Em

Ca

b

c

a

ARRAA

KCSSPGF

1

Mpt

fA

h0

Ecological Modelling 288 (2014) 203212

Contents lists available at ScienceDirect

Ecological Modelling

journa l h om epa ge: www.elsev ier .com/ locate /eco lmodel

xploring succession within aspen communities using a habitat-basedodeling approach

ody M. Mittancka,, Paul C. Rogersb, R. Douglas Ramseyb, Dale L. Bartosc, Ronald J. Ryelb

CNL Environmental Consultants, Inc., 14981 South Eagle Crest Drive, Draper, UT 84020, USADepartment of Wildland Resources and the Ecology Center, Utah State University, Logan, UT 84322-5230, USARocky Mountain Research Station, USDA Forest Service, Logan, UT 84321, USA

r t i c l e i n f o

rticle history:eceived 17 March 2014eceived in revised form 10 June 2014ccepted 12 June 2014vailable online 2 July 2014

eywords:onifer encroachmenteraltableureIS modelingunctional types

a b s t r a c t

Quaking aspen (Populus tremuloides Michx.) forest communities play a crucial ecological role acrosswestern North America. However, there is increasing evidence that these communities have diverg-ing ecological roles across aspens expansive range. Previous studies show evidence for both seral andstable aspen functional types. This leads us to believe that the pathway of these systems may not alwayslead to a climax conifer sere, but in many cases results in a stable community dominated by aspen. Thisstudy is an attempt to use a static model, based on large-scale environmental variables, to account forsuccessional dynamics within aspenconifer systems and predict distributions of aspen functional typesacross large landscapes. Environmental factors influencing aspenconifer succession have been observedin past research but not fully explored. Our study methodologies and application of model results werespecifically designed to aid land managers in identifying extent and function of aspen forest communitiesin order to plan restoration projects. Four study sites were chosen within Utah in order to capture thewidest geographic variance. Photointerpretation of National Agriculture Imagery Program (NAIP) colorinfrared imagery was used to classify dominant forest cover at approximately 250 plots within eachsite. At each plot, variables were calculated and derived from DAYMET data, digital elevation models,and soil surveys and assessed for precision and ability to model forest type distributions. A generalizedlinear model was used to assess habitat overlap between aspen and conifer in order to explore succes-sional dynamics and predict areas where stable aspen communities are likely to occur. Model resultsindicate an interaction between topographic position and moisture influence the probability of coniferencroachment but do not preclude conifers entirely. The highest probability for stable aspen communi-

ties occurs between 60 and 90 cm of total annual precipitation on topographic positions receiving greaterthan 4500 W h/m2/d of solar radiation. Prediction-conditioned fallout-rates were used to partition thecontinuous model output into a hard classification. These results were applied in an overlay analysiswith Southwest Regional Gap landcover data, indicating 19% of aspen forests across Utah are potentiallystable functional types, whereas the remaining 81% are vulnerable to conifer encroachment.

2014 Elsevier B.V. All rights reserved.. Introduction

In western North America, quaking aspen (Populus tremuloides

ichx.) occurs with conifer in mixed stands as well as in adjacenture communities. Within this landscape aspen is often assumedo be seral to conifer species (Baker, 1918; Bartos et al., 1983). The

Abbreviations: NAIP, National Agriculture Imagery Program; DAYMET, Daily Sur-ace Weather and Climatological Summaries; SWReGAP, Southwest Regional Gapnalysis Project; UFRWG, Utah Forest Restoration Working Group. Corresponding author. Tel.: +1 8013672230.

E-mail address: cody.mittanck@gmail.com (C.M. Mittanck).

ttp://dx.doi.org/10.1016/j.ecolmodel.2014.06.010304-3800/ 2014 Elsevier B.V. All rights reserved.successional trajectory of aspen to conifer is described as deter-ministic, where aspen requires disturbance or will eventually bereplaced by encroaching conifer (Shepperd and Jones, 1985). How-ever, many studies show evidence for not only a seral but a stableaspen community or functional type (Langenheim, 1962; Bettersand Woods, 1981; Mueggler, 1988; Romme et al., 2000; Rogerset al., 2014a), describing such a community as one that persists freeof conifers and is self-regenerating. This forest community has a dis-proportionately high ecological role to play within arid regions of

the Intermountain West of the United States. High organic matter,nutrient rich soils, light, and water dynamics, specifically associatedwith these communities, provides an environment for a diverseassemblage of plants and wildlife as well as domestic livestock

dx.doi.org/10.1016/j.ecolmodel.2014.06.010http://www.sciencedirect.com/science/journal/03043800http://www.elsevier.com/locate/ecolmodelhttp://crossmark.crossref.org/dialog/?doi=10.1016/j.ecolmodel.2014.06.010&domain=pdfmailto:cody.mittanck@gmail.comdx.doi.org/10.1016/j.ecolmodel.2014.06.010

2 cal Modelling 288 (2014) 203212

(saaip

ii12aFsoseordt(u

mhtehaeetaa(isemb

tttoclfist2Zbtaa

liclpasp

2014a). We refined our target population for two reasons: (1)according to the UFRWG, the upland aspen type is one of the major04 C.M. Mittanck et al. / Ecologi

Debyle and Winokur, 1985), in addition to other environmentalervices relating to water yield and ecosystem function (LaMalfand Ryel, 2008). With aspen cover potentially decreasing (Bartosnd Campbell, 1998; Rogers, 2002; Di Orio et al., 2005), there is anmportant need to classify stable and seral aspen sites to effectivelylan aspen restoration projects (Rogers et al., 2014a).While aspen is being encroached by conifer in many areas, there

s intriguing evidence that it may be persisting and even expandingts distribution in others (Langenheim, 1962; Betters and Woods,981; Mueggler, 1988; Romme et al., 2000; Manier and Laven,002; Shepperd et al., 2006). These temporal studies suggest thatt some sites aspen has remained in a persistent or stable state.urthermore, despite light-limiting requirements of young aspenhoots, juvenile aspen have been shown to grow under the canopyf pure aspen stands and in gaps, enabling self-regeneration anduggesting a persistent or stable aspen community type (Kurzelt al., 2007). With an aim to develop a more detailed typologyf persistent and seral stand structures, Kurzel et al. (2007)eport that over 70% of the aspen-dominant stands they surveyedid not require stand-replacing disturbance events; alternatively,hey regenerated through a variety of modes, with the majority60%) regenerating episodically with a large pulse of suckeringnrelated to course scale disturbance.Though individual studies have documented stable aspen com-

unities, environmental factors describing stable aspen habitatave not been thoroughly explored on a landscape scale. It is likelyhat successional processes in aspenconifer systems are influ-nced by both broad and fine-scale mechanisms. Recent studiesave uncovered broad patterns of seral and stable aspen typesccording to easily measured environmental variables (Rogerst al., 2014a). On the Owyhee Plateau in southwestern Idaho, Strandt al. (2009) found that 14% of their pure aspen plots seemedo occupy south-facing slopes above 1900 m. These aspen standslso showed characteristics of persistent stands being uneven-ged and self-regenerating. Over a 30-year period, Crawford et al.1998, p.201) did not notice any appreciable conifer encroachmentnto pure aspen stands in a study performed in the montane andubalpine forests of Gunnison County, Colorado. Where coniferstablishment did occur in aspen stands it was on the cooler,oister north-facing slopes, where they have been observed toe seral to coniferous forests.These studies suggest that environmental variables, such as

opographic position, act as a surrogate for distinct conditionshat may influence conifer encroachment. Therefore, we believehat using a static model approach, i.e. habitat-based, and relyingn environmental variables to model distributions of aspen andonifer would provide an ecologically meaningful and landscape-evel classification of seral and stable aspen communities. Whilene-scale or site-specific mechanisms, such as grazing, fire regimes,oils, and genetics, likely play an important role in determininghe successional dynamics of a particular aspen stand (Mittanck,012), such dynamics are difficult to quantify (Mueggler, 1988;immerman et al., 2007). On the other hand static models withroader GIS-derived environmental variables may account forhese fine-scale mechanisms while also providing key benefits suchs cost-effectiveness, ability to apply across multiple spatial scales,s well consistency and repeatability.Our primary objectives were twofold. First, using a generalized

inear model approach and GIS-derived environmental variables,dentify stable and seral aspen habitat. If this approach is suc-essful, our second objective is to apply the model to largerandscapes. We believe there is great utility in landscape map-ing of functional aspen types for the purpose of ecologically soundspen restoration. Working at these large scales, we chose remote

ensing and GIS applications as the most parsimonious way toroceed.Fig. 1. The geographic context of the four study sites sampled. Site boundaries weredelineated in order to capture entire site-scale distributions of aspen.

2. Methods

2.1. Study areas

Four sites within the state of Utah were chosen for this study(Fig. 1). At these sites prior aspen research had been conducted byresearchers at Utah State University (Kusbach, 2010; Rogers et al.,2010). Much of the data from this prior research was adapted andused as ground-reference data for this current study. Althoughthe selection of study sites was not random, we hoped to cap-ture the majority of both the site-scale and regional-scale rangeof upland aspen by sampling within these climatically distinctstudy locations. The size of the study sites ranged from 18,000to 33,100 ha with elevations ranging from 1770 to 3160 m. Treespecies at these sites reflect the wide geographic and elevationalrange, and include: quaking aspen, Doug-fir (Psuedotsuga men-ziesii var. glauca), subalpine fir (Abies lasiocarpa), Englemann spruce(Picea englemannii), pinyon (Pinus edulis), juniper (Juniperus scop-ulorum), limber pine (Pinus flexilis), curl-leaf mountain mahogany(Cercocarpus ledifolius Nutt. ex Torr. & Gray), Gamble oak (Quercusgambelii), and Rocky Mountain maple (Acer glabrum). The topogra-phy, soils, precipitation, dominant plant communities, and historicand current land uses at each site were distinct and are discussedin detail by Mittanck (2012).

2.2. Sample design

In this study only the upland aspen stand type is considered, asdefined by the Utah Forest Restoration Working Group (UFRWG)(OBrien et al., 2010). This stand type includes primarily large con-tiguous stands that are not restricted to hydrologic features, such asriparian stringers and snow-pocket aspen types (Rogers et al.,ones for which management or restoration decisions are repeat-edly being made on the National Forests of Utah (OBrien et al.,

cal Mo

2hdm

iseimLLetccptofsoft

2

(hwtSt(DcrftkvttfmtwwaCtibtewgTsttSwo

C.M. Mittanck et al. / Ecologi

010, p.8); (2) and by not including aspen types defined by uniqueydrologic features allowed us to circumvent using resource gra-ients, instead using direct and indirect gradients within theodel, as defined by Austin (1980).A stratified random sample design was used to generate approx-

mately 250 30 30 m plots within forested areas at each studyite. Forest and non-forest areas were stratified by creating a for-st mask using an ISODATA unsupervised classification algorithmn ENVI image processing software (version 4.8, Exelis Visual Infor-ation Solutions, URL: http://www.exelisvis.com) on a July 2008andsat 5 TM image acquired from the MRLC (Multi-Resolutionand Characteristics) Landsat database (http://www.mrlc.gov/). Atach plot canopy cover was interpreted using National Agricul-ure Imagery Program (NAIP) color infrared 2009 imagery. Canopyover was then classified into forest type groups. Groups werelassified as: Aspen if plots contained >90% aspen, Conifer iflots had >90% presence of conifer species, Mixed if plots con-ained aspen and conifer but neither dominated with >90% cover,Other if plots contained >90% cover of species other than aspenr conifer. An accuracy assessment of forest type groups was per-ormed using approximately 80 ground-reference plots for eachite. Ground-reference data was gathered by using on-the-groundcular estimates of canopy cover by species and then classified intoorest type groups. The average overall accuracy of photointerpre-ation sampling methods across all four sites was 93%.

.3. Generating climatic, physical, and biophysical variables

All variables used in this study were derived using Rversion 2.14.2, R Foundation for Statistical Computing, URL:ttp://www.r-project.org/) and Python (version 2.7, Python Soft-are Foundation, URL: http://www.python.org/), integrated within

he ArcGIS Model Builder environment (version 10, Environmentalystems Research Institute, Inc., Redlands, CA, US.). Environmen-al variables were grouped into climate, physical, and biophysicalTable 1). Raw climate data was acquired for each site fromAYMET (Thornton et al., 1997). DAYMET temperature and pre-ipitation data are 18-yr averages (19801997) supplied in 1-kmaster format interpolated from field weather stations. In orderor the model to be sensitive to fine scale variability in forestype distributions, the raw daily average air temperature at 1-m resolution was downscaled to 30-m resolution to match otherariables (Zimmermann and Roberts, 2007). Although precipita-ion is also related to elevation and could be downscaled usinghe same methods as used for temperature, it was decided toorego downscaling of precipitation. Precipitation events are muchore widespread and irregular in nature, thus we elected to keep

hem at a coarser resolution. Solar radiation, slope, and aspectere calculated using ArcGIS Spatial Analyst tools. Aspect (degrees)as further transformed into an ecologically meaningful vari-ble based on a symmetric radiation wetness index (Roberts andooper, 1989). Potential evapotranspiration was calculated usinghe empirical equation of Jensen and Haise (1963), which is specif-cally calibrated for the arid western United States. Site wateralance (SWB) is a running sum of the difference between poten-ial evapotranspiration and precipitation, while not allowing anyxcess water to exceed the site-specific soil water capacity. Soilater capacity for each pixel is derived from the Soil Survey Geo-raphic Database (SSURGO) (http://soildatamart.nrcs.usda.gov).his bucket method for calculating site water balance is veryimilar to Zimmermann and Roberts (2007), which was based onhe concept developed by Grier and Running (1977). However,

his method differs by running the water balance from January-eptember (the end of the growing season) without concern forhen recharge (e.g. precipitation exceeds evapotranspiration)ccurs. This metric for soil water balance essentially gives an indexdelling 288 (2014) 203212 205

of site dryness directly after the period when plants need moisturethe most.

2.4. Data analysis

Forest Type groups were transformed into a binary dependentvariable by assigning a 1 for Aspen Forest Type (referred to asaspen) and a 0 for Conifer and Mixed Forest Types (referred toas conifer-present). A generalized linear model (GLM) of the bino-mial family was then fit using 100% of the data from all four sites.All analyses were conducted in R (version 2.14.2, R Foundation forStatistical Computing, URL: http://www.r-project.org/).

Multivariate outliers were located by calculating a Euclideandistance matrix from scaled environmental variables as recom-mended by McCune and Grace (2002). The average distances ofa particular sample unit to all other sample units were then plot-ted in a frequency histogram and all sample units that fell abovethree standard deviations were vetted individually for data collec-tion and data processing errors. No errors were found and thereforeno outliers were removed.

Other modeling approaches were considered for this study,including discriminant function analysis (DA) and classificationand regression trees (CART). While the former is very similar tologistic regression in a GLM, in the regard that both require cat-egorical groups as the response variable, DA differs primarily byaiming to uncover the correlative structure between the groups,whereas logistic regression emphasizes prediction (Tabachnick andFidell, 1996). GLMs are also very flexible in their statistical assump-tions as the linear predictor is related to the dependent variablethrough a link function, which allows for non-normal distributionsand unequal variances to be transformed to linearity (Guisan et al.,1999). DA on the other hand has strict assumptions not often met byplant species data (McCune and Grace, 2002). CART models are alsononparametric in nature and have become fairly common in eco-logical applications that involve geographic information systemsdue to the binary, if. . .then. . ., statements as represented withinthe model tree (Moore et al., 1991). These trees act much the sameas a dichotomous key used for plant identification, and thereforeexhibit a familiar format and explanatory model for researchersto follow. For our dataset, it was determined that due to the highdegree of environmental overlap in the categorical response vari-able, classification trees were both unstable and left the task ofpruning (i.e. variable selection) the tree as arbitrary and difficultdecisions to make.

2.5. Model calibration

Model calibration as defined by Guisan and Zimmermann (2000)involves optimizing the model by assessing fit, as well as predic-tor variable selection and transformation. Since the goals of theanalysis were to both explain and predict patterns in the data, theselection of independent variables hinged on both the explanatorypower of the final variables and the model fit. First, we filteredout GIS-derived variables by considering their precision, such assources of error that affect their spatial accuracy. We also consid-ered their ability to act as surrogates for resources that explain plantdistributions and removed any redundant variables by looking forcollinearity issues. After this step, we used deviance reductionmethods to select those variables for best model fit.

Surface climate maps of temperature and precipitation havemultiple sources of error, such as interpolation errors and theinability to track spatially unique microclimates. In contrast, topo-

graphic variables, derived from an accurate digital elevation model,are comparatively very precise, especially in mountainous terrain(Guisan and Zimmermann, 2000). These variables such as elevation,slope, aspect, and in our case solar radiation, have been shown to

http://www.exelisvis.com/http://www.mrlc.gov/http://www.r-project.org/http://www.python.org/http://soildatamart.nrcs.usda.gov/http://www.r-project.org/

206 C.M. Mittanck et al. / Ecological Modelling 288 (2014) 203212

Table 1Variables considered for use in GLM. Variables were calculated using R (version 2.14.2, R Foundation for Statistical Computing, URL:http://www.r-project.org/http://http://)/http://)/) and Python (version 2.7, Python Software Foundation, URL: http://www.python.org/http://http://)/http://)/) scriptinglanguages and integrated within ArcGIS Model Builder (version 10, Environmental Systems Research Institute, Inc., Redlands, CA).

Climate Physical Biophysical

DAAT (daily average air temperature): annualdaily average air temperature averaged overan 18 year period. Downscaled to 30 mresolution

SolRad (solar radiation): annual daily averageradiation (WH/m2/day)

SWB (site water balance): running sum of the differencebetween TP and pET, never exceeding soil available watercapacity (cm/cm; negative values indicate deficit)

GDD (growing degree-days): annualsummation of daily average air temperatures>0.0 C

Elevation: 30 m National Elevation Dataset pET (potential evapotranspiration): Annual daily averagepotential evapo-transpiration (cm/day)

n rang

bervwGf(

mcgiowswPluAbl

vwtnwa

a

u

2

wpasduoo

e(a

TP (total precipitation): total accumulatedprecipitation over a year (cm/yr)

Aspect: cosine transformatioto 1 (wetdry)Slope (percent)

e effective surrogates for resource gradients (Austin, 1980; Mooret al., 1991; Stage and Salas, 2007). However, over large scales theseelationships break down and the skewed response curves to theseariables may be difficult to model. Site water balance (SWB) comesith a degree of spatial uncertainty when calculated remotely usingIS methodologies. The greatest source of spatial uncertainty stemsrom course-scale soil maps used to derive soil water availabilityMittanck, 2012).

A trade-off between precision and spatial uncertainties wasade. In order to retain some degree of large-scale predictiveapability within the model, total annual precipitation (TP) androwing-degree-days (GDD) were selected, along with all phys-cal variables of solar radiation, elevation, aspect, and slope inrder to capture fine scale distribution patterns. Collinearity issuesere avoided within the model by eliminating those variables thathowed a strong linear dependency. A linear dependency occurshen one variable is a weighted average of another variable(s).otential evapotranspiration was removed as it was essentially ainear adjustment to solar radiation, indicated by a Pearsons prod-ct moment correlation value between the two variables of 0.99.

quadratic term for total precipitation was included, as it woulde expected for tree species to show a unimodal response across aarge-scale moisture gradient.

To optimize the model all possible combinations of selectedariables were fit within the regression model. The best-fit modelas determined by the amount of deviance reduction estimated

hrough adjusted deviance squared (adjusted D2). This takes theumber of observations (n) and predictors (p) into account byeighting D2 by the residual degrees of freedom (Weisberg, 2005),s:

djusted D2 = 1 [n 1n p

]

[1 D2

]. (1)

Significant differences in deviance reduction were assessedsing a 2 approximation with = 0.05.

.6. Model evaluation

There are many methods for evaluating a binary linear model,hich can essentially be thought of as quantifying in some way theredictive accuracy of the model (Harrell et al., 1996). Predictiveccuracy is often divided by quantifying the fit of the model throughome criterion of goodness-of-fit and assessing how well the modeliscriminates between sites that are occupied and sites that arenoccupied (McKenzie et al., 2003). Though the two measures areften related, Manel et al. (2001) point out the need to use a rangef criteria to assess model performance.

To assess the stability of the model fit, we used an optimism

stimate of residual deviance (D2) with 200 bootstrapped samplesHarrell et al., 1996; Guisan and Zimmermann, 2000). Although D2

lso gives us a measure of predictive accuracy, binary linear modelsing from 0

with predictions stated as probabilities can be further explored byusing contingency tables and/or receiver operating characteristic(ROC) curves. For our particular model the predicted response rep-resents probability of a stable aspen habitat. Therefore the modelsability to discriminate between aspen and conifer refers only to thedegree of overlap between their respective distributions in environ-mental space. We used ROC curves to assess general overlap. Thetrue positive rate was presented as a function of the false positiverate, with the true positive rate being the proportion of correctlypredicted presences and the false positive rate being the proportionof incorrectly predicted presences. A models ability to identify sitesthat are occupied is always a trade-off with the rate that it incor-rectly identifies presences. In our case, if the model has no ability todiscriminate between aspen and conifer the ROC curve would be a45 line and the area under the line would be equal to 0.5. To quan-tify the discrimination ability of the model the area under the ROCcurve was approximated by the WilcoxonMannWhitney statis-tic, which gives the probability of the model correctly identifying arandomly drawn sample (Sing et al., 2005).

Data were further organized into strip plots of aspen and coniferpresence for elevation, growing degree days, and total precipitationover the entire range of the site in order to determine the distri-butional limits to aspen and conifer imposed by site boundaries,potential distributional limits imposed by ecological tolerance, andto compare the range of the variables amongst sites. For thesegraphs presence data of aspen and conifer plots was derived fromforest type groups. Plots with aspen cover >10% were grouped asaspen. Plots with conifer cover >10% were grouped as conifer pres-ence. According to this classification scheme, mixed aspenconiferplots are classified as both aspen and conifer. In addition, samplesize by forest type groups was compared across sites to deter-mine if certain sites may be disproportionately influencing the GLMresults; a preponderance of observations of the response variableas either 0 or 1 has been shown to lead to biased estimated proba-bilities of presence in GLMs (McKenzie et al., 2003).

2.7. Model application

Because of the strong potential for overlap between aspen andconifer distributions, an analytical approach was needed to dis-cretely classify the model output. If we consider the proportionof conifer presence as a metric for interpreting stable aspen habitatwe can make the decision on how to classify stable habitat accord-ing to the prediction-conditioned fallout-rate (PFrate) (Sing et al.,2005). In terms of this study, PFrate is the rate of conifer-presentplots to all plots predicted by the model to be aspen habitat:PFrate = FPTP

+ FP, (2)

where FP = number of plots falsely predicted to be aspen, andTP = number of plots correctly predicted to be aspen. This rate can

http://www.r-project.org/http://http://http://http://www.python.org/http://http://http://

C.M. Mittanck et al. / Ecological Modelling 288 (2014) 203212 207

F , and (i al sam

bTtbm

tNpvovciwbTlrr7e

3

rglmet

ig. 2. Range of (a) total annual precipitation (TP), (b) growing degree days (GDD)ndicated by A, conifer presence by C. Each line in stripchart represents individu

e assessed for any given probability cutoff point (e.g. threshold).he optimum threshold to classify stable aspen habitat will be onehat has the lowest FPrate. This method provides an analyticallyased criteria to discretely classify the continuous output from theodel.The results of the hard classification of the model were applied

o the Southwest regional Gap landcover dataset (SWReGap) (USGSational Gap Analysis Program, 2004) in order to arrive at cover ofotential stable aspen communities across the state of Utah. Modelariables at a statewide scale were attained using the same meth-ds as those for the study sites. To achieve accurate solar radiationalues across the entire range of the state, solar radiation was cal-ulated within 1 degree latitudinal zones (5 zones across the state),n order to account for differences in solar zenith angles. The modelas only used to delineate stable and seral aspen in areas classifiedy the SWReGap as Rocky Mountain Aspen Forest and Woodland.hese are upland forests and woodlands dominated by P. tremu-oides without a significant conifer component (i.e.

208 C.M. Mittanck et al. / Ecological Modelling 288 (2014) 203212

Table 2Generalized linear model (GLM) results. Four terms were included in the final model:solar radiation (SolRad), total annual precipitation (TP), unimodal response to TP(TP2), and an interaction term between SolRad and TP (SolRad TP). GLM wasparameterized using aspen (1) and conifer-present plots (0) from all four sites.

Estimate Std. error z-Value p-Value

Intercept 0.7467 3.242 0.230 0.8171SolRad 0.004037 0.0009802 4.118

C.M. Mittanck et al. / Ecological Modelling 288 (2014) 203212 209

Table 3Percentage of plots by site. Forest Types are grouped according to binary response variable used in the generalized linear model. Total numbers are in parentheses.

Forest types DLL (199) Franklin Basin (238) Book Cliffs (128) Cedar Mountain (228) Total (793)

UaialSccweeoSs1

4

4

steina

Fiap

Aspen (GLM input = 1) 56% (111) 27% (68)Conifer & mixed combined (GLM input = 0) 44% (88) 73% (170)

sing an overlay analysis this hard classification scheme waspplied to SWReGap aspen landcover (Fig. 5). A total of 6328 km2

s classified as Rocky Mountain Aspen Forest and Woodlandccording to the SWReGap dataset; representing 3% of the totaland area within the state. An application of our model to thisWReGap class indicates 19% of stands are potentially stable aspenommunities, whereas the remaining 81% should be seral aspenommunities. It should be noted that our sample dataset fromhich the model was calibrated did not cover the entire param-ter range of the SWReGap aspen dataset and therefore somextrapolation occurred near the edges. The range of SolRad withinur sample was 17605500 W h/m2/d; the range of SolRad withinWReGap was 14535571 W h/m2/d. The range of TP within ourample was 40140 cm; the range of TP within SWReGap was3169 cm.

. Discussion

.1. Interpreting a stable aspen habitat

The habitat-based modeling approach used in this study istatic. It relies on input of observational data at a specific point inime as opposed to dynamic simulation modeling which is param-

terized according to the physiological behavior of the speciesnvolved as well as detailed knowledge of successional mecha-isms (Guisan and Zimmermann, 2000). Dynamic modeling inspenconifer systems may be presumptuous since mechanisms

ig. 5. Map depicting statewide potential stable and seral aspen cover. The generalized lin.e. proportions of aspen to conifer. This habitat classification was applied to aspen covernd Cedar Breaks area (left) and all of Utah (right), show current aspen cover that has therobability predicted by the model) and current seral aspen cover that is the most vulner14% (18) 56% (128) 41% (325)86% (110) 44% (100) 59% (468)

that affect succession in these systems are poorly understood. Astatic modeling approach, on the other hand, assumes empiricalobservations of a species presence are subject to biotic interac-tions and competitive exclusion, therefore accounting for, and oftenuncovering, successional mechanisms. However, to interpret sta-ble aspen habitat using a static model approach, we made theassumption that aspenconifer distributions are near an equilib-rium in late succession. In other words, we assumed that mostsample plots represent near climax or stable communities. For thisreason systems influenced by successional dynamics are difficultto model using a static approach (Lees and Ritman, 1991; Guisanet al., 1999). To compensate for this analytical drawback, a highnumber of plots were sampled across four large-extent study sitesunder the assumption that general patterns would be evident inthe results. There is support that aspen and conifer forests acrossthe Intermountain West, in general, have shifted to late succes-sional stages due to limited fire events caused by moist climatesand wide-scale fire suppression over the past century (Bartos et al.,1983; Gruell, 1983; Kay, 2003; Kulakowski et al., 2004; Kusbach,2010; Rogers et al., 2011). Therefore, we have assumed that alarge-extent intensive sampling efforts should uncover generalrelationships of stable aspenconifer distributions to environmen-tal variables, despite the analytical noise induced by non-stablestates, such as recent disturbances to conifer dominated stands that

resulted in early-succession pure aspen stands. It is likely that someof the overlap in environmental space between pure aspen plotsand conifer-present plots is due to the analytical noise mentionedabove. However, despite this noise there appears to be a strong

ear model output was hard classified using prediction-conditioned fallout-rates, classified according to the SWReGap dataset. The resulting maps, Cedar Mountain

highest potential to be stable (GLM probabilities 0.700.97; 0.97 is the maximumable to conifer encroachment (GLM probabilities 00.70).

2 cal Mo

pams

rteeramb6arpatwidhecIeams2

idpthihuMwsdti(cc5tItaadepeisast

g

10 C.M. Mittanck et al. / Ecologi

attern indicating a relationship between environmental variablesnd conifer encroachment within aspen communities as evinced byuch higher proportions of pure aspen plots in the environmental

pace classified as stable aspen habitat.If the environmental variables used in this model were entirely

epresentative of ecological mechanisms that explain the persis-ence of a stable aspen community or the prevention of coniferncroachment, we would expect to see aspen plots inhabiting annvironmental space which is exclusive from conifers. Instead, theesults indicate that an interaction between topographic positionnd moisture progressively influence the probability of encroach-ent but do not preclude conifers entirely. This gradient cane observed in a bivariate plot (Fig. 4). Within the TP range of0100 cm, the proportion of conifer to aspen generally decreaseslong the vertical gradient represented by solar radiation. This sameelationship is shown in the spatial representation of the modelredictions (Fig. 3), where, within the same TP range, higher prob-bilities of aspen occur with increasing solar radiation, equivalento moving onto plateau tops and south-facing slopes. In addition, ife consider that we have taken a snap-shot at a particular point

n time, conifer to aspen proportions along this topographic gra-ient likely indicate probability of encroachment. In these terms,abitat delineated at the furthest end of this gradient has the high-st proportion of aspen to conifer and can be interpreted as aspenommunities that are the least vulnerable to conifer encroachment.t is likely that the rate of conifer encroachment into these stands isxtremely slow, along the lines of hundreds to thousands of years,nd that the rate of conifer establishment does not exceed coniferortality. For management purposes, such stands are often clas-

ified as stable aspen communities (Mueggler, 1988; Rogers et al.,014a).While our results indicate that topographic position strongly

nfluences the ability of conifer to encroach into aspen stands, theegree of its influence is dependent upon the amount of total annualrecipitation (TP). This outcome shows aspen shifting from rela-ively steep north facing slopes, found at the Book Cliffs site, toigher solar radiation sites on plateau tops and gentle south fac-ng slopes as we move to higher precipitation values, while theighest probability for stable aspen communities is found at thepper range of solar radiation within 60 to 90 cm of TP (Fig. 3).any plant distribution studies have found similar interactions,here plant species environmental requirements can be met byhifting topographic positions along an elevational or climatic gra-ient (Daubenmire, 1943; Whittaker, 1967). Although aspen hashe ability to grow in many disparate environmental conditions,ts distribution has been noted to be water-limited. In Mueggler1988) extensive survey of both aspen and mixed aspenconiferommunities in the Intermountain Region, he found that theseommunities required at least 38 cm, but more commonly over1 cm of total annual precipitation. Our results showing the dis-ributions of aspen-present plots align well with this (Fig. 2ac).nterestingly, however, in our habitat-based analysis, we foundhat at sites with less than 60 cm of TP, it was more likely forspen to share the same topographic position as conifer species,nd therefore more likely to be encroached and, barring significantisturbance, converted to a conifer climax community. This is mostvident at the Book Cliffs site, where large contiguous stands ofure upland aspen are much less common and conifer stands mayxist in lower elevation, north facing, locations. Topographic pos-tions, such as gently sloping areas, plateau tops, and south-facinglopes, that support stable aspen communities at the other sites,re instead inhabited primarily by Gambel oak and the occasional

parse and shrub-like aspen. According to our model this appearso be a result of low total annual precipitation.

On the other end of the TP spectrum, at sites with TP valuesreater than 90 cm, conifer dominates, with no aspen or even mixeddelling 288 (2014) 203212

aspenconifer plots found above 115 cm (precipitation values thatare only found at the Franklin Basin site). This likely does not indi-cate a limit of ecological tolerance to moisture; aspen is often foundin riparian areas with very high soil water availability (Mueggler,1988). It is possible that this pattern is due to the length of the grow-ing season as distributions of aspen are thought to be constrainedby short growing seasons at high elevations (Mueggler, 1988). It isalso possible that TP is a proxy for other factors that have a strongpositive influence on conifer encroachment. Such a factor may below fire-return intervals at high elevations, resulting in long succes-sional timelines allowing conifer dominance. Fire return intervalsinto the subalpine zone (considered to be approximately 2750 mfor latitudes similar to the Franklin Basin site) have been estimatedby various studies to be greater than 150 years, though probablymore along the lines of 300400 years (Romme, 1982; Gruell, 1983;Wadleigh and Jenkins, 1996; Bigler et al., 2005), allowing plenty oftime for conifer to dominate even at a slow rate of encroachment.

The interaction between solar radiation and precipitation in ourmodel indicates that stable aspen communities may be the result ofa very specific moisture regime (Rogers et al., 2014a). This suggeststhat habitat differentiation between stable and seral aspen mayrequire more temporal and spatial resolution than our course-scalevariables provide. In addition, we are lacking soil characteristicswithin the model. Soil types and resulting soil water availabilitiesmay be related to aspenconifer successional dynamics. Pure aspencommunities have been shown to be correlated to specific soil sub-orders within the order of Cryoborolls (Debyle and Winokur, 1985;Cryer and Murray, 1992). These soils are characterized by high inputof soil organic matter (SOM), warm soil temperatures in upper hori-zons, and good soil moisture content through much of the growingseason, all in turn promote microbial decomposition and thereforeincreased nutrient availability. These warm soil conditions havebeen shown to promote suckering and expansion of aspen clones(Maini and Horton, 1966; Williams, 1972; Bailey, 1974), providinga mechanism for non-fire related regeneration.

These same soil conditions also support productive understoreycommunities, which may play a role in inhibiting conifer establish-ment. Thick aspen and understorey leaf litter at the soil surfacemay inhibit seeds from coming into contact with the mineral soiland thus hinder germination of conifer seedlings (Coates et al.,1991). Additionally, Langenheim (1962) reported that there wereincreases in conifer encroachment where the understorey was thincompared to where it was thick. According to such accounts, heavylong-term ungulate grazing may provide the disturbance neededfor conifers to establish and recruit into the overstory at sites thatwould normally support stable aspen communities.

4.2. Application of aspen habitat modeling

Due to recent issues raised over decline of aspen cover, in partrelated to conifer encroachment, aspen restoration projects areplanned across many forested lands in Utah (OBrien et al., 2010). Animportant step in determining a management action is identifyingthe aspen community type (Rogers et al., 2014a). While stand-levelcharacteristics such as structure, soils, understorey community,grazing and fire regimes are useful in this identification process,these data are often not available at a large landscape scales andtherefore problematic for predictive modeling. However landscapescale decisions are often needed in regards to forest management.Considering this need, we applied our model across the state of Utahto areas classified as Rocky Mountain Aspen Forest and Wood-land according to the SWReGap landcover dataset (Fig. 5). Such

an application delineates potential stable and seral aspen commu-nities and would be useful for the initial planning stages of largeaspen restoration projects, other planning purposes, and policydevelopment.

cal Mo

msamecmacttwigteps

biicadeaTsfim2tItsdveprtoeH(n

mcrimci

A

EM2ta

C.M. Mittanck et al. / Ecologi

This application indicates the probability of a stable aspen com-unity at a particular site. Since this metric is on a continuouscale, decisions can be made for aspen stands relative to otherspen stands, either within the same management area or acrossanagement areas. Stands with a low model probability may bexpected to have high rates of encroachment and therefore requireontinuous treatment or disturbance to maintain an aspen com-unity. These stands are likely driven by successional processesnd in the absence of a stand-replacing fire or selective removal ofonifer trees, they will convert to conifer dominance. Identifyinghese stands early on in the successional process may be impor-ant in avoiding a threshold that leads to a transition to coniferoodlands that ultimately exclude aspen. However, our results also

ndicate that conifer encroachment follows a gradient. This sug-ests the potential for a persistent mixed aspenconifer forestype (Zier and Baker, 2006). Within these stands, conditions mayxist where conifer cover does not reach a threshold sufficient toromote stand-replacing fire. Such stands may persist in a mixedtate and may not require treatments to prevent conifer dominance.

The model only indicates probabilities of a site supporting sta-le or seral aspen communities, and therefore should be consideredn concert with other criteria when classifying and determin-ng appropriate management actions for a specific stand. Suchriteria for stable aspen communities may include stand-level vari-bles such as: age-cohort structure (Mueggler, 1988), regenerationynamics, such as gap dynamics or episodic suckering (Kurzelt al., 2007), and soil characteristics that indicate aspen has beent the site for multiple generations (Debyle and Winokur, 1985).hese characteristics along with the environmental setting stronglyupport classification of a stable aspen community. This classi-cation is important, as management strategies for stable aspenay be entirely different from seral communities (Rogers et al.,014a). Stable aspen is not considered to be a flammable forestype (Debyle and Winokur, 1985; Fechner and Barrows, 1976).n fact, recent work suggests that there are multiple aspen fireypes that range widely in the frequency and intensity of burnusceptibility (Shinneman et al., 2013). In place of stand-replacingisturbance, stable aspen has been shown to sufficiently regenerateia gap dynamics and episodic suckering (Kurzel et al., 2007; Rogerst al., 2010). This suggests management strategies should focus onromoting sufficient regeneration without fire, which may requireeducing or removing ungulate grazing pressures, or selective cut-ing to mimic gap dynamics (Rogers et al., 2014b). Observationsf 1200 stems/ha have been suggested as sufficient aspen regen-ration within stable communities (Debyle and Winokur, 1985).owever, such regeneration may need to be protected, as Burton2004) found that browsing >20% of the young aspen leaders canegatively affect regeneration in the stand.Accurate delineation of seral and stable aspen types will ulti-

ately improve assessment, management, and restoration of theseritical forest communities. Models relying on GIS-derived envi-onmental datasets allow large-scale applications that can assistn identifying aspen functional types across the landscape. Theseethods may be useful to forest managers faced with rapidlyhanging aspen conditions in the Rocky Mountains of North Amer-ca.

cknowledgements

This work was funded by grants from the Utah Agriculturalxperiment Station, USU Cedar Mountain Initiative, USDA Rocky

ountain Research Station, Utah BLM, USDA NRI Grant no. UTAR-007-01475, and the Utah State University Ecology Center. Wehank Andy Leidolf and Tony Kusbach for helping with field worknd providing field data for accuracy assessments. Special thanksdelling 288 (2014) 203212 211

to Chris McGinty, Chris Garrard, and Alexander Hernandez at theUSU GIS/Remote Sensing Laboratory for their technical expertise.

References

Austin, M., 1980. Searching for a model for use in vegetation analysis. Plant Ecol. 42,1121.

Bailey, A.W., 1974. Aspen invasion in a portion of the Alberta parklands. J. RangeManage. Arch. 27, 263266.

Baker, F.S., 1918. Aspen as a temporary forest type. J. For. 16, 294300.Bartos, D.L., Ward, F., Innis, G., 1983. Aspen Succession in the Intermountain West: A

Deterministic Model. USDA Forest Service [General Technical Report INT-153].Intermountain Forest and Range Experiment Station, Ogden, UT, US.

Bartos, D.L., Campbell, R.B., 1998. Decline of quaking aspen in the Interior West:examples from Utah. Rangelands 20, 1724.

Betters, D.R., Woods, R.F., 1981. Uneven-aged stand structure and growth of RockyMountain aspen. J. For. 79, 673676.

Bigler, C., Kulakowski, D., Veblen, T.T., 2005. Multiple disturbance interactions anddrought influence fire severity in Rocky Mountain subalpine forests. Ecology 86,30183029.

Burton, D., 2004. Aspen Delineation Project. BLM, U.S. Forest Service, and CaliforniaDept. of Fish, http://www.blm.gov/ca/pa/biology/aspen.html.

Coates, K., Emmingham, W., Radosevich, S., 1991. Conifer-seedling successand microclimate at different levels of herb and shrub cover in aRhododendronVacciniumMenziesia community of south central BritishColumbia. Can. J. For. Res. 21, 858866.

Crawford, J., McNulty, S., Sowell, J., Morgan, M., 1998. Changes in aspen com-munities over 30 years in Gunnison County, Colorado. Am. Midl. Nat. 140,197205.

Cryer, D.H., Murray, J.E., 1992. Aspen regeneration and soils. Rangelands 14,223226.

Daubenmire, R.F., 1943. Vegetational zonation in the Rocky Mountains. Bot. Rev. 9,325393.

Debyle, N.V., Winokur, R.P. (Eds.), 1985. Aspen: Ecology and management in thewestern United States. USDA Forest Service [General Technical Report RM-119].Rocky Mountain Forest and Range Experiment Station, Fort Collins, CO, US.

Di Orio, A.P., Callas, R., Schaefer, R.J., 2005. Forty-eight year decline and fragmenta-tion of aspen (Populus tremuloides) in the South Warner Mountains of California.For. Ecol. Manage. 206, 307313.

Fechner, G.H., Barrows, J.S., 1976. Aspen stands as wildfire fuel breaks. USDA For-est Service [Eisenhower Consortium Bulletin no. 4]. Rocky Mountain Forest andRange Experiment Station, Fort Collins, CO, US.

Grier, C.G., Running, S.W., 1977. Leaf area of mature Northwestern coniferousforests: relation to site water balance. Ecology 58, 893899.

Gruell, G.E., 1983. Fire and Vegetative Trends in the Northern Rockies: Interpre-tations from 18711982 Photographs. USDA Forest Service [General TechnicalReport INT-158]. Intermountain Research Station, Ogden, UT, US.

Guisan, A., Weiss, S.B., Weiss, A.D., 1999. GLM versus CCA spatial modeling of plantspecies distribution. Plant Ecol. 143, 107122.

Guisan, A., Zimmermann, N.E., 2000. Predictive habitat distribution models in ecol-ogy. Ecol. Modell. 135, 147186.

Harrell Jr., F.E., Lee, K.L., Mark, D.B., 1996. Multivariable prognostic models: issues indeveloping models. Evaluating assumptions and adequacy, and measuring andreducing errors. Stat. Med. 15, 361387.

Jensen, M.E., Haise, H.R., 1963. Estimating evapotranspiration from solar radiation.J. Irrig. Drain. Div. 89, 1541.

Kay, C.E., 2003. Long-Term Vegetation Change on Utahs Fishlake National Forest: AStudy in Repeat Photography. Utah State University Extension, Logan, UT, US.

Kulakowski, D., Veblen, T.T., Drinkwater, S., 2004. The persistence of quakingaspen (Populus tremuloides) in the Grand Mesa area, Colorado. Ecol. Appl. 14,16031614.

Kurzel, B.P., Veblen, T.T., Kulakowski, D., 2007. A typology of stand structure anddynamics of Quaking aspen in northwestern Colorado. For. Ecol. Manage. 252,176190.

Kusbach, A., 2010. Terrestrial ecosystem classification in the Rocky Mountains,Northern Utah. In: PhD. Thesis. Utah State University, Logan, UT, US.

LaMalfa, E.M., Ryel, R.J., 2008. Differential snowpack accumulation and waterdynamics in aspen and conifer communities: implications for water yield andecosystem function. Ecosystems 11 (4), 569581.

Langenheim, J.H., 1962. Vegetation and environmental patterns in the Crested Buttearea, Gunnison County, Colorado. Ecol. Monogr. 32, 249285.

Lees, B.G., Ritman, K., 1991. Decision-tree and rule-induction approach to integrationof remotely sensed and GIS data in mapping vegetation in disturbed or hillyenvironments. Environ. Manage. 15, 823831.

Lowry, J., Ramsey, R., Thomas, K., Schrupp, D., Sajwaj, T., Kirby, J., Waller, E., Schrader,S., Falzarano, S., Langs, L., et al., 2007. Mapping moderate-scale land-cover oververy large geographic areas within a collaborative framework: a case study ofthe Southwest Regional Gap Analysis Project (SWReGAP). Remote Sens. Environ.108, 5973.Maini, J.S., Horton, K.W., 1966. Vegetative propagation of Populus spp. 1: influenceof temperature on formation and initial growth of Aspen suckers. Can. J. Bot. 44,11831189.

Manel, S., Williams, H.C., Ormerod, S.J., 2001. Evaluating presenceabsence modelsin ecology: the need to account for prevalence. J. Appl. Ecol. 38, 921931.

http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0005http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0010http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0015http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0015http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0015http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0015http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0015http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0015http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0015http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0015http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0015http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0015http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0015http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0015http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0020http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0025http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0030http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0035http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0040http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0045http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0050http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0055http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0055http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0055http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0055http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0055http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0055http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0055http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0055http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0055http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0060http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0060http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0060http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0060http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0060http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0060http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0060http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0060http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0060http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0060http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0060http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0060http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0065http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0070http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0075http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0080http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0085http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0090http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0095http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0095http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0095http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0095http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0095http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0095http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0095http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0095http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0095http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0095http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0095http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0095http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0095http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0100http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0105http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0105http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0105http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0105http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0105http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0105http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0105http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0105http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0105http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0105http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0105http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0105http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0105http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0110http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0115http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0120http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0125http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0130http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0135http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0140http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0145http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0150http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155http://refhub.elsevier.com/S0304-3800(14)00292-0/sbref0155

2 cal Mo

M

M

M

M

M

M

M

O

R

R

R

R

R

R

R

R

12 C.M. Mittanck et al. / Ecologi

anier, D.J., Laven, R.D., 2002. Changes in landscape patterns associated with thepersistence of aspen (Populus tremuloides Michx.) on the western slope of theRocky Mountains, Colorado. For. Ecol. Manage. 167, 263284.

cCune, B., Grace, J., 2002. Analysis of Ecological Communities. MJM SoftwareDesign, Gleneden Beach, OR, US.

cCune, B., Keon, D., Marrs, R.H., 2002. Equations for potential annual direct incidentradiation and heat load. J. Veg. Sci. 13, 603606.

cKenzie, D., Peterson, D., Thornton, P., 2003. Climatic and biophysical controls onconifer species distributions in mountain forests of Washington State, USA. J.Biogeogr. 30, 10931108.

ittanck, C.M., 2012. Exploring a stable aspen niche within aspenconifer forests ofUtah. In: M.S. Thesis. Utah State University, Logan, UT, US.

oore, D., Lees, B., Davey, S., 1991. A new method for predicting vegetation distribu-tions using decision tree analysis in a geographic information system. Environ.Manage. 15, 5971.

ueggler, W.F., 1988. Aspen Community Types of The Intermountain Region. USDAForest Service [General Technical Report INT-250]. Intermountain ResearchStation, Ogden, UT, US.

Brien, M., Rogers, P., Mueller, K., MacWhorter, R., Rowley, A., Hopkin, B., Chris-tensen, B., Dremann, P., 2010. Guidelines for aspen restoration on the NationalForests in Utah. Utah Forest Restoration Working Group, Western AspenAlliance. Utah State University, Logan, UT, US.

oberts, D.W., Cooper, S.V., 1989. Concepts and Techniques of Vegetation Map-ping. USDA Forest Service [General technical Report INT-257]. IntermountainResearch Station, Ogden, UT, US.

ogers, P., 2002. Using Forest Health Monitoring to assess aspen forest cover changein the southern Rockies ecoregion. For. Ecol. Manage. 155, 223236.

ogers, P., Leffler, A., Ryel, R., 2010. Landscape assessment of a stable aspen com-munity in southern Utah, USA. For. Ecol. Manage. 259, 487495.

ogers, P.C., BartosF D.L., Ryel, R.J., 2011. Historical patterns in lichen communitiesof montane quaking aspen forests. Adv. Env. Res. 15, 3364.

ogers, P.C., Landhusser, S.M., Pinno, B.D., Ryel, R.J., 2014a. A functional frame-work for improve