Presentation is loading. Please wait.

Presentation is loading. Please wait.

Vegetation classification using map coincidence and pattern recognition   By George Wooten and Dave Demyan, Pacific Northwest Division, Planetary Science.

Similar presentations


Presentation on theme: "Vegetation classification using map coincidence and pattern recognition   By George Wooten and Dave Demyan, Pacific Northwest Division, Planetary Science."— Presentation transcript:

1 Vegetation classification using map coincidence and pattern recognition
By George Wooten and Dave Demyan, Pacific Northwest Division, Planetary Science Institute, Winthrop, Washington & Hans Smith, Pacific Biodiversity Institute, Winthrop, Washington This study improved existing maps of vegetation, fuels and canopy structure for the Sinlahekin Wildlife Area in North-Central Washington, as part of a plan to restore fire to the ecosystem. Models were classified using a combination of ASTER and orthophoto imagery, patch analysis, pattern recognition, and coincidence mapping of overlapping cells. The coincidence was compared among these and other existing classifications to derive an optimum approximation. Coincidence mapping of disparate data sources is an expedient tool for improving the confidence of classifications. Presented to Northwest Scientific Association (2004) By George Wooten and Dave Demyan, Pacific Northwest Division, Planetary Science Institute, Winthrop, Washington, and Hans Smith, Pacific Biodiversity Institute, Winthrop, Washington

2 Fire regimes Restoration planning was based on dynamic landscape models, ecology exams and GIS analysis 2. The purpose of the Sinlahekin fuels assessment and treatment analysis was to develop scientifically based prescriptions for restoration of historical fire regimes and fuel reduction activities on a 12,000 acre wildlife area.

3 Orthophoto images at 1m Determination of canopy structure was based on patch pattern recognition of gray-scale orthophoto images 3. Orthophoto images. Desired pattern recognition test areas are shown in blue polygons.

4 Patches were determined by brightness
4. Patches were determined by 16 brightness levels.

5 Patch thickness was the primary stand metric
5. Patch thickness. Shape, size, thickness, and inter-patch distance were used to derive pattern elements. These were combined to give the patch pattern

6 Distance Inter-patch distance was the primary metric for canopy openings 6. Distance Canopy openings were categorized into 3 main categories by inter-patch distance and type of adjacent patches.

7 Combined Patches & Openings
7. Patches combined with openings. The combination of types of openings and types of patches was made into a single map.

8 Regions of Canopy Openings
8. Regions of openings. 200 different combinations of canopy openings, aspect shading and canopy type were assigned different factors for canopy fraction within canopy shadows.

9 Patch Pattern Data – 5 classes
9. Result – Patch Patterns. The patch pattern factors were used to inform a canopy cover classification (green shades) grouped here by regions of average cover and overlaid with the cells representing trees and shadows. Patch Pattern Data – 5 classes

10 Utah State Data – 5 classes
10. Utah State Data. In contrast, the Utah State Data shows numerous stray cells of different canopy levels representing (and sometimes misrepresenting) patches of trees. The green color ramp has the same level of intervals as the previous image, but there are large areas with zero values interspersed with stray pixels of high cover value. Utah State Data – 5 classes

11 Canopy cover coincidence
11. Canopy cover coincidence. The final canopy cover map was classified in the same intervals as that of the Utah State canopy cover map and compared by counting the number of coincident cells.

12 Patch Pattern Data – 100 classes
12. Patch pattern – 100 intervals. The patch pattern cover classification of orthophoto images has good potential to become a widely used method of determining accurate canopy cover, which can be averaged over various areas. Patch Pattern Data – 100 classes

13 Shrub-steppe Map Coincidence: NLCD
13. NLCD Shrub-steppe coincidence. The use of map coincidence was useful for other layers on the SWA, by combining up to 4 layers representing a single feature and determining the amount of coincidence. This illustrates the coincidence of the NLCD shrub-steppe layer with a merged shrub-steppe layer from 4 different sources of shrub-steppe maps. The coincidence of the NLCD layer was 486,310 / 655,643 (74%) of the merged shrub-steppe layer. The percentage of coincidence of all shrub-steppe layers was 257,232 or 39% of the merged shrub-steppe layer.

14 Shrub-steppe Map Coincidence: NCGBE
14. NCGBE coincidence. This illustrates the coincidence of the NCGBE shrub-steppe layer with the merged shrub-steppe layer.

15 Shrub-steppe Map Coincidence: USU
15. USU coincidence. This illustrates the coincidence of the Utah State shrub-steppe layer with the merged shrub-steppe layer.

16 Shrub-steppe Map Coincidence: NHI
16. NHI coincidence. This illustrates the coincidence of the NHI shrub-steppe layer with the merged shrub-steppe layer.


Download ppt "Vegetation classification using map coincidence and pattern recognition   By George Wooten and Dave Demyan, Pacific Northwest Division, Planetary Science."

Similar presentations


Ads by Google