Morris highres asprs_pecora_final

Morris highres asprs_pecora_final

• 1.
  High Spatial ResolutionLand Cover Development for the Coastal United States Eric Morris (Presenter) Chris Robinson The Baldwin Group at NOAA Office for Coastal Management Nate Herold NOAA Office for Coastal Management
• 2.
  Coastal Change AnalysisProgram (C-CAP) • 25% of contiguous U.S., authoritative source for coastal landcover (30m moderate res and 1-5m higher res) • Coastal expression of the NLCD (National Land Cover Database) • NLCD is 90%+ C-CAP in coastal areas • Standard data and methods • Inventory of intertidal areas, wetlands and adjacent uplands • Updated every five years
• 3.
  High Resolution C-CAPLand Cover “Our goal is to provide consistent, accurate, nationally relevant data at a spatial scale more appropriate for support of increasingly detailed, site- specific, management decisions.” • Since 2006, direct response to customer demands – Uses the C-CAP Nat’l framework for producing local level data – Selected based on need and data availability • Developed through partnerships with private industry
• 4.
  Why Map ata Higher Resolution? Small geography of interest •Islands, counties, watersheds •management reserves Extraction of land cover components •Impervious Surfaces •Invasive species •Specific habitats Site specific issue •Local level analysis < 1m 1m to 5m 5m to 10m 10m to 30m Site Specific Mapping Application Specific Mapping Landsat SPOT SPOT (Pan) IKONOS SPOT (Pan) Quickbird IRS (Pan) Digital Aerial Cameras Mod Res C-CAP
• 5.
  Spectral Resolution • 4Band Imagery • Near Infrared, Spectral Derivatives, NIR Vegetation • Middle Infrared • Natural Color as ancillary data Leaf On, Tide controlled Leaf Off, no tide control
• 6.
  • Accuracy • Scale –Usually lower res • Vintage – Usually older Why needed? – Spectral data insufficient – Features are subdued at the time of acquisition • Sources – National Wetland Inventory (USFWS) – SSURGO Soils (USDA) – Lidar Ancillary Data
• 7.
  Lidar - DerivativesBare Earth DEM •Slope •Curvature •Wetlands and other vegetation types Digital Surface Model •Used with Bare Earth DEM •Normalized Digital Surface Model (nDSM)
• 8.
  Image Processing Considerations Highspatial res. ≠ easier = detail Increased spectral classes per thematic Class •Traditional (Pixel based) Classifiers •Noise and poor accuracy Segmentation •Network of homogenous areas • Image Objects •eCognition: Multi-resolution Segments (Baatz & Schape, 2000) Worldview2 Landsat
• 9.
  Hierarchical Approach (vs.All at Once) Major distinctions first •Woody vs. Herbaceous •Forest vs. Scrub Automated Classification •Classification and Regression Tree Analysis (CART) – Rule Set/Tree output – A lot of training data Spatial Modeling •Regional recursive rules Ex: If Object = Forest (class 10) and nDSM < 4m Then Scrub/Shrub (class 12) Manual Edits for unique & rare features
• 10.
  Change Detection Process Steps •Baselinedata •Identify areas (i.e. via Change Mask) •Collect training data •Classify change area •Insert into baseline map • Map “Change Only” areas • Instead of post classification • Object based approach • Methods guided by available imagery • (Niemeyer et al., 2008), (Duro et al.,2013)
• 11.
  Imagery Considerations
• 12.
  Change Detection Mean NIR– Class MeanDate 1 Land Cover High : 106.826 Low : -75.1246 Date 1 Date 2 Segments
• 13.
  Recap: High ResolutionChange Mapping
• 14.
  Questions Eric Morris [email protected] www.coast.noaa.gov/digitalcoast/data/ccaphighres