Land Cover Data Overview

Science Center Objects

The GAP/LANDFIRE National Terrestrial Ecosystems data set includes detailed vegetation and land cover patterns for the continental United States. Links to access GAP/LANDFIRE National Terrestrial Ecosystems 2011 Data, Metadata and Web Services.

<< Back to Land Cover

Terrestrial Ecosystems Data and Modeling

The GAP/LANDFIRE National Terrestrial Ecosystems data set includes detailed vegetation and land cover patterns for the continental United States. The data set incorporates the Ecological System classification system developed by NatureServe to represent natural and semi-natural vegetation. The 584 unique classes in the data set can be displayed at three levels of detail, from general (8 classes) to most detailed. The data set can be used to identify those places in the country with sufficient habitat to support wildlife, a key step in developing sound conservation plans.

The GAP/LANDFIRE National Terrestrial Ecosystems data set is mainly focused on habitat identification.  The USGS National Land Cover Database (NLCD) is also available and it provides information about general land cover types with 16 land cover classes (for example, Low Intensity Developed, Deciduous Forest).  Learn more about NLCD 2011 >>

 

Why are these data important?

GAP/LANDFIRE National Terrestrial Ecosystems data set provides information on the distribution of native vegetation types, modified and introduced vegetation, developed areas, and agricultural areas of the United States. For all areas of the county except Hawaii and Puerto Rico, native vegetation areas are classified to the Ecological System types developed by NatureServe. Ecological Systems provide detailed information on the vegetative communities of an area that is not available in most other regional or national land cover products. This level of thematic detail makes possible the construction of wildlife habitat distribution models, and the construction of complicated hydrology and fire dynamics models, and many other applications. Information about terrestrial ecosystems is a key component of effective conservation planning and the management of biological diversity as it is used to build predictive models of wildlife distribution and biodiversity across large geographic areas. When used in conjunction with protected areas data (PAD-US), terrestrial ecosystems data can be used to identify habitat types that may be under-protected so management activities can be adjusted. These maps and data can be used to identify those places in the country with sufficient habitat to support wildlife, a key step in developing sound conservation plans.

 

How National Terrestrial Ecosystems data are being used

Information about terrestrial ecosystems is used to develop models showing where there is appropriate habitat for the nation’s vertebrate species. This information is important for decision makers, planners, researchers, private interests and others:

  • Biodiversity: National Terrestrial Ecosystems data is key to helping land conservation decision makers better match biodiversity goals to land protection programs and activities.
  • Habitat Loss: Human population in the U.S. is predicted to grow by 25% in the next 50 years. This population increase, coupled with our land consumption patterns, means that there will be significant decreases in habitat for other species. Knowing where large tracts of intact habitat occur is key to targeting the most effective lands for biodiversity conservation which, in turn, can offset some of the effects of habitat loss.
  • Climate: Accelerating climate change is elevating the importance of effectively targeted species protection efforts. For many species, warming climates could push them to the brink of extinction unless habitat migration corridors can be set aside. Information about land cover is critical to understanding where to focus such corridor planning.
  • Management: Agencies and non-profits that manage protected areas often lack complete information about the land cover types that might be present or could be restored on their lands. Land cover maps and data are crucial to the development of improved land management practices that support continued biodiversity.

 

Description of the data

Citation:  

U.S. Geological Survey Gap Analysis Program, 20160513, GAP/LANDFIRE National Terrestrial Ecosystems 2011: U.S. Geological Survey, https://doi.org/10.5066/F7ZS2TM0.

Overview

The GAP/LANDFIRE National Terrestrial Ecosystems data represents a detailed vegetation and land cover classification for the Conterminous U.S., Alaska, Hawaii, and Puerto Rico.  The data are provided as a 30 m by 30 m raster.  GAP/LF 2011 Ecosystems for the Conterminous U.S. is an update of the National Gap Analysis Program Land Cover Data - Version 2.2.  Alaska ecosystems have been updated by LANDFIRE to 2012 conditions (LANDFIRE 2012).  Hawaii and Puerto Rico data represent the 2001 time-frame (Gon et al. 2006, Gould et al. 2008).  The classification scheme used for the Alaska and the lower 48 states is based on NatureServe’s Ecological System Classification (Comer et al. 2003), while Puerto Rico and Hawaii’s map legend are based on island specific classification systems (Gon et al. 2006, Gould et al. 2008).

2011 CONUS Update

The process for this update was tested and documented for three of the nine geoareas in the conterminous U.S. (McKerrow et al. 2014).  These geoareas are aggregations of the 66 Multi-Resolution Land Characteristics Consortium (MLRC mapping zones that have been used by NLCD, GAP and LANDFIRE as mapping zones in their 2001 national mapping efforts (Homer et al. 2004). For each geoarea the ArcGIS Combine operation was used to generate unique combinations of the pixels from the 2001 National GAP Vegetation data (USGS GAP 2011), 2011 NLCD (Homer et al. 2015), and 2010 LANDFIRE disturbance data (Nelson et al. 2013).  Those combinations were then evaluated against high resolution imagery to determine those needing to be updated either because of a change in the land conditions or a refinement of the initial classification.

The 2011 NLCD label was accepted into the GAP/LF 2011 Ecosystems update layer for row crop, pasture/hay, water, and the four developed land cover classes without further investigation. The 2001 GAP class label was maintained if it matched the NLCD general land cover class. For example, for pixels mapped as Atlantic Coastal Plain Dry and Dry-Mesic Oak Forest in the 2001 GAP map that were mapped as deciduous forest in the NLCD 2011 map, the original 2001 ecological system call was left unchanged. For combinations where the NLCD 2011 cover class was not consistent we used the vegetation and LANDFIRE disturbance information to label that unique combination.  High resolution imagery from Google Earth or ArcGIS base maps was used as the reference imagery to assign the updated 2011 cover class. For example, inland Northwest combinations of a forest type in the 2001 GAP map, shrub in the NLCD 2011 map that had been burned according to the 2010 LANDFIRE disturbance data were assigned to the class Recently Burned Forest in the 2011 GAP/LF 2011 Ecosystems update. 

In some geoareas the 10-digit Hydrologic Unit Code (HUC; USGS National Gap Analysis Program. 2011) boundary data derived from the USGS Watershed Boundary Dataset (USGS NGTOC 2011) was used to label a subset of the pixels in a combination to a new land cover class. For example, the mapping of the Northern Atlantic Coastal Plain Pitch Pine Barrens had been previously under-mapped in the 2001 GAP Land Cover. It was improved by reclassifying combinations that included pixels by NLCD as shrubland with low intensity fires in LANDFIRE disturbance data, within a subset of the HUCs representing the Pitch Pine Barrens region.  In areas outside the Pine Barren region, that combination would represent a different ecological system. 

In the Midwestern geoareas the National Agricultural Statistics Service Cropland Data Layer (2011) was used to recode the Modified/Managed Southern Tall Grassland from the 2001 Gap Land Cover Pasture/Hay for pixels where the CDL had mapped them as Grassland/Pasture or Other Hay/Non Alfalfa.

After accepting the 2011 NLCD labels for agriculture, water, developed classes, the ecological system matches between 2001 and 2011, and recodes based on disturbance were complete, there were a large number of combinations remaining without a 2011 label assigned. The combination dataset allowed each analyst to visualize the distribution of the pixels within the un-assigned combinations and evaluate the potential source of difference between the two dates.   For each combination, information about the distribution and pattern of the pixels as well as the type of transition (i.e. forest to grassland) were used to prioritize which combinations were evaluated for recoding at this step.

These were typically: 1) combinations represented by small groups of pixels scattered across the geoarea and thus problematic to assess and assign a transition path or 2) combinations whose recode options included more than one possible ecological system assignment. In these situations, we recoded the pixels based on their proximity to labelled pixels in the 2001 GAP map. For example, pixels comprising a combination of 2001 GAP non-forest, 2011 NLCD deciduous forest, and no disturbance, the recoding would be driven by the neighborhood majority of candidate pixels for deciduous forest types in the geoarea.

There were unlabeled pixel combinations that persisted after the proximity assignment step. This occurred when single pixels or small groups of pixels were isolated with respect to physiognomy type and no neighborhood majority of candidate ecological systems occurred (e.g., an island of forest in a large patch of grassland). In these situations, the 2001 GAP map classes were used in the updated map.

After each analyst completed the update for a geoarea, a second individual reviewed the resulting land cover classification.  Each individual map class location was checked against high resolution reference imagery using Google Earth or ArcGIS base maps to evaluate the pattern and extent of the mapped classes across the geoarea. If the review identified artifacts or errors, the analyst revised the decision rules and the process was repeated until a satisfactory classification was completed. 

The nine geoareas were then mosaicked in ArcGIS and a class by class check was conducted to look for errors introduced by the mosaicking process.  Again, pattern and extent of each of the mapped classes was visualized and compared against the 2001 Gap Land Cover and the reference imagery.  Logic checks were conducted at each process step to ensure that the map classes impacted by a specific decision rule or process were the only ones changing between intermediate versions of the mosaic.

It is important to keep in mind that this dataset represents an update and not a change detection. Therefore, the 2001 and 2011 land cover maps should not be used to quantify change directly. Our goal was to generate a detailed land cover product representing the 2011 timeframe. Differences between the 2001 and 2011 GAP maps do not always represent on-the-ground changes in vegetation communities. Some of the differences are the product of corrections to mis-classifications in the original 2001 map. An area that contained the same vegetation in 2001 and 2011, but was incorrectly mapped in 2001 would show up as “changed.”

Finally, the concepts of the Ecological Systems used to define the map legend are inherently variable with a range of physiognomic and phenological conditions possible in a single system.  As such, there are cases where the Ecological System land cover class may have remained “unchanged” but the general land cover class had changed between 2001 and 2011.  For example, some areas correctly mapped as shrub in the NLCD layer (based on the NLCD definition) are best mapped as the Northern Rocky Mountain Ponderosa Pine Woodland and Savanna ecological system in the GAP map; therefore the 2001 Woodland and Savanna label would be retained.

2012 Alaska Update

Updates to the LANDFIRE’s Existing Vegetation layer for Alaska have been implemented by the LANDFIRE Team.  After completion of the 2001 land cover, biennial updates have been done using a disturbance database to identify areas of change.  Learn More about  LANDFIRE 2012

2001 Base Mapping

Classification Scheme

Previous land cover mapping projects made clear the need for a nationally consistent classification scheme mappable at a meso-scale. In response to this need, NatureServe developed the Terrestrial Ecological Systems Classification framework (Comer et al. 2003). Ecological systems are defined as “groups of plant community types that tend to co-occur within landscapes with similar ecological processes, substrates and/or environmental gradients” (Comer et al. 2003). Although distinct from the US-NVC, the vegetation component of an ecological system is described by one or more NVC alliances or associations. While the ecological system concept emphasizes existing dominant vegetation types, it also incorporates physical components such as landform position, substrates, hydrology, and climate. 

The Alaska and Continental U.S. portion of the data set contains 680 Ecological systems and 28 land use, introduced vegetation or disturbed classes. The Hawaii data set contains 28 natural vegetation classes and nine land use, introduced vegetation or disturbed classes (Gon et al. 2006).  Puerto Rico’s data set includes 70 unique vegetation and land use classes. Frequently, this high number of classes provides a level of detail that exceeds a user’s needs. To accommodate these users, we have cross-walked the ecological system level data to the six highest levels of the National Vegetation Classification System (USNVC). The vegetation features used to distinguish these classes range from growth form, and climate regimes at the Class level to regional differences in substrate and hydrology at the Macrogroup level (Table 1; http://usnvc.org/). The NVC levels provide the user with a variety of options allowing the choice of making a map of the Continental U.S. with eleven classes at the NVC Class level to 583 classes at the Ecological system level.

 

         Table 1. Features used to delineate National Vegetation Classification (NVC) levels

         Class: dominant general growth forms adapted to basic moisture, temperature, and/or substrate or aquatic

         Subclass: global macroclimatic factors driven primarily by latitide and continental postion, or reflect overriding substrate or  aquatic condtions

         Formation: global macroclimatic conditions as modified by altitide, seasonality of precipitation, substrates, hydrological conditions

         Division:  continental differences in mesoclimate, geology, substrates, hydrology, disturbance regimes

         Macrogroup: sub-continental to regional differences in mesoclimate, geology, substrates, hydrology, disturbance regimes

         Group: regional differences in mesoclimate, geology, substrates, hydrology, disturbance regimes                

         

Imagery

The 2001 National GAP Land Cover was derived using Landsat TM satellite imagery from 1999-2001 as its base. Much of the imagery was acquired from the USGS National Center for Earth Resources Observation and Science (EROS) through the Multi-Resolution Land Characteristics Consortium (MRLC). EROS did the preprocessing and atmospheric correction of the images and developed three clear, low cloud images for all areas of the United States. These mosaics represented the spring, summer, and fall seasons. Having images of vegetation during these three different times of the year is very helpful for correctly classifying vegetation types that look the same during one season but different in another.  Imagery that was used for the Midwest, Northeast and Alaska had been selected and preprocessed by the LANDFIRE Team at EROS.  In Hawai’I it was necessary to supplement the MRLC imagery in order to deal with cloud cover issues in the standard datasets (Gon et al. 2006).  In Puerto Rico a cloud free mosaic of Landsat 7 ETM+ imagery was compiled for the project (Martinuzzi et al. 2006).

A variety of other dataset were used in all or some of the regions to help with the land cover classification process. Normalized Difference Vegetation Index (NDVI), and brightness, greenness and wetness indices were created using Landsat ETM+ coefficients from Huang et al. (2002). Digital elevation model-derived data sets were also used and included elevation, slope, aspect and landform. Other ancillary data sets used depend on the region but include digital data on soils, geology, stream and wetland location, point locations for rare plant communities and fire and tree harvest information.

Training Data

The models predicting the land cover distribution for the regional projects used field data collected by gap personnel specifically for the project and vegetation data collected for other projects to instruct the model. GAP field samples were collected by traversing navigable roads in a mapping zone and opportunistically selecting plots that met criteria of appropriate size (1-hectare minimum) and composition (stand homogeneity). Plot data were collected using ocular estimates of biotic and abiotic land cover elements, including percent cover of dominant species by life form (i.e. trees, shrubs, grasses, and forbs) and physical data such as elevation, slope, aspect and landform. Visual interpretation of aerial photography, digital orthophoto quads, or other remotely sensed imagery was also used to generate training data when sufficient information could be obtained to make and Ecological system determination. This technique was most often used for rare systems with insufficient training data and where additional ancillary data was available to identify image features.

Additional vegetation data was collected from Federal, State, and Non-Governmental agencies and Ecological Systems labels were assigned to their existing plots. The LANDFIRE Program has an extensive national dataset of existing vegetation data in the LANDFIRE Reference Database (LFRDB) which has been used to support mapping.

Modeling Techniques

Classification and regression trees (CART) are a valuable tool for discriminating complex relationships among environmental variables. Decision trees use a binary partitioning algorithm to successively split a multidimensional “cloud” of explanatory data into increasingly homogenous subsets. Each binary split is considered a single rule in a chain of rules defining the characteristics of the response variable. For land cover mapping, explanatory variables are the spectral and ancillary data sets and the response variable is the land cover classes (see Lowry et al. 2005 for more information on CART modeling techniques).  Decision tree classifiers are well suited for land cover mapping as they do not require normally distributed training data, they can accommodate a wide variety of predictor variables, and they have demonstrated improved accuracies over the use of traditional classifiers (Hansen et al. 1996, Pal and Mather 2003). 

CART modeling techniques were used to map the majority of the Ecological systems in the Southwest and Northwest regional projects, as well as for LANDFIRE. CART tended to work best for widespread matrix systems with sufficient training data. For rarer, patch systems and systems in the Southeast regional project where topographic relief was not sufficient and training data were limited a combination of CART modeling and decision rules were used for mapping the existing vegetation.  The Hawaii Gap Project conducted supervised classification of the Landsat data and object based classification of high resolution imagery and for some classes (Gon et al. 2006).  The Puerto Rico GAP used unsupervised classification to identify major spectral classes and ancillary data (e.g. climate, substrate, and topography) to stratify the spectral data

Data Limitations

The GAP/LANDFIRE National Terrestrial Ecosystems data attempts to map as accurately as possible the fine vegetative details of the nation’s vegetation. However, there are limitations to the data that users should keep in mind. The data set uses a 30 meter pixel cell and in most areas a minimum mapping unit of 0.4 ha (1 acre) this means that small patches of vegetation can be missed in the modeling process. By nature of their patchy distributions and frequently small extents wetlands, riparian habitats and rare habitat types can be the most frequently missed types. Some Ecological Systems have spectral signatures, and occur at similar elevations, aspects etc. to other Ecological Systems. Because of these similarities, the modeling process used to create the land cover data has difficulty differentiating between these types of systems and a significant amount of confusion may occur.

The GAP/LANDFIRE Terrestrial Ecosystem dataset is based on a range of imagery dates for the different geographies covered. In areas of rapid change, the data may need additional updating before use or may be inappropriate for use.

References

Comer, P., D. Faber-Langendoen, R. Evans, S. Gawler, C. Josse, G. Kittel, S. Menard, S. Pyne, M. Reid, K. Schulz, K. Snowand, J. Teague, 2003. Ecological systems of the United States: A working classification of U.S. terrestrial systems. NatureServe, Arlington, Virginia.

Gon, S.M., A. Allison, R. J. Cannarella, J. D. Jacobi, K. Y. Kaneshiro, M. H. Kido, M. Lane-Kamehele, S. E. Miller.  2006.  The Hawaii Gap Analysis Project Final Report.  487 pp.

Gould, W. A. C. Alarcon, B. Fevold, M.E. Jimenez, S. Martinuzzi, G. Potts, M. Quinones, M. Solorzona, E. Ventosa.  2008.  The Puerto Rico Gap Analysis Project. Volume 1: Land cover, vertebrate species distribution, and land stewardship.  Gen. Tech. Rep. IITF-GRT-39. Rio Piedras, Pr.  USDA, Forest Service, International Institute of Tropical Forestery.  165. p.

Hansen, M., R. Dubayah, and R. DeFries, 1996. Classification trees: An alternative to traditional land cover classifiers. International Journal of Remote Sensing 17(5): 1075-1081.

Homer, C.G., Dewitz, J.A., Yang, L., Jin, S., Danielson, P., Xian, G., Coulston, J., Herold, N.D., Wickham, J.D., and Megown, K., 2015,Completion of the 2011 National Land Cover Database for the conterminous United States-Representing a decade of land cover change information. Photogrammetric Engineering and Remote Sensing, v. 81, no. 5, p. 345-354. http://www.asprs.org/a/publications/pers/2015journals/PERS_May_2015/HTML/index.html#346/z

Homer, C., C. Huang, L. Yang, B. Wylie, and M. Coan, 2004.   Development of a 2001 National Land-Cover Database for the United States.  Photogrammetric Engineering & Remote Sensing, Number 7 / July 2004, pp. 829-840(12). DOI: https://doi.org/10.14358/PERS.70.7.829

Lowry, J. H, Jr., R. D. Ramsey, K. Boykin, D. Bradford, P. Comer, S. Falzarano, W. Kepner, J. Kirby, L. Langs, J. Prior-Magee, G. Manis, L. O’Brien, T. Sajwaj, K. A. Thomas, W. Rieth, S. Schrader, D. Schrupp, K. Schulz, B. Thompson, C. Velasquez, C. Wallace, E. Waller and B. Wolk. 2005. Southwest Regional Gap Analysis Project: Final Report on Land Cover Mapping Methods, RS/GIS Laboratory, Utah State University, Logan, Utah.

Martinuzzi, S., W. A. Gould, O. M. Ramos-González. 2006. Creating cloud-free Landsat ETM+ data sets in tropical landscapes: cloud and cloud-shadow re-moval. Gen. Tech. Rep. IITF-32. Río Piedras, PR: U.S. Department of Agriculture, Forest Service, International Institute of Tropical Forestry. 12 p. Accessed On line August 2016: http://www.fs.fed.us/global/iitf/pubs/iitf-gtr32.pdf.

McKerrow, A.J., A. Davidson, T. S. Earnhardt, and A. L. Benson.  2014.  Integrating Recent Land Cover Mapping Efforts to Update the National Gap Analysis Program's Species Habitat Map.  Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XL-a 245-252, 2014.  http://doi.org/10.5194/isprsarchives-XL-1-245-2014.).

Nelson, K. J., J. Connot, B. Peterson, J. J. Picotte. 2013. LANDFIRE 2010 - Updated Data to Support Wildfire and Ecological Management.  IEEE Earthzine. http://earthzine.org/2013/09/15/landfire-2010-updated-data-to-support-wi....

Pal, M. and P. M. Mather, 2003. An assessment of the effectiveness of decision tree methods for land cover classification. Remote Sensing of Environment 86, 554-565

USDA National Agricultural Statistics Service Cropland Data Layer. 2011. Published crop-specific data layer [Online]. Available at https://nassgeodata.gmu.edu/CropScape/. USDA-NASS, Washington, DC.

U.S.Geological Survey Gap Analysis Project.  2011 Additional Data – Hydrologic Unit Codes [HUCS]. [Online]. Available at https://www.sciencebase.gov/catalog/item/56d496eee4b015c306f17a42 .

U.S. Geological Survey, National Geospatial Technical Operations Center, 20140924, USGS National Watershed Boundary Dataset (WBD) 20140924 National Shapefile: U.S. Geological Survey: Reston, VA, http://nhd.usgs.gov/wbd.html.

 

Access to previous versions of this dataset

For access to previous versions of the land cover dataset please contact Alexa McKerrow: amckerrow@usgs.gov 

Contacts for GAP Land Cover Data >>