Description: <DIV STYLE="text-align:Left;"><DIV><DIV><P><SPAN>This dataset identifies the impervious surfaces within the State of Delaware including, but not exclusively, paved roads, dirt roads, buildings, sidewalks. These data, in addition to being a stand-alone dataset, were produced to aid the Delaware 2007 LU/LC Update.</SPAN></P></DIV></DIV></DIV>
Description: <DIV STYLE="text-align:Left;"><DIV><DIV><P><SPAN>The State’s first Impervious Surface (IS) dataset was created as a by-product of the previous vendor’s “change detection” image processing update methodology. The resulting impervious surface data, while not of the highest accuracy, was deemed an acceptable representation of IS in the state. The impervious surface data for the 2012 Update was produced to the same level as the State’s previous datasets, per the project requirements. The work was performed by Aerial Information Systems, Inc. (AIS), located in Redlands, California. The 2012 Impervious Surface data layer was created using a combination of Automated Feature Extraction (AFE) techniques and ArcGIS raster data editing tools. Approximately 2,000 image tiles provided complete coverage of the State. These images were composed of four bands (red, blue, green and IR) at a resolution of 1 foot (0.33 meter). The automated feature extraction involved a number of processing steps and the use of two software products: Genie Pro 2.4, by Obervera, Inc., was used to perform the automated feature extraction and Esri’s ArcGIS was used to perform image cleanup and editing tasks. Using Genie Pro 2.4, training sets and solution files were created and prepared on a representative sample of image tiles. Then the training sets and solution files were combined into one master solution file. Using a python script, all image tiles were processed with the master solution file to create new image tiles. Each of the new image tiles was reviewed to verify that processing step was successfully completed. Where processing problems occurred, the image tiles were re-processed. The analyst revised the original training sets and created new solution files to produce a new result image tile. This process was repeated until the processing errors were corrected. Using ArcMap raster processing tools, four raster catalogs were created which closely matched the extent of the 2007 Impervious Surface raster images. Each of these layers were made up of approximately 500 image tiles. The following series of processing steps were run on each of the four layers: The images were resampled from 0.33 meters to match the 1 meter resolution of the original 2007 IS data. Small mis-identified pixels were removed from each of the four images by running them through the following five processes: Majority Filter, Boundary Clean, Region Group, Set Null, and Nibble. The analyst performed a manual review and edit of each image using ArcMap and the ArcScan extension. The final Impervious Surface geodatabase was created and each of the four regional impervious surface raster data files were imported to the geodatabase.</SPAN></P></DIV></DIV></DIV>
Copyright Text: Mapped by Aerial Information Systems, Inc. of Redlands, California for the Delaware Office of State Planning
Description: <DIV STYLE="text-align:Left;"><DIV><DIV><P STYLE="margin:0 0 0 0;"><SPAN>The impervious surface data created by AIS for the 2017 Update was produced to the same level as the State’s previous datasets, per the project requirements. The 2017 Impervious Surface data layer was created using a combination of ArcGIS’ Image Classification and raster data editing toolsets. The State provided AIS with their 2017 state-wide, 4-band image (in SID format) which was clipped using the boundary of each production modules (45 in all). Each production module image was resampled to a resolution of one meter; then Arcmap’s Segment Mean Shift function was using to produce a segmented image. A training set was created for each production module segmented image and Arcmap’s Maximum Likelihood Classification tool was used to produce a classified image with three pixel values ( 1 = Impervious Surface, 2 = Pervious, and 3 = Shadow). Each classified image was further processed to add water pixels, derived from NHD waterbody polygons, and road pixels, based off of the State’s centerline data. Using Arcmap’s Reclass tool, each image had the Pervious class and Water class combined and the Impervious class and Road class combined. The shadow class processed by converting all shadow pixels to noData, then using the image processing tool Nibble, the noData values were converted to either Pervious or Impervious values. The classified images, now with two pixel values (value 3 was removed), were now ready for visual review and editing.</SPAN></P></DIV></DIV></DIV>
Copyright Text: Mapped by Aerial Information Systems, Inc. of Redlands, California for the Delaware Department of Transportation
Description: <DIV STYLE="text-align:Left;"><DIV><DIV><P><SPAN>This map was derived from a 12-class land-cover product developed for the Chesapeake Bay Watershed. It was extracted from a map showing land-cover conversion in the Chesapeake Bay Watershed during the period 2017/2018-2021/2022. The change-detection map was modeled with an existing 2017/2018 (Time Period 2, or T2) land-cover map containing the following classes: Water, Emergent Wetlands, Tree Canopy, Scrub\Shrub, Low Vegetation, Barren, Impervious Structures, Other Impervious, Impervious Roads, Tree Canopy Over Impervious Structures, Tree Canopy Over Other Impervious, and Tree Canopy Over Impervious Roads. Using object-based image analysis mapping techniques, the T2 map was first refined, where necessary, using all remote-sensing imagery and GIS datasets available for 2021\2022, including LiDAR, multispectral imagery, and thematic layers (e.g., roads, building footprints). The refined T2 map was then compared to imagery and GIS data corresponding to the period 2021/2022 (Time Period 3, or T3), identifying features that had changed during the analysis interval. Each change combination was assigned to a unique category that indicated both the "from" and "to" classes, enabling monitoring of specific types of change. Features that had not changed during the analysis interval were also incorporated into the change map, making it possible to reconstitute full land cover for either T2 or T3. Draft output was manually reviewed and edited to eliminate obvious errors of omission and commission, focusing on the most important types of landscape change: 1) tree-canopy loss; and 2) impervious-surfaces gain. The final map contained 79 change classes plus the original 12 classes for features that had not changed. The specific analysis periods for the watershed's 7 states\districts were: Delaware, 2018-2021; District of Columbia, 2017-2021; Maryland, 2018-2021; New York, 2017-2022; Pennsylvania, 2017-2022; Virginia, 2018-2021; and West Virginia, 2018-2022. The 12-class extract of the change-detection map was created by reassigning all change classes to their T3 status.</SPAN></P><P><SPAN>For specific use in land-use\land-cover (LULC) analyses for Delaware, the T3 map produced for the Chesapeake Bay Watershed was reclassifed into two classes: Impervious (1); and Pervious (2). It was reviewed and edited against 2022 leaf-off imagery to update it to 2022 conditions. </SPAN></P><P><SPAN>This map was based on a draft version of Chesapeake Bay Watershed land cover. A publicly-available version of the final Chesapeake Bay map will be available in autumn 2024, through the Chesapeake Conservancy (https://www.chesapeakeconservancy.org/).</SPAN></P></DIV></DIV></DIV>
Copyright Text: The University of Vermont Spatial Analysis Laboratory developed this map in cooperation with the Chesapeake Conservancy and the U.S. Geological Survey.