LETTER Projected climate-driven faunal movement routes

J. J. Lawler,1* A. S. Ruesch,1

J. D. Olden2 and B. H. McRae3

AbstractHistorically, many species moved great distances as climates changed. However, modern movements willbe limited by the patterns of human-dominated landscapes. Here, we use a combination of projected cli-mate-driven shifts in the distributions of 2903 vertebrate species, estimated current human impacts on thelandscape, and movement models, to determine through which areas in the western hemisphere species willlikely need to move to track suitable climates. Our results reveal areas with projected high densities of cli-mate-driven movements – including, the Amazon Basin, the southeastern United States and southeasternBrazil. Some of these regions, such as southern Bolivia and northern Paraguay, contain relatively intactlandscapes, whereas others such as the southeastern United States and Brazil are heavily impacted byhuman activities. Thus, these results highlight both critical areas for protecting lands that will foster move-ment, and barriers where human land-use activities will likely impede climate-driven shifts in species distri-butions.

KeywordsAmphibians, birds, climate change, connectivity, corridors, mammals, movement, range shifts.

Ecology Letters (2013)

INTRODUCTION

In the past, as the Earth’s climates changed, species adapted to newenvironments, moved to track suitable climates, or went extinct.Many species have already responded to modern climate change ina similar fashion (Parmesan et al. 1999; Pounds et al. 2006; Moritzet al. 2008), and projected changes in climate for the coming centurywill likewise result in distributional shifts of many species (Thuilleret al. 2005; Ara!ujo et al. 2006) and the loss of others (Thomas et al.2004). Mounting concern regarding the rapidity of climate changehas intensified scientific and conservation interest in connectivityand the need for improved land management and restoration topromote permeable landscapes.To persist in a rapidly changing climate, much of today’s biota will

need to traverse extensive human-dominated landscapes (Kareivaet al. 2007), and thus species attempting to shift their distributions willface many challenges that did not confront their ancestors. By someestimates, less than 17% of the terrestrial landscape has escaped thedirect impact of human activities (Sanderson et al. 2002), approxi-mately half of the land surface has been converted to agriculture (Mil-lennium Ecosystem Assessment 2005), and less than 1% of theworld’s rivers remain unaffected by humans (V€or€osmarty et al. 2010).Organisms attempting to relocate to more suitable climates must con-tend with expansive agricultural landscapes, road networks, dammedrivers and growing human settlements. These and other human altera-tions to the landscape create well-documented barriers to movementfor some species and reduce movement of others (Sieving et al. 1996;Forman & Alexander 1998).Conservation efforts that seek to facilitate species movement

through these human-dominated landscapes will require an under-standing of where species will likely need to move to track suitableclimates and what the available and most suitable routes for

enabling these movements will be. Projections of both the velocityand the direction of shifting climatic conditions provide a coarseapproximation of how many species might need to move to tracksuitable climates (Loarie et al. 2009; Ackerly et al. 2010). Projectionsof changes in climatic suitability for individual species provide amore targeted assessment of the areas to which species may need tomove to track climates (Thuiller et al. 2005; Ara!ujo et al. 2006;Huntley et al. 2008; Cheung et al. 2009). Finally, projected changesin species-specific climatic suitability at multiple time periods canilluminate climatic barriers that may prevent species from trackingclimate change (Phillips et al. 2008; Early & Sax 2011). Despitethese recent advances, considerable uncertainty remains regardinghow human land use may alter movement routes defined by climatesuitability alone.Here, we investigate potential movement routes for vertebrates

across human-dominated landscapes of the western hemisphere inresponse to projected changes in end-of-century climate. By cou-pling recent advances in species distribution modeling with land-scape connectivity analyses, we map routes that would allow 2903species of mammals, birds and amphibians to track suitable climateswhile avoiding, when possible, areas that are heavily impacted byhuman activities. We analyse the resulting maps to highlight areasthat will likely provide for species movements in the future and con-versely areas through which many species will likely need to movebut in which opportunities for movement will be more limited.

MATERIALS AND METHODS

Approach

To identify critical areas for fostering climate-driven species move-ments, we first projected changes in the distribution of climatically

1School of Environmental and Forest Sciences, University of Washington,

Seattle, WA, 98195, USA2School of Aquatic and Fishery Sciences, University of Washington, Seattle,

WA, 98195, USA

3The Nature Conservancy, North America Region, Seattle, WA, 98101, USA

*Correspondence: E-mail: jlawler@uw.edu

© 2013 John Wiley & Sons Ltd/CNRS

Ecology Letters, (2013) doi: 10.1111/ele.12132

suitable areas for each species under 10 different projected futureclimates. We then used a connectivity-modelling approach to plotroutes for each species from areas that were projected to becomeclimatically unsuitable to areas that were projected to be climaticallysuitable and from areas that are currently climatically suitable toareas that were projected to become newly climatically suitable.These routes were mapped in such a way as to avoid the mostheavily human-impacted parts of the landscape. For each 50-km by50-km grid cell in the western hemisphere, we then quantified theprojected magnitude of movement across species, the average direc-tion of movement in a cell and the degree of agreement in thedirection of movements across species.

Data

Climate data were downscaled from the University of East AngliaClimatic Research Unit’s (CRU’s) CL 1.0 (New et al. 1999), CL 2.0(New et al. 2002) and TS 2.1 (Mitchell & Jones 2005) climate datasets to a 50-km by 50-km grid using locally weighted, lapse-rate-adjusted interpolation. We then averaged monthly climate data from1961 to 1990 to produce a base-line time period. Future climatedata were taken from 10 general circulation model (GCM) simula-tions archived by the World Climate Research Programme’s(WCRP’s) Coupled Model Intercomparison Project phase 3(CMIP3) (Table S1). Climate projections were averaged over a30-year period from 2071 to 2100 and represent climates simulatedfor a mid-to-high (SRES A2) greenhouse-gas emissions scenario(Nakicenovic et al. 2000). In all, 39 bioclimatic variables were usedin the modeling process (Lawler et al. 2009).Species distribution data were taken from digital range maps for

birds (Ridgely et al. 2003), mammals (Patterson et al. 2003) andamphibians (data available online, www.globalamphibians.org). Forbirds, only breeding ranges were included. Species ranges weremapped to the 50-km by 50-km grid and all grid cells overlapping arange were treated as ‘presences’ and all grid cells that did not over-lap the species’ range were considered to be ‘absences’.To model potential species movements, we used a map of the

human influence index representing the cumulative effects of land-use change (e.g. built environments and agriculture), human popula-tion density, accessibility (e.g. distance to roads) and night-timelighting (Sanderson et al. 2002). We mapped the human influenceindex data to the 50-km by 50-km grid used for the climate datausing a mean aggregation statistic and a bilinear interpolation tech-nique for projection and re-sampling (Environmental SystemsResearch Institute 2009).

Species distribution models

Bioclimatic models were built using an ensemble-based machine-learning approach (random forest predictors, Cutler et al. 2007) for6933 species as part of a previous study (Lawler et al. 2009). Ran-dom forest predictors were chosen because of their ability to cap-ture interactions among variables without specifying thoseinteractions a priori, because of their ability to handle correlated vari-ables, and most importantly, because of their performance relativeto other modeling approaches when tested on the datasets used inthis study (Lawler et al. 2006; Olden et al. 2008). One model wasbuilt for each species using presence–absence as a binary responsevariable and the 39 bioclimatic variables as potential predictors.

Each random forest was based on 100 classification trees built withsubsets of the data and predictor variables. We used 80% of thepresences and 80% of the absences for each species to build themodels and used the remaining data for model testing. For eachmodel, a threshold for predicting presence was determined usingthe receiver-operating characteristic curve with the assumption thatcorrectly predicting a presence was as important as predicting anabsence (Fielding & Bell 1997). The models were evaluated basedon their ability to correctly predict the presences and absences inthe test data sets. Of the total number of species examined, modelsfor 2954—1818 birds, 723 mammals and 413 amphibians were ableto correctly predict at least 90% of the absences and 80% of thepresences in the reserved test data sets. These are the models thatwere used in this study.We used the models to project areas that will likely be climatically

suitable under the 10 simulated future climates. For each species,we identified areas that are currently climatically suitable and areprojected to become climatically unsuitable in a future climate. Wetreated these as areas of potential range contraction. Similarly, weidentified areas that are climatically unsuitable today, but are pro-jected to become climatically suitable in the future. We treated theseas areas of potential range expansion.

Species

Of the 2954 species for which we had relatively accurate models,2903 species met a set of additional criteria that allowed us to usethem in this study. On average, across the 10 GCM projections,2765 species were used for calculating species movements into areasof potential range expansion, and 2621 species were used for calcu-lating species movements out of areas of projected range contrac-tion. We did not use any species for which the bioclimatic modelsprojected a complete elimination of climatically suitable space (anaverage of 133 species across all 10 GCM projections). For model-ling movements into areas of potential range expansion, we did notuse species for which no range expansions were projected (an aver-age of 43 species across all 10 GCMs) nor did we include speciesfor which historical ranges were disconnected from projected areasof range expansion (i.e. we excluded areas that were projected to benewly climatically suitable but contained no cells that were adjacentto any cell in the species’ historical range). This latter restrictionapplied to an average of 160 species across the 10 GCM projec-tions. Similarly, for modeling movements out of areas of range con-traction, we did not include species for which projected futureranges were disconnected from areas of projected range contraction(an average of 56 species across the 10 GCM projections). By calcu-lating movement routes only for range shifts that resulted in contig-uous current and future ranges, we avoided modeling routes toareas that would likely be inaccessible to dispersing individuals.However, this restriction may have obscured some of the routesthat species will take to move longer distances as climates change.Finally, all species had some projected range contraction across all10 GCM projections.

Movement models

For each species, we modelled the potential routes of organismsfrom each grid cell in which the corresponding species-distributionmodel projected the climate to become unsuitable by the end of the

© 2013 John Wiley & Sons Ltd/CNRS

2 J. J. Lawler et al. Letter

century to an area in which climate was projected to remain suit-able. We likewise mapped potential routes from areas that were suit-able in the historical time period (1961–1990) to areas that wereprojected to become newly climatically suitable by the end of thecentury.We used current flow models from electronic circuit theory to

plot movement routes for each of the species using Circuitscapesoftware (McRae & Shah 2009). Circuitscape treats landscapes asconductive surfaces, replacing cells connected to their neighbourswith nodes connected by resistors. Patterns of current flow can beused to efficiently predict the movement patterns of random walk-ers across a landscape, where walkers are proportionally more likelyto move through low-resistance cells than high-resistance cells(McRae et al. 2008). When 1 Amp of current travels from a sourcecell to a target cell, the current flowing through each interveningcell represents the likelihood that a random walker would passthrough the cell if it started at the source and moved until itreached the target (Newman 2005; McRae et al. 2008). The result isa continuous map of movement probabilities across all possibleroutes, rather than a single, least-cost path. At the extent of thestudy area, high densities of current indicate areas through whichhigher numbers of successful dispersers are predicted to pass.Resistance to movement through the landscape was based on the

distance travelled and the degree to which the landscape along amovement route was dominated by human uses as described by thehuman influence index (Sanderson et al. 2002). Rook-like move-ments from cell to cell were assigned a resistance of one and diago-nal movements were assigned a resistance of the square root of twoto account for the difference in the distance between cell centres.The distance-associated resistances were multiplied by the averagesquared human influence index value of the adjoining cells in ques-tion. We chose to use the squared values of the human influenceindex to ensure that human activities would affect movements. Weexplored alternative weightings for individual species (e.g. raisingthe index values to the 3rd or 6th power), which further con-strained movements and, in the aggregate, would result in moreconcentrated projected movement routes.For mapping movements in potential contraction zones (areas

that are climatically suitable today, but projected to become unsuit-able in the future), we used each individual cell in that zone as asource, injecting 1A of current into the cell. Areas predicted toremain climatically suitable were set to ground, so that currentwould flow from contraction zone cells to cells predicted to remainsuitable. All routes were constrained to areas that were historicallysuitable. We applied a similar procedure for modeling movementsinto expansion zones (areas projected to become climatically suit-able as climates change): each cell in the expansion zone was usedas a target, with a current source of -1A applied to draw currentinto the cell from areas that were suitable in the historical time per-iod.To map movement vectors within each cell, we designed a two-

step computation process. First, for each species, we computed agridded map of current flow directions and magnitudes by calculat-ing the vector sum of currents leaving each cell in each of eightpossible directions (Fig. 1). For each cell, we then calculated themean direction across all species, resulting in a gridded map of vec-tor directions describing a central tendency of species movement(Batschelet 1981). In addition to a directional mean, we also calcu-lated the circular concentration (directional agreement) across

species. Second, we added current vector magnitudes across specieswithin each cell. The result was gridded surfaces that described theresultant magnitude and direction of flow across species for eachcell location. To account for differences in species richness, wedivided the current flow magnitude in a cell by the number of spe-cies ranges that currently intersect the cell. We calculated thesethree statistics for all cells for each of the 10 future climate projec-tions and then produced one set of statistics by averaging theresults for the 10 projections.

RESULTS

Several locations in North and South America will likely be criticalfor species movements in a changing climate (Fig. 2). For example,large numbers of movements are expected relative to the numberof species currently at high northern latitudes, in the southeasternUnited States, northeastern North America, the Amazon Basin andsoutheastern Brazil. In the southeastern United States, species areprojected to track shifting climates by moving northeastward intothe southern Appalachian Mountains (Fig. 2a). In central Argentina,species are projected to move poleward into the southern Pampasas well as into the Sierras de C!ordoba and Andes (Fig. 2b). Severalother mountainous areas are also projected to facilitate movementsincluding the mountain ranges of Central America, the RockyMountains of western North America and the southern Brazilianhighlands (Fig. 3). Other landforms, such as the Great Lakes inNorth America and the Atacama Desert in Chile will likely con-strain movements as well (Fig. 2).To address the question of how human land use will likely shape

the movements of species in response to climate change, we com-pared our predictions of species movements that include land-usepatterns (Fig. 2) to similar predictions based solely on climaticchanges (Fig. S1). The differences in these two forecasts highlightareas to which species movements will be constrained due tohuman activities (blue areas in Fig. 4a) and conversely, areas inwhich human activities will likely impede climate-driven shifts inspecies distributions (red areas in Fig. 4a). These patterns emphasisethe effects of potential barriers to movement in the Amazon Basin,south-central Brazil, and the eastern Unites States. They also high-light areas that are likely to be movement corridors due to the rela-tive lack of human activity in the Amazon, along the borderbetween Bolivia and Paraguay, and through the Ozarks and Appala-chian Mountains in the eastern United States.Although the areas with many projected movements reflect the

constraints imposed by human land use, they are not necessarilypristine landscapes or even conducive to movement. We identifiedpotential areas for protecting movement and areas in which landuse will likely act as a barrier to movements by overlaying a map ofprojected species movements (from Fig. 2) with a map of the samemeasures of human impact used in our original analyses. The farnorthern latitudes, the Amazon Basin, and southeastern Bolivia andnorthern Paraguay (orange areas in Fig. 4b) are important for move-ment and less impacted by human activities. In contrast, large por-tions of the southeastern United States, southeastern Brazil and thePampas lowlands are projected to experience many movementsthrough highly impacted landscapes (the maroon-coloured areas inFig. 4b).As expected, the relative amount of climate-driven movement

and the general direction of movement varies regionally. When

© 2013 John Wiley & Sons Ltd/CNRS

Letter Climate-driven movements 3

expressed as a proportion of the current species richness, the Tun-dra, Boreal Forest and Taiga, North Temperate Broadleaf andMixed Forests, and the Tropical and Subtropical Moist BroadleafForests are projected to experience the greatest flux of speciesmovements compared to less-trafficked southern hemisphere Mon-tane Grasslands and Shrublands and Temperate Broadleaf andMixed Forests, as well as the Tropical and Subtropical ConiferousForests (Fig. 5a). When absolute numbers of movements are con-sidered, the patterns, at least in part, reflect patterns of vertebratespecies richness, with the tropical regions projected to experiencethe most movement (Fig. S2).We used simple linear models to test whether the number of move-

ments (adjusted for current species richness) and the variance in thedirection of movements were associated with elevation. In general,mountainous biomes (those with larger ranges of elevations) tend tohave fewer projected movements than biomes with less topographicalrelief (P < 0.01, r2 = 0.44, Fig. 5b). Our models also forecast greatervariance in the direction of movement in mountainous regions(P < 0.05, r2 = 0.42, Fig. 5c). The ecoregions with the greatest vari-ability in the directions that species are projected to move are in thetropics – the Tropical and Subtropical Dry Broadleaf Forests, south-ern hemisphere Deserts and Xeric Shrublands, and the Tropical andSubtropical Coniferous Forests. Indeed, in the tropics, species are notconsistently projected to move poleward as in the temperate regions

(where, as expected, species are projected to move north and southrespectively, Fig. 5a). Movements in the tropical broadleaf forests andthe montane grasslands tend to be westward, in part towards higherelevations in the Andes.Finally, we found that when analysed separately, the three taxo-

nomic groups that we modelled all showed similar patterns ofmovement across much of the western hemisphere (Fig. S3–S5).Nonetheless, the taxa differed with respect to movements in partic-ular areas. For example, movements into the Andes and across thesouthern portion of the Great Basin in the southwestern UnitedStates were particularly important for amphibians and there were farfewer movements projected in the Amazon Basin for mammals.

DISCUSSION

Projected movement

Our results depict projected climate-driven, land-use constrainedmovements of almost 3000 species across the western hemisphere.Several areas are likely to experience high concentrations of move-ments either because many species will experience changes in cli-matic suitability there or because land use will constrain theirmovements. There are potentially simple explanations for many ofthe specific patterns highlighted by the models in our study. Many

(a) (b)

(c) (d)

Figure 1 Illustration of movement modelling for the waxy monkey leaf frog (Phyllomedusa sauvagii). Areas projected to become climatically unsuitable (contraction zones)

and newly climatically suitable (expansion zones) by the end of the century based on a single climate change projection are represented by purple and green areas

respectively (a). Modelled movements out of contraction zones and into expansion zones are depicted by arrows (dots indicate no movements; b and d). Patterns of

human impacts (lighter areas have more human impacts as represented by squared human influence index values; c), alter the magnitude and mean direction of

movements driven solely by climate gradients (b) to produce movements that reflect the effects of both climate change and current land-use patterns (d).

© 2013 John Wiley & Sons Ltd/CNRS

4 J. J. Lawler et al. Letter

of the movements are to higher elevations or latitudes and manyskirt areas with intensive agriculture or development. The southernAppalachian Mountains in the southeastern United States and theAtlantic Forest in Brazil were two prominent areas highlighted byour models as likely to have high concentrations of species move-ments. Both of these areas are suspected to have served as climaterefugia in the past. For example, the central and northern AtlanticForest in southeastern Brazil have been identified as refugia inwhich these forest systems likely persisted through the Pleistocene(Carnaval & Moritz 2008). Whereas the projected movements intosoutheastern Brazil may imply that this area will again be a strong-hold for species in a changing climate, the movements in the south-ern Appalachians are northward, out of the areas that served asrefugia in the past.The seemingly large impacts of land use on projected movements

at high northern latitudes largely reflect that we have mapped thenumber of movements as a fraction of the current species richnessin a cell (Figs 2 and 4). The high latitudes have few species but areprojected to experience a large influx of species (Lawler et al. 2009)and thus the number of movements – and the difference between

the number of movements accounting for and ignoring human landuse – is high relative to the number of species in a cell. Patterns ofraw species movements are more reflective of hemispheric patternsof species richness (Figs. S2 and S6).The higher number of movements projected for flatter landscapes

relative to more mountainous regions likely reflects, in part, the pro-jected velocity of climate change (the instantaneous local rate ofmovement needed to maintain constant climatic conditions; Loarieet al. 2009). In mountainous areas where climate gradients are steep(e.g. coniferous forests and Montane Grasslands and Shrublands),climates will move more slowly across the landscape and thus spe-cies will need to move shorter distances to track suitable climates.In flatter regions (e.g. Tropical and Subtropical Broadleaf Forestsand the Northern Temperate Broadleaf and Mixed Forests) withhigher velocities of climate change, more species will be required tomove greater distances and thus we might expect to see moremovement in these areas. Similarly, the contrast between the moreconsistently projected latitudinal movements in ecoregions at mid tohigh latitudes and the less consistent directional movements of thespecies in the tropics likely reflect that elevational temperature gra-

(a)

(b)

Figure 2 Projected climate-driven species movements averaged across 10 future climate projections. Arrows represent the direction of modelled movements from

unsuitable climates to suitable climates via routes that avoid human land uses. The sizes of the arrows represent the number of projected species movements (current

flow) as a proportion of current species richness. The colours of the arrows reflect the level of agreement in the direction of movement across species and routes. The

insets are maps of (a) a high concentration of movements through southeastern North America and into the Appalachian Mountains and (b) areas of movement through

the Sierra De C!ordoba and into the Andes and the southern Pampas. Lighter blue shading in these topographical overlays indicates more intensive human activity.

© 2013 John Wiley & Sons Ltd/CNRS

Letter Climate-driven movements 5

dients are generally stronger than latitudinal temperature gradientsin the tropics and thus are more likely to drive species movementsin the future (Bush 2002).In general, our projected movement patterns have some expected

similarities with studies that have modelled connectivity based solely

on measures of human impact on the landscape (e.g. Theobald et al.2012). Theobald et al. (2012) identified several corridors in the south-ern and eastern United States and parts of corridors in the westernUnited States that correspond with the routes that we identified. Thedifferences between the maps produced by these two studies are

(c) (d)

(a) (b)

Figure 3 Projected climate-driven species movements averaged across 10 future climate projections for three specific areas (a). These are the same data presented in

Fig. 2. The squares on the map in panel (a) correspond to areas in Central America (b), the Rocky Mountains (c) and the southern Brazilian Highlands (d).

(a) (b)

Figure 4 Areas in which human land use will likely impede climate-driven movements (a, red areas) and where human land use will likely concentrate movements (a, blue

areas). Values are based on differences in the number of projected climate-driven faunal movements between models that include human land use and those that do not.

Some areas with many projected movements are highly impacted by human land use (b, dark purple areas) and some are much less impacted (b, light yellow areas). In

both maps, movements are calculated as the amount of current passing through a cell divided by the number of species’ current ranges that overlap the cell.

© 2013 John Wiley & Sons Ltd/CNRS

6 J. J. Lawler et al. Letter

largely due to the fact that our routes account for climate-drivenmovements. The results of our analysis of where climate-drivenmovement routes are likely to be more or less impacted by humanland uses (Fig. 4b) also correspond well with areas that have beenidentified as providing more or less connectivity for terrestrial mam-malian carnivores in general (Crooks et al. 2011).

A matter of scale

This study provides a continental-scale perspective of climate-drivenspecies movement routes. Our goal of intensively modelling move-ment for many species across a large extent required that we limitour analyses to a relatively coarse resolution to facilitate computa-tion. Nonetheless, there is ample opportunity for analytical innova-tions that could lead to more fine-grained analyses in the future.The spatial resolution of our analyses likely obscures finer scale pat-terns that might highlight other areas through which species will beable to move. The circuit theory-based algorithm we used took anaverage value of human impact as a measure of resistance to move-ment through each 50-km by 50-km grid cell. Thus, even cells withrelatively high average human impacts could have areas, and possi-bly contiguous areas, with low human impacts. Conversely, gridcells with low average human impacts could have substantial barri-ers to movement that would only be resolved at finer spatial scales.Given the coarse resolution of the data used in our analyses, the

areas we identified as being important for the movement of speciesin a changing climate are indeed very likely to be important formany species. However, they will not be the only important areasfor providing connectivity. Finer scale regional analyses will likelyreveal additional (and possibly alternative) routes where conserva-

tion efforts can be directed. Thus, although conservation invest-ments in the areas highlighted by our analyses will likely be useful,finer scale analyses will be important for revealing additional invest-ment options – choices that in some cases will be more cost-effec-tive. Moreover, the analyses presented here are most appropriatefor locating general areas in which connectivity could be preservedor enhanced to facilitate species movements in a changing climate.They could, for example, be used to identify where new, large-scaleprojects such as the Yellowstone to Yukon (Gailus 2001) effortcould be located or for directing resources to such ongoing efforts.However, given the limitations of their coarse resolution, they can-not be used to identify where in those general regions specificinvestments should be made. Finer scale analyses will be necessaryfor informing such decisions. Nonetheless, broad-scale analysessuch as those presented here are critical for addressing the impactsof climate change, which, in many cases, will be unconstrained byregional and national borders (Hannah 2010).

Bioclimatic and movement models

Previous studies have projected the potential rates and directions oflocal climatic changes (Loarie et al. 2009; Ackerly et al. 2010). Incontrast, the bioclimatic models used in our study integrate multipleclimatic variables to provide species-specific forecasts. In addition,our movement models, which account for coarse landscape pat-terns, provide a more biologically meaningful estimate of where spe-cies will likely need to move in response to climate change.Nonetheless, much like their application in past studies, these mod-els make several simplifying assumptions. First, bioclimatic modelslike those used in this study can provide an informed, but imperfect

Tund

ra

Trop

ical

& s

ubtro

pica

l

moi

st b

road

leaf

fore

sts

Bore

al fo

rest

s/ta

iga

Trop

ical

& s

ubtro

pica

l gra

ssla

nds,

sava

nnas

& s

hrub

land

s

Nor

th te

mpe

rate

bro

adle

af

& m

ixed

fore

sts

Sout

h te

mpe

rate

gra

ssla

nds

sava

nnas

& s

hrub

land

s

Trop

ical

& s

ubtro

pica

l dry

broa

dlea

f for

ests

Nor

th d

eser

ts &

xer

ic s

hrub

land

s

Nor

th te

mpe

rate

gra

ssla

nds

sava

nnas

& s

hrub

land

s

Tem

pera

te c

onife

r for

ests

Trop

ical

& s

ubtro

pica

l

coni

fero

us fo

rest

s

Sout

h de

serts

& x

eric

shr

ubla

nds

Sout

h m

onta

ne

gras

slan

ds &

shr

ubla

nds

Sout

h te

mpe

rate

bro

adle

af

& m

ixed

fore

sts

0

2

4

6

8

10

12

14

16

18

Mov

emen

t (cu

rren

t flo

w/s

peci

es)

N

N

N

NN N N

N NN

N

N

N

N

02468

1012141618

Mov

emen

t (cu

rren

t flow

/spe

cies

)

00.10.20.30.40.50.60.70.80.9

0500

10001500

20002500

Var

ianc

e in

dire

ctio

n

Elevation range (m)

0500

10001500

20002500

Elevation range (m)

(a) (b)

(c)

Figure 5 Average number and direction of climate-driven species movements per 50-km by 50-km grid cell across major biomes (a), the relationship between the average

number of movements and elevation range in each biome (b), and the relationship between the variance in movement directions across each biome and the range of

elevations in the biome (c). Movements are calculated as current flow through a cell divided by the number of species’ ranges that overlap the cell. Biomes are depicted

in Fig. S9.

© 2013 John Wiley & Sons Ltd/CNRS

Letter Climate-driven movements 7

approximation of how some species will move in response to cli-mate change (Ara!ujo et al. 2005) because they rely on empiricallyderived relationships between species distributions and climatic con-ditions and because they do not directly model biotic interactions –for example, predator–prey interactions or habitat associations – orevolutionary processes (Botkin et al. 2007). Second, our models donot account for dispersal distances and thus some of themovements included in our results may be unrealistic as they maybe in areas that a given species will not be able to reach in a 100-year period. Many moles, shrews and primates, for example, maynot be able to disperse fast enough to keep pace with climatechange in the Western Hemisphere (Schloss et al. 2012). Third, wehave used static models to investigate a dynamic process; moredynamic modelling could potentially reveal different patterns of spe-cies movements and barriers to those movements (Early & Sax2011). Fourth, our models use a static map of the human influenceindex. Given that human land use will continue to change, andpotentially have a greater influence than climate change on speciesin some regions (Jetz et al. 2007), using projected changes in thepatterns of human influence on the landscape would likely improvethe ability to project critical movement corridors and to anticipatewhere connections could be lost due to future development.Perhaps most importantly, our models do not account for spe-

cies-specific habitat requirements. Thus, some of the areas of pro-jected climatic suitability may not have suitable habitat. In addition,by not accounting for species’ habitat requirements, our modelsmake the simplifying assumption that all species respond to humanland use in the same way and that human impacts on the landscape(and large rivers which are incorporated into the human influenceindex we used to represent human impacts) are the only factors thatlimit movement. This clearly is not the case. Many birds, for exam-ple, will be able to fly over areas that will serve as barriers to manyamphibians and small mammals. Because we do not take species-specific habitat requirements, barriers to movement and dispersalbehaviours into account, our models likely overestimate the numberof movements through a given cell. Nonetheless, although it neces-sarily obscures the routes that will be needed to provide connectiv-ity for specific species, using a simple measure of the condition ofthe landscape has the potential to reflect species movement patternswhen summarising potential movements across a wide range of spe-cies with disparate habitat preferences (WHCWG 2010).Finally, our analyses gave equal weight to movements in areas of

potential range expansions and those in areas of potential rangecontractions. However, for many species, the critical movementswill be those into areas that become newly climatically suitable.Movements out of areas that become climatically unsuitable andinto already occupied areas will likely be less important for speciespersistence. When examined separately, many of the projectedmovements into areas of potential range expansions are similar tothe projected movements highlighted in Figs 2 and 3 (Fig. S7).Nonetheless, there are some significant differences between the pro-jected movements into potential expansion zones and those out ofpotential contraction zones (Figs. S7 and S8).

CONCLUSIONS

Fostering species movements by increasing connectivity is one ofthe most often-cited adaptation strategies for conserving biodiversityin a changing climate (Heller & Zavaleta 2009). Thus, identifying

where large numbers of species will need to move has the potentialto guide critical land use and conservation planning. To date, mostconnectivity planning has focused on connecting patches of suitablehabitat for one or more focal species. However, planning for con-nectivity that will allow species to track changes in climate willrequire new approaches that address multiple species, changinglandscapes and plan for movements outside of species’ currentdistributions (Phillips et al. 2008; Nu~nez et al. 2013). Broad-scaleconservation efforts such as the Yellowstone to Yukon and Meso-american Biocorridor projects have the potential to facilitate thetypes of movements projected by our models. However, our resultshighlight other areas in which there are no clear corridors andwhere restoration efforts and land-management practices could beused to either develop such corridors or to create more permeablelandscapes in general. What is perhaps most clear, is that many spe-cies will likely move in response to climate change, and humanactivities on the landscape have the potential to greatly facilitate orinhibit those movements.

ACKNOWLEDGEMENTS

We thank Peter Kareiva, Carlos Carroll and several anonymousreferees for comments that have greatly improved the manuscript.We acknowledge the Programme for Climate Model Diagnosisand Intercomparison (PCMDI) and the WCRP’s Working Groupon Coupled Modelling (WGCM) for their roles in making avail-able the WCRP CMIP3 multi-model data set. This work was sup-ported in part by the US Environmental Protection AgencyScience to Achieve Results (STAR) Programme (grant 833834),the US Fish and Wildlife Service North Pacific Landscape Conser-vation Cooperative and the Wilburforce Foundation. Data havebeen deposited in the Dryad Repository: http://dx.doi.org/10.5061/dryad.81537.

AUTHORSHIP

JL conceived of the study, AR and BM conducted the analyses, andJL, AR, JO and BM designed the analyses and wrote the manuscript.

REFERENCES

Ackerly, D.D., Loarie, S.R., Cornwell, W.K., Weiss, S.B., Hamilton, H.,

Branciforte, R., et al. (2010). The geography of climate change: implications

for conservation biogeography. Divers. Distrib., 16, 476–487.Ara!ujo, M.B., Pearson, R.G., Thuiller, W. & Erhard, M. (2005). Validation of

species–climate impact models under climate change. Glob. Change Biol., 11,

1504–1513.Ara!ujo, M.B., Thuiller, W. & Pearson, R.G. (2006). Climate warming and the

decline of amphibians and reptiles in Europe. J. Biogeogr., 33, 1712–1728.Batschelet, E. (1981). Circular Statistics in Ecology. Academic Press, New York, USA.

Botkin, D.B., Saxe, H., Ara!ujo, M.B., Betts, R., Richard, H.W.B., Cedhagen, T.,

et al. (2007). Forecasting the effects of global warming on biodiversity.

Bioscience, 57, 227–236.Bush, M.B. (2002). Distributional change and conservation on the Andean flank:

a palaeoecological perspective. Glob. Ecol. Biogeogr., 11, 463–473.Carnaval, A.C. & Moritz, C. (2008). Historical climate modelling predicts

patterns of current biodiversity in the Brazilian Atlantic forest. J. Biogeogr., 35,

1187–1201.Cheung, W.W.L., Lam, V.W.Y., Sarmiento, J.L., Kearney, K., Watson, R. &

Pauly, D. (2009). Projecting global marine biodiversity impacts under climate

change scenarios. Fish Fish., 10, 235–251.

© 2013 John Wiley & Sons Ltd/CNRS

8 J. J. Lawler et al. Letter

Crooks, K.R., Burdett, C.L., Theobald, D.M., Rondinini, C. & Boitani, L. (2011).

Global patterns of fragmentation and connectivity of mammalian carnivore

habitat. Philos. Trans. R. Soc. Lond. B Biol. Sci., 366, 2642–2651.Cutler, D.R., Edwards, T.C. Jr, Beard, K.H., Cutler, A., Hess, K.T., Gibson, J.,

et al. (2007). Random forests for classification in ecology. Ecology, 88, 2783–2792.

Early, R. & Sax, D.F. (2011). Analysis of climate paths reveals potential

limitations on species range shifts. Ecol. Lett., 14, 1125–1133.Environmental Systems Research Institute (2009). ArcGIS Desktop v.9.3.1.,

Redlands, CA.

Fielding, A.H. & Bell, J.F. (1997). A review of methods for the assessment of

prediction errors in conservation presence/absence models. Environ. Conserv.,

24, 38–49.Forman, R.T.T. & Alexander, L.E. (1998). Roads and their major ecological

effects. Annu. Rev. Ecol. Syst., 29, 207–C2.Gailus, J. (2001). Yellowstone to Yukon. Altern. J., 27, 36–39.Hannah, L. (2010). A global conservation system for climate-change adaptation.

Conserv. Biol., 24, 70–77.Heller, N.E. & Zavaleta, E.S. (2009). Biodiversity management in the face of

climate change: a review of 22 years of recommendations. Biol. Conserv., 142,

14–32.Huntley, B., Collingham, Y.C., Willis, S.G. & Green, R.E. (2008). Potential

impacts of climate change on European birds. PLoS ONE, 3, e1439.

Jetz, W., Wilcove, D.S. & Dobson, A.P. (2007). Projected impacts of climate and

land-use change on the global diversity of birds. PLoS Biol., 5, e157.

Kareiva, P., Watts, S., McDonald, R. & Boucher, T. (2007). Domesticated nature:

shaping landscapes and ecosystems for human welfare. Science, 316, 1866–1869.Lawler, J.J., White, D., Neilson, R.P. & Blaustein, A.R. (2006). Predicting

climate-induced range shifts: model differences and model reliability. Glob.

Change Biol., 12, 1568–1584.Lawler, J.J., Shafer, S.L., White, D., Kareiva, P., Maurer, E.P., Blaustein, A.R.,

et al. (2009). Projected climate-induced faunal change in the western

hemisphere. Ecology, 90, 588–597.Loarie, S.R., Duffy, P.B., Hamilton, H., Asner, G.P., Field, C.B. & Ackerly, D.D.

(2009). The velocity of climate change. Nature, 462, 1052–1055.McRae, B.H. & Shah, V.B. (2009). Circuitscape Connectivity Analysis Software. The

University of California, Santa Barbara, USA. Available at: http://www.

circuitscape.org. Last accessed 17 May 2013.

McRae, B.H., Dickson, B.G., Keitt, T.H. & Shah, V.B. (2008). Using circuit

theory to model connectivity in ecology, evolution, and conservation. Ecology,

89, 2712–2724.Millennium Ecosystem Assessment (2005). Ecosystems and Human Well-being

Biodiversity Synthesis. World Resources Institute, Washington, D.C.

Mitchell, T.D. & Jones, P.D. (2005). An improved method of constructing a

database of monthly climate observations and associated high-resolution grids.

Int. J. Climatol., 25, 693–712.Moritz, C., Patton, J.L., Conroy, C.J., Parra, J.L., White, G.C. & Beissinger, S.R.

(2008). Impact of a century of climate change of small-mammal communities

in Yosemite National Park, USA. Science, 322, 261–264.Nakicenovic, N., Alcamo, J., Davis, G., Vries, B.D., Fenhann, J. & Gaffin, S.,

et al. (2000). Special report on emissions scenarios. A Special Report of

Working Group III of the Intergovernmental Panel on Climate Change

Cambridge, UK. Available at: http://www.grida.no/publications/other/ipcc%

5Fsr/?src=/climate/ipcc/emission/index.htm.

New, M., Hulme, M. & Jones, P.D. (1999). Representing twentieth century

space-time climate variability. Part 1: development of a 1961–90 mean

monthly terrestrial climatology. J. Clim., 12, 829–856.New, M., Lister, D., Hulme, M. & Makin, I. (2002). A high-resolution data set

of surface climate over global land areas. Clim. Res., 21, 1–25.

Newman, M.E.J. (2005). A measure of betweenness centrality based on random

walks. Soc. Networks, 27, 39–54.Nu~nez, T.A., Lawler, J.J., McRae, B., Pierce, J., Krosby, M., Kavanagh, D., et al.

(2013). Connecting planning to address climate change. Conserv. Biol., 27, 407–416.

Olden, J.D., Lawler, J.J. & Poff, N.L. (2008). Machine learning without tears: a

primer for ecologists. Q. Rev. Biol., 83, 171–193.Parmesan, C., Ryrholm, N., Stefanescu, C., Hill, J.K., Thomas, C.D., Descimon,

H., et al. (1999). Poleward shifts in geographical ranges of butterfly species

associated with regional warming. Nature, 399, 579–583.Patterson, B.D., Ceballos, G., Sechrest, W., Tognelli, M.F., Brooks, T. & Luna,

L., et al. (2003). Digital Distribution Maps of the Mammals of the Western Hemisphere,

Version 1.0. NatureServe, Arlington, Virginia.

Phillips, S.J., Williams, P., Midgley, G. & Archer, A. (2008). Optimizing dispersal

corridors for the cape proteaceae using network flow. Ecol. Appl., 18, 1200–1211.

Pounds, A.J., Bustamante, M.R., Coloma, L.A., Consuegra, J.A., Fogden, M.P.L.,

Foster, P.N., et al. (2006). Widespread amphibian extinctions from epidemic

disease driven by global warming. Nature, 439, 161–167.Ridgely, R.S., Allnutt, T.F., Brooks, T., McNicol, D.K., Mehlman, D.W. &

Young, B.E., et al. (2003). Digital Distribution Maps of Birds of the Western

Hemisphere, Version 1.0. NatureServe, Arlington.

Sanderson, E.W., Jaiteh, M., Levy, M.A., Redford, K.H., Wannebo, A.V. &

Woolmer, G. (2002). The human footprint and the last of the wild. Bioscience,

52, 891–904.Schloss, C.A., Nu~nez, T.A. & Lawler, J.J. (2012). Dispersal will limit ability of

mammals to track climate change in the Western Hemisphere. Proc. Natl Acad.

Sci. USA, 109, 8606–8611.Sieving, K.E., Willson, M.F. & Santo, T.L.D. (1996). Habitat barriers to

movement of understory birds in fragmented south-temperate rainforest. Auk,

113, 944–949.Theobald, D.M., Reed, S.E., Fields, K. & Soul!e, M. (2012). Connecting natural

landscapes using a landscape permeability model to prioritize conservation

activities in the United States. Conserv. Lett., 5, 123–133.Thomas, C.D., Cameron, A., Green, R.E., Bakkenes, M., Beaumont, L.J.,

Collingham, Y.C., et al. (2004). Extinction risk from climate change. Nature,

427, 145–148.Thuiller, W., Lavorel, S., Ara!ujo, M.B., Sykes, M.T. & Prentice, I.C. (2005).

Climate change threats to plant diversity in Europe. Proc. Natl Acad. Sci. USA,

102, 8245–8250.V€or€osmarty, C.J., McIntyre, P.B., Gessner, M.O., Dudgeon, D., Prusevich, A.,

Green, P., et al. (2010). Global threats to human water security and river

biodiversity. Nature, 467, 555–561.WHCWG (2010). Washington Wildlife Habitat Connectivity Working Group Connected

Landscapes Project: Statewide Analysis. Washington Departments of Fish and

Wildlife, and Transportation, Olympia, WA, USA.

SUPPORTING INFORMATION

Additional Supporting Information may be downloaded via the onlineversion of this article at Wiley Online Library (www.ecologyletters.com).

Editor, N. HaddadManuscript received 28 November 2012First decision made 3 January 2013Manuscript accepted 28 April 2013

© 2013 John Wiley & Sons Ltd/CNRS

Letter Climate-driven movements 9