Forest Research: Open Access
Open Access

Comparing Canopy Metric Estimations Using Three Conifer Species in the Netherlands

Alan Duncan Hibler¹, Brian P. Oswald^1*, Nienke Brouwer², Ester Willemsen² and Hans M. Williams^{1 *}Correspondence: Brian P. Oswald, Arthur Temple College of Forestry and Agriculture, Stephen F. Austin State University, Nacogdoches Texas, USA, Tel: +9366457990, Email:

Author info »

Introduction

Wildfires within or adjacent to the Wildland-Urban Interface (WUI) are an important issue in a number of countries. For example, WUI fires have accounted for nine of the 25 largest fire loss incidents in the United States of America’s history, ranging from $0.5 - $2.4 billion in direct losses per fire (in 2008 constant dollars, NFPA 2009) and WUI fires continue to cause financial loss to resources and structures. Utilizing knowledge in fire ecology and fire effects research over decades by many authors [1-3] for the creation of fuel loads estimation models [4-7], photo guides for appraising surface fuels [8,9], and fire behavior prediction models including BEHAVE and BEHAVEPLUS [10-12], much about fire behavior and the key variables required to accurately model it has been learned. However, both wildland fire and its potential interaction with WUI communities’ remains little studied and present potential for significant loss of human lives and property in many parts of the world today.

In the last decade The Netherlands has begun to examine the behavior of wildland fires across its substantial wildland urban interface (WUI). Wildland fires in 2009 (Schoorl), 2010 (Bergen, Strabrechtse Heide), and 2014 (De Hoge Veluwe National Park) have shown the threat is real, with fires leading in 2014 to evacuations of about 500 people. In many cases the threat comes from canopy (crown) fires, which are inherently more difficult to estimate potential behavior because of the challenges of quantifying canopy fuels [13].

Modelling potential canopy fire behavior requires estimates of canopy fuel loads. Four canopy parameters are required: base height, height, cover, and canopy bulk density (CBD), which is the quantity of canopy fuel/canopy volume and represents how compacted canopy fuels are [13], which are then used in many fire spread models [14]. Various direct and indirect methods have been developed to obtain CBD. Either gap fraction and leaf area index (LAI) can be used to calculate and extrapolate canopy fuel data using a correlation between gap fraction and canopy density. LAI (a unit less measurement of single-sided leaf area per unit ground surface area) is a relatively accurate quantitative measure of canopy foliage [15,16]. Both gap fraction and LAI can be estimated by several methods, including using a spherical densiometer (for gap fraction, but not LAI), LI-COR LAI 2200c plant canopy analyzer (LI-COR Environmental, Lincoln, Nebraska, USA), or a hemispherical canopy photograph (HCP).

When measuring the gap fraction using LI-COR systems, five angles of view are computed by dividing the below-canopy by the above-canopy readings. The LI-COR’s light sensor includes a filter to limit the spectrum of received radiation to <490 nm, minimizing the effect of light scattered by foliage. Directly illuminated foliage will scatter more light in the canopy than will be calculated, reducing LAI values up to 50% [17,18].

HCP utilizes a fish-eye lens under the canopy that can obtain a measure of canopy structure within a 180-degree projection. Photos are interpreted by classifying pixels into sky or canopy, and then converted into indices using an inversion model based on Beer’s Law, using the observed gap fraction distribution throughout the photo [19,20]. Gap fractions are computed from HCP by determining the fraction of exposed sky using software such as HemiView (Delta-T Devices, Cambridge, United Kingdom) canopy analysis software to obtain canopy structure information such as LAI, Gap Fraction and canopy opening distributions.

While these three different non-destructive methods are available to estimate canopy bulk densities (CBD), a comparison of the effectiveness of these methods is lacking. Considering the need to obtain estimates of canopy bulk densities (CBD) of three common tree species in the Netherlands that are subjected to recurring wildfire events, our objective was to compare these indirect, nondestructive methods with the hypothesis that HCP would provide the best correlation with CBD simulated by commonly used software. If successful, our results could provide a recommendation of the most time effective, accurate, non-destructive method of estimating CBD that could easily be converted into data that is enterable into fire prediction software, providing land managers tools to more accurately estimate fire behavior and reduce the potential risk to people and property.

Statistical Analysis

Due to the limited extent of individual forest cover types in the Netherlands, location and species were assumed confounded, and any inferences relating to species thus cannot be distinguished from potential edaphic or climatic variation. Since species were not replicated at each site, we analyzed crown bulk density, gap fraction, and LAI with one-way ANOVAs to determine potential differences in species/location; separate one-way ANOVAs by the three methods were conducted to examine instrument differences. Statistics were performed in SAS 9.2 (SAS Institute, Cary, North Carolina, USA) using PROC GLM or PROC REG with an alpha of 0.05. All data met assumptions of normality and heteroscedasticity, so transformations were unnecessary. Tukey’s test was conducted for post-hoc comparisons. Simple linear regressions were created to 1) characterize differences between instruments and, 2) examine the relationship between canopy bulk density and other canopy metrics. Since the purpose of regression analyses was to examine correlations rather than create predictive models, Pearson’s correlation coefficient is presented and trendlines are not shown.

Results

The instruments did vary (Table 1) in their estimation of both gap fraction and LAI (p<0.0001). For gap fraction, the densiometer and LI-COR both differed from HemiView, which estimated gap fractions 168–189% greater than the other methods, respectively. However, while the gap fraction means were similar between the LI-COR and densiometer, when regressed between the two instruments, no significant correlation was observed. The same was true for gap fraction estimations between the LI-COR and HemiView (Figure 2). Despite their dissimilar means, there was a strong correlation for the HemiView and densiometer estimates of gap fraction. LAI estimated with the LI-COR was more than two-times higher than that estimated by the HemiView (Table 1). There was a moderate-to-weakly significant correlation between LAI values estimated with both instruments (Figure 2).

Figure 2: Estimated Canopy Bulk Density with R2s derived from plots using (A) HemiView Gap Fraction, (B) LI-COR Gap Fraction, (C) Densiometer Gap Fraction, (D) HemiView LAI and (E) LI-COR LAI. Douglas-firs are represented by triangles, Scots pines by circles, and black pines by squares.

Table 1: Mean Gap Fractions (%), LAI (m² m^-2), and CBD (kg m^-3) recorded by a densiometer, and Hemiview and a LI-COR in the Netherlands.

There were no significant correlations between gap fraction and CBD regardless of instrument used. The only variable to demonstrate a significant but moderate strength correlation with CBD was LAI estimated with HemiView. The stands did vary substantially in LAI estimated by HemiView, with the minimum LAI stand having only 25% of the magnitude of the maximum LAI stand (Figure 3), apparent in the HCPs with the lowest LAI values in stands with large central canopy gaps for all species.

Figure 3: Gap Fraction comparisons with R2s. (A) Densiometer vs. HemiView Gap Fractions; (B) Densiometer vs. LI-COR Gap Fractions; (C) HemiView vs. LI-COR Gap Fractions; (D) HemiView vs. LI-COR LAIs. Douglas-firs are represented by triangles, Scots pines by circles, and black pines by squares.

CBD and LAI did not vary by species (p=0.6341; p=0.1664), although gap fraction (p=0.0322) for Scots pine was greater than Douglas-fir (Table 2).

Table 2: Mean Gap Fractions (%), LAI (m² m^-2), and CBD (kg m^-3) recorded for Black Pine, Scots Pine and Douglas-fir.

Discussion

That none of the species varied in LAI or CBD was expected given that all were between late stem exclusion and early understory reinitiation stages of stand development. None of these methods appear ideal for estimating CBD in these ecosystems; if accurate estimates of CBD are desired, additional measurements need to be made and/or destructive harvesting will be needed.

Our LAI values were within reported ranges for these species. Black pine LAI in Spain was most similar to the HemiView estimate, but Scots pine LAI in Belgium was more similar to that from the LI-COR [24,25]. North American LAI for Douglas-fir was most similar to our LI-COR LAI. Both Scots pine and black pine have been observed in Europe with LAI ranging up to 2.7 to 3.0 [26-28].

Differences in gap fraction between the densiometer and HemiView could be from the portion of the canopy observed with each. While both take their data from a circular projection, the area used to calculate canopy density on the spherical densiometer mirror comes from a grid 49% of the total surface area, not the full area. This does result in a smaller overall canopy section being analyzed and potential errors have been noted [29,30]. HemiView software utilizes the entire area of the photograph.

Conclusion

Differences in gap fraction and LAI between the LI-COR and HemiView could be caused by the sampling protocol. While the HCP and densiometer estimates were taken from a single point at plot center, the LI-COR data were collected from 30 different sample locations within each stand. However, it should be noted that the LI-COR and densiometer were in close agreement regarding gap fraction estimates. The positive relationship between denser canopies and lower gap fractions found in this study will not only impact canopy fire behavior, but also on understory vegetation growth and densities, and therefore on fuel availability on the forest surface. In the interface between urban development and nature, the reduction in canopy density may therefore result in an increase in fire hazard driven by an increase in surface fuels.

Acknowledgement

We are grateful for the support provided by the Arthur Temple College of Forestry and Agriculture at Stephen F. Austin State University and the Institute Fysieke Veiligheid for financial and logistical support for this project. Sincere thanks go to Dr. Bob Keane of the Rocky Mountain Research Station for the lending of equipment and consultation on model programming, and to Dr. Jeremy Stovall at the Arthur Temple College of Forestry and Agriculture for his assistance throughout the project. Scarlet van Os and Sean Hoes assisted during the Netherlands field work and Haley Rabbe entered the data.

Disclosure Statement

No potential conflict of interest was reported by the authors.

References

1. Pyne S, Andrews P, Laven R. Introduction to wildland fire. 2nd edition John Wiley & Sons, Inc., New York, New York, USA. 1996.
2. Sugihara N, Van Wagtendonk J, Shaffer K, Fites-Kaufman J, Thode A. Fire in California's ecosystems. University of California Press, Berkley, California, USA. 2006.
3. Wright H, Bailey A. Fire ecology: United States and Southern Canada. John Wiley & Sons, Inc., New York, New York, USA. 1982.
4. Albini F. Estimating wildfire behavior and effects. US department of agriculture, forest service, intermountain forest and range experiment station, general technical report INT-30, Ogden, Utah. 1976.
5. Anderson H. Aids to determining fuel models for estimating fire behavior. US department of agriculture, forest service, intermountain forest and range experiment station, general technical report INT-122, Ogden, Utah. 1982.
6. Brown J. Handbook for inventorying downed woody material. US department of agriculture, forest service, Intermountain forest and range experiment station, general technical report INT-16, Ogden, Utah. 1974.
7. Brown J, Oberheu R, Johnston C. Handbook for inventorying surface fuels and biomass in the interior west. US department of agriculture, forest service, intermountain forest and range experiment station, general technical report INT-129, Ogden, Utah. 1982.
8. Ottmar R, Vihnanek R, Wright C. Stereo photo series for quantifying natural fuels: Volume VI: longleaf, pocosin and marshgrass types in the southeastern united states. National wildfire coordinating group, Boise, Idaho. 2000.
9. Reeves H. Photo guide for appraising surface fuels in east Texas, center for applied studies at school of forestry. 1988:43-51.
10. Andrews P, Bevins C, Seli R. BehavePlus fire modeling system: Version 3.0. US department of agriculture, forest service, Intermountain forest and range experiment station, general technical report RMRS-GTR-106WWW revised, Ogden, Utah. 2005.
11. Bradshaw L, Deeming J, Burgan R, Cohen J. The 1978 national fire-danger rating system: technical documentation US department of agriculture, forest service, intermountain forest and range experiment station, general technical report INT-169, Ogden, Utah. 1983.
12. Burgan R, Rothermel R. BEHAVE: Fire behavior prediction and fuel modeling system - FUEL subsystem. US department of agriculture, forest service, intermountain forest and range experiment station, Ogden, Utah. 1984.
13. Keane R, Reinhardt E, Scott J, Gray K, Reardon J. Estimating forest canopy bulk density using six indirect methods. Can J For Res. 2007;35:724-739.
14. Finney M. FARSITE: Fire Area Simulator—model development and evaluation. US dept agric. For Serv Intermit for and Range Exp.Stn, Res.Pap. RMRS-RP-4 revised, Ogden, Utah. 1998.
15. Jonckheere I, Fleck S, Nackaerts K, Muys B, Coppin P, Weiss M, et al. Review of methods for in situ leaf area index determination: Part I. Theories, sensors and hemispherical photography. For Meteorol. 2004;121:19-35.
16. Weiss M, Baret F, Smith G, Jonckheere I, Coppin P. Review of methods for in situ leaf area index (LAI) determination: Part II. Estimation of LAI, errors and sampling. For Meteorol. 2004.
17. Welles J, Cohen S. Canopy structure measurement by gap fraction analysis using commercial instrumentation. J Exp Bot. 1996.
18. Welles J, Norman J. Instrument for indirect measurement of canopy architecture. Agron J. 1991;83:818-825.
19. Hale S, Edwards C. Comparison of film and digital hemispherical photography across a wide range of canopy densities. For Meteorol. 2002;112: 51-56.
20. Rich P. Characterizing plant canopies with hemispherical photographs. Rem Sens Rev. 1990;5:13-29.
21. Lutes D, Benson N, Keifer M, Caratti J, Streetman S. FFI: A software tool for ecological monitoring. Int J Wild Fire. 2009;18:310-314.
22. Reinhardt E, Lutes D, Scott J. FuelCalc: A method for estimating fuel characteristics. In: Andrews PL, Butler BW (eds) Fuels management - How to measure success: Conference proceedings, Portland, Oregon,. U.S. Dept.Agric., For.Serv. Rocky Mtn Res. Stn. Proc. RMRS-P-41, Fort Collins, Colorado. 2006;pp:273-282.
23. Nobis M, Hunziker U. Automatic thresholding for hemispherical canopy-photographs based on edge detection. For Meteorol. 2005;128:243-250.
24. Cerrillo R, Beira J, Suarez J, Xenakis G, Salguero R, Clemente R. Growth decline assessment in Pinus sylvestris L. and Pinus nigra Arnold. Forest by using 3-PG model. For Syst. 2016;25:3.
25. Xiao C, Janssens I, Curiel-Yuste J, Ceulemans R. Variation of specific leaf area and upscaling to leaf area index in mature Scots pine. Trees. 2006;20:304.
26. Balster N, Marshall J. Eight-year responses of light interception, effective leaf area index and stemwood production in fertilized stands of interior Douglas-fir (Pseudotsuga menziesii var. glauca). Can J For Res. 2000;30(5):733-743.
27. Bealde C, Talbot H, Jarvis P. Canopy structure and leaf area index in a mature Scots pine forest. Int J For Res. 1982;55:105-123.
28. Forrester D, Bonal D, Dawud S, Gessler A, Granier A, Pollastrini M, et al. Drought responses by individual tree species are not often correlated with tree species diversity in European forests. J Appl Eco. 2016;53:1725-1734.
29. Englund S, O'Brien J, Clark D. Evaluation of digital and film hemispherical photography and spherical densiometry for measuring forest light environments. Can J For Res. 2000;30:1999-2005.
30. Vales D, Bunnell F. Comparison of methods for estimating forest overstory cover. I. Observer effects.  Can J For Res. 1988;18(5):606-609.