Documenting Land Cover and Vegetation Productivity Changes in the NWT using the Landsat Satellite Archive Fraser, R.H 1, Olthof, I. 1, Deschamps, A. 1, Pregitzer, M. 1, Kokelj, S. 2, Lantz, T. 3,Wolfe, S. 4, Brooker, A. 5, Lacelle, D. 5, and Schwarz, S. 6 (1) Canada Centre for Remote Sensing, Natural Resources Canada, Ottawa, ON (2) NWT Geoscience Office, Govt of the Northwest Territories, Yellowknife, NWT (3) School of Environmental Studies, University of Victoria, Victoria, BC (4) Geological Survey of Canada, Natural Resources Canada, Ottawa, ON (5) Department of Geography, University of Ottawa, Ottawa, ON (6) NWT Centre for Geomatics, Govt of the Northwest Territories, Yellowknife, NWT
Outline 1. Need for large-area monitoring in the Northwest Territories 2. Potential of Landsat image archive for northern monitoring 3. Change detection using dense Landsat image stacks 4. NWT study regions and methods 5. NWT Trend detection results Landscape disturbances and tundra greening
1. Need for large-area monitoring in NWT NWT is a large and sparsely populated territory facing cumulative impacts from development and climate change Inuvik Annual Temperature Means (NWT Environment and Natural Resources) NWT Mines and Exploration Projects (NWT Geoscience Office)
1. Need for large-area monitoring in NWT Strong need for comprehensive environmental monitoring recently highlighted in the Rosenberg International Forum report on the Mackenzie River Basin: a strong, well-designed and ongoing monitoring program an absolutely essential precondition for effective management of the Mackenzie River Basin. NWT Cumulative Impacts Monitoring Program (CIMP) Supports numerous initiatives for building monitoring capacity including three projects with CCRS/NRCan involvement aimed at expanding scale of monitoring using EO The NRCan TRACS project also using EO to assess terrain sensitivity to permafrost degradation and potential impacts on transportation networks
2. Potential of Landsat Image Archive for Northern Monitoring Landsat has a spatial grain (30 m) and extent (185 km) ideal for large-area monitoring. A rich 28-Year (1984-2012) archive exists for Northern Canada Baseline monitoring and retrospective change analysis can be followed by forward monitoring USGS Archive Holdings WRS-2 Frame Overlap (CIMP study region)
3. Change Detecting Using Dense Landsat Image Stacks LandTrendr (Kennedy et al.) With the opening of the Landsat archive, more change detection initiatives have exploited dense time series of imagery (Wulder et al. 2012, RSE) Kennedy, Cohen et al. (LandTrendr), Huang et al., Masek et al., (Vegetation Change Tracker), Vogelman et al., Goodwin et al., Schroeder et al. Most have studied temperate forested ecosystems few the North Masek et al. 2012, Fraser et al., 2012 CCRS and Parks Canada investigated potential for using Landsat image stacks to monitor northern parks (ParkSPACE) ParkSPACE Ivvavik National Park Shrub
4. Current NWT Landsat Analysis Landsat Study Regions Goal: Investigate potential to use Landsat archive for monitoring a range of landscape changes in NWT CIMP Study Regions: 1.Peel Plateau and Mackenzie Delta (NWT CIMP projects) 2.Great Slave Geological Province (TRACS projects) TRACS
4. Image Stack Change Method 1. Build 25-year Landsat Image Stack Cloud/shadow/SLC masking, TOA reflectance, peak-phenology screening using AVHRR/MODIS 2. Extract Pixel Time Series Values Unique database for each pixel of six Vegetation Indices 3. Derive Linear Trends in Landsat Indices (e.g. Tasseled Cap Brightness, Greenness, Wetness) 6. Relate TC Trends to Changes in Vegetation Composition (Scale up high res training data using regression trees) 5. Create Trend Images (RBG= TCB, TCG, TCW) 6. Change Classification Product (Decision tree classifier with ancillary GIS data)
Methods: RGB Composite Trend Images for Visualizing Physical Changes 1985-2012 TC Brightness Trend 1985-2012 TC Greenness Trend 1985-2012 TC Wetness Trend Interpretation Key Red = B G W (e.g. veg bare, dev.) Yellow = B G W (e.g. water veg, fire regen) RGB Composite Image draining lakes B=Brightness Change (red channel) G=Greenness Change (green channel) W=Wetness Change (blue channel) Blue = B G W (e.g. forest succession, slump disturbance) Light Blue = B G W (e.g. veg growth over bare) fire gravel pit
Mean = 17 5. Results: Density of Growing Season Landsat Observations (1985-2012)
5. Results: Landscape Disturbance Examples Wildfires Landsat TC Trends (1985-2012) Dates indicated from GNWT fire mapping polygons
Landsat TC Trends (1985-2012) Inuvik Area B=Brightness Change (red channel) G=Greenness Change (green channel) W=Wetness Change (blue channel) 1968 fire Inuvik 2003 fire
Landsat TC Trends (1985-2012) Peel Plateau Thaw Slumps Trend trajectories related to decadal evolution of retrogressive slumps Recent 20m SPOT imagery
Landsat Trends (1985-2012) Fire and Shallow Lake Drainage Higher prevalence of lake drainage in post-fire areas likely related to permafrost degradation Landsat TC Trends (1985-2012) Yellow = B G W (e.g. water veg) GNWT ELC (2005)
Landsat Trends (1985-2012) Norman Wells Area (35km extent) Regenerating Seismic Lines Old Disturbance and Forest Succession Old Canol Road
1984 Landsat Ch. 3 Yellowknife Area RGB Composite Change Image (1985-2011) Dark Blue = B G W (e.g. veg water) Red = B G W (e.g. development) Light Blue = B G W (e.g. veg growth) Yellow = B G W (e.g. water veg) 2006 Landsat Ch. 3
Along Yellowknife Highway RGB Composite Change Image (1985-2011) Dark Blue = B G W Red = B G W New dev. 1971 Fire regen Light Blue = B G W Old hwy Great Slave Lake Drying wetlands Yellow = B G W
Drying and Greening of Wetlands on Great Slave Lake June 18, 2012 (south is up) Google Earth June 28, 2004. Landsat Wetness Trend Great Slave Lake water levels since 1934 Landsat Period
Mining Operations (New and Abandoned) 22 26 24 30
Landsat NDVI Trends 1984-2011 (positive trends in green, negative trends in red) Increasing Tundra Productivity / Shrub Growth Pouliot et al. 2009 AVHRR NDVI Trends
Increasing Tundra Productivity Near-Anniversary Date Landsat Images 8 Years Apart (RGB=SWIR TOA,NIR TOA,Red TOA, non-stretched, leafy vegetation appear green) July 25, 1992 July 23, 2000 Alder thicket visible in 1950
Validation: Will reacquire 1:2,000 Colour-Infrared Air Photos Captured in 1980 (Sims, 1983) Aug 6, 1980 Aug 8, 2013 Shrub Changes? Lichen Change? Alder thicket visible in 1950 90 m
Training Database Change Classes (718 polygons) Next Step: Developing Change Classification Products from Landsat Trends Training database Landsat VI Trends Binary Change Mask (Threshold) Decision Tree Classification Expert Decision Rules GIS Layers For Spatial Context Alder thicket visible in 1950 Change Classification Product
Changes Detected Over NWT Study Regions using Landsat Natural Wildfires and regen Thaw slumps Lake drainage and erosion Greening / increased growth of shrubs Succession of old disturbances and Vegetation flooding Anthropogenic Municipal developments Mining new footprint and regeneration of abandoned mines Highways new and regeneration of borrow pits
Thank You Support from: NWT CIMP NRCan TRACS Project Polar Continental Shelf Project