King County Impervious/Impacted Surface Interpretation based on 2000 imagery

Metadata also available as

Frequently-anticipated questions:

What does this data set describe?

King County Impervious/Impacted Surface Interpretation based on 2000 imagery
Impervious and heavily-impacted landcover surface interpretation derived from spectral classification of a three-part multispectral image dataset. Limitations to classification extraction required substantial editing of initial results to improved data quality and reach required accuracy (approximately 90%).Includes mostly surfaces with high to completed impermeability, such as concrete, asphalt, roofing materials and other sealed surfaces.Examples include:(a) building roof tops regardless of composition or construction,(b) roadways, highways and parking lots constructed of concrete or asphalt,(c) parking areas with a high density of parked vehicles as represented by the imagery,(d) sidewalks, pedestrian walkways and malls constructed of concrete, asphalt or brick,(e) other prepared surfaces such as bicycle paths, tennis courts and running paths.In addition to spectral extraction limitations, visual interpretation during editing resulted in categorization encompassing both impervious surfaces as well as unvegetated, high-reflectance surfaces that are not covered by impervious materials.Often these surfaces are the result of anthropic operations that have resulted in removal of vegetation and compaction of the surface. This "impacted surface" category includes gravel roads and parking lots, equipment staging areas, and sites prepared during pre-construction. However it may also include naturally occurring sites that had little or no vegetation, and not necessarily been altered by man. The original spectral interpretation had a high commission error, meaning areas clearly not impervious by definition, such as water bodies, streams, shadows, and dark (wet) grass areas, were incorporated into the interpretation. This error was intentionally allowed in order to maximize the inclusion of true impervious areas that are of small areal extent or narrow width or may have been missed otherwise due to marginal spectral signature. Most of the value-added editing focused on removing the gross commission error, particularly that of large areal extent. Not all commission error has been removed. Multispectral data used in this interpretation was acquired under leaf-on conditions.Impervious surfaces such as rooftops, roadways and parking lots are often partially obscured by the leaf cover. These areas are not interpreted as impervious and are part of the omission error. To reduce the omission error, a high accuracy road centerline dataset was incorporated into the original interpration to 'fill-in' roads where leaf overhang may have masked the surface.
These impervious surface distributions were created by Marshall and Associates, Inc. as part of a multiple part effort to interpret impervious surface categories from high-resolution multispectral imagery.
The interpretation was derived from three different multispectral imagery datasets covering the entire project area. The datasets had different spectral characteristics and resolutions resulting in differences in the output, both in quality, definition and resolution.
Vashon Island was interpreted from 4-band, 12-bit rgb-infrared, 1.64 meter ground sample distance (gsd) data. Western King County and southwestern Snohomish County from Range 03 to Range 07 was interpreted from 3-band, 8-bit rg-infrared, 2 foot gsd data. Eastern King County from Range 08 to Range 11 was interpreted from 4-band, 16-bit, rgb-infrared 4-meter data.
At the primary user-access level (township-range, idxptrmbr tiling level) the Vashon and western King County data are stored at 2-foot grids. The eastern King County data is stored at 15-foot grids.
Background Project Review documents:
Classification and Product Specifications:
Imagery Assessment and Processing Procedures:
  1. How should this data set be cited?

    County, King, 200405, King County Impervious/Impacted Surface Interpretation based on 2000 imagery: King County, King County, WA.

    Online Links:

    • None

    This digital data was created by Marshall and Associates, Inc,under contract to King County, contract number: T01393.
    Marshall and Associates 1603 Cooper Point Rd NW Olympia, WA 98502-8325
    In 2002, the edited Vashon Island portion of impervious surface was edited and posted to Spatial Data Warehouse as Version 1. To facilitate access to other portions of King County the full dataset was posted without edits in late 2003. Editing continued through early 2004 when the current version (3) was posted.

  2. What geographic area does the data set cover?

    West_Bounding_Coordinate: -122.4505
    East_Bounding_Coordinate: -121.6212
    North_Bounding_Coordinate: 47.9738
    South_Bounding_Coordinate: 47.1155

  3. What does it look like?

  4. Does the data set describe conditions during a particular time period?

    Beginning_Date: 24-May-2000
    Ending_Date: 23-Sep-2000
    Currentness_Reference: Imagery Acquisition Date

  5. What is the general form of this data set?

    Geospatial_Data_Presentation_Form: Raster digital data

  6. How does the data set represent geographic features?

    1. How are geographic features stored in the data set?

      Western King County, Southwestern Snohomish County, and Seattle WA
      This is a Raster data set.

    2. What coordinate system is used to represent geographic features?

      Grid_Coordinate_System_Name: State Plane Coordinate System
      SPCS_Zone_Identifier: 4601 (FIPS)
      Standard_Parallel: 47.5
      Standard_Parallel: 48.7333
      Longitude_of_Central_Meridian: -120.8333
      Latitude_of_Projection_Origin: 47.0
      False_Easting: 500,000 M
      False_Northing: 0

      Planar coordinates are encoded using Coordinate Pair
      Abscissae (x-coordinates) are specified to the nearest Not determined
      Ordinates (y-coordinates) are specified to the nearest Not determined
      Planar coordinates are specified in U.S. Survey Feet


  7. How does the data set describe geographic features?

    txxryy_imp.pat or txxryy_imp.dbf
    Feature attribute table for each respective coverage or shapefile. The primary user-access tile level is the township-range tiling scheme designated by txxryy where xx is the twnshp and yy is the rge. 'imp' implies impervious layer. (Source: King County internal coding. Classification schema and definitions were determined through consensus of a working group comprised of representatives from multiple county departments. All King County raster-based data follows this tiling/naming convention.)

    Impervious/impacted surface (Source: Remap of grid-values of 1,2,5 to single value of 1)

    1Impervious/impacted surface

    Township-Range integer identifier (Source: Public Land Survey System)

    Two-digit township cocantenated with two-digt range

    Uppercase township-range map label (Source: Public Land Survey System)

    Of form, T##R##

    Lowercase township-range map label (Source: Public Land Survey System)

    Of form, t##r##

    Integer value for township only (Source: Public Land Survey System)

    Two digit integer

    Range number only (Source: Public Land Survey System)

    One or two digit integer

    Feature attribute table for each respective grid. The primary user-access tile level is the township-range tiling scheme designated by txxryy where xx is the twnshp and yy is the rge. 'imp' implies impervious layer. RRR is the resolution of the grid, 002 = 2 foot for Vashon Island and western King County, 015 = 15 feet for eastern King County. (Source: King County internal coding. Classification schema and definitions were determined through consensus of a working group comprised of representatives from multiple county departments. All King County raster-based data follows this tiling/naming convention.)

    Landcover value (Source: User-coded)

    1Impervious surface (high confidence)
    2Impervious surface (low-confidence/mixed classes)
    515 foot road buffer conditionally added where grid not intially 1 or 2.

    The impervious/impacted surface interpretation is stored as native ArcInfo grids with VAT feature attribute table. It is also stored as simplified coverages and shapefiles with a PAT, and dbf table, respectively.
    The VAT table stores the maximum detail in the form of three classes, high-confidence impervious, low-confidence impervious and roadway values included to reduce commission error.
    The PAT (and shapefile dbf) contain a single attribute remapped from the grid-code which combines all impervious classes from the grids into a single item value. This reduces the number of individual polygons generated from the grid to poly conversion while maintaining the full impervious/impacted interpretation.
    Entity_and_Attribute_Detail_Citation: Marshall and King County

Who produced the data set?

  1. Who are the originators of the data set? (may include formal authors, digital compilers, and editors)

  2. Who also contributed to the data set?

  3. To whom should users address questions about the data?

    Mike Leathers
    King County GIS Center
    GIS Data Coordinator
    201 S Jackson Street
    Seattle, WA 98104

    206-263-4867 (voice)

Why was the data set created?

Provide a project-wide (King County and southwest Snohomish county) generalized interpretation of imperviuos and impacted surface. By exclusion, null value areas imply vegetated or water. Can to be used, within limitations of accuracy, to determine the distribution of these types of surfaces across the project area and to make first-order areal calculations. This dataset is coincident in extent with other ESA/SAO datasets acquired during this three-year project scope, such as lidar elevation data and orthophotography. Opportunities exist for synergistic applications.

How was the data set created?

  1. From what previous works were the data drawn?

    Marshall (source 1 of 1)
    Marshall, Inc, 200205, Landcover classification of Multispectral imagery for King County and Southwestern Snohomish County: Marshall, Inc, King County.

    Other_Citation_Details: None
    Type_of_Source_Media: Raster digital data
    Source_Contribution: Original contract deliverables

  2. How were the data generated, processed, and modified?

    Date: 01-May-2002 (process 1 of 10)
    For Vashon Island see process document "<>"
    For western King County see process document "<>"
    For eastern King County see process document "<>"

    Data sources used in this process:

    • isurfemer_xx, isurfdais_xx, isurfikonos_xx

    Date: 2003 (process 2 of 10)
    Grid data delivered by vendor was mosaiced from original tiles. Data was retiled into township-range tiling scheme. Best-available road network database was buffered to 15-feet and merged to grid tiles. Conditional statement was used so that road value (id =5) was added only where there was a null or zero grid value.

    Data sources used in this process:

    • isurfemer_xx, isurfdais_xx, isurfikonos_xx

    Date: 2004 (process 3 of 10)
    Prepared grids were imported into ERDAS as img files. Split screen arrangement was used to display visible version of source imagery along side the classified grid.
    Several editors were employed. One lead editor was used to instruct other editors to maintain as great as consistency between tiles as possible. The main purpose of the editing was to remove gross commission errors.
    Based on interpretation of the associated imagery areas of commission were outlined as Areas of Interest (AOI) and recoded as non-impervious (i.e., set to NULL).
    General guidelines and observations used in editing:
    1. Most areas of heavy forest appeared as impervious and were recoded.
    2. Rivers and streams also appeared as impervious and were recoded.
    3. Bare earth and fallow agricultural fields show mixed to full impervious and were recoded.
    4. Areas of obvious construction were not recoded.
    Ancillary imagery was sometimes employed to reconcile ambiguous situations. If area still remained ambiguous and could not be edited per rules, then 50% of area was recoded. In areas of dense urbanization, this recoding average would be reduced, leaving more of the area as impervious.
    Large bodies of water had previously been removed by the vendor employing a coincident Landsat TM interpretation. This coarse editing technique errored towards not clipping the water/shore boundary but in some instances this could not be avoided. Again not omission correction was applied. In areas water/shore areas where additional editing was done, wooden/floating docks were removed. However, permanent docks/piers were not recoded.
    Other areas of omission error, shadowed or shade-covered impervious features, (e.g., features in shadow of tall building) were not corrected.

    Data sources used in this process:

    • txxryy_impRRR ArcInfo grid

    Date: 2004 (process 4 of 10)
    After editing was completed in Erdas, ArcView 3.x spatial analyst was used to convert the img file back to an ArcInfo grid. This grid was then processed to convert any zero values (as recoded in Erdas) to NULLs so that the only valid values left in the grid would be 1, 2, and 5.
    Projection header was defined per King County standards and grid was posted to Spatial Data Warehouse.
    Copy of the grid was remapped to new grid collapsing all non-null values to a value of 1, representing impervious/impacted surface regardless of original coded value.
    Grid converted to polygon file using GRIDPOLY command.
    Polygon coverage converted to shapefile using SHAPEARC command with Polygon argument
    ArcView used to display and quality assess the shapefile interpretation and the original grid together. Shapefile reselected for grid-code <> -9999 to remove all null islands created during the GRIDPOLY conversion.
    Reselect shapefile moved to Spatial Data Warehouse, projection file added, and final inspection made.

    Data sources used in this process:

    • txxryy_imp002.img

    Date: Mar-2005 (process 5 of 10)
    During GIS application use of the dataset, the following quality issues were noted and addressed as indicated:
    1. Some shapefiles contained attribute (item) names that were not consistent across all files.
    A python script was used to query all shapefile table objects to identify the item inconsistency. As the grid-code item is the only required field for the data all fields after the system-defined and before grid-code were dropped from the shapefiles. This was more effective than renaming the items and reduced the file size.
    2. T25R06 grid tile did not cover the full extent of that tiles MBR.
    The subject township and all adjacent townships were merged and clipped to the correct MBR. The grid was processed through to shapefile using the same process steps as described in the original process stanzas
    3. Projection headers of some of Grid tiles was incorrect
    The grid headers were updated using the ArcGIS Workstation commandline Projectdefine command with the following parameters. Projection Stateplane Zone 5601 Datum HPGN Units FEET Spheroid GRS1980
    This corrected the erroneously defined zone of 5610 and 5106.
    All grids and shapefiles were reinspected for correct extent, attribution and attribute values.

    Date: 13-Apr-2008 (process 6 of 10)
    The reg_imp shapefile, derived from the reg_imp030 grid, had drawing errors when used in ArcMap.The layer would display all feature polygons when zoomed out to large extents, but failed to draw all polygons when zoomed in. This was likely due to the large, contiguous polygons of impervious (GRID_CODE = 1) that extended well beyond the draw window. To break up these large polygons, the shapefile was converted (SHAPE2POLY) to a coverage and then UNIONED with a coverage version of IDXPTRMID, the township-range tiling index. The resulting union created intersect lines where a township boundary occurred, thus breaking up the large polygons. Although this created intersect lines in the overall layer, it now displays properly in ArcMap. The SHAPE2POLY and UNION actions also added several auto-generated attributes as well as the user attributes from the IDXPTRMID coverage. The unnecessary auto-attributes were dropped so as to match the original schema of the REG_IMP.SHP shapefile. However the user attributes describing the township and range locations were left in the attribute table to allow query and selection by township-range, if necessary.

    Date: Oct-2008 (process 7 of 10)
    The database was moved from individual township-range grids to a seamless SDE GDB raster mosaic at 2-foot resolution. This was part of the process of migrating large raster databases to the SDE environment so that any updates could be made to the entire database and then subset products, such as the township-tiled grids, could be extracted from the master.

    Date: Oct-2008 (process 8 of 10)
    The schema of the SDE GDB raster mosaic was updated.
    The original schema contained a single item VALUE which encoded one of three domain values: 1 = high confidence impervious surface, 2 = lower confidence impervious surface, and 5 = added from the transportation network layers (early version of TRANS_NETWORK).
    As part of the plan to update the 2000 impervious surface product to a 2007 product, the schema was modified so that the VALUE field encodes an integer that reflects the data source:
    1. Value = 1 indicates impervious surface comes from 2000 DAIS imagery 2. Value = 2 indicates impervious surface comes from 2000 Emerge imagery 3. Value = 3 indicates impervious surface comes from 2001 Ikonos imagery 4. Value = 4 indicates impervious surface comes from 2000 TRANS_NETWORK geometry.
    An additional item IMP was added and encoded to 1 so that all cells, regardless of source, are indicated as impervious. This also makes it possible to symbolize the grid by a single impervious value.

    Date: 1906 (process 9 of 10)
    External city shapefile data was added with new source Values: Value 5, Woodinville buildings dated 2004. Value 6, Seattle buildings dated 1999 Value 7, Tukwila parking lots dated 1999 Value 8, Tukwila Buildings dated 1999 Value 9, SeaTac buildings dated 2003 Value 10, Redmond buildings dated 1999 Value 11, Kirkland impervious dated 2003 Value 12, Kirkland buildings dated 2002 Value 13, Kent impervious dated 1999 Value 14, Kent buildings dated 1999 Value 15, Federal Way impervious dated 2005 Value 16, Federal Way buildings dated 2002 Value 17, Des Moines buildings dated 2002 Value 18, Bothell public buildings dated 2004 Value 19, Bothell commercial buildings dated 2004 Value 20, Bellevue buildings dated 2004

    Date: Not complete (process 10 of 10)
    Erased errors in following townships: T28R04 T28R05 T27R03 T27R04 T27R05 T27R06 T26R03 T26R04 T26R13 T25R03 T25R13 T24R11 T21R02 T19R11

    Data sources used in this process:

    • 2007 Pictometry

  3. What similar or related data should the user be aware of?

How reliable are the data; what problems remain in the data set?

  1. How well have the observations been checked?

    The attributes assigned to the Grid-code item represent the best spectral interpretation of the pixel to one of the predefined classes, plus value-added editing. Classification accuracy goal was specified as:
    Impervious surface 90%
    On the average, this accuracy goal was obtained after editing of tiles was completed. Editing procedures was standardized, however, accuracy varies by township. Some of this variation is also due to quality of original interpretation which was strongly influenced by variations in imagery spectral quality.
    Quality evaluation consisted of correlating known conrol points (i.e., impervious or not-impervious) with both the original grid interpretation and the value-added (i.e., edited) grid interpretation. The results of the analysis show the number of correct classifications, incorrect classifications and improvement in classification due to editing.
    The results of an accuracy analysis of 1855 checkpoints can be view in the following spreadsheet table "<>"
    The AccuracyStat worksheet shows the results of the analysis and accuracy calculations. The AccuracySites worksheet provides the x,y locations of all control points with the interpreted surface type used in the quality assessment analysis The EditGraphic worksheet provides a link to a jpg image of each township's edits to give some idea of the amount of commission error removed from each tile.
    The analysis was performed by township tile. A graphic representation of the average impervious classification accuracy by township tile can be seen in: "<>"

  2. How accurate are the geographic locations?

    Impervious surface locations are only accurate to the accuracy of the underyling imagery.
    See "<>" for Vashon Island area
    See "<>" for western King County
    See "<>" for eastern King County

  3. How accurate are the heights or depths?

    Not Applicable

  4. Where are the gaps in the data? What is missing?

    There are no gaps in the data. All cells were classified as one of the values found in the entity section.
    The entire project area is covered by one or other of the datasets. The bounds of the township-range tiling scheme did not agree exactly with the extents of the underlying imagery. The western King County imagery fell just short of the eastern boundary of the Range 07 tiles. Required portions of the eastern King County 15-foot data were resampled to 2-feet and appended to those Range 07 tiles.
    A similar situation occurred in the Range 08 tiles where the eastern King County data fell slightly short of the western edge of the tiles. In these cases the western King County 2-foot data was resampled to 15-feet to append to those tiles.
    The results is that, to the extent of the project area, all tiles contain a complete impervious surface interpretation.
    Township-Ranges T27R04, T27R05, T28R04, T28R05 were not covered by the same road network that was used to 'patch-in' road features. Townships east of Range 08, inclusive, were updated with the road network. However the alignment between the road network and the original impervious grid was not as good due to the less accurate imagery used in the impervious interpretation.

  5. How consistent are the relationships among the observations, including topology?

    To accomodate the differences in pixel resolution of the spectral imagery, the Vashon island data was resampled from 1.64 meter to 2 foot to make it consistent with the western King County data. The eastern King County data was resampled to 15-ft to make it an even integer value and to improve consistency with other datasets in this area.

How can someone get a copy of the data set?

Are there legal restrictions on access or use of the data?

Online access is restricted to authenticated King County users. Others may obtain the data through KC GIS Center Client Services.
King County data are made available with the understanding that they shall be used exclusively by the obtainer or his/her authorized agents. Digital produce may not be reproduced or redistributed in any form or by any means without the express written authorization of King Co. GIS Center. This dataset is not accurate enough to support parcel level analyses and mapping. It is suitable for watershed level uses only.
No Representations Marshall and Associates, Inc. (hereinafter MARSHALL) and its s upplier make no representations or warranties with respect to any imagery or products derived from imagery supplied to a Client with the exception of those specified in a legal contract agreed upon both by MARSHALL and the Client.
Limited Warranty The Client shall indemnify and hold harmless MARSHALL and its supplier from any circumstances that arise from the use of imagery or information supplied by MARSHALL for any other purpose or use besides that expressly outlined in the MARSHALL /Client contract
Limitation of Liability Neither MARSHALL nor its supplier, their affiliates, officers, directors, shareholders, agents, and employees shall be liable for any loss or damage suffered by the Client, or any party claiming through the Client, as a result of or related to the provision by MARSHALL of any data, information, or images.
Reselling and Distributing Data The Client shall not resell to any third party any data, information, or images provided by MARSHALL unless the Client obtains from such third party a disclaimer of representations and a confirmation of limitations of liability and the Client shall ensure that MARSHALL and its supplier is a beneficiary of such disclaimers and limitations (even if only indirectly named as a supplier of the data, information, or images). The Client shall indemnify and hold harmless MARSHALL and its supplier from and against any claims made by third parties to whom the Client resells the data, information, or images supplied by MARSHALL.

  1. Who distributes the data set? (Distributor 1 of 1)

    Cheryl Wilder
    King County GIS Center
    Training Coordinator
    201 S. Jackson St, Room 706
    Seattle, WA 98104

    206-263-5220 (voice)

  2. What's the catalog number I need to order this data set?

    Internal King County users can access the data through: Township-Range (idxptrmbr) products: (see "<>" 1. "<\\gisdw\kclib\plibrary3\idxptrmbr\lndcov_imp\>" as txxryy_imp002 (2-foot grids), as txxryy_imp shapefiles created from grids for Vashon and western King County (Ranges 02-07), 2. "<\\gisdw\kclib\plibrary3\idxptrmbr\lndcov_imp\>" as txxryy_imp015 (15-foot grids). as txxryy_imp shapefiles created from grids for for eastern King County, Zone-level (idxpzone) mosaics: zxx_imp004 and zxx_imp015, 4-foot and 15-foot, respectively, grids, and zxx_imp.shp shapefile derived from 15-foot grid. The four foot grid is provided as a consistent, single-resolution interpretation that covers the entire county. It degrades the resolution, thus accuracy, of the western King County interpretation available in the township tiles. It displays the eastern King County data at 4-foot resolution, but this is simply resampled from 15-feet, and does not represent an improvement in data accuracy or quality. The shapefile zxx_imp is derived only from the 15-ft grid. A shapefile from the 4-foot grid was not generated due to performance considerations. KeyRegion mosaic "<\\gisdw\kclib\plibrary3\keyregion\landcover\>" contains reg_imp030, a grid resampled from zone-level grids to 30 foot resolution. This serves only as a general background interpretation for regional analysis and mapping applications. A shapefile reg_imp.shp is extracted from this grid.

  3. What legal disclaimers am I supposed to read?

    King County disclaims any warranty of use of any digital product or data beyond that for which it was designed.

  4. How can I download or order the data?

Who wrote the metadata?

Last modified: 03-Jun-2004
Metadata author:
Mike Leathers
King County GIS Center
GIS Data Coordinator
201 S Jackson St
Seattle, WA 98104

206-253-4867 (voice)

Metadata standard:
FGDC Content Standards for Digital Geospatial Metadata (FGDC-STD-001-1998)

Generated by mp version 2.9.6 on Wed Apr 10 12:34:02 2013