Skip to main content
2023 ESA Annual Meeting Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Back

Contributed Talk

Session: : Remote Sensing And Image Analysis 2

COS 268-3 - Annual Science products from National Land Cover Database Tree Canopy Cover – methods and lessons learned

Thursday, August 10, 2023
10:30 AM – 10:45 AM PDT
Location: E144

    Presenting Author(s)

  • BR

    Bonnie Ruefenacht

    Remote Sensing Analyst
    RedCastle Resources - Geospatial Technology and Applications Center
    Salt Lake City, Utah, United States

    • Coauthor(s)

  • JH

    Josh Heyer

    Remote Sensing Analyst
    RedCastle Resources - Geospatial Technology and Applications Center
    Salt Lake City, Utah, United States

  • IH

    Ian Housman

    Remote Sensing Analyst
    RedCastle Resources - Geospatial Technology and Applications Center
    Salt Lake City, Utah, United States

  • KS

    Karen Schleeweis

    Research Ecologist
    USDA Forest Service Rocky Mountain Research Station, Delaware, United States

  • SB

    Stacie Bender

    Physical Science Information Specialist
    USDA Forest Service Geospatial Technology and Applications Center
    Salt Lake City, Utah, United States

  • KM

    Kevin Megown

    RMIM Program Manager
    USDA Forest Service Geospatial Technology and Applications Center
    Salt Lake City, Utah, United States

Abstract: Large-area annual mapping programs are becoming more common. The USDA Forest Service produced the National Land Cover Database’s (NLCD) tree canopy cover (TCC) images for the years 2001, 2011, and 2016. While these data provide valuable information on forest dynamics, the infrequent release limits their effectiveness. The development of cloud computing services, such as Google Earth Engine, and the accessibility to a plethora of optical remote sensing data, such as Landsat and Sentinel 2, now make it possible to produce annual NLCD TCC datasets. TCC random forest models were calibrated with photo-interpreted 2011 TCC estimates. The LandTrendr temporal-smoothing algorithm was applied to an annual Landsat and Sentinel2 medoid composite time-series. This served as the primary predictor dataset, along with topographical information. The TCC response data were intersected with the predictor datasets creating model calibration datasets. Thirty percent of the model calibration dataset was withheld from model development and used for error assessment. Applying the holdout datasets returned a root mean square error of 12.9 percent TCC. Additionally, an analysis of stable pixels through time illustrated modeled TCC values remained relatively stable through time using a single response year dataset. While the 2021 NLCD TCC release will not include annual products, the raw annual Science products will be distributed. This is the first step to delivering annual NLCD TCC products in the future.