Elsevier

Remote Sensing of Environment

Volume 113, Issue 9, September 2009, Pages 1926-1938
Remote Sensing of Environment

Estimating the quantity and quality of coarse woody debris in Yellowstone post-fire forest ecosystem from fusion of SAR and optical data

https://doi.org/10.1016/j.rse.2009.05.001Get rights and content

Abstract

The Coarse Woody Debris (CWD) quantity, defined as biomass per unit area (t/ha), and the quality, defined as the proportion of standing dead logs to the total CWD quantity, greatly contribute to many ecological processes such as forest nutrient cycling, tree regeneration, wildlife habitat, fire dynamics, and carbon dynamics. However, a cost-effective and time-saving method to determine CWD is not available. Very limited literature could be found that applies remote sensing technique to CWD inventory. In this paper, we fused the wall-to-wall multi-frequency and multi-polarization Airborne Synthetic Aperture Radar (AirSAR) and optical Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) to estimate the quantity and quality of CWD in Yellowstone post-fire forest ecosystem, where the severe 1988 fire event resulted in high spatial heterogeneity of dead logs. To relate backscatter values to CWD metrics, we first reduced the terrain effect to remove the interference of topography on AirSAR backscatter. Secondly, we removed the influence of regenerating sapling by quadratic polynomial fitting between AVIRIS Enhanced Vegetation Index (EVI) and different channels backscatters. The quantity of CWD was derived from Phh and Phv, and the quality of CWD was derived from Phh aided by the ratio of Lhv and Phh. Two maps of Yellowstone post-fire CWD quantity and quality were produced. The calculated CWD quantity and quality were validated by extensive field surveys. Regarding CWD quantity, the correlation coefficient between measured and predicted CWD is only 0.54 with mean absolute error up to 29.1 t/ha. However, if the CWD quantity was discretely classified into three categories of “≤ 60”, “60–120”, and “≥ 120”, the overall accuracy is 65.6%; if classified into two categories of “≤ 90” and “≥ 90”, the overall accuracy is 73.1%; if classified into two categories of “≤ 60” and “≥ 60”, the overall accuracy is 84.9%. This indicates our attempt to map CWD quantity spatially and continuously achieved partial success; however, the general and discrete categories are reasonable. Regarding CWD quality, the overall accuracy of 5 types (Type 1—standing CWD ratio ≥ 40%; Type 2—15%  standing CWD ratio < 40%; Type 3—7%  standing CWD ratio< 15%; Type 4—3%  standing CWD ratio < 7%; Type 5—standing CWD ratio < 3%) is only 40.32%. However, when type 1, 2, 3 are combined into one category and type 4 and 5 are combined into one category, the overall accuracy is 67.74%. This indicates the partial success of our initial results to map CWD quality into detailed categories, but the result is acceptable if solely very coarse CWD quality is considered. Bias can be attributed to the complex influence of many factors, such as field survey error, sapling compensation, terrain effect reduction, surface properties, and backscatter mechanism understanding.

Introduction

The term Coarse Woody Debris (CWD) commonly refers to the sound and rotting logs and stumps that provide habitat for plants, animals and insects and a source of nutrients for soil development with material generally greater than 8–10 cm in diameter (Stevens, 1997). Standing dead snags are usually excluded; however, in our study they are included as CWD. CWD contributes to many ecological processes as a standing snag and fallen woody material (Brown et al., 2003), important for forest nutrient cycling (Trundell et al., 2004), tree regeneration (Takahashi et al., 2000, Ripple and Larsen, 2001), wildlife habitat (Bütler & Schlaepfer, 2004), fire dynamics (Schmoldt et al., 1999), and carbon dynamics (Ståhl et al., 2001, Bütler and Schlaepfer, 2004). Moreover, CWD quantity (t/ha) and quality (standing vs. downed) have different roles in fire studies (Hackett, 1989, Brown et al., 2003), wildlife conservation (Brown et al., 2003, Bate et al., 2004), and nutrient cycling studies (Remsburg & Turner, 2006). As a result, there is an optimum quantity of CWD in natural resource management; however, the optimum quantity for soil and fire considerations refers to downed CWD while the optimum quantity for wildlife involves both standing and downed CWD (Brown et al., 2003). With the increased focus on CWD, CWD assessments are now integral parts in many large scale forest inventories and also becoming increasingly frequent in various regional monitoring programs (Ståhl et al., 2001).

Despite the increased attention on the importance of CWD quantity and quality for such a broad spectrum of ecological processes, the CWD assessment has not been well studied. CWD tends to be ignored in many forest carbon budgets (Brown, 2002), and there is no long-term, comprehensive monitoring data available for carbon in forest floor and Downed Woody Debris (DWD) of U.S. forests (Birdsey, 2004). Current DWD estimates in forest health monitoring programs in the U.S. are made using a variety of modeling techniques supported by various data sets that partially represent the range of forest conditions and recovery from forest disturbances. Modeled estimates have unknown errors when extrapolated to areas and conditions not represented by the underlying data (Birdsey, 2004). CWD inventories are currently generally carried out using labour-intensive and expensive field methods, which is difficult in remote areas, rough terrain or steep slopes (Bütler & Schlaepfer, 2004). A limited number of sample plots may not provide reasonable estimates on the heterogeneity in forest floor and downed wood, particularly if forest disturbance history is unknown (Birdsey, 2004). A more practical, cost-effective, reliable method based on remote sensing techniques to quantify CWD on large areas would be a valuable contribution (Bütler & Schlaepfer, 2004). This is especially important for disturbance events such as logging, fire, insect mortality, windstorms, and timber harvest because large inputs of CWD occur following major disturbances.

The most common remote sensing based methods in practical CWD surveys are manual aerial photo interpretation, manual or automatic interpretation of satellite images, and ocular estimation from airplanes or helicopters. These methods seldom provide direct estimates of CWD key quantities but instead provide auxiliary data that can be used in the estimation phase or to effectively allocate the field sampling effort (Ståhl et al., 2001). Bütler and Schlaepfer (2004) used 23 cm color infrared aerial photos to map and quantify large-diameter (≥ 25 cm) standing dead snags by visually interpreting the spectral reflectance difference of living, damaged and dead trees. Their method has shown limited effectiveness for the inventory of broken snags. Marcus et al. (2002) attempted to map the spatial and temporal distributions of woody debris in limited streams of the Greater Yellowstone Ecosystem. Their attempts to map woody debris with 1-m resolution digital four-band imagery were generally unsuccessful, largely because of spectral confusion between gravel and wood in mixed pixels and coregistration problems. However, Marcus et al. (2003) used 128-band hyperspectral imagery with 1-m resolution to achieve an overall accuracy of 83% on matched filter mapping of woody debris in the fifth-order Lamar River, but there were logistical obstacles when wood is obstructed by water, vegetation, or shadows. All these limited studies examined the large-scale mapping; however, for applications at the landscape level, mapping with less detail but over larger areas would be preferable (Bütler & Schlaepfer, 2004). This is a particular problem for fire studies because the vast majority of scientific data has been collected at scales of 10−1 to 10 km2 and our ability to understand and manage for the effects of large fires has been limited by a lack of data for large spatial scales (Schmoldt et al., 1999). This problem was brought into sharp focus in 1988 during and after the large fires in the Yellowstone National Park (YNP) region (Schmoldt et al., 1999).

In 1988 Yellowstone was severely burnt by fires. The surprising variation in post-fire Lodgepole pine regeneration and understory herbaceous vegetation were studied by many scientists (e.g. Anderson and Romme, 1991, Turner et al., 1997, Turner et al., 1999, Romme and Turner, 2004, Reed et al., 1999, Turner et al., 2004). However, the spatial heterogeneity of post-fire CWD is not clear. Very few articles could be found reporting on the mapping of post-fire CWD. The ability to use remote sensing methods to classify and map spatially explicit structural characteristics of Yellowstone post-fire CWD was only examined by Marcus et al., 2002, Marcus et al., 2003. They were unsuccessful in mapping CWD with 1-m 4-band imagery but achieved overall accuracy of 83% using 1-m 128-band hyperspectral imagery. These two studies were conducted on limited streams. At a broad scale, little work has been done to quantify the spatially complex patterns that fire creates on the landscape. The ability of remote sensing to map post-fire CWD is rarely accomplished and is poorly documented, and thus not well incorporated into process-based ecological models. It is particularly difficult to find research that uses synthetic aperture radar (SAR) data to measure forest spatial patterns, especially with respect to fire history (Henry & Yool, 2005).

In this study, we seek to investigate the capability of the fusion of multi-frequency, multi-polarization Airborne Synthetic Aperture Radar (AirSAR) and optical data for post-fire CWD quantity and quality in Yellowstone. To do this, we first reduced the terrain effect and sapling contribution to backscatter by using local incidence angle and vegetation index compensation. The backscatter mechanisms were qualitatively analyzed and appropriate channels were used. We validated the results using our field survey data.

Section snippets

Study site

Our study area comprises YNP, the world's first and best-known national park, and immediately surrounding lands in northwest Wyoming, southwest Montana, and southeast Idaho, USA (Fig. 1) with elevation ranging from 1540 m to 3760 m. Approximately 83% of the total forested area of YNP is dominated by Lodgepole pine (Pinus contorta) ranging in age from seedlings to 300–400 years. Subalpine fir (Abies lasiocarpa), Engelmann spruce (Pica engelmannii), and whitebark pine (Pinus albicaulis) are also

Datasets

NASA/Jet Propulsion Lab (JPL) operated a multi-frequency (P: 0.44 GHz, 68 cm wavelength; L: 1.225 GHz, 24 cm; and C: 5.3 GHz, 5.6 cm) topographical SAR system on board a DC-8 aircraft (Zebker et al., 1992), which was used to collect 21 North–South radar stripes on July 12, 2003 for our study. The local weather report showed that there was only 1.8 cm precipitation occurring on June 14–24 during the period of June 1–July 12, 2003, indicating the ground conditions were dry and soil moisture was

Field surveys

In summer 2007, we inventoried 186 forest plots burnt in 1988 (Fig. 1). The center location of each plot was positioned by Trimble GPS for coordinates that were differentially corrected with an error around 1–2 m. Standing at the center of each plot, we used a digital camera to take photos in each cardinal direction determined by compass. We used the line intersect method (Brown, 1974, Van Wagner, 1968), which is widely used for estimating dead wood and debris on the forest floor, to inventory

Results

The CWD quantity and quality in Yellowstone post-fire forest ecosystem were mapped (Fig. 9). Fig. 9a shows the CWD quantity is the lowest in the area close to the town of West Yellowstone. The CWD quantity is high to the south of Mammoth, to the northwest of Cooke City, to the northwest of Pahaska, around Old Faithful, and at the southwest and southeast corners of Yellowstone Lake. In other areas, the CWD quantity is at a moderate level. Fig. 9b shows the post-fire dead logs are almost downed

Discussion

The results from our initial study of CWD with remote sensing data indicate that only partial success was achieved. This might be caused by the main deficiencies and limitations of our study, including the field survey method, the sapling backscatter compensation, the reduction of terrain effect on radar backscatter, the ignored surface properties, and the most important one—our limited understanding on backscatter mechanism. All these factors are discussed below.

First, various applications of

Conclusion

The CWD, both its quantity and quality, attracts more and more attention due to its significance to many ecological processes such as wildlife conservation, carbon management, fire simulation, and invasive species intrusion. However, there are no cost-effective and time-efficient ways to assess CWD. The quantity of dead wood does not generally correlate with any index of stand quality (Brown, 2002). Remote sensing can provide a landscape view of a specific site repeatedly, remotely, and

Acknowledgements

This work was done with financial support from Air Force Research Lab (No. F33615-03-C-1432) and NASA (No. NNA07CN19A and No. NNS06AA23G). This research was also supported by an appointment to the NASA Postdoctoral Program at the NASA Ames Research Center, administered by Oak Ridge Associated Universities through a contract with NASA. The authors would like to thank NASA JPL for their AirSAR and AVIRIS data acquisition and preprocessing in 2003 and 2006. The authors give special thanks to Mr.

References (58)

  • AlbrightT.P. et al.

    Classification of surface types using SIR-C/X-SAR, Mount Everest area, Tibet

    Journal of Geophysical Research

    (1998)
  • AndersonJ.E. et al.

    Initial floristics in Lodgepole pine (Pinus contorta) forests following the 1988 Yellowstone fires

    International Journal of Wildland Fire

    (1991)
  • BaskinY.

    Yellowstone fires: A decade later

    BioScience

    (1999)
  • BirdseyR.

    Data gaps for monitoring forest carbon in the United States: An inventory perspective

    Environmental Management

    (2004)
  • BrownJ.K.

    Handbook for inventorying downed woody material

    (1974)
  • ChristensenN.E. et al.

    Interpreting the Yellowstone fires of 1988

    BioScience

    (1989)
  • ColeD.M.

    A cubic-foot stand volume equation for lodgepole pine in Montana and Idaho. Research Note INT-150

    (1971)
  • DirksR.A. et al.

    The climate of Yellowstone and Grand Teton National Parks

  • DobsonM.C. et al.

    Estimation of forest biophysical characteristics in Northern Michigan with SIR-C/X-SAR

    IEEE Transactions on Geoscience and Remote Sensing

    (1995)
  • FaurotJ.L.

    Estimating merchantable volume and stem residue in four timber species: Ponderosa pine, lodgepole pine, western larch, Douglas-fir

  • GeschD. et al.

    The national elevation dataset

    Photogrammetric Engineering and Remote Sensing

    (2002)
  • GoldsteinB.

    The struggle over ecosystem management at Yellowstone

    BioScience

    (1992)
  • GreenR.M.

    Relationships between polarimetric SAR backscatter and forest canopy and sub-canopy biophysical properties

    International Journal of Remote Sensing

    (1998)
  • HackettT.

    The Yellowstone story

    Journal of Forestry

    (1989)
  • HenryM.C. et al.

    The sensitivity of SIR-C backscatter to fire-related forest spatial patterns

    GIScience & Remote Sensing

    (2005)
  • HoekmanD.H. et al.

    Land cover type and biomass classification using AirSAR data for evaluation of monitoring scenarios in the Colombian Amazon

    IEEE Transactions on Geoscience and Remote Sensing

    (2000)
  • Huang, S., Crabtree, R., Swanson, A., Halligan, K., & Harmsen, J., (submitted to publication). Error analysis and...
  • KaramM.A. et al.

    Electromagnetic scattering from a layer of finite length, randomly oriented, dielectric, circular cylinders over a rough interface with application to vegetation

    International Journal of Remote Sensing

    (1988)
  • LeckieD.G. et al.

    Forestry applications using imaging radar

  • Cited by (48)

    • Towards a non-intrusive method employing digital twin models for the assessment of complex large wood accumulations in fluvial environments

      2022, Journal of Hydrology
      Citation Excerpt :

      The need for developing innovative surveying methods and more refined assessment tools is continuously addressed throughout large wood (LW) research (Harmon et al., 1986, Huang et al., 2009, Vaughan et al., 2009, Ruiz-Villanueva et al., 2016b).

    • F<sup>3</sup>: Simulating spatiotemporal forest change from field inventory, remote sensing, growth modeling, and management actions

      2018, Forest Ecology and Management
      Citation Excerpt :

      These are the main reasons why we chose both Landsat and LiDAR data for TNF. When F3 is adopted for other applications, we suggest users choose LiDAR and/or long-wavelength radar data, if available, in addition to the widely-available optical data, because LiDAR and long-wavelength radar data are good at depicting forest structure characteristics (Huang et al., 2009). Following the same line of reasoning, for those species that do not have strong relationships with either biotic or abiotic factors, our F3 will not perform well.

    • Review of native vegetation condition assessment concepts, methods and future trends

      2017, Journal for Nature Conservation
      Citation Excerpt :

      Among all three categories of attributes, RS is more capable of measuring structure, specially canopy cover, density and height. However, it is less successful in recognizing fallen logs and woody debris (Huang et al., 2009). The two attributes that can be easily mapped at various scales (global, regional and individual) are vegetation cover and extent.

    • Site-based and remote sensing methods for monitoring indicators of vegetation condition: An Australian review

      2016, Ecological Indicators
      Citation Excerpt :

      Of the compositional, structural and functional attributes typically used to monitor vegetation condition, remote sensing has been most successful at measuring structure, particularly vegetation cover, height, and stem density. There has been little success in remotely sensed measures of tree hollows (Catena et al., 1990) or in detecting fallen logs and woody debris (Huang et al., 2009). Vegetation cover and density are the attributes most commonly derived in remote sensing studies of vegetation, and have been measured at scales ranging from global (Defries and Townshend, 1994; Hansen et al., 2000) through regional studies (Nagler et al., 2001) down to individual trees (Leckie et al., 2003).

    View all citing articles on Scopus
    View full text