snippet:
|
2022 Land cover data for all urban areas in the State of Wisconsin, as identified in the 2010 Census, as well as adjacent lands. |
summary:
|
2022 Land cover data for all urban areas in the State of Wisconsin, as identified in the 2010 Census, as well as adjacent lands. |
accessInformation:
|
Wisconsin Department of Natural Resources, Division of Forestry (Bob Smail, Research Scientist) |
thumbnail:
|
|
maxScale:
|
5000 |
typeKeywords:
|
[] |
description:
|
<DIV STYLE="text-align:Left;"><DIV><DIV><P><SPAN>2022 land cover data for all urban areas in the State of Wisconsin, as identified in the 2010 Census, as well as adjacent lands. Pixels have a 1-m2 resolution and use the below values:</SPAN></P><P STYLE="margin:0 0 0 0;"><SPAN>1 – Tree</SPAN></P><P STYLE="margin:0 0 0 0;"><SPAN><SPAN>2 – Turf/grass</SPAN></SPAN></P><P STYLE="margin:0 0 0 0;"><SPAN><SPAN>3 – Impervious</SPAN></SPAN></P><P STYLE="margin:0 0 0 0;"><SPAN><SPAN>4 – Water</SPAN></SPAN></P><P STYLE="margin:0 0 0 0;"><SPAN><SPAN>5 – Indeterminable</SPAN></SPAN></P><P STYLE="margin:0 0 0 0;"><SPAN /></P><P STYLE="margin:0 0 0 0;"><SPAN>The classification was derived from imagery collected by the U.S. Department of Agriculture’s National Agriculture Imagery Program in 2022 using a supervised machine learning classifier. The land cover data was produced by ensembling two models that classified each pixel based on its spectral signature. The full landcover model was designed to capture all landcover classes whereas the second model was specifically tuned to identify only trees. The second model was created to improve the full landcover model accuracy because it tended to overpredict trees. The ensemble model classified ‘indeterminable’ pixels where the tree model predicted trees not included in the tree-specific model. A statewide training and test dataset composed of tree, turf, building, road, and water data was utilized to build the model. Quality analysis of the results showed that there were no systematic biases based on geography or community size.</SPAN></P><P><SPAN>For more information contact: </SPAN><SPAN STYLE="font-weight:bold;">Dan Buckler, Wisconsin DNR Urban Forest Assessment Specialist, </SPAN><A href="mailto:daniel.buckler@wisconsin.gov:25" STYLE="text-decoration:underline;"><SPAN STYLE="font-weight:bold;">daniel.buckler@wisconsin.gov</SPAN></A></P></DIV></DIV></DIV> |
licenseInfo:
|
<DIV STYLE="text-align:Left;"><DIV><DIV><P STYLE="margin:0 0 0 0;"><SPAN><SPAN>Because of the wide geographic spread of this classification and the need to mosaic together photographs from different times of the year, times of the day, and from different angles, there are many sources of error inherent in the underlying source imagery. These issues include significant discrepancies in shadow angle, direction and intensity, and wide variations in phenology or climate based on the underlying geography or habitat. The use of the “indeterminable” value reflects this uncertainty for many pixels.</SPAN></SPAN></P><P /><P STYLE="margin:0 0 0 0;"><SPAN><SPAN>It is not recommended to use these and similar data for change detection; that is – the comparison from one time period to another. There is too much uncertainty within a given year’s imagery, and that uncertainty is exacerbated when comparing multiple years. </SPAN></SPAN></P><P /><P STYLE="margin:0 0 0 0;"><SPAN>Furthermore, this classification was trained on urban elements – trees, grass, and impervious surfaces in highly developed spaces; the fields and wetlands that are sometimes on urban peripheries were not part of this study’s focus and thus there may be discrepancies in those types of spaces between the model and reality. Use caution before conducting any additional analyses.</SPAN></P></DIV></DIV></DIV> |
catalogPath:
|
|
title:
|
Urban Land Cover 2022 |
type:
|
|
url:
|
|
tags:
|
["tree canopy","land cover","urban","forestry","impervious"] |
culture:
|
en-US |
name:
|
|
guid:
|
|
minScale:
|
625000 |
spatialReference:
|
|