20230401
09321400
1.0
ISO 19139 Metadata Implementation Specification
FALSE
science_se_CONUS_2017_v2021-4.tif
file://science_se_CONUS_2017_v2021-4.tif
Local Area Network
-130.23282801589895
-73.59459648889016
22.07673063066848
48.70739591304975
1
002
PROJCS["Albers Conical Equal Area",GEOGCS["WGS 84",DATUM["WGS_1984",SPHEROID["WGS 84",6378137,298.257223563,AUTHORITY["EPSG","7030"]],AUTHORITY["EPSG","6326"]],PRIMEM["Greenwich",0],UNIT["degree",0.0174532925199433,AUTHORITY["EPSG","9122"]],AUTHORITY["EPSG","4326"]],PROJECTION["Albers_Conic_Equal_Area"],PARAMETER["latitude_of_center",23],PARAMETER["longitude_of_center",-96],PARAMETER["standard_parallel_1",29.5],PARAMETER["standard_parallel_2",45.5],PARAMETER["false_easting",0],PARAMETER["false_northing",0],UNIT["metre",1,AUTHORITY["EPSG","9001"]],AXIS["Easting",EAST],AXIS["Northing",NORTH]]
2
Area
Percent tree canopy cover standard error
Image
0.0
65534.0
Percent
16
2
Area
Image
TRUE
DEFLATE
1
GTiff
TRUE
discrete
UInt16
65535
20230401
09321400
20230401
12010700
150000000
5000
FGDC
Kevin Megown
USDA Forest Service Geospatial Technology and Applications Center (GTAC)
Program Lead: Resource, Mapping, Inventory and Monitoring
801-975-3826
801-975-3478
125 S. State Street, Suite 7105
Salt Lake City
UT
US
sm.fs.tcc@usda.gov
84138
0800 - 1600 MT, M – F
20230401
ArcGIS Metadata
1.0
Kevin Megown
USDA Forest Service Geospatial Technology and Applications Center (GTAC)
Program Lead: Resource, Mapping, Inventory and Monitoring
801-975-3826
801-975-3478
125 S. State Street, Suite 7105
Salt Lake City
UT
US
sm.fs.tcc@usda.gov
84138
0800 - 1600 MT, M – F
Downloadable data
ONLINE_LINKAGES
Raster Dataset
USFS CONUS 2017 Percent Tree Canopy (Standard Error of Science Version)
20230401
v2021-4
raster digital data
USFS Percent Tree Canopy Cover (Standard Error of Science Version)
none
USFS Percent Tree Canopy Cover (Standard Error of Science Version)
Kevin Megown
USDA Forest Service Geospatial Technology and Applications Center (GTAC)
Program Lead: Resource, Mapping, Inventory and Monitoring
801-975-3826
801-975-3478
125 S. State Street, Suite 7105
Salt Lake City
UT
US
sm.fs.tcc@usda.gov
84138
0800 - 1600 MT, M - F
Kevin Megown
The USDA Forest Service (USFS) builds two versions of percent tree canopy cover data, in order to serve needs of multiple user communities. These datasets encompass conterminous United States (CONUS), Coastal Alaska, Hawaii, and Puerto Rico and U.S. Virgin Islands (PRUSVI). The two versions of data within the v2021-4 TCC product suite include: The initial model outputs referred to as the Science data; And a modified version built for the National Land Cover Database and referred to as NLCD data. The Science data are the initial annual model outputs that consist of two images: percent tree canopy cover (TCC) and standard error. These data are best suited for users who will carry out their own detailed statistical and uncertainty analyses on the dataset, and place lower priority on the visual appearance of the dataset for cartographic purposes. Datasets for the nominal years of 2008 through 2021 are available. The Science data were produced using a random forests regression algorithm. For standard error data, the initial standard error estimates that ranged from 0 to approximately 45 were multiplied by 100 to maintain data precision in unsigned 16 bit space (e.g., 45 = 4500). Therefore, standard error estimates pixel values range from 0 to approximately 4500. The value 65534 represents the non-processing area mask where no cloud or cloud shadow-free data are available to produce an output, and 65535 represents the background value. The Science data are accessible for multiple user communities, through multiple channels and platforms. For information on the NLCD TCC data and processing steps see the NLCD metadata. Information on the Science data and processing steps are included here.
The goal of this project is to provide CONUS and OCONUS with complete, current and consistent public domain tree canopy cover information.
Funding for this project was provided by the U.S. Forest Service (USFS). RedCastle Resources produced the dataset under contract to the USFS Geospatial Technology and Applications Center.
Kevin Megown
USDA Forest Service Geospatial Technology and Applications Center (GTAC)
Program Lead: Resource, Mapping, Inventory and Monitoring
801-975-3826
801-975-3478
125 S. State Street, Suite 7105
Salt Lake City
UT
US
sm.fs.tcc@usda.gov
84138
0800 - 1600 MT, M - F
U.S.
USA
United States of America
Lower 48
Conterminous United States
CONUS
United States of America
Tree Density
Digital Spatial Data
Tree Canopy Cover
Continuous
Percent Tree Canopy Standard Error
Remote Sensing
GIS
Change
Landsat
Sentinel-2
LandTrendr
NGDA Portfolio Themes
NGDA
National Geospatial Data Asset
ISO 19115 Category
BaseMaps
EarthCover
Imagery
Environment
BaseMaps
EarthCover
Imagery
Digital Spatial Data
Continuous
NGDA
Remote Sensing
National Geospatial Data Asset
Percent Tree Canopy
Environment
GIS
The USDA Forest Service makes no warranty, expressed or implied, including the warranties of merchantability and fitness for a particular purpose, nor assumes any legal liability or responsibility for the accuracy, reliability, completeness or utility of these geospatial data, or for the improper or incorrect use of these geospatial data. These geospatial data and related maps or graphics are not legal documents and are not intended to be used as such. The data and maps may not be used to determine title, ownership, legal descriptions or boundaries, legal jurisdiction, or restrictions that may be in place on either public or private land. Tree Canopy Cover changes may or may not be depicted on the data and maps, and land users should exercise due caution. The data are dynamic and may change over time. The user is responsible to verify the limitations of the geospatial data and to use the data accordingly.
none
n/a
These data were collected using funding from the U.S. Government and can be used without additional permissions or fees. If you use these data in a publication, presentation, or other research product please use the following citation:
USDA Forest Service. 2023. USFS Percent Tree Canopy (Standard Error of Science Version) CONUS v2021-4. Salt Lake City, UT.
Corner Coordinates (center of pixel, meters): upper left: -2493045.0 (X), 3310005.0 (Y); lower right: 2342655.0 (X), 177285.0 (Y).
Google Earth Engine v0.1.321; GDAL 3.4.3
The Multi-Resolution Land Characteristics continental United States study area without Alaska
1
-130.23282801589895
-73.59459648889016
22.07673063066848
48.70739591304975
20170601
20170901
The USDA Forest Service makes no warranty, expressed or implied, including the warranties of merchantability and fitness for a particular purpose, nor assumes any legal liability or responsibility for the accuracy, reliability, completeness or utility of these geospatial data, or for the improper or incorrect use of these geospatial data. These geospatial data and related maps or graphics are not legal documents and are not intended to be used as such. The data and maps may not be used to determine title, ownership, legal descriptions or boundaries, legal jurisdiction, or restrictions that may be in place on either public or private land. Tree Canopy Cover changes may or may not be depicted on the data and maps, and land users should exercise due caution. The data are dynamic and may change over time. The user is responsible to verify the limitations of the geospatial data and to use the data accordingly.
The USFS Forest Inventory and Analysis (FIA) program photo-interpreted percent tree canopy cover (TCC) response data. Photointerpretation (PI) measured TCC using a custom ArcGIS plug-in tool (Goeking et al., 2012) from 105-point grids placed in 90x90 squares centered on USFS FIA plot design. A total of 55,242 PI plots were used in TCC modeling.
20120101
Creation of Digital Elevation Model (DEM) derivatives. A CONUS-wide terrain dataset used a predictor layer was provided by the USGS 3D Elevation Program (U.S. Geological Survey, 2019). Slope, aspect, and the sine and cosine of aspect were calculated for each pixel following industry standards.
20220901
Creation of cropland data layer (CDL) binary mask. The annual binary agriculture data were produced by classifying all non-tree CDL crops as agriculture and everything else as non-agriculture.
20130101
Two sets of annual medoid composites were created. Set 1 does not include any Landsat 7 data occurring after 2002. Set 2 includes all available Landsat 7 data through 2015. To generate annual composites Landsat and Sentinel 2 imagery were collected from 1984-2022 from Julian day 153-273 for 1984-2015, and Julian day 182-244 for 2016-2022. Landsat 7 imagery were used from 1999-2002, and not used after 2002 due to scan line correction failure in 2003. For Landsat image collections, the CFmask cloud masking algorithm, an implementation of Fmask 2.0 was applied (Zhu and Woodcock 2012; Foga et al., 2017), and the cloudScore algorithm (Chastain et al., 2019). For Sentinel-2 data, the s2Cloudless algorithm was used to mask clouds (Zupanc, 2017). We used the Temporal Dark Outlier Mask (TDOM) method to mask cloud shadows in both Landsat and Sentinel-2 (Chastain et al., 2019). For each year, the annual geometric medoid was computed to summarize the data into a single annual composite for each of the 54 tiles that span the CONUS.
20221001
The Landsat-based detection of Trends in Disturbance and Recovery (LandTrendr) temporal segmentation algorithm was applied to the two sets of composite time series in Google Earth Engine (GEE) (Kennedy et al., 2018; Cohen et al., 2018). The resulting two sets of LandTrendr time-series fitted values were used as independent predictor variables in random forest models (Breiman 2001). Stripping artifacts were observed in preliminary modeling of TCC when LandTrendr set 2 visible bands - derived from composite set 2 data that includes all Landsat 7 data through 2015 - were included as predictor layers. To avoid stripping artifacts the visible bands from LandTrendr set 2 fitted values were not used in modeling.
20221015
Creation of the percent tree canopy cover (TCC) Science dataset (main process). The FS Science 2011 TCC dataset was created for the CONUS. For CONUS, 54 tiles were used in a 5x5 moving window where model calibration data was gathered from the moving windows and random forest models were created. The random forest models were applied applied to the center tiles. The final dataset is a mosaic of TCC values.
Five major steps were employed to map TCC and produce the Science product: 1) collection of reference data, 2) acquisition and/or creation of predictor layers, 3) calibration of random forests regression models for each mapping area using response data and predictor layers, 4) application of those models to predict per-pixel TCC across the entire mapping area, and 5) exporting Science images from Google Earth Engine (GEE) to local computers for further post-processing that includes the creation of the CONUS-wide mosaic. The methodology is described further below, in the technical methods document (Housman et al., 2023), and in an upcoming manuscript in preparation (Heyer et al., 2023).
Step 1: Reference data, consisting of estimated TCC at each of the {} FIA plot locations, were generated via aerial image interpretation of high spatial resolution images collected and supplied by the U.S. Forest Service Forest Inventory and Analysis (FIA) program. The spatial distribution of the sample points follows the FIA systematic grid (Brand et al. 2000). Low quality FIA PI observations were removed for a total of 55,242 FIA plots used in modeling
Step 2: Predictor layers include two sets of LandTrendr fitted images spectral derivatives. Set 1 (no Landsat 7 data after 2002) includes all optical bands and indices. Set 2 (includes all Landsat 7 data through 2015) excludes Landsat 7 visible bands to avoid stripping artifacts. Other predictor layers include binary agriculture layer (1=agriculture, 0 = non-agriculture), elevation data, and terrain derivatives (slope, aspect, sine of aspect, cosine of aspect). The processes for creating the derived layers are described separately (see related Process Steps).
Step 3: For each 480 km x 480 km moving window tile, a random forest model was built from 2011 response and predictor data that fell over a 5x5 tile neighborhood for that tile. For each model, the variable selection R package VSURF (Genuer et al., 2015) was used to determine the number of variables to randomly sample at tree splits (mtry). Models were generated locally using the random forest regression algorithm "sklearn.ensemble.RandomForestRegressor" from the Scikit-Learn package in python (Pedregosa et al. 2011).
Step 4: In Google Earth Engine (GEE), models were applied to each tile for CONUS, producing a 2-layered Science image. The first layer was the random forests mean predicted TCC value and the second layer was the standard error, which is the per-pixel standard error of the random forests regression predictions from the individual regression trees. In addition to the model output, a statistic that normalized the expected error, which we refer to as tau (Coulston et al., 2016), was calculated for each processing area. The tau statistics can be used to mask erroneous pixels as described in Coulston et al. (2016). The tau statistics are provided in the supplemental metadata.
Step 5: Following model application, the Science TCC images were exported from GEE to local computers for post-processing that includes creation of mosaics. During post-processing a colormap was applied to each image. For each image, the non-area processing value is 254, and the background value is 255. The standard error non-area processing value is 65534, and the background value is 65535.
20230201
All years are modeled separately. As such, measurements from one year to another are not inherently dependent on other years
See the data quality report for methods
Data extend across the lower 48 conterminous United States
Model performance metrics including mean of squared residuals and percent variability explained were obtained from the 54 random forest regression models (Breiman, 2001; R Core Team 2020), that were used to derive tree canopy cover estimates. The maximum mean of squared residuals was 206.5 and the minimum was 91.3. The maximum percent variability explained was 89.7 and the minimum was 60.1. All model performance metrics can be found in a supplemental accuracy text file included. Additionally, 500 random forest trees were used to derive a tree canopy cover prediction for each pixel. Standard errors were calculated for each pixel using the estimates from the 500 trees. The standard errors provide information on the certainty of TCC predictions. We conducted an independent error assessment over the 2011 NLCD TCC v2021.4 map output. Thirty percent of our response data was withheld from model calibration to be used to assess error. In order to account for inconsistencies between the FIA plot location density between different states, along with plots that we did not use due to quality assurance measures, we estimated the area weight of each plot by computing Thiessen polygons for each plot centroid. In order to avoid extremely large weights for edge plots, any plot on the edge assumed the average of the weight of neighboring non-edge plots polygon weights. We then computed the weighted root mean squared error (RMSE) and mean absolute error (MAE) using the response TCC value as the truth, and the final NLCD TCC 2011 map value as the predicted value.
20230401
20230401
20230401
For CONUS the weighted map RMSE is 12.8% TCC.
1
161190
30.0
104424
30.0
2
1
0
-2493045.0 177285.0
-2493045.0 3310005.0
2342655.0 3310005.0
2342655.0 177285.0
-75195.0 1743645.0
Percent tree canopy cover standard error
65534.0
0.0
16
Percent tree canopy cover standard error
False
False
False
False
False
science_se_conus_2017_v2021-4.tif
ENTITY_ATTRIBUTE_DESCRIPTION
OID
ObjectID Field
Short integer
4 bytes
5
0
Value
Percent tree canopy cover standard error
Short integer
4 bytes
5
0
Count
Total number of pixels per classification
Long integer
8 bytes
10
0
Blue
Color ramp red value
Short integer
1 byte
3
0
Green
Color ramp green value
Short integer
1 byte
3
0
Red
Color ramp blue value
Short integer
1 byte
3
0
UNREPRESENTABLE_DOMAIN
RANGE_DOMAIN_MINIMUM
RANGE_DOMAIN_MAXIMUM
dataset