Oecologia DOI 10.1007/s00442-014-2888-8

Community ecology - Original research

Moisture status during a strong El Niño explains a tropical montane cloud forest’s upper limit Shelley D. Crausbay · Abby G. Frazier · Thomas W. Giambelluca · Ryan J. Longman · Sara C. Hotchkiss 

Received: 30 July 2013 / Accepted: 11 January 2014 © Springer-Verlag Berlin Heidelberg 2014

Abstract Growing evidence suggests short-duration climate events may drive community structure and composition more directly than long-term climate means, particularly at ecotones where taxa are close to their physiological limits. Here we use an empirical habitat model to evaluate the role of microclimate during a strong El Niño in structuring a tropical montane cloud forest’s upper limit and composition in Hawai‘i. We interpolate climate surfaces, derived from a high-density network of climate stations, to permanent vegetation plots. Climatic predictor variables include (1) total rainfall, (2) mean relative humidity, and (3) mean temperature representing non-El Niño periods and a strong El Niño drought. Habitat models explained species composition within the cloud forest with non-El Niño rainfall; however, the ecotone at the cloud forest’s upper Communicated by Tim Seastedt. Electronic supplementary material The online version of this article (doi:10.1007/s00442-014-2888-8) contains supplementary material, which is available to authorized users. S. D. Crausbay · S. C. Hotchkiss  Department of Botany, University of Wisconsin-Madison, Madison, WI 53706, USA S. D. Crausbay · S. C. Hotchkiss  Center for Climatic Research, University of Wisconsin-Madison, Madison, WI 53706, USA Present Address: S. D. Crausbay (*)  Department of Horticulture and Landscape Architecture, Colorado State University, Fort Collins, CO 80523, USA e-mail: [email protected] A. G. Frazier · T. W. Giambelluca · R. J. Longman  Department of Geography, University of Hawai‘i-Ma¯ noa, Honolulu, HI 96822, USA

limit was modeled with relative humidity during a strong El Niño drought and secondarily with non-El Niño rainfall. This forest ecotone may be particularly responsive to strong, short-duration climate variability because taxa here, particularly the isohydric dominant Metrosideros polymorpha, are near their physiological limits. Overall, this study demonstrates moisture’s overarching influence on a tropical montane ecosystem, and suggests that short-term climate events affecting moisture status are particularly relevant at tropical ecotones. This study further suggests that predicting the consequences of climate change here, and perhaps in other tropical montane settings, will rely on the skill and certainty around future climate models of regional rainfall, relative humidity, and El Niño. Keywords  Drought · Ecotone · El Niño/Southern Oscillation · Habitat model · Hawai‘i

Introduction The relative effect of long-term climate means versus short-duration climate events is a growing research focus in community ecology that includes experimental and observational approaches (Jentsch et al. 2007; Jentsch and Beierhuhnlein 2008; Smith 2011; Lloret et al. 2012; Letten et al. 2013). Climate events can rapidly shift a species’ range limit with record summer heat (Battisti et al. 2006), reorganize species assembly with a flood (Thibault and Brown 2008), and induce widespread tree mortality across vegetation types with a strong drought (Breshears et al. 2005). Including climate variability in species distribution models can significantly improve model fit for tree species distribution (Zimmerman et al. 2009). Species distribution models of an endangered marsupial that include weather

13

Oecologia

events plus climate means suggest this species’ distribution is driven by short-duration drought and low-temperature events (Bateman et al. 2012). Likewise, distribution models based on climate means coupled with simulations of drought events show that while populations appear resistant to shifts in mean climate, they are vulnerable to increased drought frequency (Barrows et al. 2010). In addition, distribution models show that El Niño-induced droughts were associated with lower richness and endemism in Borneo (Raes et al. 2009). Overall, these empirical modeling studies suggest that short-duration climate events can strongly affect plant distribution, survival, and diversity (Reyer et al. 2013). Ecotones may be particularly responsive to short-duration climate events because many taxa at these range margins are at their physiological limits (Allen and Breshears 1998; Parmesan et al. 2000; Will et al. 2013). For example, a dynamic global vegetation model driven by gridded climate observations shows that the locations of major forestgrassland ecotones globally are largely attributable to climate variability, where increased variability reduces forest cover and increases grass cover by affecting relative growth rates, precipitation-interception rates, respiration costs, and rooting depths (Notaro 2008). In another forest-grassland ecotone study, increased vapor pressure deficit during an experimental drought intensified physiological stress of tree seedlings and increased desiccation-related mortality (Will et al. 2013). These few studies suggest climate variability and short-term events may be particularly important at ecotones, but our ability to predict where, when, and how climate variability will affect vegetation remains limited. A prominent source of climate variability globally is the El Niño/Southern Oscillation (ENSO), which drives climate extremes including drought, flood, and extreme heat across many different regions. ENSO is the dominant driver of interannual variation in moisture availability in the tropics (Chu and Chen 2005; Dore 2005; Giambelluca et al. 2008; Martínez et al. 2011), and emerging evidence suggests that ENSO-driven moisture variations influence tropical ecosystems and can lead to persistent ecological impacts. Recent studies show that strong ENSO events affect tropical leaf phenology (Pau et al. 2010), flower and seed production (Wright and Calderon 2006), and biomass (Rolim et al. 2005). Strong El Niño droughts have induced higher than background rates of tree mortality across the tropics—in Africa, Southeast Asia, Central and South America (Phillips et al. 2010 and references therein). Continuous, long-term data sets from Barro Colorado Island demonstrate that a single, widespread, ENSO-induced mortality event can herald a persistent long-term change in species composition (Feeley et al. 2011a). Testing whether El Niño events matter to tropical ecosystems is important, but challenging, and will require a combination of long-term

13

data sets, opportunistic observational studies, and empirical distribution models that include ENSO data (e.g., Raes et al. 2009). Growing evidence suggests short-duration climate events may drive community structure and composition far more directly than climate means, particularly at ecotones where taxa are close to their physiological limits. Here we evaluate the role of a strong El Niño drought versus more average non-El Niño periods (e.g., interannual variability) in structuring a tropical montane cloud forest’s (TMCF) upper limit and composition. TMCFs are among the most vulnerable tropical ecosystems to climate change, and they harbor high levels of endemism while providing essential ecosystem services (Loope and Giambelluca 1998; Foster 2001). Our study is based in Hawai‘i (Fig.  1a), where ENSO drives strong temperature and rainfall anomalies during the winter season and El Niño has a well-established relationship with winter drought (Chu 1989; Chu and Chen 2005). We hypothesized that a strong El Niño would be more closely associated with the ecotone at the upper limit of cloud forest, relative to species composition within the cloud forest, because species at the ecotone are close to their physiological limits. We used a 15-station climate network, distributed across a small area that encompasses a broad climate gradient (~7,000–3000 mm of rainfall), to create climate surfaces for mean temperature, total rainfall, and mean relative humidity from (1) a strong El Niño winter, and (2) non-El Niño periods. Because El Niño affects the entire winter season, we focus on mean or total microclimate characteristics during this time period, rather than more traditional measures of an extreme climatic event (e.g., Smith 2011). We interpolated the climate surfaces to vegetation plots, and then modeled the distribution of (1) species composition within the cloud forest, and (2) the elevation of the cloud forest’s upper elevational limit. This approach allows us to better understand the current climate controls on TMCF in Hawai‘i and is a necessary step before predicting future distributions and informing conservation planning.

Materials and methods Study area Haleakala¯ volcano rises to 3,055 m and our study area (20.7–20.8°N, 156.1–156.2°W), from 1,900 to 2,300 m elevation on the windward, northeast slope, brackets the cloud forest’s upper limit (Fig. 1b, c). Microclimate variation is driven by north-easterly trade winds, which bring humid air masses, clouds, and orographic rainfall to portions of the slope below the trade wind inversion

Oecologia Fig. 1  Map of a the Hawaiian Islands showing locations of b Haleakala¯ volcano, East Maui, Hawai‘i and c close-up of the study area on the northeast windward slope. Symbols represent micrometeorological stations located from 2,470 to 1,650 m (stars) and vegetation plot locations [cloud forest (black circles), subalpine shrubland (white squares)]. Forest line occurs between adjacent white squares and black circles, at ~2,200 m in the east and ~2,000 m in the west. Map in c shows position in the Universal Transverse Mercator coordinate system and highlights the east– west sections (eastern, central, and western) that correspond to the stratified group of elevational transects. Elevation for the local peak, Pohaku Palaha (20.73° N; 156.14° W), and highest micrometeorological station at 2,470 m is shown

a

22 o N

b

Maui

21o N

20 o N

19o N

Easting (km)

c 797

798

799

800

801

802

803

Western Central Eastern

(TWI). The TWI, a synoptic subsidence inversion that establishes a sharp decrease in relative humidity above its mean base height of 2,076–2,255 m across the Hawaiian Islands (Cao et al. 2007), corresponds to the elevational range of the cloud forest’s upper limit (Kitayama and Mueller-Dombois 1992; Crausbay and Hotchkiss 2010). A climate station monitored since 1992 shows that strong El Niño events induce anomalously low winter rainfall and relative humidity and high temperature near this ecotone (Fig. 2).

The cloud forest is strongly dominated (>80 % cover) by Metrosideros polymorpha Gaudich. Above the forest’s upper limit, a subalpine shrubland is dominated by the shrub Leptecophylla tameiameiae (Cham. and Schltdl.) C.M. Weiller, the tree fern Sadleria cyatheoides Kaulf, and the tussock grass Deschampsia nubigena Hillebr. Vegetation here is thought to be strongly driven by climate gradients, with little association between edaphic variables (substrate age, texture, and slope) and species composition or ecotone position (Crausbay and Hotchkiss 2010).

13

Oecologia 30

a

Relative humidity

Vegetation structure and composition were quantified in 98 plots (Fig. 1c) which comprise 15-m-long transects that were established along nine elevational transects with a stratified (in three east–west sections; Fig. 1c) random approach (Crausbay and Hotchkiss 2010). The point-intercept method (Levy and Madden 1933) was used to quantify species presence every 25 cm along each data transect, in five height classes (0–1, 1–2, 2–3, 3–5, and >5 m). All vascular plants were identified to the species or variety level, but bryophytes and lichens were not differentiated further. Forest line position was defined by the sharp discontinuity where the >5-m height class dropped from >60 cover to 5-m height class was chosen because trees near the forest line on Haleakala¯ are 5–8 m tall (Kitayama and Mueller-Dombois 1992).

20

10

0

-10

-20

-30 40

b

Vegetation sampling

Rainfall

30

Anomaly

20

Vegetation response variables 10

0

-10

-20 6

c

Temperature

4

2

0

-2

-4

Non-El Niño Winter season

El Niño Winter season

Fig. 2  Box plots showing the distribution of winter (December–January–February) microclimate anomalies recorded from 1992 to 2011 at a single HaleNet climate station at the forest line ecotone within the study area on northeast windward Haleakala¯, Maui. Data show anomalies for a means for relative humidity, b totals for recorded rainfall, and c means for temperature from non-El Niño winters (left) and from all six strong El Niño winters that have occurred since 1992 (right), including the strong El Niño in the winter of 2009–2010 that is the focus of this study

13

The forest line occurs between each adjacent white square and black circle in Fig. 1c, and sits at ~2,200 m elevation in the eastern section and ~2,000 m in the western section. A subset of plots that evenly bracket the forest line was used in the analysis of forest line—all four elevational positions above forest line (n = 36, white squares; Fig. 1c) and the first four elevational positions below the forest line for each of the nine elevational transects of points (n  = 36, black circles; Fig. 1c). Vegetation plots above the forest line were assigned a value of 0 and those below forest line were assigned a value of 1. Cloud forest analysis includes the subset of all plots between the lowest elevation and the forest line (n  = 62, all black circles; Fig. 1c). To represent cloud forest species composition as a single variable, we used the primary axis from a non-metric multi-dimensional scaling (NMS) ordination. NMS is a multivariate ordination technique that reduces variability in multi (k) dimensional space to fewer orthogonal dimensions (axes) through an iterative process (i.e., 500 iterations, 50 % randomized runs, with instability criterion of 1.0 × 10−7). Beginning with two dimensions, additional dimensions are added only if they reduced the final stress by 5 % or more. Final stress was lower than that for 95 % of the randomized runs (i.e., p  0.4 are indicated. Plots on the positive end of axis 1 generally occur on the wetter, eastern end of the study area (Fig. 3c). Taxonomy for flowering plants is based on Wagner et al. (1999); taxonomy for ferns is based on Palmer (2002)

2040 2010

Cloud forest species composition (NMS axis 1 score)

1980

1.50

E Epiphyte, EF epiphytic fern, F terrestrial fern, G graminoid, H herb, S shrub, T tree, TF tree fern

c

(Crausbay and Hotchkiss 2010). The gradient in cloud forest species composition is described by a three-dimensional NMS ordination and axis 1 explains 40 % of the variation in species composition (Table 1). Axis 1 generally distinguishes plots from east to west, with positive scores in more eastern, wetter plots (Fig. 3a, c). Epiphytic ferns and bryophytes load on the positive, wetter end of axis 1, whereas tree ferns and graminoids load on the negative, drier end of axis 1 (Table 2).

0.75

0.00

-0.75

-1.50

Microclimate predictor variables 798

799

800

801

Easting (km)

Fig. 3  Relationship between east–west landscape position and a total annual rainfall during a non-El Niño period, interpolated to each vegetation plot, b elevation of the cloud forest’s upper limit (forest line), described in Crausbay and Hotchkiss (2010) and c score along axis 1 of a non-metric multidimensional scaling (NMS) ordination of cloud forest species abundance (see Table 1). Three east–west sections (eastern, central, and western) correspond to the stratified group of elevational transects in Fig. 1c

Vegetation response variables Forest line occurs between 2,000 and 2,200 m elevation and is oriented approximately east–west (Figs. 1c, 3b)

NPMR selected predictor variables to represent non-El Niño periods for total rainfall, mean relative humidity, and mean air temperature, based on cross-validated fit, separately for forest line and cloud forest models (Table 3). For the cloud forest model, NPMR selected DJF 2007–2008 to represent non-El Niño total rainfall, SON 2009 to represent non-El Niño mean relative humidity, and mean annual data from 2009 to represent non-El Niño mean annual temperature. For the forest line model, NPMR selections were the same for non-El Niño total rainfall and mean annual temperature, but NPMR selected MAM 2009 to represent non-El Niño mean relative humidity. For both models, we represent a strong El Niño with total rainfall, mean relative humidity, and mean temperature from DJF 2009 to 2010. Collinearity between these six predictor variables

13

Oecologia Table 3  Fit between microclimate predictor variables from non-El Niño periods and the quantitative response variable describing cloud forest (CF) species composition (NMS axis 1; Table 1) and the binary forest line (FL) response variable on windward Haleakala¯ Rainfall

CF × R2

FL log B

Relative humidity

CF × R2

FL log B

Temperature

CF × R2

FL logB

SON-2005 JJA-2007 SON-2007 DJF-2007–2008 MAM-2008 JJA-2008 SON-2008 DJF-2008–2009 Annual-2008 MAM-2009 JJA-2009 SON-2009 Annual-2009 MAM-2010 Contin. rainless days

0.48 0.48 0.48 0.52a 0.49 0.51 0.50 0.48 0.50 0.49 0.50 0.48 0.49 0.51 0.47

0.55 0.37 1.80 9.56a 0.52 5.22 2.63 2.62 5.79 1.12 4.41 1.53 2.01 4.33 3.42

DJF-2008–2009 MAM-2009 JJA-2009 SON-2009 Annual-2009 MAM-2010

0.45 0.20 0.42 0.45a 0.43 0.42

4.31 10.23a 5.87 5.39 8.09 6.76

DJF-2008–2009 MAM-2009 JJA-2009 SON-2009 Annual-2009 MAM-2010

0.39 0.25 0.32 0.36 0.42a 0.35

5.35 5.91 4.77 0.74 6.86a 4.51

Rainless days/year

0.49

3.42

Fit is estimated for each predictor variable individually with non-parametric multiplicative regression using the cross-validated pseudo-R2 (×R2) for the CF species composition or the log10 of likelihood ratio (log B) for the binary FL. Larger log B and ×R2 signify a better fit. Predictor variables include annual and seasonal data during non-El Niño periods for rainfall (since 2005), relative humidity (since 2008), and temperature (since 2008). Seasonal variables represent 3-month periods; December–January–February (DJF), March–April–May (MAM), June–July–August (JJA), September–October–November (SON) a

  Variables with the best fit

Habitat models The cloud forest model started with six microclimate predictor variables—total rainfall, mean relative humidity, mean air temperature representing non-El Niño periods, and total rainfall, mean relative humidity, and mean air temperature from the 2009 to 2010 strong El Niño winter. NPMR modeled axis 1 of the cloud forest ordination (Fig.  3c) with a simple one-predictor model, non-El Niño rainfall, with an ×R2  = 0.51, N*  = 14.4, and P  = 0.001 (Fig. 4). Because non-El Niño rainfall is highly correlated with another moisture-based predictor variable, non-El Niño relative humidity (r = 0.91; Online Resource 4), we interpret this model as a strong relationship between the cloud forest’s species composition and moisture status during non-El Niño periods.

13

1.5

Cloud forest species composition (NMS axis 1 score)

was lowest between temperature and moisture-based variables, with an absolute value of r ranging from 0.15 to 0.69 (mean = 0.48) during non-El Niño periods and 0.32–0.81 (mean  = 0.59) during El Niño (Online Resources 4–5). Collinearity was highest between the moisture-based variables rainfall and relative humidity with r  = 0.91 during non-El Niño periods for the cloud forest model and r  = 0.91 and 0.92 during El Niño for the forest line and cloud forest models, respectively (Online Resources 4–5).

1.0

0.5

0.0

-0.5

-1.0

-1.5 2200 2400 2600 2800 3000 3200 3400 3600 3800

Non-El Niño total rainfall (mm)

Fig. 4  Fitted relationship between cloud forest species composition on northeast Haleakala¯ (Table 1; Fig. 3c) and total rainfall during non-El Niño periods, with 5th and 95th percentile variability bands. The model began with the quantitative response variable (axis 1 of a NMS ordination) and six microclimate predictor variables

The forest line model also started with six microclimate predictor variables—total rainfall, mean relative humidity, mean air temperature representing non-El Niño

Oecologia

periods, and total rainfall, mean relative humidity, and mean air temperature from the 2009 to 2010 strong El Niño event. The modeling process selected two predictor variables to provide a robust model of forest line with a logB = 15.95, AUC = 0.98, COR = 0.88 and P = 0.001 (Fig.  5). The two predictor variables include (1) relative humidity during a strong El Niño, and (2) rainfall during non-El Niño periods. The most important of these predictors was relative humidity during the strong El Niño (Fig.  5). El Niño relative humidity is highly correlated with another moisture-based predictor variable, El Niño rainfall (r  = 0.91: Online Resource 5); therefore, we interpret this model as a strong relationship between the ecotone and moisture status during El Niño. The NPMR response contours highlight two rather distinct sections of the ecotone, explained by non-El Niño rainfall (which is highly correlated with non-El Niño relative humidity, r  = 0.92; Online Resource 5; Fig. 5). Samples from the drier, western ecotone inhabit areas with higher humidity during El Niño, and samples from the wetter, eastern and central ecotone inhabit areas with lower humidity during El Niño (Fig. 5).

Non-El Niño total rainfall (mm)

3400

Subalpine shrubland Cloud forest

3200 Eastern and Central

3000

2800

2600 Western

2400

50

52

54

56

58

60

62

El Niño mean relative humidity (%)

Fig. 5  Response contour map from a two-predictor non-parametric multiplicative regression (NPMR) model of forest line on northeast Haleakala¯ , with observed points overlaid. The model was developed with a binary forest or not-forest response variable and six microclimate predictor variables. The gradient from gray to black indicates likelihood of forest occurrence, with the grayest shade indicating the most favorable climate space for forest occurrence. Sensitivity measures of predictor variables describe relative importance (highest  = most important); El Niño mean relative humidity = 1.57 and non-El Niño total rainfall = 0.48. This two-predictor NPMR model had a log10 of likelihood ratio = 15.95, area under the receiver operating characteristic curve = 0.98, Pearson correlation coefficient = 0.88, minimum average neighborhood size = 4.9, P = 0.001

Discussion Influence of a strong El Niño event The cloud forest’s upper limit on windward Haleakala¯ is best explained by moisture status during a strong El Niño (Fig. 5). Previous work showed that ecotones may be particularly responsive to climate events because many taxa at these range margins are near their physiological limits (Allen and Breshears 1998; Parmesan et al. 2000; Notaro 2008; Will et al. 2013). We hypothesized that this ecotone would be associated with El Niño-induced drought because plants here are indeed near their physiological limits. Several lines of evidence supported this hypothesis. First, nearly 40 % of the plant taxa in this study reach their elevational limits at the forest line and nearly 50 % are significant indicators to positions above or below the forest line (Crausbay and Hotchkiss 2010). Second, sharp changes in atmospheric conditions near the TWI drive high evaporative demand at the ecotone. Data since 1992 show that climate at the ecotone differs from lower elevations within the cloud forest zone in that average potential evapotranspiration is 33 % higher, vapor pressure deficit is 70 % higher, and relative humidity is 16 % lower (R. J. Longman, unpublished data). Third, a recent study of M. polymorpha, the dominant cloud forest canopy tree, across this ecotone found clear differences in water relations and showed that high vapor pressure deficit affects stomatal conductance and transpiration, particularly above the cloud forest boundary where greater atmospheric demand causes water loss despite stomatal regulation (Gotsch et al., in review). M. polymorpha is an isohydric plant (Cornwell et al. 2007) which can respond to water stress by closing its stomates to reduce water loss, but this action may eventually lead to mortality through carbon starvation (e.g., McDowell et al. 2008). Further, atmospheric demand for water is the most significant source of abiotic stress near the ecotone, as temperatures here do not reach levels that cause freezing stress (Kitayama and Mueller-Dombois 1992; Melcher et al. 2000). Cloud-forest upper limits in the trade-wind belt cooccur with the mean TWI on multiple high mountains—in Hawai‘i, the Canary Islands, the Dominican Republic, and Malaysia (Kitayama and Mueller-Dombois 1992; Fernández-Palacios and de Nicolás 1995; Kitayama 1995; Martin et al. 2007; Crausbay and Hotchkiss 2010). As a result, research has suggested the dramatic elevational change in relative humidity that characterizes mean climate near the TWI controls this ecotone. Our data suggest moisture status may in fact control this cloud forest’s upper limit, but it does so through strong El Niño-driven droughts. Proximity to the sharp changes in atmospheric conditions around the mean TWI puts species closer to their physiological

13

Oecologia

limits, likely rendering them more vulnerable to shortduration climate events. In addition, El Niño causes a lower TWI height and more frequent TWI presence (Cao et al. 2007). The low relative humidity signature during El Niño reported here (Online Resource 2) likely reflects the TWI’s greater persistence and lower elevation. Whenever the TWI moves to a lower elevation, it causes a sudden, sharp decline in relative humidity at higher elevations, which in turn causes isohydric plants to close their stomates, cease gas exchange, and conserve water. Our data imply that moisture conditions during past strong El Niño periods structured this sharp ecotone boundary, and ultimately suggest that ENSO-TWI interactions may set the cloud forest’s upper limit on mountains in the trade-wind belt. Moisture status Moisture status, as indicated by rainfall and relative humidity, during the strong El Niño was the best predictor of the ecotone, but it was followed by moisture conditions during non-El Niño periods as an additional explanatory variable (Fig. 5). Phillips et al. (2010) report that long-term mean rainfall modulates tropical trees’ response to severe droughts, such that drier forests are most vulnerable. We see a similar pattern across this ecotone, where the wetter eastern forest’s ecotone is defined at 54 % relative humidity during an El Niño drought and the somewhat drier western forest’s ecotone is defined at 58 % relative humidity, suggesting the drier ecotone section (in terms of mean rainfall) is vulnerable to a less severe drought. This pattern supports suggestions by Crausbay and Hotchkiss (2010) that ecotone response to climate change may vary from place to place and may be predicted by position along the mean cross-slope rainfall gradient. Microclimate during a strong El Niño emerges as an important explanatory variable only for the cloud forest’s upper limit and not for species composition within the cloud forest (Figs. 4, 5). Moisture during more average non-El Niño conditions explained this TMCF’s species composition, likely because these species are not as close to their physiological limits as those near the ecotone (e.g., Allen and Breshears 1998). Previous work suggested that this cloud forest’s composition and upper limit are driven primarily by moisture, rather than temperature (Crausbay and Hotchkiss 2010). However, this suggestion had not been tested with quantitative microclimate data before now. In the cloud forest, species composition clearly follows a cross-slope rainfall gradient that persists seasonally and annually during non-El Niño periods (Figs. 3, 4). Cloud forest species composition is characterized by more tree ferns in the drier western forest, and higher tree diversity and more epiphytes in the wetter eastern forest (Table  2). The strong relationship between composition

13

and moisture status may seem counterintuitive in a wet site (7,755–3,100 mm rainfall annually, east to west). However, the soils here are highly responsive to rainfall and exhibit a rapid drying response to even a single day with little rainfall (S. D. Crausbay, unpublished data). Previous work found that episodic soil drying occurs at this site, particularly after rainless periods, and that dry soils can induce wilting in M. polymorpha trees (Menard 1999). Together these data indicate that interactions between atmospheric moisture status and soil water availability likely structure this TMCF’s species composition. In addition, the study covers ~250 m of elevation within the cloud forest, equivalent to ~1.5 °C mean temperature difference and similar to the best estimate for warming at the end of the twentyfirst century in the Intergovernmental Panel on Climate Change’s (IPCC) AR4 B1 emission scenario (IPCC 2007). Temperature was not associated with species composition at this scale and was not well correlated with other moisture-related variables (Online Resource 4), suggesting the relationship between TMCF species composition and moisture is robust. This study suggests that moisture is incredibly relevant to tropical montane ecosystems, and short-term climate events are particularly relevant at tropical ecotones. The overall effort to quantify risk to tropical montane ecosystems associated with changes in moisture status has paled relative to efforts focused on rising temperature—for good reason. Relative to temperature, accurate precipitation data are scarce, patterns across elevation are complex, and future projections are highly uncertain. However, in one study of bioclimate models that included precipitation change on mountains, extirpation risk increased ten-fold relative to temperature-based models (McCain and Colwell 2011). Extirpation risk in these models of montane vertebrate populations generally increased with warming and drying, but regardless of the direction of projected precipitation trends, the spatial discordance between temperature and precipitation trends elevated risk considerably above temperature changes alone (McCain and Colwell 2011). Tropical taxa are migrating upslope in response to warmer temperatures at rates of 2–9 m year−1 across a broad spread of different kingdoms and classes in the Andes, Malaysia, and Madagascar (Raxworthy et al. 2008; Chen et al. 2009; Feeley et al. 2011b). However, moisture availability can powerfully constrain upslope migration, particularly on mountains that exhibit opposite temperature and precipitation changes. For example, recent high-elevation warming rates in Hawai‘i are among the fastest globally (0.27 °C/ decade)—due to a combination of global warming and a warm phase of the Pacific decadal oscillation (Giambelluca et al. 2008)—but plant species have shown no significant upslope migration in the past 50 years, likely due to increased aridity (Juvik et al. 2011). In fact, increasing

Oecologia

frequency of water stress is causing climate-associated decline in plants such as the iconic alpine Haleakala¯ silversword, perhaps driven ultimately by carbon starvation (Krushelnycky et al. 2013). Incorporating moisture in studies of climate change impacts is essential for anticipating the response of tropical montane species to climate change, especially on mountains with conflicting trends in future temperature and precipitation, as is often the case with mountains under the influence of the TWI.

Conclusion We constructed habitat models with microclimate data interpolated to vegetation plots to show that moisture during a strong El Niño defines the ecotone position on this tropical mountain, suggesting that this and perhaps other ecotones where plants are near their physiological limits are vulnerable to short-duration climate events. This study further demonstrates moisture’s overarching influence on tropical montane ecosystems and stands in contrast to the assumption that the TMCF in Hawai‘i will migrate upslope in response to warming temperatures. This research instead suggests that predicting the consequences of climate change here will rely on the skill and certainty around future climate models of regional rainfall, relative humidity, and ENSO. Acknowledgments This work was supported by the USGS Biological Resources Discipline Global Change Research Program, the USGS Pacific Island Ecosystems Research Center, US Fish and Wildlife Service in support of the Pacific Islands Climate Change Cooperative (PICCC; award number 12170-B-G100), and a National Science Foundation Dissertation Improvement grant (award number DEB-0808466). Partial support of HaleNet field observations and data management was provided through NSF EPSCoR 0903833 (J. Gaines, PI). We thank Haleakala¯ National Park, the Hanawī Natural Area Reserve, Corie Yanger, Gregor Schuurman, Michael Nullet, John DeLay, Lloyd Loope, Sarah Bogen, and Janice Poehlman for assistance. Anonymous reviewers, Jennifer Schmitz, Gregor Schuurman, Monica Turner, Michael Tweiten, and Jack Williams provided helpful comments on this manuscript. Data collection complies with the current laws of the USA in which the collection was performed.

References Allen CD, Breshears DD (1998) Drought-induced shift of a forestwoodland ecotone: rapid landscape response to climate variation. Proc Natl Acad Sci 95:14839–14842 Barrows CW, Rotenberry JT, Allen MF (2010) Assessing sensitivity to climate change and drought variability of a sand dune endemic lizard. Biol Conserv 143:731–736 Bateman BL, VanDerWal J, Johnson CN (2012) Nice weather for bettongs: using weather events, not climate means, in species distribution models. Ecography 35:306–314 Battisti A, Stastny M, Buffo E, Larsson S (2006) A rapid altitudinal range expansion in the pine processionary moth produced by the 2003 climatic anomaly. Glob Change Biol 12:662–671

Breshears DD, Cobb NS, Rich PM, Price KP, Allen CD, Balic RG, Romme WH, Kastens JH, Floyd ML, Belnap J, Anderson JJ, Myers OB, Meyer CW (2005) Regional vegetation die-off in response to global-change-type drought. Proc Natl Acad Sci 102:15144–15148 Cao G, Giambelluca TW, Stevens DE, Schroeder T (2007) Inversion variability in the Hawaiian trade wind regime. J Clim 20:1145–1160 Chen I-C, Shiu H-J, Benedick S, Holloway JD, Chey VK, Barlow HS, Hill JK, Thomas CD (2009) Elevation increases in moth compositions over 42 years on a tropical mountain. Proc Natl Acad Sci 106:1479–1483 Chu P-S (1989) Hawaiian drought and the southern oscillation. Int J Climatol 9:619–631 Chu P-S, Chen H (2005) Interannual and interdecadal rainfall variations in the Hawaiian Islands. J Clim 18:4796–4813 Cornwell WK, Bhaskar R, Sack L, Cordell S, Lunch CK (2007) Adjustment of structure and function of Hawaiian Metrosideros polymorpha at high vs. low precipitation. Funct Ecol 21:1063–1071 Crausbay SD, Hotchkiss SC (2010) Strong relationships between vegetation and two perpendicular climate gradients high on a tropical mountain in Hawai‘i. J Biogeogr 37:1160–1174 Da Silva SC (2012) High altitude climate of the Island of Hawai‘i. Master thesis, Department of Meteorology, University of Hawai‘i at Ma¯ noa, Honolulu Dore MHI (2005) Climate change and changes in global precipitation patterns: what do we know? Environ Int 31:1167–1181 Feeley KJ, Davies SJ, Perez R, Hubbell SP, Foster RB (2011a) Directional changes in the species composition of a tropical forest. Ecology 92:871–882 Feeley KJ, Silman MR, Bush MB, Farfan W, Cabrera KG, Malhi Y, Meir P, Revilla NS, Quisiyupanqui MNR, Saatchi S (2011b) Upslope migration of Andean trees. J Biogeogr 38:783–791 Fernández-Palacios JM, de Nicolás JP (1995) Altitudinal pattern of vegetation variation on Tenerife. J Veg Sci 6:183–190 Foster P (2001) Potential negative impacts of global climate change on tropical montane cloud forests. Earth Sci Rev 55:73–106 Frazier AG (2012) Month–year rainfall maps of the Hawaiian Islands. Master thesis, Department of Geography, University of Hawai‘i at Ma¯ noa, Honolulu Giambelluca TW, Diaz HF, Luke MSA (2008) Secular temperature changes in Hawai‘i. Geophys Res Lett 35:L12702 IPCC (2007) Climate change 2007: the scientific basis, technical summary of the Working Group I report. Cambridge University Press, Cambridge Jentsch A, Beierkuhnlein C (2008) Research frontiers in climate change: effects of extreme meteorological events on ecosystems. Geoscience 340:621–628 Jentsch A, Kreyling J, Beierkuhnlein C (2007) A new generation of climate change experiments: events, not trends. Front Ecol Environ 5:315–324 Juvik JO, Rodomsky BT, Price JP, Hansen EW, Kueffer C (2011) “The upper limits of vegetation on Mauna Loa, Hawaii”: a 50thanniversary reassessment. Ecology 92:518–525 Kitayama K (1995) Biophysical conditions of the montane cloud forests of Mount Kinabalu, Sabah, Malaysia. In: Bruijnzeel LA, Scatena FN, Hamilton JO (eds) Tropical montane cloud forests. Cambridge University Press, Cambridge, pp 183–197 Kitayama K, Mueller-Dombois D (1992) Vegetation of the wet windward slope of Haleakala, Maui, Hawaii. Pacif Sci 46:197–220 Krushelnycky PD, Loope LL, Giambelluca TW, Starr F, Starr K, Drake DR, Taylor AD, Robichaux RH (2013) Climateassociated population declines reverse recovery and threaten future of an iconic high-elevation plant. Glob Change Biol 19:911–922

13

Oecologia Letten AD, Ashcroft MB, Keith DA, Gollan JR, Ramp D (2013) The importance of temporal climate variability for spatial patterns in plant diversity. Ecography 36:001–009 Levy EG, Madden EA (1933) The point method of vegetation analysis. N Z J Agric 46:267–279 Lloret F, Escudero A, Iriondo JM, Martínez-Vilalta J, Valladares F (2012) Extreme climatic events and vegetation: the role of stabilizing processes. Glob Change Biol 18:797–805 Loope LL, Giambelluca TW (1998) Vulnerability of Island tropical montane cloud forests to climate change, with special reference to East Maui, Hawai‘i. Clim Change 39:503–517 Mair A, Fares A (2011) Comparison of rainfall interpolation methods in a mountainous region of a Tropical Island. J Hydrol Eng 16:371–383 Martin PH, Sherman RE, Fahey TJ (2007) Tropical montane forest ecotones: climate gradients, natural disturbance, and vegetation zonation in the Cordillera Central, Dominican Republic. J Biogeogr 34:1792–1806 Martínez R, Ruiz D, Andrade M, Blacutt L, Pabón D, Jaimes E, León G, Villacís M, Quintana J, Montealegre E, Euscátegui C (2011) Synthesis of the climate of the Tropical Andes. In: Herzog SK, Martínez R, Jørgensen PM, Tiessen H (eds) Climate change and biodiversity in the Tropical Andes. Inter-American Institute for Global Change Research (IAI) and Scientific Committee on Problems of the Environment (SCOPE), pp 97–109 McCain CM, Colwell RK (2011) Assessing the threat to montane biodiversity from discordant shifts in temperature and precipitation in a changing climate. Ecol Lett 14:1236–1245 McCune B (2006) Nonparametric habitat models with automatic interactions. J Veg Sci 17:819–830 McDowell N, Pockman WT, Allen CD, Breshears DD, Cobb N, Kolb T, Plaut J, Sperry J, West A, Williams DG, Yepez EA (2008) Mechanisms of plant survival and mortality during drought: why do some plant survive while others succumb to drought? New Phytol 178:719–739 Melcher PJ, Cordell S, Jones TJ, Scowcroft PG, Niemenura W, Giambelluca TW, Goldstein G (2000) Supercooling capacity increases from sea level to tree line in the Hawaiian tree species Metrosideros polymorpha. Int J Plant Sci 161:369–379 Menard T (1999) Ecological and hydrological effects of a rainless period on a montane cloud forest treeline on Haleakala¯ , Maui. Master thesis, Department of Geography, University of Hawai‘i at Ma¯ noa, Honolulu Notaro M (2008) Response of the mean global vegetation distribution to interannual climate variability. Clim Dyn 30:845–854 Palmer DD (2002) Hawaii’s ferns and fern allies. University of Hawaii Press, Honolulu Parmesan C, Root TL, Willig MR (2000) Impacts of extreme weather and climate on terrestrial biota. Bull Am Meteorol Soc 81:443–450 Pau S, Okin GS, Gillespie TW (2010) Asynchronous response of tropical forest leaf phenology to seasonal and El Niño-driven drought. PLoS One 5:e11325

13

Phillips OL, van der Heijden G, Lewis SL, López-González G, Aragão LEOC, Lloyd J, Malhi Y, Monteagudo A, Almeida S, Dávila EA, Amaral I, Andelman S, Andrade A, Arroyo L, Aymard G, Baker TR, Blanc L, Bonal D, de Oliveira ACA, Chao K-J, Cardozo ND, da Costa L, Feldpausch TR, Fisher JB, Fyllas NM, Freitas MA, Galbraith D, Gloor E, Higuchi N, Honorio E, Jiménez E, Keeling H, Killeen TJ, Lovett JC, Meir P, Mendoza C, More A, Vargas PN, Patiño S, Peh KS-H, Cruz AP, Prieto A, Quesada CA, Ramírez F, Ramírez H, Rudas A, Salamão R, Schwarz M, Silva J, Silveira M, Slik JWF, Sonké B, Thomas AS, Stropp J, Taplin JRD, Vásquez R, Vilanova E (2010) Drought-mortality relationships for tropical forests. New Phytol 187:631–646 Raes N, Ferry Slik JW, van Loon E, ter Steege H (2009) Botanical richness and endemicity patterns of Borneo derived from species distribution models. Ecography 32:180–192 Raxworthy CJ, Pearson RG, Rabibisoa N, Rakotondrazafy AM, Ramanamanjato J-B, Raselimanana AP, Wu S, Nussbaum RA, Stone DA (2008) Extinction vulnerability of tropical montane endemism from warming and upslope displacement: a preliminary appraisal for the highest massif in Madagascar. Glob Change Biol 14:1703–1720 Reyer CPO, Leuzinger S, Rammig A, Wolf A, Bartholomeus RP, Bonfante A, de Lorenzi F, Dury M, Gloning P, Jaoudé RA, Klein T, Kuster TM, Martins M, Niedrist G, Riccardi M, Wohlfahrt G, de Angelis P, de Dato GA, François L, Menzel A, Pereira M (2013) A plant’s perspective of extremes: terrestrial plant responses to changing climate variability. Glob Change Biol 19:75–89 Rolim SG, Jesus RM, Nascimento HEM, do Couto HTZ, Chambers JQ (2005) Biomass change in an Atlantic tropical moist forest: the ENSO effect in permanent sample plots over a 22-year period. Oecologia 142:238–246 Smith MD (2011) The ecological role of climate extremes: current understanding and future prospects. J Ecol 99:651–655 Thibault KM, Brown JH (2008) Impact of an extreme climatic event on community assembly. Proc Natl Acad Sci 105:3410–3415 Wagner WL, Herbst DR, Sohmer SH (1999) Manual of the flowering plants of Hawaii, revised edition. Bishop Museum Press, Honolulu Will RE, Wilson SM, Zou CB, Hennessey TC (2013) Increased vapor pressure deficit due to higher temperature leads to greater transpiration and faster mortality during drought for tree seedlings common to the forest-grassland ecotone. New Phytol. doi:10.1111/nph.12321 Wolter K, Timlin MS (2011) El Niño/Southern Oscillation behaviour since 1871 as diagnosed in an extended multivariate ENSO index (MEI.ext). Int J Climatol 31:1074–1087 Wright SJ, Calderón O (2006) Seasonal, El Niño and longer term changes in flower and seed production in a moist tropical forest. Ecol Lett 9:35–44 Zimmermann NE, Yoccoz NG, Edwards TC, Meier ES, Thuiller W, Guisan A, Schmatz DR, Pearman PB (2009) Climatic extremes improve predictions of spatial patterns of tree species. Proc Natl Acad Sci 106(Suppl 2):19723–19728