Succession of Woody Plants After Large, Severe Forest Fires in Arizona and New Mexico

THE SOUTHWESTERN NATURALIST 53(2):146–161 JUNE 2008

INHABITANTS OF LANDSCAPE SCARS: SUCCESSION OF WOODY PLANTS AFTER LARGE, SEVERE FOREST FIRES IN ARIZONA AND NEW MEXICO

SANDRA L. HAIRE* AND KEVIN MCGARIGAL

Department of Natural Resources Conservation, University of Massachusetts, Amherst, MA 01003 *Correspondent: [email protected]

ABSTRACT—Understanding consequences of changes in climate and fire regimes for succession in plant communities is critical for conservation planning at broad spatial and temporal scales. We selected two sites that burned in high-severity fire decades ago and studied succession in the woody plant community and its variations across two environmental gradients; elevation and distance from a lower- severity/unburned edge. By overlaying an ordination of data for woody species on the modeled environmental gradient most closely related to variation in communities, we analyzed the interaction of life-history traits of species and landscape heterogeneity at each study site. Species that resprout from surviving roots were widespread across the distance gradient 28 years after the La Mesa fire in New Mexico. Species that reproduce from off-site seed, including Pinus ponderosa, were more prevalent where resprouters (e.g., Quercus) were less important in defining communities. At Saddle Mountain, Arizona, 45 years post-fire, we observed neighborhood interactions across the elevation gradient, for example, where shade-tolerant conifers (e.g., Abies concolor) occurred in understories of Populus tremuloides.At both sites, greater cover of woody plants that reproduce from off-site seed at shorter distances from a lower-severity/unburned edge suggested migration of these species following the model of wave-form succession. In contrast to studies that emphasize undesirable effects when forest transitions to openings and alternative habitats, our research elucidates the need for further consideration of both young forest communities and the persistent species and communities described as landscape scars in conservation plans for forest systems of the southwestern United States.

RESUMEN—Entender el papel que juegan los cambios de clima y de re´gimen de incendios en la sucesioń de comunidades de plantas es crucial para planear la conservacioń a grandes escalas temporales y espaciales. Seleccionamos dos sitios que sufrieron severos incendios hace dećadas y estudiamos los patrones de sucesioń de sus comunidades de plantas lenõsas y sus variantes a lo largo de dos gradientes ambientales: elevacioń y distancia al borde de un parche con baja incidencia de incendios o no quemado. Al superponer los resultados de la ordenacioń de los datos de especies lenõsas sobre el gradiente ambiental del modelo ma´s relacionado con la variacioń en las comunidades, analizamos la interaccioń de las caracterı´sticas de la historia de vida de las especies con la heterogeneidad del paisaje en cada sitio de estudio. Veintiocho anõs despue´s del incendio de La Mesa en Nuevo Me´xico, las especies que rebrotan de raıćes supervivientes se distribuyeron a lo largo de todo el gradiente de distancia. Las especies que se reproducen por semillas procedentes de sitios exo´genos, incluyendo Pinus ponderosa, fueron ma´s prevalecientes allı´ donde las que rebrotan (por ejemplo, Quercus) tendieron a ser menos decisivas en la definicioń de las comunidades. En Saddle Mountain, Arizona, 45 anõs despue´s del incendio, observamos interacciones entre vecinos a lo largo del gradiente de elevacioń. Por ejemplo, conı´feras tolerantes a la sombra (como Abies concolor) se hallaron en el sotobosque de Populus tremuloides. En ambos sitios, la mayor cobertura de plantas lenõsas reproducidas por semillas exo´genas a poca distancia de bordes con parches con baja incidencia de incendios o no quemados, sugiere que la migracioń de dichas especies siguiera el modelo de sucesioń ‘‘forma de ola.’’ Al contrario de estudios que enfatizan los efectos no deseados de la transicioń de bosque a a´reas abiertas y ha´bitats alternativos, nuestra investigacioń establece la necesidad de considerar conjuntamente las comunidades de bosque temprano al igual que las especies y comunidades persistentes descritas como cicatrices del paisaje en los planes de conservacioń de los sistemas de bosques del sudoeste de los Estados Unidos. June 2008 Haire and McGarigal—Succession of plants after forest fires 147

The mosaic of species and communities on and time frames. A better understanding of the forested landscapes is influenced greatly by large complexities of vegetational response to large, fires (Allen et al., 1998; Johnson et al., 1998; severe disturbance is needed (Clark, 1993; Turner and Dale, 1998; Morgan et al., 2003). In Turner et al., 1998) to evaluate the potential recent years, characteristics of large fires in parts importance of post-fire plant communities to of the southwestern United States differ from conservation efforts. those documented by historical evidence in that The purpose of our research was to better effects of fire are more severe over a greater understand plant succession after severe fire proportion of the area burned (Swetnam and events in the southwestern United States, given Baisan, 1996; Schoennagel et al., 2004). Conser- the possibility that these landscapes occupy an vationists have expressed concern regarding the important place in long-term variability of uncertain consequences of recent large fires for ecosystems. To that end, we selected two sites biodiversity in general and Pinus ponderosa that burned in high-severity fire decades ago and specifically (P. D. McCarthy and S. L. Yanoff, in studied succession of woody plant communities litt.). Conversion of P. ponderosa to persistent, and its variations across two environmental nonforested grass or shrub communities has gradients; elevation and distance from potential been discussed as an undesirable outcome of seed source, or lower-severity/unburned edge. severe fire in forests of P. ponderosa (Savage and In this way, we examined how successional Mast, 2005; Strom and Fule´, 2007). These outcomes varied given different pre-fire commu- nonforest communities, or landscape scars, can nities and in relation to heterogeneity of remain from decades to centuries because seeds landscapes created by the fire. of P. ponderosa have relatively short dispersal The influence of post-fire spatial heterogene- distances, and establishment of seedlings is ity on succession after severe fires (often called sporadic (Allen et al., 2002). stand-replacing, or crown fires) has been docu- Recent evidence suggests, however, that over mented in northern Rocky Mountain coniferous long time frames large, severe fires occurred forests (e.g., Turner et al., 1997, 1998). Crown historically in some forests of P. ponderosa in the fires burn with variable intensity (e.g., Van southwestern United States. Several fire-related, Wagner, 1983), producing a remarkably hetero- debris-flow deposits dating to the late Holocene geneous mosaic of burn severities, or ecological recorded high-severity fires in P. ponderosa-dom- effects, on the landscape (Christensen et al., inated, mixed-conifer forests of the Sacramento 1989). The spatial heterogeneity of burn severity Mountains, New Mexico (Pierce et al., 2004; J. D. leads to different successional changes, depend- Frechette and G. A. Meyer, in litt.). In that case, ing on life-history strategies (Turner and forests of P. ponderosa persisted in a fire regime Romme, 1994; Turner et al., 1997; Frelich, that included fires that were large and severe, 2002). We defined functional life-history groups and plant communities emerging after high- of species as follows: 1) species that primarily severity fire represent an important stage in sprout from surviving roots and root collars, 2) variability of ecosystems related to climate species that reproduce primarily by on-site seed change. that is cached and sometimes require scarifica- On the other hand, if present and future tion, and 3) species that reproduce exclusively by climates differ in important ways from those seed dispersed from off-site via wind or animals. experienced in the past (Millar and Woolfenden, The following expectations were based on 1999; Miller, 2003), it is possible that changes studies in forest systems that have been charac- observed on contemporary landscapes may rep- terized by large, severe fires. First, we expected resent unique communities outside of our that rate of establishment of plants that depend current understanding of variability in forest on seed dispersal, including P. ponderosa, would systems. Both models (McKenzie et al., 2004) be influenced by the spatial distribution of and empirical data (Westerling et al., 2006) surviving individuals (Turner and Romme, suggest that large, severe fires throughout the 1994; Frelich, 2002). In contrast, plants that Rocky Mountains region are directly related to resprout from residual living tissue (e.g., Quercus effects of changing climate. All of this poses a or Populus tremuloides) or that reproduce from difficult conundrum for those concerned with scarified seed (e.g., Arctostaphylos) can be distrib- conservation of natural systems over large areas uted widely across gradients of distance and 148 The Southwestern Naturalist vol. 53, no. 2 elevation because they survive regardless of Forest, and the Los Alamos National Laboratory, a position relative to source of seed (Frelich, United States Department of Energy facility. Because both fires burned in areas with active post-fire 2002). Second, neighborhood effects can be management programs, salvage logging, seeding of positive or negative depending on the species non-native grasses and planting of seedlings of P. involved (Frelich and Reich, 1995; Ponge et al., ponderosa occurred in some places (C. Allen and D. 1998). We expected to observe regeneration of Steffensen, pers. comm.). Subsequent wildland and prescribed fires have burned in portions of both study Pinus edulis shaded by other plants (Floyd, 1982), sites (National Park Service, unpubl. data; United and shade-tolerant conifer species, including Abies States Forest Service, unpubl. data). concolor, benefitting from quickly resprouting P. Data Collection and Analysis—For La Mesa, we tremuloides (Brown and DeByle, 1987; Keyser et al., obtained aerial photos taken in 1973, 1975, 1977, 1981, 2005). Conversely, abundant resprouting shrubs and 1983 that covered portions or all of the burn. Photos varied in scale from 1:600 to 1:24,000. Photos (e.g., Quercus) might limit resources for off-site obtained of the Saddle Mountain burn were taken in seeders, including P. ponderosa. 1957 and 1963 at a scale of 1:15,840. Using the pre-fire and post-fire aerial photos, we mapped areas that had MATERIALS AND METHODS—Study Areas—We select- changed from forest to non-forest corresponding to ed two study sites where large fire events resulted in fire perimeters available from the United States Forest extensive areas of dead trees: the 1960 Saddle Service and the National Park Service. We did not Mountain fire on the Kaibab Plateau in northern distinguish mortality of trees that was a direct result of Arizona and the La Mesa fire of 1977 on the Pajarito crown fire from mortality caused by surface-fire, or Plateau in northern New Mexico. The fires occurred trees that initially survived the fire but later succumbed across an elevational gradient that encompassed major to damage or disease. Therefore, areas where all trees community types, including P. edulis-Juniperus wood- were killed were labeled as high-severity, and areas of land, P. ponderosa, and mixed-conifer (P. ponderosa- surviving trees were labeled as lower-severity/unburned Pseudotsuga menziesii-A. concolor-P. tremuloides) forest. because they could have experienced low, moderate, or mixed effects of fire, or could represent unburned The study sites represented diverse conditions in the islands within the perimeter. The minimum mapping geographic range of P. ponderosa; thus, providing highly unit for areas of surviving trees within high-severity variable landscapes in which we expected to observe a patches was two live trees. We used ArcInfo version 9.0 diversity of successional outcomes following high- (Environmental Systems Research Institute, Redlands, severity fire. California) for spatial analyses. Both sites were located on regional plateaus with We sampled 68 plots at La Mesa and 79 plots at markedly different geologic histories. Ash-flow tuffs, Saddle Mountain between 16 May and 30 June 2005. erupted from the Jemez Mountains, define the Pajarito Locations were chosen within high-severity patches at Plateau; its alternating broad mesas and steep canyons greatest distances from lower-severity/unburned edge drain eastward to White Rock Canyon of the Rio across the full range of sizes and shapes of high-severity Grande (Reneau and McDonald, 1996). In contrast, patches. We distributed the sample plots throughout the Kaibab Plateau was formed from sedimentary rock burns to the greatest extent possible (Figs. 1 and 2). layers deposited with shifts in sea level (Hopkins, However, access was limited in some places by Los 1990). The topography creates dramatic relief; steep Alamos National Laboratory, and we also avoided scarp slopes, or combs, are adjacent to narrow stream locations where tree planting was documented (e.g., bottoms, with sheer walls on the south forming the on the Mesa del Rito at the La Mesa burn). Plots were Nankoweap Rim. circular with 25-m radius (0.2 ha). Climate patterns were similar in both areas, with We conducted point-intercept surveys at each plot fluctuations at a decadal scale influenced by the El along two 50-m transects positioned North-South and Nin˜o Southern Oscillation (Swetnam and Betancourt, East-West. At each 1-m mark, we recorded all woody 1998). Annually, frequent, strong thunderstorms occur species present above the meter-mark on the transect during July through early September (Woodmencey, tape, i.e., that would be intercepted by a vertical pin 2001). Storms in winter (December–March) also bring extending up from the ground. Cover for each species moisture, but in lesser amounts than received during was derived by summing its frequency for a given the summer monsoon. Snow accumulates in winter at sample plot; therefore, total cover per plot for a elevations .1,500 m; below-freezing, overnight lows particular species could exceed 100. In the case of occur throughout winter and are possible in any season species that tend to hybridize (Quercus), or if similar- (Woodmencey, 2001). ities led to difficulty in identification to the species Land use and management at the two sites have level (Arctostaphylos [Arsp1], Artemisia [Arsp2], Gutierre- included many influences. These areas are traditional zia, Rubus, Sambucus, and Symphoricarpos), data were and current homelands to many Native American presented at the genus level (Table 1). We also peoples. The Saddle Mountain burn is entirely within recorded additional woody species present outside of the Saddle Mountain Wilderness on the Kaibab the line transects but within the plot. National Forest, except for a small area in Grand We verified origin of forest opening (i.e., the high- Canyon National Park. La Mesa falls under several severity patch) by presence of downed wood, stumps, jurisdictions including Bandelier National Monument or snags. Also, we field-checked maps for accurate and the Dome Wilderness, the Santa Fe National representation of surviving trees, and modified maps in June 2008 Haire and McGarigal—Succession of plants after forest fires 149

FIG. 1—Locations of 68 field plots at the La Mesa burn, New Mexico. High-severity patches, mapped from aerial photos, are shown in gray, with black dots representing the center of the 25-m radius field plots. The Pinus ponderosa-mixed-conifer zone is at the northwest, with a general decrease in elevation and associated change in forest types toward the southeast. Forest Service lands are on the western side, and Los Alamos National Laboratory properties are at the northern edge of the burn. The central and eastern portions of the burn are located in Bandelier National Monument and the Dome Wilderness. a few cases upon our return from the field. Photos were and Saddle Mountain, respectively). The final config- taken from center of plots in the four cardinal uration was rotated so that the variance of points was directions. maximized on the first dimension. We determined that To determine how species form communities, and the rotation allowed minimal loss of information when describe relationships among functional groups and multi-dimensional solutions are displayed in two- the environment, we first conducted a plant-commu- dimensional plots. nity ordination using non-metric multidimensional Then, we chose a single environmental-variable scaling. We deleted uncommon species (i.e., occurring model that explained the greatest amount of deviance at ,5% of plots) to emphasize major compositional in ordination for each site using generalized additive patterns in the dataset. Data were square-root trans- models. The models are driven by characteristics of the formed and Wisconsin double-standardization was used data, rather than parametric classes so that generalized to equalize emphasis among sample units and among additive models can take any smooth shape (Yee and species. To derive the species matrix of dissimilarities, Mitchell, 1991; Oksanen and Minchin, 2002). The we used the Bray-Curtis algorithm (Bray and Curtis, regression surface is the sum of the smooth functions 1957). We chose a three-dimensional model based on for each variable. Selection of smoothing parameters goodness of fit, evaluated with a stress statistic that was accomplished through minimizing a generalized reflected the linear or non-linear correlation between cross-validation score that incorporated multiple pen- original dissimilarities and ordination distances alties because the model attempts to accommodate (Clarke, 1993; stress 5 13.2 and 11.6, for La Mesa every variation in the data (Wood, 2000). 150 The Southwestern Naturalist vol. 53, no. 2

FIG. 2—Locations of 79 field plots at the Saddle Mountain burn, Arizona. High-severity patches, mapped from aerial photos, are shown in gray, with black dots representing the center of the 25-m radius field plots. Elevation decreases from the southwest Populus tremuloides-mixed-conifer zone along the boundary of Grand Canyon National Park to the northeast (Pinus edulis-Juniperus woodland zone). Most of the burn is within the Saddle Mountain Wilderness, Kaibab National Forest.

The environmental variable chosen was different at on the distance gradient. At Saddle Mountain, the the two sites. At La Mesa, the generalized additive generalized additive model with elevation as the model with distance as the dependent variable ex- dependent variable explained a greater proportion of plained a greater proportion of variation in the data deviance (adjusted r2 5 0.85) compared to the distance (adjusted r2 5 0.42), compared to elevation (adjusted model (adjusted r2 5 0.19), and so we superimposed r2 5 0.25). Therefore, we superimposed the ordination the ordination on the elevation gradient. We used the June 2008 Haire and McGarigal—Succession of plants after forest fires 151

R statistical package (R Development Core Team, of distance; R. neomexicanus, for example, was http://www.R-project.org) for all statistical analyses, fairly ubiquitous, occurring at locations up to ca. the R-vegan library functions for ordination using non- metric multidimensional scaling ( J. Oksanen, http:// 225 m. Pinus ponderosa was recorded at a maxi- cc.oulu.fi/,jarioksa), and the R-mgcv library for mum distance of ca. 222 m. Ribes viscosissimum, generalized additive models (S. N. Wood, http://www. Quercus, and C. fendleri were at peak cover levels maths.bath.ac.uk/,sw283/simon/mgcv.html). at intermediate distances. Quercus was present Defining functional groups in relation to reproduc- across the range of distances from 10 to ca. tive strategies is challenging because most plants demonstrate more than one strategy depending on 300 m. Juniperus monosperma, Cercocarpus monta- conditions at a specific time and place. For our analysis, nus, Artemisia (Arsp2), and Gutierrezia were we used several references that described life-history located optimally at the highest modeled dis- characteristics (Epple, 1995; Foxx and Hoard, 1995; tances of any species. Distribution of these Ecological Restoration Institute, 2004; United States species was fairly dispersed, compared to that Forest Service, http://www.fs.fed.us/database/feis/), chose the most-likely strategy based on available of species with greatest cover at the shortest information, and assigned a functional group accord- distances. ingly (Table 1). We enhanced interpretation of the Species that reproduce primarily by resprout- arrangement of species, functional groups, and com- ing played an important role in defining munities in ordination space using two strategies. First, we displayed a four-letter abbreviation of each species ecologically similar locations at La Mesa centered at the location of its peak cover (i.e., (Fig. 4). Greatest variation in cover for resprou- weighted-average scores); nearby locations had less ters occurred at shorter distances from center of cover and farther distances in the ordination space sample plot to edge of lower-severity/unburned, contain low levels or no cover of the species. This and sample plots with greater cover of resprou- enabled an overview of distributions of species along the modeled environmental gradients in the ordina- ters clustered in one area of the ordination tion space (Figs. 3 and 5). Second, we used gray circles space. Sample sites located at the edge of the scaled in size relative to the total cover of species in distance gradient, at both small and large each functional group to symbolize each sample site. In distances, tended to have relatively sparse cover this way, we examined the distribution of functional of resprouters. Robinia neomexicanus, Quercus, F. groups as they varied in cover across the environmental gradient and in terms of the role of each group in paradoxa, and R. woodsii were major contributors defining similarities among sample sites (Figs. 4 and to cover of this functional group. Populus 6). tremuloides occurred at one sample plot at La Mesa. RESULTS—We identified 52 species of native The off-site seeders at La Mesa made a marked trees and shrubs in sample plots at the study contribution to similarities in sample sites at areas (Table 1). Several species were abundant locations where resprouters tended to be less in terms of both cover and frequency of prevalent (Fig. 4). In ordination space, distribu- occurrence across sample plots at both study tion of sample sites with maximum cover of off- sites (e.g., Quercus, Robinia neomexicanus). Others site seeders was separate from the cluster of sites were observed frequently only at La Mesa (e.g., containing the greatest cover of resprouters. In P. ponderosa, Artemisia [Arsp2], Rosa woodsii)or addition, sites with more abundant cover of off- Saddle Mountain (A. concolor, Arctostaphylos site seeders were located in two distinct areas of [Arsp1], Berberis [Mahonia] repens, P. tremuloides, ordination space; one at the low end of the Symphoricarpos rotundifolius). Species observed distance gradient and the other at intermediate infrequently with low mean total cover at either distances. At the smaller distances, P. ponderosa study site included Ceanothus greggii, Garrya was the greatest contributor to cover of off-site flavescens, Holodiscus dumosus, Prunus virginiana, seeders; Artemisia and Gutierrezia had greater and Pinus flexilis. cover at intermediate distances in this group. La Mesa—For La Mesa, one species of tree and Off-site seeders exhibited a general decreasing 11 species of shrubs were represented in the trend across the range of increasing distances ordination diagram (Fig. 3). Minimum distance (Fig. 4, scatter plot). from lower-severity/unburned edge shown on Cover was relatively low across the range of the gradient corresponds to optimal locations distances for on-site seeders at La Mesa. Location for Berberis fendleri, P. ponderosa, Fallugia paradoxa, of the sample with abundant cover of C. fendleri R. woodsii, and R. neomexicanus. Distribution of (largest gray-shaded area in Fig. 4, on-site seed- these species extended across the entire gradient ers plot) was exceptional for this functional 152

TABLE 1—Cover (mean and SD) and frequency (%) of woody species observed at La Mesa, New Mexico, and Saddle Mountain, Arizona. Designations of functional groups are as follows: 1 5 resprouters; 2 5 on-site seeders; 3 5 off-site seeders.

La Mesa (n 5 68) Saddle Mountain (n 5 79) Functional Scientific name (common name) Abbreviation Cover Mean (SD) Frequency % Cover Mean (SD) Frequency % group Abies concolor (white fir) Abco 0.01 (0.12) 0.01 2.92 (8.55) 0.29 3 Acer glabrum (Rocky Mountain maple) Acgl 0.19 (1.12) 0.04 1 Acer grandidentatum (big tooth maple) Acgr 0.13 (0.81) 0.03 1 Amelanchier utahensis (Utah serviceberry) Amut 0.03 (0.24) 0.01 1.44 (4.50) 0.18 1 Arctostaphylos pringlei (Pringlei manzanita) Arsp1 14.99 (21.87) 0.49 2 Arctostaphylos pungens (pointleaf manzanita) Artemisia carruthii (wormwood) Arsp2 2.19 (3.44) 0.46 0.27 (1.00) 0.09 3 Artemisia dranunculus (false tarragon)

Artemisia frigida (fringed sage) Naturalist Southwestern The Artemisia ludoviciana (wormwood) Artemisia tridentata (big sagebrush) Artr 0.53 (2.15) 0.11 3 Berberis fendleri (Colorado barberry) Befe 0.09 (0.41) 0.06 2 Berberis (Mahonia) repens (creeping barberry) Bere 1.19 (2.39) 0.37 1 Ceanothus fendleri (Fendler ceanothus) Cefe 1.90 (6.54) 0.18 0.82 (3.32) 0.14 2 Ceanothus greggii (Gregg ceanothus) Cegr 0.01 (0.11) 0.01 2 Cercocarpus montanus (true mountain mahogany) Cemo 0.22 (0.77) 0.10 0.85 (3.71) 0.08 1 Chrysothamnus nauseousus (golden rabbit brush) Chna 0.39 (1.22) 0.18 3 Cowania mexicana (cliff-rose) Come 0.54 (1.65) 0.11 3 Ephedra trifurea (longleaf ephedra or Mormon tea) Eptr 0.03 (0.16) 0.03 1 Fallugia paradoxa (Apache-plume) Fapa 0.66 (3.75) 0.07 1 Garrya flavescens (ashy silktassel) Gafl 0.08 (0.68) 0.01 1 Gutierrezia sarothrae (snakeweed) Gusp 0.75 (3.14) 0.10 0.01 (0.11) 0.01 3 Holodiscus dumosus (rock-spirea) Hodu 0.03 (0.23) 0.01 3 Juniperus communis (common juniper) Juco 0.06 (0.33) 0.04 3 Juniperus deppeana (alligator juniper) Jude 0.01 (0.12) 0.01 1 Juniperus monosperma (one-seed juniper) Jumo 0.06 (0.24) 0.06 3 Juniperus osteosperma (Utah juniper) Juos 0.03 (0.16) 0.03 3 Juniperus scopulorum (Rocky Mountain juniper) Jusc 0.07 (0.43) 0.03 3 Pachystima (Paxistima) myrsinites (mountain lover) Pamy 1.57 (4.68) 0.22 2 2 no. 53, vol. Physocarpus monogynus (mountain ninebark) Phmo 0.14 (0.78) 0.04 1 Picea engelmannii (Engelmann spruce) Pien 0.22 (1.00) 0.06 3 Pinus edulis (pin˜on pine)a Pied 0.06 (0.29) 0.05 3 ue20 ar n caia—ucsino lnsatrfrs ie 153 fires forest after plants of McGarigal—Succession and Haire 2008 June

TABLE 1—Continued.

La Mesa (n 5 68) Saddle Mountain (n 5 79) Functional Scientific name (common name) Abbreviation Cover Mean (SD) Frequency % Cover Mean (SD) Frequency % group Pinus flexilis (limber pine) Pifl 0.01 (0.12) 0.01 3 Pinus ponderosa (ponderosa pine) Pipo 5.34 (9.96) 0.54 1.30 (3.98) 0.24 3 Populus tremuloides (quaking aspen) Potr 0.10 (0.85) 0.01 20.90 (28.21) 0.47 1 Pseudotsuga menziesii (Douglas fir) Psme 0.03 (0.24) 0.01 0.20 (0.72) 0.10 3 Ptelea trifoliata (narrowleaf hoptree) Pttr 0.11 (0.60) 0.04 2 Prunus virginiana (chokecherry) Prvi 0.01 (0.12) 0.01 2 Purshia tridentata (antelopebrush) Putr 0.05 (0.22) 0.05 2 Quercus gambelii (Gambel oak) Qusp 12.00 (13.89) 0.78 8.38 (14.53) 0.46 1 Quercus turbinella (shrub live oak)b Quercus undulata (wavyleaf oak) Rhus trilobata (skunk bush) Rhtr 0.06 (0.34) 0.03 0.01 (0.11) 0.01 2 Ribes cereum (wax currant) Risp 0.38 (2.27) 0.06 2 Ribes inerme (gooseberry) Ribes viscosissimum (sticky currant) Rivi 0.43 (2.61) 0.05 2 Robinia neomexicanus (New Mexico locust) Rone 10.54 (11.01) 0.79 16.18 (17.78) 0.76 1 Rosa woodsii (wild rose) Rowo 1.07 (2.56) 0.29 0.20 (1.09) 0.04 1 Rubus parviflorus (raspberry or thimbleberry) Rusp 1.11 (4.18) 0.13 2 Rubus neomexicanus (thimbleberry) Sambucus racemosa (elderberry) Sasp 0.51 (1.50) 0.16 2 Symphoricarpos rotundifolius (snowberry) Sysp 2.92 (7.28) 0.33 2 a This species was observed at both sites but did not contribute to cover at La Mesa. b This species was observed at Saddle Mountain only but did not contribute to cover. 154 The Southwestern Naturalist vol. 53, no. 2

FIG. 3—Results of ordination with non-metric multidimensional scaling of species at the La Mesa burn, New Mexico, displayed on a modeled surface of distance to lower-severity/unburned edge (m). Location of the 4-letter code for a species represents the optimum, or peak of response in the ordination space along the distance gradient. Locations at the lower end of the distance gradient were optimal for Rosa woodsii, Robinia neomexicanus, and Pinus ponderosa, among others. Higher distances along the gradient corresponded with locations of peaks for species including Quercus, Juniperus monosperma, and Cercocarpus montanus. group. Otherwise, B. fendleri, P. virginiana, Rhus species were at much higher elevations; we trilobata, and Ribes were present at low total-cover recorded Arctostaphylos (Arsp1), for example, at values with no apparent interpretation for sample plots that ranged from 2,021 to 2,680 m, similarities in locations of samples. and A. utahensis occurred from 2,117 to 2,658 m. Saddle Mountain—At Saddle Mountain, opti- Symphoricarpos rotundifolius reached its peak at mal locations for 6 species of trees and 17 species a group of sample locations at middle elevations, of shrubs were distributed across an elevational set apart from optima of other species. Its gradient (Fig. 5). Composition of communities distribution overlapped with lower-elevation varied with greatest cover of Cowania mexicana, and higher-elevation species ranging from Artemisia tridentata, Chrysothamnus nauseousus, 2,111 to 2,707 m. Species of shrubs with optimal and P. edulis occurring at lowest elevations. locations above the 2,500-m contour included Communities in which Quercus, Purshia tridentata, some common plants such as C. fendleri and R. and C. montanus had relatively abundant cover neomexicanus, as well as some less-frequently were nearby at the lower end of the elevation observed species; Pachystima (Paxistima) myrsinites, gradient. Some species exhibited a broader Rubus, Sambucus, and B. repens. Maximum cover distribution than others; for example, C. nause- of P. ponderosa was located at the next position as ousus occurred at elevations from 2,021 to the gradient ascended; its range of occurrence 2,382 m, Quercus was observed from 2,021 to was 2,261 to 2,713 m. Locations with maximum 2,680 m, and P. tridentata was only at locations cover of R. viscosissimum, P. tremuloides, A. concolor, from 2,317 to 2,372 m. Optima for Artemisia Picea engelmannii, and P. menziesii, were clustered (Arsp2), Amelanchier utahensis, and Arctostaphylos at the upper end of the elevation gradient. (Arsp1) were located on the next contour, Observations of these species were confined to moving up along the gradient. These three sample plots within a fairly narrow range of June 2008 Haire and McGarigal—Succession of plants after forest fires 155

FIG. 4—The distance gradient (m) at La Mesa, New Mexico, modeled with the Generalized Additive Model, and symbols (gray shaded circles) for sample sites are scaled relative to total cover of species in a functional group. Off- site seeders define ecologically similar locations at low and intermediate distances, and these sites tend to be separate from sites where resprouters predominate. On-site seeders had high cover in only one location. The scatter plot shows a decreasing trend in total cover for off-site seeders in relation to distance from center of plot to lower-severity/unburned edge. elevation, with P. tremuloides and A. concolor cover in this functional group. At Saddle observed above 2,459 m, P. menziesii and R. Mountain, abundant cover of resprouters and viscosissimum observed only above ca. 2,550 m off-site seeders coincided, in contrast to patterns and P. engelmannii $2,673 m. at La Mesa. Variation in plant communities at Saddle Off-site seeders were limited in cover across Mountain was influenced by total cover of the gradient of elevation except for one cluster resprouting species, which reached their greatest of sample sites at the higher end of the gradient values at higher-elevation sites, and also were (Fig. 6). The species with greatest maximum influencial at several mid-elevation sites (Fig. 6). cover among the off-site seeders was A. concolor. Cover of P. tremuloides was a major contributor to Pinus ponderosa, P. engelmannii, and P. menziesii ecological similarity in sample sites at high also contributed to cover of off-site seeders at elevations. Across the gradient of elevation, higher-elevation sites. Cover of off-site seeders Quercus and R. neomexicanus contributed to total decreased with increasing distance to lower- 156 The Southwestern Naturalist vol. 53, no. 2

FIG. 5—Results of ordination with non-metric multidimensional scaling of species at the Saddle Mountain burn, Arizona, displayed on a modeled elevation surface (m). Location of the species abbreviation represents the optimum, or peak of response in the ordination space. Cowania mexicana and Pinus edulis are among species reaching their highest cover at lower elevations. Species with peak locations at higher elevations included Sambucus racemosa, Berberis (Mahonia) repens, Populus tremuloides, and Pseudotsuga menziesii. severity/unburned edge at Saddle Mountain; heterogeneity created by the fire, with its partic- greatest variation in cover was observed at the ular configurations of seed sources and neigh- shortest distances whereas moderate-to-large borhood conditions. For example, variation in cover was never observed at farther distances to abundance and diversity of species that spread lower-severity/unburned edge (Fig. 6, scatter from a refuge of seed sources remaining after the plot). Maximum cover of off-site seeders oc- fire (i.e., off-site seeders) influenced the spatial curred at sample plots where on-site seeders were dynamic of community composition across the less prevalent. landscape. Our finding of greater cover of off-site Distribution of on-site seeders also contributed seeders at shorter distances from lower-severity/ to defining similarities of sites, with greater cover unburned edge suggests gradual migration of occurring at sites all along one side of the these species following the model of wave-form gradient in elevation (Fig. 6). Moderate and low succession (Frelich, 2002; Fig. 7). Interacting levels of on-site seeders were scattered at sites in environmental conditions that allow establish- other areas. Arctostaphylos exhibited the largest ment of seedlings (e.g., for P. ponderosa; Bonnet et maximum cover among on-site seeders. Other al., 2005) likely influence the temporal and influential species in the on-site seeder group directional dynamic of migration into high- were S. rotundifolius, C. fendleri, P. myrsinites, severity patches by off-site seeders. Rubus, and R. viscosissimum. In addition, woody communities at our study areas resulted from legacies of spatial pattern in DISCUSSION—Community composition of woody the pre-disturbed landscape including distribu- species within high-severity patches at La Mesa tion of species that survive within the disturbance and Saddle Mountain was a function of spatial perimeter. This successional dynamic was evi- June 2008 Haire and McGarigal—Succession of plants after forest fires 157

FIG. 6—For Saddle Mountain, Arizona, the elevation surface (m) modeled with the Generalized Additive Model, and symbols (gray shaded circles) for sample sites are scaled relative to cover of species in a functional group. Highest total cover of resprouting species occurred at higher-elevations sites. Off-site seeders were low in cover across the gradient of elevation, except for one cluster of sample sites at the higher end of the gradient. Plots with higher cover of off-site seeders had lower cover of on-site seeders. The scatter plot illustrates a general decrease in cover of off-site seeders with increasing distance to lower-severity/unburned edge. dent in distribution of species that survive and site seeders that move into openings through resprout within a high-severity patch, such as time. Although Quercus can initially limit regen- Quercus or P. tremuloides, which corresponds with eration of trees after fire, succession to pine and their distribution before the fire, and depends other conifers has been documented (Moir and on environmental conditions in the pre-fire Ludwig, 1979; Moir et al., 1997). Differing landscape. As expected, resprouting species were distributions of on-site and off-site seeders at widespread across the full range of distances and Saddle Mountain could represent a similar elevations (Figs. 4 and 6). Abundant cover of process in which abundant cover of Arctostaphylos resprouting species generally did not coincide precedes movement of conifers into openings with greater cover of off-site seeders at La Mesa, over long time frames (Skau et al., 1970). suggesting potential neighborhood effects be- We observed possible neighborhood effects tween resprouters that establish quickly and off- from shading at some places. Some seedlings of 158 The Southwestern Naturalist vol. 53, no. 2

FIG. 7—Locations at La Mesa, New Mexico (upper photos), and Saddle Mountain, Arizona (lower photos), where Pinus ponderosa (an off-site seeder) regenerated in openings created by the fire. Adult trees appearing in the background of each photo survived the fire event, and represent refugial sources of seeds of ponderosa pine in lower-severity/unburned islands within the fire perimeter. Woody plant communities at these locations included Quercus and Robinia neomexicanus (La Mesa), Populus tremuloides, Ceanothus fendleri, and Arctostaphylos (Saddle Mountain). Arctostaphylos, a shrub in the on-site seeder group, formed a dense understory in some places (lower right).

P. edulis were observed growing in the shade of loides than of A. concolor (Fig. 5). At farthest Juniperus, although Juniperus and P. edulis were distances from lower-severity/unburned edge, uncommon at both study areas. Pinus edulis- no species at La Mesa reached maximum cover Juniperus woodlands are relatively slow to return (Fig. 3) and cover was low for all functional to a mature state, and based on historic evidence, groups (Fig. 4). Significance of this observation disturbance regimes are characterized by infre- is uncertain because of the limited number of quent stand-replacing fires in some places (Floyd sample locations at the greatest distances. et al., 2004). Regeneration by Abies concolor was Differences due to spatial and temporal common in understories with P. tremuloides at location were important in interpreting the Saddle Mountain; this observation is consistent relationship of successional outcomes to envi- with predicted neighborhood effects. Dense ronment. Important biogeographical differences regeneration of P. ponderosa has been document- in study sites may have resulted in stronger ed in small stands of P. tremuloides after fire association with one environmental gradient (Keyser et al., 2005). We observed some regen- rather than another. It is possible that the broad eration of P. ponderosa in larger forest openings mesas on the Pajarito Plateau provided a better of P. tremuloides, but optimal locations for P. environment in which to observe effects of ponderosa at Saddle Mountain were farther away distance from lower-severity/unburned edge in from locations with greatest cover of P. tremu- structuring plant communities after the La Mesa June 2008 Haire and McGarigal—Succession of plants after forest fires 159 fire. In contrast, the greater topographic relief food sources, a plethora of medicinal remedies, and dissimilarity of species associated with lower and ceremonial uses (Dunmire and Tierney, and higher elevations at Saddle Mountain 1995, 1997; United States Forest Service, http:// resulted in a better correspondence of plant www.fs.fed.us/database/feis/). In contrast to communities with changes in elevation. studies that emphasize undesirable effects when Certain underlying environmental gradients forests transition to openings and alternative can become more influential through time. Soon habitats, our research elucidates the need for after the Yellowstone fires of 1988, patterns of further consideration of both young forest burn severity strongly affected vegetational re- communities, and the persistent species and sponse, but the importance of other abiotic communities described as landscape scars, in factors has increased through time (Turner et conservation plans for forest systems of the al., 2003). Topography, therefore, could have southwestern United States. more influence on communities at Saddle Mountain, 45 years post-fire, while the influence We thank J. Stone for skillful interpretation and of spatial heterogeneity resulting from the fire mapping from historic photos and R. Franklin, J. Riling, D. Lively, M. Sperry, and M. Brachman for still persists at La Mesa, only 28 years post-fire. outstanding work in collection of data in the field. The environmental gradients we examined Permits, data, and logistical support from personnel at undoubtedly interact with multi-scale temporal Bandelier National Monument, Santa Fe National and spatial variables, including regional pools of Forest, Kaibab National Forest, Grand Canyon National species, precipitation, and heterogeneity of Park, and Los Alamos National Laboratory are greatly resources (Keeley et al., 2005). Further investi- appreciated. This research was supported by The gation is needed to understand the role of these Nature Conservancy, the United States Geological factors and their interactions in structuring plant Survey, the University of Massachusetts, and a fellow- communities at La Mesa and Saddle Mountain. ship from the American Association of University Conclusion—Understanding landscape change Women. Reviews by S. DeStefano, S. Lerman, C. Miller, and two anonymous reviewers improved the quality of after severe fire events is of vital importance, the manuscript. We are grateful to P. McCarthy, M. given predictions of a warming climate accom- Savage, and W. Romme, who served as reviewers, as well panied by an increase in frequency, size, and as consultants throughout the project. duration of fires. Explicit consideration of spatial patterns of burning in relation to life-history LITERATURE CITED characteristics is key to understanding complex successional outcomes after severe disturbance. ALLEN, C. D., J. L. BETANCOURT, AND T. W. SWETNAM. 1998. In large fires, broad gradients affect spatial Landscape changes in the southwestern United patterns of burning, and vegetation patterns States: techniques, long-term data sets, and trends. are influenced in turn (Turner and Romme, Pages 71–84 in Perspectives on the land-use history 1994). If these factors are not considered, of North America: a context for understanding our important variability in species and communities changing environment (T. D. Sisk, editor). BSR- 1998-0003. Biological Science Report, United States is likely to be missed. Geological Survey, Biological Resources Division. Detailed information about species can benefit ALLEN, C. D., M. SAVAGE,D.A.FALK,K.F.SUCKLING,T.W. conservation assessments (Brooks et al., 2004), SWETNAM,T.SCHULKE,P.B.STACEY,P.MORGAN,M. especially when variability in ecosystems is likely HOFFMAN, AND J. KLINGEL. 2002. Ecological restora- changing with climate. Individual species and tion of southwestern ponderosa pine ecosystems: a the communities they compose play ecological broad perspective. Ecological Applications 12: and cultural roles on contemporary landscapes 1418–1433. (sensu Nabhan, 1997, and Center for Sustainable BONNET, V. H., A. W. SCHOETTLE, AND W. D. SHEPPERD. Environments, 2002). Areas burned in severe fire 2005. Postfire environmental conditions influence at Saddle Mountain and La Mesa included the spatial pattern of regeneration for Pinus ponderosa. Canadian Journal of Forest Research 35:37–47. communities that might diversify function of BRAY, J. R., AND J. T. CURTIS. 1957. An ordination of the landscapes through creation of early successional upland forest communities in southern Wisconsin. habitats for wildlife (Romme and Knight, 1982). Ecological Monographs 27:325–349. In addition, woody species at the study sites have BROOKS, T., G. A. B. DA FONSECA, AND A. S. L. RODRIGUES. a wide range of traditional and current uses; 2004. Species, data, and conservation planning. basketry and other building material, important Conservation Biology 18:1682–1688. 160 The Southwestern Naturalist vol. 53, no. 2

BROWN, J. K., AND N. V. DEBYLE. 1987. Fire damage, landscape patterns and ecosystem management. mortality, and suckering in aspen. Canadian Jour- Journal of Vegetation Science 9:603–610. nal of Forest Research 17:1100–1109. KEELEY, J. E., C. J. FOTHERINGHAM, AND M. BAER-KEELEY. CENTER FOR SUSTAINABLE ENVIRONMENTS. 2002. Safeguard- 2005. Factors affecting plant diversity during post- ing the uniqueness of the Colorado Plateau an fire recovery and succession of Mediterranean- ecoregional assessment of biocultural diversity. climate shrublands in California, USA. Diversity Northern Arizona University, Flagstaff. Distributions 11:525–537. CHRISTENSEN, N. L., J. K. AGEE,P.F.BRUSSARD,J.HUGHES, KEYSER, T. L., F. W. SMITH, AND W. D. SHEPPERD. 2005. D. H. KNIGHT,G.W.MINSHALL,J.M.PEEK,S.J.PYNE, Trembling aspen response to a mixed-severity F. J. SWANSON,J.W.THOMAS,S.WELLS,S.E.WILLIAMS, wildfire in the Black Hills, South Dakota, USA. AND H. A. WRIGHT. 1989. Interpreting the Yellow- Canadian Journal of Forest Research 35:2679–2684. stone fires of 1988. BioScience 39:678–685. MCKENZIE, D., Z. GEDALOF,D.L.PETERSON, AND P. MOTE. CLARK, J. S. 1993. Paleoecological perspectives on 2004. Climatic change, wildfire, and conservation. modeling broad-scale responses to global change. Conservation Biology 18:890–902. Pages 315–332 in Biotic interactions and global MILLAR, C. I., AND W. B. WOOLFENDEN. 1999. The role of change (P. M. Kareiva, J. G. Kingsolver, and R. B. climate change in interpreting historical variability. Huey, editors). Sinauer Associates, Sunderland, Ecological Applications 9:1207–1216. Massachusetts. MILLER, C. 2003. Simulation of the effects of climatic CLARKE, K. R. 1993. Non-parametric multivariate anal- change on fire regimes. Pages 69–94 in Fire and yses of changes in community structure. Australian climatic change in temperate ecosystems of the Journal of Ecology 18:117–143. western Americas. Volume 160 (T. T. Veblen, W. L. DUNMIRE, W. W., AND G. D. TIERNEY. 1995. Wild plants of Baker, G. Montenegro, and T. W. Swetnam, the Pueblo Province. Museum of New Mexico Press, editors). Ecological Studies, Springer, New York. Santa Fe. MOIR, W. H., AND J. A. LUDWIG. 1979. A classification of DUNMIRE, W. W., AND G. D. TIERNEY. 1997. Wild plants spruce-fir and mixed habitat types of Arizona and and native peoples of the Four Corners. Museum of New Mexico. Research Paper RM-207. United States New Mexico Press, Santa Fe. Department of Agriculture. Forest Service, Rocky ECOLOGICAL RESTORATION INSTITUTE. 2004. Plants of Mountain Forest and Range Experiment Station, northern Arizona forests: a comprehensive guide Fort Collins, Colorado. for the Mogollon Plateau coniferous forest bior- MOIR, W. H., B. W. GEILS,M.A.BENOIT, AND D. SCURLOCK. egion in Arizona. Northern Arizona University, 1997. Ecology of southwestern ponderosa pine Flagstaff. forests. Pages 3–27 in Songbird ecology in south- EPPLE, A. O. 1995. A field guide to the plants of western ponderosa pine forests: a literature review Arizona. Globe Pequot Press, Guilford, Connecti- (W. M. Block and D. M. Finch, editors). General cut. Technical Report RM-GTR-292. United States De- FLOYD, M. E. 1982. The interaction of pin˜on pine and partment of Agriculture Forest Service. Rocky Gambel oak in plant succession near Dolores, Mountain Forest and Range Experiment Station, Colorado. Southwestern Naturalist 27:143–147. Fort Collins, Colorado. FLOYD, M. L., D. D. HANNA, AND W. H. ROMME. 2004. MORGAN, P., G. E. DEFOSSE, AND N. F. RODRIGUEZ. 2003. Historical and recent fire regimes in pin˜on-juniper Management implications of fire and climate woodlands on Mesa Verde, Colorado. Forest Ecol- changes in the western Americas. Pages 413–440 ogy and Management 198:269–289. in Fire and climatic change in temperate ecosys- FOXX, T. S., AND D. HOARD. 1995. Flowering plants of the tems of the western Americas. Volume 160 (T. T. southwestern woodlands. Otowi Crossing Press, Los Veblen, W. L. Baker, G. Montenegro, and T. W. Alamos, New Mexico. Swetnam, editors). Ecological Studies, Springer, FRELICH, L. E. 2002. Forest dynamics and disturbance New York. regimes: studies from temperate evergreen-decidu- NABHAN, G. P. 1997. Cultures of habitat: on nature, ous forests. Cambridge University Press, Cam- culture, and story. Counterpoint, Washington, D. C. bridge, United Kingdom. OKSANEN, J., AND P. R. MINCHIN. 2002. Continuum theory FRELICH, L. E., AND P. B. REICH. 1995. Spatial patterns revisited: what shape are species responses along and succession in a Minnesota southern-boreal ecological gradients? Ecological Modeling 157: forest. Ecological Monographs 65:325–346. 119–129. HOPKINS, R. L. 1990. Kaibab formation. Pages 225–245 PIERCE, J. L., G. A. MEYER, AND A. J. T. JULL. 2004. Fire- in Grand Canyon geology (S. S. Beus and M. induced erosion and millennial-scale climate Morales, editors). Oxford University Press, New change in northern ponderosa pine forests. Nature York. 432:87–90. JOHNSON, E. A., K. MIYANISHI, AND J. M. H. WEIR. 1998. PONGE, J-F., J. ANDRE´ ,O.ZACKRISSON,N.BERNIER, M-C. Wildfires in the western Canadian boreal forest: NILSSON, AND C. GALLET. 1998. The forest regenera- June 2008 Haire and McGarigal—Succession of plants after forest fires 161

tion puzzle, biological mechanisms in humus layer TURNER,M.G.,AND V. H. DALE. 1998. Comparing large, and forest vegetation dynamics. BioScience 48: infrequent disturbances: what have we learned? 523–530. Ecosystems 1:493–496. RENEAU, S. L., AND E. V. MCDONALD.(EDITORS). 1996. TURNER, M. G., AND W. H. ROMME. 1994. Landscape Landscape history and processes on the Pajarito dynamics in crown fire ecosystems. Landscape Plateau, northern New Mexico LA-UR-96-3035. Los Ecology 9:59–77. Alamos National Laboratory, Los Alamos, New TURNER, M. G., W. H. ROMME, AND D. B. TINKER. 2003. Mexico. Surprises and lessons from the 1988 Yellowstone ROMME, W. H., AND D. H. KNIGHT. 1982. Landscape fires. Frontiers in Ecology and Environment diversity: the concept applied to Yellowstone Park. 1:351–358. BioScience 32:664–670. TURNER, M. G., W. L. BAKER,C.J.PETERSON, AND R. K. SAVAGE, M., AND J. N. MAST. 2005. How resilient are PEET. 1998. Factors influencing succession: lessons southwestern ponderosa pine forests after crown from large, infrequent natural disturbances. Eco- fires? Canadian Journal of Forest Research 35: systems 1:511–523. 967–977. TURNER, M. G., W. H. ROMME,R.H.GARDNER, AND W. W. HARGROVE. 1997. Effects of fire size and pattern on SCHOENNAGEL, T., T. T. VEBLEN, AND W. H. ROMME. 2004. The interaction of fire, fuels, and climate across early succession in Yellowstone National Park. Rocky Mountain forests. BioScience 54:661–676. Ecological Monographs 67:411–433. VAN WAGNER, C. E. 1983. Fire behavior in northern SKAU, C. M., R. O. MEEUWIG, AND T. W. TOWNSEND. 1970. conifer forests and shrublands. Pages 65–80 in The Ecology of eastside Sierra chaparral: a literature role of fire in northern circumpolar ecosystems (R. review. R71. University of Nevada, Max C. Fleisch- W. Wein and D. A. MacLean, editors). John Wiley mann College of Agriculture, Agricultural Experi- and Sons, New York. ment Station, Reno. WESTERLING, A. L., H. G. HIDALGO,D.R.CAYAN, AND T. W. STROM, B. A., AND P. Z. FULE´ . 2007. Pre-wildfire fuel SWETNAM. 2006. Warming and earlier spring increas- treatments affect long-term ponderosa pine forest es western U.S. forest wildfire activity. Science dynamics. International Journal of Wildland Fire 313:940–943. 16:128–138. WOOD, S. N. 2000. Modeling and smoothing parameter SWETNAM, T. W., AND C. H. BAISAN. 1996. Historical fire estimation with multiple quadratic penalties. Journal regime patterns in the southwestern United States of the Royal Statistical Society B 62(Part 2):413–428. since 1700. Pages 11–32 in Fire effects in south- WOODMENCEY, J. 2001. Weather in the Southwest. western forests: proceedings of the second La Mesa Southwest Parks and Monuments Association, fire symposium (C. D. Allen, editor). United States Tucson, Arizona. Forest Service General Technical Report RM-286, YEE, T. W., AND N. D. MITCHELL. 1991. Generalized Los Alamos, New Mexico. additive models in plant ecology. Journal of SWETNAM, T. W., AND J. L. BETANCOURT. 1998. Mesoscale Vegetation Science 2:587–602. disturbance and ecological response to decadal climatic variability in the American Southwest. Submitted 21 August 2006. Accepted 25 October 2007. Journal of Climate 11:3128–3147. Associate Editor was Chris Lauver.