Articles published for Earthzine's Forest Resource Information theme (March. 20 - June 20, 2012) address…
LANDFIRE 2010 – Updated Data to Support Wildfire and Ecological Management
- Published on Sunday, 15 September 2013 12:16
- Kurtis J. Nelson, Joel Connot, Birgit Peterson, and Joshua J. Picotte connot
- 0 Comments
Kurtis J. Nelson1, Joel Connot2, Birgit Peterson3, and Joshua J. Picotte3
1U. S. Geological Survey (USGS) Earth Resources Observation and Science (EROS) Center, Sioux Falls, South Dakota, email@example.com
2Stinger Ghaffarian Technologies, contractor to USGS EROS, Sioux Falls
3ASRC InuTeq LLC, contractor to USGS EROS, Sioux Falls
Wildfire is a global phenomenon that affects human populations and ecosystems. Wildfire effects occur at local to global scales impacting many people in different ways (Figure 1). Ecological concerns due to land use, fragmentation, and climate change impact natural resource use, allocation, and conservation. Access to consistent and current environmental data is a constant challenge, yet necessary for understanding the complexities of wildfire and ecological management. Data products and tools from the LANDFIRE Program help decision-makers to clarify problems and identify possible solutions when managing fires and natural resources. LANDFIRE supports the reduction of risk from wildfire to human lives and property, monitoring of fire danger, prediction of fire behavior on active incidents, and assessment of fire severity and impacts on natural systems   . LANDFIRE products are unique in that they are nationally consistent and provide the only complete geospatial dataset describing vegetation and wildland fuel information for the entire U.S. As such, LANDFIRE data are useful for many ecological applications . For example, LANDFIRE data were recently integrated into a decision-support system for resource management and conservation decision-making along the Appalachian Trail.
LANDFIRE is a joint effort between the U.S. Department of the Interior Office of Wildland Fire, U.S. Department of Agriculture Forest Service Fire & Aviation Management, and The Nature Conservancy. To date, seven versions of LANDFIRE data have been released, with each successive version improving the quality of the data, adding additional features, and/or updating the time period represented by the data. The latest version, LANDFIRE 2010 (LF 2010), released mid-2013, represents circa 2010 landscape conditions and succeeds LANDFIRE 2008 (LF 2008), which represented circa 2008 landscape conditions. LF 2010 used many of the same processes developed for the LF 2008 effort .
Ongoing refinement of the LANDFIRE vegetation and fuel data is necessary to improve the quality and usability of the data and to capture landscape disturbance. LANDFIRE relies on Landsat multi-spectral imagery to produce and update vegetation and fuel data. The deep Landsat archive provides data needed for vegetation classification, change analysis, and historical disturbance characterization, for which LANDFIRE has used more than 24,000 image scenes since the program’s inception. In addition, LF 2010 used airborne and spaceborne lidar, and spaceborne synthetic aperture radar (SAR) to map vegetation structure in areas where ground-based field information was lacking, including Alaska and U.S.-affiliated islands in the Caribbean and the Pacific. The mapping of insular areas is new for the 2010 data release; previous versions of LANDFIRE were limited to the conterminous U.S., Alaska, and Hawaii.
A major component of the LANDFIRE program is its reliance on field reference data for modeling, mapping, and validation. Data are continually collected from national programs and databases, such as the Forest Inventory and Analysis Program, and solicited from many federal, state, tribal, and local government agencies, as well as universities, nonprofit organizations, and private groups. The LANDFIRE Reference Database contains more than 800,000 individual georeferenced data points from more than 650 sources. Each point describes the vegetation and/or fuel conditions at that site. This unique collection is distributed by the LANDFIRE Program for use by other projects across the country (including only data for which LANDFIRE has explicit permission to share).
In addition, the LANDFIRE Events Database consists of points and polygons that contain information about landscape treatments and natural and anthropogenic disturbances. Each feature contains the spatial location, type of event, and in some cases the severity or magnitude of change within the feature. The Events Database for LF 2010 contains more than 200,000 features from more than 360 sources. Finally, an Exotics Database contains more than 105,000 plot-level descriptions of exotic and invasive species locations from 34 different sources.
Disturbance and Vegetation Transition
For the conterminous U.S., spatial disturbance data from the Events Database were combined with Remote Sensing of Landscape Change (RSLC) data to produce disturbance products depicting disturbance type, time since disturbance, and disturbance severity.
The RSLC process was not used in Alaska and Hawaii because of a lack of suitable imagery (e.g., frequent clouds, data anomalies), subsequently disturbance products for these areas were developed using only the Events Database. The RSLC process involved selecting two Landsat image pairs, representing leaf-on and leaf-off conditions, for each year from 2007-2011 for every Landsat path/row in the conterminous U.S. Data were selected from before and after the 2008-2010 update period to ensure disturbances from the entire period were captured. Each image was processed to surface reflectance using the Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS) software  .
The images processed with LEDAPS included a series of data quality masks (e.g., clouds, shadows, water) that were applied to the imagery prior to disturbance detection, which minimized falsely detected changes. A modified version of the Multi-Index Integrated Change Analysis method, was used to produce maps showing areas of increasing and decreasing biomass by tracking several spectral indices to identify changes in vegetation over time  . An example is shown in Figure 2, comparing LF 2008disturbance with LF 2010 disturbance. Additional disturbances are shown in the LF 2010 disturbance layer in part because additional image data were available and different change detection methods were used. In LF 2010, more than 18 million hectares were mapped as disturbed between 2008 and 2010, including more than 5 million hectares of wildfire. Figure 3 shows the distribution of disturbance by geographic area per year.
Areas that were disturbed, both in LF 2010 and previously, were transitioned to new vegetation type and structure classes based on the disturbance type, severity, and time. In addition, non-disturbed areas were adjusted to account for vegetation growth and succession. A vegetation transition database developed for LF 2008 was updated to include new disturbance pathways and longer times since disturbance. In forested areas, vegetation transitions were modeled using the Forest Vegetation Simulator . In non-forested areas, staff ecologists and regional vegetation experts defined transition rulesets for each vegetation type.
Several enhancements were made to the LF 2010 existing vegetation type layer. Urban expansion was mapped using data from the National Land Cover Database (NLCD) 2006 land cover layer, which showed an increase of 8.6 percent of land mapped as urban between 2001 and 2006  (Figure 4). Refinement of agricultural lands was accomplished using the National Agricultural Statistics Service’s Cropland Data Layer 2011 . Wetland extents were enhanced using NLCD 2006 land cover and National Wetlands Inventory data.
In Alaska, forest canopy height was re-mapped to take advantage of additional data sources and to provide more detailed forest height data. Spaceborne lidar data from the Geoscience Laser Altimeter System (GLAS) were used to provide discrete training samples statewide. Because sloping terrain can impact height derivation from the GLAS waveforms, methods presented by Pang et al.  were adapted for use in areas of high relief, incorporating airborne lidar data from the Kenai Peninsula. The height estimates from the GLAS waveforms were used to develop regression tree models with Landsat imagery from the Web-enabled Landsat Data project , elevation data and derivatives from the National Elevation Dataset (NED), and NLCD 2001 land cover  as independent variables. The final forest height map was classified into 5-meter increments. A qualitative review indicated spatial patterns fit reasonably well with the imagery and expected height patterns (Figure 5).
Surface fuel models in all 50 states were updated based on defined rulesets that incorporate potential vegetation types, the updated vegetation type and structure layers, exotics information, and topographic data. Changes in canopy fuels were modeled using processes developed for LF 2008. The updated fuel datasets were then used to model regional fire behavior characteristics that were compared to previous data versions to quantify the effects of fuel updates.
An additional component of LF 2010 was the mapping of U.S.-affiliated islands in the Caribbean and the Pacific. A subset of the LANDFIRE products, including topographic layers, existing vegetation, and fuels, were produced for: American Samoa, Guam, the Northern Mariana Islands (NMI), the Federated States of Micronesia, Palau, the Marshall Islands, Puerto Rico, and the U.S. Virgin Islands. Elevation data were acquired from NED. Vegetation type maps were acquired from the U.S. Forest Service and cross-walked to common legends separately for the Pacific and Caribbean islands. Vegetation height was mapped using regression tree models with height estimates from GLAS waveforms as training data and Landsat composites, elevation data and derivatives, landcover from NLCD 2001 or National Oceanic and Atmospheric Administration Coastal Change Analysis Program, and ALOS PALSAR L-band SAR backscatter as independent variables. Airborne lidar data were available for Guam, the island of Saipan (NMI), the U.S. Virgin Islands, and portions of Puerto Rico (Figure 6), and were used for terrain correction of the GLAS data, as in Alaska. Canopy cover was modeled using the airborne lidar where it was available. Regression tree models were developed to relate the lidar-derived canopy cover to the same independent variables used for height. Two models were developed, one for the islands in the Pacific and another for the Caribbean islands. Surface and canopy fuels were mapped by creating rulesets based on existing vegetation products. The result was a subset of the LANDFIRE product suite for each of the islands. Since this was the initial mapping of these areas, disturbance products were not created.
LANDFIRE data distribution policy specifies that the three most current versions of the data be available from the LANDFIRE data distribution site. With the release of LF 2010, the three datasets available are LANDFIRE 2001, LF 2008, and LF 2010. All previous versions have been archived and may be made available upon special request. Along with the release of LF 2010 data, a re-designed program website has also been released with updated content, enhanced usability, and more efficient navigation.
Future of the LANDFIRE Program
Numerous process improvements are underway as LANDFIRE prepares for the next round of product updates. Data sharing agreements are being negotiated with the Natural Resource Conservation Service’s National Resources Inventory program to add numerous high-quality data points to the LANDFIRE Reference Database in rangeland areas, and with the Bureau of Indian Affairs to obtain additional plots on tribal lands. Upgrades to the data distribution system are planned, including a Web Coverage Service interface that will enable direct access to the LANDFIRE data.
With the decommissioning of Landsat 5 and well-known data gap anomalies in Landsat 7, the next round of LANDFIRE RSLC processing will be challenging. Data available for 2012 will be predominantly from Landsat 7. Algorithm development is underway to maximize the utility of the Landsat 7 Scan Line Corrector-Off data. With the successful launch of Landsat 8, high-quality data for 2013 and beyond are expected to be available, though existing change detection algorithms may need to be modified to work with the new sensor. Landsat 8’s Operational Land Imager (OLI) collects data in wavelength bands that are slightly modified from previous Landsat instruments, and at an increased radiometric resolution. Therefore, the sensitivity of change detection algorithms to detect various disturbances using OLI data may be different, and the algorithms may need to be adjusted accordingly. A comparison between Landsat 7 and Landsat 8 imagery over a wildfire area is shown in Figure 7.
LF 2010 represents the most current comprehensive database of vegetation and wildland fuel information for the entire U.S. Multiple Earth-observing datasets were instrumental in the creation of new maps, improvement of previous products, and subsequent update of the products. The resulting data products from LF 2010 provide a robust array of information for managers and decision-makers to help improve strategies and tactics to mitigate fire risk, anticipate fire behavior, and promote fire recovery to protect people, homes, and neighborhoods from wildland fires. Similarly, as a national consistent dataset, LANDFIRE provides needed vegetation characteristics to support ecological assessments, monitoring, and planning.
While LANDFIRE datasets are produced only in the U.S., the methods used to produce LANDFIRE data could be applicable to any area. LANDFIRE mapping and updating methods depend on Landsat imagery, topographic data, and ground plot information. Many countries have national inventories of timber and other lands that would likely be suitable for LANDFIRE-style mapping. Earth observation datasets are available globally, enabling similar approaches for generating consistent and up-to-date vegetation and fuel data.
We acknowledge leadership direction and funding from the U.S. Department of the Interior Office of Wildland Fire and US Department of Agriculture Forest Service Fire & Aviation Management. We are grateful to all who have contributed data and information as well as our partners in sharing data to improve LANDFIRE. We further acknowledge all of the LANDFIRE teams and resources for their work and contributions.
The work of J. Connot was performed under USGS contract G10PC00044. The work of B. Peterson and J. Picotte was performed under USGS contract G13PC00028.
 M. G. Rollins, “LANDFIRE: a nationally consistent vegetation, wildland fire, and fuel assessment,” International Journal of Wildland Fire, vol. 18, pp. 235-249, 2009.
 K. C. Ryan and T. S. Opperman, “LANDFIRE – A national vegetation/fuels data base for use in fuels treatment, restoration, and suppression planning,” Forest Ecology and Management, vol. 294, pp. 208-216, 2013.
 K. J. Nelson, J. Connot, B. Peterson, and C. Martin, “The LANDFIRE Refresh strategy: updating the national dataset,” Fire Ecology, vol. 9, pp. 80-101, 2013.
 J. G. Masek, E. F. Vermote, N. E. Saleous, R. Wolfe, F. G. Hall, K. F. Huemmrich, et al., “A landsat surface reflectance dataset for North America, 1990-2000,” IEEE Geoscience and Remote Sensing Letters, vol. 3, pp. 68-72, 2006.
 G. Schmidt, C. Jenkerson, J. Masek, E. Vermote, and F. Gao, “Landsat ecosystem disturbance adaptive processing system (LEDAPS) algorithm description,” US Geological Survey, Reston, VA, Open File Report 2013-1057, 2013.
 J. A. Fry, G. Xian, S. Jin, J. A. Dewitz, C. G. Homer, L. Yang, et al., “Completion of the 2006 national land cover database for the conterminous United States,” Photogrammetric Engineering and Remote Sensing, vol. 77, pp. 858-864, 2011.
 S. Jin, L. Yang, P. Danielson, C. Homer, J. Fry, and G. Xian, “A comprehensive change detection method for updating the National Land Cover Database to circa 2011,” Remote Sensing of Environment, vol. 132, pp. 159-175, 2013.
 G. E. Dixon, “Essential FVS: A user’s guide to the Forest Vegetation Simulator,” U.S. Department of Agriculture Forest Service, Forest Management Service Center, Fort Collins, CO, Internal Report,2002.
 C. Boryan, Z. Yang, R. Mueller, and M. Craig, “Monitoring US agriculture: the US Department of Agriculture, National Agricultural Statistics Service, Cropland Data Layer Program,” Geocarto International, vol. 26, pp. 341-358, 2011.
 Y. Pang, M. Lefsky, H.-E. Andersen, M. E. Miller, and K. Sherrill, “Validation of the ICEsat vegetation product using crown-area-weighted mean height derived using crown delineation with discrete return lidar data,” Canadian Journal of Remote Sensing, vol. 34, pp. S471-S484, 2008.
 D. P. Roy, J. Ju, K. Kline, P. L. Scaramuzza, V. Kovalskyy, M. Hansen, et al., “Web-enabled Landsat Data (WELD): Landsat ETM+ composited mosaics of the conterminous United States,” Remote Sensing of Environment, vol. 114, pp. 35-49, 2010.
 C. Homer, C. Huang, L. Yang, B. K. Wylie, and M. J. Coan, “Development of a 2001 National Land Cover Database for the United States,” Photogrammetric Engineering and Remote Sensing, vol. 70, pp. 829-840, 2004.