Technical ReportPDF Available

Conserving Nature’s Stage: Mapping Omnidirectional Connectivity for Resilient Terrestrial Landscapes in the Pacific Northwest

Authors:
CO N S E R V I N G N A T U R E S ST A G E :
MA P P I N G OM N I D I R E C T I O N A L CO N N E C T I V I T Y
F O R RE S I L I E N T TE R R E S T R I A L LA N D S C A P E S
IN T H E PA C I F I C NO R T H W E S T
PAC I FIC N O R THWES T A ND NOR T HERN C A L IFO R NIA
FIN A L REPOR T
TO THE DORIS DUKE CHARITABLE FOUNDATION
JUNE 2016
BRAD MCRAE
KEN POPPER
AARON JONES
MICHAEL SCHINDEL
STEV E BUTTRICK
KIMBERLY HALL
BOB UNNASCH
JIM PLATT
Cover Photo Credits:
Top photo- Oak savanna, Willamette Valley ©Rick McEwan
Second photo- Sandy River old growth ©Harold E. Malde
Third photo- Zumwalt Prairie ©Michael Durham
Bottom photo- Boardman Grassland ©Rick McEwan
Please cite as: McRae, B.H., K. Popper, A. Jones, M. Schindel, S. Buttrick, K. Hall, R.S. Unnasch,
and J. Platt. 2016. Conserving Nature’s Stage: Mapping Omnidirectional Connectivity for
Resilient Terrestrial Landscapes in the Pacific Northwest. The Nature Conservancy, Portland
Oregon. 47 pp. Available online at: http://nature.org/resilienceNW June 30, 2016.
Table of Contents
Acknowledgements ............................................................................................................................................. 3
Project History, Scope, and Setting ..................................................................................................................... 4
Introduction ......................................................................................................................................................... 6
Landscape connectivity and Conserving Nature’s Stage ............................................................................. 6
Mapping connectivity at multiple scales ..................................................................................................... 6
Modeling broad-scale connectivity: a new approach ......................................................................................... 8
The Omnidirectional Circuitscape (OmniScape) algorithm ....................................................................... 10
Resistance and source weight modeling ................................................................................................... 10
Moving window algorithm ........................................................................................................................ 12
Using blocks of target pixels as a computational shortcut ........................................................................ 14
Mapping current flow across the study area ............................................................................................ 14
Regional flow potential ............................................................................................................................. 19
Normalized current flow............................................................................................................................ 19
Pilot climate gradient analysis ................................................................................................................... 22
Discussion and Guidance for Use ...................................................................................................................... 24
Incorporating climate data: a pilot analysis .............................................................................................. 24
How to use these products ........................................................................................................................ 24
Caveats and potential enhancements ....................................................................................................... 28
Data Products .................................................................................................................................................... 31
Report, Appendices and Maps .................................................................................................................. 31
GIS data...................................................................................................................................................... 31
Scripts ........................................................................................................................................................ 31
Literature Cited .................................................................................................................................................. 32
Appendix A: Description of Resistance and Source Weight Modeling .............................................................. 36
Base land-cover data ................................................................................................................................. 36
Other data sources .................................................................................................................................... 37
Resistance scores ....................................................................................................................................... 38
Source weights .......................................................................................................................................... 39
Appendix B: Resistance and Source Weight scores ........................................................................................... 41
Appendix C: Further Detail on Moving Window Algorithm and Computational Shortcut ................................ 45
Scaling flow by target weight or by source and target weights ................................................................ 45
Computational shortcut ............................................................................................................................ 46
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
1
The Nature Conservancy- June 2016
Figure 1: 'Wall-to-wall' Circuitscape method………………………………………………………………………………………………9
Figure 2: Illustration of the moving window method……………………………………………………………………………………12
Figure 3: Illustration of the moving window method as applied in this study……………………………………………… 15
Figure 4: Summing individual moving window results to create a seamless current flow map…………………….16
Figure 5: Examples cases in which landscape configuration results in low current flow……………………………….18
Figure 6: Current flow patterns in three landscapes with differing landscape composition………………………….18
Figure 7: Examples of high normalized flow scores, where flow has been channeled by natural or
anthropogenic barriers………………………………………………………………………………………………………………………………..26
Figure 8: Areas with above-average resilience overlaid on connectivity results in the northern Columbia
Plateau……………………………………………………………………………………………………………………………………………………..…28
Figure C1: Illustration of the moving window method applied to a target block of pixels rather than a single
target pixel………………………………………………………………………………………………………………………………………………….46
Map 1: Study Area/Figure Index …….……………………………………………………………………………………………………………5
Map 2: Terrestrial Resistance……………………………………………………………………………………………………………………..11
Map 3: Source Weight…………………….………………………………………………………………………………………………………..13
Map 4: Current Flow……………………….………………………………………………………………………………………………………..17
Map 5: Regional Flow Potential…………………………………………………………………………………………………………………..20
Map 6: Regional Connectivity..……….…………………………………………………………………………………………………………..21
Map 7: Current Flow across Climate Gradients….………………………………………………………………………………………..23
Map 8: Regional Connectivity and Terrestrial Resilience Density….……………………………………………………………..27
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
2
The Nature Conservancy- June 2016
Acknowledgements
We thank the Doris Duke Charitable Foundation whose generous grants to The Nature
Conservancy made this work possible. We also thank the Wilburforce Foundation for supporting
development of the OmniScape methods and code, which were considerably improved through a
collaborative project with The Nature Conservancy’s California Chapter. We also thank those who
provided input on our modeling methods, feedback on this report, and counsel on applied conservation
needs including Mark Anderson, Dick Cameron, Carlos Carroll, Melissa Clark, Joe Fargione, Josh Lawler,
Caitlin Littlefield, Catherine Macdonald, Julia Michalak, Carrie Schloss, and David Theobald. Thanks to
Dani O’Brien for her formatting and production of the final documents.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
3
The Nature Conservancy- June 2016
Project History, Scope, and Setting
This report completes a larger project to identify and map sites that contribute to climate change
resilience in the Pacific Northwest, all funded by the Doris Duke Charitable Foundation. Previous
work (reported in Buttrick et al. 2015) focused on mapping sites likely to be resilient to climate
change based on local permeability and topoclimate diversity. Those sites that were more locally
intact and topoclimatically diverse were considered more resilient to climate change because they
would have higher potential to allow organisms to access climatically suitable areas by moving
short distances. The previous analyses purposefully considered the local scale, not looking beyond
a 3-km window when measuring terrestrial resilience characteristics. Results were stratified by
ecoregion and by geophysical setting (land facets”) to identify portions of land facets more likely
to be resilient to climate change.
The broad-scale landscape connectivity analysis reported here complements these previous
analyses by identifying areas likely to facilitate ecological flowparticularly movement, dispersal,
gene flow, and distributional range shifts for terrestrial plants and animalsover large distances
and long time periods. Similar to the local permeability analyses (Buttrick et al. 2015), this analysis
is not species-specific. Rather, it focuses on structural connectivity of natural lands, with resistance
to movement modeled as a function of landscape naturalness. This analysis shifts the focus to
identifying areas important for longer-distance movementsup to 50 kmcomplementing the
local permeability analyses which identified areas well-connected within a 3-km radius. This effort
does not incorporate projections of future climates, nor does it address connectivity for aquatic
species. The results identify broad, intact areas where movement of terrestrial organisms is largely
unrestricted by human modifications to the landscape, as well as constricted areas where
fragmentation has reduced movement options and further habitat loss could isolate remaining
natural lands. We provide guidance on how these results can be combined with the resilient sites
analyses of Buttrick et al. (2015), as well as other conservation priorities.
Our project area covers 97.3 million hectares (240.4 million acres) of the Pacific Northwest and
northern California. This includes 92 million hectares (227 million acres) analyzed in Buttrick et al.
(2015) namely, the California North Coast, Klamath Mountains, Sierra Nevada, West Cascades,
East Cascades/Modoc Plateau, Columbia Plateau, and Middle Rockies/Blue Mountains ecoregions
as well as the U.S. portion of the Pacific Northwest Coast, Willamette Valley/Puget Trough, North
Cascades and Canadian Rockies ecoregions. This connectivity study also encompasses an additional
5.3 million hectares (13.4 million acres) comprising the U.S extent of the Okanagan ecoregion and
the extent within Idaho of the Utah-Wyoming Rocky Mountains ecoregion (Map 1).
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
4
The Nature Conservancy- June 2016
C A N AD A
U N I T E D S T A T E S
NORTH
CASCAD ES
B R I T I S H C O L U M B I A
N E V A D A
C A L I F O R N I A
I D A H O
O R E G O N
M O N T A N A
W A S H I N G T O N
U T A H
W Y O M I N G
STUDY AREA /
FIGURE INDEX MAP
The Nature Conservanc y in Oregon, 2016Map produced by
0 50 100 150 200 Miles
0 50 100 150 20 0 Kilome ters
Map 1: Study Area / Figure Index
O
c
e
a
n
P
a
c
i
f
i
c
Fig.5 c
Fig.7 c
Fig.4
Fig.7 b
Fig.6 c
Figs .6 a, 7a , 8
Fig.5 b
Figs .3 ,
5a, 6b
The geographic extents of report figures are indicated in this map to help orient the reader. Generalized land use/land
cover data are also shown for reference. Total project area was 97.3 million hectares, including all or part of 13
ecoregions.
Lan d Us e
Developed Lands Non-forested
Forested
Agriculture
Open Water
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
5
The Nature Conservancy- June 2016
Introduction
Landscape connectivity and Conserving Nature’s Stage
The “Conserving Nature’s Stage” strategy focuses on identifying places that are likely to be good
conservation investments now and under future climate change. Maintaining landscape
connectivity is a key part of the Conserving Nature’s Stage approach (Anderson et al. 2012, 2014,
Beier et al. 2015). One of the most important ways species have responded to past climatic
changes has been to shift their ranges to track suitable climates (Jackson et al. 2000, Krosby et al.
2010, Blois et al. 2013, Moritz et al. 2013, Gill et al. 2015) Rapid warming projected for the next
century will likely require many species and populations to adapt in similar ways or face extinction
(Thuiller et al. 2005, Lawler et al. 2013). Many species are already moving in response to rapid
warming (Chen et al. 2011).
Increased high levels of anthropogenic habitat loss and fragmentation mean that many species will
likely encounter barriers that weren’t present during past periods of climate change (Warren et al.
2001, Thomas et al. 2010, Corlett and Westcott 2013, Gill et al. 2015). This, combined with rapid
climate change projected for the coming century, means that many species may not be able to
move quickly enough or far enough to keep up as suitable climates shift across the landscape
(Loarie et al. 2009, Schloss et al. 2012, Lawler et al. 2013). Moreover, maintaining gene flow and
genetic diversity through dispersal will be increasingly important for species adapting to climate
change in situ (Hoffmann & Sgrò 2011, Sexton et al. 2011, Sgrò et al. 2011). For these and other
reasons, conserving connectivity is the most recommended strategy for conserving biodiversity
under climate change (Heller and Zavaleta 2009).
Mapping connectivity at multiple scales
As with climate resilience analyses in the eastern USA (Anderson et al. 2012), we have identified a
need to map areas that contribute to the ability of species to adapt to climate change through both
local and long-distance movements. The previously completed local terrestrial permeability
analysis for this study area (Buttrick et al. 2015) quantified local connectedness of the immediate
neighborhood surrounding every pixel in the study area, measuring connectedness of that pixel to
its neighborhood out to a maximum distance of 3 km. By doing so, that analysis estimated the
ability of species to move short distances in order to find suitable habitats or microclimates under
climate change. The broad-scale connectivity analysis described in this report complements the
local connectedness analysis by modeling the potential for movements among natural lands
separated by distances up to 50 km. It estimates how flow patterns at this scale may become
diminished, redirected, or concentrated through certain areas due to the spatial arrangements of
cities, towns, farms, roads, open water, and natural land.
Our definition of connectivity (modified from Meiklejohn et al. 2010) is: the degree to which
regional landscapes, encompassing a variety of natural, semi-natural and developed land cover
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
6
The Nature Conservancy- June 2016
types, will sustain ecological processes and are conducive to the movement of many types of
organisms. Thus, we focus on areas that will be important for facilitating the local- and regional-
scale terrestrial ecological reorganization expected from climate change, involving many types of
organisms, over long time periods, among all types of natural and semi-natural habitats. Our
assumption is that maintaining a connected landscape, in conjunction with protecting and restoring
sufficient areas of high-quality habitat, will facilitate climate-induced range shifts and community
reorganization.
Species respond individualistically to climate change, and do not always move upslope or poleward
to cooler areas; instead, some have moved downslope in response to gradients in water availability
and other climate variables (Jackson and Overpeck 2000, Crimmins et al. 2011, Rapacciuolo et al.
2014, Gill et al. 2015). Thus our primary analysis represents a coarse-filter approach (Noss 1987)
that sought to quantify the existing structural connectivity of natural lands with no consideration of
predicted changes in temperature, precipitation, or other climate variables. Avoiding explicit
projections of how the climate will change was also in keeping with the Conserving Nature’s Stage
approach, which seeks to identify sites likely to contribute to climate resilience in a way that is
robust to uncertainties about how climate change will play out on the landscape (Anderson et al.
2012, 2014, Beier et al. 2015, Buttrick et al. 2015, Lawler et al. 2015). However, we also include a
pilot analysis that connects across climate gradients – from warm to cool areasto demonstrate
the flexibility of our methods and their applicability with additional climate data.
Both the local and broad-scale models are based on measures of human modification to the
landscape, with natural lands presenting the least resistance to movement, and developed lands
and human-created barriers such as highways causing the most resistance to movement. Although
both analyses focused on natural and semi-natural lands, we recognize that species respond
differently to anthropogenic land use, and that in fact there are species that thrive in heavily-
modified landscapes. Such species were not the target of this analysis. Connectivity for aquatic
species was also not addressed. Rather, this report identifies areas important for maintaining
connectivity for terrestrial species dependent on natural landscapes for movement, survival, and
reproduction.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
7
The Nature Conservancy- June 2016
Modeling broad-scale connectivity: a new
approach
We used Circuitscape (McRae et al. 2013a; http://www.circuitscape.org/) with a novel moving-
window analysis to quantify flow among all natural and semi-natural lands up to a distance of 50
km. Circuitscape models connectivity using electric circuit theory and leveraging mathematical
connections between circuit and random walk theories. It incorporates all possible pathways
between movement sources and destinations and identifies movement via low-resistance routes,
i.e., routes presenting relatively low movement difficulty and mortality risk. Circuitscape works by
treating landscapes as resistive surfaces, where high-quality movement habitat has low resistance
and barriers have high resistance. When two features on the landscape are to be connected,
electrical current flows from one (the source) to the other (the target). Patterns of current flow
through intervening areas help identify important routes for movement.
Previous applications have shown that three basic patterns can be seen in the products produced
by Circuitscape. Current flow will 1) avoid (be impeded by) areas with strong movement barriers, 2)
concentrate (intensify) in key linkages where flow accumulates or is channeled through pinch-points
(bottlenecks), and 3) spread out (diffuse) in highly intact areas with few barriers (Anderson et al.
2012). A primary use of Circuitscape has been to identify high-flow areas, particularly pinch-points,
where the loss of a small amount of movement habitat could disproportionately compromise
connectivity (e.g., Dickson et al. 2013).
Traditional applications of Circuitscape for conservation planning have typically focused on
connecting pairs of core areas or patches (e.g., Dickson et al. 2013, Brodie et al. 2015, Vasudev and
Fletcher 2015). This requires breaking the landscape into discrete core areas to be connected and
matrix lands between them.
Our development of the moving window approach was inspired by recent efforts that have used
Circuitscape to create ‘wall-to-wall’ connectivity maps, particularly Anderson et al. (2012, 2014),
Koen et al. (2014), and Pelletier et al. (2014). These methods modeled electrical current passing
through a given region as it flowed between sources and destinations placed in buffer areas
surrounding the region. For example, the wall-to-wall method employed by Anderson et al.
(2012, 2014) and Pelletier et al. (2014) used a tiling approach, in which a landscape is broken down
into square tiles surrounded by a buffer area, and current is passed across each tile from a source
beyond one edge of the tile to a ground beyond the opposite edge (Fig. 1). This is repeated for each
of the four cardinal directions and the four current maps are summed. Tiles are then reassembled
to create continuous, omnidirectional connectivity maps. The resulting mosaics highlight pinch-
points, where movement appears to be channeled through the landscape.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
8
The Nature Conservancy- June 2016
Figure 1. ‘Wall-to-wall’ Circuitscape
method, in which current typically flows
across each tile in each of the four cardinal
directions (in this case from West to East)
with the four results summed to produce a
single current map. Buffer areas are then
removed and tiles are stitched together.
Adapted from Anderson et al. (2012).
This approach represented a significant advance because it created seamless maps of broad-scale
connectivity without the need to divide the landscape into a binary representation of matrix lands
and core areas to be connected. This is important because arbitrary decisions about how core areas
are defined, e.g., minimum size requirements, can strongly influence connectivity modeling results
(Carroll et al. 2010, Koen et al. 2014, Pelletier et al. 2014). Identifying core areas to connect can be
desirable in some cases, e.g., when discrete animal populations are well mapped or when the goal
is to connect existing protected areas in a network (e.g., Brodie et al. 2015, Dutta et al. 2015).
Delineating areas to connect can be problematic, however, in studies such as ours where the goal is
to model connectivity for different kinds of processes across a large region; the approach can also
obscure important connectivity routes within core areas.
We built on the wall-to-wall methods in a way that retained the ability to map connectivity without
the need to delineate discrete core areas, but still allowed us to define what types of lands were
connected to one another, how strongly they were connected, and over what distances. In other
words, even with continuously-mapped landscape features we wanted to:
explicitly connect those features (e.g., natural areas) that represented important
conservation targets for those using our products;
adjust flow depending on site characteristics, e.g., allowing more flow to emanate
fromand travel tolands in more natural condition;
map flow only between areas close enough to one another to be connected for most
movement processes within realistic planning horizons and planning scales.
Thus our method built on previous efforts by adding the capacity to define which features were to
be connected while still maintaining a continuous, “core-free” connectivity modeling approach.
Being explicit about what is to be connected has several advantages. It recognizes that sources and
destinations for movement are essential to the concept of connectivity, and that some areas are
more important to connect than others. Moreover, the ability to specify what is being connected is
critical for any future efforts to use our method to model connectivity across climate gradients. For
example, applications that focus on identifying pathways that connect warm areas to cooler areas
(as in Nuñez et al. 2013, McGuire et al. 2016) or connect sites that have a specific set of climate
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
9
The Nature Conservancy- June 2016
conditions to sites projected to have those climate conditions in the future (as in Littlefield et al. in
review) require that source-target pairs be explicitly defined.
The Omnidirectional Circuitscape (OmniScape) algorithm
We developed an algorithm that modeled connectivity between natural and semi-natural areas
using a circular moving window. We prioritized connecting pixels representing natural lands,
reflecting our assumption that these areas were the most important to connect. We also
connected semi-natural areas, such as agricultural lands and urban open spaces, but adjusted the
flow originating from and arriving at such areas based on their level of human modification. In
other words, natural areas generated and received more flow than semi-natural areas, and these in
turn generated and received more flow than heavily-modified areas.
Following the work of Nuñez et al. (2013), which also sought to identify areas important for
connectivity under climate change, we chose to connect natural and semi-natural lands within 50
km of one another. This balanced the desire to examine broad-scale connectivity with
computational tractability, but also focused our analyses on movements that fall within realistic
conservation planning scales and time horizons.
The algorithm required first defining what types of land uses were to be connected, and assigning
resistances to different landscape features. It then used a moving window to connect all pixels
within the 50-km radius to one another using Circuitscape, summing up results from each moving
window into a cumulative current map. We describe the method and resulting maps in detail
below.
Resistance and source weight modeling
As in other applications of Circuitscape, the algorithm we developed represents a landscape as a
resistive surface. Landscape features conducive to movement are given low resistances, and
features that act as barriers to movement are given higher resistances.
We created a resistance raster surface at 180 m resolution using a process and input data similar to
those used for resistance modeling to support the local landscape permeability analyses reported
by Buttrick et al. (2015). This involved combining data on land use, roads, energy infrastructure,
housing density, and other features. Details about this process are in Appendix A. The resulting
resistance raster is shown in Map 2.
In addition to variability in resistance to movement, we also assumed that landscape features vary
in their importance for being connected. For example, natural areas may provide better habitat for
native species and therefore act as more important sources and destinations for movement.
Connecting such areas is likely to be of greater importance to conservation managers than
connecting non-natural areas. To represent the variation in importance for connecting different
areas, we created a surface of source weights which reflect the differing value we place on
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
10
The Nature Conservancy- June 2016
C A N AD A
U N I T E D S T A T E S
NORTH
CASCAD ES
W A S H I N G T O N
B R I T I S H C O L U M B I A
C A L I F O R N I A
U T A H
M O N T A N A
N E V A D A
O R E G O N
I D A H O
W Y O M I N G
TERRESTRIAL RESISTANCE
The Nature Conservanc y in Oregon, 2016Map produced by
0 50 100 150 200 Miles
0 50 100 150 20 0 Kilome ters
For a h igh-reso lution ver sion of t his map, see: http ://w ww.natu re.o rg/res ilienc eNW
Map 2: Terrestrial Resistance
O
c
e
a
n
P
a
c
i
f
i
c
Open wate r
This map shows resistance to movement, the first of two datasets which underpin our regional connectivity analysis.
We modeled resistance based on the degree of human modification to the landscape, with low resistance values
assigned to natural lands and high values assigned to developed lands, open water, and human-created barriers such
as highways.
Res ista nce Values
2-12 13-35 36-98 99-267 268-772 723-1,0001
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
11
The Nature Conservancy- June 2016
connecting different types of landscape features, with more natural pixels having higher source
weights.
We used the same input data layers and a similar procedure to create a source weight raster as we
did to create the resistance raster, in this case assigning greater source weights to land uses we
considered more likely to support natural populations now or in the future, resulting in more flow
to and from them. For example, areas consisting of entirely natural vegetation were given the
maximum source weight of 1, whereas semi-natural areas were given lower source weights. Open
water and completely developed areas were assigned a source weight of zero. The resulting
source-weight raster, also created at 180 m resolution, is shown in Map 3.
The resistance and source weight rasters formed the inputs for subsequent analyses, with higher
current flow occurring between areas with high source weights and along paths of low resistance.
More detail about our resistance and source-weight modeling can be found in Appendix A. A table
with resistance and source-weight scores for different land cover/land use classes is in Appendix B.
Figure 2. Illustration of the moving
window method. In the simplest case,
a moving window is passed over the
resistance and source weight rasters,
centering in turn on each pixel. If the
center pixel meets the naturalness
criteria for being a destination for
movement, it is treated as a target for
movement. All pixels within the moving
window radius that meet the same
criteria are considered sources. Current
flows from all source pixels to the
target pixel, with more current flowing
from more natural source pixels.
Moving window algorithm
Once resistance and source weights were mapped, our algorithm modeled connectivity between
pixels with non-zero source weights using a 50-km circular moving window. In the simplest
formulation of the algorithm, the moving window passes over the resistance and source weight
rasters described above, centering in turn on each pixel (Fig. 2). If the center pixel is not open water
or completely developed, it meets the criteria for being a destination for movement. If these
criteria are not met, i.e., the target pixel has a source weight of zero, the window moves on to the
next target pixel without performing any calculations. All pixels within the moving window radius
that meet these criteria are considered sources, but they can have different weights. Calculations
are performed only for the moving window area; prior to calculating flow, the algorithm masks out
all areas of the resistance layer that are outside of the 50-km circle. Masking increases
computational efficiency while also limiting movements to no greater than 50 km from
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
12
The Nature Conservancy- June 2016
C A N AD A
U N I T E D S T A T E S
NORTH
CASCAD ES
W A S H I N G T O N
B R I T I S H C O L U M B I A
C A L I F O R N I A
U T A H
M O N T A N A
N E V A D A
O R E G O N
I D A H O
W Y O M I N G
SOURCE W EIGHT
The Nature Conservanc y in Oregon, 2016Map produced by
0 50 100 150 200 Miles
0 50 100 150 20 0 Kilome ters
For a h igh-reso lution ver sion of t his map, see: http ://w ww.natu re.o rg/res ilienc eNW
Map 3: Source Weight
O
c
e
a
n
P
a
c
i
f
i
c
Open wate r
Sou rc e We ig ht
Low High
(Eq ua l inte rval )
This map depicts source weights, the second dataset underpinning our regional connectivity analysis. Source weights were
derived from similar input data layers and procedures as the resistance values, with greater source weights assigned to
pixels in more natural condition. We modeled connectivity between all pixels with non-zero source weights (i.e., excluding
only completely developed lands and open water), with higher current flow occurring between areas with high source
weights.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
13
The Nature Conservancy- June 2016
the target. Once the area of analysis is defined, the algorithm calls Circuitscape. Circuitscape injects
current from each source pixel, with more current flowing from more natural pixels. The target
pixel is set to ground, so that current flows across the subsetted landscape from sources to the
center target.
Note that each pixel that meets the criteria for being a source/target will be a source for many
moving window iterations (i.e., as many as there are sources within 50 km), and a target for one
iteration. Note also that in this model, barriers do not absorb or “kill” current; instead, they only re-
route current. Current will take the best route possible, “punching through” barriers if needed.
The result is a current map for each target pixel showing areas important for connecting the source
pixels within 50 km to the target pixel (Fig. 3). The moving window then shifts one pixel to the right,
centering on the next target pixel; if that pixel meets the naturalness criteria (i.e., it is not entirely
developed and is not open water), all other pixels meeting the criteria in the radius will be
connected to it, and so on. Current maps are summed across all moving windows to create a
cumulative current flow map among all sources and targets (Fig. 4).
Using blocks of target pixels as a computational shortcut
In practice, our method proved computationally prohibitive when calculations had to be repeated
with the moving window centered on each and every natural or semi-natural pixel in the study
region. To speed up processing, we employed a computational shortcut in which the moving
window centered on square blocks of pixels rather than on each individual pixel. Each computation
solved for all targets in a block; in this way, a single computation replaced many individual
computations, speeding up the algorithm without lowering the resolution of the resistance data.
More detail on this method can be found in Appendix C.
Mapping current flow across the study area
We used the above methods to produce a current flow surface for the study area, indicating where
concentrations of natural land and barriers interacted to produce differing patterns of flow (Map
4).
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
14
The Nature Conservancy- June 2016
Figure 3. Illustration of the
moving window method as
applied in this study. a) Top left
panel shows subset of resistance
layer, with a circular moving
window centered on a natural
pixel in Forest Park, Portland,
OR. Natural and semi-natural
lands have low resistance, and
human-modified lands have high
resistance. b) Top right panel
shows pixel source weights for
the same area, with natural
pixels (greens) having higher
weight. Pixels with source
weights > 0 (i.e., those that are
not in entirely developed or
open water classes) are treated
as sources, except for the center
pixel, which is treated as the
target. c) The bottom panel
shows resulting current flow
pattern when 1 Amp of current
is apportioned among all source
pixels in proportion to their
naturalness and allowed to flow
along low resistance routes to
the target (yellow representing
highest current flow).
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
15
The Nature Conservancy- June 2016
Figure 4. Summing individual moving window results
to create a seamless current flow map. Top two panels
show a) locations and b) results for two 50-km-radius
moving windows (centered on the Portland, OR area
and the less-human-modified area around Mount
Adams, WA). In both windows, current concentrates
toward the center of the window. But flow is less
constrained and thus more evenly spread
throughout the Mount Adams area. c) Bottom panel
shows the same subset of the study area, with summed
current flow from moving windows passed over the
entire study area. Flow is lower in heavily modified
areas like Portland because: 1) high resistance causes
flow to divert around them when other routes are
available; and 2) there are fewer natural areas to
connect within 50 km, and thus current sources and
targets are fewer and weaker.
In these raw current flow results, the patterns
typically produced by Circuitscape are evident,
with current avoiding areas with strong
movement barriers, concentrating where flow is
channeled through pinch-points, and diffusing in
highly intact/highly permeable areas. Large
urban centers often have low scores, both
because flow is diverted around these areas by
anthropogenic barriers and because naturalness
scores tend to be low, on average, within 50 km
of these centers (e.g., Portland; Fig. 5a). Flow
can also be low in large agricultural areas with
little natural land, e.g., the Palouse Prairie area
south of Spokane, WA (Fig. 5b), or areas along
the outer coast, even with relatively intact
landscapes (Fig. 5c).
Areas with highest current flow tend to be those
where natural or artificial barriers channel and
concentrate flow. This is particularly evident in
agricultural areas where linear stretches of
natural land form corridors conducive to
movement, e.g., in the northern portion of the
Columbia Plateau Ecoregion in Washington (Fig.
6a). Similarly, natural, linear features that are
surrounded by development form conduits,
concentrating flow (e.g., Forest Park in Fig. 6b).
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
16
The Nature Conservancy- June 2016
C A N AD A
U N I T E D S T A T E S
NORTH
CASCAD ES
B R I T I S H C O L U M B I A
N E V A D A
C A L I F O R N I A
O R E G O N
I D A H O
M O N T A N A
W A S H I N G T O N
U T A H
W Y O M I N G
CURRENT FLOW
The Nature Conservanc y in Oregon, 2016Map produced by
0 50 100 150 200 Miles
0 50 100 150 20 0 Kilome ters
Map 4: Current Flow
O
c
e
a
n
P
a
c
i
f
i
c
This map shows current flow for terrestrial connectivity among all natural and semi-natural pixels within 50 km of one
another, created using the OmniScape moving window algorithm. The algorithm is illustrated in Figs. 2-4. Current flow is
highest between areas with high source weights and along paths of low resistance.
For a h igh-reso lution ver sion of t his map, see: http ://w ww.natu re.o rg/res ilienc eNW
Open wate r
Curre nt F low
1-10 11-20 21-30 31-40 41-50 51-60 61-70 71-80 81-90 91-100
(Perc enti le )
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
17
The Nature Conservancy- June 2016
Flow is also channeled around natural barriers. For example, Lake Chelan separates highly natural
lands to the east and west (Fig. 6c). Moving windows that straddle but also include land to the
north of the lake produce flow between eastern and western sides of the lake via the northern tip
of the lake.
Figure 5. Example cases in which landscape configuration results in low current flow. Areas
that have relatively few natural lands to connect and many barriers to movement include: a)
urban areas such as Portland, OR; and b) intensive agricultural areas such as the Palouse Prairie.
c) Areas along the outer coast tend to have low flow because they are not situated between
large natural areas.
Figure 6. Current flow patterns in three landscapes with differing landscape composition. a)
Current is channeled along natural features (flood-scoured channels in scablands) surrounded by
agricultural lands in the Northern Columbia Plateau in eastern Washington. b) Current flow is all but
blocked in the Portland, OR area, and there are relatively few natural areas to connect in the area,
but some current concentrates in natural areas like Forest Park. c) Current flows around the
northern end of Lake Chelan, WA, because open water and developed areas to the south have high
resistance.
As with traditional Circuitscape results, diffuse flow in large, intact areas was harder to discern with
these results. The moving window approach helped to highlight intact areas somewhat, because
large natural areas have many sources and targets to connect. Still, local pinch-points tend to have
even higher flow, overshadowing the more intact portions of our study area. Moreover, the flow
through a given area is the product of two factors: the amount of natural land to connect within
the search radius, and the configuration of movement routes available between those natural
lands. The effects of these two factors can be difficult to distinguish within a single map of current
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
18
The Nature Conservancy- June 2016
flow, as low levels of flow can arise from several different mechanisms (e.g., spread of current
across larger areas, fewer natural areas to connect, impeded flow, and proximity to coasts).
Because conservation strategies would potentially differ among these different contexts, it is
important to try to distinguish their locations in the landscape. To help users do this, we produced
two additional maps, representing regional flow potential and normalized current flow,
respectively, designed to create more interpretable results. We describe these next.
Regional flow potential
To further distinguish intact (diffuse flow) areas from areas where flow is locally channeled, we
developed a map of regional flow potential. By this we mean, given the amount and configuration
of natural pixels available to connect within 50 km, how much flow would be expected in the
absence of barriers? We produced this map by running the same OmniScape analysis but setting all
resistances to the lowest score of 1. We used the same source weight raster to determine how
much current flowed to and from pixels. As a result of these modifications, areas with higher
current flow were located between larger expanses of natural land (i.e., areas that serve as sources
or destinations for moving organisms), and thus flow indicates their potential to connect natural
lands in the absence of barriers.
Map 5 shows results with all resistances set to the lowest score of 1. This map serves as a baseline,
or null model, against which we can compare flow patterns impeded or channeled by landscape
features. Pixels surrounded by highly natural areas, particularly those away from lakes and coasts,
have the most natural land to connect within 50 km, and thus the highest flow potential. Areas
where natural lands have been converted show lower flow potential, as do areas adjacent to large
water bodies, because there are fewer natural lands to connect via those areas.
Normalized current flow
We divided current flow by regional flow potential to produce a map of normalized current flow
(Map 6). This map helps to tease apart the mechanisms behind different flow rates, and better
distinguishes broadly natural areas with diffuse flow from areas where barriers are blocking flow or
channeling flow through pinch-points. If flow is lower than would be expected without barriers,
then barriers are blocking flow from the area. This is evident in urban centers, which have low
scores. If flow is higher than would be expected without barriers (i.e., current flow is high relative
to regional flow potential), then barriers are channeling flow into the area and potentially creating
pinch-points. These areas often show where the best movement options still exist in fragmented
landscapes, e.g., in the scablands in the northern Columbia Plateau (Fig. 7a). In areas where
barriers are having little effect on current flow patterns, current flow and regional flow potential
will be approximately equal (i.e., the ratio of actual to expected flow will be close to 1). These are
diffuse flow areas (Map 6).
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
19
The Nature Conservancy- June 2016
C A N AD A
U N I T E D S T A T E S
NORTH
CASCAD ES
W A S H I N G T O N
B R I T I S H C O L U M B I A
N E V A D A
C A L I F O R N I A
U T A H
M O N T A N A
O R E G O N
I D A H O
W Y O M I N G
REGIONAL FLOW POTENTIAL
The Nature Conservanc y in Oregon, 2016Map produced by
0 50 100 150 200 Miles
0 50 100 150 20 0 Kilome ters
Map 5: Regional Flow Potential
O
c
e
a
n
P
a
c
i
f
i
c
Given the amount and configuration of natural pixels available to connect within the analysis window, this map shows
how much flow would be expected in the absence of barriers, which we term regional flow potential. The dataset was
produced by running OmniScape with the original source-weight dataset and a null resistance surface with all cells
assigned the lowest resistance of 1. Pixels surrounded by highly natural areas have the most natural land to connect
within the search radius, and thus the highest flow potential.
Open wate r
Reg io nal Flow P oten ti al
Low High
(Eq ua l inte rval )
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
20
The Nature Conservancy- June 2016
C A N AD A
U N I T E D S T A T E S
NORTH
CASCAD ES
W A S H I N G T O N
B R I T I S H C O L U M B I A
C A L I F O R N I A
U T A H
M O N T A N A
N E V A D A
O R E G O N
I D A H O
W Y O M I N G
REGIONAL CONNECTIVITY
Normalized Current Flow
The Nature Conservanc y in Oregon, 2016Map produced by
0 50 100 150 200 Miles
0 50 100 150 20 0 Kilome ters
For a h igh-reso lution ver sion of t his map, see: http ://w ww.natu re.o rg/res ilienc eNW
Map 6: Regional Connectivity
O
c
e
a
n
P
a
c
i
f
i
c
Open wate r
N o r m a l i z e d
C ur r en t f l o w
R e g i on al F l o w Po t en ti al
Di ffus e
Im p ede d
Ch a nne l e d
In ten s i fi e d
HighestLowest
Integrating data from Maps 4 and 5 yields a fuller picture of regional connectivity patterns. Here, current flow (Map 4)
was normalized by dividing it by regional flow potential (Map 5). Upper and lower extremes of regional flow potential
are overlaid in order to highlight areas with the most and least natural land to connect, respectively. This helps to
distinguish 1) broad, intact areas where movement is diffuse or largely unrestricted, 2) channeled areas or pinch points
where further habitat loss could isolate natural areas, and 3) areas where flow is impeded by barriers.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
21
The Nature Conservancy- June 2016
Large, intact areas can be easily discerned using Maps 5 and 6. They have high regional flow
potential in Map 5 and fall in the diffuseclass in Map 6.
Pilot climate gradient analysis
Over longer timeframes, the degree to which landscapes are connected across climate gradients
will likely be a key factor in mediating range shifts (Hannah et al. 2014, Nuñez et al. 2013). To
demonstrate how our method can be adapted to explicitly incorporate climate, we conducted a
pilot analysis that combined our algorithm with present-day climate data connecting each natural
and semi-natural pixel to cooler pixels (if available) within 50 km (see Nuñez et al. 2013 for
rationale behind climate gradient connectivity analyses). Similar to Nuñez et al. (2013), we used the
30-year mean of mean annual temperature (MAT); in our case, we used means from 1961 to 1990,
available at 1 km2 resolution from AdaptWest (AdaptWest Project 2015). In our study region,
gradients in mean annual temperature are broadly correlated with those of more direct ecological
relevance, such as growing-degree days, average temperature of the coldest month, and moisture
deficit (Nuñez et al. 2013). We resampled our MAT layer to 180 m resolution, and connected pixels
that differed by 1 C and < 5 C. We consider this an experimental application of our methods to
demonstrate how they could be used with climate data; there are many data and parameter
decisions that we did not have time to explore, such as use of climate data other than MAT, the use
of more finely-downscaled climate data (as done by WHCWG 2013), and appropriate temperature
differences to connect.
Map 7 shows current flow when only pixels that differ by 1-5 C were connected. Compared with
Map 4, flow is diminished in areas with fewer options for moving to significantly cooler areas within
50 km, e.g., portions of the Columbia Plateau and the Oregon Coast range.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
22
The Nature Conservancy- June 2016
C A N AD A
U N I T E D S T A T E S
NORTH
CASCADES
B R I T I S H C O L U M B I A
N E V A D A
C A L I F O R N I A
I D A H O
O R E G O N
M O N T A N A
W A S H I N G T O N
U T A H
W Y O M I N G
The Nature Conservanc y in Oregon, 2016Map produced by
0 50 100 150 200 Miles
0 50 100 150 20 0 Kilome ters
Map 7: Cur r ent Flo w ac ro ss Climate Gradients
O
c
e
a
n
P
a
c
i
f
i
c
Op en w ater
Cur r ent Flo w
1-10 11-20 21-30 31-40 41-50 51-60 61-7071-80 81-9091-100
(Perc enti le )
CURRENT FLOW
CLIMATE GRADIENTS
acro ss
This map p ro vides an early lo o k at ho w inc o rp o rating c limate data c an c hange o mnidirec tio nal c o nnec tivity results.
Current flo w is used to p redic t mo vement r o utes fr o m w ar m to c o o l ar eas ac ro ss c limate gr adients. In this c ase,
mo vement w as allo w ed o nly betw een p ixels w here temp eratur es differ ed by 1-5 C. We exp ec t future OmniSc ap e
analyses w ill sup p o r t investigatio n o f ar eas likely to be imp o r tant fo r c o nnec ting ac ro ss c limate gradients, betw een
analo go us p resent-day and p r o jec ted futur e c limates, and/o r betw een p resent and p ro jec ted future sp ec ies ranges.
W A S H I N G T O NW A S H I N G T O N
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
23
The Nature Conservancy- June 2016
Discussion and Guidance for Use
As in previous applications of Circuitscape (e.g., Anderson et al. 2012, 2014, Koen et al. 2014,
Pelletier et al. 2014), our maps can help to identify areas where landscape features are likely to
block or constrain movement. Our methods allow us to explicitly define what is being connected,
allowing more flow among more natural areas, while still preserving the ability to model
connectivity in a continuous, core-free framework. This means that total current flow will be higher
in natural landscapes than in more human-modified landscapes, and more flow will be modeled
among large natural patches than among small ones.
Moreover, our ratio of flow to potential flow (Map 6) enhances the ability to highlight broadly
connected lands where flow is likely to be unconstrained. These areas of diffuse flow through intact
natural lands are typically difficult to distinguish in connectivity maps, but maintaining such areas
may often be the most cost-effective way to maintain functioning natural landscapes.
Incorporating climate data: a pilot analysis
Although our results incorporating climate data (Map 7) can give users an idea of which portions of
the study area could promote movement across significant climate gradients, we emphasize that
this effort was experimental and more work needs to be done to determine the effects of data
choices and decisions made parameterizing the model. For example, the temperature data we used
are at a coarse spatial scale. Had we used climate data that included finer-scale topoclimatic
variation, more temperature matches would have been found overall, including in areas with less-
steep temperature gradients (Gillingham et al. 2012). Moreover, although gradients in mean
annual temperature are correlated with biologically important gradients, different climatic
variables are likely limiting for different species in different portions of our study area (Wang and
Price 2007, Nuñez et al. 2013). Thus, this map should be considered a proof-of-concept rather than
a definitive map of important areas for movement across climate gradients. Still, the map
illustrates how explicitly modeling which areas would provide access to cooler climates might
change prioritizations.
How to use these products
Maintaining well-connected landscapes is important for many processes, including daily
movements, dispersal from natal areas, gene flow, recolonization of vacant habitat, and range
shifts under climate change. These processes are highly complex and vary among species and types
of movement. The areas important for conservation will depend on what is being connected, the
process that is being conserved, and the timeframe over which that process is expected to occur.
Thus it is impossible to create a single map that captures all areas important for maintaining all
types of connectivity. In this light, these products cannot be taken as simple maps of which areas
are most important or which strategies are appropriate in different parts of the study area. Instead,
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
24
The Nature Conservancy- June 2016
we recommend that users carefully examine the maps and data layers provided, including current
flow, regional flow potential, and normalized flow, as well as resistance and source weight maps.
These maps can best guide conservation strategies when the user identifies what is being
connected, understands how well the process to be conserved matches model assumptions, and
combines the results with other priorities.
We suggest that these products may best be used to create narratives describing the connectivity
value of different sites under consideration for conservation actions. For any area under
consideration, our three connectivity maps (current flow, potential flow, and normalized flow) can
be used in concert to help users consider not only the relative amount of ecological flow likely to be
in the landscape, but what is being connected by that flow natural lands, semi-natural lands, etc.
Additional data can be combined with these products to determine whether high-flow areas cross
climatic gradients, connect resilient lands, or connect habitat for particular species of concern.
In developing such narratives, we suggest users:
1) Give special consideration to intact areas with high amounts of natural lands to connect.
These areas have high regional flow potential scores and normalized current flow levels
that fall in the “diffuseclass across large areas. Examples include central Idaho, northern
Nevada, southeastern Oregon, the Cascades in Washington, and the Kalmiopsis and
Siskiyou areas of Oregon and California (Map 6).
2) Carefully evaluate areas with channeled flow; these could indicate pinch-points. High
normalized flow scores indicate areas where flow has been channeled by barriers, and as
such they often occur where landscapes are fragmented by water bodies or human
development. Some of the clearest examples in our study area include the northern
Columbia Plateau Ecoregion, where coulees and flood-scoured scablands form linear
connections across large agricultural areas (Fig. 7a; see also WHCWG 2012); the Kitsap
Peninsula, where a relatively small isthmus connects the peninsula to the mainland (Fig.
7b); and halosof high normalized flow where current skirts around agricultural areas,
such as the Snake River Plain in Idaho (Fig. 7c).
Pinch-points can indicate areas that are critical for maintaining connectivity — they may
be the last routes connecting natural lands and their loss may sever such connections.
But they must also be interpreted carefully. Because they are associated with
fragmentation, flow through them may also be crossing barriers or traveling large
distances to circumvent barriers. Recall that in our model, current may pierce a barrier if
it represents the best route possible.
Users should carefully evaluate the viability of connections through pinch-points, and
consider whether they are viable targets for protection (to maintain their existing
connectivity functions) or restoration (to provide alternate routes and alleviate
constrictions in flow). Restoration should also be considered where high flow crosses
restorable barriers.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
25
The Nature Conservancy- June 2016
Figure 7. Examples of high normalized flow scores, where flow has been channeled by natural
or anthropogenic barriers. a) The northern Columbia Plateau Ecoregion, where coulees and
flood-scoured scablands form linear connections across large agricultural areas. b) The Kitsap
Peninsula, where a relatively small isthmus connects the peninsula to the mainland. c) ‘Halos’ of
high normalized flow where current skirts around the Snake River Plain in Idaho.
3) Pay attention to intact coastal areas. Coastal areas typically show low amounts of flow
(e.g., Fig. 5c) and also score low on regional flow potential, because by definition they
don’t fall between large concentrations of natural lands. Centrality approaches such as
ours will discriminate against such areas, but for many reasons connecting coasts to
inland areas may still be an important conservation goal. The normalized flow map helps
identify coastal areas with high degrees of naturalness and normalized flow scores in the
diffuseclass, indicating opportunities for achieving this.
4) Compare results with local permeability analyses. Many of the same patterns in our
results can also be detected in the local permeability results (Map 7.2 in Buttrick et al.
2015); in particular, both will allow users to detect large, intact areas and highly
converted areas. But this broad-scale connectivity analysis emphasizes how areas
contribute to connectivity over larger distances, thus providing complementary
information. Areas of agreement between the two approaches should be given extra
consideration.
5) Compare results with terrestrial resilience data to identify resilient linkages and/or
linkages between resilient areas. Similar to other examples, areas where the two
analyses agree (i.e., a resilient area that also has good connectivity) should be given
special consideration (Map 8). Linkages between high-resilience areas may also be very
important, especially if alternative movement routes do not exist. One example of this is
in the Columbia Plateau region of eastern Washington (Fig. 8), where flood-scoured
scablands connect resilient areas, with these linkages identified in the regional
connectivity data as intensified or channeled. These linkages also connect areas differing
in mean annual temperature (Map 7), and protection or restoration could provide
multiple conservation benefits.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
26
The Nature Conservancy- June 2016
C A N AD A
U N I T E D S T A T E S
NORTH
CASCAD ES
B R I T I S H C O L U M B I A
N E V A D A
C A L I F O R N I A
I D A H O
O R E G O N
M O N T A N A
W A S H I N G T O N
U T A H
W Y O M I N G
The Nature Conservanc y in Oregon, 2016Map produced by
0 50 100 150 200 Miles
0 50 100 150 20 0 Kilome ters
For a h igh-reso lution ver sion of t his map, see: http ://w ww.natu re.o rg/res ilienc eNW
Map 8: Regional Connectivity and Terrestrial Resilience Density
O
c
e
a
n
P
a
c
i
f
i
c
N o r m a l i z e d
C ur r en t f l o w
R e g i on al F l o w Po t en ti al
Di ffus e
Im p ede d
Ch a nne l e d
In ten s i fi e d
HighestLowest
Open wate r
Hig h resilien ce d ensity
This map overlays results of this analysis with resilience data from Buttrick et al. (2015). Areas with above-average
terrestrial resilience are shown on top of the regional connectivity results from Map 6. This provides one example of how
connectivity data can be combined with other priorities, in this case showing linkages that may be important for
connecting highly resilient areas.
REGIONAL CONNECTIVITY
RESILIENCE DENSITY
and
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
27
The Nature Conservancy- June 2016
Figure 8. Areas with above-average resilience overlaid on
connectivity results in the northern Columbia Plateau.
Combining our results with previous analyses or other
prioritizations can help inform conservation decisions. In this
case, linkages that connect highly resilient sites are identified.
6) Consider other sources of connectivity information. Independent analyses of connectivity
(e.g., WHCWG 2010, 2013, Krosby et al. 2014, Littlefield et al. in review), movement data,
or landscape genetic data can complement these analyses, particularly to support species-
specific conservation efforts. Users should place more confidence in areas where different
modeling efforts agree, and more weight should be given to conservation actions
supported by different analytical approaches, data sources, or conservation goals (e.g.,
restoration of riparian habitat in an area identified by multiple analyses as important for
connectivity or climate resilience).
7) Consider other priorities. Similar to the rest of the Conserving Nature’s Stage
methodology and datasets (www.nature.org/resilienceNW), it is possible to use these
data in concert with almost any other prioritization dataset. Identifying which portions of
existing protected areas are most at risk of losing connectivity, which unprotected areas
may be most important to connect, where the most connected areas are for a potential
species reintroduction programs, or which areas are most important to prioritize for
improved management because they demonstrate high connectivity value are but a few
of the myriad of ways these data could be combined with other conservation priorities.
Caveats and potential enhancements
As with many connectivity analyses (e.g., WHCWG 2010, 2012), modeled routes may pass over
barriers and these routes must be evaluated for viability. As described earlier in this report, current
in our model will take the best route possible to connect all sources and targets within 50 km of
one another, and will “bore through” barriers if alternative, low-resistance routes are unavailable.
Similarly, current will flow distances greater than 50 kmas long as they remain within the circular
windowto reach targets. Thus, careful interpretation and on-the-ground validation of movement
routes should be conducted before conservation actions are taken.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
28
The Nature Conservancy- June 2016
Both the resistance and source-weight modeling processes relied on expert opinion, involving many
subjective decisions, and we emphasize that no single resistance or naturalness scoring scheme will
be ideal for all individual species of conservation concern. Rather, our model focuses on keeping
natural lands structurally connected to one another via the most natural movement routes,
focusing on connecting the “stages” upon which a diverse set of species are most likely to be found
now and into the future (Buttrick et al. 2015). Users should carefully examine our resistance and
naturalness score maps to determine the degree to which the maps are compatible with their
conservation goals. We further note that we considered large water bodies to be significant
barriers, did not consider effects of steep terrain (including cliffs), and connected pixels among
dramatically different vegetation types and biomes. These are all modeling decisions that could
have been made differently depending on goals and assumptions.
An example of how our models are sensitive to resistance scores and other parameters is the
previously mentioned zone of high flow to the north of Lake Chelan (Fig. 6c). This zone exists
because we are connecting large blocks of natural land on opposite sides of the lake. Moving
windows completely encircle the lake, and current tends to flow around the northern tip of the
lake from sources on one side to the other because open water has high resistance and movement
around the southern end of the lake is hindered by roads and the town of Chelan and roads.
Moreover, flow farther north is somewhat impeded by Ross Lake and US Highway 20 (Fig. 6c). Thus,
flow between eastern and western sides of the lake is somewhat constrained and concentrated at
the northern tip. This pattern would be diminished had we limited the total linear distance that
current could travel, or had we used a lower resistance for open water (because more current
would have flowed across the lake rather than around it).
Note that channeled”/”intensified”/”diffuse”/”impeded” designations based on our normalized
current scores are scaled based on how much natural land is available to connect, and this varies
considerably across our study area. As such, the results shown in Map 6 must be viewed in context
and in conjunction with those shown in Maps 4 and 5. For example, roads within channeled areas
may have normalized current scores close to 1 (placing them in the diffuseclass), simply because
flow is locally avoiding them and thus their flow scores are more in line with expectations based on
regional potential.
Our moving window method is experimental, but is promising in several respects. First, it readily
identifies where large concentrations of intact natural lands exist, and where organisms occupying
natural lands could move a user-specified distance to reach other natural areas. Second, it
produces continuous maps and does not require identifying discrete patches of natural lands to
connect. Third, the moving window method is flexible, with potential for many additional
enhancements.
Current flow could be scaled such that flow is lower from sources that are more distant
from targets. For example, within a 50-km window, sources 5 km from the target would
produce more flow to the target than sources 40 km from the target.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
29
The Nature Conservancy- June 2016
For species-specific applications, source weights could reflect habitat suitability rather than
naturalness, resulting in more flow modeled between large, high-quality habitat blocks than
between smaller or lower-quality blocks.
In specific climate connectivity applications, the method could be used to connect across
climate gradients (as in our example in Map 7 and in an early application of our code by
Anderson et al. 2015), or present-day climate data and future climate projections could be
used to connect pixels to targets that have analogous climates under future projections (as
in Littlefield et al. in review).
Connectivity along riparian and freshwater habitats could be modeled by limiting analyses
to only those habitat types. Alternatively, flow through riparian areas and valley bottoms
could be extracted from our results and examined separately to identify riparian areas that
contribute highly to regional connectivity.
Flow from sources could be scaled with cost distance or effective resistance, so that current
would be diminished between sources and targets separated by strong barriers. In the
present analysis, total current leaving a source was not affected by movement difficulty, but
simply flowed along the best route possible, even if that route crossed strong barriers.
Restoration opportunities could be evaluated in two ways. Voltage maps (see McRae et al.
2008) could be produced by the algorithm and used to identify restoration opportunities as
suggested in McRae et al. (2012); such applications require further exploration. More
simply, the model could be rerun with a set of proposed restoration projects burned into
the resistance layer to evaluate how connectivity values could change following restoration.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
30
The Nature Conservancy- June 2016
Data Products
Data, maps, and computer code created by this project are included in a small set of files available
for download from this Conservation Gateway site http://nature.org/resilienceNW along with any
updates to this report.
Report, Appendices and Maps
Two files are available which include:
1. The main report and written appendices.
2. High-resolution (600 dpi) versions of the report maps.
GIS data
GIS data created for the project, including resistance, source weight, and currentflow maps, are
available at www.nature.org/resilienceNW.
Scripts
Computer code created for the project, including OmniScape and land cover pre-processing scripts,
are available upon request. Please see www.nature.org/resilienceNW for contact information.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
31
The Nature Conservancy- June 2016
Literature Cited
AdaptWest Project. 2015. Gridded current and projected climate data for North America at 1-km
resolution, interpolated using the ClimateNA v5.10. Available at adaptwest.databasin.org.
Anderson, M. G., A. Barnett, M. Clark, C. Ferree, A. Olivero Sheldon, and J. Prince. 2014a. Resilient
Sites for Terrestrial Conservation in the Southeast Region. The Nature Conservancy, Boston,
MA.
Anderson, M. G., M. Clark, and B. H. McRae. 2015a. Permeable Landscapes for Climate Change. The
Nature Conservancy, Boston, MA.
Anderson, M. G., M. Clark, and A. O. Sheldon. 2012. Resilient Sites for Terrestrial Conservation in
the Northeast and Mid-Atlantic Region. The Nature Conservancy, Eastern Conservation
Science. The Nature Conservancy, Boston, MA.
Anderson, M. G., M. Clark, and A. O. Sheldon. 2014b. Estimating climate resilience for conservation
across geophysical settings. Conservation Biology 28:959970.
Anderson, M. G., P. J. Comer, P. Beier, J. J. Lawler, C. A. Schloss, S. Buttrick, C. M. Albano, and D. P.
Faith. 2015b. Case studies of conservation plans that incorporate geodiversity. Conservation
Biology 29:680691.
Beier, P., M. Hunter, and M. Anderson. 2015. Special section: Conserving nature’s stage.
Conservation Biology 29:613617.
Blois, J. L., P. L. Zarnetske, M. C. Fitzpatrick, and S. Finnegan. 2013. Climate change and the past,
present, and future of biotic interactions. Science 341:499504.
Brodie, J. F., A. J. Giordano, B. Dickson, M. Hebblewhite, H. Bernard, J. Mohd-Azlan, J. Anderson,
and L. Ambu. 2015. Evaluating multispecies landscape connectivity in a threatened tropical
mammal community. Conservation Biology 29:122132.
Buttrick, S., K. Popper, M. Schindel, B. H. McRae, B. Unnasch, A. Jones, and J. Platt. 2015.
Conserving Nature’s Stage: Identifying Resilient Terrestrial Landscapes in the Pacific
Northwest. The Nature Conservancy, Portland, Oregon.
Chen, I.-C., J. K. Hill, R. Ohlemüller, D. B. Roy, and C. D. Thomas. 2011. Rapid range shifts of species
associated with high levels of climate warming. Science 333:10241026.
Commission for Environmental Cooperation (CEC). 2013. 2010 North American Land Cover at 250
m spatial resolution. Produced by Natural Resources Canada/Canadian Center for Remote
Sensing (NRCan/CCRS), United States Geological Survey (USGS); Insituto Nacional de
Estadística y Geografía (INEGI), Comisión Nacional para el Conocimiento y Uso de la
Biodiversidad (CONABIO) and Comisión Nacional Forestal (CONAFOR).
Compton, B. W., K. McGarigal, S. A. Cushman, and L. R. Gamble. 2007. A Resistant-Kernel Model of
Connectivity for Amphibians that Breed in Vernal Pools. Conservation Biology, 21:788-799.
Corlett, R. T., and D. A. Westcott. 2013. Will plant movements keep up with climate change? Trends
in Ecology and Evolution 28:482488.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
32
The Nature Conservancy- June 2016
Crimmins, S. M., S. Z. Dobrowski, J. A. Greenberg, J. T. Abatzoglou, and A. R. Mynsberge. 2011.
Changes in climatic water balance drive downhill shifts in plant species’ optimum elevations.
Science 331:324–327.
Dickson, B. G., G. W. Roemer, B. H. McRae, and J. M. Rundall. 2013. Models of regional habitat
quality and connectivity for pumas (Puma concolor) in the Southwestern United States. PLoS
ONE 8.
Dutta, T., S. Sharma, B. H. McRae, P. Roy, and R. DeFries. 2015. Connecting the dots: mapping
habitat connectivity for tigers in central India. Regional Environmental Conservation:115.
Gill, J. L., J. L. Blois, B. Benito, S. Dobrowski, M. L. Hunter, and J. L. Mcguire. 2015. A 2.5-million-year
perspective on coarse-filter strategies for conserving nature’s stage. Conservation Biology
29:640648.
Gillingham, P. K., Palmer, S. C., Huntley, B., Kunin, W. E., Chipperfield, J. D. and Thomas, C. D. 2012.
The relative importance of climate and habitat in determining the distributions of species at
different spatial scales: a case study with ground beetles in Great Britain. Ecography 35:831-
838.
Hannah, L., L. Flint, A. D. Syphard, M. A. Moritz, L. B. Buckley, and I. M. McCullough. 2014. Fine-
grain modeling of species’ response to climate change: Holdouts, stepping-stones, and
microrefugia. Trends in Ecology and Evolutionrends in Ecology and Evolution 29:390397.
Heller, N. E., and E. S. Zavaleta. 2009. Biodiversity management in the face of climate change: A
review of 22 years of recommendations. Biological Conservation 142:1432.
Hoffmann, A., and C. Sgrò. 2011. Climate change and evolutionary adaptation. Nature 470:479
485.
Jackson, S. T., and J. T. Overpeck. 2000. Responses of plant populations and communities to
environmental changes of the late Quaternary. Paleobiology 26:194220.
Jin, S., Yang, L., Danielson, P., Homer, C., Fry, J., and Xian, G. 2013. A comprehensive change
detection method for updating the National Land Cover Database to circa 2011. Remote
Sensing of Environment, 132: 159-175.
Krosby, M. B., R. Norheim, D. M. Theobald, and B. McRae. 2014. Riparian Climate-Corridors:
Identifying Priority Areas for Conservation in a Changing Climate. Climate Impacts Group,
University of Washington, Seattle, WA.
Krosby, M., J. Tewksbury, N. M. Haddad, and J. Hoekstra. 2010. Ecological connectivity for a
changing climate.
Lawler, J. J., D. D. Ackerly, C. M. Albano, M. G. Anderson, S. Z. Dobrowski, J. L. Gill, N. E. Heller, R. L.
Pressey, E. W. Sanderson, and S. B. Weiss. 2015. The theory behind, and the challenges of,
conserving nature’s stage in a time of rapid change. Conservation Biology 29:618629.
Lawler, J. J., A. S. Ruesch, J. D. Olden, and B. H. McRae. 2013. Projected climate-driven faunal
movement routes. Ecology Letters 16:10141022.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
33
The Nature Conservancy- June 2016
Littlefield, C. E., B.H. McRae, J. Michalak, J. J. Lawler, and C. Carroll. (In review). Missed connections:
Tracking climates through time and space to predict connectivity under climate change.
Conservation Biology.
Loarie, S. R., P. B. Duffy, H. Hamilton, G. P. Asner, C. B. Field, and D. D. Ackerly. 2009. The velocity of
climate change. Nature 462:10521055.
McGuire, J. L., J. J. Lawler, B. H. McRae, T. Nuñez, and D. M. Theobald. 2016. Achieving climate
connectivity in a fragmented landscape. Proceedings of the National Academy of Sciences of
the United States of America.
McRae, B. H., B. G. Dickson, T. H. Keitt, and V. B. Shah. 2008. Using circuit theory to model
connectivity in ecology, evolution, and conservation. Ecology 89:27122724.
McRae, B. H., S. A. Hall, P. Beier, and D. M. Theobald. 2012. Where to restore ecological
connectivity? Detecting barriers and quantifying restoration benefits. PLoS ONE 7.
McRae, B. H., V. B. Shah, and T. K. Mohapatra. 2013a. Circuitscape 4 User Guide. The Nature
Conservancy, Fort Collins, CO.
McRae, B.H., A.J. Shirk, and J.T. Platt. 2013b. Gnarly Landscape Utilities: Resistance and Habitat
Calculator User Guide. The Nature Conservancy, Fort Collins, CO. Available at:
http://www.circuitscape.org/gnarly-landscape-utilities.
Meiklejohn, K., R. Ament, and G. Tabor. 2009. Habitat Corridors & Landscape Connectivity:
Clarifying the Terminology. Center for Large Landscape Conservation, Bozeman, MT.
Moritz, C., and R. Agudo. 2013. The future of species under climate change: resilience or decline?
Science 341:504–508.
Noss, R. F. 1987. From plant communities to landscapes in conservation inventories: A look at the
nature conservancy (USA). Biological Conservation 41:1137.
Nuñez, T. A., J. J. Lawler, B. H. McRae, D. J. Pierce, M. B. Krosby, D. M. Kavanagh, P. H. Singleton,
and J. J. Tewksbury. 2013. Connectivity planning to address climate change. Conservation
Biology 27:407416.
Rapacciuolo, G., S. P. Maher, A. C. Schneider, T. T. Hammond, M. D. Jabis, R. E. Walsh, K. J. Iknayan,
G. K. Walden, M. F. Oldfather, D. D. Ackerly, and S. R. Beissinger. 2014. Beyond a warming
fingerprint: Individualistic biogeographic responses to heterogeneous climate change in
California. Global Change Biology 20:28412855.
Schloss, C. a., T. a. Nunez, and J. J. Lawler. 2012. Dispersal will limit ability of mammals to track
climate change in the Western Hemisphere. Proceedings of the National Academy of Sciences
2012:16.
Sexton, J. P., S. Y. Strauss, and K. J. Rice. 2011. Gene flow increases fitness at the warm edge of a
species’ range. Proceedings of the National Academy of Sciences 108:1170411709.
Sgrò, C. M., A. J. Lowe, and A. A. Hoffmann. 2011. Building evolutionary resilience for conserving
biodiversity under climate change. Evolutionary Applications 4:326337.
Thomas, C. D. 2010. Climate, climate change and range boundaries. Diversity and Distributions
16:488495.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
34
The Nature Conservancy- June 2016
Thuiller, W., S. Lavorel, M. B. Araujo, M. T. Sykes, and I. C. Prentice. 2005. Climate change threats to
plant diversity in Europe. Proceedings of the National Academy of Sciences 102:82458250.
Vasudev, D., and R. J. Fletcher. 2015. Incorporating movement behavior into conservation
prioritization in fragmented landscapes: An example of western hoolock gibbons in Garo Hills,
India. Biological Conservation 181:124132.
Wang, A., and D. T. Price. 2007. Estimating global distribution of boreal, temperate, and tropical
tree plant functional types using clustering techniques. Journal of Geophysical Research-
Biogeosciences 112.
Warren, M. S., J. K. Hill, J. A. Thomas, J. Asher, R. Fox, B. Huntley, D. B. Roy, M. G. Telfer, S.
Jeffcoate, P. Harding, G. Jeffcoate, S. G. Willis, J. N. N. Greatorex-Davies, D. Moss, and C. D.
Thomas. 2001. Rapid responses of British butterflies to opposing forces of climate and habitat
change. Nature 414:6569.
Washington Wildlife Habitat Connectivity Working Group (WHCWG). 2010. Washington Connected
Landscapes Project: Statewide Analysis. Washington Departments of Fish and Wildlife and
Transportation, Olympia, WA.
Washington Wildlife Habitat Connectivity Working Group (WHCWG). 2012. Washington Connected
Landscapes Project: Analysis of the Columbia Plateau Ecoregion. Washington Departments of
Fish and Wildlife and Department of Transportation, Olympia, WA.
Washington Wildlife Habitat Connectivity Working Group (WHCWG). 2013. Washington Connected
Landscapes Project: Columbia Plateau Climate-Gradient Corridors Analysis. Washington
Departments of Fish and Wildlife and Department of Transportation, Olympia, WA.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
35
The Nature Conservancy- June 2016
Appendix A: Description of Resistance and
Source Weight Modeling
Several datasets and processing steps were used in the creation of the resistance and source
weight layers. Although most of these datasets had been used in our earlier local permeability
analysis (Buttrick et al. 2015), the search larger radius required to model regional connectivity (50
km vs 3 km) necessitated refreshing all datasets to increase their extent into a buffer region around
our study area to prevent edge effects.
The same general processing workflow was also followed, with some enhancements, from the 2015
local permeability analysis. A wall-to-wall land-cover map was created as the basis of the resistance
and source weight surfaces. These base land-cover layer were then augmented with finer-scaled
data, such as data on electrical transmission lines and roads, from ancillary datasets. This approach
allowed us to incorporate the best local data for many of the important features that can affect
terrestrial regional connectivity.
Base land-cover data
To mitigate for edge effects, the base land-cover data needed to extend well beyond our project
footprint. Within the U.S., the 2011 National Land Cover Dataset (NLCD; Jin et. al. 2013) covered
the entire project area including the buffer. However, the buffer region north of the international
boundary was not represented. To fill this gap, we used the Canadian portion of the North
American Land Cover Dataset (Commission for Environmental Cooperation 2013). These data were
resampled from their native 250-m pixel size to match the 30-m pixel size of the NLCD, then
reclassified to crosswalk to the land-cover class values in NLCD. These two datasets were then
merged as the basis for subsequent processing.
To maximize the benefits of incorporating ancillary data it was first necessary to remove the
vestiges of those that appeared in the land-cover map, especially roads. NLCD, for example,
inconsistently represented roads as various developed types, and missed many altogether. These
road artifacts were problematic for two reasons. First, existing road fragments were assigned to
various development classes, with differing resistance scores, which would erroneously affect
resistance values. Second, any misalignment between datasets would allow double-counting of
roads in cases where multiple parallel road features appeared in the final data layer.
Similarly, bridges over water bodies often appear in land-cover data, complicating our efforts to
develop a resistance model that included information on distance from shore. Without removing
bridges from land-cover data, water pixels adjacent to bridges could not be distinguished from
water pixels adjacent to shore.
To remedy these issues, as well as to reclassify water into distance bands to represent increasing
difficulty many terrestrial species would experience when crossing wider bodies of water, we
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
36
The Nature Conservancy- June 2016
developed a python script. The first task the script performed was to reclassify water bodies in the
NLCD/NALC layer to include information on distance from shore. This necessitated first removing
bridges, so that water pixels adjacent to bridges would not be assigned to low distance classes. We
used a modification of methods used to remove roads by the Washington Wildlife Habitat
Connectivity Working Group (WHCWG 2013). This was accomplished with the ArcGIS Shrink
command, which we used to contract all developed areas by 2 pixels, replacing those cells with
their nearest neighbor values. This resulted in a land-cover map with bridges removed. The script
then removed barrenand herbaceous wetlandclasses (which in many cases corresponded to
tidal flats or sand bars), and calculated Euclidean distances from the nearest remaining non-water
classes. Water pixels in the original NLCD/NALC layer were then reclassified to reflect these
Euclidean distances, creating a new water raster with classes representing several distance bands.
Note that this meant that water pixels immediately adjacent to urban pixels were assigned to
higher distance classes because urban areas were shrunk by two pixels. This allowed us to later
penalize movements through pixels immediately adjacent to urban shorelines, which would have
provided unrealistically conducive movement routes skirting developed shorelines.
The second task the script performed was to remove roads from the original NLCD/NALC layer,
using both the ArcGIS Shrink and Expand commands. As with removing bridges, we removed roads
using the Shrink operation, but in this case only contracting by one pixel and then re-expanding
remaining developed pixels back into their pre-contracted locations up to a distance of one pixel. In
other words, if there were developed pixels that were shrunk but immediately adjacent to
developed pixels that remained after the Shrink operation, the pre-shrunken pixels were re-
assigned their original developed class. This process eliminated developed features less than two
pixels wide but retained them otherwise. Thus, we were able to remove roads without losing
developed pixels at the edges of urban or other developed areas. Shrunken developed pixels were
assigned new codes identifying their original and replacement cover types (e.g., 21041 for pixels
originally in class 21, Developed, open space,that had been contracted and replaced by class 41,
Deciduous forest). These unique codes allowed us to assign appropriate resistance values to each
combination of pre- and post-process land-cover types.
Other data sources
To represent features that were not captured in our base land-cover data, we incorporated several
additional datasets. These data represented features that were not well represented in the
NLCD/NALC dataset, such as roads, energy infrastructure, and low-density housing. Each of these
vector datasets was converted to a 30m raster and snapped to the base land-cover grid.
Roads were represented by multiple datasets. TIGER roads data (U.S. Census Bureau 2013) were
available for the entire USA, and these were used to identify freeways and highways, as well as
low-use roads. However, we found that low-use roads were incomplete in this dataset. This is
problematic because in rural, non-agricultural areas away from highways and buildings, low-use
roads are often the only readily mapped features that indicate human land uses. We therefore
included roads data from the USDA Forest Service, the Bureau of Land Management, the California
Timber Harvest Program, the State of Idaho, and the Canadian National Road network. In each
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
37
The Nature Conservancy- June 2016
case, we dropped the relatively small portion of road segments that were clearly identified as
closed or decommissioned. No one dataset was comprehensive for any state, so we used all
datasets that were of reasonable quality in combination.
Railroads were captured from the U.S. Census Bureau's 2013 TIGER database. Railroads include
main, spur, and yard rail lines; carline, streetcar track, monorail, and mass transit lines; and cog rail
line, incline rail, and tram lines. No rail lines were represented in the Canadian buffer zone.
Energy infrastructure, including all significant transmission lines, wind towers and natural gas
pipelines within the U.S. portion of the project extent, were represented by the EV Energy Map
layer (Ventyx 2015). Transmission lines in these data are grouped into voltage classes, so each
could be given unique resistance weights reflecting the differing footprints on the landscape from
different transmission capacities. Energy infrastructure was not represented in the Canadian
portions of the buffer zone.
The 2010 Population and Housing Unit Counts Report, produced by the U.S. Census Bureau, were
used in conjunction with census tract polygons (clipped to private lands) to calculate 8 classes of
Block Housing Density(BHD) on private lands across the study area. These data, obtained from
David Theobald (Conservation Science Partners), were included to represent the non-specific
impacts associated with increasing human densities. As described in our 2015 report, housing
densities are a good surrogate for a number of anthropogenic impacts, such as noise, predation by
pets and non-native landscaping that reduce connectivity potential and which don’t appear in
standard land-cover classifications.
Resistance scores
We developed expert-based resistance scores for all classes in each input layer, representing the
estimated resistance to movement created by each landscape feature (Appendix B). Resistance
values for the 2015 local permeability analyses were based on accumulated cost-weighted
distances (Compton et al. 2007). However, circuit-theoretic analyses are based on probabilities of a
random walker moving into a pixel, so resistance values developed for one framework are not
necessarily appropriate for the other. More work needs to be done to determine best practices for
assigning resistance scores to different features; however, based on previous experience with
broad-scale connectivity analyses (e.g., WHCWG 2010, 2013) and previous experience with
developing resistance scores for Circuitscape, we developed a steeper scoring scheme to create
more differentiation between permeable and impermeable land cover types. We began with scores
that were roughly the square of values used in the previous permeability analyses, adjusting scores
as needed to achieve values deemed appropriate for a circuit-theoretic framework. In the end,
entirely “natural” pixels were assigned a resistance of 1, and anthropogenic or natural barriers
were assigned resistances up to a maximum of 1000.
In most cases, highly resistant features were either only represented in a single dataset, or
overlapped in ways that one dataset would “eclipse” the other (e.g., urban classes from NLCD and
housing density classes from BHD). However, our road data were compiled from many sources with
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
38
The Nature Conservancy- June 2016
varying degrees of overlap. Because our resistance values are derived by taking the maximum,
rather than the sum, of resistances from overlapping layers, having the same road represented in
more than one layer would yield the same resistance values as long as the pixels aligned. However,
misalignment between these source data often caused the same road to be represented in slightly
different positions in the various datasets, typically running parallel within 30m, but sometimes
farther apart.
To mitigate the effects of these misalignments, we maintained the target resistance of 9 for low-
use roads, but spread the resistance across a larger area (90 m instead of 30 m). We accomplished
this by reducing low-use roads resistance from 9 to 3, but assigned the value of 3 to roads that had
been expanded (widened) by one 30-m pixel in each direction (resulting in a total resistance of 9
for crossing a road). This expansion meant that a single road represented as two side-by-side
features from different road data layers typically had a large proportion of overlapping pixels from
the two datasets. The extra resistance from parallel features was thus reduced except for a
minority of cases where misalignment was greater than 60 m.
We were not concerned about these issues for roads in cities and towns, as urban features in the
housing density and NLCD layers represent higher resistance values than low-use roads, and thus
took precedence in those areas. Major roads and highways were still derived from a single dataset
(TIGER), and assigned high resistances. Shrunken developed features in the NLCD/NALC were
assumed to be low-use roads and were typically assigned either the resistance of low-use roads or
the class that replaced the shrunken pixels, whichever was higher.
All scores were recorded in a Microsoft Excel spreadsheet, which referenced the layer, class, and
resistance score. These data were input into the Resistance and Habitat Calculator of Gnarly
Landscape Utilities (McRae et al. 2013b, http://www.circuitscape.org/gnarly-landscape-utilities).
Road features were ‘fattened’ using the ‘expand cells’ setting (see details on low-use road
treatment above). The resulting raster represented the maximum resistance across all input layers
at a 30 m pixel size. We then aggregated to 180 m taking the mean value of all 30-m pixels within
each 180-m pixel to produce our final resistance surface.
Source weights
As described in the main report, our approach required that each pixel be given a source weight,
representing the weight that would be given to connections to and from that pixel (i.e., the amount
of flow to and from the pixel). We gave greater source weights to land uses we considered to be
more likely to support natural populations now or in the future, resulting in more current flowing
to and from them. Pixels consisting of entirely natural vegetation were given the highest weight of
1, and entirely developed pixels were given a weight of zero. Semi-natural pixels that had some
likelihood of supporting native species now or in the future were given intermediate values; for
example, the NLCD pasture/hayclass was given a source weight of 0.5, and croplands were given
a value of 0.2. Unvegetated but still natural pixels were treated somewhat differently. Perennial
snow/icewas given a value of 0.75, reflecting the assumption that these areas are inhospitable to
most species now, but may become snow and ice free under climate change and thus become
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
39
The Nature Conservancy- June 2016
targets for range shifts. Barren landsin NLCD were also given a value of 0.75, since these often
represent tidal areas or sand bars and in some cases are barren due to human land uses such as
mining. Open waterpixels were not considered sources or targets for movement, and were
assigned a source weight of 0.
As with resistance modeling, shrunken developed features in NLCD/NALC data were assumed to be
low-use roads. These were typically assigned either the source weight of low-use roads or the class
that replaced the shrunken pixels, whichever was lower.
These data were input into a modified version of the Resistance and Habitat Calculator of Gnarly
Landscape Utilities, producing a source raster reflecting the minimum source weight across all
input layers at a 30m pixel size. The modification allowed us to fatten road features while taking
the minimum value across inputs. We then aggregated to 180 m taking the mean value of all 30-m
pixels within each 180-m pixel to produce our final source weight surface.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
40
The Nature Conservancy- June 2016
Appendix B: Resistance and Source Weight
scores
Table B1. Source weight and resistance scores assigned to different classes in input layers.
Class ID
Class Description
Source
wt
Resistance
Expand
Cells
11
Water
0
4
0
12
Perennial Ice/Snow
0.75
2
0
21
Developed, Open Space
0.3
16
0
22
Developed, Low Intensity
0.2
81
0
23
Developed, Medium Intensity
0.1
400
0
24
Developed, High Intensity
0
1000
0
31
Barren Land
0.75
2
0
41
Deciduous Forest
1
1
0
42
Evergreen Forest
1
1
0
43
Mixed Forest
1
1
0
52
Shrub/Scrub
1
1
0
71
Grassland/Herbaceous
1
1
0
81
Pasture/Hay
0.5
16
0
82
Cultivated Crops
0.2
49
0
90
Woody Wetlands
1
1
0
95
Emergent Herbaceous Wetlands
1
1
0
21011
Shrunken from 21 to Open Water
0
4
0
21012
Shrunken from 21 to Perennial Ice/Snow
0.75
3
0
21031
Shrunken from 21 to Barren Land
0.75
3
0
21041
Shrunken from 21 to Deciduous Forest
0.75
3
0
21042
Shrunken from 21 to Evergreen Forest
0.75
3
0
21043
Shrunken from 21 to Mixed Forest
0.75
3
0
21052
Shrunken from 21 to Shrub/Scrub
0.75
3
0
21071
Shrunken from 21 to Grassland/Herbaceous
0.75
3
0
21081
Shrunken from 21 to Pasture/Hay
0.5
16
0
21082
Shrunken from 21 to Cultivated Crops
0.2
49
0
21090
Shrunken from 21 to Woody Wetlands
0.75
3
0
21095
Shrunken from 21 to Emergent Herbaceous
0.75
3
0
22011
Shrunken from 22 to Open Water
0
4
0
22012
Shrunken from 22 to Perennial Ice/Snow
0.75
3
0
22031
Shrunken from 22 to Barren Land
0.75
3
0
22041
Shrunken from 22 to Deciduous Forest
0.75
3
0
22042
Shrunken from 22 to Evergreen Forest
0.75
3
0
22043
Shrunken from 22 to Mixed Forest
0.75
3
0
22052
Shrunken from 22 to Shrub/Scrub
0.75
3
0
22071
Shrunken from 22 to Grassland/Herbaceous
0.75
3
0
22081
Shrunken from 22 to Pasture/Hay
0.5
16
0
22082
Shrunken from 22 to Cultivated Crops
0.2
49
0
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
41
The Nature Conservancy- June 2016
22090
Shrunken from 22 to Woody Wetlands
0.75
3
0
22095
Shrunken from 22 to Emergent Herbaceous
0.75
3
0
23011
Shrunken from 23 to Open Water
0
4
0
23012
Shrunken from 23 to Perennial Ice/Snow
0.75
3
0
23031
Shrunken from 23 to Barren Land
0.75
3
0
23041
Shrunken from 23 to Deciduous Forest
0.75
3
0
23042
Shrunken from 23 to Evergreen Forest
0.75
3
0
23043
Shrunken from 23 to Mixed Forest
0.75
3
0
23052
Shrunken from 23 to Shrub/Scrub
0.75
3
0
23071
Shrunken from 23 to Grassland/Herbaceous
0.75
3
0
23081
Shrunken from 23 to Pasture/Hay
0.5
16
0
23082
Shrunken from 23 to Cultivated Crops
0.2
49
0
23090
Shrunken from 23 to Woody Wetlands
0.75
3
0
23095
Shrunken from 23 to Emergent Herbaceous
0.75
3
0
24011
Shrunken from 24 to Open Water
0
4
0
24031
Shrunken from 24 to Barren Land
0.75
3
0
24041
Shrunken from 24 to Deciduous Forest
0.75
3
0
24042
Shrunken from 24 to Evergreen Forest
0.75
3
0
24043
Shrunken from 24 to Mixed Forest
0.75
3
0
24052
Shrunken from 24 to Shrub/Scrub
0.75
3
0
24071
Shrunken from 24 to Grassland/Herbaceous
0.75
3
0
24081
Shrunken from 24 to Pasture/Hay
0.5
16
0
24082
Shrunken from 24 to Cultivated Crops
0.2
49
0
24090
Shrunken from 24 to Woody Wetlands
0.75
3
0
24095
Shrunken from 24 to Emergent Herbaceous
0.75
3
0
0
Gap 1, 2 or 3 lands
1
1
0
1
Undeveloped
1
1
0
2
Residential - rural low (0.001-.0.006 dua)
0.75
1.4
0
3
Residential - rural (0.006-0.025 dua)
0.5
2.3
0
4
Residential - exurban low (0.025-0.1 dua)
0.3
6.3
0
5
Residential - exurban (0.1-0.4 dua)
0.2
16
0
6
Residential - low (0.4-1.6 dua)
0.1
49
0
7
Residential - med (1.6-10 dua)
0
256
0
8
Residential - high (>10 dua)
0
400
0
1100
Interstate
0
400
1
1200
State and local highways, major secondary
0
100
1
1400
City and rural streets
0.75
3
1
1500
Unpaved and AWD
0.75
3
1
1630
Highway interchange ramp
0
400
1
1640
Service Drive
0.75
3
1
1710
Walkway/Pedestrian Trail
0.75
3
1
1720
Stairway
0.75
3
1
1730
Alley
0.75
3
1
1740
Private Road for service vehicles
0.75
3
1
1750
Internal U.S. Census Bureau use
0.75
3
1
1780
Parking Lot Road
0.75
3
1
1820
Bike Path or Trail
0.75
3
1
1830
Bridle Path
0.75
3
1
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
42
The Nature Conservancy- June 2016
3
Road
0.75
3
1
1
Road
0.75
3
1
0
Trail
0.75
3
1
1
Road
0.75
3
1
0
Obliterated or decommisioned road
0.9
2
1
1
Closed road
0.9
2
1
2
Road
0.75
3
1
0
Water
1
1
1
1
Rails to trails
0.75
3
1
2
Road
0.75
3
1
1
Road
0.75
3
1
1
Road
0.75
3
1
3
Road
0.75
3
1
1
Road
0.75
3
1
2
Converted, decommissioned, planned
0.75
2
1
1
Road
0.75
3
1
2
Proposed or abandoned road
0.9
2
1
10
Highway
0
400
1
12
Primary highway
0
400
1
13
Secondary highway
0
400
1
20
Road
0.75
3
1
21
Arterial
0.1
9
1
22
Collector
0.2
9
1
23
Local
0.75
3
1
24
Alley/Lane/Utility
0.75
3
1
25
Connector/Ramp
0.2
9
1
26
Reserve/Trail
0.75
3
1
29
Strata (housing developments)
0.2
9
1
80
Bridge/Tunnel
0.75
3
1
90
Unknown (mostly logging roads, etc)
0.75
3
1
0
Railroad- active
0
25
0
1
Wind tower
0
100
0
90
Inner wind tower buffer (< 90m)
0.5
10
0
180
Outer wind tower buffer (90-180m)
0.75
5
0
102
Transmission line - 100-161 Volts
0.75
9
0
103
Transmission line - 230-300 Volts
0.5
16
0
104
Transmission line - 345 Volts
0.4
25
0
105
Transmission line - 500 Volts
0.4
25
0
106
Transmission line - DC Line
0.4
25
0
107
Transmission line - Step-Up
0.75
9
0
108
Transmission line - Under 100 V
0.75
9
0
1
Natural Gas Pipelines
0.75
9
0
1030
Open Water 0 to 30 m from shore
0
4
0
1060
Open Water 30 to 60 m from shore
0
33
0
1090
Open Water 60 to 90 m from shore
0
66
0
1120
Open Water 90 to 120 m from shore
0
100
0
1150
Open Water 120 to 150 m from shore
0
133
0
1180
Open Water 150 to 180 m from shore
0
166
0
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
43
The Nature Conservancy- June 2016
1210
Open Water 180 to 210 m from shore
0
200
0
1240
Open Water 210 to 240 m from shore
0
233
0
1270
Open Water 240 to 270 m from shore
0
266
0
1300
Open Water 270 to 300 m from shore
0
300
0
1330
Open Water 300 to 330 m from shore
0
333
0
1360
Open Water 330 to 360 m from shore
0
366
0
1390
Open Water 360 to 390 m from shore
0
400
0
1420
Open Water 390 to 420 m from shore
0
433
0
1450
Open Water 420 to 450 m from shore
0
466
0
1480
Open Water 450 to 480 m from shore
0
500
0
2000
Open Water > 480 m from shore
0
500
0
Notes: Table adapted from Excel worksheets used to calculate source weight
and resistance rasters using Gnarly Landscape Utilities. Expand cells column
indicates whether a class was expanded and by how many pixels.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
44
The Nature Conservancy- June 2016
Appendix C: Further Detail on Moving Window
Algorithm and Computational Shortcut
In this appendix we further describe the moving window algorithm, particularly the use of the
computational shortcut to speed processing.
Scaling flow by target weight or by source and target weights
The moving window algorithm can be run in two ways: either more current can flow to each
target when there are more source pixels, or a fixed amount of current can be apportioned
among all sources. In the former case, the flow in a landscape (after results from all moving
windows are added up) will scale with the square of the amount of natural land. Double the
natural land, and you quadruple the flow (and the inferred importance of keeping it
connected). In the latter case, flow scales linearly with the amount of natural land (double the
amount of natural land, and flow doubles).
The two cases produce similar maps, but the former case (flow scales with the square of natural
land) emphasized intact landscapes at the expense of coastal areas and landscapes with any
appreciable degree of human use. We felt the latter case, where each target accepts a fixed
amount of current regardless of the number of sources, produced more useful maps of
connectivity for our primary analyses. The approach still clearly identified intact landscapes
(Map 6), but better highlighted more subtle patterns of connectivity in coastal areas and in
working landscapes, where there were still valuable natural lands to connect. In other words,
scaling flow by natural land rather than the square of natural land meant that the signal from
intact landscapes did not overwhelm that of partially developed and coastal landscapes.
By contrast, we used the former case (more current flows to each target when there are more
source pixels) for the pilot climate gradient analysis. This was simpler conceptually, because we
could scale flow by the number of temperature matches (with a match being defined by a
source cell connecting to a target cell that was at least 1 C cooler but no more than 5 C cooler).
However, this is experimental and an argument could still be made for scaling flow by the
weight of the target pixels. One could think of this weight as a ‘carrying capacity,’ with each
target pixel having a fixed capacity to receive immigrants. Parameterization decisions such as
these must be more fully explored.
Note that because we used different approaches (case 2 and case 1 above, respectively) for our
primary and pilot climate analyses, maps 4 and 7 are not directly comparable. They can still be
compared qualitatively however.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
45
The Nature Conservancy- June 2016
Computational shortcut
Our moving window method required prohibitive amounts of processing time when the
window centered on every pixel in the study area. Each time the moving window centered on a
natural or semi-natural pixel required exporting rasters and calling Circuitscape, a relatively
time-intensive program, to map current flow within the window. Analyzing our study area in
this way could easily have required months of processing time.
To speed up processing, we employed a computational shortcut in which the moving window
centered on blocks of pixels rather than on each individual pixel. Pixels in the block with source
weights > 0 were considered potential targets for flow, and those in the remainder of the 50-
km radius area were considered potential sources (Fig. C1). For each block, we summed the
source weights of all potential target pixels, and the target block was assigned a weight equal to
this value. A total amount of current equal to this target weight was then allocated to the
source pixels in the window in proportion to their individual source weights. This resulted in an
amount of current emanating from each source pixel equal to the summed target weights
multiplied by the source pixel weight divided by the summed weight of all source pixels. The
total flow emanating from all sources was thus equal to the target weight. In this way, the total
flow in a landscape scales linearly with the amount of natural lands to connect. As with the
simpler case above, the model can alternatively be parameterized such that the total flow is
equal to the product of source and target weights, but we found this unreasonably penalized
coastal areas and landscapes with any appreciable degree of human use.
Figure C1. Moving window (solid red circle)
centered on a target block of 31 x 31 pixels (solid
red square) instead of a single target pixel. All
pixels with source weights > 0 inside the window
but outside of the center target block are treated as
current sources. A single pixel at the center of the
target block is set to ground, which acts as the
destination for all flow. Total current flow from
sources equals the summed source weights of all
pixels in the target block. After solving for this
moving window, the window would move 31 pixels
to the right (dashed dark red circle), and the
process would be repeated for the next target block
(dashed dark red square).
A single raster with all sources and associated weights surrounding each block and a raster
representing a single grounded pixel at the center of the block were then saved to disk. These
and a subset of the resistance raster with all cells outside of the radius set to NoData were used
as inputs to Circuitscape in advanced mode, producing a single current map for the moving
window area, as in Fig. 3. After each computation, the window would move to the next block
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
46
The Nature Conservancy- June 2016
and the process would begin again. Each current map from Circuitscape was multiplied by a
correction raster (see below) with the result added into a cumulative current raster.
We experimented with different block sizes, and found that even fairly large block sizes yielded
results similar to those without blocking. As reported elsewhere (e.g., McRae et al. 2008),
current flow patterns at coarser pixel sizes also approximated those at finer scales. We used a
block size of 31 x 31 pixels and a pixel size of 180 m because this struck a reasonable balance
between minimizing processing time and reducing artifacts from blocks. Note that flow was not
modeled among pixels within a block, only between block pixels and the area surrounding the
block within 50 km of the block center (Fig. C1). Each pixel with a non-zero source weight was
part of a target block once, and was a source for many computations. In several test
landscapes, the results closely approximated results achieved with running calculations with the
moving window centered on each pixel. Using blocks cut computation time dramatically,
replacing as many as 961 (31 x 31) calls to Circuitscape with a single call. This speeded up the
algorithm more than 100-fold, without lowering the resolution of the resistance data.
Block analyses still represented approximations of those centered on each pixel, and artifacts
were created by the analyses. We employed a simple procedure to remove the artifacts. At the
beginning of each run, the OmniScape code calculated the expected current flow pattern using
null inputs (a resistance raster with all resistances set to 1, and a source weight raster with all
source weights set to 1). From this, we derived the expected current flow pattern to all pixels in
a block (in this case adding up 31 x 31 = 961 null current maps, with one map centered on each
of the pixels in the block). This formed the null expectation for current flow to the block area
under the basic case in which the moving window centered on each pixel. We then calculated
the current flow that would be derived with the block code invoked (one calculation with the
moving window centered on the block center, and a total of 961 amps injected into pixels
outside of the block but inside the 50-km moving window, with the center pixel set to ground).
This formed the null expectation for current flow to the block area when the computational
shortcut was invoked.
We then divided the null expectation under the case in which the moving window centered on
each pixel by the null expectation when the computational shortcut was invoked. This provided
a raster that could be used to correct observed current flow for each moving window iteration.
Current maps from each iteration were multiplied by the correction raster to remove artifacts
before results were summed across all moving windows.
Mapping Omnidirectional Connectivity For Resilient Terrestrial Landscapes In The Pacific Northwest
47
The Nature Conservancy- June 2016
... For example, some types of connectivity needs may be associated with particular types or patterns of land ownership, such as protection or restoration of riparian corridors in an agricultural matrix or reducing forest fragmentation in industrial timberland (Hilty & Merenlender, 2004;Lindenmayer et al., 2000). Categorizing output from a gradient-based assessment by factors related to conservation urgency and feasibility could help practitioners identify and prioritize suites of implementation strategies that collectively can help reduce fragmentation of the landscape (McRae et al., 2016;Schloss et al., 2021). Using spatial analytical tools to map where different conservation strategies could have an effect can also support conservation planning in myriad ways, including impact quantification, stakeholder identification, and decision support (Tallis et al., 2021). ...
... To facilitate the development of conservation priorities and strategies, we categorized Omniscape output by normalizing the continuous current flow, C, with a null model of potential flow C p that was based only on relative naturalness in the source flow, following (McRae et al., 2016). The null model represents the connectivity potential of the landscape in the absence of barriers to movement and was generated by running Omniscape with a constant for R,1. ...
... Impeded areas (≤0.7) and areas with Limited movement potential were places where land uses or infrastructure (e.g., highways) presumably fragment or greatly restrict movement, respectively. We visually evaluated 54 combinations of category thresholds and determined that the same thresholds that were applied in the Pacific Northwest (McRae et al., 2016) best characterized the connectivity categories in California. ...
Article
Full-text available
Protecting or restoring habitat connectivity in landscapes undergoing rapid environmental change requires multiple conservation and restoration strategies. These strategies have different risk profiles, costs, and require various types of technical expertise to conduct. This diversity in landscape context and strategic approach requires more nuance and flexibility than traditional connectivity conservation plans have supported. We present a novel, spatially‐explicit framework for developing connectivity conservation priorities and strategies based on Omniscape, an adaptation of Circuitscape, a common tool for mapping habitat connectivity. Using California (USA) as a case study, we mapped structural connectivity and developed a classification of connectivity conservation and restoration categories across the gradient of land use intensities, as well as by land ownership. The most constrained areas with highly concentrated flow (movement potential) make up 3% of the state and occur primarily on private lands. Conversely, intact areas with diffuse flow that indicate multiple connectivity options cover 55% of the state, including the majority of the desert and mountain ecoregions. This “strategy mapping” approach can be used to identify priority areas for conservation investment and suites of potential implementation mechanisms and partners, which in turn may improve the efficiency and effectiveness of connectivity conservation in this era of global change. Conserving and restoring habitat connectivity in complex landscapes requires diverse skills, strategies, and partnerships. We present a strategy mapping framework for connectivity using a novel model and illustrate its application in California.
... Our analyses utilize current flow in a wall-to-wall framework using the Omniscape algorithm (McRae et al., 2016;Landau et al., 2021). Omniscape is a newer application of circuit theory that does not rely on definition of core habitat patches; instead, connectivity is assessed continuously across the landscape. ...
Article
Full-text available
Roads are not the only determining factor for wildlife movement across the landscape, but due to the extensive distribution of the road network their impact can be dramatic. Although it has been well documented that roads decrease habitat connectivity for wildlife due to animal-vehicle collisions, habitat fragmentation, and avoidance behavior, approaches for identifying connectivity across the landscape often do not fully examine the barrier effect of roads. Here, we explored the extent of the impact of roadways on wildlife connectivity by using Omniscape to model connectivity including and without the barrier effect of roads, then evaluating the difference between these two models. We created these connectivity models for three organisms that represent different taxa, movement types, and habitat requirements: northern red-legged frog, Pacific-slope flycatcher, and Columbian black-tailed deer. We found that roads had a strong impact on connectivity for all three species. Change in flow was most pronounced on the roads, especially where they ran through permeable habitat for a species. Roads also influenced connectivity well beyond the footprint of the roadway, affecting flows intersecting the roads and diffusely around them. The extent and nature of this impact depended on the species, road density, and surrounding habitat. The different effects across species highlight the importance of considering different taxa simultaneously while planning. Moreover, the ability to assess modeled wildlife habitat connectivity in the absence of existing widespread linear infrastructure allows for critical evaluation of where mitigation activities, such as wildlife crossing structures and fencing, may be most beneficial. Hence, this novel approach has practical application for increasing connectivity for wildlife across roads.
... points and for the whole landscape using a circuit theory approach (Anantharaman et al., 2020;McRae et al., 2016). Borrowing from circuit theory, Circuitscape models landscape connectivity based on a resistance matrix and calculates the connectivity between two locations as the cumulative current between the two points of the circuit. ...
Article
Full-text available
Ecological processes and biodiversity patterns are strongly affected by how animals move through the landscape. However, it remains challenging to predict animal movement and space use. Here we present our new r package enerscape to quantify and predict animal movement in real landscapes based on energy expenditure. enerscape integrates a general locomotory model for terrestrial animals with GIS tools in order to map energy costs of movement in a given environment, resulting in energy landscapes that reflect how energy expenditures may shape habitat use. enerscape only requires topographic data (elevation) and the body mass of the studied animal. To illustrate the potential of enerscape, we analyse the energy landscape for the Marsican bear (Ursus arctos marsicanus) in a protected area in central Italy in order to identify least‐cost paths and high‐connectivity areas with low energy costs of travel. enerscape allowed us to identify travel routes for the bear that minimize energy costs of movement and regions that have high landscape connectivity based on movement efficiency, highlighting potential corridors. It also identifies areas where high energy costs may prevent movement and dispersal, potentially exacerbating human–wildlife conflicts in the park. A major strength of enerscape is that it requires only widely available topographic and body size data. As such, enerscape permits a first cost‐effective way to estimate landscape use and movement corridors even when telemetry data are not readily available, such as for the example with the bear. enerscape is built in a modular way and other movement modes and ecosystem types can be implemented when appropriate locomotory models are available. In summary, enerscape is a new general tool that quantifies, using minimal and widely available data, the energy costs of moving through a landscape. This can clarify how and why animals move in real landscapes and inform practical conservation and restoration decisions.
... This work reviews a pathway toward climate change adaptation planning in a region currently facing tangible threats from climate change. Future climate change refugia conservation in the Sierra Nevada might incorporate regional landscape connectivity work (Buttrick et al., 2015;McRae et al., 2016), considerations for increased humanwildlife conflict and zoonotic disease (e.g., Hammond, Liebman, Payne, Pigage, & Padgett, 2020;MacDonald, McComb, O'Neill, Padgett, & Larsen, 2020), additional existing climate change refugia conservation approaches in the region (Buhler et al., 2019), and much more. Placing priorities and resources into an actionable framework provides ideas for near-term application, and can stimulate additional collaboration to meet the challenges of climate change adaptation in the Sierra Nevada ecoregion. ...
Article
Full-text available
Abstract Climate change uncertainty poses serious challenges to conservation efforts. One emerging conservation strategy is to identify and conserve climate change refugia: areas relatively buffered from contemporary climate change that enable persistence of valued resources. This management paradigm may be pursued at broad scales by leveraging existing resources and placing them into a tangible framework to stimulate further collaboration that fosters management decision‐making. Here, we describe a framework for moving toward operationalizing climate change refugia conservation at an ecoregion scale with an analysis for the Sierra Nevada ecoregion (CA, USA). Structured within the Climate Change Refugia Conservation Cycle, we identify a preliminary suite of conservation priorities for the ecoregion, and demonstrate how existing mapping, data, and applications could be used for identifying, prioritizing, managing, and monitoring refugia. We focus on six stakeholder‐identified conservation priorities, including two process‐based refugial priorities (snow and fire), and four ecosystem‐based refugial priorities (meadows, giant sequoia, old growth forests, and alpine communities). This pilot overview of concepts and resources provides a foundation for both near‐term implementation and further discussion in moving from science to conservation practice. Such an approach may provide new practical insights for ecosystem management at ecoregion scales in the face of climate change.
Article
Global policies call for connecting protected areas (PAs) to conserve the flow of animals and genes across changing landscapes, yet whether global PA networks currently support animal movement-and where connectivity conservation is most critical-remain largely unknown. In this study, we map the functional connectivity of the world's terrestrial PAs and quantify national PA connectivity through the lens of moving mammals. We find that mitigating the human footprint may improve connectivity more than adding new PAs, although both strategies together maximize benefits. The most globally important areas of concentrated mammal movement remain unprotected, with 71% of these overlapping with global biodiversity priority areas and 6% occurring on land with moderate to high human modification. Conservation and restoration of critical connectivity areas could safeguard PA connectivity while supporting other global conservation priorities.
Technical Report
Full-text available
Migration is widespread across taxonomic groups and increasingly recognized as fundamental to maintaining abundant wildlife populations and communities. Many ungulate herds migrate across the western United States to access food and avoid harsh environmental conditions. With the advent of global positioning system (GPS) collars, researchers can describe and map the year-round movements of ungulates at both large and small spatial scales. The migrations can traverse landscapes that are a mix of different jurisdictional ownership and management. Today, the landscapes migrating herds traverse are increasingly threatened by fencing, high-traffic roads, oil and gas development, and other types of permanent development. Through the use of GPS collars, a model of science-based conservation emerged in which migration corridors, stopovers, and winter ranges can be mapped in detail, thereby allowing threats and conservation opportunities to be identified and remedied. In 2018, the U.S. Geological Survey (USGS) assembled a Corridor Mapping Team (CMT) to work collaboratively with western states to map migrations of Odocoileus hemionus (mule deer), Cervus canadensis (elk), and Antilocapra americana (pronghorn). Led by the USGS Wyoming Cooperative Fish and Wildlife Research Unit, the team consists of Federal scientists, university researchers, and biologists and analysts from participating State and Tribal agencies. The first set of maps described a total of 42 migrations across 5 western states and was published in 2020 as the first volume of this report series. This second volume describes an additional 65 migrations mapped within 9 western states and select Tribal lands. As the western United States continues to grow, this report series and the associated map files released by the USGS will allow for migration maps to be used for conservation planning by a wide array of State and Federal stakeholders to reduce barriers to migration caused by fences, roads, and other development.
Article
Although matrix improvement in fragmented landscapes is a promising conservation measure, matrix permeability (willingness of an organism to enter the matrix) and movement survival in the matrix are usually aggregated. Consequently, it is unknown which matrix property needs to be improved. It also remains unclear whether matrix upgrading from dispersal passage to providing reproduction opportunities has large conservation benefits and whether there are interactive effects between habitat and matrix management. We examined matrix effects on regional populations across a gradient of habitat loss and fragmentation using simulation experiments that integrated demographic processes and movement modelling based on circuit theory. We separately modified the levels of matrix permeability and movement survival to evaluate their individual effects. We also altered the amount and configuration of not only habitat but also improved matrix to assess their effects on population vital rates (size, survival and density). In binary landscapes comprising habitat and unimproved matrix, matrix movement survival had larger effects on population vital rates than matrix permeability. Increasing movement survival increased vital rates, yet, increasing matrix permeability decreased vital rates. Increased permeability required corresponding increased movement survival to offset potential negative population outcomes. When subsets of the matrix functioning as dispersal passage only (where no reproduction opportunities existed) were improved, increasing matrix permeability but holding movement survival constant reduced all vital rates, especially with increasing habitat fragmentation. In contrast, when movement survival increased, vital rates increased given strong habitat fragmentation. The benefits of upgrading dispersal passage to provide reproduction opportunities for population survival were greatest when habitat amount was moderate. We also found synergetic effects between amounts of habitat and improved matrix, and the benefits of matrix improvement were promoted when improvement was achieved in a spatially aggregated manner. Synthesis and applications. Matrix improvement and connectivity modelling aimed at increasing movement survival will likely bring larger conservation benefits than those for improving permeability alone. Buffering and connecting habitat remnants with improved matrix could provide benefits as long as movement survival is increased. Simultaneous implementation of habitat management and matrix improvement would yield synergistic conservation benefits. 分断化景観でのマトリックスの改善は有望な保全手法だが、マトリックスの透過性(生物がマトリックスに入る確率)とマトリックスでの移動生存率は通常まとめて扱われ、いずれのマトリックスの属性を改善すべきか不明である。また、マトリックスを移動通路から繁殖の機会を提供するまで改善すると大きな保全上の便益が得られるのか、生息地とマトリックスの管理の間に交互作用があるのかも明らかになっていない。 私たちは、人口統計学的過程とサーキット理論に基づいた移動モデルを統合したシミュレーション実験を行ない、生息地の消失と分断化の傾度上で、地域個体群にマトリックスが及ぼす影響を調査した。ここで異なるマトリックスの属性としてマトリックスの透過性と移動生存率を個別に変化させ、個々の影響を評価した。また生息地だけでなく改善したマトリックスの量と配置も変化させ、個体群統計量(サイズ、生存、密度)に及ぼす影響を調査した。 生息地と改善されていないマトリックスからなる二値の景観では、マトリックスの移動生存率は透過性よりも個体群統計量に大きな影響を及ぼしていた。移動生存率の増加は個体群統計量を増加させたが、透過性の増加は個体群統計量を減少させた。 移動通路としてのみ機能するマトリックス(繁殖の機会は存在しない)の一部を改善する場合、移動生存率を一定にしてマトリックスの透過性を増加させるとすべての個体群統計量が特に生息地が分断化した場合に大きく減少した。対照的に、移動生存率が増加すると、個体群統計量は強度の生息地の分断化下で増加した。マトリックスを移動通路から繁殖の機会を提供するまで改善すると、生息地の量が中程度の際に個体群生存率への便益は最も大きかった。生息地と改善マトリックスの量の間には相乗効果があり、マトリックス改善は空間的にまとめて実施すると便益が大きくなった。 まとめと応用:マトリックスの改善と連結性のモデリングは移動生存率の増加を目指して行なうと、透過性のみを改善する場合に比べて保全上の便益が大きいだろう。改善マトリックスで残存生息地の緩衝帯を設け、残存生息地間を連結させることは、移動生存率が増加するならば便益を生むだろう。生息地の管理とマトリックスの改善を同時に実施すれば、相乗的な保全上の便益がもたらされるだろう。 Matrix improvement and connectivity modelling aimed at increasing movement survival will likely bring larger conservation benefits than those for improving permeability alone. Buffering and connecting habitat remnants with improved matrix could provide benefits as long as movement survival is increased. Simultaneous implementation of habitat management and matrix improvement would yield synergistic conservation benefits.
Preprint
Full-text available
Governments around the world have acknowledged the importance of conserving ecological connectivity to help reverse the decline of biodiversity. In this study we employed recent methodological developments in circuit theory to conduct the first pan-Canadian analysis of multi-species connectivity for all terrestrial regions of the country, at a spatial grain sufficient to support local land-management decisions. We developed a movement cost surface with a limited number of thematic categories using the most recently updated land cover data available for the country. We divided the country into 17 tiles and used a wall-to-wall, omnidirectional mode of Circuitscape on each tile in order to assess ecological connectivity throughout entire landscapes as opposed to strictly among protected areas. The resulting raw current density map of Canada revealed heterogenous patterns of current density across the country, strongly influenced by geography, natural barriers, and human development. We included a validation analysis of the output current density map with independent wildlife data from across the country and found that mammal and herpetofauna locations were predicted by areas of high current density. We believe our current density map can be used to identify areas important for connectivity throughout Canada and thereby contribute to efforts to conserve biodiversity.
Article
As both plant and animal species shift their ranges in response to a changing climate, maintaining connectivity between present habitat and suitable habitat in the future will become increasingly important to ensure lasting protection for biodiversity. Because the temporal period commensurate with planning for mid-century change is multi-generational for most species, connectivity designed to facilitate climate adaptation requires pathways with 'stepping-stones' between current and future habitat. These areas should have habitats suitable not only for dispersal, but for all aspects of species lifecycles. We integrated present-day land use, topographic diversity, and projections of shifting climate regimes into a single connectivity modeling approach to identify pathways for mid-century shifts in species ranges. Using Omniscape we identified climate linkages, or areas important for climate change-driven movement, as the areas with more current flow than would be expected in the absence of climate considerations. This approach identified connectivity potential between natural lands in the present climate and natural lands with future analogous climate following topo-climatically diverse routes. We then translated the model output into a strategic framework to improve interpretation and to facilitate a more direct connection with conservation action. Across modified landscapes, pathways important to climate-driven movement were highly coincident with the last remaining present-day linkages, reinforcing their importance. Across unfragmented lands, the presence of climate-adapted pathways helped inform the prioritization of conservation actions in areas where multiple connectivity options still exist. Many climate linkages follow major watercourses along elevational gradients, highlighting the importance of protecting or managing for these natural linear pathways that provide movement routes for climate adaptation. By integrating enduring landscape features with climate projections and present-day land uses, our approach reveals "no-regrets" pathways to plan for a connected landscape in an uncertain future.
Preprint
Full-text available
Rapid environmental change threatens to isolate the world's wildlife populations and intensify biodiversity loss. Global policies have called for expanding and connecting the world's protected areas (PAs) to curtail the crisis, yet how well PA networks currently support wildlife movement, and where connectivity conservation or restoration is most critical, have never been mapped globally. Here, we map the functional connectivity (how animals move through landscapes) of the world's terrestrial PAs for the first time. Also, going beyond existing global connectivity indices, we quantify national PA-connectedness using an approach that meaningfully represents animal movement through anthropogenic landscapes. We find that reducing the human footprint may improve national PA-connectivity more than adding new PAs; however, both strategies are critical for improving and preserving connectivity in places where the predicted flow of animal movement is highly concentrated. We show that the majority of critical connectivity areas (CCAs) (defined as globally important areas of concentrated animal movements) remain unprotected. Of these, 72% overlap with previously-identified global conservation priority areas, while 3% of CCAs occur within moderate to heavily modified lands. Conservation and restoration of CCAs could safeguard connectivity of the world's PAs, and dovetail with previously identified global conservation priorities.