3128 V
OLUME
11JOURNAL OF CLIMATE
Mesoscale Disturbance and Ecological Response to Decadal Climatic Variability in the
American Southwest
T
HOMAS
W. S
WETNAM
Laboratory of Tree-Ring Research, The University of Arizona, Tucson, Arizona
J
ULIO
L. B
ETANCOURT
Desert Laboratory, U.S. Geological Survey, Tucson, Arizona
(Manuscript received 14 January 1997, in final form 31 July 1997)
ABSTRACT
Ecological responses to climatic variability in the Southwest include regionally synchronized fires, insect
outbreaks, and pulses in tree demography (births and deaths). Multicentury, tree-ring reconstructions of drought,
disturbance history, and tree demography reveal climatic effects across scales, from annual to decadal, and from
local (,10
2
km
2
) to mesoscale (10
4
–10
6
km
2
). Climate–disturbance relations are more variable and complex
than previously assumed. During the past three centuries, mesoscale outbreaks of the western spruce budworm
(Choristoneura occidentalis) were associated with wet, not dry episodes, contrary to conventional wisdom.
Regional fires occur during extreme droughts but, in some ecosystems, antecedentwet conditions play a secondary
role by regulating accumulation of fuels. Interdecadal changes in fire–climate associations parallel other evidence
for shifts in the frequency or amplitude of the Southern Oscillation (SO) during the past three centuries. High
interannual, fire–climate correlations (r 5 0.7 to 0.9) during specific decades (i.e., circa 1740–80 and 1830–
60) reflect periods of high amplitude in the SO and rapid switching from extreme wet to dry years in the
Southwest, thereby entraining fire occurrence across the region. Weak correlations from 1780 to 1830 correspond
with a decrease in SO frequency or amplitude inferred from independent tree-ring width, ice core, and coral
isotope reconstructions.
Episodic dry and wet episodes have altered age structures and species composition of woodland and conifer
forests. The scarcity of old, living conifers established before circa 1600 suggests that the extreme drought of
1575–95 had pervasive effects on tree populations. The most extreme drought of the past 400 years occurred
in the mid–twentieth century (1942–57). This drought resulted in broadscale plant dieoffs in shrublands, wood-
lands, and forests and accelerated shrub invasion of grasslands. Drought conditions were broken by the post-
1976 shift to the negative SO phase and wetter cool seasons in the Southwest. The post-1976 period shows up
as an unprecedented surge in tree-ring growth within millennia-length chronologies. This unusual episode may
have produced a pulse in tree recruitment and improved rangeland conditions (e.g., higher grass production),
though additional study is needed to disentangle the interacting roles of land use and climate. The 1950s drought
and the post-1976 wet period and their aftermaths offer natural experiments to study long-term ecosystemresponse
to interdecadal climate variability.
1. Introduction
Climatic variables such as radiation, temperature,and
precipitation determine rates of ecosystem processes
from net primary productivity to soil development. They
predict a wide array of biogeographic phenomena, in-
cluding soil carbon pools, vegetation physiognomy, spe-
cies range, and plant and animal diversity. Climate also
influences ecosystems indirectly by modulating the fre-
quency, magnitude, and spatial scales of natural distur-
Corresponding author address: Dr. Thomas W. Swetnam, Labo-
ratory of Tree-Ring Research, The University of Arizona, Building
#58, Tucson, AZ 85721.
E-mail: tswetnam@ltrr.arizona.edu
bances (Clark 1988; Overpeck et al. 1990; Swetnam
1993)
Disturbance is any discrete event that results in the
sudden mortality of biomass on a timescale significantly
shorter than that of its accumulation (Huston 1994). A
broader definition would include any abrupt event that
disrupts community structure and changes the physical
environment, resources, or availability of space (White
and Pickett 1985). Much of the focus in disturbance
ecology has been on the patch dynamics produced by
resetting of succession within portions of an ecosystem
(so-called gap disturbances). Emphasis has shifted from
how disturbances produce landscape mosaics to how
these mosaics in turn predispose the landscape to further
disturbance. Both spatial and temporal scale enter the
D
ECEMBER
1998 3129SWETNAM AND BETANCOURT
discussion about whether disturbance helps to maintain
or prevent ecosystems from ever reaching equilibrium.
The current paradigm favors a nonequilibrium per-
spective, but steady states may be scale dependent. The-
oretically, if areas affected by disturbance are suffi-
ciently small, then a self-reproducing steady state can
exist as an average condition over relatively larger areas
(Sousa 1984; Rogers 1996; Turner et al. 1993). A re-
gional steady state is unlikely wherever mesoscale dis-
turbances (10
4
–10
6
km
2
) recur on timescales shorter
than the successional cycles necessary to reach and
maintain equilibrium across a region (Romme 1982;
Barden 1988; Sprugel 1991; Swetnam 1993; Turner et
al. 1993). One such place is the American southwest,
where rates of plant and population growth are sluggish;
soil properties accumulate only gradually (10
3
–10
5
yr)
and are usually in disequilibrium with current climate
and vegetation; meager soil water and organic matter
cannot buffer climatic extremes; and episodic mortality
from drought, insect outbreaks, and fire is often syn-
chronized and broadscale.
In this paper, we explore the effects of climatic vari-
ability and mesoscale disturbance on nonriparian eco-
systems of the American southwest. We emphasize de-
cadal and regional scales for three reasons. First, the
mean generation times of most of the woody perennials
that dominate from the lowland deserts to upland forests
are from 15 to 50 yr; hence these are the minimal time
spans to observe their population dynamics. Second,
there is ample evidence for climate behavior on decadal
scales (15–30-yr period) in both the instrumental (Cayan
et al. 1998; Dettinger et al. 1998; Cayan and Webb,
1992; Webb and Betancourt 1992) and tree-ring records
of the American Southwest (Fritts 1991; Cleaveland et
al. 1992; Dettinger et al. 1998; Grissino-Mayer 1996).
Last, local and shorter time frames have dominated em-
pirical and experimental work in ecology, as well as
resource management, while regional and historical per-
spectives have been neglected (Ricklefs 1987; Brown
1995). Experimentation is valuable for clarifying mech-
anisms, but is generally impractical or impossible at
regional or decadal to centennial timescales. In the
Southwest, long-term perspectives on ecosystem re-
sponse to climate and disturbance are abetted by a long
tradition of vegetation monitoring (Goldberg and Turner
1984; Turner 1990; McCormick and Galt 1993; Martin
1986; Avery et al. 1976; Brown and Heske 1990; Pake
and Venable 1995), and the ability to cross-date and
map past climatic variations, disturbances, and tree de-
mography at interannual resolution using tree rings
(Fritts and Swetnam 1989).
Our goals in this paper are to 1) explore the com-
plexity of climate–disturbance relationships, 2) dem-
onstrate the synchroneity and spatial extent of fires and
drought-related plant mortality (and establishment), and
3) use extreme episodes such as the 1950s drought and
the post-1976 climatic shift to illustrate the impacts of
decadal-scale climate on ecosystems and the present
challenge of disentangling natural and cultural factors.
2. Disturbance climates of the American southwest
a. Seasonal variability
In the American southwest, seasonal precipitation is
characterized by a highly variable winter–early spring
(November–March), an arid late spring and foresummer
(April–June), monsoonal rains in July and August, and
a dry autumn (September–October; Fig. 1). In general,
cool season precipitation recharges soil moisture and
controls woody plant growth, whereas summer rains
drive the annual grass production that supports the live-
stock industry. Synergism between seasonal precipita-
tion and lightning activity yields a vigorous fire regime
in late spring–early summer. During the arid foresum-
mer (Fig. 1) fuels are dry, convective storms begin to
generate lightning, and the maximum area is burned in
June. Subsequently, full development of the monsoonal
pattern in July leads to a maximum in numbers of light-
ning-ignited fires, but increasing fuel moisture results
in reduced area burned (Fig. 2).
Although the influence of seasonal climatic variability
on wildland fires is intuitively obvious—droughts are
correlated with fire—the predictive aspects of this as-
sociation are poorly understood. Recently, there has
been a surge of interest and research in fire climatology,
driven in part by recognition of ENSO teleconnections
and concern over potential impacts of future climatic
change on wildfire activity (Simard et al. 1985; Swet-
nam and Betancourt 1990; Brenner 1991; Balling et al.
1992; Flannigan and van Wagner 1991; Johnson and
Wowchuk 1993; Larsen and McDonald 1995; Knapp
1995; Price and Rind 1994). In contrast to traditional
emphasis on short-term and local weather (temperature,
relative humidity, and wind speed) effects on fire (Roth-
ermel 1983), more recent studies focus on relatively
large spatial and long temporal scales, that is, regional
to continental and seasonal to decadal.
In the Southwest, plant communities experience ep-
isodic recruitment and mortality of seedlings and adults
(Betancourt et al. 1993). To survive its first year, a new
seedling must endure frost and seasonal drought. If win-
ter is dry, there may be insufficient soil moisture for
fast growth in the spring–early summer. If summer rains
fail, the arid foresummer extends into the hottest time
of the year, and even the hardiest of seedlings will wither
and die. One outcome of seasonal or longer droughts is
a lull in plant recruitment. Conversely, wet and/or warm
conditions favor above-average seed germination and
survival of seedlings leading to recruitment pulses.
These pulses may be dominated by single-year events,
such as the year 1919 in many Southwestern ponderosa
pine stands (Pearson 1950; Savage et al. 1996), or they
may be clustered in favorable decades (White 1985).
3130 V
OLUME
11JOURNAL OF CLIMATE
F
IG
. 1. Three-dimensional time series plots of monthly precipitation totals from Tucson, Arizona, 1870–1995 (upper plot), and Las Cruces,
New Mexico, 1853–1995 (lower plot). Areas shaded black are months with less than approximately 2.5 cm of precipitation. Note the
persistence of winter and summer drought in Las Cruces during the 1950s and the post-1976 increase in cool season precipitation in both
Tucson and Las Cruces.
D
ECEMBER
1998 3131SWETNAM AND BETANCOURT
F
IG
. 2. Monthly distribution of numbers and area burned by light-
ning fires in Arizona and New Mexico national forests, 1940–74
(Barrows 1978).
b. Interannual and interdecadal climate variability
Interannual variability in southwestern precipitation
is linked to shifts in the upper-air westerlies and ENSO.
During the warm SO phase, warm waters in the eastern
Pacific provide the necessary energy for development
of troughs along the west coast of North America; the
warm waters also weaken the trade wind inversion, re-
sulting in stronger subtropical westerlies. Although
there appears to be no single canonical response in the
polar jet stream, the midlatitude winter storm track is
usually displaced southward during El Nin˜o episodes.
In general, an expanded circumpolar vortex, with
strengthened westerly winds centered about 308N and
positive 700-mb height anomalies at higher latitudes,
prevails during decades when the Southern Oscillation
approaches a biennial cycle (Cayan and Webb 1992;
Webb and Betancourt 1992; Cayan et al. 1998). In the
Southwest, El Nin˜o (La Nin˜a) conditions are associated
with wetter (drier) winters and springs (Andrade and
Sellers 1988) and drier (wetter) summers in year 11
after the onset of El Nin˜o (Harrington et al. 1992).
Teleconnections between the tropical Pacific and
American southwest have been shown by correlations
between indices such as Southern Oscillation index,
equatorial sea surface temperatures, and Line Island
rainfall against southwestern precipitation (Douglas and
Englehart 1984; Andrade and Sellers 1989; Cayan and
Webb 1992), streamflow (Cayan and Webb 1992; Webb
and Betancourt 1992; Kahya and Dracup 1994), and
annual area burned in Arizona and New Mexico (Swet-
nam and Betancourt 1990; Price and Rind 1994). These
teleconnections tend to be lagged, promising some pre-
dictive capability a season or more in advance. Although
linear regressions provide weak predictive capability
(about 25%–30% of the variance) for precipitation,
streamflow, and fire, the mesoscale responses to extreme
phases of the SO are highly consistent. This predict-
ability, for example, allows managers to schedule con-
trol fires or to improve fire readiness depending on the
onset of El Nin˜o or La Nin˜a conditions, respectively
(Swetnam and Betancourt 1990). For example, based
on La Nin˜a conditions in fall 1995 and National Oceanic
and Atmospheric Administration (NOAA) forecasts
(Climate Prediction Center) for a dry Southwestern win-
ter and spring, emergency funds were obtained and used
to combat a heavy 1996 fire season (Albuquerque Jour-
nal, 14 April 1996, and The New York Times, 11 May
1996).
Evidence for climate behavior on interdecadal scales
(15–30-yr period) has been demonstrated empirically
from historical climate datasets (Namias et al. 1988;
Rasmusson et al. 1990; Halpert and Ropelewski 1992;
Xu 1993; Ghil and Vautard 1994; Latif and Barnett
1994; Trenberth and Hurrell 1994; Miller et al. 1994;
Mann et al. 1995; Cayan et al. 1998; Dettinger et al.
1998) and high-resolution (annual) climate proxies in-
cluding tree rings, corals, marine and lake sediment
varves, and ice cores (Ebbesmeyer et al. 1991; Diaz and
Pulwarty 1994; Dunbar et al. 1994; Meko et al. 1993;
Meko et al. 1995; Slowey and Crowley 1994; Stocker
and Mysak 1992; Fritts 1991; Cleaveland et al. 1992;
Stahle and Cleaveland 1993). In the American south-
west, three different tree-ring reconstructions of the
Southern Oscillation suggest that the frequency and am-
plitude of the SO, and/or the strength of its telecon-
nections with winter precipitation in this region, has
varied on decadal to multidecadal timescales (Lough
and Fritts 1985; Michaelsen and Thompson 1992; Stahle
and Cleaveland 1993). During the twentieth century,
decadal trends are evident in time series of monthly
mean sea level pressure in the North Pacific and SOI
(Trenberth and Hurrell 1994). In the Southwest, twen-
tieth century climatic trends stemming from the inter-
decadal behavior of the westerlies and the tropical Pa-
cific include wet winters in the early part of the century
(1905–30), a midcentury dry period (1942–64), and
warm, wet winters and erratic summers since 1976 (Fig.
1).
In sections to follow, we illustrate and discuss a range
of ecological responses to interannual and interdecadal
climatic variability in the American southwest.
3. Regionally synchronized insect outbreaks
Outbreaks of phytophagous insects are one of the
most dramatic and poorly understood mesoscale eco-
logical phenomena (Barbosa and Schultz 1987). The
cyclical or aperiodic eruption of insect populations—
sometimes from rare, endemic levels, to broadscale out-
breaks covering 10
4
–10
5
km
2
—has been known at least
since biblical times (e.g., locust plagues), and associated
anecdotally with drought. The drought connection was
formalized by White (1976) in the ‘‘stress’’ hypothesis,
3132 V
OLUME
11JOURNAL OF CLIMATE
which stated that moisture and heat-stressed plants have
a higher food quality than nonstressed plants. Highly
variable and sometimes contradictory results of both
observational and experimental studies, however, have
fostered continued debate about insect, plant, and
drought interactions (Larsson 1989; Martinat 1987;
Price 1991; Mopper and Whitham 1992). Part of the
difficulty in identifying the role of climatic fluctuations
in outbreak dynamics has been the lack of sufficiently
long and large-scale records of climate or insect pop-
ulations. This is particularly true for major forest de-
foliators, such as the western spruce budworm (Chor-
istoneura occidentalis) and the eastern North American
species (Choristoneura fumiferana), which have re-
peatedly erupted over regional areas at frequencies of
only four or fewer outbreaks in the twentieth century
(Sanders et al. 1985).
Tree-ring reconstructions of multicentury length bud-
worm outbreak histories are possible because past de-
foliation of host trees by the budworm larvae cause
distinctive growth reductions in tree-ring width series
of surviving trees (Blais 1962). When nonhost tree spe-
cies (pines) are sampled in nearby sites and their ring
series examined in comparison with the host tree species
(firs and spruces), effects of climatic variation can be
distinguished from the host-specific defoliation effects
(Swetnam and Lynch 1993). These approaches have
yielded 300-yr reconstructions of budworm outbreaks
in the central and southern Rockies of Colorado and
New Mexico (Swetnam and Lynch 1993; Hadley and
Veblen 1993; Weber and Schweingruber 1995) and the
Blue Mountains of eastern Oregon (Swetnam et al.
1995). Independently derived tree-ring reconstructions
of June Palmer drought severity index (PDSI) from non-
host tree species show that, over the past three centuries,
budworm outbreaks generally coincided with wet pe-
riods, whereas low budworm population levels corre-
sponded to droughts (Fig. 3). These tree-ring recon-
structed patterns are supported by similar comparisons
between twentieth century time series of defoliated area
(from aerial and ground measurements) and meteoro-
logical data in the twentieth century (Swetnam and
Lynch 1993).
The positive association between precipitation or
drought indices and budworm population proxies over
such a long period and large spatial scale supports a
‘‘vigor’’ rather than a stress hypothesis (Price 1991).
This finding is supported by experimental results show-
ing that, in the case of some leaf or bud feeding insects,
food quality and quantity may be enhanced by increased
moisture (Larsson 1989). In contrast, drought inhibits
the resin production that defends a tree against cambium
feeders (e.g., bark beetles) (Craighead 1925; Lorio
1986). For example, bark beetle outbreaks were asso-
ciated with broadscale tree mortality during the 1986–
92 drought in the Sierra Nevada (Ferrell 1996).
Increasing drought stress and susceptibility due to
increasing leaf area is touted as the leading cause of
declining ‘‘forest health’’ in the Mountain West (e.g.,
Langston 1995; Rogers 1996), but clearly, there can be
a variety of responses by mesoscale disturbance agents
to both wet and dry periods. Another layer of com-
plexity is added when we consider that insect outbreaks
also influence subsequent fire activity by increasing
highly flammable dead leaves and woody fuels (Scho-
walter et al. 1986; Stocks 1987). Although these syn-
ergisms are widely noted, their temporal and spatial
dynamics are virtually unexplored and unknown.
4. Regionally synchronized wildfires
In most terrestrial ecosystems, fire is a keystone pro-
cess that is heavily influenced by climatic variability.
Seasonal to interdecadal variability in precipitation,
temperature, wind, and lightning regimes determine fuel
dynamics and ignition rates. The role of climate is most
evident in the recurrence of regionally synchronized
wildfires during particular years. No other factor can
explain the prevalence and synchroneity of such unusual
fire activity.
The American southwest has the longest and most
detailed records of drought and fire history in the world.
Multicentury networks of ring-width and fire-scar data
allow assessment of fire–climate relations at seasonal to
century scales, as well as estimation of the magnitude,
extent and frequency of regional, synchronous fires. For
our analyses we used 13 gridpoint reconstructions of
June PDSIs extending back to 1700. These reconstruc-
tions were derived from more than 100 ring-width chro-
nologies (Meko et al. 1993; Cook et al. 1996) from
throughout the region. Sixty-three fire-scar chronolo-
gies, specifying dates of fires during the last 400 yr,
have been compiled from ponderosa pine, mixed-co-
nifer, and pine-oak woodlands in 25 mountain ranges
(Fig. 4) (Swetnam and Baisan 1996). Dates were de-
termined by obtaining cross sections from 10 or more
fire-scarred trees in each stand of 10–100 ha, crossdating
ring-width patterns, and determining the year of fire-
scarring (formation of heat-caused lesions with the xy-
lem tissue). Fire-scar dates from more than 900 trees
are included in this dataset.
The composite fire-scar record across the Southwest
shows that regionally synchronous fires have recurred
for centuries (Fig. 5). Fire frequency within individual
sites averaged about one fire per 7.5 yr from 1700 to
1900. At this frequency, strictly by chance, we would
expect about one coincidence of the same fire date in
21 of the 63 sites (one-third) in about a 35 000-yr period.
Yet, 15 different years met or exceeded this criterion in
the 201-yr period (Fig. 5). The probability of 41 of 63
sites recording the same fire date by chance, as in 1748,
is astronomically low (one chance in billions)! As in
1910 (Fig. 6), this synchroneity of fires across such a
large region must reflect regional to subcontinental-
scale drought for one or more seasons. Direct compar-
ison of the fire-scar records with independent PDSI re-
