Prototyping A Methodology To Develop Regional-Scale Forest Aboveground Biomass Carbon Maps Predicted From Landsat Time Series, Trained From Field and Lidar Data Collections, And Independently Validated With FIA Data
Current Monitoring Reporting and Verification (MRV) needs cannot be met by using only available NASA satellite data products, but must be integrated with commercial off-the-shelf technologies. The exceptional sensitivity of commercial, airborne scanning lidar data to forest canopy structure has made it the best remote sensing technology for predicting vegetation attributes, including biomass. We propose to use multiple, landscape-level lidar datasets, previously acquired in conjunction with project-level field plot datasets for model calibration/validation, to predict aboveground biomass stores across representative vegetation types in the northwestern USA. The predicted biomass maps will serve as training area for upscaling biomass carbon predictions to the regional level, as predicted from Landsat time series imagery processed through LandTrendr. Regional maps will be validated with FIA data summarized at the county level, along with error statistics. Bias between biomass predictions and FIA observations summarized for the representative vegetation types will be quantified, and bias corrections applied, with the goal of maintaining a transparent record of bias corrections at the county level. We envision a lidar and field plot database that can continue to be updated as new project-level forest inventory data are collected. This strategy will actively engage forest managers by utilizing existing data collected by and maintained by land managers of the US Forest Service (USFS) and other public and private stakeholders. Our chosen study region is the northwestern USA, where multiple commercial lidar and field plot datasets exist, LandTrendr data products are farthest along in the production line, and steep environmental gradient provide an exceptional diversity in vegetation types. The cumulative area of LiDAR collections across multiple ownerships in the northwestern USA has reached the point that land managers of the USFS and other stakeholders need to develop a strategy for how to utilize LiDAR for improved regional inventory, and because these inventories are the initial conditions for simulation modeling of future conditions, the strategy will result in more accurate estimates of projected conditions. We have assembled and consistently processed field plot and lidar datasets at >21 landscape-level project areas distributed along a broad climate gradient across the northwestern USA from temperate rainforest to cold desert. We propose to employ imputation as our predictive modeling strategy because it assigns actual ground observations at representative sample locations, to unsampled locations. Further, imputation modeling is firmly ensconced within the forest management community, and has been used for decades to assign stand attributes from reference stands to target stands. Therefore, forest and rangeland managers of the USFS and other major public and private land management stakeholders will have little difficulty buying in to our proposed methodology, and would benefit enormously by making more effective use of available LiDAR and ground inventory data. Fortunately, the USFS has also developed a carbon management capability with greater utility to local forest managers: the carbon accounting tool of the Forest Vegetation Simulator (FVS) (http://www.fs.fed.us/fmsc/fvs/). FVS remains freely available, is now open source (Open-FVS), is approved by the American Carbon Registry to estimate carbon stock changes, and provides the option of climate change projections using Climate-FVS. Our chosen modeling methods and tools lend themselves to transparency and verifiability. Our goal is to develop a prototype CMS that works with acceptable accuracy, objectivity, transparency, and reproducibility in the northwestern USA, it will be ready for replication and application elsewhere in the USA, and globally with ties to SilvaCarbon and REDD+.
Product Title: Annual Aboveground Biomass Maps for Forests in the Northwestern USA, 2000-2016
Start Date: 01/2000End Date: 12/2016 (2000-2016)
Description: This dataset provides annual maps of aboveground biomass (AGB, Mg/ha) for forests in Washington, Oregon, Idaho, and western Montana, USA, for the years 2000-2016, at a spatial resolution of 30 meters. Tree measurements were summarized with the Fire and Fuels Extension of the Forest Vegetation Simulator (FFE-FVS) to estimate AGB in field plots contributed by stakeholders, then lidar was used to predict plot-level AGB using the Random Forests machine learning algorithm. The machine learning outputs were used to predict AGB from Landsat time series imagery processed through LandTrendr, climate metrics generated from 30-year climate normals, and topographic metrics generated from a 30-m Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM). The non-forested pixels were masked using the PALSAR 2009 forest/nonforest mask.
Status: Archived
CMS Science Theme(s): Decision Support; Land Biomass; MRV
Input Data Products: Commercial airborne lidar data, field plot datasets, LandTrendr (Landsat time series), climate metrics derived from 1961-1990 normals, topographic metrics derived from void-filled 30 m SRTM digital elevation model, Simard et al. (2011) global canopy height product
Algorithm/Models Used: LandTrendr, Random Forests regression modeling
Evaluation: Regional maps will be calibrated with FIA plot data
Intercomparison Efforts/Gaps: Comparison to Kennedy et al. GNN-derived aboveground biomass carbon maps in WA and OR
Uncertainty Estimates: Maps (30 m resolution) of standard deviation from the random forest model as a measure of precision; Comparison to independent FIA estimates as a measure of bias.
Potential Users: Federal land management agencies (e.g., US Forest Service), state land and other public and private forest managers
Stakeholders: Confederated Tribes of the Colville Reservation (Point of Contact: Cody Desautel, cody.desautel@colvilletribes.com); Mason, Bruce & Girard, Inc. (Point of Contact: pgould@masonbruce.com); Northwest Management, Inc. (Point of Contact: Mark Corrao, mcorrao@nmi2.com); The Nature Conservancy (Point of Contact: Ryan Haugo, rhaugo@TNC.ORG); U.S. Forest Service Region 4 (Point of Contact: Jed Gregory, jed.gregory@usda.gov); U.S. Forest Service Region 6 (Point of Contact: Jim Muckenhoupt, jim.muckenhoupt@usda.gov); USFS Region 1 (Point of Contact: Renate Bush, renate.bush@usda.gov); Washington Department of Natural Resources (Point of Contact: Luke Rogers, lwrogers@uw.edu)
Current Application Readiness Level: 5
Start Application Readiness Level: 4
Target Application Readiness Level: 7
Future Developments: Phase 2 (funded) will expand spatially to entire U.S. West and temporally to 1984-2020
Limitations: Local inaccuracies can be caused by the choice of Forest/Non-Forest mask used; we used a F/NF mask independently derived from 2009 PALSAR data and publicly available
Archived Data Citation: Fekety, P.A., and A.T. Hudak. 2019. Annual Aboveground Biomass Maps for Forests in the Northwestern USA, 2000-2016. ORNL DAAC, Oak Ridge, Tennessee, USA. DOI: 10.3334/ORNLDAAC/1719
Bounding Coordinates:
West Longitude:
-127.52000
East Longitude:
-110.31000
North Latitude:
50.79000
South Latitude:
39.81000
Product Title: CMS: Pinyon-Juniper Forest Live Aboveground Biomass, Great Basin, USA, 2000-2016
Start Date: 01/2000End Date: 12/2016 (2000-2016)
Description: This dataset provides annual maps of live aboveground tree biomass (Mg/ha) for pinyon-juniper forests across the Great Basin of the Western USA for the years 2000-2016 at a spatial resolution of 30 meters. Biomass estimates are limited to areas of the Great Basin defined as a pinyon-juniper ecosystem type by the 2016 Landfire Existing Vegetation Type map. The estimates of biomass were based on a linear relationship with pinyon-juniper canopy cover and crown-based allometrics developed from field data in Nevada and Idaho. Canopy cover was estimated from remote sensing by using annual composites of Landsat imagery, which were temporally segmented with the LandTrendr algorithm, along with biologically-relevant climate variables, and topographic indices in a Random Forest regression model. Models of canopy cover were trained from semi-automatic extraction of tree crowns from 2011 - 2013 high resolution imagery (1 m) from the National Agriculture Imagery Program, which were validated with photo interpretation. Maps of the standard deviation of biomass estimates from decision trees in the Random Forest model are provided as an indicator of uncertainty. Biomass estimates were calibrated to estimates from the Forest Inventory and Analysis program (FIA) on an annual basis and corrections applied.
Input Data Products: Landsat time series, gridded 30-year climate normals, topography derived from the National Elevation Dataset
Algorithm/Models Used: LandTrendr, Random Forest, Spatial Wavelet Analysis
Evaluation: Field-based estimates from the SageSTEP and Forest Inventory and Analysis programs
Intercomparison Efforts/Gaps:
Uncertainty Estimates: Maps of standard deviation from the random forest model. Comparisons to SageSTEP and FIA.
Uncertainty Categories: ensemble and model-data comparison
Application Areas: MRV, Land management, Cap-and-trade
Relevant Policies/Programs: US National Greenhouse Gas Inventory (NGHGI) baseline reporting to the UN Framework Convention on Climate Change (UNFCCC), U.S. Carbon Cycle Science Program (USCCSP), North American Carbon Program (NACP)
Potential Users: USFS, BLM, EPA, stakeholders in land management
Stakeholders: The Nature Conservancy (Point of Contact: Ryan Haugo, rhaugo@TNC.ORG); U.S. Forest Service Region 4 (Point of Contact: Jed Gregory, jed.gregory@usda.gov)
Current Application Readiness Level: 5
Start Application Readiness Level: 3
Target Application Readiness Level: 7
Future Developments: - Continued methods development
- Share results with forest service partners
Limitations: - Some exclusion of areas with low tree cover and prior disturbance
Archived Data Citation: Filippelli, S.K., M.J. Falkowski, A.T. Hudak, and P.A. Fekety. 2020. CMS: Pinyon-Juniper Forest Live Aboveground Biomass, Great Basin, USA, 2000-2016. ORNL DAAC, Oak Ridge, Tennessee, USA. DOI: 10.3334/ORNLDAAC/1755
Start Date: 01/2002End Date: 12/2016 (2002 through 2016)
Description: This dataset provides maps of aboveground forest biomass (AGB) of living trees and standing dead trees in Mg/ha across portions of Northwestern United States, including Washington, Oregon, Idaho, and Montana, at a spatial resolution of 30 m. Forest inventory data were compiled from 29 stakeholders that had overlapping lidar imagery. The collection totaled 3805 field plots with lidar imagery for 176 collections acquired between 2002 and 2016. Plot-level AGB estimates were calculated from tree measurements using the default allometric equations found in the Fire Fuels Extension (FFE) of the Forest Vegetation Simulator (FVS). The random forest algorithm was used to model AGB from lidar height and density metrics that were generated from the lidar returns within fixed-radius field plot footprints, gridded climate metrics obtained from the Climate-FVS Ready Data Server, and topographic estimates extracted from Shuttle Radar Topography Mission (SRTM) 1 Arc-Second Global elevation rasters. AGB was then mapped from the same lidar metrics gridded across the extent of the lidar collections at 30-m resolution. The standard deviation of estimated AGB of the terminal nodes from the random forest predictions was also mapped to show pixel-level model uncertainty. Note that the AGB estimates are, for the most part, a single snapshot in time and that the forest conditions are not necessarily representative of the larger study area.
Status: Archived
CMS Science Theme(s): Land Biomass
Keywords:
Spatial Extent: Northwestern United States: Washington, Oregon, Idaho, and part of western Montana
Spatial Resolution: 30 m
Temporal Frequency: annual
Input Data Products:
Algorithm/Models Used:
Evaluation:
Intercomparison Efforts/Gaps:
Uncertainty Estimates:
Uncertainty Categories:
Application Areas:
Relevant Policies/Programs:
Potential Users:
Stakeholders: Confederated Tribes of the Colville Reservation (Point of Contact: Cody Desautel, cody.desautel@colvilletribes.com); Mason, Bruce & Girard, Inc. (Point of Contact: pgould@masonbruce.com); Northwest Management, Inc. (Point of Contact: Mark Corrao, mcorrao@nmi2.com); The Nature Conservancy (Point of Contact: Ryan Haugo, rhaugo@TNC.ORG); U.S. Forest Service Region 4 (Point of Contact: Jed Gregory, jed.gregory@usda.gov); U.S. Forest Service Region 6 (Point of Contact: Jim Muckenhoupt, jim.muckenhoupt@usda.gov); USFS Region 1 (Point of Contact: Renate Bush, renate.bush@usda.gov); Washington Department of Natural Resources (Point of Contact: Luke Rogers, lwrogers@uw.edu)
Archived Data Citation: Fekety, P.A., and A.T. Hudak. 2020. LiDAR Derived Forest Aboveground Biomass Maps, Northwestern USA, 2002-2016. ORNL DAAC, Oak Ridge, Tennessee, USA. DOI: 10.3334/ORNLDAAC/1766
Bounding Coordinates:
West Longitude:
-125.58000
East Longitude:
-112.28000
North Latitude:
49.35000
South Latitude:
41.66000
Product Title: Tree and stand attributes for "A carbon monitoring system for mapping regional, annual aboveground biomass across the northwestern USA"
Start Date: 01/2002End Date: 12/2017 (These data were collected between 2002 and 2017)
Description: These data represent a portion of the forest inventory data used in Hudak et al. (in review) "A carbon monitoring system for mapping regional, annual aboveground biomass across the northwestern USA". This study used forest inventory data located in lidar units along with Landsat data, topographic metrics, and climate metrics to create maps of forested biomass across the northwestern USA (Washington, Oregon, Idaho, and western Montana.) The data requirements for inclusion in this study included: 1. fixed-area plots, 2. plots centers were recorded using a global navigation satellite system receiver (e.g., a GPS receiver) capable of differential correction, and 3. plots were located in a lidar unit where tree data were collected within 3 years of the lidar collection. A shapefile of the lidar units can be found in Fekety and Hudak (2020, https://doi.org/10.3334/ORNLDAAC/1766). The forest inventory data presented here (n = 2,680 plots) include all data that could be made publicly available and have been compiled from numerous existing datasets. The forest inventory data were collected using project-specific sampling plans and therefore these data have been formatted to be read by the Forest Vegetation Simulator (FVS; https://www.fs.fed.us/fvs/). The forest inventory data in this dataset were collected between 2002 and 2017 and located in Idaho, Oregon, and Washington, USA.
Status: Archived
CMS Science Theme(s): Land Biomass
Keywords:
Spatial Extent: Idaho, Oregon, and Washington, USA.
Spatial Resolution: Inventories so measurements on an individual tree level
Temporal Frequency:
Input Data Products:
Algorithm/Models Used:
Evaluation:
Intercomparison Efforts/Gaps:
Uncertainty Estimates:
Uncertainty Categories:
Application Areas:
Relevant Policies/Programs:
Potential Users:
Stakeholders: Washington Department of Natural Resources (Point of Contact: Luke Rogers, lwrogers@uw.edu)
Archived Data Citation: Fekety, P.A., A.T. Hudak and B.C. Bright. 2020. Tree and stand attributes for 'A carbon monitoring system for mapping regional, annual aboveground biomass across the northwestern USA'. Fort Collins, CO: Forest Service Research Data Archive. DOI: 10.2737/RDS-2020-0026.
Bounding Coordinates:
West Longitude:
-124.42690
East Longitude:
-115.02440
North Latitude:
48.98967
South Latitude:
42.10223
Publications:
Fekety, P. A., Crookston, N. L., Hudak, A. T., Filippelli, S. K., Vogeler, J. C., Falkowski, M. J. 2020. Hundred year projected carbon loads and species compositions for four National Forests in the northwestern USA. Carbon Balance and Management. 15(1). DOI: 10.1186/s13021-020-00140-9
Fekety, P. A., Falkowski, M. J., Hudak, A. T. 2015. Temporal transferability of LiDAR-based imputation of forest inventory attributes. Canadian Journal of Forest Research. 45(4), 422-435. DOI: 10.1139/cjfr-2014-0405
Fekety, P. A., Falkowski, M. J., Hudak, A. T., Jain, T. B., Evans, J. S. 2018. Transferability of Lidar-derived Basal Area and Stem Density Models within a Northern Idaho Ecoregion. Canadian Journal of Remote Sensing. 44(2), 131-143. DOI: 10.1080/07038992.2018.1461557
Fekety, P. A., Sadak, R. B., Sauder, J. D., Hudak, A. T., Falkowski, M. J. 2019. Predicting forest understory habitat for Canada lynx using LIDAR data. Wildlife Society Bulletin. 43(4), 619-629. DOI: 10.1002/wsb.1018
Filippelli, S. K., Falkowski, M. J., Hudak, A. T., Fekety, P. A., Vogeler, J. C., Khalyani, A. H., Rau, B. M., Strand, E. K. 2020. Monitoring pinyon-juniper cover and aboveground biomass across the Great Basin. Environmental Research Letters. 15(2), 025004. DOI: 10.1088/1748-9326/ab6785
Fusco, E. J., Rau, B. M., Falkowski, M., Filippelli, S., Bradley, B. A. 2019. Accounting for aboveground carbon storage in shrubland and woodland ecosystems in the Great Basin. Ecosphere. 10(8). DOI: 10.1002/ecs2.2821
Hudak, A. T., Fekety, P. A., Kane, V. R., Kennedy, R. E., Filippelli, S. K., Falkowski, M. J., Tinkham, W. T., Smith, A. M. S., Crookston, N. L., Domke, G. M., Corrao, M. V., Bright, B. C., Churchill, D. J., Gould, P. J., McGaughey, R. J., Kane, J. T., Dong, J. 2020. A carbon monitoring system for mapping regional, annual aboveground biomass across the northwestern USA. Environmental Research Letters. 15(9), 095003. DOI: 10.1088/1748-9326/ab93f9
Sanchez-Lopez, N., Boschetti, L., Hudak, A. 2018. Semi-Automated Delineation of Stands in an Even-Age Dominated Forest: A LiDAR-GEOBIA Two-Stage Evaluation Strategy. Remote Sensing. 10(10), 1622. DOI: 10.3390/rs10101622
Sanchez-Lopez, N., Boschetti, L., Hudak, A. T. 2019. Reconstruction of the disturbance history of a temperate coniferous forest through stand-level analysis of airborne LiDAR data. Forestry: An International Journal of Forest Research. DOI: 10.1093/forestry/cpz048
Stitt, J. M., Hudak, A. T., Silva, C. A., Vierling, L. A., Vierling, K. T. 2021. Characterizing individual tree-level snags using airborne lidar-derived forest canopy gaps within closed-canopy conifer forests. Methods in Ecology and Evolution. 13(2), 473-484. DOI: 10.1111/2041-210X.13752
Stitt, J. M., Hudak, A. T., Silva, C. A., Vierling, L. A., Vierling, K. T. 2022. Evaluating the Use of Lidar to Discern Snag Characteristics Important for Wildlife. Remote Sensing. 14(3), 720. DOI: 10.3390/rs14030720
Tinkham, W. T., Mahoney, P. R., Hudak, A. T., Domke, G. M., Falkowski, M. J., Woodall, C. W., Smith, A. M. 2018. Applications of the United States Forest Inventory and Analysis dataset: a review and future directions. Canadian Journal of Forest Research. 48(11), 1251-1268. DOI: 10.1139/cjfr-2018-0196
Deo, R. K., Froese, R. E., Falkowski, M. J., Hudak, A. T. 2016. Optimizing Variable Radius Plot Size and LiDAR Resolution to Model Standing Volume in Conifer Forests. Canadian Journal of Remote Sensing. 42(5), 428-442. DOI: 10.1080/07038992.2016.1220826
Archived Data Citations:
Fekety, P.A., and A.T. Hudak. 2019. Annual Aboveground Biomass Maps for Forests in the Northwestern USA, 2000-2016. ORNL DAAC, Oak Ridge, Tennessee, USA. DOI: 10.3334/ORNLDAAC/1719
Filippelli, S.K., M.J. Falkowski, A.T. Hudak, and P.A. Fekety. 2020. CMS: Pinyon-Juniper Forest Live Aboveground Biomass, Great Basin, USA, 2000-2016. ORNL DAAC, Oak Ridge, Tennessee, USA. DOI: 10.3334/ORNLDAAC/1755
Fekety, P.A., A.T. Hudak and B.C. Bright. 2020. Tree and stand attributes for 'A carbon monitoring system for mapping regional, annual aboveground biomass across the northwestern USA'. Fort Collins, CO: Forest Service Research Data Archive. DOI: 10.2737/RDS-2020-0026.
Fekety, P.A., and A.T. Hudak. 2020. LiDAR Derived Forest Aboveground Biomass Maps, Northwestern USA, 2002-2016. ORNL DAAC, Oak Ridge, Tennessee, USA. DOI: 10.3334/ORNLDAAC/1766
2015 NASA Carbon Cycle & Ecosystems Joint Science Workshop Poster(s)
Developing an Ecoregion-level Imputation Model From LiDAR-derived Biomass Maps
-- (Andrew Thomas Hudak, Patrick A Fekety, Michael J Falkowski, Robert E Kennedy, Alistair Matthew Stuart Smith)
[abstract]