Embed Size (px)
Citation preview
This content has been downloaded from IOPscience. Please scroll down to see the full text.
Download details:
IP Address: 47.143.95.101
This content was downloaded on 10/01/2017 at 17:47
Please note that terms and conditions apply.
Fine-scale modeling of bristlecone pine treeline position in the Great Basin, USA
View the table of contents for this issue, or go to the journal homepage for more
2017 Environ. Res. Lett. 12 014008
(http://iopscience.iop.org/1748-9326/12/1/014008)
Home Search Collections Journals About Contact us My IOPscience
OPEN ACCESS
RECEIVED
14 September 2016
REVISED
28 November 2016
ACCEPTED FOR PUBLICATION
16 December 2016
PUBLISHED
10 January 2017
Original content fromthis work may be usedunder the terms of theCreative CommonsAttribution 3.0 licence.
Any further distributionof this work mustmaintain attribution tothe author(s) and thetitle of the work, journalcitation and DOI.
Environ. Res. Lett. 12 (2017) 014008 doi:10.1088/1748-9326/aa5432
LETTER
Fine-scale modeling of bristlecone pine treeline position in theGreat Basin, USA
Jamis B Bruening1, Tyler J Tran1, Andrew G Bunn1, Stuart B Weiss2 and Matthew W Salzer3
1 Department of Environmental Sciences, Western Washington University, Bellingham, WA 98225, United States2 Creekside Center for Earth Observation, Menlo Park, CA 94025, United States3 Laboratory of Tree-Ring Research, University of Arizona, Tucson, AZ 85721, United States
E-mail: [email protected]
Keywords: dendroclimatology, topoclimate, dendrochronology
AbstractGreat Basin bristlecone pine (Pinus longaeva) and foxtail pine (Pinus balfouriana) are valuablepaleoclimate resources due to their longevity and climatic sensitivity of their annually-resolvedrings. Treeline research has shown that growing season temperatures limit tree growth at and justbelow the upper treeline. In the Great Basin, the presence of precisely dated remnant wood abovemodern treeline shows that the treeline ecotone shifts at centennial timescales tracking long-termchanges in climate; in some areas during the Holocene climatic optimum treeline was 100metershigher than at present. Regional treeline position models built exclusively from climate data mayidentify characteristics specific to Great Basin treelines and inform future physiological studies,providing a measure of climate sensitivity specific to bristlecone and foxtail pine treelines. Thisstudy implements a topoclimatic analysis—using topographic variables to explain patterns insurface temperatures across diverse mountainous terrain—to model the treeline position of threesemi-arid bristlecone and/or foxtail pine treelines in the Great Basin as a function of growingseason length and mean temperature calculated from in situ measurements. Results indicate: (1)the treeline sites used in this study are similar to other treelines globally, and require a growingseason length of between 147–153 days and average temperature ranging from 5.5°C–7.2°C, (2)site-specific treeline position models may be improved through topoclimatic analysis and (3)treeline position in the Great Basin is likely out of equilibrium with the current climate,indicating a possible future upslope shift in treeline position.
1. Introduction
The treeline ecotone on a mountain is the transitionzone between closed montane forest and treelessalpine landscape, encompassing the highest locationswhere mature trees are found (Wardle 1971, Scuderi1987, Jobbagy and Jackson 2000, Körner 2012). In theabsence of disturbance-related conditions and sub-strate prohibiting tree growth, the treeline positionrepresents a boundary between areas in which climaticconditions allow for physiological activity in maturetrees, and areas where tree growth is not possible.Research suggests this life-form boundary is climate-limited; regardless of species, elevation, or latitude,treeline positions globally share common climatologicalcharacteristics (Wardle 1971, Jobbagy and Jackson 2000,Körner 2012, Weiss et al 2015). Two independent
© 2017 IOP Publishing Ltd
studies (Körner and Paulsen 2004, Paulsen and Körner2014) provide evidence of a common growing-seasonisotherm around 5°C–6°C present at many differenttreeline sites globally.
Accordingly, climate-limited treelines are valued aspaleoclimatic indicators of environmental change asregional treeline positions have been shown to trackcentennial-scale changes in climatic conditions (Scuderi1987, Lloyd and Graumlich 1997, Salzer et al 2013). Inthe American southwest, Great Basin bristlecone pine(Pinus longaeva, D. K. Bailey) forms climate-limitedtreelines throughout Nevada and California. Thisspecies is a valuable climate proxy due to its extremelylong-livednature (e.g. Currey 1965) and the tendency ofits annual rings to correlate with the most growth-limiting environmental factor. Ring-width chronologiesfrom the upper-forest-border (at and just below
MWA
CSL
115°W
115°W
120°W
120°W125°W
40°N
40°N
35°N
35°N
SHP
0 160 320Km
Figure 1. Locations of treeline sites used in this study.
Environ. Res. Lett. 12 (2017) 014008
treeline) have been widely used as a proxy fortemperature (e.g. LaMarche Jr 1974), while ring-widthchronologies from the more arid lower-forest-borderhave been used a proxy for summer precipitation (e.g.Hughes and Funkhouser 2003). These findings indicatethe primary growth-limiting factor operates on agradient, changing from moisture limitation at thelower-forest-border to temperature limitation at theupper-forest-border (Kipfmueller and Salzer 2010).
Past research has shown topography influencesclimate—and subsequently biological systems—on thescale of tens to hundreds of meters (Weiss et al 1988,Lookingbill and Urban 2003, Dobrowski et al 2009,Geiger et al 2009, Adams et al 2014). This phenomena isreferred to as topoclimate, and has been the subject ofour recent research regarding the climate response ofnear-treeline bristlecone pine (Bunn et al 2011, Salzeret al 2013, 2014). Bunn et al (2011) discovered thattopographic position affects the growth response oftrees; individual trees growing well below the upper-forest-border in areas of cold air pooling displayeddistinctly different ring-width patterns from nearbytrees (within tens of meters) outside areas of cold airpooling. Further, the climate signal of low-elevationtrees in areas of cold air pooling was very similar to theclassic temperature-limited signal characteristic of theupper-forest-border. Salzer et al (2014) built on Bunnet al (2011) by constructing treeline and below-treelinechronologies from north and south-facing aspects. Theauthors identified a divergence in growth patternsbetween north and south facing aspects, as well as aclimate-response-threshold between moisture andtemperature limitation approximately 60–80 verticalmeters below treeline.
2
This study models treeline positions from atopoclimatic perspective. Combining evidence ofclimate-driven treeline formation with in situ temper-ature measurements, we present three site-specificmodels in the Great Basin predicting bristlecone pinetreeline position as a function of topoclimate.
2. Data and methods
2.1. Study areasWe chose three Great Basin treeline sites for thisanalysis (figure 1); (1) Mount Washington, SnakeMountain Range, NV (MWA, 38.91°N. lat., 114.31°W.long., treeline position approximately 3400m.a.s.l.),(2) Chicken Spring Lake, Sierra Nevada, CA (CSL,36.46°N. lat., 118.23°W. long., treeline positionapproximately. 3600m.a.s.l.), and (3) Sheep Moun-tain, White Mountains, CA (SHP, 37.52°N. lat.,118.20°W. long., treeline position approximately.3500m.a.s.l.). Sites MWA and SHP support GreatBasin bristlecone pine treelines, while the CSL treelineis formed by mostly foxtail pine (Pinus balfournaia,Grev. & Balf.), a closely related species to bristleconepine with a slightly shorter life-span and similarclimate growth-response (Lloyd and Graumlich 1997).
2.2. Topoclimate analysisAt each treeline site hourly temperatures wererecorded at 50 unique locations using iButtonthermochron sensors (Maxim Integrated, San JoseCA model DS1922L-F5); October 2013–September2014 at MWA, and October 2014–September 2015 atCSL and SHP. Sensors weremounted at a height of one
° C
● ●●
●
●
●
● ●
●
●
●
●
●● ●
●
●
●
●
● ●
●●
●
●
●
Tmin 1895 − 2015Average Tmin 1895 − 2015Tmin Oct '13 − Sep '14
MWA
Nov '13 Jan '14 Mar '14 May '14 Jul '14 Sep '14−20
−10
0
10
20
−20
−10
0
10
20
° C
● ●●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
Tmax 1895 − 2015Average Tmax 1895 − 2015Tmax Oct '13 − Sep '14
MWA
Nov '13 Jan '14 Mar '14 May '14 Jul '14 Sep '14−10
0
10
20
30
−10
0
10
20
30
° C
● ●●
●
●
●
● ●
●
●
●
●
●● ● ●
●
● ●
●
●
●
●
●
●
●
Tmin 1895 − 2015Average Tmin 1895 − 2015Tmin Oct '14 − Sep '15
CSL
Nov '14 Jan '15 Mar '15 May '15 Jul '15 Sep '15−20
−10
0
10
−20
−10
0
10
° C
● ●●
●
●
●
● ●
●
●
●
●
●
●● ● ●
●●
●
●
●
●
●
●
●
Tmax 1895 − 2015Average Tmax 1895 − 2015Tmax Oct '14 − Sep '15
CSL
Nov '14 Jan '15 Mar '15 May '15 Jul '15 Sep '15
0
10
20
30
0
10
20
30
° C
● ● ●
●
●
●
● ●
●
●
●
●
●● ●
●
●
● ●●
●
●
●
●
●
●
Tmin 1895 − 2015Average Tmin 1895 − 2015Tmin Oct '14 − Sep '15
SHP
Nov '14 Jan '15 Mar '15 May '15 Jul '15 Sep '15
−20
−10
0
10
−20
−10
0
10
° C
● ●●
●
●
●
●●
●
●
●
●●
● ● ●●
●●
●
●
●
●
●
●
●
Tmax 1895 − 2015Average Tmax 1895 − 2015Tmax Oct '14 − Sep '15
SHP
Nov '14 Jan '15 Mar '15 May '15 Jul '15 Sep '15−10
0
10
20
−10
0
10
20
Figure 2. Minimum (light blue) andmaximum (light orange) monthly temperatures during the period of iButton deployment at eachsite plotted against a 120 climate normal of minimum (dark blue) and maximum temperatures (dark orange). Annual monthlytemperatures 1895–2015 are shown in the background for reference (grey). The anomalies used to adjust the hourly iButton data arerepresented by the difference between the light and dark curves in each plot.
Environ. Res. Lett. 12 (2017) 014008
meter in living trees, dispersed across varyingtopographic features within a 1 km2
–2 km2 area.Our primary goal was to capture differences intemperature between different topographic positions,so the relative differences in temperature between allthe sensors at a given site were equally as important tothe raw recorded temperatures. Because we recordedtemperatures for only one calendar year at any givensite, our data reflect the weather conditions at that sitespecific to the period of deployment, rather than along-term climatic average (figure 2). To moreaccurately represent the average climate at eachtreeline location, we calculated monthly anomaliesbetween the temperature during deployment and theclimate normal for each location, and applied thesecorrections to our sensor data (PRISM Climate Group2004). This provided a data set that captured relativedifferences in temperature due to topography, whichthe raw values representative of the average climate,rather than anomalous weather during the period ofdeployment (figure 2).
We then applied a warming correction to our dataset to more accurately represent the climate whenGreat Basin treelines stabilized their current positions.Salzer et al (2013) report treeline positions in the GreatBasin moved downslope up to 100meters below theirhighest positions during the Holocene climaticoptimum, and established their current positions(well below the maximum positions during theclimatic optimum) in the early 1300s A.D. (also seeCarrara and McGeehin 2015). The authors present amulti-millennial Great Basin climate reconstruction
3
from bristlecone pine chronologies of previousSeptember–August temperature anomalies relativeto a baseline period A.D. 1000–1990, which showsan approximate warming of 1.5°C between the periodwhen treeline positions in the Great Basin stabilized inthe early 1300’s and present day. Therefore, wesubtracted 1.5°C from our observed temperatures sothat the topoclimate dataset would most accuratelyrepresent the climate that influenced the currenttreeline positions when they established in the early1300’s, rather than today’s climate that has noinfluence on treeline positions formed in the past.
From the observed hourly temperatures, wecalculated values of two climate variables unique toeach sensor: average monthly temperatures werecalculated by averaging all hourly values within agiven month, yielding twelve values per sensor; annualsum of degree hours above 5°C was also calculated,yielding one value per sensor. We used lasso regressionmodels (Kuhn 2015) (10-fold cross-validation, tenrepeats per fold) to model each climate variable as afunction of topographic variables at ten meterresolution. The topographic variables used forprediction are: elevation, slope, aspect-derived East-ness and Southness indices, topographic position andconvergence indices, and solar radiation loads. Themodels were used to predict the variables across areasabove 3000m.a.s.l. at each site, yielding thirteentopoclimate raster surfaces for each study locationrepresenting values of average monthly temperatureand degree hours above 5°C. Model skill was relativelyhigh but fluctuated between variables, and relied most
alpine
treeline
subalpine
0 350175Meters
LGS (days)
109
212
SMT (°C)
5.5
9.9
(A) (B) (C)
Figure 3. Positions of the subalpine and alpine regions located on Wheeler Peak in the Snake Range, NV (a), with overlaid treelinevariables representing the length of the growing season (b) and seasonal mean temperature (c). In all frames the treeline position isdisplayed red, with the 250m wide subalpine and apline regions respectively on the left and right of the treeline.
Environ. Res. Lett. 12 (2017) 014008
on elevation and solar radiation values as predictors(see appendix A for more on this process andmeasuresof model skill).
2.3. Treeline position modelsPaulsen and Körner (2014) present a model thatpredicts treeline positions globally as a function ofthree parameters; a threshold temperature (DTMIN,measured in °C) above which physiological activity ispossible, a growing season length (LGS, measured indays) that includes all days with an average dailytemperature above DTMIN, and a seasonal meantemperature (SMT, measured in °C) that is the averageof all days within the growing season. Using theauthors’ best fit value of DTMIN (0.9°C), we adoptedtheir methods to calculate LGS and SMT rastersurfaces at each site from our predicted monthlytopoclimate surfaces. We used cubic splines tointerpolate daily temperatures from the modeledmonthly topoclimate rasters, and summed thenumber of days with average temperatures above0.9°C for the growing season length, and averaged thedaily temperatures of all days within the growingseason to find the seasonal mean temperature.
We built classification models using the LGS andSMT raster variables to predict treeline position as theboundary between two mountainous biomes; asubalpine region of closed montane forest, and atreeless alpine region above the upper-forest-border(figure 3 panel (a)). We defined the boundaries of eachbiome around the treeline position through multi-stepprocess: (1) Using Google Earth we digitized treelineposition at the landscape scale (the red line in figure 3panel (a)). Conventions set by Körner (2007, 2012)define treeline position at a larger scale by connectingstraight lines between the upper reaches of maturetrees. We altered this method because our 10meter
4
resolution topoclimate variables allowed for a moreresolved definition of treeline position. We were verydeliberate in the areas of treeline used to build themodels, selecting only stretches of treeline that wereobviously climate-limited, and not influenced bydisturbances such as slope, rockfall, lack of substrate,etc. (2) We then set a 25meter upslope and downslopebuffer for the boundary of each biome nearest totreeline, to ensure a conservative separation betweenthe upper boundary of the subalpine and the lowerboundary of the alpine, and set the width of eachbiome to 250meters.
With the biome regions delineated, we obtainedtraining data for the classificationmodels by extractingvalues of LGS and SMT specific to each biome fromrandomly spaced points with a density of 500 pointsper square kilometer. Classification models were thendeveloped through an iterative process at each site; wegenerated threemodels with maximum branch lengthsof one, two, and three splits, and compared theaccuracy, complexity, and cost of adding additionalsplits between each model. The simplest, mostaccurate model was chosen by balancing the predic-tion accuracy and the complexity of each model, withthe fewest number of splits and terminal nodesrepresenting the simplest model. For example, if theprediction accuracy was similar between models ofdifferent complexities (one split vs two or three splits),preference was given to the model with the fewestnumber of splits.
3. Results and discussion
3.1. Treeline predictionThe classification trees (figure 4) at all sites suggestseasonal mean temperature is the best predictor of
Table 1. Confusion matrices for each classification model.Columns represent actual classifications, while rows indicatepredicted classifications. At all sites the models tend to overpredict the subalpine biome.
MWA CSL SHP
alpine subalpine alpine subalpine alpine subalpine
alpine 839 89 1060 113 616 174
subalpine 146 1056 532 1621 369 811kappa 0.78 0.61 0.45
yes noIs seasonal mean
temperature < 7.2 °C?
alpine839 89
subalpine146 1056
yes no
MWA
yes noIs seasonal mean
temperature < 5.4 °C?
alpine1060 113
subalpine532 1621
yes no
CSL
yes noIs seasonal mean
temperature < 6.2 °C?
alpine616 174
subalpine369 811
yes no
SHP
Figure 4. Classification trees by site. The numbers in each leaf represent the amount of alpine and subalpine points (respectively)included in that leaf.
Environ. Res. Lett. 12 (2017) 014008
treeline position. SMT thresholds of 7.2°C at MWA(89% overall accuracy), 5.4°C at CSL (81% overallaccuracy), 6.2°C at SHP (72% overall accuracy)separate the alpine and subalpine regions. Whensecondary and tertiary splits were allowed the overallaccuracy increased by 0.4%, 0.6%, and 0.1%respectively, thus we present the simplest model ateach site with a single threshold in SMTseparating thebiomes.
The Snake Range (MWA) model was the mostaccurate (table 1 and figure 5). Rates of misclassifica-tion were different between the predicted biomes(producer accuracy: alpine 85%, subalpine 92%)indicating a model slightly biased towards subalpineprediction. Cohen’s kappa statistic—a measure ofhow different a prediction is from a randomizedclassification-is 0.78, indicating this classification isdifferent than random with substantial agreement.The Sierra Nevada Range (CSL) model was slightly lessaccurate than the MWA model (table 1 and figure 6),with a slightly larger bias towards subalpine prediction(producer accuracy: alpine 66%, subalpine 93%).Cohen’s kappa indicates substantial agreement theprediction is different than random. The WhiteMountains (SHP) model provided the least accurateprediction (producer accuracy: alpine 63%, subalpine82%), yet was consistent with other sites in a biastoward subalpine prediction, and Cohen’s kappaindicates moderate agreement the prediction isdifferent than random. It is important to note thatat all sites higher rates of producer accuracy for thesubalpine region come at the cost of overpredictingthis region. This results in a tendency to predict thetreeline position slightly higher than a model withoutthis bias, and is most pronounced at CSL and SHP.
A drawback to this method lies in the discontinuityof easily identifiable climate limited treelines. In theabsence of other unrelated factors that influence
5
treeline position, the entire treeline at a site would beclimate-limited. However this is rarely the case, at eachsite there are many stretches of treeline that are clearlydriven by factors other than climate-limitation, such assubstrate availability, slope, or snowpack. Thechallenge lies in identifying enough continuous areathat display obvious climate limitation to allow atreeline position model. As a result, appropriatestretches of treeline for modeling were usually limitedto a single aspect due to the geographies of each siteand locations of climate limited treelines: at MWA theclimate-limited treeline faces west, at CSL the climatelimited foxtail pine treeline faces southwest; and atSHP the climate limited treeline faces east. Top-oclimate modeling requires a dense network of sensorsin a single area, and thus we needed to locate oursensors on a single side of each mountain range nearthe climate limited treelines to maximize outpredictive power (Bruening 2016). While this methodallows for accurate climate prediction near the sensors,the models lose predictive power on the other sides ofthe mountain ranges. Consequently, these models areinherently site-specific and are in no way intended as acomprehensive treeline position model for the GreatBasin. This analysis is different from other treelinemodels conducted at larger scales, yet our resultsregarding the physiological constraints of treelineposition at these sites are consistent with the globalmodels.
3.2. Global model comparisonsThis analysis provides comparisons to previoustreeline studies while accounting for topocli-maticeffects on treeline position (Körner and Paulsen 2004,Körner 2012, Paulsen and Koorner 2014). We modelthe length of the growing season and its averagetemperature at a scale previously unavailable totreeline researchers, which have provided insightsinto the physiology of near-treeline bristlecone pinegrowth in the Great Basin. In a related analysis, Tranet al (2017) examined the climate-growth response ofbristlecone pine at MWA, CSL, and SHP, and usedcluster analysis to identify the primary growth patternsin the ring-widths of trees sampled at these sites. Theyfound that Bristlecone pine growth nearest to treelinewas controlled by SMT (calculated via the samemethods as described in this analysis), whileBristlecone pine growth farther downslope and awayfrom the upper-forest-border was more influenced by
0 31.5Kilometers
(A) (B)
Alpine
Subalpine
Figure 6. (a) Sierra Nevada Range, CA, (site CSL) prediction of the current treeline position from the climate during its stabilizationand, (b) prediction of the current treeline position potential based on today’s climate.
0 31.5Kilometers
Alpine
Subalpine
(A) (B)
Figure 5. (a) Snake Range, NV, (site MWA) prediction of the current treeline position from the climate during its stabilization and,(b) prediction of the current treeline position potential based on today’s climate.
Environ. Res. Lett. 12 (2017) 014008
moisture availability, and identified an SMT threshold7.4°C–8°C that separates the different climate-growthresponses. These results corroborate previous growthlimitation research (LaMarche Jr 1974, Hughes andFunkhouser 2003, Salzer et al 2009, Kipfmueller and
6
Salzer 2010, Salzer et al 2014), and are unique inproviding specific physiological thresholds betweenthe two dominant modes of Bristlecone pine growth inthe Great Basin. In combination with the classificationmodel results from this analysis, we can identify a
Table 2. Values of LGS and SMT from from the treelinepositions at each site, compared to a global model’s best fitpresented by Paulsen and Korner (2014).
MWA CSL SHP TREELIM
LGS [days] 147 153 149 94þSMT [°C] 7.2 5.5 6.3 6.4
Environ. Res. Lett. 12 (2017) 014008
window of temperature sensitivity at each site; areaswith SMT values above our models’ reportedthreshold and below the 7.4°C–8°C thresholdreported by Tran et al (2017) are most suitablefor temperature limited Bristlecone pine growth—ideal for paleoclimate temperature reconstructionsthroughout the Holocene. These data, whilecharacteristically site specific, allow for deeperunderstanding of bristlecone pine physiology inthe harshest conditions and identification of themost desirable samples for paleoclimatic inference.
Paulsen and Körner’s (2014) model (TREELIM)calculates values of LGS and SMT that predict a set of376 treeline positions formed by various species acrossall ecozones (See table 2. Note: the authors report thatthe LGS value is at least 94 days, and may be longerdepending on the climate region). These values are notspecific to any one species or ecozone, and are mostlyconsistent with our topoclimate calculations. Thesimilarity in seasonal mean temperatures suggests thatGreat Basin treelines are likely subject to a similar SMTisotherm as other sites globally, despite differences ingrowing season length.
According to Körner (2012) the length of thegrowing season at treeline varies significantly betweenclimate regions, shown by plots of daily mean rootzone temperature from 32 different treeline sites fromvarious different climate regions (pages 40–47). Eachsite has an estimated length of the growing seasoncalculated from root-zone and air temperatures, withvalues spanning from around 100 days in the subarcticand boreal zones up to 365 days in the equatorialtropics. Growing season lengths fromwarm-temperateand cool-temperate climate zones (the Great Basinfalling somewhere in between these climate zones)range between 122–190 days, consistent with ourmodeled values of LGS. For further validation of ourmodeled treeline variables, we obtained observed rootzone temperatures on Mount Washington (MWA)which allowed for independent calculations of SMTand LGS (Scotty Strachan, unpublished data). Wecalculated LGS to be 152 days (modeled value is147 days, table 2) and SMT to be 7.7°C (modeled valueis 7.2°C, table 2).
While the length of the growing season andseasonal mean temperature may be useful predictorsof treeline (as defined and calculated by Körner(2012), Paulsen and Körner (2014) and in thisanalysis), these variables explain little about treephysiology. Two treeline sites may have comparable
7
season lengths and average temperatures withcontrasting annual mean temperature profiles (Körner2012). Paulsen and Körner (2014) guard againstinterpreting these variables too literally—TREELIMcalculates a best fit of 6.4°C as ‘a common isotherm oflow temperature for forest limits’, and they stress theabsence of physiological relevance represented by thisvalue. While it is close to the physiological limit forwoody biomass accumulation, they argue the seasonalmean temperature ‘reflects an arithmetic mean that issubsuming the combined action of low temperature,integrated over time, on a suite of processes associatedwith tissue formation, from root tips to apicalmeristems’.
3.3. Treeline position models, treeline potential, andgrowth-limiting factorsModeled treeline positions were obtained for eachmountain range by applying the SMT thresholdcalculated in each classification model to the SMTraster surface at each site (figures 5–7). We developedmodeled treeline positions under two different climatescenarios.
The first ((a) in figures 5–7) represents predictionsof the current treeline position using SMT representa-tive of the climate from the early 1300s (the period ofcurrent treeline stabilization, see section 2.2), the SMTraster used to build the models. The second ((b) infigures 5–7) represents the same models applied totoday’s climate, which has warmed 1.5°C since thetreeline positions in the Great Basin stabilized theircurrent position according to Salzer et al (2013).
The differences between the treeline positionmodel projections (figure 4) are the result of arelatively sensitive threshold between the subalpineand alpine regions. By projecting our model usingclimate data representative of the current climate, weconclude treeline positions at our sites are likely out ofequilibrium with the Great Basin climate; at all sitesthe treeline positionmodeled from today’s climate ((b)in figures 5–7) moved upslope from its currentposition ((a) in figures 5–7). It is important to notethat b is a projection of treeline position potentialbased on warming at these sites since the currenttreelines established their positions. We speculate thatcurrent treeline positions have yet to ‘catch up’ withthe the current climate, a result of the slow nature oftreeline dynamics (Körner 2012). The demographicprocesses that cause an upslope shift in treelineposition are slow and lag behind changes in climate(Scuderi 1987, 1994, Lloyd and Graumlich 1997),however an in-depth discussion of how and why thislag exists is outside the scope of this analysis. It isunknown exactly how long behind changes in climatetreeline position lags, but it is speculated to be as longas hundreds of years depending on the ecozone,species, slope, etc.
As the Great Basin climate gradually warms,uninhabitable areas above treeline start to become
0 21Kilometers
Alpine
Subalpine
(A) (B)
Figure 7. (a) White Mountains, CA, (site SHP) prediction of the current treeline position from the climate during its stabilizationand, (b) prediction of the current treeline position potential based on today’s climate.
Environ. Res. Lett. 12 (2017) 014008
favorable for seedling recruitment. A recent study oftreeline demographics in nearby regions in the GreatBasin documented Bristlecone pine and limber pine(Pinus flexilis E. James) seedling recruitment abovetreeline up to 75m and 225m respectively (Millar et al2015). These observations corroborate previousBristlecone pine species distribution models underdifferent warming scenarios for the White Mountains,CA by Van de Ven et al (2007). The authors provideBristlecone pine range projections indicating a singledegree of warming could be enough to initiate anupslope migration of Bristlecone pine treeline by tensto hundreds of meters. This is supported by ourtopoclimate analysis and the sensitivity of the treelineposition models (figure 4). An increase in seasonalmean temperature would force the position ofthe SMT isotherm upslope and make conditionsmore favorable for Bristlecone pine growth above thecurrent treeline position ((b) in figures 5 to 7). Theextent to which the SMT isothermwill move upslope isdependent on the topoclimatology of each treelineposition; for example a gradual slope would enablefarther upslope migration of treeline position than asteep slope. An in depth analysis of microrefugia andtopoclimatology in the Great Basin would benefit thestudy of bristlecone pine treeline dynamics.
4. Conclusion
We predicted the position of three Great Basintreelines formed by bristlecone and foxtail pine
8
exclusively from climate and topography data.Through a topoclimatic analysis we capturedlandscape-scale effects of topography on climate—and consequently on treeline position—that areindependent of changes in elevation. Our resultsindicate the average temperature throughout thegrowing season (SMT) is the most dominant factorinfluencing treeline position on the landscape,regardless of species or elevation, in agreement withprevious research Jobbagy and Jackson (2000), Körner(2012), Paulsen and Koorner (2014). At the sites in thisanalysis, treelines form in areas on a mountain slopewhere average growing season temperatures rangebetween 5.5°C and 7.2°C, and the growing seasonlength is approximately 150 days (defined by all dayswith an average temperature above 0.9°C). We alsoprovide an estimate of the climate sensitivity of GreatBasin treelines, and demonstrate that the treelines inthis analysis are likely out of equilibrium with thecurrent climate.
Dendroclimatological studies would benefit from acomprehensive investigation of these findings. Inconjunction with the recommendations of Salzer et al(2014), our results suggest that over time the spatialwindow of near-treeline temperature-sensitivity shiftson a centennial-millennial timescale. Treeline positionmodeling throughout the Holoscene would enable amore accurate isolation of a purely temperature-limited signal. For any given year, chronologiesreconstructing temperatures should only use treesthat are within the temperature-sensitive window. Astreeline position shifts up and down, individual trees
Table 3. Measures of model skill (R2 and root mean squareerror) for topoclimate models described in section 2.2; monthlyaverage temperatures and the annual sum of degree hours above5°C (DH5C).
MWA CSL SHP
R2 rmse R2 rmse R2 rmse
Jan 0.64 1.60 0.47 1.57 0.85 1.54
Feb 0.54 1.57 0.40 1.59 0.94 1.52
Mar 0.39 1.62 0.52 1.55 0.89 1.54
Apr 0.62 1.57 0.47 1.53 0.91 1.53
May 0.59 1.60 0.67 1.52 0.92 1.53
Jun 0.66 1.53 0.62 1.52 0.89 1.53
Jul 0.72 1.53 0.64 1.52 0.89 1.53
Aug 0.70 1.52 0.58 1.53 0.90 1.53
Sep 0.72 1.54 0.54 1.53 0.90 1.53
Oct 0.79 1.53 0.43 1.53 0.92 1.52
Nov 0.67 1.55 0.50 1.53 0.90 1.53
Dec 0.58 1.58 0.59 1.54 0.91 1.53
DH5C 0.68 4196 0.48 3836 0.91 4130
Environ. Res. Lett. 12 (2017) 014008
should be added and removed from the reconstructivechronology depending on their location relative to thetemperature-sensitive window. This would enable thecleanest temperature signal and potentially moreaccurate calibration for global climate models.
Appendix A
The topoclimate models described in section 2.2 arenot the main focus of this study, but are a necessarystep in the larger task of predicting treeline position asa function of topoclimate. Thus we present the resultsof these models as an appendix, to ensure sufficientclarity and transparency without distracting the readerfrom our main objectives. The results of each modelare shown in table 3; typically the summer monthsyielded more accurate prediction than other seasons.This is likely due to snowpack covering sensors duringthe winter months, which decreased predictive ability.At all sites elevation was the most consistent andimportant predictive variable, however overall modelaccuracy between sites is inconsistent due to thevarying degrees at which other topographic features(excluding the influence of elevation) influenced theobserved temperatures. The SHP models were themost accurate, likely a result of the strong elevation-dependent trend in the observed temperatures at thissite. Aside from elevation, solar radiation loads andaspect derivatives eastness and south-ness providedthe most predictive power. For more specificinformation regarding the theory and methods behindthe topoclimate models see Bruening (2016).
References
Adams H R, Barnard H R and Loomis A K 2014 Topographyalters tree growth-climate relationships in a semi-aridforested catchment Ecosphere 5 1–16
9
Bruening J M 2016 Fine-scale topoclimate modeling and climatictreeline prediction of Great Basin bristlecone pine (Pinuslongaeva) in the American southwest Master’s thesisEnvironmental Science, Western Washington University,Bellingham, WA
Bunn A G, Hughes M K and Salzer M W 2011 Topographicallymodified tree-ring chronologies as a potential meansto improve paleoclimate inference Clim. Change 105627–34
Carrara P E and McGeehin J P 2015 Evidence of a higher late-Holocene treeline along the Continental Divide in centralColorado The Holocene 25 1829–37
Currey D R 1965 An ancient bristlecone pine stand in easternNevada Ecology 46 564–6
Dobrowski S Z, Abatzoglou J T, Greenberg J A and Schladow S2009 How much influence does landscape-scalephysiography have on air temperature in a mountainenvironment? Agr. Forest Meteorol. 149 1751–58
Geiger R, Aron R H and Todhunter P 2009 The Climate Nearthe Ground (Lanham, MD: Rowman & Littlefield)
Hughes M K and Funkhouser G 2003 Frequency-dependentclimate signal in upper and lower forest border tree ringsin the mountains of the great Basin Clim. Change 59233–44
Jobbagy E G and Jackson R B 2000 Global controls of forest lineelevation in the northern and southern hemispheres GlobalEcol. Biogeogr. 9 253–68
Kipfmueller K F and Salzer M W 2010 Linear trend andclimate response of five-needle pines in the westernUnited States related to treeline proximity Can. J. For.Res. 40 134–42
Körner C 2007 Climatic treelines: conventions, global patterns,causes (klimatische baumgrenzen: konventionen, globalemuster, ursachen) Erdkunde 61 316–24
Körner C 2012 Alpine Treelines: Functional Ecology of the GlobalHigh Elevation Tree Limits (Berlin: Springer)
Körner C and Paulsen J 2004 A world-wide study of highaltitude treeline temperatures J. Biogeogr. 31 713–32
Kuhn M 2015 caret: Classification and Regression Training Rpackage version 6.0–47
LaMarche Jr V C 1974 Paleoclimatic inferences from long tree-ring records Science 183 1043–48
Lloyd A H and Graumlich L J 1997 Holocene dynamics oftreeline forests in the Sierra Nevada Ecology 781199–210
Lookingbill T R and Urban D L 2003 Spatial estimation ofair temperature differences for landscape-scale studiesin montane environments Agr. Forest Meteorol. 114141–51
Millar C I, Westfall R D, Delany D L, Flint A L and Flint L E2015 Recruitment patterns and growth of high-elevationpines in response to climatic variability 1883-2013, inthe western Great Basin, USA Can. J. For. Res. 451299–312
Paulsen J and Körner C 2014 A climate-based model to predictpotential treeline position around the globe Alpine Bot. 1241–12
PRISM Climate Group 2004 Oregon State University (http://prism.oregonstate.edu)
Salzer M W, Bunn A G, Graham N E and Hughes M K 2013Five millennia of paleotemperature from tree-rings in theGreat Basin, USA Clim. Dynam. 42 1517–26
Salzer M W, Hughes M K, Bunn A G and Kipfmueller K F 2009Recent unprecedented tree-ring growth in bristlecone pineat the highest elevations and possible causes. Proc. Natl.Acad. Sci. USA 106 20348–53
Salzer M W, Larson E R, Bunn A G and Hughes M K 2014Changing climate response in near-treeline bristlecone pinewith elevation and aspect Environ. Res. Lett. 9 114007
Scuderi L A 1987 Late-Holocene upper timberline variation inthe southern Sierra Nevada Nature 325 242–44
Scuderi L A 1994 Solar influences on Holocene treeline altitudevariability in the Sierra Nevada Phys. Geogr. 15 146–65
Environ. Res. Lett. 12 (2017) 014008
Tran T J, Bruening J M, Bunn A G, Weiss S B and SalzerM W 2017 Cluster analysis and topoclimatemodeling to examine bristlecone pine tree-ringgrowth signals in the Great Basin, USA Environ.Res. Lett. 12 014007
Van de Ven C M, Weiss S and Ernst W 2007 Plant speciesdistributions under present conditions and forecasted forwarmer climates in an arid mountain range Earth Interact.11 1–33
10
Wardle P 1971 An explanation for alpine timberline NZ J. Bot. 9371–402
Weiss D J, Malanson G P and Walsh S J 2015 Multiscalerelationships between alpine treeline elevation andhypothesized environmental controls in the western UnitedStates Ann. Assoc. Am. Geogr. 105 437–53
Weiss S B, Murphy D D and White R R 1988 Sun, slope, andbutterflies: topographic determinants of habitat quality forEuphydryas editha Ecology 69 1486–96
