Historical and current landscape‐scale ponderosa pine and mixed conifer forest structure in the Southern Sierra Nevada

Many managers today are tasked with restoring forests to mitigate the potential for uncharacteristically severe fire. One challenge to this mandate is the lack of large-scale reference information on forest structure prior to impacts from Euro-American settlement. We used a robust 1911 historical dataset that covers a large geographic extent (>10,000 ha) and has unbiased sampling locations to compare past and current forest conditions for ponderosa pine and mixed conifer forests in the southern Sierra Nevada. The 1911 dataset contained records from 18,052 trees in 378 sampled transects, totaling just over 300 ha in transect area. Forest structure was highly variable in 1911 and shrubs were found in 54% of transects. Total tree basal area ranged from 1 to 60 m2 ha−1 and tree density from 2 to 170 ha−1 (based on trees >30 cm dbh). K-means cluster analysis divided transects into four groups: mixed conifer-high basal area (MC High BA), mixed conifer-average basal area (MC Ave BA), mixed conifer-average basal ...


INTRODUCTION
Throughout much of the drier, low-to midelevation coniferous forests in western North America fires historically burned at intensities that often left mature trees unaffected or scarred by fire, but seldom killed (Allen et al. 2002, Collins andStephens 2007) even when preceded by multi-year drought (Stephens et al. 2008). Mid-elevation forests in the Sierra Nevada, such as those historically dominated by ponderosa pine (Pinus ponderosa), Jeffrey pine (P. jeffreyi ), and other mixed conifer species commonly burned many times per century (Van de Water and Safford 2011) and this produced generally open stand structures (Figs. 1 and 2). These low to moderate intensity fires burned for months at a time and collectively covered very large areas (Skinner and Chang 1996).
While it is clear that fires historically had a strong influence on forest structure at the standscale and vegetation patterns at the landscape scale, there are topographic controls on vegetation, which can operate independent of fire (Show andKotok 1929, Rundel et al. 1977). These topographic controls affect available moisture and temperature patterns, which in combination influence vegetation composition and structure. Elevation has long been noted as a driver of vegetation composition, which in the Sierra Nevada results in greater Pinus dominance in lower elevations of the montane zone and greater Abies dominance in the mid-to upper montane zone (Show and Kotok 1929). However what is actually a stronger predictor of forest composition in the Sierra Nevada is the trade-off between evaporative demand, which is affected by slope aspect and elevation, and water supply, which is affected by soil properties and precipitation patterns (Stephenson 1998).
In the late 19th and early 20th centuries, policies of fire exclusion, introduction of livestock grazing, and elimination of Native American ignitions greatly reduced fire frequencies in ponderosa pine and mixed conifer forests (Stephens et al. 2007a). With these changes many contemporary forests have substantially increased in their vulnerability to uncharacteristically severe fire (Agee and Skinner 2005), exhibited primarily through the creation of large high severity patches. These large patches, where most if not all of the trees are dead, can impede conifer tree regeneration (Barton 2002, Goforth and Minnich 2008, Roccaforte et al. 2012, Collins and Roller 2013, Crotteau et al. 2013, Crotteau et al. 2014) and predispose them for repeated high severity fire and potentially, type conversion (Van Wagtendonk et al. 2012, Collins andSkinner 2014). As a result, many forest managers today are tasked with mitigating the potential for uncharacteristic severe fire (North et al. 2009, Stephens et al. 2013.
The methods used by managers to reduce fire hazards and restore US coniferous forests are prescribed fire, mechanical treatments, and managed wildfire ). US National Park Service policy mandates incorporation of natural disturbances into management plans to maintain parks in an unimpaired state (van Wagtendonk 2012). One obvious challenge to this mandate is the lack of large-scale reference information on forest structure prior to the impacts from Euro-American settlement. Where feasible, fire history data provides historical reference information in the context of the natural range of variability of disturbance regimes (Allen et al. 2002, Reynolds et al. 2013. Based on the records preserved in trees rings (fire scars and tree ages) researchers have reconstructed fire regimes (Kilgore and Taylor 1979, Caprio and Swetnam 1995, Swetnam et al. 2009) and past forest structure (North et al. 2007, Hurteau andNorth 2010) from forests in the southern Sierra Nevada. Additionally, there is one study in giant sequoia (Sequoiadendron giganteum)-mixed conifer forests that reported forest structure based on a set of plots measured in 1900-1901(Stephens and Elliott-Fisk 1998; however, plot locations used in this study were biased and are not representative of southern Sierra Nevada mixed conifer forest as a whole. In many cases these studies were intended to inform managers and policy makers of the historical or natural range of variability operating on the landscape which can assist in the development of restoration plans. One of the biggest limitations of forest structure reconstructions based on dendrochronology is their limited spatial extents. In addition, treering based reconstructions are subject to an inherent limitation brought about by only using extant data. When there is a considerable amount of time between the reconstruction period and data collection, as there is with most historical forest reconstructions in the western US (.100 years), uncertainty increases owing to sample losses from fire, insects, disease, and decomposition (Collins et al. 2011).
Large historical datasets, collected from the field and archived, are an alternative source of information used in forest reconstructions (e.g., Leiberg 1902, Wieslander et al. 1933). These datasets sometimes allow for detailed quantitative comparisons of current versus historical forest structure and composition (Lutz et al. 2009). However, there are a number of concerns associated with historical datasets: (1) limited geographic extent, (2) unknown or unrepeatable study site selection and inventory methodolo-gies, and (3) limited temporal depth. Finding a historical dataset that addresses all of these concerns would be extremely difficult, if not impossible (Collins et al. 2011, Hagmann et al. 2013. That stated, we have identified a historical dataset that covers a large geographic extent (.10,000 ha) and has unbiased sampling locations. This particular dataset consists of early timber inventories conducted by the US Forest Service (USFS) and includes robust tree measurements over large areas, which contrasts with the reconstruction of forest structure from very low density tree measurements from General Land Office records (e.g., Baker 2014). In California, these inventories were conducted ca. 1910 and were part of the first organized assessment of all timber resources within the then, new agency v www.esajournals.org (Collins et al. 2011).
In this study we take advantage of a portion of these early USFS timber inventories that were conducted on what was then the Kern National Forest (now part of the Sequoia National Forest). A comparison of current forest conditions to those based on the early inventories can provide information on potential changes driven primarily by 20th century fire exclusion policies and harvesting. Such comparisons can provide information to managers on how close current forest structure and composition is to a time when natural processes operated on the landscape. Our objectives were to: (1) summarize historical timber inventory records, (2) identify distinct forest types based on forest structure and species composition, and (3) compare past and current forest conditions for ponderosa pine and mixed conifer forests at the landscape scale in the southern Sierra Nevada.

STUDY AREA
The forest sampled in 1911 was in the Greenhorn Mountains of the southern Sierra Nevada. This area is characterized as west-slope Sierra Nevada ponderosa pine and mixed conifer forests (Figs. 1 and 2) consisting of ponderosa pine, sugar pine (Pinus lambertiana), white fir (Abies concolor), incense-cedar (Calocedrus decurrens), California black oak (Quercus kelloggii ), and canyon live oak (Q. chrysolepis). The climate is Mediterranean with cool, wet winters and hot, dry summers. Annual precipitation ranges from 9 to 74 cm year À1 , with an average of 32 cm year À1 ; mean monthly temperatures range from 0.98C in January to 20.38C in July (data from the Piutes Remote Automated Weather Station, 2005Station, -2011. The study site covers 11,500 ha and topography varies within this area. Elevation in the study area ranges between 1430 and 2270 m. The southern portion of the area is of lower elevation  while the northern area is higher . Soils in this area are predominately coarse-loamy well-drained dystric xerochrepts (under taxonomic order of inceptosols). Prior to the late 1800s fire was common in this region, with a mean point fire return interval of approximately 5-20 years (Kilgore and Taylor Fig. 2. Historic conditions in a ponderosa pine stand on the Sierra National Forest (ca. 1917;Sierra National Forest Photo HP03137).
v www.esajournals.org 1979, Caprio and Swetnam 1995). With the onset of fire exclusion fire was largely removed from the study area.
Early land use information from the Greenhorn Mountains is provided by records from the Sequoia National Forest (USDA 2013;Jim Whitfield, Ecosystem Staff Officer, Sequoia National Forest, personal communication, 2014). Small portions of the Greenhorn Mountains were logged using only wagons and horses in the late 19th and early 20th centuries. There was no railroad logging in this area. The nearest record of railroad logging is on the Hume Lake Ranger District which is approximately 75 km northwest of the study site (Otter 1963, Johnston 1997. Most of the earliest logged areas in the Greenhorn Mountains were located near mining camps around Lake Isabella that are not included in the 1911 transects. Other areas logged prior to 1911 were located at lower elevations that were more accessible by early roads and were used to construct houses in the growing town of Bakersfield, California. The historic inventory records collected for this study included 15 survey transects that mentioned prior cutting; these were excluded from the study.

Historical forest inventory data
All of the historical forest inventories within our study area were conducted in 1911. The inventories were conducted by two-person crews, with one person compass and distancing and the other assessing the forest. Their assessments involved recording all live conifer trees !30.5 cm (12 inches) in diameter at breast height (dbh) within belt transects that were 20.1 m (66 feet or 1 chain) wide and 402 m (1320 feet or 20 chains) long. Trees were tallied by species into dbh classes of 5.1 cm (2 inches). Recording of smaller trees (,30.5 cm dbh) was inconsistent across transects. In addition to inventorying trees the following observations were made in each transect: rock type and exposure, soil texture, ground cover, and shrub cover. General notes on forest condition, including insect depredation, fire damage, stand development stage, and the spatial distribution of the trees were also recorded.
Transect locations were based systematically on the Public Land Survey System (PLSS), which was the primary method used to survey rural or undeveloped land in the western US. Transects started and ended at the mid-points of quarterquarter sections, which we refer to as lots, with one belt transect sampled in each 15.7 ha (40 acres) lot for the area surveyed. Transects were placed as strips across a section (an area comprised of 16 lots) and generally oriented either N-S or E-W, against dominant contour within a section. Some transects were placed for 10 chains in a N-S direction followed by 10 chains in an E-W direction allowing surveyors to connect strips within a section. Each transect covered 0.9 ha (2 acres).
We assembled a suite of variables to characterize forest structure and composition using tree lists generated from the 1911 belt transect data. The tree lists were standardized for area sampled such that outputs represented per hectare values. The constructed variables were: tree density for all trees and partitioned into three dbh classes, proportion of stand basal area by three species groups, and total basal area. Basal area is the total cross-sectional area of all live tree stems in a stand (similar basal area can be produced by a few large trees or many small trees). The three dbh classes used to report partitioned tree density were: 30.5-61.0 cm, 61.1-91.4 cm, .91.4 cm, and were based on a stand classification scheme used in forest management throughout the Sierra Nevada (USDA 2004) and in previous work (Lutz et al. 2009, Collins et al. 2011. The species groups that we investigated for potential changes in proportion of stand basal area were: (1) pine-ponderosa pine and sugar pine, (2) true fir-white fir and red fir (Abies magnifica), and (3) incense-cedar.
In addition to recording information on forest structure and site characteristics, during the surveys the field crew took detailed notes on vegetation conditions along transects (see Appendix for a copy of all notes recorded during this inventory); these records were used to estimate the incidence of high severity fire. Surveyors typically noted where they entered and exited patches of contrasting vegetation, such as chaparral or dense ''immature'' timber, but did not record explicit descriptions of fire severity in terms of damage to timber; we assumed these areas were burned by high v www.esajournals.org severity fire. We used the length of these discrete forest patches along survey transects to estimate the proportion of high severity fire on the landscape (average proportion of transect in high severity patch multiplied by percentage of total transects with evidence suggesting high severity fire).
The presence of chaparral in these forests can be both a successional change brought about by high severity fire (Kauffman and Martin 1991) or it can be a product of edaphic conditions that limit tree establishment/growth (e.g., dry sites with shallow soils) (Nagel and Taylor 2005). Given the uncertainty as to what was driving the observed chaparral, we bracketed our estimates of high severity fire. We noted transects in which observed chaparral may have been environmentally and/or edaphically controlled (see Appendix). These included transects at the edge of the timber survey area, which corresponded with the extent of the timber belt, as well as transects that had distinct topographic features, such as a ridge tops or steep canyons. Our bracketed high severity fire estimates included and excluded these potentially environmentally and/or edaphically controlled transects. In all cases the high severity patches were smaller than the length of the transect, or did not result in complete mortality across a transect, because all transects contained trees .61 cm dbh.

Current forest inventory data
The Forest Inventory and Analysis (FIA) program provides a nationwide inventory of forestlands in the US and is executed by the USFS. The FIA program delivers inventory and monitoring data related to forest structure, composition, and health, based on a set of systematically randomized plots across the country. Standard FIA plot densities are one plot per 2428 ha (6000 ac). FIA plots (673 m 2 ) are divided into four 7.3 m (24 ft) radius circular subplots, with one central subplot and three peripheral subplots arranged at 120-degree angles from plot center. Peripheral subplots are situated at a distance of 36.6 m (120 ft) from the subplot center to the plot center. For all subplots, every tree .12.7 cm (5 in) is measured (dbh, height) and identified to species. Each subplot also contains a 2.1 m (6.8 ft) radius circular microplot for the measurement of tree regeneration (all trees ,12.7 cm dbh, seedlings and saplings). More information pertaining to FIA protocol can be found in Bechtold and Patterson (2005), Woudenberg et al. (2011), and the FIA program website (http://www.fia.fs.fed.us/). All FIA plot data used here was based on the most recent FIA protocol.

Data analysis
Vegetation in the 1911 transects was summarized by basal area, stem density of tallied trees, and shrub cover. Shrubs, as they were denoted by 1911 surveyors, included mountain misery (Chamaebatia foliolosa), manzanita (Arctostaphylos spp.), black oak, canyon live oak, scrub oak (Q. berberidifolia, or shrubby form of Q. kelloggii or Q. chrysolepis), chaparral, chinquapin (Chrysolepis sempervirens), willow (Salix spp.), mountain mahogany (Cercocarpus betuloides), wild plum (Prunus spp.), and Sierra gooseberry (Ribes roezlii ). We converted qualitative assessments of cover, assigning a cover of 10%, 20%, 50% or 90% for each species based on the description provided by the surveyor (e.g., sparse, scattered, moderate, and dense, respectively). It is worth noting that for the entire 1911 area included in our analysis there were only two sets of observers; therefore, a reasonable assumption of consistency in the written descriptions across the survey area.
Topographic variables were derived using a digital elevation model (Gesch et al. 2002, Gesch 2007. Raster datasets for slope, aspect, and curvature were generated using Spatial Analyst tools in ArcMap 10.1. Aspect was classified so that values ranged from 0 (xeric) to 20 (mesic) (Parker 1982). Pixels were categorized by four topographic position index (TPI) classes (valley bottom, gentle slope, steep slope, ridgetop) using the CorridorDesigner toolbox (Majka et al. 2007). The default settings of the ''create topographic position raster'' tool were used so that pixel elevation was assessed relative to neighboring pixels within 200 m, with those being lower or higher by 12 m compared to the neighborhood average classified as valley bottom or ridgetop, respectively. The cutoff point for gentle versus steep slope was 68 (10.5%).
A raster of topographic relative moisture index (TRMI) was calculated using TPI, aspect, slope, and curvature following the method of Parker v www.esajournals.org (1982). Actual evapotranspiration (AET) and annual climatic water deficit were calculated using soil water holding potential, temperature, and precipitation (Churchill et al. 2013). We used the soil water holding capacity in the top 150 cm of the soil (Soil Survey Staff 2013) and weather data from the period 1901-1910, obtained from Climate WNA (Wang et al. 2011). All topographic variables were calculated at a 10 m resolution except AET and water deficit, which were calculated at a 30 m resolution then resampled to a 10 m pixel size to match that of the other variables. Mean transect values for continuous topographic variables were extracted from 10-m raster layers using zonal statistics. For TPI, proportion of transect area in each class was calculated, and a TPI class was assigned to each transect based on the majority area in each class.
Transects were classified into groups using kmeans cluster analysis (Hartigan and Wong 1979). Clustering was based on the following overstory and understory structure variables: cover of mountain misery, total cover of all other shrub species, basal area by tree species (yellow pine [ponderosa and Jeffrey pine], sugar pine, incense-cedar, white fir, and total), and stem density by size class (30.4-61, 61.1-91.4, .91.4 cm dbh and total). Variables were z-score standardized prior to clustering to account for differences in scale (McCune and Grace 2002). To select the optimum number of clusters we compared results for up to a 12 cluster solution to 250 runs done using randomized input data. Graphs of the within-group sum of squared error (SSE) and the absolute difference in SSE from the average SSE of the randomized data were examined. The cluster solution that maximized the difference between sum of squared error for the chosen cluster solution and the average of the randomized runs was selected (Peeples 2011).
To investigate whether identified vegetation groups were associated with different topographic characteristics we used a two-stage analysis. First, we used a random forest analysis to identify which, if any, of the constructed topographic variables explained the occurrence of the vegetation groups. This was performed in R version 3.0.2, using the conditional inference method ''cforest'' (Hothorn et al. 2006). This method constructs a suite of conditional inference trees, using a randomly selected subset of predictor variables and a random subsample of plots for each tree. Examining a large number of trees allows for identification and ranking of influential variables, and averages out the instability of individual regression trees that can exhibit large changes in structure due to random variation in the data (Strobl et al. 2009). Variable importance was assessed using the conditional method developed by Strobl et al. (2008). The second part of this analysis was to compare values for the top three topographic variables, as indicated by the variable importance rankings from the random forest analysis, among the vegetation groups using an ANOVA model that accounted for unequal variances. When ANOVA models were significant (a ¼ 0.05) a Games-Howell comparison was used to compare means among vegetation group (Games and Howell 1976). This was performed using ''proc MIXED'' in SAS version 9.4.
Following the cluster analysis, the Forest Vegetation Simulator (FVS; Dixon 2002) was used to estimate canopy cover for the historical inventory transects. FVS uses tree species and size to generate crown radii for all trees in a given belt transect. FVS then calculates the percentage of ground area directly covered with tree crowns, correcting for canopy overlap (Dixon 2002). To investigate how well FVS canopy cover estimates approximate field-based observations in old-growth forests, we compared field-measured canopy cover from an intact Jeffrey pine-mixed forest in the Sierra San Pedro Mártir, Mexico (Stephens and Gill 2005), to FVSestimated canopy cover for the same plots. This was done for 25 0.1-ha circular plots, which were established on a regular 125-m grid. Within each plot the following information was collected for each tree .1 cm dbh: status (live/dead), species, and dbh (to nearest 0.1 cm). Canopy cover was assessed with GRS densitometer based on a 5 3 5 grid, with approximately 3 m between grid points. The number of points that intersected tree crowns was multiplied by four to obtain percent canopy cover. The trees recorded on each plot were entered into FVS to generate modeled estimates of canopy cover for comparison to field based measurements.
FIA plot data was obtained from all west-side mixed conifer and ponderosa pine forests on the Sequoia and Sierra National Forests (data colv www.esajournals.org lected from 2001 to 2008) to compare to the 1911 forest structure data. The same forest structure variables were computed for the FIA data as described above for the 1911 historical data. Note that we only included trees .30.5 cm dbh from the FIA data to match the minimum diameter used for the 1911 data. We tested for differences in forest structure variables among time periods (1911 against recent FIA) using Welch's varianceweighted one-way ANOVA, which does not require an underlying assumption of homogeneity of variances (Welch 1951). This analysis was done with Proc GLM in SAS version 9.4.

RESULTS
Overall the 1911 dataset contained 18,052 trees in 378 transects, totaling just over 300 ha in transect area. Forest structure was highly variable in this forest. Total tree basal area ranged from 1 to 60 m 2 ha À1 and tree density ranged from 2 to 170 ha À1 (based on trees .30 cm dbh). On average the basal area was comprised of 25% white fir, 29% incense-cedar, 8% sugar pine, and 38% ponderosa pine. For each species, the proportion of transects in which it was absent were: 18% (white fir), 19% (incense-cedar), 35% (sugar pine) and 8% (ponderosa pine). Stems 30-60 cm dbh were about twice as abundant on average as those in the 60-90 and .90 cm size classes (Table 1). Shrubs (all species including mountain misery) were found in 54.4% of transects.
K-means cluster analysis divided transects into four groups (Table 1). The first group, mixed conifer-high basal area (MC High BA) was characterized by high total tree basal area and stem density, with high basal area of white fir, incense-cedar, and sugar pine but low ponderosa pine basal area and density. This group accounted for 15% of the total survey area. White fir and incense-cedar were the dominant species in this group and mountain misery was entirely absent. The second group was mixed conifer-average basal area (MC Ave BA) and had values close to the overall averages for all forest structure variables in the sampled landscape, and accounted for 34% of the total area. Incense-cedar had the highest tree basal area, but white fir and ponderosa pine were also co-dominant (based on tree basal area). The third group, mixed coniferaverage basal area-high shrubs (MC Ave BA Shrubs) had very high cover of mountain misery and other shrubs, above average density of trees ,91.4 cm dbh, and high basal area of pine species, particularly ponderosa pine. This group accounted for 10% of the 1911 survey area. The two species with the greatest overall tree basal area in this group were ponderosa pine and incense-cedar. The last group, which accounted for the largest percentage of area (41%), was dominated by ponderosa pine (Pond Pine) and had low conifer density, and tended to have lower than average tree basal area of all species except ponderosa pine. K-means groups tended to be spatially clustered within the study area  . 3); the southwestern portion of the study area was predominantly in the Pond Pine group. Density of trees .91 cm dbh was the highest in MC High BA and basal area was approximately 42 m 2 ha À1 . The two MC classes with average basal areas (MC Ave BA, MC Ave BA Shrubs) differed by understory shrub cover (26% versus 76%) and the dominance of mountain misery (5% versus 62%), respectively. Interestingly the average mixed conifer areas with high shrubs were primarily located in the northern and northeastern portions of the study area (Fig. 3). The shrub dominated areas had a higher proportion of ponderosa pine that was co-dominated by incense-cedar; the areas with lower shrub cover were dominated incense-cedar with a co-dominance of white fir (Table 1). Tree density in the 30-61 cm dbh class was twice as large in the MC Ave BA Shrub group when compared to the MC Ave BA group even though average total basal areas were very similar.
The three highest ranking variables from the random forest analysis were AET, Elevation, and Aspect (Appendix: Fig. A1). ANOVA results indicated significant differences among groups for all three variables (Table 2). For all three v www.esajournals.org topographic variables, Pond Pine was different from the other three groups, having higher AET, lower elevation, and more southwesterly aspects (average transformed value of 7.6, corresponding to 1408 and 2678) ( Table 2). MC High BA occurred on higher elevations and northeasterly aspects (average transformed value of 11.8, corresponding to 1068 and 3058). MC Ave BA Shrubs occurred on sites with the lowest AET, but had similar elevations and aspects to those for the other two MC groups ( Table 2).
The 1911 surveyor's notes indicated that low fire damage was the most common observation in these forests followed by moderate fire damage (Table 3). Collectively these two categories covered approximately 92% of all three mixed conifer structural groups and 76% of the Pond Pine structural group (Pond Pine also had 17% of plots with no recent fire recorded) (Table  3). High fire damage, including that inferred by the presence of chaparral or immature timber, was recorded on very few transects (Table 3). The percentage of this landscape that experienced high severity fire in the last decade or two (or Table 2. Topographic conditions (means 6 SD) in each K-means group for the 1911 Kern National Forest data. Letters denote Games-Howell significance groupings between K-means groups (P , 0.05). AET, actual evapotranspiration. Statistical comparisons for aspect were done prior to back-transforming values. Transformed values were 7.6 6 3.4, 9.9 6 4.2, 108 6 4.5, and 11.8 6 3.7 for Pond Pine, MC Ave, MC Shrubs, and MC High, respectively.  (Table 3). Tree regeneration was recorded as ''none'', ''scarce'', or ''low'' in approximately 40% of the Pond Pine, MC High BA, and MC Ave BA structural groups; this increased to 92% in the MC Ave BA Shrubs group. Conversely, approximately 20% of the MC Ave BA and Pond Pine groups had ''high'' conifer tree regeneration (Table 3). Oaks were present in many plots, with fewer in the MC Ave BA Shrubs and MC High BA structural groups (Table 3). Tree distributions within the belt transects were diverse with approximately one-third of MC Ave BA, MC Ave BA Shrubs, and Pond Pine structural groups having a ''scattered'' distribution (Table 3). Pond Pine forests were noted for their openness and MC High BA areas for their high density (41% of transects). Trees were recorded as ''distributed in groups'' in 26% of the MC Ave BA Shrubs structural group with MC High BA having the fewest groups (2.6%) (Table 3). Approximately 20% of the transects in MC High BA, MC Ave BA, and Pond Pine structural groups had a ''patchy'' forest distribution with ''uniform'' forests most common in the MC High BA group (23%); in the other forests groups ''uniform'' forests were relatively rare (Table 3).
The average difference between SSPM plotmeasured and FVS-estimated canopy cover is reported in Table 5. Both the average overall difference, which includes positive and negative values, and the average absolute difference, which only includes absolute values of the differences are reported. Average canopy cover estimates projected by FVS were slightly below that measured in reference forests (24.8% vs. 26.1%, Table 5). However, a paired t-test on the two canopy cover estimates indicated no significant difference (P ¼ 0.57) suggesting that our FVS estimates of canopy cover are accurate. It is worth noting that both the plot-based and FVSprojected canopy cover estimates agree with a previously reported canopy cover estimate (25.3%) from a larger sampling area in the same forest of northern Baja California, Mexico, which was based on ten 100 m canopy line intercept transects (Stephens and Gill 2005).
Tree density increased markedly between 1911 and the recent sampling effort (FIA plots), especially in ponderosa pine forests (430% increase) (Table 6). This increase was highly  (Table 6). Total tree basal area in mixed conifer forests was unchanged between 1911 and the present (approximately 30 m 2 ha À1 ) but increased by approximately 50% in ponderosa pine forests (Table 6). Sugar pine basal area remained relatively constant in mixed conifer forests whereas incense-cedar basal area was two times greater in 1911 as compared to today. White fir basal area in mixed conifer forests increased greatly in the recent measurement versus 1911. Average forest canopy cover increased from 25% to 49% in mixed conifer forests and from 12% to 49% in ponderosa pine forests; canopy cover in current forests is similar between types versus in 1911 when mixed conifer forests had twice the canopy cover as ponderosa pine forests.

DISCUSSION
The southern Sierra Nevada ponderosa pine and mixed conifer forests sampled in 1911 had low tree densities but there was important variation by structural groups (Table 1). The Pond Pine forest group tended to be located in the lower elevation portion of the study area and had the lowest tree densities and basal areas (25 trees ha À1 and 11 m 2 ha À1 , respectively). Ponderosa pine made up 58% of tree basal area in 1911 followed by approximately equal amounts of white fir and incense-cedar (18% each). Shrub cover was important in this group and averaged 14% with less than one-tenth being mountain misery.
In Jeffrey pine dominated mixed conifer forests in the Sierra San Pedro Mártir, Mexico, which have not experienced wide-spread fire suppression or harvesting, average tree basal area was 20 m 2 ha À1 (Stephens and Gill 2005) and this is larger than that found in 1911 ponderosa pine forests in the southern Sierra Nevada. Jeffrey pine dominated the SSPM forests contributing 67% of basal area followed by smaller amounts of white fir and sugar pine (23% and 8% of basal area, respectively). In ponderosa pine forests in northern Arizona, Fulé et al. (2009) found that ponderosa pine dominated stand basal area (64%) in 1870 but contributed only 36% in the same forest in 2003 after fire exclusion and harvesting. Stoddard (2011) did an extensive review of historical forest information from the Southwestern US (Arizona, New Mexico, and small portions of Utah and Colorado). Information is given on pre Euro-American forest conditions v www.esajournals.org from ponderosa pine and mixed conifer forests from studies that varied in spatial scale (1-1000 ha) and reference date . Selecting ponderosa pine studies that provided information from larger areas (.80 ha) and estimated structural characteristics before the major impacts of Euro-Americans (before 1880) resulted in a basal area 10.4 m 2 ha À1 which is similar to the southern Sierra Nevada forests studied here. Stoddard (2011) also summarized pre Euro-American mixed conifer forest basal area of 16.2 m 2 ha À1 , which is lower than all three mixed conifer structural groups in the southern Sierra Nevada (Table 1), probably reflecting lower ecosystem productivity in the Southwestern US. Reynolds et al. (2013) provided another recent review of historical forest conditions from the Southwestern US and report ponderosa pine and mixed conifer tree density and basal area of 30-315 trees ha À1 and 5-20 m 2 ha À1 , and 53-251 trees ha À1 and 9-28 m 2 ha À1 , respectively. The values from Reynolds et al. (2013) overlap those found in our 1911 data, but the basal areas from our 1911 estimates in the southern Sierra Nevada are at the high end of their distributions, especially for mixed conifer forests. The description of forest conditions in 1911 support previous assertions on the critical role of fire in these ecosystems (Kilgore and Taylor 1979, Skinner and Chang 1996, Scholl and Taylor 2010. Past fire evidence was clearly seen in approximately 95% of all mixed conifer structural groups and on 80% of ponderosa pine areas (Table 3). The vast majority of fire evidence was described as low to moderate damage (Table 3) with low percentages of high severity fire on this landscape (1-6%) ( Table 4). Another piece of evidence of landscape-scale dominance of lowmoderate severity fire is the ubiquity of large trees in all of the sampled transects (Table 1). This indicates that these trees persisted for a long period of time (200-400 years) without experiencing a stand-replacing disturbance. Comparing forest inventory data from 1911 to the present (Table 6) indicates that current forests have changed greatly, particularly in tree density, canopy cover, density of large trees, dominance of white fir in mixed conifer forests, and the similarity of tree basal area in contemporary ponderosa pine and mixed conifer forests. The vast majority of published studies on historic mixed-conifer forests document lower tree densities and a more open structure comprised of a higher proportion of old and large trees that were more spatially heterogeneous (having gaps and patches of trees) and more uneven-aged compared to current conditions (Fulé et al. 2002, 2009, Moore et al. 2004, Stephens and Gill 2005, North et al. 2009, Collins et al. 2011. Topography also influenced the historical distribution of vegetation in our study area. Ponderosa pine forests were associated with lower elevation, more southwesterly aspects, and greater values of AET. The influence of elevation on vegetation type, with ponderosa pine forests occurring below mixed conifer forests, has long been recognized (Show and Kotok 1929) and the association of greater pine dominance with more southerly aspects has also been observed (Fites-Kaufman et al. 2007, Lydersen andNorth 2012). However the association with AET was unexpected, as others have found this variable to be associated with greater productivity (Stephenson 1998;Kane et al., in press). The southern portion of the study area, which was predominantly ponderosa pine forest ( Fig. 3) had higher water holding capacity, leading to greater values of AET. While water holding capacity and AET were relatively lower in the rest of the study area, perhaps the trees may be able to access deeper water stores in weathered bedrock (Meyer et al. 2007) allowing these areas to sustain more productive forest types.
One important artifact is the scale at which the 1911 data were recorded (0.9 ha transect area) such that any fine-scale patterning cannot be discerned (Hagmann et al. 2013). The majority of the variability in structure in frequent-fire forests has been observed at spatial scales smaller than 0.4 ha Churchill 2012, Fry et al. 2014), and topographic characteristics that influence forest structure may also be observed at a smaller scale (Lydersen and North 2012). The scale at which the 1911 inventory data were recorded homogenizes this patchiness, which has been shown to include widely spaced individuals, clusters of large trees, dense patches of regeneration, and small openings (North et al. 2002, Franklin and Van Pelt 2004, Larson and Churchill 2012, Lydersen et al. 2013, Fry et al. 2014). This fine-scale patchiness, and the scale at v www.esajournals.org which topography affects forest patterns, are important characteristics that warrant further investigation.
The fact that the southern Sierra forests we studied were more open historically relative to current conditions is not surprising given the number of previous studies that have already demonstrated this trend (e.g., Debenedetti 1979, Scholl andTaylor 2010). What is more surprising is just how open these forests were in 1911, with average canopy cover being 25% and 12% in mixed-conifer and ponderosa pine types, respectively (Table 6). Fornwalt et al. (2002) modeled ponderosa pine reference canopy cover conditions of 13-22% on the Colorado Front Range and canopy cover in unmanaged Jeffrey pine-mixed conifer forests in northern Baja California, Mexico, was 27% (Stephens et al. 2007b). In addition, other canopy cover studies including White (1985), Covington and Sackett (1986), and Covington et al. (1997) reported 21.9%, 19.0%, and 17.3% canopy cover for ponderosa pine reference conditions on the Fort Valley Experimental Forest, Arizona, respectively (Reynolds et al. 2013); these values are similar to canopy cover estimates from southern Sierra Nevada ponderosa pine forests in 1911 (Table 1).
The 1911 forest inventory data depict a forest with high heterogeneity (Tables 1 and 3). As such, the historical data we present do not support the idea of basing management goals for restoration and forest resilience treatments on average values (Stephens andGill 2005, North et al. 2009). Perhaps the ranges documented in this work in canopy cover and tree densities by size classes can assist in the creation of restoration plans to increase forest resilience (Collins et al. 2011). The second point related to the historical distributions found here is that common restoration goals of mixed conifer and ponderosa pine forests, particularly canopy cover and tree density, are on the upper end of or entirely exceed the values we report based on the 1911 data (Table 1). Consequently, forest restoration goals in southern Sierra Nevada mixed conifer and ponderosa pine forests are probably misaligned with the historic range of variation in stand structure. We note that our analysis is at the landscape scale, which may not capture finer scale heterogeneity brought about by openings and denser forest patches, however our estimates of landscape conditions are robust.

Relevance to dry forest management
The 1911 forest inventory data describes a forest that is a product of a range of fire effects over a long period of time (i.e., a functioning fire regime), as well as the underlying topographic/ moisture availability influence. The changes in forest structure and composition relative to 1911 indicate severely altered present forest conditions. The present forest in the western portions of Sequoia and Sierra National Forests is characterized by much higher overall tree densities, shifted species dominance from pine to fir (especially in mixed conifer forests), lower density of large trees, and higher canopy cover. These findings are consistent with those from previous studies in the Sierra Nevada (Parsons and Debenedetti 1979, Ansley and Battles 1998, Scholl and Taylor 2010, Collins et al. 2011, Dolanc et al. 2014. Periods of frequent fire in mixed-conifer and pine-dominated forests gave fire-resistant species a competitive advantage, allowing them to establish dominance (Stephens et al. 2008). During ''fire-free'' or less frequent-fire periods, pines persisted due to their dominant positions in the forest canopy (Fulé et al. 2009). However, extensive fire-free periods, such as that associated with fire exclusion, coupled with grazing, selective logging, and favorable climatic conditions for young tree establishment in the early 20th century has created atypical stand compositions and structures in many of today's Sierra Nevada ponderosa pine and mixed-conifer forests. In many locations, large, dominant ponderosa pine trees have been significantly reduced leaving today's stands dominated by small trees. This has also been documented in other studies in mixed-conifer and ponderosa pine forests in the western US (Swetnam and Baisan 2003, Moore et al. 2004, North et al. 2009).
Recently a new procedure has been used to estimate past forest structure based on the establishment of a public land survey system (PLSS) by the General Land Office (GLO), the advantage of this dataset is it covers most of the western US. Baker (2012Baker ( , 2014 used this system (GLO), which used eight trees per section (259 ha) that were marked to assist in the relocation of survey section corners, to reconstruct historical forest conditions in eastern Oregon and the Sierra Nevada. Four townships (or 144 sections) in eastern Oregon (Baker 2012) overlap an area that used similar historical transect data analyzed in this work (Hagmann et al. 2013). GLO survey data collected in 1866-1895 would include a record of approximately 1152 trees marking section and quarter section corners in this four township area while the historical timber transect inventory included 163,558 trees on 1355 transects (Hagmann et al. 2013). More recently Baker (2014) used similar GLO procedures to estimate past forest structure and fire severity in the Sierra Nevada. A comparison of the results from the GLO procedures to four different study sites from northern Oregon to the southern Sierra Nevada demonstrates that the GLO methods overestimate historic tree density (Table 7). One challenge in the comparison of these results is possible differences in lower diameter limits in these studies. The transect studies (Collins et al. 2011, Hagmann et al. 2013, 2014, this work) all have explicit lower diameter limits whereas the GLO reconstructions do not include this infor-mation (Table 7). One study that also compared these two methods determined that the large differences in tree densities between the direct inventory and GLO could not be reconciled by possible differences in diameter limits of the two datasets (Hagmann et al. 2013). Another limitation of all historical datasets is they are from one brief time period which characterizes past fire regimes with limited temporal depth.
Our estimate of past mixed conifer and ponderosa pine high severity fire (1-6%), which are similar to estimates from other work (Stephens et al. 2007a, Mallek et al. 2013, are much lower than those reported by Baker (2014) for the Sierra Nevada (31-39%). Perhaps information from contemporary restored fire regimes, for which fire severity patterns can be quantified more robustly, can provide insight. Working in mesic upper-elevation mixed conifer forests in the Sierra Nevada that have been burned repeatedly by lighting fires, Collins and Stephens (2010) found that high severity fire accounted for 15% of the total burned area in recent large fires. These fires occurred 30 years after the start of the v www.esajournals.org natural fire program in this area, suggesting a relatively intact fire regime. However, despite having an intact modern fire regime, this area previously experienced nearly 100 years of fire exclusion, which corresponded with a large increase in tree establishment (Collins and Stephens 2007). It is reasonable to assume this large pulse of tree establishment led to higher levels of severe fire than occurred historically. Furthermore, the higher elevation and associated shifts in species composition (high dominance of red/white fir and lodgepole pine [Pinus contorta sub. murrayana]) may also contribute to the greater proportion of high severity fire reported by Collins and Stephens (2010). Collectively, these points suggest that 15% high severity fire is likely an overestimate for historical high severity proportions in ponderosa pine-mixed conifer forests, indicating that our 1-6% estimates are likely closer to actual conditions than the 31-39% posed by Baker (2014).
Changing climates are already warming temperatures in the Sierra Nevada so a specific goal to recreate past conditions is not advisable ). However, the information from this work could be used to inform the production of desired landscape conditions because the forests sampled in 1911 were highly resilient to many of the same environmental processes affecting today's forest ecosystems, including insects, diseases, and fire. Increasing forest resiliency is the most common goal that forest mangers desire in a world of changing climates (Fulé 2008) and this work could be used to inform this goal.  A1. Importance ranking of the topographic variables based on their association with vegetation class of historical survey transects, as determined by random forest analysis. Variables included in the analysis were actual evapotranspiration (AET), elevation, transformed aspect, water deficit, topographic relative moisture index (TRMI), slope steepness, and proportion of transect area in each of the four topographic position index categories (% Steep Slope, % Ridgetop, % Valley Bottom and % Gentle Slope). The three variables that appear in Table 2 (AET, elevation, aspect) are the three most highly ranked variables.
v www.esajournals.org Timber is very uneven aged and very much scattered more or less singly. Good clear length to the pine trees (yel þ sug) of 16 0 of trees over 24 00 in D.
Stand is very open with a little dead timber standing and an understory of sparsely scattered scrub black oak and with a forest floor of very dense bear clover. Insect depredation very prevalent.
none n/a n/a low insect none scattered

MC Shrub
Timber is very uneven aged; it is more or less uniformly distributed thru the forty.
Stand is very open but the forest floor is not thickly covered with underbrush to prevent reprodbut very little reprod present. The clear length to the yel and sug p exceeds 16 0 on ave. and on a few trees there 2 and 3 16 0 lengths. Considerable fire damage but not injuring the butts of the nearly mature trees much. Much insect depredation as many dead trees standing and fallen trees. low n/a n/a none insect scarce uniform

MC Shrub
Timber is very uneven aged and more or less widely scattered in small groups. The stand is very openreproduction being poor and forest floor of bear clover thicker than it has been. Clear length to the yel and sug p about 24 00 and over 16 0 good. Timber very slightly fire scarred. There has been much insect depredation present as much timber dead and yet standing. low n/a n/a none insect low groups

MC Shrub
Timber is very uneven aged, there being clear length to the sug and yel p of 24 00 and over of 16 0 . The stand is very open and more or less uniformly distributed accept in the NE corner where there seems to be nothing but manzanita and chaparral and boulders. A bad fire been thru but it has only injured the cedar trees badly. Also insect depredation, shown by the dead standing trees and the many stagheaded trees. moderate n/a n/a none insect scarce uniform v www.esajournals.org low n/a n/a low none scarce groups

MC Shrub
The timber is very uneven aged but distributed uniformly thru the forty; clear length to the nearly mature pine is 16 0 and over. Has been a bad fire but it has not done much damage to the yel or sug p. insect depredation present -many dead standing trees and few fallen trees. Forty very open with forest floor of thick clumps of chaparral and bear clover -there no reproduction. low n/a n/a none none scarce uniform

MC Shrub
Timber is very uneven aged and more or less scattered in groups. Clear length to the yel and sug p 16 0 and over (for nearly mature trees 24 00 and over). Timber been fire scarred but not very badly except the cedar. Insects have been present as there are many dead trees standing and many fallen. The stand is very open and the reproduction seems to get hold better but only in very few places. A very little immature oak scattered. low n/a n/a low insect scarce groups 25320508 MC Shrub first few chains the timber more or less uniformly distributed but at top of ridge about 6 ch come to practically nothing but young scrub oak which follows us about 10 chains (with a little scattering coniferous trees). The pine about 24 00 has clear length of 16 0 . Timber is slightly fire scarred and insect depredation present as usual -few dead trees standing. Get fair mature timber thru first few ch and then oak for few ch and then young timber.
high chaparral (along ridge) 10 moderate insect none groups

MC Shrub
Timber is not of especially good development, especially the younger trees as many are wind topped and stagheaded. Very uneven aged stand and widely scattered, singly and in groups. Clear length to the older pine (yel and sug) good for 16 0 . Much fire damage to trees but not affecting the butts far up the trunk. Much insect depredation present; dead trees standing -all species. Reprod confined to latter part of forty low n/a n/a low (understory) insect low scattered

MC Shrub
Timber is very uneven aged there being clear length to the older sug and yel p of 16 0 and more. Timber has been badly fire scarred but only affecting a few very badly, the pine and cedar mostly. Much dead standing timber present -proof of insect depredation. The stand is very open and more or less the trees scattered in groups. But even then there very little reprod even though the forest floor is not densely covered with brush. low n/a n/a none insect scarce groups v www.esajournals.org Timber is very uneven aged and is very evenly distributed thru the forty. Clear length to the yel. p and sug p. about 16 0 . all timber has fire scars marked on it but in general they are not bad. Insect depredation present as many stagheaded trees present; also dead trees standing. We pass 2 seeps one at 7 chains and another about at the end of the forty. low n/a n/a none insect none uniform v www.esajournals.org Timber is in fairly good condition and nearly mature. Clear length to the Yel. P. 16 0 . All timber is badly fire scarred, pine and cedar worse than the fir. There is very little insect depredation. Much oak only along the first part of the forty and then disappear -very good young reprod. The timber is more or less scattered in groups. moderate n/a n/a moderate insect high groups

MC Shrub
Timber is very widely scattered thru the first part of the forty but is in fairly dense groups thru the last few chains of the forty. Though the pine is very badly fire scarred would say that there is clear length of 16 0 . All timber is more or less stunted and stagheaded and also insects have killed a few trees -yet standing. There is little to no reproduction the forty being very open and exposed. Timber is mostly middle aged and has no clear length. It is more or less scattered in groups with very immature black and cañon live oak as understory. There has been much damage by fire and also insect depredation has been present as there are many trees stagheaded and dead. The country is very scattered with coniferous timber with oak as a thick understory. low n/a n/a high insect none groups

MC Ave
Timber for first few chains fairly dense but not uniform in size. After passing across creek get timber much more scattered thru boulders and of uniform size mostly. There a clear length to the pine of 16 0 and over. Considerable damage by fire to the butts of the nearly mature timber. Insect depredation present thru the last 10-14 chains as many stagheaded trees and many already dead. low n/a n/a none insect none patchy v www.esajournals.org The timber is more or less scattered in groups leaving the ground cover very much exposed to the sun's rays. To the nearly mature pine timber may get 16 0 of clear length but most of the timber except in the draws are very much stunted in growth or wind topped or killed or nearly so by insects (as there be many trees that are dead and standing). In general very poor timber thru forty. none n/a n/a low (understory) insect none groups

MC Shrub
Timber is widely scattered and in bad shape. There is no clear length. All timber is very much stunted in growth, stagheaded and wind topped. And some is very badly fire scarred. There is no clear length. There is quite a little cañon live oak but it is very small and scrubby. Boulder outcrops are very abundant all thru the forty. The country is very very open, sun having easy access to soil. none n/a n/a moderate none none scattered

MC Ave
Timber is very much scattered in groups more or less. There is good clear length to the nearly mature sug and yel p. timber thru the last part of the forty is gets stunted in growth (even the young timber) and is more or less wind topped. Also thru the last part of forty there must be insect depredation present as the pine trees -young ones are getting stagheaded and dying. Only the nearly mature pine has clear length of 16 0 and over. The nearly mature timber is scattered in groups. Thru first 10 ch the old timber ought to be cut and the younger growth thinned out. Also the ground floor needs to be cleaned up of dead wood. The last 10 ch consist mostly of dense young oak in clumps. It quite immature. Thru first 10 ch very little fire or insect damage to trees. low n/a n/a high insect high groups v www.esajournals.org Nearly mature timber is more or less uniformly distributed thru forty. As there is not much young reproduct. This nearly mature timber ought to be heavily cut and the ground cleared up. The timber is not very badly damaged by fire except the real large fir and sug p. trees. Very little insect disease apparent as all the timber is in good shape. All ground practically shaded by the moderately dense stand. low n/a n/a none insect scarce uniform v www.esajournals.org Timber is very scattered thru first 10 ch. and then get a little denser stand for 2-3 ch. and then into a mess of very young and immature black oak. Clear length to the pine is a good 16 0 length. A good deal of this pine is nearly mature. The nearly mature trees show rather bad fire scar but they do not extend very far up from the base of the trunks. There is a little timber blown over by the wind and a few standing dead trees killed probably by insects. Timber is more or less scattered being found both singly and in groups. There is no clear length on the average. Considerable fire damage noted on the butts and trunks of the nearly mature timber tallied. Much moss found growing on the bark and dead boughs of the timber. There is a little oak scattered thru forty but most of it is toward the end of the forty and is very young and immature. There a few dead stubs found standing.
low n/a n/a low none moderate scattered

MC Ave
Timber is very much scattered but most of it is near maturity; no clear length. Very much damaged by fire and considerable moss found growing on the bark and limbs of the trees. A little young and immature oak found scattered thru the forty.
moderate n/a n/a low moss moderate scattered

MC Ave
Timber is very much scattered and is nearly mature. There has been a fire thru here but has not done a great deal of damage. No clear length. Much moss found growing on the bark and limbs of the timber tallied. Timber is very very poor as many standing dead butts. low n/a n/a none moss moderate scattered

MC Ave
Timber is yet quite immature and is very much scattered; it not of good quality along fist 12 chains but after ridge passed and descend east slope the young as well as the nearly mature timber has fine form. Timber in good condition to be cut as it is nearly mature. There is more clear length on the pine than the fir as it tends to cling to lower branches even when of large size. Best use saw logs. There a good deal of sugar pine here. Much damage done to all mature timber by fire as the scarred butts are present but to no great height. low n/a n/a none none low excluded v www.esajournals.org Country pretty open so the trees seem to hold their lower branches. There considerable pine reproduction about 6 00 -12 00 in D and also fir. Best use for timber is saw logs. Virgin stand. Very much is the nearly mature timber damaged by fire, it being clearly marked on the butts and up to a distance (in some cases) to 15 0 -20 0 . Also the fir tree bothered with moss growing on the bark. Very heavy growth of fir of a good and nearly mature age. Clear length about 16 0 log. Best use for the timber saw logs; virgin stand. Very much damaged by fire. Great amt of moss growing on the bark, more so than anywhere else found. There a few scraggling oak and pine trees present. Thus a great amt of litter has accumulated. low n/a n/a low moss low dense v www.esajournals.org The nearly mature fir, cedar and pine has fair good form but the cedar is very badly fire scarred. The pine fairly so but damage not so far up the trunk. The fir has hollow butts but most has grown over. Best use for the timber is saw logs. Where indicated, oak prevalence was assessed based on its presence recorded in notes on transect understory.