification of conifer, deciduous, and mixedwood land cover, the accuracies increase as indicated in the table below:

EOSD class category

N

EOSD c.2000

EOSD c.2007

Prod %

User %

Prod %

User %

Conifer

31

84

67

90

90

Deciduous

9

44

100

78

88

Mixedwood

16

0

0

81

81

Overall Accuracy: c.2000 = 53%; c.2007 = 86%

Phase 2 (c.2010)

Class Number | Class description | User Accuracy c. 2010 EOSD | User Accuracy c. 2000 EOSD | Improvement

0 | NoData | Accuracy not assessed

11 | Shadow | Accuracy not assessed

12 | Cloud | Accuracy not assessed

20 | Water | 89% | 68% | +22%

31 | Snow / Ice | Accuracy not assessed

32 | Rock / Rubble | 71% | 5% | +67%

33 | Exposed land | 92% | 15% | +77%

34 | Roads | Accuracy not assessed, not part of original EOSD schema

40 | Bryoids | 92% | 43% | +49%

51 | Shrub tall | 58% | 19% | +39%

52 | Shrub low | 49% | 9% | +40%

81 | Wetland-treed | 28% | 21% | +7%

82 | Wetland-shrub | 50% | 26% | +24%

83 | Wetland-herb | 60% | 18% | +42%

100 | Herb | 33% | 0% | +33%

211 | Coniferous Dense | 53% | 52% | +1%

212 | Coniferous Open | 32% | 26% | +6%

213 | Coniferous Sparse | 34% | 25% | +9%

221 | Broadleaf Dense | 53% | 65% | -13%

222 | Broadleaf Open | 25% | 0% | +25%

223 | Broadleaf Sparse | Deemed absent in Taiga Plains

231 | Mixedwood Dense | 35% | 25% | +10%

232 | Mixedwood Open | 40% | 11% | +29%

233 | Mixedwood Sparse | Deemed absent in Taiga Plains

Phase 3 (c.2010)

0 | NoData | Accuracy not assessed

11 | Shadow | Accuracy not assessed

12 | Cloud | Accuracy not assessed

20 | Water | 61% | 68% | -7%

31 | Snow / Ice | Accuracy not assessed

32 | Rock / Rubble | 100% | 5% | +95%

33 | Exposed land | 79% | 15% | +64%

34 | Roads | Accuracy not assessed, not part of original EOSD schema

40 | Bryoids | 86% | 43% | +43%

51 | Shrub tall | 48% | 19% | +29%

52 | Shrub low | 55% | 9% | +46%

81 | Wetland-treed | 23% | 21% | +2%

82 | Wetland-shrub | 47% | 26% | +21%

83 | Wetland-herb | 40% | 18% | +22%

100 | Herb | 0% | 0% | 0%

211 | Coniferous Dense | 37% | 52% | -15%

212 | Coniferous Open | 54% | 26% | +28%

213 | Coniferous Sparse | 48% | 25% | +23%

221 | Broadleaf Dense | 80% | 65% | +15%

222 | Broadleaf Open | 40% | 0% | +40%

223 | Broadleaf Sparse | Deemed absent in Taiga Plains

231 | Mixedwood Dense | 33% | 25% | +8%

232 | Mixedwood Open | 50% | 11% | +39%

233 | Mixedwood Sparse | Deemed absent in Taiga Plains



Name: Mosaics/WMS_DBO_IBL_ENR_FMD_MVI_Landcover_VegetationTypeDensity

Description:

An update to the land cover over Phase 1 of the Northwest Territories study area was undertaken using Landsat TM imagery collected from 2006 to 2008 (c.2007); the Phase 2 and Phase 3 area consisted of Landsat TM imagery collected from 2007 to 2013 (c.2010). The image classification procedure followed the methods of Earth Observation of Sustainable Development of Forests (EOSD), which employs unsupervised classifications and cluster labeling techniques to classify twenty-four land cover classes. The main objectives for updating the land cover were: i) to replace the original circa 2000 EOSD land cover with an EOSD product more current to the natural disturbances of the landscape (ex., fire and forest regeneration), and ii) to improve upon some of the issues in creating the original land cover data (e.g., NoData, seamlines, misclassifications).

A large collection of aerial survey data (including field and air plots provided by Ducks Unlimited and the NWT Ecological Land Classification photo library), forest inventory datasets, and high resolution satellite imagery was referenced for labeling clusters. In Phase 2 a total of 2,692 calibration locations were used, and classification accuracy was assessed for 17 classes using 954 independent validation locations (stratified random sampling). In Phase 3 a total of 1,340 calibration locations were used, and classification accuracy was assessed for 17 classes using 535 independent validation locations (stratified random sampling). Calibration and validation data locations were captured within the last ten years prior to the Landsat scenes and were removed if located in areas that were subsequently burned (> 2003). Any changes in relative abundance of vegetative species between the calibration/validation data and the image data were expected to be minor because of the slow growth and successional rates in the NWT.

Data gaps due to cloud cover were filled with additional Landsat TM and OLI scenes using the same EOSD classification process. An iterative filtering technique was applied to the final product to reduce granularity. Hydrology and road layers (CANVEC 1:50K, NRCAN National Road Network) were used to further update and delineate linear features to ensure compatibility with other datasets. Mountains were manually delineated.

The results of this image classification project were a more consistent and accurate land cover map of the vegetation composition and density relative to c.2000 EOSD product. Based on the independent validation in Phase 2, overall accuracy improved from 29% (c. 2000 EOSD) to 50% for the original EOSD classification schema (95% confidence interval: 47 % to 53%). The same independent data (n = 954 samples across 17 classes) were used to assess the c. 2000 EOSD classification. Based on the independent validation, in Phase 3, overall accuracy improved from 29% (c. 2000 EOSD) to 52% for the original EOSD classification schema (95% confidence interval: 47% to 56%).

Classification schema:

Phase 1 (c.2007)

Field data collected in 2007 for independent assessment of the remote sensing inventory products was used to assess the accuracy of the updated circa 2007 land cover. A classification error matrix of the EOSD forested classes was generated below:

EOSD class (#)

N

EOSD c.2000

EOSD c.2007

Prod %

User %

Prod %

User %

Wetland Treed (81)

3

0

0

67

100

Conifer Dense (211)

8

38

16

63

56

Conifer Open (212)

15

27

22

73

61

Conifer Sparse (213)

5

20

20

40

100

Deciduous Dense (221)

6

33

100

50

75

Deciduous Open (222)

3

67

100

100

75

Mixedwood Dense (231)

11

0

0

82

60

Mixedwood Open (232)

5

0

0

20

100

Overall Accuracy: c.2000 = 21%; c.2007 = 64%

By combining the density classes of the table above and assessing the broad-level classification of conifer, deciduous, and mixedwood land cover, the accuracies increase as indicated in the table below:

EOSD class category

N

EOSD c.2000

EOSD c.2007

Prod %

User %

Prod %

User %

Conifer

31

84

67

90

90

Deciduous

9

44

100

78

88

Mixedwood

16

0

0

81

81

Overall Accuracy: c.2000 = 53%; c.2007 = 86%

Phase 2 (c.2010)

Class Number | Class description | User Accuracy c. 2010 EOSD | User Accuracy c. 2000 EOSD | Improvement

0 | NoData | Accuracy not assessed

11 | Shadow | Accuracy not assessed

12 | Cloud | Accuracy not assessed

20 | Water | 89% | 68% | +22%

31 | Snow / Ice | Accuracy not assessed

32 | Rock / Rubble | 71% | 5% | +67%

33 | Exposed land | 92% | 15% | +77%

34 | Roads | Accuracy not assessed, not part of original EOSD schema

40 | Bryoids | 92% | 43% | +49%

51 | Shrub tall | 58% | 19% | +39%

52 | Shrub low | 49% | 9% | +40%

81 | Wetland-treed | 28% | 21% | +7%

82 | Wetland-shrub | 50% | 26% | +24%

83 | Wetland-herb | 60% | 18% | +42%

100 | Herb | 33% | 0% | +33%

211 | Coniferous Dense | 53% | 52% | +1%

212 | Coniferous Open | 32% | 26% | +6%

213 | Coniferous Sparse | 34% | 25% | +9%

221 | Broadleaf Dense | 53% | 65% | -13%

222 | Broadleaf Open | 25% | 0% | +25%

223 | Broadleaf Sparse | Deemed absent in Taiga Plains

231 | Mixedwood Dense | 35% | 25% | +10%

232 | Mixedwood Open | 40% | 11% | +29%

233 | Mixedwood Sparse | Deemed absent in Taiga Plains

Phase 3 (c.2010)

0 | NoData | Accuracy not assessed

11 | Shadow | Accuracy not assessed

12 | Cloud | Accuracy not assessed

20 | Water | 61% | 68% | -7%

31 | Snow / Ice | Accuracy not assessed

32 | Rock / Rubble | 100% | 5% | +95%

33 | Exposed land | 79% | 15% | +64%

34 | Roads | Accuracy not assessed, not part of original EOSD schema

40 | Bryoids | 86% | 43% | +43%

51 | Shrub tall | 48% | 19% | +29%

52 | Shrub low | 55% | 9% | +46%

81 | Wetland-treed | 23% | 21% | +2%

82 | Wetland-shrub | 47% | 26% | +21%

83 | Wetland-herb | 40% | 18% | +22%

100 | Herb | 0% | 0% | 0%

211 | Coniferous Dense | 37% | 52% | -15%

212 | Coniferous Open | 54% | 26% | +28%

213 | Coniferous Sparse | 48% | 25% | +23%

221 | Broadleaf Dense | 80% | 65% | +15%

222 | Broadleaf Open | 40% | 0% | +40%

223 | Broadleaf Sparse | Deemed absent in Taiga Plains

231 | Mixedwood Dense | 33% | 25% | +8%

232 | Mixedwood Open | 50% | 11% | +39%

233 | Mixedwood Sparse | Deemed absent in Taiga Plains



Single Fused Map Cache: false

Extent: Initial Extent: Full Extent: Pixel Size X: 30.0

Pixel Size Y: 30.0

Band Count: 1

Pixel Type: U8

RasterFunction Infos: {"rasterFunctionInfos": [{ "name": "None", "description": "A No-Op Function.", "help": "" }]}

Mensuration Capabilities:

Has Histograms: true

Has Colormap: false

Has Multi Dimensions : false

Rendering Rule:

Min Scale: 0

Max Scale: 0

Copyright Text: Citation: Natural Resources Canada and Government of Northwest Territories. (2017). Monitoring forests in the Northwest Territories – Earth Observation for Sustainable Development of Forests (EOSD) land cover updates Phase 1, 2 and 3. Data access through the Data Coordinator, NWT Centre for Geomatics: gnwtmaps_admin@gov.nt.ca. Credits: Northern Forestry Centre, Canadian Forest Service, Natural Resources Canada Forest Management Division, Government of the Northwest Territories NWT Centre for Geomatics, Government of the Northwest Territories Ducks Unlimited Canada

Service Data Type: esriImageServiceDataTypeThematic

Min Values: 0

Max Values: 255

Mean Values: 102.71785605272763

Standard Deviation Values: 83.84273095425013

Object ID Field: OBJECTID

Fields: Default Mosaic Method: Northwest

Allowed Mosaic Methods: NorthWest,Center,LockRaster,ByAttribute,Nadir,Viewpoint,Seamline,None

SortField:

SortValue: null

Mosaic Operator: First

Default Compression Quality: 75

Default Resampling Method: Nearest

Max Record Count: 1000

Max Image Height: 4100

Max Image Width: 15000

Max Download Image Count: 20

Max Mosaic Image Count: 20

Allow Raster Function: true

Allow Compute TiePoints: false

Supports Statistics: true

Supports Advanced Queries: true

Use StandardizedQueries: true

Raster Type Infos: Has Raster Attribute Table: true

Edit Fields Info: null

Ownership Based AccessControl For Rasters: null

Child Resources:   Info   Raster Attribute Table   Histograms   Key Properties   Legend   MultiDimensionalInfo   rasterFunctionInfos

Supported Operations:   Export Image   Query   Identify   Measure   Compute Histograms   Compute Statistics Histograms   Get Samples   Compute Class Statistics   Query Boundary