Forest Inventory and Analysis (FIA) in Alaska

Use of High-Resolution LIDAR for Forest Inventory

Credits and Responsiblities

Project Sponsor

Forest Inventory and Analysis Program

University of Washington
School of Environmental and Forest Sciences
Precision Forestry Cooperative
Box 352100
Seattle, WA 98195-2100
USDA Forest Service
PNW Research Station
Silviculture and Forest Models Team
Seattle, WA


Hans-Erik Andersen
Forest Inventory and Analysis, USDA Forest Service, Pacific Northwest Research Station, Anchorage, AK

Robert J. McGaughey
PNW Research Station
Stephen Reutebuch
PNW Research Station


The USDA Forest Service, Forest Inventory and Analysis (FIA) program is the Nation's forest census. The Pacific Northwest Research Station (PNW) is one of five Research Stations with an FIA program. It is FIA's job to determine the extent and condition of forest resources and analyze how these resources change over time. The inventories are conducted across all ownerships in Alaska, California, Hawaii, Oregon, Washington, and the US Pacific Territories regardless of management policies. FIA consists of a nationally consistent core program, which can be enhanced at the regional, state, or local level by collecting additional data to address special interests. The national core consists of three phases:

Phase 1 uses remote sensing imagery or aerial photography to classify land into forest or non-forest and to identify landscape patterns such as fragmentation and urbanization. Historically, this phase was accomplished exclusively with aerial photographs. Current methods are shifting to a system based on remote sensing imagery.

Phase 2 consists of permanently established field plots distributed across the landscape with approximately one plot every 6,000 acres. In the West, field crews visit 10 percent of forested sample locations each year to collect a variety of forest ecosystem data. A typical plot usually takes a 2-3 person field crew one full day to complete. Factors such as the steepness of the ground, the size of trees, amount of understory vegetation, and the length of the hike to the plot all factor into the time commitment. When a field crew visits a plot some of the information includes:

This proposal is submitted under Joint Fire Sciences Program Solicitation 2004-1, Task 2 which solicits studies that use pre-fire fuels conditions, post-fire data, and fire behavior for 2003 fires, with emphasis on sites where pre-fire data are available.

Tree diameter, length, damage, amount of rotten or missing wood, and tree quality

Counts of tree regeneration

General land use

Stand characteristics such as forest type, stand age, and disturbance

Changes in land use and general stand characteristics

Estimates of growth, mortality, and removals (determined by revisiting plots every ten years)

Vegetation diversity and structure

Down woody debris

Phase 3 is designed to assess forest health by sampling a subset of Phase 2 plots. Plots are visited only during the growing season and the entire Phase 3 inventory cycle is completed in 5 years. Phase 3 measurements relate to forest ecosystem function, condition, and health.

In this study, we will explore how airborne laser scanning (also known as light detection and ranging or LIDAR) and high-resolution imagery can be used for estimating some forest structure and composition variables in these three sample phases over scattered FIA forest inventory plots. LIDAR data have been shown to accurately measure forest overstory and high-resolution image data are useful in determining forest composition and structure. By fusing these two remote sensing data sources, more reliable and more complete vegetation measures may be obtained.

Likewise, sampling (as opposed to total landscape data coverage) with LIDAR and high-resolution imagery may provide an efficient and economical method for supplementing future FIA inventories, particularly in remote regions such as Alaska, where costs to install and remeasure ground plots are impeding full inventory implementation. We propose two stages within this study:

Stage 1—Alaska Pilot Study

This stage of the study will commence in Spring 2004. LIDAR data will be collected over 100-200 recently measured FIA plots on the Chugach National Forest in Alaska. In addition, high-resolution imagery will be acquired over a subset of the LIDAR plots. Correlations between FIA plot data and LIDAR- and high-resolution imagery-derived vegetation structure and composition variables will be analyzed and recommendations for integrating LIDAR sampling techniques into an interior Alaska remote sensing framework will be developed.

Stage 2—Washington, Oregon, N. California Study

This stage of the study will commence in Spring 2005. A similar, but expanded study is proposed for FIA plots (several hundred) within Oregon, Washington, or Northern California. This follow-up study will build on the knowledge gained from the Alaska pilot study. The objectives of the study will be to design and test LIDAR and high-resolution imagery collection and analysis techniques over the more diverse forest types found in the broader PNW-FIA sampling region. Final study design and data collection and analysis protocols for this second stage will not be complete until early 2005, after results from the Alaska pilot study are available. Plots would be chosen from the subset of the WA/OR/CA plots that are planned for remeasurement in FY05 by the PNW-FIA field crews.

Objectives and Scope

1. Test the feasibility of collecting high-resolution LIDAR data over scattered FIA plots

2. Test the feasibility of collecting georeferenced, high-resolution imagery over the same scattered plots

3. Develop methods for aligning field plot data with LIDAR data

4. Develop a method for fusing high-resolution imagery and LIDAR data

5. Test for correlations between LIDAR-derived metrics and FIA ground measurements

6. Test for correlations between high-resolution imagery and FIA ground measurements

7. Test the ability of LIDAR to distinguish between hardwoods and conifers

8. Examine how LIDAR and high-resolution imagery data can be used to distinguish between overstory tree species

9. Examine how LIDAR and high-resolution imagery data can be used to identify tree mortality or decline in tree vigor

Methods and Analysis


FIA inventory plot measurements:

Alaska Pilot stage—Recent ground data has been collected on FIA ground plots in the Kenai peninsula region. Accurate GPS positions of FIA plots will be collected (by PNW-FIA, Alaska) in the 2004 field season for those plots where it is not possible to accurately align LIDAR data with existing plot stem maps. LIDAR data will be flown over the plots in leaf-off conditions in spring 2004.

Washington, Oregon, N. California stage—During the 2005 field season, PNW-FIA crews will remeasure ground plot on several hundred plots. They will carefully GPS plot locations so that LIDAR data may be accurately aligned with plots. For a subset of these remeasured plots LIDAR data will be collected in leaf-off conditions in fall 2005 or winter 2006.


LIDAR-based estimation of inventory parameters:

The utility of small-footprint LIDAR for estimation of stand inventory parameters has been well-established in previous studies (Means et al, 2000; Naesset, 2002). The preliminary results of several research studies conducted by the UW Precision Forestry Cooperative and PNW Research Station has provided further validation of these findings, indicating that LIDAR can be used to generate accurate, plot-level estimates of basal area, stem volume, dominant height, stem density, and canopy fuel characteristics. Previous work has also shown that computer vision algorithms can be applied to high-density LIDAR data to extract accurate measurements of individual tree height, crown area, and crown base height (Persson et al., 2002).

In this study, regression analysis will be used to develop predictive models relating LIDAR-derived metrics to the following field-based inventory parameters:

Stem volume


Basal area

Mean diameter

Dominant height

Stem density

Vegetation cover

In addition, direct measures of crown diameter and height to base of crown will be made.

Analyses will be conducted at both the plot-level and individual tree-level. The predictive value of the regression models will be assessed through a model validation procedure.

It should be noted that the area covered by LIDAR at each plot will be approximately 9 hectares, which when compared to the field plot size of 0.07 ha (1/6 acre), represents a 13,000% increase in the sampling intensity over the current FIA sampling design. The larger size of the LIDAR plots will decrease sampling error and allow for more accurate estimation of spatially-variable stand characteristics, e.g., vegetation patch size and patterns not evident at the current FIA plot sample scale. If wide-area IFSAR or multispectral image data are available, the study will explore the potential for developing a multiphase sampling design using field data, LIDAR, digital imagery and possibly IFSAR (Hazard et al., 1995).

Species recognition and vegetation condition using LIDAR and high-resolution imagery:

Recent studies have shown that species class (i.e. decidous vs. coniferous) can be recognized using the intensity information from LIDAR data collected in leaf-off conditions (Brandtberg et al., 2003). Preliminary results of research conducted at the UW Precision Forestry Cooperative and PNW Research Station have also shown the potential utility of LIDAR for species recognition and condition classification. It is expected that fusion of the spectral content of digital imagery and the structural (and NIR reflectance) information contained in LIDAR will further improve the accuracy of species recognition algorithms (Leckie et al., 2003).

In this study, the utility of LIDAR for species recognition and condition classification will be assessed through comparison to the intensive measurements acquired at the field plots.


Brandtberg, T., T. Warner, R. Landenberger, and J. McGraw. 2003. Detection and analysis of individual leaf-off tree crowns in small footprint, high sampling density lidar data from the eastern deciduous forest in North America. Remote Sensing of Environment 85(3):290-303.

Hazard, J.H., H. Schreuder, and V. LaBau. 1995. The Alaska four-phase forest inventory sampling design using remote sensing and ground sampling. Photogrammetric Engineering and Remote Sensing 61(3): 291 - 302.

Leckie, D. F. Gougeon, D. Hill, R. Quinn, L. Armstrong, and R. Shreenan. 2003. Combined high-density lidar and multispectral imagery for individual tree crown analysis. Canadian Journal of Remote Sensing 29(5):633-649.

Means, J., S. Acker, B. Fitt, M. Renslow, L. Emerson, and C. Hendrix. 2000. Predicting forest stand characteristics with airborne scanning lidar. Photogrammetric Engineering & Remote Sensing 66(11): 1367-1371.

Naesset, E. 2002. Predicting forest stand characteristics with airborne scanning laser using a practical two-stage procedure and field data. Remote Sensing of Environment 80: 88-99.

Persson, A., J. Holmgren, and U. Soderman. 2002. Detecting and measuring individual trees using an airborne laser scanner. Photogrammetric Engineering and Remote Sensing 68(9): 925-932.