The National Land Cover Database: Methods to Improve Impervious Surface Data

Keywords: impervious surface, land cover
Session Type: Virtual Poster Abstract
Day: Friday
Session Start / End Time: 2/25/2022 03:40 PM (Eastern Time (US & Canada)) - 2/25/2022 05:00 PM (Eastern Time (US & Canada))
Room: Virtual 39

Abstract

The National Land Cover Database (NLCD) is the definitive Landsat-based, 30-meter resolution land cover dataset for the United States. NLCD provides spatial reference and descriptive data for thematic classes (e.g., urban, agriculture, and forest), as well as continuous data such as percent impervious surface. A continuing priority for the NLCD project is to improve the impervious surface data layers with each data release, while increasing automation and reducing hand-editing. The 2016 and 2019 versions of NLCD greatly enhanced the spatial accuracy and precision of roads and energy features through the use of ancillary datasets. Starting with the 2016 data release we also added an impervious descriptor layer that labels certain types of impervious surface (e.g., road type, wind turbines, and well pads) for pixels in “developed” classes. However, as we accelerate the frequency of data releases to a targeted 1-2 years it has become increasingly necessary to reduce and expedite the postprocessing of models that predict new impervious features. We are now building on our previous methods by applying new techniques (i.e., machine learning algorithms and spectral/temporal models created in Google Earth Engine), and incorporating additional ancillary datasets. Future plans also include collecting training data from high-resolution maps of impervious surface. These new technologies and datasets are being developed and acquired in cooperation with our Multi-Resolution Land Characteristics (MRLC) Consortium partners and with the expertise of other remote sensing scientists. The advancements will allow us to produce higher-quality data more frequently while reducing overall production time.