Comparing three sampling techniques for estimating fine woody down dead biomass

Designing woody fuel sampling methods that quickly, accurately and efficiently assess biomass at relevant spatial scales requires extensive knowledge of each sampling method’s strengths, weaknesses and tradeoffs. In this study, we compared various modifications of three common sampling methods (planar intercept, fixed-area microplot and photoload) for estimating fine woody surface fuel components (1-, 10-, 100-h fuels) using artificial fuelbeds of known fuel loadings as reference. Two modifications of the sampling methods were used: (1) measuring diameters only and both diameters and lengths and (2) measuring diameters to (a) the nearest 1.0mm, (b) traditional size classes (1 h1⁄4 0–6mm, 10 h1⁄4 6–25mm, 100 h1⁄4 25–76mm), (c) 1-cm diameter classes and (d) 2-cm classes. We statistically compared differences in sampled biomass values to the reference loading and found that (1) fixed-area microplot techniques were slightly more accurate than the others, (2) the most accurate loading estimates were when fuel particle diameters were measured and not estimated to a diameter class, (3) measuring particle lengths did not improve estimation accuracy, (4) photoloadmethods performed poorly under high fuel loads and (5) accurate estimate of fuel biomass requires intensive sampling for both planar intercept and fixed-area microplot methods. Additional keywords: fixed-area plot, fuel inventory, fuel loading, line intersect, monitoring, photoload, planar intercept. Received 13 March 2013, accepted 22 May 2013, published online 23 August 2013 Introduction Successful wildland fuel management activities will ultimately depend on the accurate inventory andmonitoring of the biomass of forest and rangeland fuels (Conard et al. 2001; Reinhardt et al. 2008). Wildland fuel loadings are important direct inputs to fire effects models and are used to create fuel classifications that are inputs to predicting fire behaviour from common fire management software such as BEHAVE, FARSITE and FLAMMAP (Finney 2004, 2006; Andrews 2008).More recently, fuel loadings have become critical inputs for estimating fuel consumption (Reinhardt and Holsinger 2010), smoke emissions (Hardy et al. 2000), soil heating (Campbell et al. 1995), carbon stocks (Reinhardt and Holsinger 2010), wildlife habitat (Bate et al. 2004) and site productivity (Hagan and Grove 1999; Brais et al. 2005). Wildland fuel biomass is the one factor that can be directly manipulated to achieve management goals, such as restoring ecosystems, lowering fire intensity, minimising plant mortality and reducing erosion (Graham et al. 2004; Ingalsbee 2005; Reinhardt et al. 2008). As a result, comprehensive and accurate estimates of fuel loadings are needed in nearly every phase of fire management including fighting wildfires (Chen et al. 2006; Ohlson et al. 2006), implementing prescribed burns (Agee and Skinner 2005), describing fire danger (Deeming et al. 1977; Hessburg et al. 2007) and predicting fire effects (Ottmar et al. 1993; Reinhardt and Keane 1998). It is often difficult to estimate surface fuel biomass for many ecological, technological and logistical reasons (Keane et al. 2012). A fuelbed consists of many fuel components, such as litter, duff, twigs, logs and cones, and the properties of each component that influence biomass measurements, such as shape, size, particle density and orientation are highly variable even within a single fuel particle. Fuel component properties vary at different spatial scales (Kalabokidis and Omi 1992; Habeeb et al. 2005; Keane et al. 2012) and fuel loadings are so highly variable that they are often unrelated to vegetation characteristics, topographic variables or climate parameters (Brown and See 1981; Rollins et al. 2004; Cary et al. 2006). However, it is the uneven distributions of fuel across spatial scales that confound many fuel sampling and mapping activities. Therefore, picking the proper method for sampling fuel biomass is important, especially when there are disparate sampling techniques for the different surface fuel components (Lutes et al. 2006, 2009). Several sampling techniques have been developed to sample downed woody fuel biomass for fire management. The most commonly used woody surface fuel sampling method is the CSIRO PUBLISHING International Journal of Wildland Fire 2013, 22, 1093–1107 http://dx.doi.org/10.1071/WF13038 Journal compilation IAWF 2013 www.publish.csiro.au/journals/ijwf planar intercept (PI) (often called line intercept) where diameters of fuel particles that intercept a vertical sampling plane are measured and converted to biomass loading (Van Wagner 1968; Brown 1971, 1974). Fixed-areamicroplots (FMs) are often used to estimate biomass for research applications where diameters and lengths of woody fuel particles are measured within microplots to compute fuel volume that is then converted to loading using the wood particle density and microplot area (Sikkink and Keane 2008). A new technique, called photoloads (PHs), provides a set of photographs of increasing fuel loadings for the user to select a photo that best matches the observed fuelbed to estimate loading (Keane and Dickinson 2007a, 2007b). Each of these techniques has strengths and weakness when applied to fuel inventory and monitoring, so determining how well each sampling technique performs under a variety of fuel loadings is critical to designing efficient sampling projects. In this study, we explored how the three surface fuel sampling methods (PI, FM and PH) compared in their ability to assess downed dead fine woody debris biomass (FWD; woody particles less than 8 cm in diameter). We also modified two of these techniques (PI and FM) to improve accuracy and compared results from the modifications across and between techniques. This study is somewhat unique in that we conducted fuel sampling on a 500-m square plot established in a parking lot within which we placed FWD in known fuel loadings in various spatial distributions. Loadings from each technique were compared with the known reference loading to determine accuracy and precision. Hazard and Pickford (1986) performed a similar experiment, but they used simulation modelling instead of real fuels. Our goal was to determine the tradeoffs involved in using each of these sampling techniques to inform the design of sampling projects for research, resource and fire management applications. Background The line transect method was originally introduced by Warren and Olsen (1964) and made applicable to measuring coarse woody debris by VanWagner (1968). Its development is rooted in probability-proportional-to-size concepts and several variations have been developed since 1968, including those that vary the line arrangements and those that apply the technique using different technologies (DeVries 1974; Hansen 1985; NemecLinnell andDavis 2002). The PImethod is a variation of the linetransect method that was developed specifically for sampling FWD and coarse woody debris (CWD) in forests (Brown 1971, 1974; Brown et al. 1982). It has the same theoretical basis as the line transect, but it uses sampling planes instead of lines. The planes are somewhat adjustable because they can be any size, shape or orientation in space and samples can be taken anywhere within the limits set for the plane (Brown 1971). The PI method has been used extensively in many fuel inventory and monitoring programs because it is fairly fast and simple to use (Busing et al. 2000; Waddell 2002; Lutes et al. 2006). It has also been used in many research efforts because it is often considered an accurate technique for measuring downed woody fuels (Kalabokidis and Omi 1998; Dibble and Rees 2005). The problem is that it is difficult to integrate PI sampling designs with other fixed-area plot designs such as those used for estimating canopy fuels because the method was designed to sample entire stands, not fixed-area plots. In contrast to probability-based methods, FMs are based on frequency concepts and have been adapted from vegetation composition and structure studies to sample fuels (MuellerDombois and Ellenberg 1974). In fixed-area sampling, a round or rectangular plot is used to define a sampling frame and all fuels within the plot boundary are measured using diverse methods that range from destructive collection to volumetric measurements (i.e. length, width, diameter) to particle counts by size class (Keane et al. 2012). Because fixed-area plots require significant investments of time and money, they are more commonly used to answer specific research questions rather than to monitor or inventory fuels for management action. In recent years, a new method of assessing fuel loading has been developed to sample fuel beds using visual techniques. The PH method uses calibrated, downward-looking photographs of known fuel loads to compare with conditions on the forest floor and estimate fuel loadings for individual fuel categories (Keane and Dickinson 2007a, 2007b). These ocular estimates can then be adjusted for diameter, rot and fuel height. There are different PH methods for logs, fine woody debris, shrubs and herbaceous material, but there are no methods for measuring duff and litter fuels. Photoload methods are much faster and easier than fixed-area and planar intercept techniques with comparable accuracies (Sikkink and Keane 2008). The PH technique differs from the commonly used photo-series technique (Fischer 1981) in that assessments are made at smaller scales using downward pointing photographs of graduated fuel loadings. It is somewhat problematic to compare surface fuel sampling techniques because of the difficulty in estimating the actual fuel biomass for reference (Sikkink and Keane 2008). Successful comparison studies contrast sampled fuelbeds with the actual known biomass, but quantifying these reference or actual fuel loadings is costly, and often impossible, because it is difficult to collect, sort, dry and weigh the heavy amount of fuel biomass within a commonly used ecological sampling frame (250–1000-m plot) located in a natural setting. As a result, most fuel sampling comparisons relied on subsampling using FMs and destructive sampling (Sikkink andKeane 2008), but the standard errors involved in subsample estimation may overwhelm the subtle differences between sampling methods. There are also major differences between sampling techniques that make statistical comparisons difficult. The commonly used PI sampling, for example, is difficult to compare with other methods because the two dimensional (length and height) sampling plane makes it difficult to relate to area-based reference sampling frames because fuel loads must be destructively sampled in three dimensions (microplot; length, width and height). Loading estimates from visually based fuel sampling methods, such as PHs (Keane and Dickinson 2007a), have major sources of error because of differences between samplers and their ability to accurately estimate fuel loadings using visual cues; and size, shape and density differences between fuel components make comparisons difficult in that each component has its own inherent ecological scale (Keane et al. 2012). 1094 Int. J. Wildland Fire R. E. Keane and K. Gray