- Top of page
- Literature Cited
Abstract: Understanding patterns of plant population mortality during extreme weather events is important to conservation planners because the frequency of such events is expected to increase, creating the need to integrate climatic uncertainty into management. Dominant plants provide habitat and ecosystem structure, so changes in their distribution can be expected to have cascading effects on entire communities. Observing areas that respond quickly to climate fluctuations provides foresight into future ecological changes and will help prioritize conservation efforts. We investigated patterns of mortality in six dominant plant species during a drought in the southwestern United States. We quantified population mortality for each species across its regional distribution and tested hypotheses to identify ecological stress gradients for each species. Our results revealed three major patterns: (1) dominant species from diverse habitat types (i.e., riparian, chaparral, and low- to high-elevation forests) exhibited significant mortality, indicating that the effects of drought were widespread; (2) average mortality differed among dominant species (one-seed juniper[Juniperus monosperma (Engelm.) Sarg.] 3.3%; manzanita[Arctostaphylos pungens Kunth], 14.6%; quaking aspen[Populus tremuloides Michx.], 15.4%; ponderosa pine[Pinus ponderosa P. & C. Lawson], 15.9%; Fremont cottonwood[Populus fremontii S. Wats.], 20.7%; and pinyon pine[Pinus edulis Engelm.], 41.4%); (3) all dominant species showed localized patterns of very high mortality (24–100%) consistent with water stress gradients. Land managers should plan for climatic uncertainty by promoting tree recruitment in rare habitat types, alleviating unnatural levels of competition on dominant plants, and conserving sites across water stress gradients. High-stress sites, such as those we examined, have conservation value as barometers of change and because they may harbor genotypes that are adapted to climatic extremes.
Resumen: El entendimiento de los patrones de mortalidad de poblaciones de plantas durante eventos climáticos extremos es importante para los planificadores de conservación porque se espera que la frecuencia de tales eventos aumente, creando la necesidad de integrar la incertidumbre climática a la gestión. Las plantas dominantes proporcionan hábitat y estructura al ecosistema, así que se puede esperar que cambios en su distribución tengan efectos de cascada en toda la comunidad. La observación de áreas que responden rápidamente a las fluctuaciones climáticas proporciona un panorama de futuros cambios ecológicos y ayudará a la definición de prioridades de esfuerzos de conservación. Investigamos los patrones de mortalidad en seis especies de plantas dominantes durante una sequía en el suroeste de Estados Unidos. Cuantificamos la mortalidad poblacional para cada especie en su área de distribución regional y probamos hipótesis para identificar los gradientes de estrés ecológico para cada especie. Nuestros resultados revelaron tres patrones mayores: (1) las especies dominantes en diversos tipos de hábitats (i.e., ribereño, chaparral y bosques de baja a alta elevación) presentaron mortalidad significativa, lo que indica que los efectos de la sequía fueron extendidos; (2) la mortalidad promedio fue diferente (Juniperus monosperma [Engelm.] Sarg.) 3.3%; Arctostaphylos pungens Kunth, 14.6%; Populus tremuloides Michx., 15.4%; Pinus ponderosa P. & C. Lawson, 15.9%; Populus fremontii S. Wats., 20.7%; y Pinus edulis Engelm., 41.4%); (3) todas las especies dominantes mostraron patrones localizados de mortalidad muy alta (24–100%) consistentes con gradientes de estrés hídrico. Los gestores de tierras deberían planificar para la incertidumbre climática mediante la promoción del reclutamiento de árboles en tipos de hábitat raros, lo que aligeraría los niveles no naturales de competencia sobre las plantas dominantes y conservaría sitios a lo largo de gradientes de estrés hídrico. Los sitios con estrés alto, como los que examinamos, tienen valor de conservación como barómetros de cambio y porque pueden albergar genotipos que están adaptados a cambios climáticos extremos.
- Top of page
- Literature Cited
Climate change is altering species distributions, thereby complicating conservation efforts. Current models predict biotic responses to climate change on a global scale and ignore the regional and short-term patterns and processes useful to conservation biologists and land managers (e.g., International Panel on Climate Change [IPCC] 2001; Thomas et al. 2004). Local and rapidly changing factors, such as extreme weather events, landscape modifications, invasive species, and changing genetic frequencies are likely to interact with long-term climate trends to cause more severe effects than any of the factors alone, but such interactions are poorly understood (Loehle & LeBlanc 1996; Gutschick & BassiRad 2003; Pounds & Puschendorf 2004). Conservation planners must understand how climate drives ecological changes at spatial and temporal scales relevant to human decision making.
The IPCC (2001) predicts an increase in extreme weather events and a 66–90% chance of increased midcontinental drought frequency; thus, more studies on the effects of climate perturbations on ecosystems are needed. As a primary objective in long-term climate change monitoring, the National Research Council (NRC 1990) called for the identification of specific sites within ecosystems to be designated as barometers of climate change. These sensitive areas will yield information about the effects of climate on all ecosystems and have high conservation value. A decade later, the National Ecological Observatory Network (NEON 2000), a large-scale environmental monitoring initiative, named ecotones and transition zones as useful barometers. According to the IPCC (2001), to assess ecosystem vulnerability and prioritize conservation efforts, in situ studies of ecosystem change are more realistic than those that suggest species migration, but all the in situ studies cited in their report are a posteriori or paleoecological, and none document a mortality event in progress. Identification of the stress gradients affecting specific dominant plants will help locate barometer sites and enable monitoring and prediction of habitat change.
High-stress locations and ecotonal regions have been recognized only recently as a worthy investment of conservation funds (NEON 2000; Channell & Lomolino 2000; Smith et al. 2001). Environmental gradients should have the highest within-species levels of adaptive variation, and extreme environments should drive selection for novel genotypes (Smith et al. 2001; Gutschick & BassiRad 2003). Peripheral habitats are essential refuge locations for many species owing to their relative lack of anthropogenic influences (Channell & Lomolino 2000). By defining stress gradients in the context of dominant plants, we can manage habitats for dependent associated communities. It is important to understand the effects of extreme events on dominant vegetation because the death of dominant plant species will have cascading effects on other trophic levels (Whitham et al. 2003).
The southwestern United States experienced an extreme drought event in 2002 (NOAA 2003), which resulted in widespread mortality of dominant plants in multiple community types, including manzanita (Arctostaphylos pungens Kunth), quaking aspen (Populus tremuloides Michx.), ponderosa pine (Pinus ponderosa P. & C. Lawson), Fremont cottonwood (Populus fremontii S. Wats.), and pinyon pine (Pinus edulis Engelm.) (Allen 2004; this paper). United States Forest Service surveys show that as of 2003, 12,000 km2 of pinyon and ponderosa pine have died in the Southwest (Breshares et al. 2005; Mueller et al. 2005).
Despite the overall high mortality across the region, localized levels of mortality were spatially heterogeneous, ranging from 0 to 100%. Little is known about what factors affect the probability of mortality in dominant woody plants during a severe drought (but see Allen & Breshears 1998; Fensham & Holman 1999; Suarez et al. 2004). Here we identify some of the major patterns of mortality associated with a record drought across diverse habitat types. Previously researchers have focused on a single ecosystem's response to water stress (Solomon & Kirilenko 1997; Allen & Breshears 1998; Horton et al. 2003), plant functional type (e.g., Condit et al. 1996; Sperry & Hacke 2002), or individual species or groups of related species (Fensham & Holman 1999; Suarez et al. 2004). To our knowledge, no studies have documented concurrent patterns of extreme drought mortality for the dominant plants that characterize a wide range of habitat types in a local region (riparian, semiarid, and low- to high-elevation forest).
The semiarid region surrounding Flagstaff, Arizona (U.S.A.), affords a unique opportunity to study local and regional patterns of drought mortality due to the presence of diverse vegetation types within a short geographic distance and the various stress gradients that have been documented previously as affecting plant water availability. Elevation varies from approximately 500–3400 m, creating a gradient of temperature and precipitation, and varied local topography creates water stress related to slope aspect (Ogle et al. 2000; Nevo 2001). Many edaphic stress gradients result from the varied age, composition, and texture of both igneous and sedimentary soils (Sullivan & Downum 1991; Cobb et al. 1997). Multiple species interactions leading to various stress-inducing or stress-relieving relationships act independently or in combination to affect patterns of mortality (Johnsen 1962; Bertness & Callaway 1994; Busch & Smith 1995). The severe drought event culminating in 2002 enabled investigations of plant mortality associated with each of these conditions.
We hypothesized that mortality would not be distributed randomly across the region but would be associated with specific factors linked to an increase in water stress. Because plant responses to stress are likely to be species-specific, the effects of extreme drought are likely to differ among species at either a regional level (mortality among species) or local scale (mortality gradients within a single species). We investigated three major hypotheses: (1) overall mortality during drought is greater in some dominant species than others, (2) stressors associated with mortality vary by species and locality, and (3) greater mortality is associated with more stressful environments. With an understanding of how drought mortality is manifested across the region, one can begin to predict future vegetation and community distributions at regional scales and incorporate climate change predictions into conservation efforts.
- Top of page
- Literature Cited
Owing to widespread tree mortality during the 2002 drought, we originally initiated several studies of the mortality of individual species. However, we saw an unprecedented opportunity to consider all of the studies simultaneously and address broader hypotheses about the effects of extreme drought on dominant plant distributions. We compiled data from the individual studies and then used a common method to compare mortality among species.
To examine local and regional patterns of drought-associated mortality, we conducted our research within an 80–km radius around Flagstaff, Arizona, between the fall of 2002 and the spring of 2004. We divided the research area into three zones across an elevational gradient and chose two dominant plants to represent each: (1) semiarid zone (500–1500 m, including riparian areas), where we measured mortality in Fremont cottonwood and manzanita, (2) midelevational woodland (1500–2300 m), where we measured mortality in one-seed juniper and pinyon pine, and (3) montane forests (2300–3000 m), where we measured mortality in quaking aspen and ponderosa pine. We sampled locations within the core ranges of these plants and avoided the elevational extremes of their distributions.
To test for differences in mortality levels among species, we counted live and dead trees at sites within the 80–km radius. A minimum of 14 sites per dominant plant were identified. We sampled one-seed juniper, pinyon pine, and ponderosa pine at forested locations within 2 km of state and interstate highways; all sites were >5 km apart. We counted quaking aspen in all stands encountered along U.S. Forest Service roads chosen for their proximity to aspen habitat. At each site, we picked two haphazard directions and sampled two straight-line transects until 100 trees of each dominant species present were encountered. Each tree or shrub along the transects was classified as living or dead, with the assumption that a lack of aboveground live biomass represented a mortality event. To count manzanita and one-seed juniper growing in areas where sensitive landscapes prevented the use of straight transects, we located preexisting trails >5 km apart and counted all trees near the trails. Owing to the rarity of Fremont cottonwood habitat in the region and limited presence of large accessible stands, we extended our searches beyond the 80-km radius to include sufficient sample sizes for this species. The 20 sites of the local Fremont cottonwood study (see below) were also used for the regional study, with 30 trees per site. We compared average mortality level per site across the six species.
To contrast mortality levels with predrought habitat abundance, we compared average percent mortality per stand to the percentage of landscape occupied by each species. The predrought area occupied by each species was determined from a digital Arizona Gap Analysis Project Vegetation Map (Halvorson et al. 2001). We estimated occupied area within an 80-km radius around Flagstaff for all plants except Fremont cottonwood, which was studied within a 170-km radius.
We compared documented factors associated with increased water stress (competition, soil age, soil type, elevation, distance to water, slope aspect) to within-site mortality levels (Table 1).
Table 1. Patterns of dominant plant mortality observed in this study and their potential mechanisms as documented in the literature.
|Patterns of mortality||Potential mechanisms||References|
| association with invasive species tamarisk||Fremont cottonwoods may depend more on surface soil moisture during water stress||Reily & Johnson 1982; Vandersande et al. 2001|
|high salinity levels can inhibit surface soil moisture use|| |
|tamarisk increase water use and growth in saline conditions, and individual plants use more water when growing in dense thickets||Busch & Smith 1995; Devitt et al. 1997; Glenn et al. 1998; Vandersande et al. 2001|
| distance from water course||there may be a drop in water table at increased distance from wash||Masek Lopez & Springer 2002|
| association with grasses||competition with understory vegetation may increase water stress||Teague et al. 2001|
| soil type||a stress gradient exists between coarse-cinder and finer-textured soils||Cobb et al. 1997|
| slope aspect||more stressful conditions exist on slope aspects oriented toward the equator||Ogle et al. 2000; Nevo 2001|
| soil depth||deep cinders act as a mulch and retain more soil moisture than shallow cinders||Sullivan & Downum 1991|
| soil age||edaphic stress decreases with soil age due to higher water availability in older cinder soils||P. Selmants, personal communication;|
| nurse association/competition||facilitation is important when abiotic stress is high, but as abiotic stress decreases, competition becomes a larger factor in plant mortality||Bertness & Callaway 1994; Callaway et al. 2002|
| elevation||temperature and precipitation gradients correlated with elevation cause higher stress at lower elevations||Allen & Breshears 1998|
|a late freeze occurred in 1999 when low-elevation trees were leafing out, compounding other stressors||M. Manthei, personal communication|
To test whether Fremont cottonwoods growing in association with the invasive species tamarisk (Tamarix sp.) were experiencing greater mortality than trees in stands with no tamarisk, we selected five river systems. Twenty sites with varying tamarisk cover were chosen. We selected 30 Fremont cottonwoods at each site and classified each as living or dead. Tamarisk cover was estimated by measuring cover along three 50-m transects established perpendicular to the river, starting at the inner edge of riparian vegetation. The percent cover along the three transects was then averaged to obtain a value for the site. We regressed cottonwood mortality levels by tamarisk cover.
To determine whether manzanita mortality increased as distance from an ephemeral wash increased, we sampled twelve, 75-m transects along washes, with six transects in the eastern direction and six transects in the western direction. We classified each shrub encountered along the transect as living or dead and measured its distance from the wash. We tested mortality levels of manzanita found growing <40 m from the wash with those found growing >40 m from the wash.
To test the effect of slope aspect on pinyon pine mortality, we chose plots encompassing 40 trees each on a north-facing slope and a south-facing slope with similar elevation, slope angles, and soil type. All pinyon pine trees within each plot were counted and recorded as live or dead. Mortality levels were compared between the two slopes. To determine whether soil depth had an effect on pinyon pine mortality, we chose two visually different cinder deposits, red and black. We established two random 20-m2 plots in both red and black cinders and measured soil depth at 25 locations within each plot. We placed plots along north- and east-facing slopes with the same elevation and slope angle to constrain these potentially confounding factors. Every pinyon pine within a plot was classified as living or dead, and pinyon mortality levels were compared between the two soils.
To investigate whether association with grasses affected mortality in one-seed juniper, we compared mortality rates of one-seed juniper growing in a grassland habitat (area visually dominated by grass cover) and an adjacent nongrassland habitat (area nearly devoid of grasses). We chose a site at the edge of a volcanic cinder field that included both grassland and nongrassland habitat. We sampled a 200-m-long, 30-m-wide belt transect across the study area. We counted 181 one-seed juniper trees along this transect and classified each tree as living or dead and as occurring in grassland or nongrassland. Mortality levels in and out of grassland were compared.
To determine whether soil age and nurse–plant associations affect mortality levels of ponderosa pine seedlings, we sampled four 100-m transects at three differently aged volcanic soil sites as defined by Moore et al. (1974): 0.92 million years old, 0.33 million years old, and 0.15 million years old. We recorded all ponderosa pine seedlings within 20 m of the transects. We noted whether each seedling was growing within 3 cm of another plant or sheltering feature (nursed) or in the open (not nursed) and classified each as living or dead. We compared the mortality levels of seedlings that were nursed and those that were not at all soil ages. To determine whether soil age affected adult ponderosa pine mortality, we classified each mature ponderosa pine along each transect as living or dead and compared mortality of adult ponderosa pine among soil ages.
To determine whether quaking aspen mortality differed along an elevational gradient, we located two high- (>2900 m), six intermediate- (2600–2900 m), and four low-elevation (2300–2600 m) sites. Four transects were established at each site, and 100 quaking aspens were counted and categorized as either living or dead. We correlated mortality levels of quaking aspen with elevational ranges.
For comparisons with categorical response and independent variables, we applied the likelihood ratio χ2 test. In one case, we had two independent variables and used a 2 × 3 × 2 χ2 test in conjunction with post hoc Wald effect tests. Independent t tests were used to analyze data with binary independent variables and continuous response variables if the assumptions of normal distribution and homoscedascity of variance were met. We tested equality of variance with Bartlett's test ( p > 0.05) and normality of the distribution with the Shapiro-Wilk test ( p > 0.05). When the data did not meet these criteria, we used the nonparametric Mann-Whitney U test. We used least squares linear regression in cases where the data consisted of a continuous response variable and continuous independent variable. To test for differences in regional mortality among dominant species, we used a Welch analysis of variance (ANOVA) because we were unable to correct the heteroscedascity of variance. A post hoc Dunnet's T3 test was used to determine which species were significantly different from one another. All statistics were performed in JMP-IN 4.0 (SAS Institute 1999), except the Dunnet's T3 test, which was completed in SPSS 11.5 (SPSS 2002).