Forest Ecology and Management 482 (2021) 118860 Available online 22 December 2020 0378-1127/© 2020 Elsevier B.V. All rights reserved. Shifting tree species composition of upland oak forests alters leaf litter structure, moisture, and flammability Jennifer K. McDaniel a , * , Heather D. Alexander b , Courtney M. Siegert c , Marcus A. Lashley d a Warnell School of Forestry and Natural Resources, University of Georgia, Athens, GA 30605, United States b School of Forestry and Wildlife Sciences, Auburn University, Auburn, AL 36849-5418, United States c Department of Forestry, Forest and Wildlife Research Center, Mississippi State University, MS 39762, United States d Department of Wildlife Ecology and Conservation, University of Florida, Gainesville, FL 23611, United States A R T I C L E I N F O Keywords: Prescribed fire Mesophication Leaf traits Fuel moisture Fire behavior A B S T R A C T In historically open-canopied and fire-dependent upland oak ( Quercus spp.) forests of the central and eastern United States, fire exclusion is contributing to an increase in competing non-oak tree species that are often shade- tolerant and fire-sensitive. As these non-oak species encroach and oak abundance declines, forests are becoming denser and will likely become cooler, moister, and less flammable through a hypothesized feedback loop termed mesophication. To better understand how this gradual shift in forest composition could affect flammability of leaf litter, the primary fuel in these systems, we measured leaf litter traits and moisture dynamics of two pyrophytic oaks and three non-oaks, and implemented experimental burns on plots (~3-m 2 ) with fuel beds comprised of single- and mixed-species leaf litter. Non-oaks produced thinner, smaller leaves with a greater specific leaf area compared to oaks, and had a higher initial fuel moisture content, traits associated with low flammability. Our experimental burns confirmed that non-oak leaf litter flammability was lower than that of oaks and that flammability decreased linearly with increasing non-oak leaf litter contribution to the fuel bed. These species-driven changes in fuelbed flammability may provide a mechanism whereby encroaching non-oak tree species create self-promoting conditions that are less favorable for regeneration of fire-dependent upland oaks. Thus, without other management interventions, our ability to reintroduce fire into these systems as a manage- ment tool to prevent further compositional shifts and improve oak regeneration will likely decline as non-oak species ’ contribution to the fine fuel bed increases. 1. Introduction Historically, fire-maintained upland oak ( Quercus spp.) woodlands and forests were prominent across the central and eastern United States since a period of warming and drying began at the end of the last glaciation (Delcourt and Delcourt, 1997). Recurring fires prevented encroachment from competing species and encouraged growth of in- termediate shade-tolerant oaks by maintaining open canopies (Abrams, 1992; Stambaugh et al., 2015). Due in part to fire exclusion beginning in the early 20th century, however, forests across the region began to experience shifting species composition and structure (McEwan et al., 2011). Open-canopy woodlands transitioned to closed-canopy forests with limited recruitment of oaks yet successful recruitment of shade- tolerant and fire-sensitive species including red maple ( Acer rubrum L.) and American beech ( Fagus grandifolia Ehrh ) and/or generalists, hereafter referred to as non-oaks, including tulip-poplar ( Liriodendron tulipifera L.), sweetgum ( Liquidambar styraciflua L.), and others (Han- berry et al., 2020; Abrams, 1992; Hutchinson et al., 2008). This transi- tion in species composition from pyrophytic oaks to non-oak species is likely a component of a process termed mesophication, whereby non- oaks species promote cooler, moister, less flammable conditions that are conducive to their own growth and persistence at the expense of more fire-tolerant species (Nowacki and Abrams, 2008). Although shifting forest dynamics are likely a result of multiple interacting factors including fire exclusion, land use changes, altered herbivore pop- ulations, and climate change (McEwan et al., 2011), finer-scale, stand- level drivers likely include changes in fuelbed moisture dynamics and flammability (i.e., the ability of a fuel to ignite and burn) that result from differences in leaf litter traits among species with differing fire toler- ances (Alexander and Arthur, 2010; Babl et al., 2020; Kreye et al., 2018; * Corresponding author. E-mail address: jkmcdan@uga.edu (J.K. McDaniel). Contents lists available at ScienceDirect Forest Ecology and Management journal homepage: www.elsevier.com/locate/foreco https://doi.org/10.1016/j.foreco.2020.118860 Received 5 October 2020; Received in revised form 7 December 2020; Accepted 8 December 2020 Forest Ecology and Management 482 (2021) 118860 2 Kreye et al., 2013; Nowacki and Abrams, 2008, Varner et al., 2016). Understanding species-level differences in leaf litter traits associated with flammability is crucial in upland oak forests because leaf litter is the primary carrier of fire in these systems (Arthur et al., 2017; Brewer and Rogers, 2006). Several leaf litter traits impact flammability, including the structural and chemical properties of leaves, moisture content, and packing or physical arrangement of the fuelbed (Varner et al., 2015). Morphological traits of individual leaves, which combine to form fuelbeds, influence flammability both directly and indirectly. For example, larger, curlier leaves burn with greater flame heights and more consumption than smaller, flatter leaves (Engber and Varner, 2012), and time to ignition decreases with increasing specific leaf area (Grootemaat et al., 2015). Indirectly, leaf traits influence fuel moisture adsorption and drying. Thin, small, flat leaves are generally associated with more tightly packed fuelbeds (Cornwell et al., 2015; Schwilk and Caprio, 2011) that gain more moisture and lose moisture more slowly than aerated, loosely packed fuelbeds made up of large, thick, curly leaves (Kreye et al., 2013). Moisture content of leaf litter has a large impact on flammability and fire behavior due to lowered probability of ignition and rate of fire spread from increased moisture content (Roth- ermel, 1972). As bulk density increases, fuel moisture also increases, and these moist and compacted fuelbeds have a lower probability of ignition (Plucinski and Anderson, 2008). Due to influences on fuel structure and porosity, characteristics of individual leaves including thickness, surface area to volume ratio, and density impact fuelbed moisture adsorption and retention (Kreye et al., 2013). Differences in leaf traits among pyrophytic oaks and less fire-tolerant non-oaks may lead to changes in both moisture content and flammability and partially explain shifting fire regimes co-occurring with changes in species composition and structure of upland oak forests. Although prescribed fire is a commonly used tool to manage upland oak forests, studies throughout the literature report a variety of impacts on oak regeneration and recruitment, both positive and negative (Brose et al., 2013), perhaps due to the changes in fuel moisture and flammability created by increases in non-oak species. Here, we sought to determine the impacts of increasing non-oak species ’ dominance in an upland oak forest in north Mississippi, USA, on flammability by measuring leaf litter traits and examining moisture dynamics then implementing experimental burns to confirm differences in flammability among species. Specific objectives were to 1) examine how varying leaf litter traits known to influence flammability differ among non-oaks and upland oaks found in north Mississippi, 2) assess how non-oaks interact with upland oaks to influence fuel moisture dy- namics in species mixtures of increasing non-oak leaf litter contribution as well as single species fuelbeds, and 3) evaluate differences in burning characteristics during dormant season plot-level experimental burns of mixed-species and single-species leaf litter from two upland oaks and three non-oaks. We tested hypotheses that 1) when compared to non- oaks, upland oaks will have morphological traits associated with fire adaptations including thicker, larger, and curlier leaves, 2) increasing non-oak leaf litter contribution to fuelbeds will lead to higher initial moisture content and slower moisture loss rate than upland oaks, 3) among single-species fuelbeds, leaf litter of non-oak species will have dampened flammability compared to that of oak species, and 4) among mixed-species fuelbeds, flammability will decrease with increasing contribution of non-oak leaf litter. Examining morphological traits and moisture dynamics known to be associated with flammability and con- firming differences in flammability of upland oak and non-oak leaf litter in field-based trials will clarify the role of these species in flammability that could hinder upland oak regeneration and recruitment in the eastern and central United States. 2. Methods 2.1. Study area Leaf litter was collected from upland hardwood forests at Spirit Hill Farm (SHF; 34 ◦ 41 ′ N, 89 ◦ 42 ′ W), located ~ 30 km west of Holly Springs, Mississippi, USA. The average annual temperature is 15.7 ◦ C and ranges from 4.0 ◦ C in January to 26.3 ◦ C in July, and the average annual pre- cipitation is 1346 mm y -1 (Arguez et al., 2010). Soils are Ultisols of the Providence-Ruston, Providence silt loam, and Ruston-Providence com- plexes and are described as moderately well-drained silty loams and sandy loams (Soil Survey Staff, 2019). Stands consisted of an oak-dominated overstory ( ≥ 20 cm diameter at breast height (dbh)) with a total basal area of 21.3 m 2 ha 1 and density of 210 trees ha 1 . Oaks, including southern red oak ( Quercus falcata Michx.; 30%) and post oak ( Q stellata Wangenh.; 25%), comprised most of the overstory density with minor contributions of sweetgum (9%), winged elm ( Ulmus alata Michx.; 9%), and hickory ( Carya spp.; 14%). The midstory (10 – 20 cm dbh) density was 213 trees ha 1 and was predominately sweetgum (23%), winged elm (20%), and hickory (19%), with minor contributions of oaks (7%). Among oaks, southern red oak and post oak were the most abundant seedling species (1072 seedlings ha 1 and 2057 seedlings ha 1 , respectively). The most abundant of all species in the seedling pool, however, were sweetgum (4674 seedlings ha 1 ) and winged elm (4005 seedlings ha 1 ). Southern red oak and post oak, in addition to black oak ( Q. velutina Lam.) and blackjack oak ( Q. marilandica Munchh.), historically dominated upland oak forests in the surrounding area, while species including sweetgum, winged elm, red maple, and blackgum ( Nyssa sylvatica Marsh.) were historically ab- sent or restricted to floodplain sites (Brewer, 2001; Surrette et al., 2008). The historic fire return interval was approximately 4 – 6 years (Frost, 1998), but there has been no timber harvesting or fire in stands for at least 50 years (B. Bowen, Landowner of SHF, personal communication). 2.2. Litter collection Leaf litter used for leaf measurements, moisture dynamics experi- ments, and experimental burns was collected using a combination of suspended collection nets and hand collection during October 2017- February 2018. We selected the two most abundant oaks (southern red oak and post oak) and three most abundant non-oak species (winged elm, sweetgum, and hickory) and collected litter of those species that exhibited no signs of decomposition. Litter was air-dried and stored in cardboard boxes until processing. Hickory leaflets, hereafter referred to as leaves, were collected from all species within the genus found at SHF (pignut hickory ( C. glabra Mill.), mockernut hickory ( C tomentosa (Poir.) Nutt.), bitternut hickory ( C. cordiformis (Wangenh.) K. Koch), and shagbark hickory ( C. ovata (Mill.) K. Koch) because of difficulty sepa- rating fallen litter to species and similarities in leaf characteristics. We collected hickory leaflets rather than intact leaves with a rachis attached because the majority of fallen leaf litter we encountered was only leaflets. 2.3. Leaf litter traits We measured individual leaf litter traits that influence flammability and moisture dynamics on a subsample (n = 50) of collected leaves of southern red oak, post oak, winged elm, sweetgum, and hickory. Leaf curling was measured as the maximum height of the leaf when laid horizontally on a flat surface without flattening and under laboratory, air-dried conditions. Leaf perimeter and one-sided surface area were determined using a flatbed scanner and threshold-based pixel count measurement in ImageJ software (Abr ` amoff et al., 2004). Leaves were oven dried at 60 ◦ C for 48 h and weighed to determine the oven-dry mass. Specific leaf area (SLA) was calculated as the one-sided surface area of the leaf divided by its oven-dry mass. Leaf volume was calculated as the one-sided surface area multiplied by thickness, and surface area to volume ratio (SA:V) was calculated as the two-sided surface area of the leaf divided by volume (Engber and Varner, 2012). The leaf dissection index (LDI), which describes the degree of dissection or lobing of a leaf, was calculated as the ratio of perimeter to the square root of area J.K. McDaniel et al. Forest Ecology and Management 482 (2021) 118860 3 (McLellan and Endler, 1998). Higher values represent more incised leaves while leaves with lower LDI values have less serrations or lobes. After bisecting the leaf halfway between the leaf base and apex and perpendicular to the midvein, we used calipers to measure the leaf thickness near the margin and the midvein and averaged the two values. All statistical analyses were conducted in R v. 3.5.0 (R Core Team, 2018). Leaf traits were compared using a Kruskal-Wallis non-parametric test followed by a Dunn test because data were not normally distributed. Because we expected leaf litter traits to be highly correlated, we com- bined traits into uncorrelated principal components using principal component analysis (PCA) and also performed a permutational multi- variate analysis of variance with 999 permutations (PERMANOVA; Anderson, 2001) using the adonis function in the vegan package to understand differences in the centroid locations of each species (Oksa- nen et al., 2019). Additionally, we analyzed the multivariate homoge- neity of dispersion of each species using the betadisp function in the vegan package to determine if there were significant differences ( ∝ = 0.05) in the multivariate spread of individual leaves around the centroid location of each species and a Tukey-Kramer HSD test to test for pairwise differences in dispersion among species. 2.4. Leaf litter moisture dynamics To examine moisture dynamics of leaf litter, we constructed single- species and mixed-species litterbeds from collected litter, soaked the litterbeds, and weighed them throughout drying until a constant mass was reached using methods described in Kreye et al. (2018; 2013). For single-species litter, we constructed 15 g litterbeds (n = 4) from leaf litter of southern red oak, post oak, winged elm, sweetgum, and hickory. For mixed-species litter, we constructed 15 g litterbeds (n = 5) of four treatments of mixed oak litter with increasing contribution from non- oak leaf litter (0%, 33%, 66%, and 100% non-oak litter; Table 1). Within mixed-species treatments, individual species proportions were based on the combined midstory and overstory density of each species, with the 33% non-oak mixture representing the current stand condition. Litterbeds were oven dried at 60 ◦ C for 48 h to determine the oven-dry weight and then allowed to equilibrate under laboratory conditions. Next, litterbeds were soaked in water for 24 h, removed, and placed in elevated metal pans with drainage holes. We weighed litter after soaking to determine initial moisture content, then litter was allowed to dry under laboratory conditions (21.1 ± 0.4 ◦ C, 46.2 ± 0.4% relative hu- midity) for approximately 48 h or until equilibrium moisture content was reached (Kreye et al., 2013). Assuming a negative exponential desorption response (Kreye et al., 2013), litterbeds were weighed more often during early stages of drying than later stages (i.e., every 30 min for the first 4 h, hourly for 5 – 12 h, every 4 h during 13 – 16 h, etc.). We calculated the fuel moisture content for each fuelbed at each drying time using the following equation: (Equation 1) m t = (mass t - mass od )/(mass od ), where m t is the moisture content at time t, mass t is the litter mass at time t, and mass od is the oven-dry mass of the litter. Fuel moisture content was then converted to relative moisture content (Fosberg, 1970) using the following equation: E = (m t -m e ) / (m i -m e ), where E is relative moisture content, m t is moisture content at time t, m e is equilibrium moisture content, and m i is initial moisture content. Litter drying typically exhibits multiple distinct time lags with the most moisture loss occurring in the first time lag that corresponds to 63% of moisture loss of the most easily evaporable moisture (Nelson and Hiers, 2008), so we used piecewise polynomial curve fitting using the function segmented in the package segmented to separate the drying curve into two time lags and determine response time, which was calculated as the negative inverse of the slope of the first time lag (Kreye et al., 2013; Viney and Catchpole, 1991). We compared single-species initial moisture contents and response times among species with an ANOVA followed by a post-hoc Tukey- Kramer HSD when significant differences were detected ( ∝ = 0.05). To understand impacts of increasing non-oak litter proportion in leaf litter mixtures, we regressed percent non-oak leaf litter with initial moisture content and with response time. 2.5. Experimental burns To determine the effects of increasing non-oak leaf litter contribution on flammability, we implemented two series of plot-level experimental burns at SHF. In 2018, we conducted mixed-species burns of four treatments with five replicates of different leaf litter combinations with increasing contributions from non-oak leaf litter (0%, 33%, 66%, and 100% non-oak litter; Table 1). Prior to plot establishment, the per- square meter annual litterfall was estimated using 15 approximately 2 × 2-m leaf litter nets installed approximately 1.5 m above the forest floor in October 2017. Leaf litter in each net was collected monthly during leaf fall, weighed, and returned to the lab. A subsample was collected in the field from each net and weighed before and after drying at 60 ◦ C for 48 h to obtain an air-dry to oven-dry conversion factor. We then calculated the estimated litterfall (234 g m 2 oven-dry mass) as the converted oven-dry mass of the litter in each net divided by the area of each net. Plots were installed in a representative upland hardwood stand at SHF in mid-February 2018 and located in areas with little to no un- derstory vegetation to reduce effects of understory fuels on flamma- bility; any existing vegetation was clipped and removed. The existing O i horizon was removed using leaf blowers, and 937 g of collected litter was added to each 1.75 × 1.75-m plot (Fig. 1A). Leaf litter treatments were randomly assigned to plots to prevent bias due to any slope or microsite variation. Plots were stabilized using 35.5 cm tall wire mesh around edges to minimize loss of litter due to wind and prevent any additional litter inputs. Additionally, we installed a 0.5-m firebreak around each burn plot using a leaf blower. Mixed species plots were ignited on 9 March 2018 between 1300 and 1530 under stable weather conditions (wind speed 1.8 ± 0.1 m s 1 ; air temperature 20.0 ± 0.2 ◦ C; relative humidity 23.8 ± 0.6%). Prior to ignition, we removed a grab sample of leaf litter to determine fuel moisture content using Equation 1. We measured litter depth in the center of each quadrant of each plot, and depths were converted to bulk density by dividing the total mass of leaf litter added to the plot by the volume the litter occupied (depth multiplied by plot area). Air temper- ature, relative humidity, and wind speed were recorded using a pocket weather meter (Kestrel ® 5500 Fire Weather Pro, Kestrel Meters, Boot- hwyn, PA) prior to ignition of each plot. Each plot was ignited individually with a head fire using a drip torch (3:1 diesel:gasoline) along a single edge. One plot with 33% non-oak leaf litter was ignited first and with a backing fire that did not spread; this plot was excluded from further analyses because other plots were suc- cessfully ignited with a head fire. We measured several flammability metrics including percent area burned, temperature, flame height, rate of spread, and flaming duration. We visually estimated percent area burned as a proxy for percent consumption, as we anticipated utilizing burned plots for future studies and avoided disturbing remaining litter post-burn. The pyrometer-indicated fire temperature, hereafter referred to as temperature, was measured using pyrometers that were installed Table 1 Proportion of leaf litter by species used in mixed-species moisture dynamics and flammability experiments. Leaf litter was collected at Spirit Hill Farm, MS, USA. 1 Hickory includes pignut hickory, mockernut hickory, bitternut hickory, and shagbark hickory. Non-oak litter (%) Oaks Non-oaks Southern red oak Post oak Winged elm Sweetgum Hickory 1 0 0.50 0.50 0.00 0.00 0.00 33 0.33 0.33 0.11 0.11 0.11 66 0.17 0.17 0.22 0.22 0.22 100 0.00 0.00 0.33 0.33 0.33 J.K. McDaniel et al. Forest Ecology and Management 482 (2021) 118860 4 pre-burn and removed immediately following all burns. Four pyrome- ters were installed in each plot in four quadrants and were attached to pin flags directly on top of the litter layer. Pyrometers were constructed using aluminum tags painted with six Tempilaq ® fire-sensitive paints ranging from 79 ◦ C to 510 ◦ C (Tempil, South Plainfield, NJ) and covered with aluminum foil (Loucks et al., 2008). Maximum temperature was recorded for each pyrometer as the highest temperature indicated, and ambient temperature (20 ◦ C) was used if the pyrometer showed no melted paint. The mean maximum temperature for each plot was calculated as the average of the four pyrometer temperatures. Flame height to the nearest 10 cm was visually estimated every 30 sec using a marked pole adjacent to the plot. Rate of spread was determined by timing the fire spread from the ignition edge to the center of the plot and opposite edge of the plot and dividing each distance by the measured time. The flaming duration was measured as the total amount of time from the initial ignition to extinction of a visible flame. In 2019, we implemented single-species burns of leaf litter of each of the two oaks and three non-oak species using the same experimental design used for mixed-species but with different sample sizes among species due to varying amounts of litter collected (post oak: 5; southern red oak: 5; hickory: 3; sweetgum: 4; winged elm: 3). Plots were ignited on 22 March 2019 between 1530 and 1630 (wind speed 1.0 ± 0.2 m s 1 ; air temperature 22.5 ± 0.4 ◦ C; relative humidity 22.1 ± 1.0%) and on 27 March 2019 between 1100 and 1400 (wind speed 0.7 ± 0.0 m s 1 ; air temperature 20.1 ± 0.5 ◦ C; relative humidity 23.5 ± 1.0%). For mixed species-plots, we first used linear regression to analyze the relationship between each flammability metric and percent non-oak leaf litter. For both mixed-species and single-species plots, we expected flammability metrics to be highly correlated (Engber and Varner, 2012; Varner et al., 2015) and confirmed this using correlation analysis. To understand patterns in overall flammability and to avoid issues with multicollinearity, PCA was performed using centered, scaled, and log- transformed values for temperature, flame height, percent area burned, and rate of spread. Data used for PCA were log-transformed to stabilize variance of each variable. After performing PCA, principal components with eigenvalues greater than one were retained, and each burn replicate ’ s Principal Component 1 score was used as a measure of overall flammability for mixed-species burns. We then used linear regression to evaluate the impact of percent non-oak leaf litter on the flammability-related Principal Component 1 score. For single-species burns, we used a PERMANOVA to understand multivariate differences among species. 3. Results 3.1. Leaf litter traits Leaf litter traits varied among the five species, with oaks generally having thicker, larger leaves than non-oaks. Post oak and southern red oak leaves were the thickest and had the lowest SLA compared to other species (P < 0.0001; Table 2). The leaf surface area of post oak was ~ 5.3 times the surface area of winged elm, which had the smallest leaves (P < 0.0001). There was less distinction among species with regards to curl, but winged elm had significantly flatter leaves than all other species (P < 0.0001). Combining measured leaf morphological characteristics using a PCA (Fig. 2) resulted in three principal components each with eigenvalues > 1 and was confirmed by a Bartlett test of sphericity (3548; P < 0.001) and a Kaiser-Meyer-Olkin (KMO) measure of sampling ade- quacy (overall MSA = 0.63). The three principal components explained a cumulative variance of 86.8%. Principal Component 1 loaded strongly on mass (0.42), volume (0.41), and perimeter (0.37; Table 3). Principal Component 2 loaded strongly on SA:V (0.52), thickness (-0.44), and LDI (0.34), and Principal Component 3 loaded strongly on curl (0.50), area (-0.41), and LDI (0.38; Table 3). The PERMANOVA indicated that the species significantly differed in measured leaf traits ( F = 67.9; P = 0.001), and the multivariate analysis of homogeneity of variance revealed significant differences in the spread around the centroid of each species ( F = 32.2; P < 0.0001). A post-hoc Tukey-Kramer HSD revealed that all pairwise comparisons were significantly different in dispersion ( P < 0.001) except southern red oak and post oak ( P = 0.83), winged elm and southern red oak ( P = 0.80), and winged elm and post oak ( P = 0.99). 3.2. Leaf litter moisture dynamics Among single species litterbeds, oaks gained less moisture initially than non-oaks and lost moisture more quickly (Fig. 3A). Winged elm and hickory initial moisture contents (449.5% and 428.5%, respectively) Fig. 1. Leaf litter flammability experimental burn plots at Spirit Hill Farm, MS, USA (A) prior to burning, (B) during burning, and (C, D, E, and F) post-burn with 0%, 33%, 66%, and 100% non-oak leaf litter, respectively. J.K. McDaniel et al. Forest Ecology and Management 482 (2021) 118860 5 were ~ 1.5-fold higher than southern red oak and post oak initial moisture contents (283.3% and 288.8%, respectively; P < 0.0001; Table 4). As expected, litter mixtures that contained more non-oak leaf litter gained more moisture initially relative to mixtures that contained primarily oak litter (Fig. 3B). Litter mixtures containing 100% non-oak litter had an average initial moisture content that was ~ 1.6x that of litter mixtures containing 0% non-oak litter (Table 4). Initial moisture content exhibited a strong positive linear relationship with percent non- oak leaf litter (R 2 = 0.92; P < 0.0001; Fig. 4), but response time was not influenced by percent non-oak litter (R 2 = 0.01; P = 0.29). 3.3. Leaf litter flammability For mixed-species burns, all individual flammability metrics except total flaming duration exhibited negative linear relationships with percent non-oak leaf litter, as plots with increasing percent non-oak litter burned with a slower rate of spread, shorter flames, lower tem- perature, and less area burned (Fig. 1). Rate of spread (R 2 = 0.67; P < 0.001; Fig. 5A) and flame height (R 2 = 0.71; P < 0.001; Fig. 5B) were most strongly correlated with percent non-oak leaf litter and decreased as percent non-oak leaf litter increased. Area burned (R 2 = 0.43; P = 0.001; Fig. 5C) and temperature (R 2 = 0.56; P < 0.001, Fig. 5D) were less strongly but significantly correlated with percent non-oak leaf litter and also decreased as percent non-oak leaf litter increased. Bulk density and moisture content were not strongly correlated with percent non-oak leaf litter (Table 5). Rate of spread, flame height, area burned, and temperature were Table 2 Morphological trait values and principal component (PC) scores (mean ( ± standard error)) measured on individual leaves (n = 50) collected from Spirit Hill Farm, MS. Values followed by a common letter are not significantly different by a Kruskal-Wallis test followed by a Dunn test ( ∝ = 0.05). SLA, specific leaf area; LDI, leaf dissection index; SA:V, surface area to volume ratio, χ 2 , Kruskal-Wallis H-value, d.f., degrees of freedom. Species Surface area (cm 2 ) Perimeter (cm) Thickness (mm) Volume (cm 3 ) Mass (g) SLA (cm 2 g 1 ) LDI Tissue density (g cm 3 ) Curl (cm) SA:V PC1 PC2 PC3 Post oak 62.5 (3.3) a 63.8 (2.0) a 0.27 (0.01) a 1.7 (0.1) a 0.68 (0.03) a 91.97 (2.23) a 8.25 (0.19) a 0.42 (0.01) ac 2.2 (0.1) bd 76.2 (2.24) b 2.51 (0.18) 0.35 (0.13) 0.79 (0.16) Southern red oak 34.3 (1.5) bc 66.9 (2.3) a 0.26 (0.01) a 0.9 (0.1) b 0.35 (0.02) b 100.68 (2.13) a 11.5 (0.23) b 0.40 (0.009) ab 3.3 (0.2) a 80.4 (2.24) b 1.45 (0.15) 0.17 (0.10) 1.22 (0.10) Hickory 1 29.0 (3.4) b 29.3 (1.6) b 0.21 (0.01) b 0.6 (0.01) c 0.20 (0.02) c 144.63 (6.58) b 5.84 (0.08) c 0.37 (0.01) b 2.2 (0.1) b 102.9 (4.39) c 0.97 (0.14) 0.92 (0.16) 0.21 (0.16) Sweetgum 40.0 (3.0) c 57.7 (2.8) a 0.15 (0.01) c 0.6 (0.01) c 0.29 (0.03) bc 183.98 (12.23) bc 9.27 (0.15) a 0.48 (0.03) c 2.9 (0.2) ad 163.7 (9.56) a 0.53 (0.29) 1.79 (0.20) 0.20 (0.17) Winged elm 11.8 (0.7) d 22.3 (0.7) b 0.17 (0.01) c 0.2 (0.01) d 0.06 (0.004) d 201.13 (7.27) c 6.62 (0.07) c 0.31 (0.01) d 1.3 (0.1) c 123.5 (3.18) a 2.46 (0.08) 0.69 (0.08) 0.42 (0.07) P value < 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001 < 0.0001 χ 2 , d.f. 140, 4 170, 4 130, 4 160, 4 170, 4 130, 4 200, 4 84, 4 97, 4 130, 4 1 Hickory includes pignut hickory, mockernut hickory, bitternut hickory, and shagbark hickory. Fig. 2. Principal component analysis (PCA) biplot of morphological traits measured on leaf litter collected at Spirit Hill Farm, MS, USA. Points indicate individual leaves measured, and the length of the vector arrow represent the strength of each trait ’ s correlation with principal components. LDI, leaf dissection index; SLA, specific leaf area; SA:V, surface area to volume ratio. Hickory includes pignut hickory, mockernut hickory, bitternut hickory, and shagbark hickory. Table 3 Eigenvectors from principal component (PC) analysis of leaf traits measured on leaves collected from upland hardwood stands at Spirit Hill Farm, MS, USA. LDI, leaf dissection index; SLA, specific leaf area; SA:V, surface area to volume ratio. Variable PC 1 PC 2 PC 3 Surface area 0.35 0.24 0.41 Perimeter 0.37 0.37 0.12 Thickness 0.33 0.44 0.14 Volume 0.41 0.01 0.33 Mass 0.42 0.08 0.24 SLA 0.35 0.30 0.30 LDI 0.23 0.34 0.38 Density 0.11 0.26 0.37 Curl 0.13 0.26 0.50 SA:V 0.29 0.52 0.14 % variance explained 48.4 18.9 14.1 J.K. McDaniel et al. Forest Ecology and Management 482 (2021) 118860 6 strongly correlated, which justified conducting a PCA for mixed-species burns. Flaming time was weakly correlated with other variables (i.e. rate of spread, r = -0.19; temperature, r = 0.04) and was excluded from the PCA because variance explained was maximized without flaming time. Temperature and flame height ( r = 0.81) and temperature and percent area burned ( r = 0.81) were the most strongly correlated. Combining temperature, rate of spread, percent area burned, and flame height using a PCA resulted in two principal components and was confirmed by the Bartlett test (48.05, P < 0.001) and the KMO index (Overall MSA = 0.84). Principal Component 1 and Principal Component 2 of the PCA explained 87.9% and 6.0% of the variation in the dataset, respectively (Table 6, Fig. 6A). More flammable, oak-dominated mixtures had higher, more positive Principal Component 1 scores while less flam- mable, non-oak dominated mixtures had lower, more negative Principal Component 1 scores (Fig. 6A). A regression of Principal Component 1 and percent fire-sensitive non-oak leaf litter revealed a significant relationship between flammability and non-oak litter contribution to the fuelbed (R 2 = 0.58, P < 0.001, Fig. 7). Flammability metrics for single-species burns were also strongly correlated, and temperature and percent area burned most strongly correlated ( r = 0.85). A PCA of temperature, rate of spread, percent area burned, and flame height resulted in two principal components (Bartlett test = 34.05; Overall MSA = 0.70). Principal Component 1 and Principal Component 2 explained 76.8% and 14.3% of the variation in the dataset, respectively (Table 6, Fig. 6B). Principal Component 1 loaded strongly on percent area burned ( r = 0.55) and temperature ( r = 0.54), and Principal Component 2 loaded strongly on rate of spread ( r = 0.73) and flame height ( r = -0.68; Table 6). A PERMANOVA revealed significant differences in flammability metrics among species ( F = 5.21; P = 0.009), but it should be noted that we lacked a large sample size for some species (e.g., winged elm, n = 3). Bulk density varied significantly among spe- cies;winged elm had a bulk density ~ 1.8x that of southern red oak (Table 5). Fuel moisture content did not vary significantly among species. Fig. 3. Drying curves of (A) single-species litter and (B) mixtures of increasing proportion of non-oak litter relative to oak litter collected at Spirit Hill Farm, MS, and soaked for 24 h and dried for 48 h under laboratory conditions. Points indicate means and error bars represent standard error. Hickory includes pignut hickory, mockernut hickory, bitternut hickory, and shagbark hickory. Table 4 Initial moisture content and response time (mean ( ± standard error)) measured during litter soaking and drying experiment of single and mixed species leaf litter collected at Spirit Hill Farm, MS from fall 2017 to winter 2018. Values sharing a common letter are not significantly different. ( ∝ = 0.05). Litter type Initial moisture content (%) Response time (h) 1 Single species Southern red oak 283.3 (0.6) a 4.1 (0.9) Post oak 288.8 (3.7) a 4.6 (1.7) Sweetgum 393.8 (10.9) b 7.6 (0.6) Hickory 2 428.5 (5.0) c 5.0 (2.0) Winged elm 449.5 (4.1) c 6.6 (2.9) P value < 0.0001 0.40 Litter mixture (% non-oak litter) 0 266.2 (7.9) 9.3 (0.4) 33 319.8 (6.2) 8.1 (0.6) 66 374.9 (6.1) 6.1 (0.8) 100 421.9 (11.5) 12.0 (1.0) P value < 0.0001 0.21 R 2 0.92 0.03 1 Response time measured as the time necessary for litter to lose 63% of moisture and transition from the first drying time lag to the second time lag as determined by piecewise polynomial curve fitting. 2 Hickory includes pignut hickory, mockernut hickory, bitternut hickory, and shagbark hickory. Fig. 4. Relationship of percent non-oak leaf litter and initial moisture content measured during leaf litter soaking and drying experiment of mixed oak and non-oak leaf litter collected at Spirit Hill Farm, MS. J.K. McDaniel et al. Forest Ecology and Management 482 (2021) 118860 7 4. Discussion Our study demonstrates that leaf litter traits, moisture dynamics, and flammability differed between upland oak and non-oak species and that flammability decreased as non-oak leaf litter contribution to the fuelbed increased. When compared to the three non-oaks (winged elm, hickory, and sweetgum), the two upland oaks (southern red oak and post oak) had larger, thicker leaves with a lower SA:V and lower SLA, which is consistent with other studies ’ findings that fire-tolerant species exhibit traits associated with higher flammability than fire-sensitive species (Engber and Varner, 2012; Grootemaat et al., 2017; Kreye et al., 2013). The three non-oaks exhibited greater moisture gain and a slower drying rate than the two oaks, and mixed species litterbeds with increasing non- oak proportion exhibited increased initial moisture content as expected given the behavior of single species litterbeds. This indicates that the addition of non-oak litter could dampen flammability and decrease ignition probability due to higher initial fuel moisture after the addition of water. We expected to find slower moisture loss rate with increasing percent non-oak leaf litter, but we instead found no clear trend. This could be due to additive effects that have been documented in litter flammability (Varner et al., 2017; Zhao et al., 2019, 2016), with the most flammable species controlling the flammability of mixtures (de Magalh ̃ aes and Schwilk, 2012). Drying rates of mixed litter types could potentially behave similarly, where the species with the quickest response time controls the response time for the mixture through increased fuelbed depth and decreased bulk density created by curlier, larger leaves. For example, our 33% and 66% non-oak fuelbed response times were most similar to the response time of 0% non-oak litter, which could indicate the oak litter in the mixed fuelbeds is at least partially controlling litter drying. Additive versus non-additive effects of litter drying deserve future work to fully understand the controls of individual species on drying of litter mixtures. Fig. 5. Relationship of percent non-oak leaf litter and (A) rate of spread (ROS), (B) flame height, (C) percent area burned, (D) temperature measured during plot-level experimental burns of mixtures of increasing proportion of non-oak litter relative to oak-litter at Spirit Hill Farm. J.K. McDaniel et al. Forest Ecology and Management 482 (2021) 118860 8 In mixed-species flammability experiments, as expected, rate of spread, percent area burned, flame height, and temperature decreased with increasing percent non-oak leaf litter. Total flaming time did not have a significant relationship with percent non-oak leaf litter, likely because plots with non-oak-dominated fuels either burned very slowly or extinguished soon after ignition and had low fuel consumption. Re- lationships between other flammability metrics and non-oak leaf litter contribution were as expected. Combining individual flammability metrics resulted in an overall flammability score that decreased with increasing percent non-oak leaf litter. This is consistent with the findings of Kreye et al. (2018), who examined laboratory flammability of mix- tures of similar species found in northern Mississippi and found that flammability metrics decreased as the contribution of sweetgum, winged elm, and dogwood ( Cornus florida L.) litter increased; this trend was strongest when litter was moist. For our single-species flammability experiment, there were significant differences in overall flammability as indicated by a PERMANOVA. Two potential mechanisms of such dampened flammability include increased moisture content and differ- ences in leaf characteristics, which we demonstrated here, that lead to increased bulk density (Kreye et al., 2013; Varner et al., 2015). Bulk density trended higher as the percent non-oak litter increased in mixed- species flammability experiments, potentially driven by winged elm ’ s increased bulk density relative to oak as found in single-species flam- mability experiments. Interestingly, bulk density was generally lower in mixed-species flammability experiments compared to single-species experiments, which may be due to interactions between litter types. For example, the bulk density of 0% non-oak litter was lower than

举报滥用

PDF 详情

联系信息

Shifting tree species composition of upland oak forests alters leaf litter structure, moisture, and flammability

Jennifer K. McDaniel