Data - References
i-Tree Landscape offers users a wide variety of data and map layers.
More information can be found in the Landscape section of the i-Tree Tools Archives.
The references are grouped in separate pages.
Tree cover estimates are derived directly from 2011 National Land Cover Data (NLCD) or 2001 NLCD data in Alaska, Hawaii and Puerto Rico (as 2011 data are not available). These data estimate percent tree cover using satellite data with a 30 meter resolution (www.mrlc.gov). The 2001 tree cover estimates are known to underestimate tree cover by an average of 9.7 percent, but the range of underestimation varies by region and land cover class (Nowak and Greenfield 2010). It is believed, based on preliminary tests, that the 2011 tree cover maps also underestimate tree cover. Therefore, the tree cover maps are likely conservative in estimating tree cover as well as ecosystem services, which are derived from tree cover. To help overcome this presumed underestimate of tree cover, high resolution tree cover maps are used where available.
Impervious cover estimates are derived directly from 2011 National Land Cover Data (NLCD) or 2001 NLCD data in Puerto Rico and Hawaii. These data estimate percent impervious cover using satellite data with a 30-meter resolution (www.mrlc.gov). The 2001 impervious cover estimates are known to underestimate tree cover by an average of 1.4 percent (Nowak and Greenfield 2010). It is believed that the 2011 NLCD impervious data provide a reasonable estimate of impervious cover.
Land cover data are derived from the 2011 and 2001 NLCD which provides a synoptic nationwide classification of land cover into 16 classes at a spatial resolution of 30 meters (www.mrlc.gov; US EPA 2015). In Hawaii and Puerto Rico the 2001 land cover data is the only available year of data.
U.S. population statistics are derived directly from the U.S. Census Bureau data (www.census.gov) and are believed to be without error. Census data are provided for each geographic unit.
- ¶Wildfire Potential: A map classified by the relative potential for wildfire that would be difficult for suppression resources to contain. The intended use of this layer is to help inform evaluations of wildfire risk or prioritization of fuels management needs across large spatial scales. This map layer was derived by the U.S. Forest Service by classifying the Wildfire Hazard Potential (WHP) continuous dataset values. The process entailed converting the WHP values to integers, evaluating the statistical distribution of WHP values and then classifying them into fire classes: very high, high, moderate, low, very low. WHP ranges for each class are unknown. The non-burnable and water values were incorporated from the LANDFIRE FBFM40 layer to produce the final classified WFP.
- ¶Pests: The data is based on range maps of insects and diseases, and indicates where species are present. Range maps are derived from the Forest Health Technology Enterprise Team (FHTET) and can be viewed at the Insect and Disease Detection Survey Data Explorer.
¶Forest to Faucets: Derived from a U.S. Forest Service project to map forested areas that are most beneficial to surface drinking water (SDW) surfaces. This data is only available if a user is analyzing HUC12 watershed areas.
- Surface Drinking Water Importance¶
- This index is calculated based on water supply, spatial flow through the landscape, and the downstream drinking water demand. Watersheds are ranked and normalized on a scale of 0-100. The greater the index value, the greater the importance.
- Forest Importance to SDW¶
- This index is used to determine the extent to which forests are currently protecting the geographic areas most important for surface drinking water. To calculate this index, Surface Drinking Water Importance values are multiplied by the percent of forest in the watershed.
- Protected Forest Importance to SDW¶
- This index is used to determine the extent to which protected forests are currently protecting the geographic areas most important for surface drinking water. To calculate this index, Surface Drinking Water Importance values are multiplied by the percent of protected forest in the watershed.
- NFS Forest Importance to SDW¶
- This index is used to determine the extent to which forest land managed by the U.S. Forest Service is currently protecting the geographic areas most important for surface drinking water. To calculate this index, Surface Drinking Water Importance values are multiplied by the percent of National Forest Service (NFS) forest in the watershed.
- Private Forest Importance to SDW¶
- This index is used to determine the extent to which private forests are currently protecting the geographic areas most important for surface drinking water. To calculate this index, Surface Drinking Water Importance values are multiplied by the percent of private forest in the watershed.
- SDW Pest Threat¶
- This index is used to identify forested areas important for surface drinking water that are likely to be affected by future increases in insects and diseases. To calculate this index, Forest Importance to Surface Drinking Water values are multiplied by the percent of the watershed that is highly threatened by insects and diseases.
- SDW Development Threat¶
- This index is used to identify forested areas important for surface drinking water that are likely to be affected by future increases in housing density. To calculate this index, Forest Importance to Surface Drinking Water values are multiplied by the percent of the watershed that is highly threatened by development.
- SDW Wildfire Threat¶
- This index is used to identify forested areas important for surface drinking water that are likely to be affected by future increases in wildland fire incidents. To calculate this index, Forest Importance to Surface Drinking Water values are multiplied by the percent of the watershed that is highly threatened by wildfire.
- ¶Air Quality: Pollution concentrations are estimated for ozone (O3) and particulate matter less than 2.5 microns (PM2.5) and derived from the U.S. Environmental Protection Agency's (EPA) Downscaler Model. Average and maximum values are estimated from the pollution concentration for all days in 2008. The month in which the maximum concentration value occurs is also reported.
- ¶UV Index: Ultraviolet (UV) radiation is emitted by the sun and while beneficial to humans in small doses, can have negative health effects when people are overexposed. The UV index scale was developed by the World Health Organization to more easily communicate daily levels of UV radiation and alert people to when protection from overexposure is needed most. UV index values are calculated by dividing UV dose estimates from Temis by 100. Average and maximum values are estimated from the UV index value at solar noon for all days between 2008 and 2012. The month in which the maximum UV index value occurs is also reported.
¶Land Surface Temperature: Determining the difference between localized surface temperatures and regional mean temperatures can help qualify the impacts of land use. Generally speaking, areas with more impervious surface tend to be warmer than average, while areas with more canopy cover tend to be cooler than average. Land surface temperatures were estimated for the United States based on Landsat 8 data and standard procedures from the literature as described below.
- Landsat Scene Selection¶
- Landsat scenes were downloaded for the U.S. and U.S. territories that were as cloud-free as possible (June, July, August) for the years 2013, 2014 and 2015. Scenes were selected to provide total U.S. coverage for each year. As scenes overlapped, many areas had two or more scenes covering a location.
- Converting Scene Data to Surface Temperature¶
- Landsat 8 Thermal Infrared Sensor (TIR) band values were converted to land surface temperature by:
- Converting satellite data to At-Satellite Brightness temperature based on equations detailed in USGS (2017).
- Calculating emissivity values from NDVI using the NDVI Thresholds method (Sobrino et al. 2001).
- Converting At-Satellite Brightness temperature and emissivity information for each 30m pixel to Land Surface temperature based on equations given in section 3.4 in Weng et al. (2004).
- Cloud and Snow Removal¶
- To remove the existing clouds for the images, Fmask software was used. The software called Fmask (Function of mask) is used for automated masking of clouds, cloud shadows, and snow for Landsat TM/ETM+ images. Pixels where clouds were removed were converted to "no data".
- Land Surface Temperature Differences¶
As Landsat scenes had various dates across the United States, each Landsat scene was processed independently to minimizing the effect of different scene dates on surface temperatures. To standardize the land surface temperature estimates, the Landsat scene’s median land surface temperature was subtracted for each pixel land surface temperature value to create a relative surface temperature difference estimate. With this approach, about one-half of scene pixels will have above median temperatures and one-half will have below median temperatures, but the amount they differ from the mean will vary based on local land surface temperatures.
Within each year’s data (e.g., 2015) all scenes were mosaicked and where scenes overlapped, an average of surface temperature differences among the multiple scenes was used to calculate the surface temperature difference at each pixel with multiple values.
To help fill in for "no data" pixels due to clouds or snow, data from the various years (2013-2015) were used. The first priority was given to data from 2015. If "no data" pixels existed in 2015, data from 2014 were used to fill in these missing pixels. If "no data" pixels existed in 2015 and 2014, data from 2013 were used to fill in these missing pixels. If "no data" pixels existed in 2015. 2014 and 2013, these pixels were labeled as "no data".
Land surface temperature difference are projected at 90-meter resolution in i-Tree Landscape.
To illustrate areas with relatively warm land surface temperatures (LST) and large amounts of growing space (to potentially cool the environment) or high human populations (areas with greatest human health risks to warm temperatures), two temperature indices were produced. In each index, the LST value used is the calculated LST difference from the scene mean as described above.
- People Index¶
This index averaged the LST index (above) with a population density index. This population density index (0-100) was developed nationally by subtracting the minimum population density (PD) based on US Census block group data from each pixel’s PD value and then dividing the value by the range of PD values (national maximum PD value minus national minimum PD value):
PD Index = (pixel PD – minimum PD) / PD range
This index was then restandardized on to a scale of 0 – 100, with with low values representing areas with lower surface temperatures and/or low population density and high values representing areas with relatively high population density that are relatively warm.
- Growing Space Index¶
To create this index, the LST values were scaled on a score of 0 – 100 nationally by subtracting the minimum LST value from each pixel’s value and then dividing the value by the range of LST values (national maximum value minus national minimum value):
LST Index = (pixel LST – minimum LST) / LST range
With this index, 0 represents the coolest pixel (lowest LST) and 100 represents the warmest pixel (highest LST).
A growing space index of 0 – 100 was also created for each pixel based on percent of the pixel that is potentially readily plantable – that is, not covered by impervious, water or tree cover
The average of the LST Index and Growing Space Index was calculated to develop the LST_Growing Space Index. This index was then restandardized on to a scale of 0 – 100, with low values representing areas with lower surface temperatures and/or low percent green space and high values representing areas with relatively large amounts of growing space that are relatively warm.
- ¶Climate Change Data: Projected climate data were obtained from the National Center for Atmospheric Research (NCAR) Community Climate System Model (CCSM) projections (https://gisclimatechange.ucar.edu/). These data were 4.5 km resolution that covered the conterminous US, but were resampled for display/analysis at 30 m resolution and projected to WGS 1984 Web Mercator Auxiliary Sphere.
Data extracted were from AR5 (IPCC Fifth Assessment Report) “ensemble average” of six model runs for Representative Concentration Pathways (RCP) 4.5 and 8.5 for decades starting in 2010 and ending in 2100 (i.e., 2010, 2020, 2030, ... 2100). These data include both the projected values and the differences between the future decadal modeled values and 2010 modeled values for annual mean air temperature, annual total precipitation, and mean January and July temperatures.
Future Climate maps were funded, in part, by the USDA Climate Hubs.
Based on the tree and impervious cover data, along with other local data, the following ecosystem services for trees are assessed for the year 2010:
Carbon storage and annual sequestration values are calculated from two separate sources depending upon location in non-forest or forest land cover. Land cover classification was determined using the National Land Cover Database (NLCD).
- Non-forest carbon¶
- For non-forest NLCD classes, total carbon storage and net annual sequestration were estimated using values from urban forests (Nowak et al., 2013). Net annual sequestration is estimates of carbon accumulation from tree growth minus estimated carbon lost through decomposition due to tree mortality. Carbon storage was estimated based on the national average storage value of 7.69 kgC/m2 tree cover (standard error (SE) = 1.36 kgC/m2). Net sequestration was based on state estimates that varied based on length of growing season and averaged 0.226 kgC m2 tree cover/yr (SE = 0.045 kgC m2 tree cover/yr). State values varied from 0.430 kgC m2 tree cover/yr (Hawaii) to 0.135 kgC m2 tree cover/yr (Wyoming) (Nowak and Greenfield 2010). These estimates per unit of tree cover are essential as these values were applied to the tree cover estimates (m2) from the tree cover map to estimate total carbon (kg).
- Forest carbon¶
For forested regions, total carbon storage and net annual sequestration were derived from U.S. Forest Service Forest Inventory and Analysis (FIA) data for each county (Special thanks to Jim Smith for extracting these county FIA data). Net annual sequestration was carbon accumulated annually between FIA re-measurements based on accumulation from tree growth and new trees minus carbon lost through tree mortality.
Note: sequestration in forests is based on field measurements of change including the influx of new trees and loss of existing trees; in non-forest areas, net sequestration is modeled based on tree growth of existing trees and estimated mortality based on tree condition over a one-year period; this estimate does not include new tree influx and only includes a partial loss of carbon from mortality due to decomposition (entire carbon from trees is not removed, only part of carbon lost to decomposition is removed).
Total carbon storage and net sequestration per hectare of land was converted to total carbon storage and net sequestration per hectare of tree cover by dividing the carbon per hectare by percent tree cover in the forest land in the county. As tree cover on FIA land was not known, tree cover estimates from NLCD forest classes were used. In counties where tree cover in forest land was less than 10 percent (19 counties), tree cover was set to 10 percent to avoid inflating carbon density values per unit of cover due to low tree cover estimates. If a county had no FIA carbon storage data, but had tree cover estimates, storage density values (kgC/m2 tree cover) from the closest county were used. FIA carbon storage densities per m2 of land area averaged 6.3 kgC/m2; carbon storage density adjusted for tree cover equaled 9.8 kgC/m2 tree cover.
Net sequestration per m2 of tree cover was calculated in the same manner as for carbon storage. For net carbon sequestration, values for some counties are missing. If a county had a missing value, sequestration density values (kgC/m2 tree cover/yr) from nearby counties in the same state were used. If the entire state had missing values, the county sequestration value was estimated based on converting the national FIA sequestration density value from all known counties to state values based on the ratio of state sequestration densities to national sequestration density for non-forest areas:
Forest sequestration density for state = national average forest density x (state non-forest sequestration density / national average non-forest density).
This procedure was used for net forest sequestration in many western states (AZ, CA, ID, MT, NM, NV, OR, UT, WA, WY). The average net sequestration value for forests was 0.14 kgC/m2 tree cover/yr (average SE = 0.10 kgC/m2 tree cover/yr)(see "i-Tree Landscape Carbon Storage and Sequestration for US Counties"). This value is about 60 percent of the non-forest sequestration value. This difference is likely due to increased growth rates in urban areas (due to more open-grown nature of trees) and differences in means of calculating net sequestration (forest estimates remove all carbon from trees that die, but in urban estimates only a small portion are removed).
The 2017 value of carbon storage and sequestration is estimated at $143 / metric ton of carbon (Interagency Working Group, 2016).
Air pollution removal and value estimates are based on procedures detailed in Nowak et al. (2014). This process used local tree cover, leaf area index, percent evergreen, weather, pollution, and population data to estimate pollution removal (g/m2 tree cover) and values ($/m2 tree cover) in urban and rural areas for each county. These values are applied to the m2 of tree cover to determine total removal and values related to carbon monoxide (CO), nitrogen dioxide (NO2), ozone (O3), particulate matter less than 2.5 microns (PM2.5), particulate matter between 2.5 and 10 microns (PM10*), and sulfur dioxide (SO2). Value estimates are based on local health impacts estimated using the U.S. EPA BenMAP model for each county (based on local population data) for all pollutants except for CO and PM10*, which use externality values ($/t) to estimate pollutant removal value.
Estimates of pollution removal varied by county. Average county removal rates are used, but have a potential maximum and minimum value (see i-Tree Landscape Pollutant Ranges) that illustrates a potential range. The minimum and maximum values on average are about 57 percent of the mean value. Average differences from the mean varied from a low of 30 percent for NO2 to a high of 106 percent for PM2.5. The maximum and minimum values are likely unreasonable values as they assume a maximum or minimum removal rate for every hour of the year. No maximum or minimum values are estimated for CO.
Estimates of transpiration, precipitation interception, and avoided runoff for each county in the conterminous United States in 2010 were developed using the i-Tree Eco model and local leaf area indices and weather data. Methods are detailed in Hirabayashi (2015), Hirabayashi and Endreny (2015) and Hirabayashi and Nowak (2015). The margin of error on these estimates is unknown.
Tree Planting Prioritization¶
To determine the best locations to plant or protect trees, tree and impervious cover data in conjunction with U.S. Census data can be used to create an index that highlights priority areas among the selected geographic units. With these index values, the higher the index value, the higher the priority of the area for tree planting or protection. The index is developed by weighting the criteria that are selected by the user, along with the associated weights, assigned by the user. The sum of the criteria weights must equal 100.
As geographic areas differ in size, all index inputs are either in percentages or standardized per unit area or person. Each non-percentage layer was standardized on a scale of 0 to 1, with 1 representing the geographic area with the highest value in relation to priority (e.g., areas with highest population density, lowest stocking density, or lowest tree cover per capita were standardized to a rating of 1).
Standardized values for population density (
PD) are calculated as:
PD = (n - m) / r
PD is the value (
0 - 1),
n is the value for the geographic area (
population / km2),
m is the minimum value for all geographic areas, and
r is the range of values among all selected areas (
maximum value - minimum value).
Standardized value for percent population below poverty line (
BPL) was calculated as:
BPL = percent_population_below_poverty_line / 100
Standardized value for tree cover per capita (
TPC) is calculated as:
TPC = 1 - [(n - m) / r]
TPC is the value (
0 - 1),
n is the value for the census block (
m2 / capita),
m is the minimum value for all census blocks, and
r is the range of values among all census blocks (
maximum value - minimum value).
Standardized value for tree stocking (
TS) is calculated as:
TS = [1 - (t / (t + g)]
TS is the value (
0 - 1),
t is percent tree cover, and
g is percent grass cover.
Individual scores were combined based on the following formula to produce an overall priority index (
PI) value, where the user selects the index layer and its weight:
PI = (index_1 * weight_1) + (index_2 * weight_2) + (index_3 * weight_3)
The final index was standardized to yield values between 0 (lowest priority) and 100 (highest priority).
A default index is given based on
TPC, where the default
index = (PD * 40) + (TS * 30) + (TPC * 30). This index is a type of "environmental equity" with areas of higher human population density and lower tree cover tending to get a higher index value.
- Hirabayashi, S., 2015. i-Tree Eco Precipitation Interception Model Descriptions. http://www.itreetools.org/eco/resources/iTree_Eco_Precipitation_Interception_Model_Descriptions.pdf [accessed Apr. 2015].
- Hirabayashi, S., Endreny, T.A., 2015. Surface and Upper Weather Pre-processor for i-Tree Eco and Hydro. http://www.itreetools.org/eco/resources/Surface_weather_and_upper_air_preprocessor_description.pdf [accessed Apr. 2015].
- Hirabayashi, S., D.J. Nowak. 2015. i-Tree Eco United States County-Based Hydrologic Estimates and Estimates of Species Differentiation
- Interagency Working Group on Social Cost of Carbon, United States Government. 2016. https://www.epa.gov/sites/production/files/2016-12/documents/sc_co2_tsd_august_2016.pdf
- Technical Support Document: Social Cost of Carbon for Regulatory Impact Analysis Under Executive Order 12866. https://www.whitehouse.gov/sites/default/files/omb/inforeg/scc-tsd-final-july-2015.pdf (accessed September 2015)
- Nowak, D.J., E.J. Greenfield. 2010. Evaluating the National Land Cover Database tree canopy and impervious cover estimates across the conterminous United States: A comparison with photo-interpreted estimates. Environmental Management. 46: 378-390.
- Nowak, D.J., E.J. Greenfield, R. Hoehn, and E. LaPoint. 2013. Carbon storage and sequestration by trees in urban and community areas of the United States. Environmental Pollution. 178: 229-236.
- Nowak, D.J. S. Hirabayashi, A. Bodine and E.J. Greenfield. 2014. Tree and forest effects on air quality and human health in the United States. Environmental Pollution 193:119-129
- U.S. Environmental Protection Agency. 2015. Report on the Environment: Land Cover http://cfpub.epa.gov/roe/indicator_pdf.cfm?i=49 [accessed Oct. 2015].
- Sobrino, J.A., N. Raissouni, Z.L. Li. 2001. A Comparative Study of Land Surface Emissivity Retrieval from NOAA Data. Remote Sensing of Environment, 75(2): 256-266. http://www.sciencedirect.com/science/article/pii/S0034425700001711 Last accessed (June 2017).
- USGS. 2017. Using the USGS Landsat 8 Product. https://landsat.usgs.gov/using-usgs-landsat-8-product Last accessed (June 2017).
- Weng, Q., D. Lu, J. Schubring. 2004. Estimation of land surface temperature–vegetation abundance relationship for urban heat island studies. Remote Sensing of Environment, 89(4): 467-483. http://www.sciencedirect.com/science/article/pii/S0034425703003390 Last accessed (June 2017).