A forest reconstruction model to assess changes to Sierra Nevada mixed-conifer forest during the fire suppression era

Forest Ecology and Management 354 (2015) 104–118

Contents lists available at ScienceDirect

Forest Ecology and Management journal homepage: www.elsevier.com/locate/foreco

A forest reconstruction model to assess changes to Sierra Nevada mixed-conifer forest during the fire suppression era Molly A.F. Barth a,⇑, Andrew J. Larson a, James A. Lutz b a b

Department of Forest Management, College of Forestry and Conservation, University of Montana, Missoula, MT 59812, USA Wildland Resources Department, Quinney College of Natural Resources, Utah State University, Logan, UT 84322, USA

a r t i c l e

i n f o

Article history: Received 17 May 2015 Received in revised form 24 June 2015 Accepted 25 June 2015 Available online 30 June 2015 Keywords: Tree growth Fire history Restoration Sugar pine Tree decay Pinus lambertiana

a b s t r a c t Fire suppression has resulted in dramatic changes to species composition and structural diversity in the Sierra Nevada mixed-conifer forests. We need a better understanding of how these forests have changed during the fire suppression era, but empirical historical datasets are rare and methodologies for developing new historical reference information are subject to limitations. We sought to develop historical reference information for the Yosemite Forest Dynamics Plot (YFDP), a research plot located in an old-growth mixed-conifer forest in Yosemite National Park. We performed a dendrochronologial fire history analysis to characterize the historical fire regime of the YFDP, resulting in an estimated pre-1900 point fire return interval of 29.5 years. We then developed two different forest reconstruction models to estimate structural and compositional forest changes since 1900, the year the last widespread fire burned the YFDP, to the present. We explored the use of two different tree growth models—a regionally parameterized competition-dependent model and a parsimonious site-specific model—as well as a decay model based on published estimates of wood decay rates. The competition-dependent growth model predicted slightly higher stem densities (175 trees ha1 vs. 112 trees ha1 in 1900) and slightly lower basal area (20.9 m2 ha1 vs. 24.1 m2 ha1 in 1900) than the site-specific growth model. Predictions about dead trees, especially large diameter sugar pines, are potentially inaccurate, both in this study and other reconstruction studies in the Sierra Nevada, due to a lack of size-specific snag and log decay rate data. Our study highlights the need for more detailed decay data for Sierra Nevada mixed-conifer forests. While reconstruction models are constrained by the data used to parameterize them, they can still produce estimates of historical conditions that are useful for understanding the direction and magnitude of forest change, as well as for planning forest management. Our analysis demonstrates dramatic changes in forest conditions since fire-exclusion—a clear ecological basis for restoration. We encourage managers to focus on restoring key ecological processes that maintain ecosystem structure and function, in this case fire, rather than attempting tree-for-tree recreations of historical structure without reintroducing fire. Ó 2015 Elsevier B.V. All rights reserved.

1. Introduction Fire suppression has resulted in dramatic changes to species composition and structural diversity in dry coniferous forests across the western United States (Abella et al., 2007). The mixed-conifer forests of the Sierra Nevada in California, including those in Yosemite National Park (Yosemite), are no exception (Scholl and Taylor, 2010). Climate change may confound or intensify existing ecological problems, and could bring a continued increase in fire frequency and severity (Miller and Urban, 1999; ⇑ Corresponding author. E-mail addresses: [email protected] (M.A.F. umontana.edu (A.J. Larson), [email protected] (J.A. Lutz). http://dx.doi.org/10.1016/j.foreco.2015.06.030 0378-1127/Ó 2015 Elsevier B.V. All rights reserved.

Barth),

a.larson@

Cansler and McKenzie, 2014, but see Lutz et al., 2011), tree species range shifts (Lutz et al., 2010), declines in forest productivity (Battles et al., 2008), drought-triggered tree mortality (Guarín and Taylor, 2005), and a loss of biological diversity (Stephenson and Parsons, 1993). Given the significant alteration of these forests and the uncertain ecological impacts of future climate change, it is imperative that we exercise timely adaptive management and restoration based on the best available science if we hope to sustain western dry forests and the ecological services they provide. Forest reconstruction is a technique in which contemporary inventories of live and dead trees are used to estimate forest structure and composition at some point in the past (Fulé et al., 1997). When historical forest structure datasets (e.g., timber inventories, historical photos, and land surveys) are rare or lacking,

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118

reconstructions often represent the best available option for obtaining historical reference information. This information is important for guiding management and for investigating changes caused by the exclusion of fire or other past land uses, as well as for understanding the mechanisms of change in forest ecosystems. It is essential to have an understanding about the accuracy, limitations, and uncertainties of reconstruction models when using this approach to assess ecological change or guide management (Harmon et al., 2015). Forest reconstructions are an invaluable research tool and are worth improving because they offer a way to obtain new information and can potentially provide more detailed, site-specific data than extant historical datasets. Reconstructions generally rely on several assumptions, including: (1) all evidence of historical forest structures is detectable during contemporary inventories; (2) the ages of snags and logs can be determined based on a field classification of tree decay; and (3) decay rates are consistent across size classes. There have been some efforts to investigate the impact of variable decay rates on reconstruction estimates through sensitivity analyses (Fulé et al., 1997; Scholl and Taylor, 2010; Taylor et al., 2014), but thorough investigations of uncertainties in decay and growth are rare despite the widespread use of reconstructions to investigate changes from historical conditions in frequent-fire forests (Arno et al., 1995; Harrod et al., 1999; North et al., 2007; Beaty and Taylor, 2007; Van de Water and North, 2011). Sierra Nevada mixed-conifer forests present an opportunity for forest scientists to investigate how management interventions (e.g., fire exclusion and grazing) and shifts in climate (e.g., the end of the Little Ice Age) may have influenced forest structure, dynamics, and ecological functions and services over the past century. With this study, we seek to expand our ability to obtain new scientific evidence needed to both improve our ecological understanding as well as address specific research needs identified in contemporary Sierra Nevada mixed-conifer management frameworks (North et al., 2009; North, 2012). Specifically, our objectives are to: (1) develop a forest reconstruction model for Sierra Nevada mixed-conifer forests and evaluate the use of two alternative tree growth models; (2) systematically investigate the consequences of uncertainty in model components, including decay rates, on reconstructed estimates of historical forest structure and composition; (3) evaluate model performance by reconstructing historical forest structure and composition of a Sierra Nevada mixed-conifer forest at the time of the last widespread fire and comparing our results to existing historical datasets.

105

2.2. Data collection Baseline tree data in the YFDP have been collected following the protocols of the Smithsonian ForestGEO network (Anderson-Teixeira et al., 2015). All live trees P 1 cm dbh, all snags P 10 cm dbh and P 1.8 m in height, and all logs P 50 cm in their largest diameter were identified and mapped (for details, see Lutz et al., 2012). We augmented the existing measurements with additional data on logs <50 cm in diameter and snags <1.8 m tall. To avoid over-sampling these dead trees, we used estimates of species-specific tree growth rates developed from local Forest Inventory and Analysis data, species-specific log decay rates (Harmon et al., 1986), as well as allometric equations relating dbh to diameter at stump height (dsh) (Walters and Hann, 1986; Weigel and Johnson, 1997) to estimate the minimum dsh that stumps and logs would have to be in the present to have been alive in 1900. For each log and short snag that met minimum dsh requirements (white fir: 30 cm; sugar pine: all; incense cedar: 10 cm; black oak: 15 cm), we recorded the species and decay class (Thomas et al., 1979), and estimated dbh prior to decay using structural clues (Fig. 1). We mapped the original rooting locations of all logs. The YFDP contemporary tree inventory includes 35,498 live trees, 2734 snags, and 696 logs. We collected tree increment cores around the perimeter of the YFDP to estimate local species-specific growth rates (sampling within plot boundaries was not permitted). We sampled individual trees from the principal species present on the plot (white fir: n = 27, sugar pine: n = 34, incense cedar: n = 35, black oak: n = 11) and sampled throughout the diameter distribution, although we could not core trees <10 cm or >130 cm dbh. Although fire suppression began regionally as early as 1891 (Rothman, 2007; van Wagtendonk, 2007), we needed to know the year of the last widespread fires on the YFDP to set an appropriate year for the reconstruction. To assess YFDP fire history, we removed cross-sections from dead fire scarred trees (n = 10 sugar pine and n = 2 incense cedar) within and immediately adjacent to the plot (see Barth, 2014 for maps detailing sample locations and study area). Cross-sections were prepared using standard dendrochronological techniques (Stokes and Smiley, 1968) and growth rings were measured using CooRecorder version 7.5 (Cybis Elektronik & Data AB, Sweden). Fire scars identified by their characteristic ring disruption (McBride, 1983). Cross-sections were cross-dated by establishing marker years using locally developed tree-ring chronologies (King and Graumlich, 1990; Barth et al., 2014)

2. Methods 2.1. Study site This study used field data from the Yosemite Forest Dynamics Plot (YFDP), a 25.6 ha (800 m  320 m) permanent plot established in an old-growth, mixed-conifer forest near Crane Flat in Yosemite National Park (Yosemite), California (Gabrielson et al., 2012; Lutz et al., 2012, 2014). The YFDP is centered at 37.77°N, 119.82°W, with an elevation of 1774–1911 m. The climate is Mediterranean, with warm dry summers and cool, wet winters. Soils are primarily metasedimentary soils of the Clarkslodge-Ultic Palexeralfs complex (Lutz et al., 2012). Major tree species on the plot include sugar pine (Pinus lambertiana), white fir (Abies concolor), incense cedar (Calocedrus decurrens), California black oak (Quercus kelloggii) and Pacific dogwood (Cornus nuttallii). Canopy emergent trees, primarily sugar pine and white fir, reach 60–67 m in height and over 200 cm in diameter at breast height (1.37 m above ground level; dbh). The forest within the YFDP has never been subject to timber harvest.

Fig. 1. Field assistant Erin Costello estimates a dead tree’s original dbh, using structural clues to account for bole loss due to decay since tree death. Photo by M. Barth.

106

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118

and individual ring formation years and fire scars were assigned to calendar years. Fire interval statistics were calculated using the FHX2 software package (Grissino-Mayer, 2001). 2.3. Objective 1: Develop forest reconstruction models We chose to develop data-driven reconstruction models (Fig. A.1) to derive historical estimates of stand structure and composition (Bakker et al., 2008), as opposed to using a dendrochronological reconstruction approach in which the presence or absence and size of all contemporary trees at some point in the past is estimated by analyzing a core from each tree (Arno et al., 1995). Modeled reconstructions are less labor intensive and more feasible for use at high-spatial resolution across large, contiguous study areas (i.e., stem map plots), such as the YFDP. We developed two different tree growth models and combined each with a single decay model, resulting in two separate reconstruction models. The models were written and implemented in R version 3.0.2 (R Development Core Team, 2014). 2.3.1. Competition-dependent tree growth model First, we used a regionally calibrated, competition-dependent tree growth model developed for the Sierra Nevada mixed-conifer forests (Das, 2012) to predict past diameter growth increment. Although the Das (2012) model was not developed specifically for the YFDP, we chose to employ it because it was a low-cost reconstruction method that could be used on a large dataset (30,000 trees) and because of the potential that our reconstruction model could then be used to reconstruct other mixed-conifer sites in the Sierra Nevada. The Das (2012) growth model is parameterized for the major conifer tree species found on the YFDP, including sugar pine, incense cedar, and white fir. Future diameter growth for a given focal tree is in part predicted by the neighborhood crowding index (NCI) surrounding the focal tree (trees that are within 618.5 m radius are considered neighbors, with the radius dependent on focal species). The NCI characterizes the interspecific and diameter-dependent competitive interactions between a focal tree and all of its neighbors (Das, 2012). We modified the Das (2012) model to predict past diameter growth at five year intervals. In addition, because the Das (2012) model is not parameterized for California black oak or Pacific dogwood, we developed five year diameter growth increment estimates from black oak cores along the perimeter of the plot (mean: 1.04 cm5yr) and estimated Pacific dogwood growth using published five year diameter growth rates of Pacific dogwood trees in Oregon (mean: 1.02 cm5yr) (Hann and Hanus, 2002). To account for edge effects while calculating crowding indices, we established an 18.5 m buffer (the maximum potential radius of neighborhood influence) around all plot edges; trees inside this buffer zone were not considered in subsequent analyses. 2.3.2. Site-specific tree growth model We were curious how the use of a more parsimonious growth model in which tree growth rates were based on locally derived tree growth estimates might affect reconstructed results. We used tree cores collected around the YFDP to develop estimates of five year diameter growth rates for sugar pine, white fir, incense cedar, and black oak. For each tree, we calculated radial growth at five-year intervals, and then averaged these rates across the entire core to obtain a mean five-year radial growth rate for each tree. These growth rates were then averaged across all cores for each species. Trees cores were measured to pith when possible, but if the pith was not present, we measured the entire length of the intact core.

In this site-specific growth model, we set tree growth rates based on the growth rates derived from the cores. Since we did not calculate diameter-dependent growth rates, modeled growth remained constant regardless of tree age. Dogwood growth rates were the same as the competition-dependent model (Hann and Hanus, 2002). We did not employ any edge adjustments for the site-specific model because tree growth was not based on spatial relationships to other trees, but only analyzed trees within the core plot area (i.e., we excluded trees in the 18.5 m buffer zone) to allow for direct comparison to results from the competition-dependent model. 2.3.3. Tree decay To ‘‘undecay’’ trees, we followed a similar methodology to other nearby reconstruction studies (North et al., 2007; Scholl and Taylor, 2010; Van de Water and North, 2011), which involved piecing together the best available decay data for each species (Table A.1). We estimated the time for sugar pine and white fir snags to transition between decay classes from transition matrices developed by Morrison and Raphael (1993) which predict genus-specific snag decay class transitions over time (Pinus rates developed from sugar pine, Jeffery pine (Pinus jeffreyi) and lodgepole pine (Pinus contorta); Abies developed from white fir and red fir (Abies magnifica)). We predicted the age of incense cedar snags using estimates for western redcedar (Thuja plicata) presented in Daniels et al. (1997). Oak snags transitioned to lesser stages of decay based on a equalized probabilistic transition matrix, due to lack of available decay information – for example, a decay class 3 oak would have an equal likelihood of remaining decay class 3 (33.3% chance) or transitioning at the next earlier interval to either decay class 2 (33.3% chance) or decay class 1 (33.3% chance). Log decay for sugar pine, white fir, and oak was modeled based on the exponential decay function (Harmon et al., 1986): Dt ¼ Do ekt (Dt is the wood density (g cm3) at time t (years), Do is the initial wood density, and k is the species-specific decay rate constant for density). Species-specific decay rate constants used in this function were derived from the literature (MacMillan, 1981; Harmon et al., 1987; Dunn, 2011), as were estimates of wood density by decay class (Harmon et al., 2008). We substituted ponderosa pine (Pinus ponderosa) decay rates for sugar pine and eastern oak decay rates for black oak. Incense cedar log time since death was estimated based on data for western redcedar presented in Daniels et al. (1997). 2.3.4. Incorporating stochasticity We used a simulation approach in which we ran each reconstruction model (competition-dependent growth model + decay model or site-specific growth model + decay model) 100 separate times, resulting in a unique historical forest for each simulation. This approach allowed us to incorporate variability in growth and decay rates across simulations as a way to imitate the natural variability between individual trees and over time. During each run of the competition-dependent model, tree growth and competition parameters (12 total), which were established once during each run, were allowed to vary based on the 2-unit support intervals presented for each parameter in Das (2012) and Barth (2014). In the site-specific model, a tree’s individual growth rate was derived by generating a random normal deviate based on the mean and standard deviation of the species-specific five-year radial growth rate measured in the YFDP tree cores. Transition times between decay classes and log-bole decay rates for each dead tree also varied between simulations by generating a random normal deviate from the mean and standard deviation of the rates as presented in the literature. We employed the R ‘‘rnorm’’ function to generate random normal deviates for each parameter.

107

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118

2.4. Objective 2: Systematically investigate uncertainties 2.4.1. Effects of stochasticity We sought to understand which types of trees (including different species and diameter classes) were most sensitive to variability in growth and decay rates. To gain insight into components of each model version, we calculated, for each tree in the dataset, the probability of being alive in 1900 based on results from all simulations. We also investigated how the predicted 1900 dbh of each individual tree varied across simulations of both models. Additionally, we sought to understand how overall stand metrics (such as live tree stem density and total basal area) changed across simulations to see if incorporating stochasticity in the individual tree growth and decay rates introduced variation into the stand level metrics. 2.4.2. Quantifying missing evidence We used the decay model to investigate the potential for missing evidence of white fir and sugar pine trees (the two most prevalent tree species, 45.0% and 45.9% of the total 2010 live tree basal area, respectively) that would need to be accounted for in our results, assuming decay rates were accepted as accurate. We decayed hypothetical trees from 1900 to 2010 and then calculated each tree’s predicted wood density (g cm3). We deemed a tree ‘‘undetectable’’ if its density in 2010 was more than one standard error below the mean decay class 5 density for its species (Harmon et al., 2008). We ran this simulation 100 times, allowing the decay rates and snag transition times to vary stochastically as in the reconstruction models, then quantified the probability of not detecting each tree in 2010, depending on the year of its death. 2.5. Objective 3: Reconstruct historical stand structure and composition For each simulated reconstruction from the competition and site-specific models, we calculated total number of stems, trees per hectare (tph), and basal area (m2/ha) for each 10 cm diameter class of each species, as well as for all species pooled. For each model, we calculated the mean, standard deviation and range of these metrics across simulations to gain insight into the variability of stand structure and composition. We compiled historical data from the published literature with which to compare and validate the results from our reconstruction. The historical data included both empirical datasets and reconstructed estimates for other fire-suppressed mixed-conifer sites within the greater Sierra Nevada region from 1865 to 1935. 3. Results 3.1. Historical fire regime The most recent widespread fire to burn the YFDP occurred in 1900, and we therefore chose 1900 as the reference year for the reconstruction. Mean PFRI during the pre-suppression period (before 1900) was 29.5 years; the fire return interval departure for this site is 2.2 (Table 1). We found evidence for small, localized fires occurring into the fire suppression era (i.e., in 1916, 1926, and 1947) (Barth, 2014). Fires occurred most often late in the growing season and after dormancy. 3.2. Model performance 3.2.1. Tree growth Analysis of tree growth rates across simulations of the competition-dependent model and comparison to growth rates derived from tree cores collected around YFDP demonstrates that

Table 1 Mean point fire return intervals (PFRI) in years as determined from samples collected from dead trees on the Yosemite Forest Dynamics Plot, Yosemite, California, USA. Time

Mean PFRI (SD)

1600s 1700s 1800s 1900s Pre-1900

17 30 29 65 30

All years Mean Median Minimum

39 (37) 23 6

(15) (25) (16) (37) (25)

the competition-dependent growth model produced growth rates in general agreement with local empirical data (Fig. 2). When averaged across all trees, the competition-dependent model resulted in 5-year growth rates that were lower than estimates based on increment cores for sugar pine (modeled: 1.22 cm yr5, empirical: 2.1 cm yr5), white fir (modeled: 0.96 cm yr5, empirical: 2.4 cm yr5), and incense cedar rates (modeled: 1.03 cm/yr5, empirical: 2.2 cm yr5). The growth increments in the site-specific model were closest to actual data, as would be expected given that in this model version tree growth rates were based on growth rates derived from the tree cores (Fig. 2). The competition-dependent model tended to produce slower diameter growth rates in the small diameter classes and faster diameter growth rates in the large diameter than the species level mean diameter growth rates based on local tree cores. The effect of these differential growth rates is apparent in the reconstructed time series of stem density and basal area (Fig. B.1), which show a faster decrease in stem density but slower decrease in basal area for the site-specific model compared to the competition-dependent model. 3.2.2. Tree decay The average of the decay model results predicted that 334 (of 2734) snags and 185 (of 696) logs would have been alive in 1900. White fir decay was fastest (average age of decay class 5 snag: 23 years), with slower rates for sugar pine (average age of decay class 5 snag: 26 years) and the slowest rates for incense cedar (average age of decay class 5 snag: 160 years). In general, snags were modeled to have decayed more slowly than logs. 3.3. Investigation of uncertainties 3.3.1. Effects of stochasticity In both reconstruction models, trees in larger diameter classes exhibited higher variability in their probability of presence in 1900 than trees in smaller diameter classes, likely due to the lower number of large-diameter trees (Fig. 3). In the site-specific model, probabilities of presence in 1900 for moderate diameter classes (10–100 cm dbh) were higher than the competition-dependent model. In the competition-dependent model, white fir and incense cedar exhibited high variability in projected historical dbh, evidenced by the large vertical spread of points for a given 2010 diameter class (Fig. 4). In the site-specific model, variability in tree growth was much lower in general and more consistent between species (Fig. 4). 3.3.2. Quantifying missing evidence Based on decay rates and the parameters in our decay model, there would be a high likelihood of not detecting trees that were alive in the reference year due to lack of evidence. The probability of not detecting sugar pines is not linearly related to year of death and there is generally a lower probability of not detecting sugar pines compared to white firs (Fig. 5).

108

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118

Fig. 2. Top panels: modeled tree growth of 50 randomly selected trees derived from the competition-dependent model (colored). Bottom panels: modeled tree growth for the same 50 random trees based on the site-specific model (colored). Left panels show growth of white fir (Abies concolor; ABCO); center panels show growth of sugar pine (Pinus lambertiana; PILA); and right panels show growth of incense cedar (Calocedrus decurrens; CADE). Grey symbols in all panels show empirical growth data from tree cores collected immediately adjacent to the Yosemite Forest Dynamics Plot. Note that for modeled tree growth some trees have a dbh of 0 cm at or after 1900, indicating they established that year. Trees that have a measureable dbh in 1900 are estimated to have established and grown to that diameter prior to 1900. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3. The probability of contemporary (2010) live trees on the Yosemite Forest Dynamics Plot being alive in 1900 based on both the competition-dependent and site-specific models. Error bars represent the standard error across each size class from 100 simulations of each model.

3.4. Reconstructed historical stand structure and composition Analysis of stand structure and composition in the 1900 reference plots revealed that competition-dependent and site-specific

models predict historical stand structures that are in general agreement with each other (Table B.1), with estimates of live tree density (stems P 1 cm dbh) of 175 tph (competition-dependent) and 112 tph (site-specific), and live basal area of 20.9 m2 ha1

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118

109

Fig. 4. Predicted historical (1900) dbh of contemporary (2010) live trees on the Yosemite Forest Dynamics Plot based on the competition-dependent model and the sitespecific model. Each point represents the estimated dbh from a single simulation; all simulations (n = 100) are displayed to show variability in reconstructed dbh across simulations.

(dbh P 100 cm) overall and for sugar pine specifically compared to the site-specific model (Table B.1). 4. Discussion 4.1. Comparison with other reconstruction studies and historical datasets

Fig. 5. Probabilities of not detecting logs (A) and snags (B) from trees that may have been alive in 1900 but have died between 1900 and 2010 on the Yosemite Forest Dynamics Plot. Estimations derived from 100 forward simulations of the decay component of the reconstruction model.

(competition-dependent) and 24.1 m2 ha1 (site-specific). The effect of size-dependent growth rates in the competitiondependent model on reconstructed historical stand structure is apparent in the relatively higher density of trees <20 cm dbh compared to the site-specific model estimates (Fig. 6), as well as in the differential reconstructed time series of stem density (Fig. B.1). The models both predict that species composition was markedly different historically compared to the contemporary (i.e., 2010) co-domination by white fir and sugar pine (45.8% and 44.8% of basal area, respectively). Modeled 1900 forest basal area (Table B.1) was 61.7% sugar pine (competition-dependent) and 74.3% sugar pine (site-specific). The competition-dependent model predicted slightly lower densities of large-diameter trees

Previous studies of historical forest conditions in Sierra mixed-conifer forests provide a basis for evaluating our model predictions through comparative analysis (Table 2). Such comparisons serve as a coarse model assessment due to the unique biophysical setting and management histories of each site (Abella and Denton, 2009). We acknowledge that there are inherent differences between sites in mixed-conifer forests across the Sierra Nevada region and that even reconstructing the YFDP to the same reference year of the other studies as we did (Table 2) does not account for all these differences. However, this comparative analysis is nevertheless a useful way to explore our model outputs and theorize about potential explanations for differences in historical conditions. Three studies reconstructed historical conditions shortly after the last widespread fires burned through their respective study areas prior to fire suppression. Scholl and Taylor (2010) reconstructed historical forest structure and composition in 1899 at Big Oak Flat, a site located downslope and north of the YFDP, and estimated similar tree density and slightly higher basal area in 1899 than we estimated for the YFDP, with 160 tph P 10 cm dbh and a basal area of 30 m2 ha1. Similarly, Van de Water and North (2011) reconstructed pre-suppression forest conditions in a northern Sierra Nevada mixed-conifer forest and estimated a historical density of 201 tph (P5 cm dbh), but estimated a basal area of 21.4 m2/ha, very similar to our estimates. North et al. (2007) reconstructed forest conditions in 1865 after the last widespread fire at their study site in the Teakettle Experimental Forest, a mixed-conifer stand in the southern Sierra Nevada. They estimated only 65 tph P 5 cm dbh, although their basal area estimate (51.5 m2 ha1) for 1865 was relatively higher than we estimated for the YFDP in 1900. Empirical historical datasets offer another opportunity to examine our results in a comparative framework. Collins et al. (2011) analyzed a rare historical dataset detailing forest conditions in 1911 in the Gin Flat and Crane Flat areas of Yosemite; some of these plots overlap Big Oak Flat (Scholl and Taylor, 2010) and are also near the YFDP. They found that there were roughly 60 tph

110

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118

Fig. 6. Contemporary diameter distribution of trees P1 cm dbh on the Yosemite Forest Dynamics Plot (upper panels) and reconstructed 1900 diameter distributions based on the competition-dependent (middle panels) and site-specific (lower panels) growth models. Error bars represent the standard error across 100 simulations of each reconstruction model. Note: diameter distributions are truncated—trees >170 cm dbh are present at low densities in the reconstruction estimates but are not shown for figure clarity.

Table 2 Comparison of other studies investigating historical conditions in Sierra Nevada mixed-conifer forests to estimates produced by our two model versions using the same reference year and diameter cutoff from each study. MC: mixed conifer; BA: basal area. Study

Location

Year

DBH cutoff (cm)

Historical conditions

Competition model

Site-specific model

Reconstruction studies Scholl and Taylor (2010) Van de Water and North (2011) North et al. (2007)

Yosemite N. Sierra Nevada Teakettle Experimental Forest

1899 1900 1865

P10 P5 P5

160 tph|30 m2/ha 201 tph|21.4 m2/ha 67 tph|56.4 m2/ha

95 tph|20.8 m2/ha 125 tph|20.9 m2/ha 125 tph|20.9 m2/ha

85 tph|24.1 m2/ha 98 tph|24.1 m2/ha 98 tph|24.1 m2/ha

Historical datasets Collins et al. (2011) Knapp et al. (2013) Lutz et al. (2009) Stephens et al. (2015) Stephens et al. (2015)

Yosemite Stanislaus Nat’l Forest Yosemite S. Sierra Nevada – MC High BA S. Sierra Nevada – MC Ave BA

1911 1929 1935 1911 1911

P15.2 P10 P10 P30.5 P30.5

60 tph|n.a. 315 tph|53.9 m2/ha 232 tph|n.a. 98 tph|41.8 m2/ha 59 tph|24.6 m2/ha

87 tph|24.0 m2/ha 149 tph|32.8 m2/ha 162 tph|35.2 m2/ha 55 tph|22.8 m2/ha 55 tph|22.8 m2/ha

83 tph|27.0 m2/ha 121 tph|34.4 m2/ha 130 tph|36.5 m2/ha 59 tph|26.0 m2/ha 59 tph|26.0 m2/ha

(P15.2 cm dbh) in 1911, marginally lower than our estimates of tree densities for this size class (Table 2). This lower density compared to Big Oak Flat (Scholl and Taylor, 2010) is possibly due to the exclusion of trees in the 5–15.2 cm dbh size class and diminished tree density due to the 1899 fire which may have burned plots prior to data collection (Collins et al., 2011). Lutz et al. (2009) investigated vegetation data collected in surveys organized by Albert E. Wieslander in Yosemite between 1932 and 1936 (Wieslander, 1935); 21 of these Wieslander plots lie within 5 km of the YFDP. These 21 plots illustrate a higher tree density in 1935 (232 tph P 10 cm dbh) than either of our models predict, and large-diameter trees (>91 cm dbh in the Wieslander surveys) were recorded as 26.5 tph as compared to model results of 4.0 tph (competition-dependent) and 7.4 tph (site-specific). Knapp et al. (2013) explored a historical dataset collected in 1929 in a mixed-conifer stand of the Stanislaus-Tuolumne Experimental Forest (STEF), located approximately 50 km north

of the YFDP, where the last widespread fire burned in 1889. They found that there were 315 tph (P10 cm dbh) with a basal area of 53.9 m2/ha. It is important to note that although the STEF data was collected 40 years after the last widespread fire, a 40-year fire interval departure was not unreasonable for the site when fires were actively burning, and therefore conditions in 1929 were likely appropriate reference conditions (Knapp et al., 2013). Additionally, Stephens et al. (2015) investigated an extensive 1911 dataset in the mixed-conifer forests of the Greenhorn Mountains of the southern Sierra Nevada, where average mixed-conifer forests exhibited conditions very close to our reconstruction (Table 2). While there are slight differences in historical conditions across the studies, our reconstruction estimates are in general agreement with these independent data sources (Table 2). Other studies of historical forest conditions and contemporary changes throughout the region confirm our model results (Parsons and DeBenedetti, 1979; Taylor, 2004; North et al., 2007; Beaty and Taylor, 2007;

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118

Dolanc et al., 2014). We conclude that our model results are generally consistent with historical datasets and represent a level of accuracy appropriate for assessing changes to forest structure and composition during the fire-suppression era. 4.2. Model assessment and model-driven sources of error Evaluation of each component in our reconstruction model provides insight as to why our estimates of historical forest stand density, basal area, and tree sizes might differ from other reconstructions and empirical historical datasets. In general, our modeled reconstruction approach is inherently less accurate than dendrochronological approaches that use tree cores collected from each live and dead (when possible) tree to determine a tree’s presence or absence and diameter during the reference year (Arno et al., 1995; Fulé et al., 1997). Our approach instead utilizes a simulation model (Bakker et al., 2008) and each model component introduces uncertainties in estimates of historical forest conditions. 4.2.1. Competition-dependent model We have the highest confidence in the growth rates and reconstruction model estimates of historical conditions that are based on the competition-dependent growth model (Das, 2012). The Das (2012) model accounts for both size and competition effects on tree growth, and was parameterized with measurements of trees that completely span the range of tree diameters in the YFDP dataset. Nevertheless, we acknowledge inherent limitations to this modeling approach. One major assumption in modeling tree growth is that growth rates are directly linked to competition and modeled rates do not take into account other environmental or biological factors that may affect growth (Das, 2012). For example, the competition-dependent model does not account for non-competitive density-dependent effects, such as exposure to pathogens, which are difficult not only to quantify, but also to predict in a modeling framework. The competition-dependent model also does not take into account the site characteristics specific to the YFDP, such as soil productivity, which would affect tree growth at the stand scale. Moreover, in this framework, the plot is treated in two-dimensional space with no regard to local environmental heterogeneity, yet trees growing with access to different light environments or belowground resources would likely differ in growth (Canham et al., 2006; Astrup et al., 2008). While the original Das (2012) growth model was developed to predict future five year radial growth rates, we employ the growth parameters to predict past five year radial growth rates. To run the model backwards in time required us to bootstrap the model using mean neighborhood crowding index (NCI) values (developed for each diameter class of each species present on the plot during a single five year period), to predict the growth rates of a focal tree’s competitors. Additionally, the data Das (2012) used to parameterize his growth model was based on tree growth rates from 2000– 2009. It is possible that growth rates during this time period are not representative of growth rates over the past century, especially for long-lived trees such as sugar pine. For example, many of the large-diameter sugar pine on the YFDP likely established during the Little Ice Age (1450–1850), when climate was cooler than the present day (Graumlich, 1993). 4.2.2. Site-specific growth model We have a relatively lower level of confidence in reconstructions based on the site-specific growth model. The site-specific model averages across tree size (predicting relatively faster and slower diameter growth rates for small and large trees, respectively, than the competition-dependent model). Another limitation to the site-specific growth model is our inability to core trees in the

111

smallest and largest diameter classes: very high and very low growth rates are likely underrepresented in the data used to parameterize the site-specific model. Nevertheless, simulations based on the very parsimonious site-specific growth model still resulted in estimates of historical structure that generally agreed with the competition-dependent simulations (Table B.1) and historical empirical datasets (Table 2). Large-diameter tree growth was slower in the site-specific model (Fig. 2) and more of the large trees were predicted to be present in 1900, contributing to the higher predicted basal area (Table B.1). Especially in the case of very old individuals, tree-specific growth rates based on increment cores may perform better than a regionally parameterized growth model. Growth rates derived from tree cores collected on site inherently incorporate the many complex, site-specific aspects of tree growth that are difficult to predict in a modeling framework, such as site and substrate characteristics and climate variations. Furthermore, with fixed growth rates, the decreasing number of trees (and therefore decreasing competition) does not cause growth rates to increase as with the competition-dependent model. 4.2.3. Tree decay model The decay model is probably the large source of error in our estimates of historical forest conditions and is the model component in which we have the lowest confidence. Our decay model predicts that few contemporary snags (mean 334 of 2734 total) and logs (mean 185 of 696 total) were alive in 1900. We believe tree decay in our model to be overall too rapid, especially for sugar pine. Dendrochronological analysis of cross-sections from large-diameter fire-scarred sugar pine snags and logs collected around the YFDP demonstrate that many decay class 3–5 sugar pine snags and logs died when fires were still actively burning the plot or shortly thereafter – indicating a residence time of over a century (Barth, 2014). A bark beetle outbreak in the early 1990s (Guarín and Taylor, 2005) left many standing sugar pine snags, most of which, after about 20 years, are only in the early stages of decay. In our model, the average age of highly decayed class 5 sugar pine snags is about 25 years, which is likely too fast, especially since many of the dead sugar pines were P100 cm dbh at the time of death. Thus, we conclude that our reconstruction model likely underestimates the historical density of large diameter sugar pines, and that this underestimate is specifically attributable to the decay model, not the growth models. Erroneous tree decay predictions could be attributed to a number of factors. Snag transition rates between decay classes and snag to log transition rates for sugar pine and white fir are based on a single study (Morrison and Raphael, 1993) that investigated snag dynamics only over a short time period (10 years). This study likely did not capture the high variation in snag decay class transition and fall rates. Additionally, the rates presented in the study are not diameter-dependent and the mean dbh of snags was 40.6 cm (Morrison and Raphael, 1993), much smaller than many snags present on the YFDP (many of which are >100 cm dbh). In reality, however, smaller trees decay faster than larger trees (Harmon et al., 1986). Additionally, the log decay component is also not size dependent and we had to substitute decay rates (k values) for species that lacked decay data (Table A.1). Although the ages of incense cedar logs and snags are based on data for western redcedar (T. plicata) (Daniels et al., 1997), we believe that incense cedar decay is well represented in our study. Frequent field observations of decay class 2 incense cedar logs on the YFDP with charred bases indicate it is likely these logs were already on the forest floor when fires were still actively burning, rendering our prediction that dead incense cedar trees of decay class 3 would not have been alive in the reference year (1900) plausible.

112

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118

4.2.4. Model variability Incorporating stochasticity in model parameters did not introduce high levels of variability across simulations (Fig. 3), and stand-level metrics remained relatively constant across all simulations for both models (Table B.1), suggesting that presentation of estimated conditions for each model based on mean results across simulations is appropriate. 4.3. Data-driven sources of error In both models, all of the Pacific dogwood trees disappeared before the reference year, but it is highly unlikely that there were no dogwood trees in 1900. A dogwood tree is a group of genetically identical ramets growing up from a central genet. While individual ramets may grow and die on decadal timescales, the genet will persist since it can resprout (Brush, 1948). Since we did not collect data on snags or logs <10 cm in diameter, it is highly likely that we therefore did not detect dead dogwood ramets that would provide evidence for genets persisting over the past century. By modeling the growth of only individual ramets we are likely underestimating historical dogwood abundance. Given that white fir trees and small diameter trees have fast decay rates (Harmon et al., 1986, 1987), it is likely that we were not able to detect the presence of trees that may have been alive in the reference year but have died and decayed substantially since, which would cause us to underestimate historical tree density. This lack of evidence represents a potentially large source of error inherent to forest reconstructions. For example, the low 1930 (149 tph P 10 cm dbh) and 1935 (162 tph P 10 cm dbh) tree densities predicted by the competition-dependent model for the YFDP compared to the 1929 historical empirical data collected in STEF (315 tph P 10 cm dbh) (Knapp et al., 2013) and the 1935 Wieslander survey data from Yosemite (232 tph P 10 cm dbh) (Wieslander, 1935) could be due in part to missing evidence of historical small diameter trees on the YFDP. In 1929 on STEF, there

were 154 small-diameter (10–20 cm dbh) tph; in 1935 on the Wieslander Yosemite plots there were on average 95 small-diameter tph (10–30 cm dbh); many of these small-diameter trees alive around 1930 on the YFDP could have already decayed by 2010. Of course, regardless of possible missing evidence on the YFDP, we must acknowledge that STEF, the Wieslander plots, and the YFDP did not necessarily have similar historical conditions and that discrepancies between YFDP reconstructed estimates and historical datasets does not necessarily signify missing evidence on the YFDP. This example, however, helps illustrate the uncertainties involved when reconstructing forests from extant evidence. 4.4. Evaluation of other reconstruction studies 4.4.1. Decay We identified that the use of a decay model based on integrating available decay data for Sierra Nevada mixed-conifer tree species are not sufficient for accurately modeling tree decay over long periods of time. Reconstruction studies, including this one, over-simplify the tree decay process, and although we understand little about tree decay in Sierra Nevada mixed-conifer forests, we know it is much more complex than current studies imply. More rigorous decay data could help improve reconstructed estimates of historical reference conditions and could also improve other forest modeling projects. For example, careful investigation of the decay components of the commonly used Forest Vegetation Simulator (FVS) model (Crookston and Dixon, 2005) reveals that it too is limited by a lack of relevant decay data. In the Fire and Fuels Extension (FFE) of FVS, snag and log decay in the Sierra Nevada variant is a simplification of decay rates derived from ‘‘some rates for Douglas-fir snags’’ taken from an unpublished study in Oregon (Reinhardt and Croolston, 2003). If outputs from reconstructions and models such as FVS are being used to make management decisions, we need to make improvements to the

Fig. A.1. Representation of the flow of data within the reconstruction model written and implemented in R.

Table A.1 Specific decay data used to ‘‘undecay’’ snags and logs in the forest reconstruction model. ‘‘Forest type’’ and ‘‘target species’’ are where, and for which species, the decay rates were originally derived in the source study. YFDP species Logs White fir

Density data (g/ cm3)

Target species

Sample tree sizes

Forest type

Source

Limitations

Log mineralization rate: k = 0.049 Density by decay classa

DC 1: 60.340, >0.305 DC 2: 60.305, >0.212 DC 3: 60.212, >0.178 DC 4/5: <0.178

White fir

>20 cm diameter

Dry mixed-conifer (CA)

Harmon et al. (1987)

Small sample size (n = 20)

Log mineralization rate: k = 0.024

DC 1: 60.369, >0.269 DC 2: 60.269, >0.221 DC 3: 60.221, >0.113 DC 4/5: <0.113

Ponderosa pine

>23 cm diameter

Dry mixed-conifer (CA)

Dunn (2011)

Western redcedar

Range: 79–250 cm dbh

Coastal rainforest, BC

Daniels et al. (1997)

White oak, eastern black oak, northern red oak combined

Range: 10–80 cm diameter

Deciduous (ID)

MacMillan (1981)

Snag fall rates by decay class

White fir and red fir combined

Mean dbh: 40.6

Sierra Nevada mixed-conifer

Morrison and Raphael (1993)

Base on single 10 year study

Snag fall rates by decay class Snag age estimated by decay class

Jeffery pine, sugar pine, lodgepole pine combined Western redcedar

Mean dbh: 40.6

Sierra Nevada mixed-conifer

Morrison and Raphael (1993) Daniels et al. (1997)

Base on single 10 year study Small sample size (n = 17)

Density by decay class

Incense cedar

Log age estimated by decay class

Black oak

Log mineralization rate: k = 0.0295 Density by decay class

Snags White fir Sugar pine Incense cedar Black oak a

DC 1: 60.611, >0.450 DC 2: 60.450, >0.382 DC 3: 60.382, >0.241 DC 4/5: <0.241

Range: 115–312 cm Southwest coastal rainforest, dbh British Columbia Snag fall rates stochastic: equal likelihood of staying in same decay class or transitioning to earlier decay class.

Small sample size (n = 15)

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118

Sugar pine

Decay data used

Derived from Harmon et al. (2008).

113

114

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118

decay components. Current research by Cousins et al. (2015) is beginning to fill this knowledge gap. A number of studies in the Sierra Nevada (see North et al., 2007; Van de Water and North, 2011), including our study, estimate snag decay rates based on the Morrison and Raphael (1993) transition matrices. A possible result, as demonstrated in our study, is an overestimation of decay rates and obfuscation of variation in decay across different species and diameter classes. Other studies in the region (e.g., Scholl and Taylor, 2010) use decay rates that are further generalized and parametrized for ponderosa pine (Rogers, 1984) and/or are based on decay rates of fire-killed trees (Kimmey, 1955), which decay differently than trees that die in the absence of fire (Harmon et al., 1986). Furthermore, for those studies that did include hardwoods, neither reference any decay data used to model the decay of black oak, the implications of which are a likely misrepresentation of historical black oak populations. The use of the decay class rating system for snags and logs (Thomas et al., 1979) is useful for field surveys, estimates of coarse woody debris amounts, and describing general decay trajectories and biomass loss over time (Grove et al., 2011). It was not, however, developed for assigning a specific calendar year to tree death. Some reconstructions seek to classify trees ages more broadly (e.g., as ‘‘pre’’ or ‘‘post’’ suppression) (Fulé et al., 1997), in which case the use of a decay class system may provide adequate accuracy. While the decay class system is used in most reconstructions, however the validity of this approach is largely untested. Furthermore, employing a simple decay class system to estimate tree ages across the different diameter classes is not accurate, as trees of smaller diameter classes will reach advanced stages of decay more rapidly than larger trees (Vanderwel et al., 2006).

4.4.2. Loss of evidence While some reconstruction studies mention the loss of evidence of historical trees as a potential bias in the results, this missing evidence effect, to the best of our knowledge, has never been quantified until now (Fig. 5). In the YFDP, where mortality is tracked annually, when structural root rots have contributed to the death of white fir, snags up to 20 cm dbh can topple in just a few years (unpublished data). A possible effect of this omission includes underestimating historical tree densities and misrepresenting historical species composition. 4.4.3. Exclusion of hardwoods Many assessments of historical forest conditions in the Sierra Nevada neglect to account for the presence of hardwoods (North et al., 2007; Van de Water and North, 2011; Collins et al., 2011; Lydersen et al., 2013), although angiosperms are indisputably important for biodiversity in mixed-conifer forest ecosystems (Schowalter and Zhang, 2005; Fontaine et al., 2009). A number of rare and threatened wildlife species depend on hardwoods for nesting, foraging, and cover: black oak, in particular, is important habitat for dusky-footed woodrats (Innes et al., 2007), the California fisher (Zielinski et al., 2004; Purcell, 2007), the spotted owl (Irwin et al., 2012), as well as other wildlife in decline (Purcell, 2007). Pacific dogwood trees are important forage for ungulates (Lawrence and Biswell, 1972) and their flowers are attractive to many insects and birds (Michael, 1928). Given the ecological importance of hardwoods, we should not overlook how their population distributions have changed during the fire suppression era and restoration efforts to reintroduce fire should also take into account impacts on populations of trees other than conifers. For example, black oak has been found to be declining

Fig. B.1. Predicted changes to tree density and basal area of the all live trees P1 cm dbh on the Yosemite Forest Dynamics Plot from 1900 to 2010 for the competitiondependent model (left panels) and site-specific model (right panels). Calculations are based on mean values across 100 simulations of each model.

Table B.1 Comparison of the contemporary (2010) and historical (1900) tree populations for the five primary species on the YFDP using two different reconstruction model approaches: the competition-dependent model and the site-specific model. (SD) [range].

All trees 2010 Competition 1900 Site-specific 1900

Site-specific 1900 Sugar pine 2010 Competition 1900 Site-specific 1900 Incense cedar 2010 Competition 1900 Site-specific 1900 Black oak 2010 Competition 1900 Site-specific 1900 Pacific dogwood 2010 Competition 1900 Site-specific 1900

Basal area (m2/ha) Stems P 1 cm dbh

% Total basal area

Density (stems/ha) P10 cm dbh

Density (stems/ha) P100 cm dbh

1391.1 174.8 (1.2) [172.0–178.3] 112.1 (0.75) [109.5–114.1]

64.0 20.9 (0.1) [20.6–21.3] 24.1 (0.16) [23.7–24.6]

– –

535.9 94.2 (0.6) [92.7–95.5] 84.9 (0.49) [83.7–86.1]

19.0 5.9 (0.2) [5.6–6.3] 8.0 (0.24) [7.5–8.6]

985.5 83.8 (1.01) [80.8–86.7] 45.8 (0.52) [44.2–47.0]

29.3 5.9 (0.04) [5.8–6.0] 4.7 (0.05) [4.5–4.8]

45.8 28.2

382.2 36.4 (0.4) [35.3–37.3] 33.7 (0.39) [32.7–34.8]

4.2 1.0 (0.07) [0.8–1.2] 0.2 (0.07) [0.1–0.4]

193.3 65.5 (0.38) [65–67] 47.5 (0.30) [47.0–48.5]

28.7 12.9 (0.09) [12.7–13.2] 17.9 (0.13) [17.5–18.3]

44.8 61.7

84.7 47.5 (0.24) [46.0–48.2] 42.8 (0.25) [42.1–43.3]

13.3 4.0 (0.16) [3.7–4.5] 7.4 (0.22) [6.9–8.0]

61.5 12.8 (0.27) [12.1–13.4] 6.1 (0.17) [5.7–6.5]

4.4 1.8 (0.09) [1.7–2.1] 1.3 (0.08) [1.1–1.4]

6.9 8.6

26.4 6.7 (0.22) [6.1–7.2] 4.8 (0.17) [4.3–5.2]

1.4 0.9 (0.07) [0.8–1.1] 0.4 (0.06) [0.2–0.5]

47.0 11.6 (0.35) [10.7–12.5] 11.8 (0.35) [11.0–12.6]

1.2 0.1 (<0.01) [0.09–0.10] 0.1 (<0.01) [0.09–0.11]

1.9 <1.0

30.0 2.8 (0.17) [2.4–3.1] 2.9 (0.17) [2.5–3.4]

0 0 (0) [0–0] 0 (0) [0–0]

103.1 0 (0) [0–0] 0 (0) [0–0]

0.28 0 (0) [0–0] 0 (0) [0–0]

0.4 0

12.2 0 (0) [0–0] 0 (0) [0–0]

0 0 (0) [0–0] 0 (0) [0–0]

–

19.5

74.3

5.4

<1.0

0

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118

White fir 2010 Competition 1900

Density (stems/ha) Stems P 1 cm dbh

115

116

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118

in Yosemite (Ripple and Beschta, 2008). Pacific dogwood may be particularly threatened due to its susceptibility to the invasive fungal pathogen dogwood anthracnose (Discula destructiva; Brown et al., 1996) as well as its limited genetic diversity, which could result in poor population adaptability as climate changes (Keir et al., 2011). While we do consider the historical presence of black oak and dogwood trees in this reconstruction study, we admit to excluding many other angiosperms from our dataset, such as Scouler’s willow (Salix scouleriana) and chokecherry (Prunus virginiana). 4.5. Management implications Our reconstruction model, using both the competitiondependent and the site-specific growth models, produced estimates of historical conditions in line with other studies of historical forest conditions in the Sierra Nevada. Our reconstructions likely underestimate tree density and diameter in the early 1900s due to overestimated decay rates, particularly for large-diameter sugar pine and Pacific dogwood. We encourage managers to integrate multiple sources of reference information, including reconstruction studies, historical datasets, and contemporary active fire regime sites (Lydersen and North, 2012; Larson et al., 2013a) in a process-based (sensu Stephenson, 1999) framework when designing forest restoration objectives and prescriptions. Restoring forests to a historical condition is not necessarily possible, or desirable. Instead of attempting to re-create forests of the past, we encourage managers to focus on reestablishing key ecological processes, especially active fire regimes. Reconstructed forest conditions at the time of fire suppression, like those we provide here, can help managers better understand past forest dynamics, and develop restoration prescriptions that facilitate resumption of an active fire regime. If reconstructed data is corroborated by other historical data and/or photos, managers might develop prescriptions directly in line with reconstructed estimates, if they feel these conditions will improve forest resiliency and help meet management objectives. But, as a general rule, we encourage managers to focus first and foremost on developing prescriptions based on the ecological processes of interest (e.g., frequent fire). With this in mind, reconstruction models can be judiciously used to learn about past forest dynamics and estimate past conditions. Managers can use uncertainty in historical conditions as an opportunity to test the effects of developing prescriptions from both conservative or liberal interpretations of reconstructed data in an adaptive management framework (Larson et al., 2013b). We encourage forest scientists to provide more nuanced discussion of reconstruction study limitations, in order to enable appropriate use of reference information in development of forest restoration prescriptions (Churchill et al., 2013) and evaluation of treatment effects (Larson et al., 2012). The results presented here are appropriate for understanding the direction and magnitude of forest change and will be useful for setting general targets in restoration planning efforts. The results from our fire history study (Table 1) and reconstruction models (Fig. 6 and Table B.1) demonstrate unequivocally that the YFDP has experienced significant changes as a result of fire-suppression. Our results are part of a growing body of evidence demonstrating that the exclusion of fire from Sierra Nevada mixed-conifer forests has contributed to a dramatic shift in forest structure and composition. The YFDP, similar to other forests throughout the region (Dolanc et al., 2014; Stephens et al., 2015), now hosts a substantially higher density of small-diameter, shade-tolerant trees, especially white-fir, than were present historically. Management actions that reduce tree density, especially that of small-diameter, fire-intolerant species, accompanied by the reintroduction of fire, will likely prepare these forests to adapt

to future climate and disturbance regimes while maintaining important ecological functions and services. Acknowledgements This research was performed under National Park Service research permits YOSE-2011-SCI-0015 and YOSE-2012-SCI-0059. We thank the Smithsonian Center for Tropical Forest Science CTFS-SIGEO Grants Program for funding the fire history portion of this research. We thank J. Meyer and the Yosemite National Park Division of Resources and Science for logistical support and B. Procotor and A. Davenport of the Yosemite Fire Crew for their time in assisting with the collection of cross-sections. D. Wright, E. Sutherland and I. Hyp at the Rocky Mountain Research Station provided advice and technical assistance. This work could not be possible without the help of YFDP volunteers and lab assistant M. Toulas. We thank Adrian Das for advice in developing the tree growth model and Alan Taylor for sharing Big Oak Flat fire years and suggestions for appropriate tree ring chronologies. Robert Logan of the University of Montana and Jeff Braun at Montana Tech provided valuable assistance and access to computational resources. Kip Van De Water provided advice in developing the tree decay model. We thank Solomon Dobrowski, Anna Sala, and three anonymous reviewers for their helpful comments and suggestions for improving this manuscript. Appendix A (see Fig. A.1 and Table A.1). Appendix B (see Fig. B.1 and Table B.1). References Abella, S., Denton, C., 2009. Spatial variation in reference conditions: historical tree density and pattern on a Pinus ponderosa landscape. Can. J. For. Res. 39, 2391– 2403. http://dx.doi.org/10.1139/X09-146. Abella, S.R., Covington, W.W., Fule, P.Z., Lentile, L.B., Meador, A.J.S., Morgan, P., 2007. Past, present, and future old growth in frequent-fire conifer forests of the Western United States. Ecol. Soc. 12, art16. Anderson-Teixeira, K.J. et al., 2015. CTFS-ForestGEO: a worldwide network monitoring forests in an era of global change. Glob. Change Biol. http:// dx.doi.org/10.1111/gcb.12712. Arno, S.F., Scott, J.H., Hartwell, M.G., 1995. Age-class structure of old growth ponderosa pine/Douglas-fir stands and its relationship to fire history. Research Paper INT-RP-481. USDA Forest Service, Intermountain Research Station, Ogden, UT. Astrup, R., Coates, K.D., Hall, E., 2008. Finding the appropriate level of complexity for a simulation model: an example with a forest growth model. For. Ecol. Manage. 256, 1659–1665. http://dx.doi.org/10.1016/j.foreco.2008.07.016. Bakker, J.D., Sánchez Meador, A. J., Huffman, D.W., Moore, M.M., 2008. Growing trees backwards: description of a stand reconstruction model. In: Olberding, S.D., Moore, M.M., (Eds.), Fort Valley Experimental Forest—A Century of Research 1908–2008. USDA Forest Service Proceedings RMRS-P-55, Rocky Mountain Research Station, Fort Collins CO. pp. 106–115. Barth, M.A.F., 2014. Use of a forest reconstruction model to assess changes to Sierra Nevada mixed-conifer forest conditions during the fire suppression era. M.S. Thesis. University of Montana, Missoula, MT. Barth, M.A., Larson, A.J., Lutz, J.A., 2014. CADE tree ring chronologies for Yosemite National Park. International Tree-Ring Data Bank, IGBP PAGES/World Data Center for Paleoclimatology, NOAA/NCDC Paleoclimatology Program, Boulder, CO. Battles, J.J., Robards, T., Das, A., Waring, K., Gilless, J.K., Biging, G., Schurr, F., 2008. Climate change impacts on forest growth and tree mortality: a data-driven modeling study in the mixed-conifer forest of the Sierra Nevada, California. Climatic Change 87, S193–S213. http://dx.doi.org/10.1007/s10584-007-9358-9. Beaty, R.M., Taylor, A.H., 2007. Fire disturbance and forest structure in old-growth mixed conifer forests in the northern Sierra Nevada, California. J. Veg. Sci. 18, 879–890. http://dx.doi.org/10.1111/j.1654-1103.2007.tb02604.x. Brown, D.A., Windham, M.T., Triglano, R.N., 1996. Resistance to dogwood anthracnose among Cornus species. J. Arboric. 22, 83–86.

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118 Brush, W., 1948. Pacific dogwood: Cornus nuttallii, Audubon. Am. Forests 54, 216– 217. Canham, C.D., Papaik, M.J., Uriarte, M., McWilliams, W.H., Jenkins, J.C., Twery, M.J., 2006. Neighborhood analyses of canopy tree competition along environmental gradients in New England forests. Ecol. Appl. 16, 540–554. http://dx.doi.org/ 10.1890/1051-0761(2006) 016[0540:NAOCTC]2.0.CO;2. Cansler, C.A., McKenzie, D., 2014. Climate, fire size, and biophysical setting control fire severity and spatial pattern in the northern Cascade Range, USA. Ecol. Appl. 24, 1037–1056. Churchill, D.J., Larson, A.J., Dahlgreen, M.C., Franklin, J.F., Hessburg, P.F., Lutz, J.A., 2013. Restoring forest resilience: from reference spatial patterns to silvicultural prescriptions and monitoring. For. Ecol. Manage. 291, 442–457. http:// dx.doi.org/10.1016/j.foreco.2012.11.007. Collins, B.M., Everett, R.G., Stephens, S.L., 2011. Impacts of fire exclusion and recent managed fire on forest structure in old growth Sierra Nevada mixed-conifer forests. Ecosphere 2. http://dx.doi.org/10.1890/ES11-00026.1, art51. Cousins, S.J.M., Battles, J.J., Sanders, J.E., York, R.A., 2015. Decay patterns and carbon density of standing dead trees in California mixed conifer forests. Forest Ecol. Manage. 353, 136–147. Crookston, N.L., Dixon, G.E., 2005. The forest vegetation simulator: a review of its structure, content, and applications. Comput. Electron. Agric. 49, 60–80. http:// dx.doi.org/10.1016/j.compag.2005.02.003. Daniels, L., Dobry, J., Klinka, K., Feller, M., 1997. Determining year of death of logs and snags of Thuja plicata in southwestern coastal British Columbia. Can. J. For. Res. 27, 1132–1141. http://dx.doi.org/10.1139/x97-055. Das, A., 2012. The effect of size and competition on tree growth rate in old-growth coniferous forests. Can. J. For. Res. 42, 1983–1995. http://dx.doi.org/10.1139/ x2012-142. Dolanc, C.R., Safford, H.D., Dobrowski, S.Z., Thorne, J.H., 2014. Twentieth century shifts in abundance and composition of vegetation types of the Sierra Nevada, CA, US. Appl. Veg. Sci. 17, 442–455. http://dx.doi.org/10.1111/avsc.12079. Dunn, C.J., 2011. Coarse woody detritus dynamics, variable decay rates and their contribution to wildland fuel succession following high-severity fire disturbance in dry-mixed conifer forests of Oregon’s eastern Cascades. Ph.D Dissertation. Oregon State University, Corvallis, OR. Fontaine, J.B., Donato, D.C., Robinson, W.D., Law, B.E., Kauffman, J.B., 2009. Bird communities following high-severity fire: response to single and repeat fires in a mixed-evergreen forest, Oregon, USA. For. Ecol. Manage. 257, 1496–1504. http://dx.doi.org/10.1016/j.foreco.2008.12.030. Fulé, P.Z., Covington, W.W., Moore, M.M., 1997. Determining reference conditions for ecosystem management of southwestern ponderosa pine forests. Ecol. Appl. 7, 895–908. http://dx.doi.org/10.2307/2269441. Gabrielson, A.T., Larson, A.J., Lutz, J.A., Reardon, J.J., 2012. Biomass and burning characteristics of sugar pine cones. Fire Ecol. 8 (3), 58–70. http://dx.doi.org/ 10.4996/fireecology.0803058. Graumlich, L.J., 1993. A 1000-year record of temperature and precipitation in the Sierra Nevada. Quatern. Res. 39, 249–255. http://dx.doi.org/10.1006/ qres.1993.1029. Grissino-Mayer, H.D., 2001. Evaluating crossdating accuracy: a manual and tutorial for the computer program COFECHA. Tree-Ring Res. 57, 205–221. Grove, S.J., Stamm, L., Wardlaw, T.J., 2011. How well does a log decay-class system capture the ecology of decomposition? A case-study from Tasmanian Eucalyptus obliqua forest. For. Ecol. Manage. 262, 692–700. http://dx.doi.org/10.1016/ j.foreco.2011.05.005. Guarín, A., Taylor, A.H., 2005. Drought triggered tree mortality in mixed conifer forests in Yosemite National Park, California, USA. For. Ecol. Manage. 218, 229– 244. http://dx.doi.org/10.1016/j.foreco.2005.07.014. Hann, D.W., Hanus, M.L., 2002. Enhanced diameter-growth-rate equations for undamaged and damaged trees in southwest Oregon. Oregon State University, College of Forestry, Forest Research Laboratory, Corvallis, OR. Harmon, M.E., Franklin, J.F., Swanson, F.J., Sollins, P., Gregory, S.V., Lattin, J.D., Anderson, N.H., Cline, S.P., Aumen, N., Sedell, J.R.others., 1986. Ecology of coarse woody debris in temperate ecosystems. Adv. Ecol. Res. 15, 302. http:// dx.doi.org/10.1016/S0065-2504(08)60121-X. Harmon, M.E., Cromack Jr., K., Smith, B.G., 1987. Coarse woody debris in mixedconifer forests, Sequoia National Park, California. Can. J. For. Res. 17, 1265–1272. http://dx.doi.org/10.1139/X06-236. Harmon, M.E., Woodall, C.W., Fasth, B., Sexton, J., 2008. Woody detritus density and density reduction factors for tree species in the United States: a synthesis. USDA Forest Service General Technical Report NRS-29, Northern Research Station, Newton Square, PA. Harmon, M.E., Fasth, B., Halpern, C.B., Lutz, J.A., 2015. Uncertainty analysis: an evaluation metric for synthesis science. Ecosphere 6. http://dx.doi.org/10.1890/ ES14-00235.1, art63. Harrod, R.J., McRae, B.H., Hartl, W.E., 1999. Historical stand reconstruction in ponderosa pine forests to guide silvicultural prescriptions. For. Ecol. Manage. 114, 433–446. http://dx.doi.org/10.1016/S0378-1127(98)00373-9. Innes, R.J., Van Vuren, D.H., Kelt, D.A., Johnson, M.L., Wilson, J.A., Stine, P.A., 2007. Habitat associations of dusky-footed woodrats (Neotonia fuscipes) in mixedconifer forest of the northern Sierra Nevada. J. Mammal. 88, 1523–1531. http:// dx.doi.org/10.1644/07-MAMM-A-002R.1. Irwin, L.L., Rock, D.F., Rock, S.C., 2012. Habitat selection by northern spotted owls in mixed-coniferous forests. J. Wildlife Manage. 76, 200–213. http://dx.doi.org/ 10.1002/jwmg.218.

117

Keir, K.R., Bemmels, J.B., Aitken, S.N., 2011. Low genetic diversity, moderate local adaptation, and phylogeographic insights in Cornus nuttallii (Cornaceae). Am. J. Bot. 98, 1327–1336. http://dx.doi.org/10.3732/ajb.1000466. Kimmey, J.W., 1955. Rate of deterioration of fire-killed timber in California. USDA Forest Service, Washington, D.C., USA. King, J.C., Graumlich, L.J., 1990. PILA Tree Ring Chronologies for Yosemite National Park. International Tree-Ring Data Bank, IGBP PAGES/World Data Center for Paleoclimatology. NOAA/NCDC Paleoclimatology Program, Boulder, CO. Knapp, E.E., Skinner, C.N., North, M.P., Estes, B.L., 2013. Long-term overstory and understory change following logging and fire exclusion in a Sierra Nevada mixed-conifer forest. For. Ecol. Manage. 310, 903–914. Larson, A.J., Stover, K.C., Keyes, C.R., 2012. Effects of restoration thinning on spatial heterogeneity in mixed-conifer forest. Can. J. For. Res. 42, 1505–1517. http:// dx.doi.org/10.1139/x2012-100. Larson, A.J., Belote, R.T., Cansler, C.A., Parks, S.A., Dietz, M.S., 2013a. Latent resilience in ponderosa pine forest: effects of resumed frequent fire. Ecol. Appl. 23, 1243– 1249. http://dx.doi.org/10.1890/13-0066.1. Larson, A.J., Belote, R.T., Williamson, M.A., Aplet, G.H., 2013b. Making monitoring count: project design for active adaptive management. J. Forest. 111, 348–356. http://dx.doi.org/10.5849/jof.13-021. Lawrence, G., Biswell, H., 1972. Effect of forest manipulation on deer habitat in giant sequoia. J. Wildlife Manage. 36, 595–605. Lutz, J.A., van Wagtendonk, J.W., Franklin, J., 2009. Twentieth-century decline of large-diameter trees in Yosemite National Park, California, USA. For. Ecol. Manage. 257, 2296–2307. http://dx.doi.org/10.1016/j.foreco.2009.03.009. Lutz, J.A., van Wagtendonk, J.W., Franklin, J.F., 2010. Climatic water deficit, tree species ranges, and climate change in Yosemite National Park. J. Biogeogr. 37, 936–950. http://dx.doi.org/10.1111/j.1365-2699.2009.02268.x. Lutz, J.A., Key, C.H., Kolden, C.A., Kane, J.T., van Wagtendonk, J.W., 2011. Fire frequency, area burned, and severity: a quantitative approach to defining a normal fire year. Fire Ecol. 7 (2), 51–65. http://dx.doi.org/ 10.4996/fireecology.0702051. Lutz, J.A., Larson, A.J., Swanson, M.E., Freund, J.A., 2012. Ecological importance of large-diameter trees in a temperate mixed-conifer forest. PLoS ONE 7, e36131. http://dx.doi.org/10.1371/journal.pone.0036131. Lutz, J.A., Schwindt, K.A., Furniss, T.J., Freund, J.A., Swanson, M.E., Hogan, K.I., Kenagy, G.E., Larson, A.J., 2014. Community composition and allometry of Leucothoe davisiae, Cornus sericea, and Chrysolepis sempervirens. Can. J. For. Res. 44, 677–683. http://dx.doi.org/10.1139/cjfr-2013-0524. Lydersen, J.M., North, M.P., 2012. Topographic variation in structure of mixedconifer forests under an active-fire regime. Ecosystems 15, 1134–1146. http:// dx.doi.org/10.1007/s10021-012-9573-8. Lydersen, J.M., North, M.P., Knapp, E.E., Collins, B.M., 2013. Quantifying spatial patterns of tree groups and gaps in mixed-conifer forests: reference conditions and long-term changes following fire suppression and logging. For. Ecol. Manage. 304, 370–382. http://dx.doi.org/10.1016/j.foreco.2013.05.023. MacMillan, P.C., 1981. Log decomposition in Donaldson’s Woods, Spring Mill State Park, Indiana. Am. Midl. Nat. 106, 335–344. McBride, J.R., 1983. Analysis of tree rings and fire scars to establish fire history. Tree-Ring Bull. 43, 51–67. Michael, C.W., 1928. The pileated woodpecker feeds on berries. Condor 30, 157. Miller, C., Urban, D.L., 1999. Forest pattern, fire, and climatic change in the Sierra Nevada. Ecosystems 2, 76–87. http://dx.doi.org/10.1007/s100219900060. Morrison, M.L., Raphael, M.G., 1993. Modeling the dynamics of snags. Ecol. Appl. 3, 322–330. http://dx.doi.org/10.2307/1941835. North, M., 2012. Managing Sierra Nevada forests. USDA Forest Service General Technical Report PSW-GTR-237, Pacific Southwest Research Station, Albany, CA. North, M., Innes, J., Zald, H., 2007. Comparison of thinning and prescribed fire restoration treatments to Sierran mixed-conifer historic conditions. Can. J. For. Res. 37, 331–342. http://dx.doi.org/10.1139/X06-236. North, M.P., Stine, P., O’Hara, K., Zielinski, W., Stephens, S., 2009. An ecosystem management strategy for Sierran mixed-conifer forests. USDA Forest Service General Technical Report PSW-GTR-220, Pacific Southwest Research Station, Albany, CA. Parsons, D.J., DeBenedetti, S.H., 1979. Impact of fire suppression on a mixed-conifer forest. For. Ecol. Manage. 2, 21–33. http://dx.doi.org/10.1016/03781127(79)90034-3. Purcell, K.L., 2007. Nesting habitat of warbling vireos across an elevational gradient in the southern Sierra Nevada. J. Field Ornithol. 78, 229–239. http://dx.doi.org/ 10.1111/j.1557-9263.2007.00107.x. R Development Core Team, 2014. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria, ISBN 3900051-07-0, http://www.R-project.org/. Reinhardt, E., Crookston, N.L., 2003. The Fire and Fuels Extension to the Forest Vegetation Simulator. USDA Forest Service General Technical Report RMRSGTR-116, Rocky Mountain Research Station, Ogden, UT. Ripple, W.J., Beschta, R.L., 2008. Trophic cascades involving cougar, mule deer, and black oaks in Yosemite National Park. Biol. Conserv. 141, 1249–1256. http:// dx.doi.org/10.1016/j.biocon.2008.02.028. Rogers, J.J., 1984. ECOSIM: a system for projecting multiresource outputs under alternative forest management regimes. Administrative Report. Rocky Mountain Forest and Range Experiment Station, Fort Collins, CO. Rothman, H.K., 2007. Blazing heritage: a history of wildland fire in the National Parks. Oxford University Press.

118

M.A.F. Barth et al. / Forest Ecology and Management 354 (2015) 104–118

Scholl, A.E., Taylor, A.H., 2010. Fire regimes, forest change, and self-organization in an old-growth mixed-conifer forest, Yosemite National Park, USA. Ecol. Appl. 20, 362–380. http://dx.doi.org/10.1890/08-2324.1. Schowalter, T.D., Zhang, Y.L., 2005. Canopy arthropod assemblages in four overstory and three understory plant species in a mixed-conifer old-growth forest in California. Forest Sci. 51, 233–242. Stephens, S.L., Lydersen, J.M., Collins, B.M., Fry, D.L., Meyer, M.D., 2015. Historical and current landscape-scale ponderosa pine and mixed conifer forest structure in the Southern Sierra Nevada. Ecosphere 6. http://dx.doi.org/10.1890/ES1400379.1, art79. Stephenson, N.L., 1999. Reference conditions for giant sequoia forest restoration: structure, process, and precision. Ecol. Appl. 9, 1253–1265. http://dx.doi.org/ 10.1890/1051-0761(1999) 009[1253:RCFGSF]2.0.CO;2. Stephenson, N.L., Parsons, D.J., 1993. A research program for predicting the effects of climatic change on the Sierra Nevada. In: Veirs, S.J., Stohlgren, T.J., SchonewaldCox, C., (Eds.), Proceedings of the 4th Conference on Research in California’s National Parks. US Department of the Interior National Park Service Transactions and Proceedings Series NPS/NRUX/NRTP-93/9, Davis, CA. pp. 93–109. Stokes, M.A., Smiley, T.L., 1968. An Introduction to Tree-Ring Dating. University of Chicago Press. Taylor, A.H., 2004. Identifying forest reference conditions on early cut-over lands, Lake Tahoe Basin, USA. Ecol. Appl. 14, 1903–1920. http://dx.doi.org/10.1890/025257. Taylor, A.H., Vandervlugt, A.M., Maxwell, R.S., Beaty, R.M., Airey, C., Skinner, C.N., 2014. Changes in forest structure, fuels and potential fire behaviour since 1873 in the Lake Tahoe Basin, USA. Appl. Veg. Sci. 17, 17–31.

Thomas, J.W., Anderson, R.J., Maser, C., Bull, E.L., 1979. Chapter 5: snags. In: Thomas, J.W., (Ed.), Wildlife Habitats in Managed Forests: The Blue Mountains of Oregon, USDA Forest Service Agriculture Handbook 533, Washington, DC. pp. 60–77. Van de Water, K., North, M., 2011. Stand structure, fuel loads, and fire behavior in riparian and upland forests, Sierra Nevada Mountains, USA; a comparison of current and reconstructed conditions. For. Ecol. Manage. 262, 215–228. http:// dx.doi.org/10.1016/j.foreco.2011.03.026. van Wagtendonk, J.W., 2007. The history and evolution of wildland fire use. Fire Ecol. 3 (2), 3–17. http://dx.doi.org/10.4996/fireecology.0302003. Vanderwel, M.C., Malcolm, J.R., Smith, S.M., 2006. An integrated model for snag and downed woody debris decay class transitions. For. Ecol. Manage. 234, 48–59. http://dx.doi.org/10.1016/j.foreco.2006.06.020. Walters, D.K., Hann, D.W., 1986. Taper equations for six conifer species in southwest Oregon. Research Bulletin, vol. 56. Forest Research Lab, College of Forestry, Oregon State University, Corvallis, OR. Weigel, D.R., Johnson, P.S., 1997. Estimating DBH of southern Indiana oaks from stump diameter. USDA Forest Service Technical Brief TB-NC-4. North Central Forest Experiment Station, Saint Paul, MN. Wieslander, A.E., 1935. A vegetation type map of California. Madroño 3, 140– 144. Zielinski, W.J., Truex, R.L., Schmidt, G.A., Schlexer, F.V., Schmidt, K.N., Barrett, R.H., 2004. Resting habitat selection by fishers in California. J. Wildlife Manage. 68, 475–492. http://dx.doi.org/10.2193/0022-541X(2004)068[0475:RHSBFI]2.0. CO;2.