Content uploaded by Yolanda Wiersma
Author content
All content in this area was uploaded by Yolanda Wiersma
Content may be subject to copyright.
This version is a postprint. It has the same peer-reviewed content as the published version, but lacks publisher
layout and branding. It is available under License Creative Commons Attribution Non-commercial. Published
version is copyrighted and available at: Bourne CM, DG Kehler, YF Wiersma, D Cote.2011. Barriers to fish
passage and barriers to fish passage assessments: the impact of assessment methods and assumptions on
barrier identification and quantification of watershed connectivity. Aquatic Ecology 45: 389-403. ISSN 1573-5125
1 Barriers to fish passage and barriers to fish passage assessments: the
2 impact of assessment methods and assumptions on barrier identification and
3 quantification of watershed connectivity
4
5 Christina M. Bourne
1,4
, Dan G. Kehler
2
, Yolanda F.Wiersma
1,*
, David Cote
3
6
7
1
Department of Biology, Memorial University, St. John’s NL A1B 3X9.
8
2
Atlantic Service Center, Parks Canada, 1869 Upper Water St. Halifax, NS, B3J 1S9
9
3
Terra Nova National Park, Parks Canada, General Delivery, Glovertown, NL, A0G 2L0
10
4
current address: Department of Fisheries and Oceans, Northwest Atlantic Fisheries Centre, St. John’s NL
11 A1C 5X1
12 *author for correspondence. Email: ywiersma@mun.ca, Ph. (709) 864-7499 Fax: (709) 864-3018
13
14
This version is a postprint. It has the same peer-reviewed content as the published version, but lacks publisher
layout and branding. It is available under License Creative Commons Attribution Non-commercial. Published
version is copyrighted and available at: Bourne CM, DG Kehler, YF Wiersma, D Cote.2011. Barriers to fish
passage and barriers to fish passage assessments: the impact of assessment methods and assumptions on
barrier identification and quantification of watershed connectivity. Aquatic Ecology 45: 389-403. ISSN 1573-5125
15 Abstract
16 Barriers (culverts and dams) can impede fish passage and affect the overall habitat connectivity of rivers.
17 However, a challenge lies in how to conceptualize and adequately measure passability at barriers. We
18 hypothesize that estimates of barrier and watershed connectivity are dependent on assumptions about the
19 nature of passability, and how it is measured. Specifically, we compare passability estimates in Terra
20 Nova National Park, Canada for individual barriers for two barrier assessment methods (a rapid
21 assessment, and one based on FishXing software), two salmonid species, different fish sizes and
22 swimming speeds, and varying hydrological conditions. Watershed connectivity was calculated using the
23 Dendritic Connectivity Index (DCI). Lastly, we test to see what the impact of the various factors is on the
24 practical goal: prioritizing barriers for restoration. Our results show that barrier passability estimates can
25 vary drastically for some barriers (0-100%). In general, the rapid field-based assessment tended to give
26 more conservative estimates of passability than those based on FishXing. Estimates of watershed
27 connectivity were not as sensitive to the assumptions and methods used (DCI: 40-83). Fish size had the
28 greatest effect on DCI. Importantly, variation in DCI had little impact on the restoration priorities. The
29 same barrier was retained as the top priority >96% of the time. Thus, managers wishing to assess barriers
30 for restoration need to carefully consider how passability is to be measured, but can reduce the impact of
31 these decision by considering barriers in their watershed context by using a connectivity index such as the
32 DCI.
33
34 Keywords: barrier assessment; connectivity; dendritic connectivity index; passability; salmonids
35
36
37
This version is a postprint. It has the same peer-reviewed content as the published version, but lacks publisher
layout and branding. It is available under License Creative Commons Attribution Non-commercial. Published
version is copyrighted and available at: Bourne CM, DG Kehler, YF Wiersma, D Cote.2011. Barriers to fish
passage and barriers to fish passage assessments: the impact of assessment methods and assumptions on
barrier identification and quantification of watershed connectivity. Aquatic Ecology 45: 389-403. ISSN 1573-5125
38 Introduction
39 Fragmentation of many of the world’s stream networks has been recognized as a serious threat to the
40 population diversity, abundance, and persistence of a variety of aquatic species (e.g., Sheldon 1988;
41 Dunham et al. 1997; Khan and Colbo 2008). Human activities are largely to blame for these connectivity
42 losses, often through the installation of physical barriers (such as dams and culverts) to movement (e.g.,
43 Morita and Yamamoto 2002; Park et al. 2008; Doehrign et al. 2011; Hall et al. 2011; Rolls 2011). While
44 many of these barriers can be eliminated or mitigated by modification, such as by the construction of
45 fishways, the process is typically expensive and budgetary constraints restrict the amount of restoration
46 that can occur (Gibson et al. 2005; Poplar-Jeffers et al. 2008). Thus, a solid understanding of the
47 ecological impacts of potential barriers is essential to prioritize restoration efforts and maximize returns
48 on limited funding. Although in simplest terms we know that barriers impact the passage of fish,
49 quantifying this impact is challenging because barrier passability is difficult to define and measure. Many
50 definitions and methods for estimating passability exist (see Kemp and O’Hanley 2010 for a recent
51 summary). Common methods include measuring or modelling the physical characteristics of a barrier and
52 comparing it to known fish physiological parameters (e.g., FishXing; USFS 2003 for culverts), through
53 mark-recapture (e.g., Helfrich et al. 1999; Porto et al. 1999 for dams; Blank et al. 2005 for culverts),
54 analysis of genetic structure of the population (Neraas and Spruell 2001; Kemp and O’Hanley 2010) or by
55 tracking individual fish attempting to navigate the barrier (Bjornn and Peery 1992; Steig et al. 2005;
56 Cahoon et al. 2007). Passability is also challenging to quantify because it is dynamic. Fish physiological
57 capacity varies by species, size, amongst individuals and across environmental conditions, while the
58 physical characteristics of barriers also vary temporally due to variations in stream flow (see Bjornn and
59 Peery 1992; Rolls 2011). Such physiological and environmental variability makes the task of defining
60 passability at a population or landscape scale challenging.
61 A second factor important to understanding barrier impacts is the need to consider the context in
This version is a postprint. It has the same peer-reviewed content as the published version, but lacks publisher
layout and branding. It is available under License Creative Commons Attribution Non-commercial. Published
version is copyrighted and available at: Bourne CM, DG Kehler, YF Wiersma, D Cote.2011. Barriers to fish
passage and barriers to fish passage assessments: the impact of assessment methods and assumptions on
barrier identification and quantification of watershed connectivity. Aquatic Ecology 45: 389-403. ISSN 1573-5125
62 which a barrier is found (Cote et al. 2009; Rolls 2011). Previous studies of aquatic barriers were based
63 largely on the effects that the barriers had on nearby portions of stream systems (e.g., Belford and Gould
4
64 1989). More recently, concepts from landscape ecology, such as fragmentation and patch dynamics, have
65 been applied to aquatic systems to investigate the impacts of barriers on entire stream networks and
66 catchments (Dunham et al. 1997; Jones et al. 2000; Park et al. 2008; Cote et al. 2009; Fullerton et al.
67 2010). This broad view is crucial to understanding and mitigating the ecological consequences of stream
68 fragmentation, as the effects of even a single barrier may have large impacts on entire stream networks,
69 and multiple barriers may lead to cumulative impacts (Kemp and O’Hanley 2010; Rolls 2011; however,
70 see Padgham and Webb 2010 for a model which suggests that multiple impacts are simply equal to the
71 sum of the parts).
72 One method for quantitatively evaluating the cumulative impacts of barriers on entire stream
73 networks is the Dendritic Connectivity Index (DCI; Cote et al. 2009), which could be a valuable tool for
74 assessing fragmentation in stream systems and for prioritizing barrier restoration. The DCI requires two
75 key data inputs; spatial location of barriers (both artificial and natural) within a stream or river network,
76 and a passability value for each individual barrier. While spatial data for the barriers are relatively simple
77 to acquire, useful barrier passability estimates not. In this paper, we examine how different barrier
78 assessment methods and definitions of passability affect i) estimates of connectivity at both the barrier
79 and the landscape scale, as measured by the DCI, as well as ii) the prioritization of restoration efforts.
80 Cote et al. (2009) suggested passability could be quantified in several different ways and noted
81 that decisions on how to define and measure passability would be important to interpret and evaluate
82 watershed connectivity. Interpretations that capture the variability in fish physiology within and among
83 species (e.g., assigning a passability of 0.5 to a barrier that is passable to 50% of the target population)
84 may be insensitive to temporal environmental variation, while definitions that account for temporal
85 variation of physical characteristics (e.g., the barrier is passable 50% of the time to fish with a defined
86 physiological capacity) may not account for variation amongst individual fish. Furthermore, once defined,
87 subsequent passability values will reflect decisions regarding the time period of the assessment (i.e.,
88 stream discharge), the species being modelled and the accuracy of the swim speeds estimates.
89 Unfortunately, the sensitivity of barrier passability estimates, and subsequent watershed connectivity
5
90 estimates, to these decisions is unknown. If these measures are highly sensitive to variation in fish
91 physiology or environmental conditions at barriers, then the utility and general applicability of watershed
92 connectivity estimates will be reduced, and managers wishing to use them will have to very careful about
93 how data are collected.
94 We used river/stream systems (hereafter “watersheds”) of Terra Nova National Park (TNNP),
95 Newfoundland and Labrador, Canada, (which ranged in area from 0.5 km
2
to 36 km
2
) as a case study to
96 examine the sensitivity of passability estimates and resulting river connectivity (watershed and park-
97 wide)to four aspects of barrier passability: i) the fish species of interest, ii) barrier assessment
98 methodology, iii) inter- and intra-annual variability in stream flow, and iv) assumptions about fish
99 swimming capacity. The results of these simulation scenarios are also evaluated in terms of their effects
100
on restoration priorities in the tested watershed. The simulations are interpreted with respect to the effect
101
on individual barrier assessments, and on the watershed connectivity using the DCI. Finally, we
102
demonstrate a means to include variation of fish physiological capacity and environmental conditions to
103
104
calculate an integrated DCI score.
105 Methods
106
Calculating the DCI
107
The barrier passability values used in DCI calculations range from 0 to 1, with 0 being impassable (a
108
complete barrier), 1 fully passable and values in between considered partially passable. We obtained
109
connectivity values for potamodromous (DCI
P
,) and diadromous (DCI
D
,) fish life histories for all park
110
catchments using fish size/swim speed and stream flow parameters. Diadromous life history refers to fish
111
that move between ocean and freshwater (in either direction) during their life cycle. The species examined
112
in this study exhibit anadromy; a form of diadromy where spawning occurs in freshwater and adults spend
113
part of their life at sea, but the form of diadromy is irrelevant for this analysis. The formula for calculating
114
potamodromous connectivity (in both upsteam and downstream directions) is taken from Cote et al.
6
i
115
(2009) and requires dividing the watershed into segments, where segments are separated by barriers. The
116
formula is:
n n
l
l
117
DCI
c
i
j
, (1)
P
i1
j 1
ij
L L
118
where l is the length of segment i and j,c
ij
is the connectivity between segments i and j, and L is the
119
total stream length. Diadromous connectivity applies to both anadromous and catadromous (migrating
120
from ocean to freshwater) cases and is calculated as follows from Cote et al. (2009):
n
l
M
u d
121
DCI
D
L
(
m
p
m
p
m
) *100 , (2)
1
i1
122
123
where l
i
is the length of segment i, and are upstream and downstream passabilities of the m
th
barrier (m=1….M) between the river mouth and section i, and L is the total stream length. Maximum
124
DCI value is 100, which indicates a fully connected watershed, with connectivity decreasing as DCI
125
126
values decrease from 100.
127
Fish species of interest
128
Barrier assessments were conducted for two different salmonid species (brook trout, Salvelinus fontinalis
129
and Atlantic salmon, Salmo salar). These species have well-studied physiology, are widely distributed in
130
the study area and are culturally and recreationally important (Scott and Crossman 1973). Though brook
131
trout and Atlantic salmon are of the same family, Atlantic salmon have superior swimming capabilities
132
(Peake et al. 1997) and diadromous individuals can attain larger sizes than those of brook trout. We based
133
our Atlantic salmon assessments on the physiology of a 50 cm (fork length; FL) individual and the
134
135
physiology of a 15 cm (FL) individual for brook trout assessments.
136
Barrier assessment methodology – rapid assessment vs. modelling
137
We used two methods to evaluate passability of all culverts in TNNP (n = 43); rapid field assessments
138
(which examine culvert passabilities during a single visit), and more detailed field data coupled with
7
139
modelling software (that integrates variation in stream flow in the evaluation of culvert passabilities).
140
Field assessments consisted of a screening process for barriers, based on a set of criteria (Fig. 1) adapted
141
from previous culvert inventories (Clarkin et al. 2005). These criteria included culvert slope, outflow drop
142
height and presence of an outflow pool. FishXing, a widely used freeware, creates hydrological models of
143
culverts based on data collected in the field (culvert shape, length (m), material, slope, installation type)
144
together with flow equations and fish movement parameters. While FishXing can model culverts using
145
minimal field data, more detailed data can be included such as the cross-section topography of the
146
tailwater control area and discharge rates for the study stream. FishXing also identifies which of three
147
mechanisms impede the passage of fish: insufficient water depth in the culvert (depth barrier), excessive
148
height for fish to jump into the culvert (height barrier), and excessive water flow for fish passage (velocity
149
barrier).
150
The data collection for the rapid assessment surveys took from 5 to 15 minutes per culvert,
151
whereas for the FishXing assessments, surveys in the field took from 20 to 40 minutes per culvert with an
152
additional time of 5-10 minutes per culvert for computer simulations (and more when default values
153
proved problematic - see below for details). We collected additional parameters (e.g., water depth and
154
water velocity in culvert) to ground-truth FishXing results, and three culverts were revisited to improve
155
congruence between field and FishXing outputs. Rapid assessment surveys were carried out in May and
156
June of 2007. Using the FishXing software also requires additional inputs of fish limitations for burst and
157
sustained swimming speed, minimum water depth and maximum outflow drop. These values were
158
obtained for our species from Peake et al. (1997) and from Peake (unpublished data).
159
We encountered some challenges when assessing culverts using FishXing. Specifically, the
160
default values provided by the software for the culvert entrance loss coefficient (K
e
) and the culvert and
161
tailwater control roughness coefficients (n) – parameters used to model water flow in open channels
162
(Brater and King 1976) – did not provide accurate approximations of field conditions. Thus, at a given
163
discharge rate, modelled values of culvert water depth and velocity were often very different from the
164
actual values measured in the field at that discharge rate – leading us to suspect that the modelled values
8
165
provided by FishXing at other discharge rates were also inaccurate. This issue has been observed in other
166
evaluations using FishXing (Blank et al. 2005; Poplar-Jeffers et al. 2008) and likely occurs because the
167
software uses K
e
values which are derived from culverts under full water and roughness coefficients
168
which are often derived from large streams and generalized to all streams without considering details such
169
as the presence of debris, inconsistencies in substrate across a small area or rapid changes in slope or
170
wetted width (R. Gubernick, FishXing design team, pers. comm.; see also Mangin et al. 2010). To more
171
accurately model the study sites, we obtained new K
e
values for partially full culverts from Straub and
172
Morris (1950ab) and back-calculated new roughness coefficient values (n) using field data from original
173
culvert surveys and from the three culverts which were revisited for ground-truthing. Though the culvert
174
parameters provided by FishXing did not always match field values exactly, our modifications to the n
175
and K
e
values did improve the precision of all culvert models. Passability estimates obtained from rapid
176
assessments (using first visit data only) and more detailed field surveys plus FishXing modelling were
177
178
used to calculate DCI
P
and DCI
D
for all catchments in Terra Nova National Park.
179
Temporal variability in stream flow
180
We investigated the effect of intra-annual stream flow variability for all park watersheds (n = 15). We
181
calculated the DCI
P
and DCI
D
for two time periods: when fish are migrating, and the whole year (Table
182
1), using daily discharge data averaged over a twenty year period from the Southwest Brook gauging
183
(48°36’27” N, 53°58’44”W station 02YS003) station located in the national park. We investigated the
184
effect of inter-annual variability in water flow on the DCI
P
and DCI
D
within a test watershed, Big Brook
185
(Fig. 2). To investigate inter-annual variability, we calculated the DCI
P
and DCI
D
for twenty different
186
years using daily discharge data (Table 1).
187
For each analysis, we scaled gauging-station hydrographs for each barrier by calculating the ratio
188
of the area draining into the stream gauge location to that of the area draining into the barrier. This
189
assumes that discharge rate is proportional to catchment size. FishXing determines whether a barrier is
9
l
N
190
passable at a range of flow values between the minimum and maximum provided. Using these results, we
191
192
determined passability as the proportion of days the flow would allow a fish of the given size to pass.
193
Variable fish swim speed
194
We modelled fish passage for a range of swimming speed scenarios in each culvert in the test watershed,
195
Big Brook (n = 18 culverts). We set a range of ‘user-defined’ burst and sustained swim speeds in
196
FishXing for our study species to model the effect of fish size and swimming ability on passability. These
197
speeds are summarized in Table 2, and are based on models for brook trout and Atlantic salmon by Peake
198
et al. (1997), who conducted swim speed tests using fish from a watershed in north central
199
Newfoundland. Though Peake’s study used forced performance models, which recent research has shown
200
to produce conservative measures (Peake and Farrell 2006), it likely represents the best available data as
201
fish were collected from an area close to the TNNP study site. We used this speed ±25% to account for
202
individual variability and uncertainty due to the fact that speeds were based on forced performance
203
204
models (Peake and Farrell 2006) (Table 1).
205
Calculating a population-integrated watershed connectivity score.
206
Using the barrier passability results for fish of different lengths, and a length-frequency distribution for a
207
population of interest, we can calculate a population-integrated DCI score using a weighted mean:
208
Weighted mean DCI =
DCI
n
l
l
(3)
209
where l is the length class, n
l
is the number of fish of that length class, and N is the total number of fish.
210
Length-frequency data and species composition were obtained from past field sampling programs
211
from ponds and streams throughout TNNP. These data were obtained from samples collected over many
212
seasons and thus represent a general characterization of fish communities in the study area. Fish
213
communities vary by habitat and life history types. Therefore we determined species composition and fish
214
lengths according to each habitat (stream vs. lake) and life history subset (potamodromous vs.
10
215
diadromous). The diadromous length-frequency distribution and relative species abundance were derived
216
from two fish counting fences of similar size to Big Brook (Minchins Brook, Cote et al. (2005); Wings
217
Brook, Potter (1989)) during the migration period. Potamodromous fish communities were characterized
218
based on electrofishing in streams (Cote 2007) and fyke netting in lakes (Cote et al. 2005; Cote et al. in
219
press) throughout TNNP. Population abundance for brook trout and Atlantic salmon was calculated using
220
available habitat in the Big Brook system and existing habitat models (Cote 2007; Cote et al. in press).
221
Finally, the integrated abundance-weighted watershed connectivity value for Big Brook was calculated
222
using equation 3. Since barrier passability values were not available for all fish lengths, we used length
223
224
categories defined by the midpoints between length values in Table 2 for each of the two species.
225
Identifying priority culverts for restoration of watershed connectivity
226
To prioritize culvert replacement based on the greatest potential gains to connectivity, we simulated
227
restoration of each culvert, individually, to full passability (i.e., barrier passability was set to 1) and then
228
re-calculated DCI values for the Big Brook watershed using all possible scenarios of inter-annual stream
229
flow variability between 1998-2008, and fish length/swim speed and for both Atlantic salmon and brook
230
trout. For each scenario, we ranked the culverts from 1 (most improvement in connectivity) to 18 (least
231
improvement in connectivity) and calculated the average rank, as well as the proportion of scenarios in
232
233
which each culvert was ranked first for restoration.
234 Results
235 We calculated passability, DCI
P
, and DCI
D
with variations in fish species, barrier assessment method,
236
stream flow period, fish length and in fish swimming ability, as described above. Here we report how
237
estimated passability varied at the barrier, and DCI at the watershed, and park scales.
238
Barrier passability
239
The definition and method of measuring passability affected the passability estimate for individual
240
barriers (Fig. 3). Furthermore, the results differed considerably among culverts, with 5 of 18 (28%)
11
241
culverts (ak, an, u, y, and z) impervious to any change in methodology and definition and consistently
242
being completely impassable, and 4 culverts (22%) varying between a passability of 0 and 1 (ao, ag, aj,
243
and w). For these barriers, the range of passabilities was much more likely to include a full barrier (0)
244
than complete passability (1). We performed a simple analysis of variance to decompose the total
245
variance in passability, as represented by sums of squares, into contributions from each factor. Fish length
246
explained the majority of the variance, once the barrier effect was removed (sum of squares (SS) =
247
236.5), followed by variation in swimming speed (SS = 8.6), hydrological year (SS = 7.7), species
248
(SS=3.1) and finally period with the year used for the analysis (SS = 1.1).
249
Single watershed scale
250
Connectivity values at the watershed scale varied less than the passability values of individual barriers
251
(Fig. 3 vs. Fig. 4). For the DCI
P
, the range of values encountered was 40-70, and for the DCI
D
the range
252
of values encountered was 62-83. There was a distinct hump-shaped pattern in DCI values when plotted
253
against fish length for both species in both the potadromous and diadromous cases (Fig. 4). The DCI was
254
lowest for very small fish, and highest for small to mid-sized fish. The DCI was also low for large fish, in
255
some cases as low as that for the smallest size classes. Variation in the DCI due to swim speed was less
256
than the variation due to different stream flows for large fish, but not for small fish (Fig. 4). The effect of
257
interannual variability was fairly constant across both species and all length classes, but tended to be
258
larger for the DCI
P
than the DCI
D
results. As with the barrier scale, we performed a simple analysis of
259
variance to decompose the total variance in DCI
D
as represented by sums of squares, into contributions
260
from each factor. Again, fish length explained the majority of the variance, (~73% , followed by variation
261
in swimming speed (~4%), hydrological year (~1.6%), species (~ 0.4%) and finally period with the year
262
263
used for the analysis (SS = 1.1). Results for the DCI
D
, are very similar.
264
National Park scale
265
Across all watersheds within Terra Nova National Park, DCI values varied depending on whether the
266
rapid field-based assessment or field assessment plus modelling in FishXing was used to estimate barrier
12
267
passability (Fig. 5). DCI values were lower for most catchments when the field assessment alone was
268
used, although the difference was not as dramatic for the diadromous case as the potadromous one. In the
269
potadromous case, 6 watersheds (40%) had DCI values between 0-40 when the field assessments were
270
used, while all watersheds had DCI of 41 or higher when passability estimates from FishXing were used.
271
Overall 12 watersheds (80%) dropped to a lower DCI category (based on categorizing DCI into intervals
272
of 20) (Fig. 5). In the diadromous case, only 5 watersheds (33%) dropped to a lower DCI category (with
273
the field assessment (Fig. 5). DCI values across park watersheds were also quite variable depending on
274
whether passability was calculated based on an annual flow period, or restricted to flow during fish
275
migration period. For example, more watersheds were in a lower category of DCI (<50) when passability
276
was calculated during trout migration period than for the whole year (Fig. 6).
277
The integrated watershed connectivity score for the fish community in Big Brook was 58.3 for
278
brook trout and 67.5 for Atlantic salmon (DCI
P
); and 77.7 for brook trout and 78.1 for salmon (DCI
D
).
279
Lower values indicate lower watershed connectivity. These values are plotted against the median length
280
281
values in Figure 4.
282
Barrier prioritization
283
Finally, the results of the prioritization exercise are shown in Table 3. Since the results for salmon vs.
284
brook trout are very similar, only those from brook trout are presented (the results for salmon are
285
available from the corresponding author on request). For both the DCI
D
and DCI
P
, culvert “al” is the
286
culvert identified as the highest priority for restoration. Culvert “al” was ranked as the priority for brook
287
trout under all combinations of stream flow/swim speed 98% of the time and for salmon under all
288
scenarios 99% of the time (Table 3 shows average data across the two species; data by species are
289
290
291
available from corresponding author by request).
292 Discussion
13
293
The preservation and restoration of aquatic connectivity has been recognized as a major conservation goal
294
in stream systems (Pringle 2003); and new methods have been developed to measure the alteration of
295
connectivity in dendritic systems. Common to all methods is the difficulty in assessing barrier passability
296
– the dynamic component of connectivity. Our results demonstrate how passability varies by species, size
297
and hydrological conditions (see also Poplar-Jeffers et al. 2008; Meizler et al. 2009; Kemp and O’Hanley
298
2010; Rolls 2011) and managers will often be forced to select a target demographic and/or target
299
conditions when evaluating barrier passability. In this study we showed the implications of making such
300
decisions (e.g., differences associated with picking a particular method, or a particular target species/size)
301
as well as the error that may be related to parameter estimates (e.g., swim speed) on passability and
302
connectivity at the watershed scale.
303
A useful result from this work is that watershed scale assessments of connectivity are less
304
sensitive to variations in passability definition or assessment method than estimates of passability for
305
individual barriers. For the DCI results, the choice of fish length had the largest impact on the
306
connectivity score. The effect of fish length on watershed connectivity yielded an unexpected hump-
307
shaped pattern, with smaller and larger fish experiencing lower values. However, this is readily explained
308
by the specific passage requirements of differing size classes. Smaller fish have lower swim speeds and
309
experience velocity barriers during high flow periods, whereas larger fish are limited by the depth of
310
water in the culvert during low flow periods. This is illustrated by the effect of swim speed assumptions
311
on the DCI for small fish, and the insensitivity of the DCI to swim speed assumptions for large fish (Fig.
312
4).
313
We found that watershed connectivity results can vary with barrier assessment methods – making
314
the choice of method a crucial and influential step in connectivity assessment. For most culverts, using a
315
simple set of criteria to do barrier field assessments produced passability values that were more
316
conservative than those calculated by computer modelling (FishXing) for fish of the same size and
317
species, which in turn led to reduced connectivity values (Figs. 3 and 4). It remains likely that the simple
318
field assessments were too conservative when compared to those provided by FishXing. Since the rapid
14
319
field-based assessments have been developed as general installation/assessment guidelines (Fig. 1; see
320
also Clarkin et al. 2005), they do not account for the variable nature of passability. Hence they are
321
necessarily precautionary and less accurate. Though the simplified field assessments did give very
322
different estimates of passability in this study, with modified criteria and further evaluations of partial
323
barriers using FishXing, they could be used more efficiently as tools to save time during culvert surveys
324
by ‘screening’ obvious barriers – a practice which has been implemented in other studies and surveys
325
(e.g., Clarkin et al. 2005). The modelling approach has an advantage in that it can account for variability
326
in passabilty through time and for different species. Unfortunately, specific biological data (i.e., fish
327
telemetry data) were not available to directly assess the accuracy of culvert passability estimates in this
328
study. Such information would enable researchers to assess key assumptions in fish passage but remains a
329
common data gap in passability assessments.
330
In this study, the assessment period (i.e., full year vs. migration period) did not have a substantial
331
impact on watershed connectivity due to the fact that stream hydrology during the migration period for
332
the two species assessed was representative of the entire year (i.e., both including floods and low water
333
events). Thus, in similar systems to TNNP, watershed connectivity estimates based on a shorter
334
hydrological time period might be reliable. These results are specific to the Terra Nova situation, but are
335
likely relevant to watersheds in elsewhere. For example, in an examination of fish community
336
assemblages above and below low-head dams in Kansas, Gillette et al. (2005) found seasonal effects.
337
Similarly, Rolls (2011) examined watersheds with and without barriers in Australia, and found a
338
significant effect of migratory period on barrier passage for some species. Both of these studies (Gillette
339
et al. 2005; Rolls 2011) did not consider overall watershed connectivity, but at the barrier scale the
340
patterns observed were similar to ours in Terra Nova, suggesting that some of our overall conclusions and
341
recommendations on assessment methods may be worth considering in other systems. The relatively
342
minimal impact of temporal scale observed here may not be the case in systems where species have more
343
restricted discharge-dependent migration periods (e.g., Pacific salmon and see Rolls 2011 for an example
344
of variation in connectivity depending on migration strategy), or in seasonally arid landscapes where
15
345
streambeds go dry for months at a time (Eby et al. 2003).Nonetheless, our assessment clearly
346
demonstrates that field assessments that evaluate barriers based on conditions for only a single day (the
347
rapid assessment method, Fig. 1) gives very different values for connectivity than those that use more
348
dynamic assessment methods to evaluate passability. Thus, barrier assessments need to be considered in
349
the context of ecological conditions at a particular study site, and researchers should choose appropriate
350
assessment methods based on the local species and hydrology.
351
Barrier assessments done for two different salmonid species demonstrated the variation in
352
passability values that can be associated with both species and size class. Though brook trout and Atlantic
353
salmon are physically similar species, their swimming capabilities differ – with Atlantic salmon being
354
able to attain higher swimming speeds (Peake et al. 1997) and larger sizes than brook trout. The highest
355
DCI scores were observed for salmon, but the relatively low DCI values obtained for large salmon
356
represents the numerous depth barriers in this system. We set the minimum culvert water depth for both
357
species at 75% of their body length, giving depth values of 11.25cm for brook trout and 37.5cm for
358
salmon. Many of the culverts in our study areas do not have water exceeding 30cm deep. These
359
evaluations were likely conservative, as large Atlantic salmon have been observed moving upstream in
360
water less than 30cm deep in TNNP (D. Cote, pers. obs.). This example demonstrates the importance in
361
choosing parameters for barrier evaluations that are accurate for the study species, and if applicable, the
362
sub-set of the population being targeted. There is a general requirement for better information on fish
363
swimming capacity and behaviour, particularly for non-salmonids (Kemp and O’Hanley 2010).
364
We demonstrate a means to calculate an integrated stream connectivity value that accounts for
365
variation in hydrology, fish size, and species variation. As such, it presents a useful approach for
366
ecosystem based management of aquatic systems. Though the data required to do this are considerable,
367
our results illustrate the difference when using a single target length in TNNP versus an integrated
368
analysis (see position of star on Fig. 4 relative to other data points). Thus, picking “target” species or sizes
369
could cause difficulty in determining a generalized connectivity value, particularly in systems with higher
370
diversity and more varied species. Wiens (2002) suggested that it could be useful to group similar species
16
371
in order to obtain fewer connectivity values per system. However, recent research on fish passage has
372
shown that taxonomic and physical similarities may not be adequate predictors of barrier sensitivity
373
(McLaughlin et al. 2006). Nonetheless, in many cases, assessing watershed connectivity for a specific
374
375
target species of management interest may be very useful and appropriate.
376
Restoration prioritization
377
Prioritization was done using the approach of systematically simulating the restoration of one culvert at a
378
time and assessing the effect on the DCI results. Connectivity in this case is based on the extent of
379
watershed (in km) that becomes available when a barrier is removed, without any consideration of habitat
380
quality (although incorporation of habitat quality is possible with these methods). This approach has the
381
benefit of examining all possible scenarios of which culvert to restore to assess the net gain in
382
connectivity with each. This facilitates a cost-benefit analysis; if the next-to-optimal culvert is
383
significantly cheaper to restore than the most optimal, then this may be the most pragmatic solution.
384
Alternative approaches have been proposed and include using integer-based programming to optimize
385
decisions (O’Hanley and Tomberlin 2005; Kemp and O’Hanley 2010, also see Kibler et al. 2010 for a
386
description of an experimental approach to assessing restoration effects). If restoration decisions were
387
based on prioritizing for the culvert with the lowest passability, then the barrier-scale results would make
388
it difficult to choose the best culvert for restoration. In this case, 5 culverts are tied for “worst” passability
389
across all scenarios but all culverts can have zero passability under some scenarios (Fig. 3). However, in
390
TNNP, considering the spatial arrangement of barriers within the watershed resulted in a consistent
391
prioritization for restoration (barrier ‘al’, Table 3) in virtually all scenarios examined. If a barrier in a key
392
location is severe enough, any assessment will conclude the same thing: that the barrier is impassable
393
394
under all conditions and the watershed connectivity may be heavily influenced by it.
395
Further work/management advice
17
396
While a useful tool, FishXing, was not without issues and limitations. As others have noted (Blank et al.
397
2005; Poplar-Jeffers et al. 2008; Mangin et al. 2010; R. Gubernick pers. comm.), FishXing uses
398
conservative modelling which does not account for all variables and, as with any model, must be used
399
with caution. Though we were able to improve the results provided by the software with field calibration,
400
it was still difficult to simulate passage for some culverts. Furthermore, there is limited behavioural
401
information available on how fish swim through culverts (e.g., to what extent they swim in the reduced
402
flow of the boundary layers) and whether they exhibit avoidance of these structures; Kemp et al. 2005;
403
Kemp and Williams 2008; Kemp et al. 2008).
404
An examination of barrier properties across TNNP suggests some modifications to the
405
preliminary screening process, based on physical characteristics of the culverts and the degree to which
406
passability was compromised based on our assessments with FishXing. For the field screening method
407
used for brook trout and salmon in our study area, we recommend altering both the maximum outflow
408
drop height and slope in the evaluation flowchart (Fig. 1). Based on both simulations using FishXing and
409
confirmed with the field data collected on multiple dates at the same site, we observed that outflow drops
410
for partial and non-barriers were significantly lower than for full barriers. Therefore, the maximum
411
outflow drop height could be changed from 30cm to 40cm (for 15cm salmonids) to compensate for the
412
potential fluctuation in drops with discharge. For field assessments, we also recommend that the slope
413
used to automatically designate a barrier as impassable be increased from 1.5 to 4.0%, based on the
414
FishXing results discussed above. Though this is steeper than most culvert assessment guides
415
recommend, the further evaluation of culverts using FishXing would be expected to identify barriers that
416
were missed by the initial field assessment. Finally, we recommend caution when determining if culverts
417
are backwatered as some culverts appeared to be passable at low flows, but were actually barriers at
418
higher discharges. Drop height and slope have been shown to be the limiting factors for juvenile fish in a
419
field experiment (Doehring et al. 2011), so we believe these parameters should be the primary focus.
420
When considering modifications to culvert structure to enhance restoration, it should be noted
421
that the type of barrier (velocity, depth or jump) varies based on discharge rates. If fish
18
422
migration/dispersal periods coincide with periods of high or low flow, than culvert modifications should
423
be prioritized to address the main barrier type. For example, at low flow rates, most culverts in TNNP
424
were depth barriers for adult/50cm salmon. Since periods of low stream flow coincide with salmon
425
migration, then modifications should aim to increase water depth within the culvert. Conversely, for
426
brook trout, most barriers at high flow rates (and some barriers for salmon) are velocity barriers, thus
427
modifications should be carried out to reduce water velocity in culverts (for example though the use of
428
flow baffles). These modifications are applicable to the system in Terra Nova National Park; similar
429
modifications to a flowchart based assessment for systems in other parts of the world would have to be
430
based on in situ assessments of local condition and species. However, our findings illustrate that coupling
431
field-based assessments with modelling can help to customize the field-based assessments to better assess
432
433
culvert passability.
434 Conclusion
435
Passability has long been acknowledged to be dynamic and specific to species physiology and
436
morphometry and environmental conditions. Our results here illustrate the importance of making
437
decisions on ecological and hydrological criteria when determining barrier passability, including the
438
errors associated with selecting target species and sizes. In our system, static models, while simpler to
439
implement, do not provide as clear a picture as dynamic models. We have shown that inter- and intra-
440
species variation affects passabilities for individual culverts, and hence for estimates of watershed
441
connectivity. Thus, future assessments of stream connectivity should attempt to be as comprehensive as
442
possible and integrate data that captures the inherent variability in both the fish community and the stream
443
444
properties.
445 Acknowledgements
446
Thanks to S. Peake for providing swim speed data and to R. Gubernick for insights into FishXing models.
447
Also thanks to staff at Terra Nova National Park for assistance with collection of field data. S. Mahlum,
19
448
P. Spaak and 3 anonymous reviewers provided helpful comments on an earlier draft of the manuscript.
449
This research was funded by Parks Canada and by grants for the Canadian Foundation for Innovation and
450
451
452
the Natural Science and Engineering Research Council grants to YFW.
20
453
References
454
Belford DA, Gould, WR (1989) An evaluation of trout passage through six highway culverts in Montana.
455
N Am J Fish Manage 9:437-445
456
Bjornn TC, Peery CA (1992) A review of literature related to movements of adult salmon and steelhead
457
past dams and through reservoirs in the Lower Snake River. Technical Report 92-1, US Fish and
458
Wildlife Service, Idaho Cooperative Fish and Wildlife Research Unit, University of Idaho,
459
Moscow.
460
Blank M, Cahoon J, Burford D, McMahon T, Stein O (2005) Studies of fish passage through culverts in
461
Montana. Conference Proceedings of the 2005 International Conference on Ecology and
462
Transportation held in San Diego, CA. pp 647-661
463
Brater EF, King HW (1976) Handbook of hydraulics for the solution of hydraulic engineering problems
464
(6 ed.) McGraw-Hill, New York
465
Cahoon JE, McMahon T, Solcz A, Blank M, Stein O (2007) Fish passage in Montana culverts: Phase II -
466
passage goals. Helena, MT: The State of Montana Department of Transportation, 64 pp.
467
Clarkin K, Connor A, Furniss MJ, Gubernick B, Love M, Moynan K, WilsonMusser S (2005) National
468
inventory and assessment procedure – for identifying barriers to aquatic organism passage at
469
road-stream crossings (Report no. 7700) San Dimas, CA: United States Department of
470
Agriculture Forest Service
471
Cote D (2005) A compilation of fisheries research in ponds of Terra Nova National Park 1997-2003.
472
Parks Canada technical reports in ecosystem science. Tech Rep Eco Sci 46: 116pp. ii
473
Cote D (2007) Measurements of salmonid population performance in relation to habitat in eastern
474
Newfoundland streams. J Fish Biol 64:1134-1147
475
Cote D, Kehler DG, Bourne C, Wiersma YF (2009) A new measure of longitudinal connectivity for
476
stream networks. Landsc Ecol 24:101-113
477
Cote D, Adams BK, Clarke KD, Langdon M (in press) Salmonid habitat and biomass relationships in
478
small lakes. Environ Biol Fish DOI: 10.1007/s10641-011-9845-8
21
479
Doehring K, Young RG, McIntosh AR (2011) Factors affecting juvenile galxiid fish passage at culverts.
480
Mar Freshw Res 62:38-45
481
Dunham JB, Vinyard GL, Rieman BE (1997) Habitat fragmentation and extinction risk of Lahontan
482
cutthroat trout. N Am J Fish Manage 17:1126-1133
483
Eby LA, Fagan WF, Minckley WL (2003) Variability and dynamics of a desert stream community. Ecol
484
Appl 13:1566-1579
485
Fullerton AH, Burnett KM, Steel EA, Flitcroft RL, Press GR, Feist BE, Torgersen CE, Miller DJ,
486
Sanderson BL (2010) Hydrological connectivity for riverine fish: measurement challenges and
487
research opportunities. Freshw Biol 55:2215-2237
488
Gibson R, Haedrich R, Wernerheim C (2005) Loss of fish habitat as a consequence of
489
inappropriately constructed stream crossings. Fisheries 30:10-17
490
Gillette DP, Tiemann JS, Edds DR, Wildhaber ML (2005) Spatiotemporal patterns of fish
491
assemblage structure in a river impounded by low-head dams. Copeia 2005:539-549
492
Hall CJ, Jordaan A, Frisk MG (2011) The historic influence of dams on diadromous fish habitat with a
493
focus on river herring and hydrologic longitudinal connectivity. Landsc Ecol 26:95-107
494
Helfrich LA, Liston C, Hiebert S, Albers M, Frazer K (1999) Influence of low-head diversion dams on
495
fish passage, community composition, and abundance in the Yellowstone River, Montana. Rivers
496
7:21-32
497
Jones JA, Swanson FJ, Wemple BC, Snyder KU (2000) Effects of roads on hydrology, geomorphology,
498
and disturbance patches in stream networks. Conserv Biol 14:76-85
499
Kemp PS, Gessel MH, Williams JG (2005) Seaward migrating subyearling Chinook salmon avoid
500
overhead cover. J Fish Biol 67: 381-1391
501
Kemp PS, Williams JG (2008) Response of migrating Chinook salmon (Oncorhynchus tshawytscha)
502
smolts to in-stream structure associated with culverts. River Res Appl 24:571-579
22
503
Kemp PS, Gessel MH, Williams JG (2008) Response of downstream migrant juvenile Pacific salmonids
504
to accelerating flow and overhead cover. Hydrobiologia 609:205-207
505
Kemp PS, O’Hanley JR (2010) Procedures for evaluating and prioritising the removal of fish passage
506
barriers: a synthesis. Fish Manag Ecol 17:297-322
507
Khan B, Colbo MH (2008) The impact of physical disturbance on stream communities: lessons from road
508
culverts. Hydrobiologia 600:229-235
509
Kibler KM, Tullos DD, Kondolf GM (2010) Learning from dam removal monitoring: challenges to
510
selecting experimental design and establishing significance of outcomes. River Res Appl DOI:
511
10.1002/rra.1415
512
Mangin SF, Stipetich M, Tritico HM (2010) Reducing the error associated with Manning’s roughness in
513
culvert design for improved fish passage. World Environment and Water Resources Congress
514
2010: Challenge of Change – Proceedings of the World Environmental and Water Resources
515
Congress, 2010. pp. 1540-1553
516
McLaughlin RL, Porto L, Noakes DLG, Baylis JR, Carl LM, Dodd HR, Goldstein JD, Hayes DB, Randall
517
RG (2006) Effects of low-head barriers on stream fishes: taxonomic affiliations and
518
morphological correlates of sensitive species. Can J Fish Aquat Sci 63:766-779
519
Meizler MS, Bain MB, Walter MT (2009) Predicting barrier passage and habitat suitability for migratory
520
fish species. Ecol Model 220:2782-2791
521
Morita K, Yamamoto S (2002) Effects of habitat fragmentation by damming on the persistence of stream-
522
dwelling char populations. Conserv Biol 16:1318-1323
523
Neraas LP, Spruell P. 2001. Fragmentation of riverine systems: the genetic effects of dams on bull trout
524
(Salvelinus confluentus) in the Clark Fork River system. Mol Ecol 10:1153-1164
525
Padgham M, Webb JA (2010) Multiple structural modifications to dendritic ecological networks produce
526
simple responses. Ecol Model 221:2537-2545
23
527
Park D, Sullivan M, Bayne E, Scrimgeour G (2008) Landscape-level stream fragmentation caused by
528
hanging culverts along roads in Alberta’s boreal forest. Can J Forest Res 38:566-575
529 Peake S, Farrel AP (2006) Fatigue is a behavioral response in respirometer‐confined smallmouth bass. J
530
Fish Biol 68:1742-1755
531
Peake S, McKinley RS, Scruton DA (1997) Swimming performance of various freshwater Newfoundland
532
salmonids relative to habitat selection and fishway design. J Fish Biol 51:710-723
533
Poplar-Jeffers IO, Petty JT, Anderson JT, Kite SJ, Strager MP, Fortney RH (2008) Culvert replacement
534
and stream habitat restoration: Implications from brook trout management in an Appalachian
535
watershed, U.S.A. Restoration Ecol 17:404-413
536
Porto MM, McLaughlin RL, Noakes DLG (1999) Low-head barrier dams restrict the movements of fishes
537
in two Lake Ontario streams. N Am J Fisheries Manage 19:1028-1036
538
Potter AJ (1989) The movements of Brook trout (Salvelinus fontinalis) on a small Newfoundland stream
539
system. B.Sc. Thesis. Memorial University of Newfoundland. St. John’s, Newfoundland. 65pp.
540
Pringle C (2003) What is hydrolgic connectivity and why is it ecologically important? Hydrol Proces
541
17:2685-2689
542
Rolls RJ (2011) The role of life-history and location of barriers to migration in the spatial distribution and
543
conservation of fish assemblages in a coastal river system. Biol Conserv 144:339-349
544
Scott WB, Crossman EJ (1973) Freshwater fishes of Canada. Bull Fish Res Board Can 184, 966p.
545
Sheldon AL (1988) Conservation of stream fishes: patterns of diversity, rarity and risk. Conserv
546
Biol 2:149-156
547
Steig TW, Skalski JR, Ransom BH (2005) Comparison of acoustic and PIT tagged juvenile Chinook,
548
steelhead and sockeye salmon (Oncorhynchus, spp.) passing dams on the Columbia River, USA.
549
In: Spedicato MT, Lembo G, Marmulla G (eds) Aquatic telemetry: advances and applications.
550
proceedings of the fifth conference on fish telemetry. 9–13 June 2003. Ustica: FAO/COISPA,
551
295 pp.
24
552
Straub LG, Morris HM (1950a) Hydraulic tests on concrete culverts (Technical paper no. 4, series B.)
553
University of Minnesota, Minneapolis
554
Straub LG, Morris HM (1950b). Hydraulic tests on corregated metal culvert pipes (Technical paper no. 5,
555
series B.) University of Minnesota, Minneapolis
556
USFS (USForest Service) (2003) FishXing 3.0, Available at www. stream.fs.fed.us/fishxing/
557
558
Wiens JA (2002) Riverine landscapes: taking landscape ecology into the water. Freshwat Biol 47:501-515
25
559
560
Figure Captions
561
Fig. 1 Flowchart for preliminary culvert evaluation based on criteria for 15 cm salmonid, (adapted from
562
Clarkin et al. 2005). The flowchart has been used as a rapid assessment strategy in one-time field visits to
563
564
culverts to assess whether they are passable, impassable or partially passable barriers.
565
Fig. 2 The Big Brook watershed of Terra Nova National Park, Newfoundland, Canada. Streams are and
566
waterbodies are shown in dark grey and roads are dashed light grey lines. Anthropogenic barriers are
567
568
indexed by letters and hexagons. The waterfall (diamond) is a complete natural barrier.
569
Fig. 3 The mean (solid bar) and range (empty rectangle) of passability values across all simulations of
570
fish length/swim speed and stream flow for barriers in Big Brook. Barrier labels are shown on the x-axis
571
572
and match those on Figure 2. Passability ranges from 0 (full barrier) to 1 (fully passable) on the y-axis.
573
Fig. 4 Mean and range of connectivity measured for the potadromous case; DCI
P
(top panels) and the
574
diadromous case; DCI
D
(bottom panels) for different scenarios of fish swim speed (indicated by groups of
575
3 points per length class) and stream flow (indicated by error bars). Left hand panels are for brook trout
576
and right hand panels for salmon. Star symbol indicates weighted mean DCI for each case based on
577
578
length-frequency data for fish sampled from the Big Brook population.
579
Fig. 5 Comparison of connectivity measured using the DCI for potamodromous (top panels) and
580
diadromous (bottom panels) in catchments in Terra Nova National Park, Newfoundland and Labrador,
581
Canada. DCI values are calculated when passability estimates are obtained via computer modelling with
582
FishXing (left hand panels) versus field evaluations (right hand panels) of culverts based on 15cm brook
583
584
26
tro
ut
dur
ing
the
mi
gra
tio
n
per
iod.
27
585
Fig. 6 Comparison of variability in DCI as a result of using different seasons, species and methods to
586
estimate passability. Figure shows the number of catchments containing culverts (n = 15) in Terra Nova
587
National Park, Newfoundland and Labrador, Canada with very low (0-25), low (26-50), moderate (51-75)
588
and high (76-100) connectivity measured using the DCI in the a. potamodromous case and b. diadromous
589
case. DCI values are based on calculating passability with FishXing during fish migration period for
590
brook trout and salmon, and with FishXing across the entire year (salmon only) as well as based on a
591
592
593
594
rapid assessment of passability using only the simplified field-based method.
