Ecological Modelling

Ecological Modelling 220 (2009) 2782–2791 Contents lists available at ScienceDirect Ecological Modelling journal homepage: www.elsevier.com/locate/e...
Author: Sharon Francis
3 downloads 0 Views 1MB Size
Ecological Modelling 220 (2009) 2782–2791

Contents lists available at ScienceDirect

Ecological Modelling journal homepage: www.elsevier.com/locate/ecolmodel

Predicting barrier passage and habitat suitability for migratory fish species Marcia S. Meixler a,∗ , Mark B. Bain a , M. Todd Walter b a b

Department of Natural Resources, Cornell University, 104 Rice Hall, Ithaca, NY 14853, USA Department of Biological and Environmental Engineering, 222 Riley-Robb Hall, Cornell University, Ithaca, NY 14853 USA

a r t i c l e

i n f o

Article history: Received 28 January 2009 Received in revised form 17 July 2009 Accepted 21 July 2009 Available online 21 August 2009 Keywords: Barriers Migratory fish passage Habitat suitability Geographical information systems Spatial modeling Freshwater fishes

a b s t r a c t Fish migrate to spawn, feed, seek refuge from predators, and escape harmful environmental conditions. The success of upstream migration is limited by the presence of barriers that can impede the passage of fish. We used a spatially explicit modeling strategy to examine the effects of barriers on passage for 21 native and non-native migratory fish species and the amount of suitable habitat blocked for each species. Spatially derived physical parameter estimates and literature based fish capabilities and tolerances were used to predict fish passage success and habitat suitability. Both the fish passage and the habitat suitability models accurately predicted fish presence above barriers for most common, non-stocked species. The fish passage model predicted that barriers greater than or equal to 6 m block all migratory species. Chinook salmon (Oncorhynchus tshawytscha) was expected to be blocked the least. The habitat suitability model predicted that low gradient streams with intact habitat quality were likely to support the highest number of fish species. The fish passage and habitat suitability models were intended to be used by environmental managers as strategy development tools to prioritize candidate dams for field assessment and make decisions regarding the management of migratory fish populations. © 2009 Elsevier B.V. All rights reserved.

1. Introduction Approximately 76,000 dams exist in the United States of America (Larinier, 2000). Dams cause fragmentation of river habitat and obstruct the dispersal and migration of fish (Morita and Yamamoto, 2002; Nilsson et al., 2005; Fukushima et al., 2007). Fish migrate to spawn, feed, reach rearing areas, and seek refuge from predators or harmful environmental conditions such as freeze-up of a lake or stream (Katopodis, 1989; Gallagher, 1999). Disruption of migration patterns can lead to injury of fish (Larinier, 2000) and the decline or even extinction of populations (Arthington and Welcomme, 1995; Larinier, 2000; Penczak and Kruk, 2000; Gehrke et al., 2002; Masters et al., 2006). Thus, several recent calls have emphasized the need for additional scientific studies to address the impact of barriers on fish passage success (Larinier, 2000; Harford and McLaughlin, 2007). Barriers to migration occur through the presence of natural or artificial obstacles, the type of topography, and the timing and magnitude of precipitation and stream flows (Bartson, 1997). Fish passage success is thus based on physical parameters (height of the barrier), hydraulic conditions (velocity, plunge pool depth) and the darting and jumping abilities of fish species during migration

∗ Corresponding author. Tel.: +1 607 227 8304; fax: +1 607 255 0238. E-mail address: [email protected] (M.S. Meixler). 0304-3800/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolmodel.2009.07.014

(Bartson, 1997; Larinier, 2000; Reiser et al., 2006). Some obstructions may be permanently insurmountable for all fish species but others may be passable by some fish species at certain times of the year under favorable hydrological conditions (Larinier, 2000). Recent studies evaluated fish passage success for individual fish species in the field (Morita and Yamamoto, 2002; Ovidio and Philippart, 2002; Holthe et al., 2005; Kondratieff and Myrick, 2006; Reiser et al., 2006) and using models (Jager et al., 2001; Labonne and Gaudin, 2006; Sheer and Steel, 2006; Baigun et al., 2007; Fukushima et al., 2007). In addition, several online models have been developed for this purpose including the Fish Passage Decision Support System (http://fpdss.fws.gov/) which evaluates the length of stream open to passage upon removal of a barrier, FishXing (http://www.stream.fs.fed.us/fishxing/) which focuses on fish passage through culverts, and CriSP (http://www.cbr.washington.edu/crisp/crisp.html), a detailed fish passage model for the Columbia River. However, none of the previous studies or models of fish passage success used geographic information systems (GIS) modeling to spatially assess the ability of migratory fish species to pass a barrier. GIS models have many advantages such as the ability to rapidly and quantitatively assess landscape scale issues over varying temporal and spatial scales. Thus, GIS models of fish passage success are important and valuable since field testing of all barriers under all possible conditions for all species is time and labor intensive (Ovidio and Philippart, 2002).

M.S. Meixler et al. / Ecological Modelling 220 (2009) 2782–2791

2783

Fig. 1. Study locations for pool depth and velocity testing (Rouge River, Ontario, Canada) and fish passage and habitat suitability modeling and testing (Sandy Creek to Little Salmon River, eastern Lake Ontario, New York, USA).

Recent work has indicated that barrier assessment should consider not just fish passage success but also upstream habitat quality (Sheer and Steel, 2006). To date, few studies have attempted a comprehensive, regional perspective of the distribution and types of habitat loss due to barriers. An understanding of the location of barriers, the effect of barriers on migratory fish passage, size of the area blocked, and impact on access to suitable habitat are important for prioritizing restoration activities (Roni et al., 2002). In this paper, we use two geospatial models to (1) examine the effects of natural and artificial barriers on native and nonnative migratory fish species passage and (2) assess the amount of suitable habitat blocked for each species. Combined results from these models allow us to predict which barriers block passage, the combined length of the streams blocked, and the amount of suitable habitat lost to each migratory fish species of interest. We present these strategy development tools to assist environmental managers locate candidate dams for field assessment and make decisions regarding the management of migratory fish populations.

2. Methods 2.1. Study site The study area includes eleven tributaries to eastern Lake Ontario from Sandy Creek south to the Little Salmon River, New York (Fig. 1) between 43◦ 46 57 N, 76◦ 13 12 W and 43◦ 20 27 N, 76◦ 00 14 W. We chose this area for its abundance of natural and artificial barriers, variation in land uses, and availability of GIS and field data. The combined watersheds have an area of 2009 km2 and are largely composed of deciduous and evergreen forest (60%), agriculture (21%), lakes and wetlands (18%) and small rural settlements (1%). Elevations range from 580 m at the western edge of the Adirondacks to 75 m at Lake Ontario. Fifty-five dams and six waterfalls exist in the study area ranging from Vs ), dive deep enough in the plunge pool to achieve maximum jump height (Pd > 2L), and jump high enough to clear the barrier (JH > H). If a species scored positively for all three, the structure was not considered a barrier to that species. We repeated this process for all 21 modeled species and all barriers. All reaches upstream from a passable barrier were marked as unblocked until another barrier or the headwaters were reached. In this way, we summed the total length of stream blocked for each modeled fish species.

where v is the seasonal stream velocity and q is the seasonal average discharge. In this way, we were able to calculate ratios of seasonal velocity to bankfull velocity for the USGS gaging stations in the study area. An analysis of bankfull and average annual discharge data (Dunne and Leopold, 1978) revealed that drainage basins less than 12,950 ha had substantially greater differences between average annual and bankfull discharge than those over 12,950 ha. Therefore, we used 12,950 ha as the cutoff between small and large basins. We extended this concept for use with bankfull velocity in each migratory season, Vs . Thus, the final seasonal velocity equation is Vs (m/s) = (0.380A0.148 )X D where X is the ratio of seasonal velocity to bankfull velocity for small or large basins as referenced above (spring, small: 0.08 of bankfull, spring, large: 0.21, fall, small: 0.05, fall, large: 0.16, summer, small: 0.04, summer, large: 0.09). Each species has different darting and jumping abilities so that a barrier for one species may be passable by others. We used maximum darting speed to determine whether a fish species could overcome the velocity of the water during its migratory season to reach the barrier and we used maximum jumping height to deter-

DS (m/s) = 9L (9L)2 JH (m) = 2g

2786

M.S. Meixler et al. / Ecological Modelling 220 (2009) 2782–2791

Fig. 3. Algorithm for the habitat suitability prediction model to determine habitat supporting each species of interest. Dotted boxes indicate spatial data inputs.

2.3. Habitat suitability prediction model Knowing whether a migratory fish species is capable of passing a barrier is only part of the information needed by environmental managers. Combining fish passage information with habitat suitability predictions answers the question: if a fish species could pass this barrier, would it find any suitable habitat upstream? We modeled habitat suitability by classifying reaches into habitat types, segregating fish species according to preferred habitat type, and then associating fish species with the appropriate reaches. We first used an AML program (algorithm displayed in Fig. 3) in ArcGIS 9.0 to obtain stream gradient and percent catchment land use for each reach. To calculate stream gradient, we obtained elevations for the furthest upstream and downstream points of each reach on the 1-arc second National Elevation Dataset and divided by the stream length between the two endpoints. We calculated percent urban-residential, agriculture, forest, water, wetland, and barren-transitional land from National Oceanic and Atmospheric Administration Coastal Change Analysis Program 2001 data for the catchments of each reach. We then undertook a literature review on the habits of fish to determine the gradient preferences (low (L), high (H), both) and habitat degradation tolerances (intolerant (I), moderately tolerant (M), highly tolerant (T)) of each species (Table 2). Classification was accomplished using information on gradient from NatureServe (2008) and on habitat degradation from a variety of state fish reference books (Pflieger et al., 1975; Lee et al., 1980; Trautman, 1981; Becker, 1983; Smith, 1985; Hocutt and Wiley, 1986; Robison and Buchanan, 1988; Etnier and Starnes, 1993; Jenkins and Burkhead, 1994) and FishBase (2008). We classified reaches with habitat supporting the preferences and tolerances of each fish species as suitable habitat. Thus, high gradient reaches, those with slopes greater than 1.5 m/km (Meixler, 2000), were classified as suitable for fish species with a preference for high gradient habitats and vice versa. Both high and low gradient habitats were classified as suitable for fish species with preferences for both gradients. Highly degraded habitats, those with >30% urban-residential (Morgan and Cushman, 2005) or >50%

agricultural area (Roth et al., 1996) in the catchment of the reach, were classified as only supporting tolerant fish species. Moderately degraded habitats, those with 10–30% urban-residential (Klein, 1979) or 40–50% agricultural area (Genito et al., 2002), were classified as supporting both moderately tolerant and highly tolerant species. Intact habitats were expected to support species of all tolerances. We summed the number of fish species in each habitat type to identify the relative diversity of all habitat types. Using an ArcGIS 9.0 AML, we calculated for each species the cumulative suitable habitat (km) above each barrier and the total percent suitable habitat behind all blocked dams. 2.4. Model validation We tested the accuracy of the plunge pool depth and velocity calculations to assess the validity of substituting the height of the structure in place of the total drop in head and the overall reliance of the model on these physical parameters. We used the Rouge River Watershed in Ontario, Canada for validity testing due to the ready availability of previously collected plunge pool depth and velocity field data. Field collected plunge pool depth (m) and barrier height (m) measurements were collected in 67 locations between June and November 2005 and July and October 2006. Similarly, velocity (m/s) field measurements were taken in 87 locations (76 in small drainage basins ( 0.05). Abbreviations are as follows: I = intolerant, M = moderately tolerant, T = highly tolerant; L = low gradient, M = medium gradient, H = high gradient. Scientific name

Habitat degradation tolerance

Gradient preference

Model success (species predicted and observed)

Model success (species not predicted or observed)

Model failure (species predicted but not observed)

Model failure (species observed but not predicted)

Fisher’s exact test p-value

Suitable habitat behind blockages (%)

Alewife Atlantic salmon Brook trout Brown trout Chinook salmon Coho salmon Creek chubsucker Fantail darter Gizzard shad Iowa darter Johnny darter Northern hog sucker Northern pike Shorthead redhorse Smallmouth bass Spottail shiner Steelhead Walleye White perch White sucker Yellow perch

Alosa pseudoharengus Salmo salar Salvelinus fontinalis Salmo trutta Oncorhynchus tshawytscha Oncorhynchus kisutch Erimyzon oblongus Etheostoma flabellare Dorosoma cepedianum Etheostoma exile Etheostoma nigrum Hypentelium nigricans Esox lucius Moxostoma macrolepidotum Micropterus dolomieu Notropis hudsonius Oncorhynchus mykiss Sander vitreus Morone americana Catostomus commersonii Perca flavescens

M M I I M M T M T I M I M I M I I T M T M

LM HLM HM HLM LM LM M HM LM LM M HM L M LM HLM HLM HLM L HLM L

17 3 35 24 3 3 0 67 1 0 0 15 50 5 16 17 35 24 15 115 103

197 20 87 86 188 190 195 14 194 185 195 84 160 201 161 93 83 0 189 0 130

115 307 202 213 129 129 134 243 133 129 132 222 82 124 116 220 202 306 117 215 29

1 0 6 7 10 8 1 6 2 16 3 9 38 0 37 0 10 0 9 0 68

0.00

4 28 67 38 3 4 3 4 4 3 4 65 2 3 3 69 38 63 4 79 4

a b

a

0.04 0.54 0.26 0.54 a

0.41 a

0.00 0.28 0.35 0.00 0.01 0.13 0.00 0.38 b

0.03 b 0.00

M.S. Meixler et al. / Ecological Modelling 220 (2009) 2782–2791

Common name

Fisher’s exact test invalid due to cells with expected counts less than 5. Could not compute Fisher’s exact test since species predicted to be present in all habitats.

2787

2788

M.S. Meixler et al. / Ecological Modelling 220 (2009) 2782–2791

Fig. 4. The possible forms of success and failure of the fish passage prediction model when compared with field collected data.

teen barriers in the eastern Lake Ontario study region. Each location was sampled one or more times between 1979 and 1999 with each site visited an average of eight times and 97% of the sites receiving at least two visits. We aggregated multiple surveys at individual locations to arrive at overall species presence data for each site. We then summed the total number of successes and failures for each migratory fish species (Fig. 4). Two forms of success were possible: (1) a fish species was expected and observed above a barrier and (2) a species was not expected and not observed above a barrier. Likewise, two forms of failure were possible: (1) a species was expected but not observed above a barrier and (2) a species was not expected but was observed above a barrier. The number of successes and failures were compared using a one-tailed sign test (Graphpad Software Inc., San Diego, CA, USA, 2005) to assess if the predicted and observed outcomes had equal probabilities. Because it is reasonable for true performance to be above chance but not to be below it, the use of one-tailed rather than two-tailed tests would appear to be appropriate. Paired differences were assumed to be independent. We also tested modeled fish species presence based on habitat preferences and tolerances against field collected fish presence data obtained from the NYSDEC Bureau of Fisheries. Fish data were collected in 330 locations in the study area. We were able to include more locations than those used in the fish passage validation since we were not restricted to only those locations above dams in this test. Each location was sampled one or more times between 1979 and 2000 with 97% of the sites receiving multiple site visits and an average of eight visits per location. Again, we aggregated multiple surveys at individual locations to arrive at overall species presence data for each site. We calculated gradient and habitat degradation values from GIS data for each survey location. Fish habitat degradation tolerances and gradient preferences were used to link fish to habitat and classify the suitability of each survey site for each species. Predicted and observed presence data were tested for independence using the Fisher’s exact test in Minitab (Minitab Release 15.1.0.0, Minitab Inc., State College, PA, USA, 2007). 3. Results 3.1. Model accuracy Field collected pool depths ranged from 0.02 to 1.5 m with a mean of 0.34 m. Mean differences between paired field collected and modeled plunge pool depths were small (0.03 m) with a 95% confidence interval of −0.09 to 0.03 m. Thus, the mean difference includes zero and indicates no meaningful difference between predicted and observed values. Field collected velocities ranged from 0.003 to 0.38 m/s for drainage basins less than 12,950 ha and 0.07–0.25 m/s for those greater than 12,950 ha. Means for the two groups were 0.06 m/s and 0.15 m/s, respectively. Mean differences between paired field collected and modeled velocities in the fall for drainage basins less than 12,950 ha were small (0.01 m/s) with a 95% confidence interval of 0–0.02 m/s. Again, the mean difference includes zero barely providing weak evidence that predicted and

observed values are equivalent overall. Mean differences between paired field collected and modeled velocities in the fall for large basins (>12,950 ha) were slightly larger (0.11 m/s) with a 95% confidence interval of 0.08–0.15 m/s. For large basins, predicted values overestimate velocities although the magnitude of this error is small and can be corrected in the model. Fish passage validation indicated that the probability of equally obtaining a successful fish passage prediction, defined as a match between predicted and observed fish presence, or a failure was significantly small (P ≤ 0.05) for 14 of the 21 species (Table 3) indicating good agreement between observed and predicted fish presence above barriers for these species. Observed and predicted match failures occurred for several abundant species and some stocked species. Match failures were the result of the model predicting presence of a species in a location where it was not observed. The test of habitat suitability indicated that fish presence was the same whether observed in the field or predicted using our models for eight species (P > 0.05; Table 2). The Fisher’s exact test could not be calculated for Atlantic salmon (Salmo salar), creek chubsucker (Erimyzon oblongus), and gizzard shad (Dorosoma cepedianum) due to lack of sufficient data. We predicted that walleye (Sander vitreus) and white sucker (Catostomus commersonii) would be found in all habitats so again the Fisher’s exact test could not be calculated for these species. 3.2. Model results We predicted that all migratory fish species would be blocked by at least one barrier (Fig. 5). We expected chinook salmon (Oncorhynchus tshawytscha) to be blocked from only 47% of the streams in the watershed and the remaining species to be blocked from 74% to 80% of the streams (Table 3). We predicted all 61 barriers in the study area to block one or more migratory fish species. Not surprisingly, low barriers blocked the fewest species but even the smallest barrier at 0.3 m in height, from base to top, still blocked the four small-bodied fish species mentioned above and creek chubsucker, white perch (Morone americana) and yellow perch (Perca flavescens). The model predicted blockage of all species by barriers greater than or equal to 6 m. The most notable finding from the habitat suitability prediction model was that low gradient streams with intact habitat quality were expected to have the highest fish species richness and high gradient, highly degraded streams the lowest expected richness. The largest decline in predicted fish species occurrence followed a shift to high gradient habitat with an expected loss of 67% of the species. This is due to the fact that there were no migratory fish in the study area specific to high gradient habitats. Degradation of habitat quality caused a smaller decrease in fish species richness with an expected loss of 52% of the species. Most fish species were predicted to lose only 2–4% of their habitat, but eight species showed losses from 28% to 79% (Table 2). We predicted six of the 61 barriers in the study region to have no upstream habitat supportive of any migratory fish species. However, twelve other barriers, if removed, were expected to have upstream habitat able to support all fish. 4. Discussion This study was aimed at modeling the effects of natural and artificial barriers on migratory fish species passage and the amount of suitable habitat blocked for each species. Results of our fish passage and habitat suitability models appear encouraging for a large number of common species in the eastern Lake Ontario watersheds.

M.S. Meixler et al. / Ecological Modelling 220 (2009) 2782–2791

2789

Table 3 Success of the fish passage model in predicting species presence above nineteen barriers in the eastern Lake Ontario study area. The sign test assesses the probability of equally obtaining a success, where predicted fish presence matches observed presence, or failure. Bolded values indicate better than random agreement between predicted and observed fish presence above barriers (p < 0.05). Common name

Scientific name

Sign test p-value

Successes (out of 19)

Failures (out of 19)

Ale wife Atlantic salmon Brook trout Brown trout Chinook salmon Coho salmon Creek chubsucker Fantail darter Gizzard shad Iowa darter Johnny darter Northern hog sucker Northern pike Shorthead redhorse Smallmouth bass Spottail shiner Steelhead Walleye White perch White sucker Yellow perch

Alosa pseudoharengus Salmo salar Salvelinus fontinalis Salmo trutta Oncorhynchus tshawytscha Oncorhynchus kisutch Erimyzon oblongus Etheostoma flabellare Dorosoma cepedianum Etheostoma exile Etheostoma nigrum Hypentelium nigricans Esox lucius Moxostoma macrolepidotum Micropterus dolomieu Notropis hudsonius Oncorhynchus mykiss Sander vitreus Morone americana Catostomus commersonii Perca flavescens

0.000 0.032 0.180 0.180 0.500 0.010 0.000 0.532 0.000 0.000 0.000 0.010 0.000 0.000 0.084 0.002 0.500 0.000 0.000 0.510 0.010

19 14 12 12 9 15 18 5 18 18 18 15 17 19 13 16 9 18 18 4 15

0 5 7 7 10 4 1 14 1 1 1 4 2 0 6 3 10 1 1 15 4

Fish passage model results indicated that all migratory fish in the eastern Lake Ontario watersheds are likely to be blocked by at least one barrier and that all barriers block at least one species of migratory fish. In general, smaller species such as shiners and darters (0.04–0.09 m total length) experienced the highest degree of blockage (81% of streams) owing to their weaker jumping abilities and darting speeds. We predicted fish of this size would be unable to pass even the smallest barriers in the watershed. Similarly, large species such as Atlantic salmon (0.69 m total length)

Stocked

× × × ×

× ×

Streams predicted to be blocked to passage (%) 78 74 78 78 47 79 80 80 78 80 80 78 79 78 78 80 78 79 80 78 80

may be blocked from passing barriers with shallow plunge pools due to their size, increased disorientation due to turbulence and decreased tail propulsion power (Powers and Orsborn, 1985). The fish passage model successfully predicted presence above barriers for most of the common, non-stocked migratory fish species in the study area. Model failure generally occurred for two groups of fish: abundant species and stocked species. Abundant species such as fantail darter and white sucker were observed in large numbers in the NYSDEC fisheries database both above and

Fig. 5. Number of species blocked by dams and waterfalls and number of species supported in each reach in eastern Lake Ontario watersheds, New York.

2790

M.S. Meixler et al. / Ecological Modelling 220 (2009) 2782–2791

below dams. It is possible that stable populations existed throughout the watershed prior to dam construction and these populations, though fragmented, remained in place after dam completion. These observations are at odds with the model, which predicted an inability for these species to pass barriers based on physical parameters and capabilities. Agreement was also low between predictions and observations of stocked fish such as brook trout (Salvelinus fontinalis), brown trout (Salmo trutta), chinook salmon (O. tshawytscha) and steelhead (Oncorhynchus mykiss). These fish were found in habitats they would not normally have access to given their darting and jumping abilities. Model failure may also have occurred due to: (1) barriers that simultaneously had portions that were passable and not passable and (2) barriers that were passable at certain flows but not at other flows. However, these situations are beyond the scope of the model’s capabilities. Based on these results, we have confidence in the ability of the model to predict presence above barriers only for common, non-stocked species. Further testing could be performed to investigate: (1) the model’s ability to predict passage of typically stocked species in watersheds with no stocking and (2) the impact of a species negotiating a barrier predicted to block that species and the consequences for upstream barrier passage. The small mean difference between paired field collected and modeled plunge pool depths indicates that the known height of the structure can replace the total drop in head in the plunge pool depth calculation. This is important as the height of the structure is often known information but the total drop in head is generally unknown and difficult to model. Likewise, the small mean differences between paired field collected and modeled velocities indicate that the model is predicting realistic velocity estimates during migration. This result demonstrates that the factors used to relate bankfull velocity to velocity during the fall migratory season, derived using data in the United States, were robust enough to produce meaningful results for fall velocities on the Canadian side of Lake Ontario. We note however that the small mean differences between model predictions and field collected observations only allow us to conclude that the model process has low bias and not that model predictions were accurate. Low velocities are difficult to predict precisely. However, the ranges we predicted for the velocities were well below the critical velocities for most fish species (Clay, 1995). Thus, velocity appears to be a less important factor at this scale though future versions of the model could include a factor to adjust for this systematic bias. We built the model using established estimators including velocity as we expected velocity to be a critical factor. We found velocity to be largely unimportant at this scale, however we left it in as it is part of other work that helped us build the model in its current form. Overall, we are likely poor at accurately predicting low velocities, but our predicted ranges were close to those found through empirical observation and velocity seems to only rarely block fish passage in natural channels. Habitat suitability model results indicated that low gradient streams with intact habitat quality were likely to support the highest number of fish species. Intact habitats can support fish both tolerant and intolerant to habitat degradation and thus the highest species richness. Further, all migratory species in the study region preferred low gradients or a combination of low and high gradient habitats. Conversely, only the most tolerant species can live in highly degraded environments and few of the modeled migratory species are both tolerant and likely to be found in high gradient habitats. Thus, highly degraded streams with high gradients were expected to support the fewest number of fish species in the study area. We predicted that white sucker and walleye would be found in all habitats in the study area. This finding is consistent with Smith (1985), which indicates that both species are found in nearly every kind of habitat in New York. We did not use stream size as a

predictor of species presence; therefore these generalist fish were predicted to occupy all sizes of streams. Not surprisingly, these species also have some of the highest predicted suitable habitat behind blockages at 79% and 63%, respectively. The two species predicted to be found in only one habitat type (intact—low gradient), Iowa darter and shorthead redhorse, are fairly rare in the study region (Smith, 1985). These species only have 3% predicted suitable habitat behind blockages. Although other factors may affect the presence of a fish species in a particular stream (e.g. species interactions, water temperature, and point-source pollution) and our model is quite coarse, the habitat suitability model successfully predicted presence above barriers for some species in the study area. However, validation of the habitat suitability model was difficult for several reasons. First, fish may become concentrated in poor habitat when high and moderate quality habitat are lacking or fully occupied resulting in observations in unexpected habitats and rarity in predicted habitats. Second, a different subset of the stream network is available to fish each year due to variations in stream flow and this also contributes to anomalies in the observed data. For these reasons predicted presence data did not match observations for several species. Model failure did not seem to relate to specific fish preferences and tolerances but was more likely the inability of this coarse model to capture the variability in habitat preferences demonstrated by some of the species (e.g. cold water may trump gradient and habitat quality for brook trout). More complicated fish habitat classification models have been created based on available field data for specific locations (Joy and Death, 2004; Fransen et al., 2006) or by using literature at the community level (Goldstein and Meador, 2004). Our attempt was aimed at developing a model that could be broadly and easily applied for a wide variety of migratory fish species. Though not accurate for all attempted species, we believe that the species classifications presented here provide an opportunity for further examination of fish species’ relations to specific physical factors. The substitution or addition of other factors such as size (Goldstein and Meador, 2004) or temperature (Schmutz et al., 2007) or the refinement of biological data (e.g. swimming/jumping ability) might improve model prediction accuracy for specific species or certain environmental conditions. Varying degrees of sampling intensity could affect the validation results for both the fish passage and habitat suitability models. Sites with few sampling visits would be expected to not represent the species composition of the location as well as those with many visits. Thus, some of the disagreement between model predictions and observations may be due to sampling adequacy. Additionally, fragmented populations above barriers may impact validation results. Some species of fish (e.g. white sucker, northern hog sucker) are better able to maintain viable populations above barriers than other species (e.g. shorthead redhorse) despite sharing similar biological characteristics. Such populations when observed in the field may be inconsistent with model predictions of fish species likely to live in the area. Auer (1996) states that environmental managers should give barrier removal or fish passage greater consideration than habitat enhancement for populations currently isolated or restricted in range. Effective fish population management requires the ability to assess current blockage of fish by barriers in order to develop strategies for conservation. However, field research is time and labor intensive and few tools exist to remotely assess barrier passage and habitat suitability for migratory fish. It is clear that the modeling procedures presented here have considerable potential to predict barrier passage and suitable habitat for common, nonstocked species of migratory fish in the eastern watersheds of Lake Ontario. Our models were built using known principles of hydraulic conditions and fish behavior instead of location based empirical data. Further, our models can be tailored to predict fish passage

M.S. Meixler et al. / Ecological Modelling 220 (2009) 2782–2791

and habitat conditions for any migratory fish species of interest. For these reasons, we believe these methods have the potential to provide information on conditions in other locations as well. These models can be used by environmental managers to effectively prioritize barriers, with appropriate field assessment, to determine the extent of blockage and the fish species most impacted. Acknowledgements This study was supported by the Great Lakes Protection Fund under Grant no. 766 and benefited from research performed in the study area under the Natural Science Foundation Grant no. 0083625. We wish to thank Doug Carlson, Gary Bowen and Jason Tam for allowing us to use their data in this study. A great number of people generously contributed their time and effort to this project and we offer our sincere thanks to all of them. References Arthington, A.H., Welcomme, R.L., 1995. The condition of large river systems of the World. In: Armantrout, N.B., Wolotira Jr., R.J. (Eds.), Condition of the World’s Aquatic Habitats. World Fisheries Congress. Science Publishers, Lebanon, NH, USA, pp. 44–75. Auer, N.A., 1996. Importance of habitat and migration to sturgeons with emphasis on lake sturgeon. Can. J. Fish. Aquat. Sci. 53, 152–160. Baigun, C.R.M., Nestler, J.M., Oldani, N.O., Goodwin, R.A., Weber, L.J., 2007. Can North American fish passage tools work for South American migratory fishes? Neotropical Ichthyol. 5, 109–119. Bartson, A., 1997. Smith River Fisheries and Ecosystem Report. Institute for River Ecosystems, Arcata, CA. Becker, G.C., 1983. Fishes of Wisconsin. University of Wisconsin Press, Madison, WI, USA, p. 1052. Clay, C.H., 1995. Design of Fishways and Other Fish Facilities. Lewis Publishers, Boca Raton, FL, USA, p. 248. Dunne, T., Leopold, L.B., 1978. Water in Environmental Planning. W.H. Freeman, San Francisco, CA, USA, 818 pp. Etnier, D.A., Starnes, W.C., 1993. The Fishes of Tennessee. University of Tennessee Press, Knoxville, TN, USA, 681 pp. FishBase, 2008. FishBase Species Profiles. Available: http://www.fishbase.org. (February 2008). Fransen, B.R., Duke, S.D., McWethy, L.G., Walter, J.K., Bilby, R.E., 2006. A logistic regression model for predicting the upstream extent of fish occurrence based on geographical information systems data. N. Am. J. Fish. Manage. 26, 960–975. Fukushima, M., Kameyama, S., Kaneko, M., Nakao, K., Steel, E.A., 2007. Modelling the effects of dams on freshwater fish distributions in Hokkaido, Japan. Freshwat. Biol. 52, 1511–1524. Gallagher, A.S., 1999. Barriers. In: Bain, M.B., Stevenson, N.J. (Eds.), Aquatic Habitat Assessment: Common Methods. American Fisheries Society, Bethesda, MD, USA, pp. 135–147. Gehrke, P.C., Gilligan, D.M., Barwick, M., 2002. Changes in fish communities of the Shoalhaven River 20 years after construction of Tallowa Dam, Australia. River Res. Appl. 18, 265–286. Genito, D., Gburek, W.J., Sharpley, A.N., 2002. Response of stream macro invertebrates to agricultural land cover in a small watershed. J. Freshwat. Ecol. 17, 109–119. Goldstein, R.M., Meador, M.R., 2004. Comparisons of fish species traits from small streams to large rivers. Trans. Am. Fish. Soc. 133, 971–983. Harford, W.J., McLaughlin, R.L., 2007. Understanding uncertainty in the effect of low-head dams on fishes of Great Lakes tributaries. Ecol. Appl. 17, 1783–1796. Hocutt, C.H., Wiley, E.O., 1986. The Zoogeography of North American Freshwater Fishes. Wiley, New York, NY, USA, 866 pp. Holthe, E., Lund, E., Finstad, B., Thorstad, E.B., McKinley, R.S., 2005. A fish selective obstacle to prevent dispersion of an unwanted fish species, based on leaping capabilities. Fish. Manage. Ecol. 12, 143–147. Jager, H.I., Chandler, J.A., Lepla, K.B., Van Winkle, W., 2001. A theoretical study of river fragmentation by dams and its effects on white sturgeon populations. Environ. Biol. Fishes 60, 347–361. Jenkins, R.E., Burkhead, N.M., 1994. Freshwater Fishes of Virginia. American Fisheries Society, Bethesda, MD, USA, 1079 pp. Joy, M.K., Death, R.G., 2004. Predictive modelling and spatial mapping of freshwater fish and decapod assemblages using GIS and neural networks. Freshwat. Biol. 49, 1036–1052.

2791

Katopodis, C., 1989. A Guide to Fishway Design. Freshwater Institute, Dept. of Fisheries and Oceans—Central and Arctic Region, Winnipeg, Manitoba, Canada. Klein, R.D., 1979. Urbanization and stream quality impairment. Water Resour. Bull. 15, 948–963. Kondratieff, M.C., Myrick, C.A., 2006. How high can brook trout jump? A laboratory evaluation of brook trout jumping performance. Trans. Am. Fish. Soc. 135, 361–370. Labonne, J., Gaudin, P., 2006. Modelling population viability in fragmented environments: contribution to the conservation of an endangered percid (Zingel asper). Can. J. Fish. Aquat. Sci. 63, 650–659. Larinier, M., 2000. Dams and Fish Migration. World Commission on Dams, Toulouse, France. Lee, D.S., Gilbert, C.R., Hocutt, C.H., Jenkins, R.E., McAllister, D.E., Stauffer Jr., J.R., 1980. Atlas of North American Freshwater Fishes. Publication of the North Carolina Biological Survey, North Carolina State Museum of Natural History, Raleigh, NC, USA, 854 pp. Masters, J.E.G., Jang, M.H., Ha, K., Bird, P.D., Frear, P.A., Lucas, M.C., 2006. The commercial exploitation of a protected anadromous species, the river lamprey (Lampetra fluviatilis (L.)), in the tidal River Ouse, north-east England. Aquat. Conserv.: Mar. Freshwat. Ecosyst. 16, 77–92. Meixler, M.S., 2000. Predicting aquatic biodiversity with geographic information systems. Master’s thesis, Cornell University, Ithaca, NY, 57 pp. Morgan, R.P., Cushman, S.E., 2005. Urbanization effects on stream fish assemblages in Maryland, USA. J. N. Am. Benthol. Soc. 24, 643–655. Morita, K., Yamamoto, S., 2002. Effects of habitat fragmentation by damming on the persistence of stream-dwelling charr populations. Conserv. Biol. 16, 1318–1323. NatureServe, 2008. NatureServe Explorer: An Online Encyclopedia of Life. Available: www.natureserve.org. (February 2008). Nilsson, C., Reidy, C.A., Dynesius, M., Revenga, C., 2005. Fragmentation and flow regulation of the world’s large river systems. Science 308, 405–408. Ovidio, M., Philippart, J.C., 2002. The impact of small physical obstacles on upstream movements of six species of fish—synthesis of a 5-year telemetry study in the River Meuse basin. Hydrobiologia 483, 55–69. Penczak, T., Kruk, A., 2000. Threatened obligatory riverine fishes in human-modified Polish rivers. Ecol. Freshwat. Fish 9, 109–117. Pflieger, W.L., Sullivan, M., Taylor, L., 1975. The Fishes of Missouri. Missouri Dept. of Conservation, Jefferson City, MO, USA, 343 pp. Powers, P., Orsborn, J., 1985. New concepts in fish ladder design: analysis of barriers to upstream fish migration, volume IV of IV. Investigation of the physical and biological conditions affecting fish passage success at culverts and waterfalls. Project No. 198201400 (BPA Report DOE/BP-36523-1), Bonneville Power Administration, Portland, OR. Reiser, D.W., Peacock, R.T., 1985. A Technique for assessing upstream fish passage problems at small-scale hydropower developments. In: Olson, F.W., White, R.G., Hamre, R.H. (Eds.), Symposium on Small Hydropower and Fisheries. American Fisheries Society, Western Division, Bethesda, MD, USA, pp. 423–432. Reiser, D.W., Huang, C.M., Beck, S., Gagner, M., Jeanes, E., 2006. Defining flow windows for upstream passage of adult anadromous salmonids at cascades and falls. Trans. Am. Fish. Soc. 135, 668–679. Robison, H.W., Buchanan, T.M., 1988. Fishes of Arkansas. University of Arkansas Press, Fayetteville, AR, USA, 536 pp. Roni, P., Beechie, T.J., Bilby, R.E., Leonetti, F.E., Pollock, M.M., Pess, G.R., 2002. A review of stream restoration techniques and a hierarchical strategy for prioritizing restoration in Pacific northwest watersheds. N. Am. J. Fish. Manage. 22, 1–20. Roth, N.E., Allan, J.D., Erickson, D.L., 1996. Landscape influences on stream biotic integrity assessed at multiple spatial scales. Landscape Ecol. 11, 141–156. Schmutz, S., Melcher, A., Frangez, C., Haidvogl, G., Beier, U., Boehmer, J., Breine, J., Simoens, I., Caiola, N., De Sostoa, A., Ferreira, M.T., Oliveira, J., Grenouillet, G., Goffaux, D., Leeuw, J.J., Roset, N., Virbickas, T., 2007. Spatially based methods to assess the ecological status of riverine fish assemblages in European ecoregions. Fish. Manage. Ecol. 14, 441–452. Sheer, M.B., Steel, E.A., 2006. Lost watersheds: barriers, aquatic habitat connectivity, and salmon persistence in the Willamette and Lower Columbia River basins. Trans. Am. Fish. Soc. 135, 1654–1669. Smith, C.L., 1985. The Inland Fishes of New York State. New York State Dept. of Environmental Conservation, Albany, NY, USA, 522 pp. Stuart, T.A., 1962. The Leaping Behaviour of Salmon and Trout at Falls and Obstructions. Freshwater and Salmon Fisheries Research. H.M. Stationery Office, Edinburgh, USA, 46 pp. Trautman, M.B., 1981. The Fishes of Ohio. Ohio State University Press, Columbus, OH, USA, 782 pp. Wittler, R.J., Annandale, G.W., Abt, S.R., Ruff, J.F., 1998. New technology for estimating plunge pool or spillway scour. In: Proceedings of the 1998 Annual Conference of the Association of State Dam Safety Officials, Las Vegas, NV.