Spatial Criteria Used in IUCN Assessment Overestimate Area of Occupancy for Freshwater Taxa

Transcription

1 Spatial Criteria Used in IUCN Assessment Overestimate Area of Occupancy for Freshwater Taxa By Jun Cheng A thesis submitted in conformity with the requirements for the degree of Masters of Science Ecology and Evolutionary Biology University of Toronto Copyright Jun Cheng 2013

2 Spatial Criteria Used in IUCN Assessment Overestimate Area of Occupancy for Freshwater Taxa Abstract Jun Cheng Masters of Science Ecology and Evolutionary Biology University of Toronto 2013 Area of Occupancy (AO) is a frequently used indicator to assess and inform designation of conservation status to wildlife species by the International Union for Conservation of Nature (IUCN). The applicability of the current grid-based AO measurement on freshwater organisms has been questioned due to the restricted dimensionality of freshwater habitats. I investigated the extent to which AO influenced conservation status for freshwater taxa at a national level in Canada. I then used distribution data of 20 imperiled freshwater fish species of southwestern Ontario to (1) demonstrate biases produced by grid-based AO and (2) develop a biologically relevant AO index. My results showed grid-based AOs were sensitive to spatial scale, grid cell positioning, and number of records, and were subject to inconsistent decision making. Use of the biologically relevant AO changed conservation status for four freshwater fish species and may have important implications on the subsequent conservation practices. ii

3 Acknowledgments I would like to thank many people who have supported and helped me with the production of this Master s thesis. First is to my supervisor, Dr. Donald Jackson, who was the person that inspired me to study aquatic ecology and conservation biology in the first place, despite my background in environmental toxicology. Don, your mentorship has always been thoughtful, motivating and well-placed, leading me through the transition and through the further research process. To my co-supervisor, Dr. Nicholas Mandrak, your depth of knowledge and enthusiasms towards fishes have constantly excited me and shaped my vision. I cannot express how grateful I am for your generous and inspiring inputs, as well as your encouraging and patient guidance. I am honored to be a graduate student of both of you. To my advisory committee members, Dr. Ken Minns and Dr. Keith Somers, I am blessed to have your insightful comments and questions that contributed tremendously to making this thesis complete. To the Jackson lab (Karen, Lifei, Brad, Jaewoo, Brie, Cindy, Sarah, and Georgina) and the extended members (Liset, Sarah, Danielle, Caren, Jonathon, Caroline, Henrique, Jordan and many more): thank you all for being great colleagues and friends that made grad school an enjoyable journey. I will always remember and deeply appreciate what I have learnt from you both in academically and personally. To my family: Mom, Dad, Grandpa, Juan and Rey. My most sincere appreciation goes to Mom and Dad. Thank you for the years of supporting my education and believing in me while I was away from home. You have given me the best experience growing up independently yet feeling supported. Dad and Grandpa, you are the initial inspirations for me to be interested in iii

4 science and research. Juan and Rey, you are the most awesome aunt and uncle, and the most awesome friends. Thank you for giving me a feeling of home ever since I came to Canada. iv

5 Table of Contents Acknowledgments...iii List of Tables.... vi List of Figures......viii List of Appendices.....x Introduction...1 Methods...8 Results...19 Discussion...43 References...62 Appendices...70 v

6 List of Tables Table 1. Summary of the five quantitative criteria used by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC) for assessing conservation status of wildlife species, which is adapted from the IUCN Red List Categories and Criteria...3 Table 2. Total number of occurrence record, number of geographically distinct localities, 2km- IAO, 1km-IAO, and 0.5km-IAO calculated using grid cells of the same position and orientation, ratio of 2km-IAO to 1km-IAO, and 2km-IAO to 0.5km-IAO in percentage for 20 at-risk freshwater fish species in Ontario Table 3. Mean, standard deviation, maximum, minimum values of 2km-IAO calculated using 13 layers of grid cells at varying positions for 20 at-risk freshwater fish species in Ontario. Also shown are COSEWIC reported 2km-IAO, and ratio of COSEWIC-reported values to the calculated mean IAO values in percentage Table 4. Habitat classification for the 20 at-risk freshwater fish species included in this study...33 Table 5. Home range estimates based on average body length reported in Ontario (Holm et al. 2010) and the equation of Woolnough et al. (2009), and the adjusted buffer scales used for BioAO calculation for the 20 at-risk freshwater fish species included in this study...34 Table 6. Stream width (m) predicted by relationship with Strahler stream order. Values are anti-log e transformed and corrected with bias estimator (Sprugel 1983) Table 7. Proposed BioAO as sum of stream occupancy calculated in terms of occupied stream length x stream width, and lake/wetland occupancy calculated in terms of suitable habitat area within circular buffer, in comparison to COSEWIC reported biological AO for 20 atrisk freshwater fish species in Ontario Table 8. Summary of species status designated under COSEWIC and after application of BioAO, and the corresponding reasons for designation for 20 at-risk freshwater fish species in Ontario vi

7 Table 9. Summary of advantages and drawbacks of each AO measurement approach for freshwater taxa...50 vii

8 List of Figures Figure 1. Study area Figure 2. Example of Index of Area of Occupancy (IAO) measurements for Black Redhorse in the Grand River, Ontario, at three spatial scales: 0.5km, 1km and 2km Figure 3. Example of BioAO measurement for stream occupancy of Black Redhorse in the Grand River, Ontario. The highlighted stream stretch indicates the segments considered occupied habitat. Localities distant from others were considered single locations..15 Figure 4. Example of BioAO calculation for lakeshore occupancy for Grass Pickerel in Long Point Bay, Lake Erie.17 Figure 5. Number of freshwater fish (a) and mollusc (b) species listed under COSEWIC in each threat category and the number of Endangered (EN) and Threatened (TH) freshwater fish (c) and mollusc (d) species designated by criterion..21 Figure 6. Primary and secondary threats identified responsible for species decline in freshwater fish (a) and mollusc (b) species of Canada that are classified as extinct, extirpated, endangered, threatened and special concerned.. 22 Figure 7. Grid-based IAO calculations for 20 freshwater fish species measured at three spatial scales: 2km X 2km (circles); 1km X 1km (triangles); 0.5km X 0.5km (squares), with species arranged on the x-axis in the order of decreasing 2km-IAO size. Dashed lines indicate threshold values for IUCN criteria Figure 8. Mean values of calculated 2km-IAO (triangles) based on 13 placements of grids, COSEWIC-reported 2km-IAO (squares), proposed BioAO (circles), COSEWIC-reported biological AO (diamonds), and HR-BioAO (crosses) for 20 at-risk freshwater fish species in Ontario, with species arranged on the x-axis in the order of decreasing mean 2km-IAO size. Dashed lines indicate threshold values for IUCN criteria. 31 viii

9 Figure 9. Linear regression (solid line) between Strahler stream order and loge-transformed stream width (m) based on 2431 DFO survey records in Ontario (open circles), P < Predicted stream width (see Table 4) at each Strahler order is estimated by the regression Figure 10. Frequency distribution of total BioAO using adjusted buffer size. 38 Figure 11. Linear regression (solid lines) between number of geographically unique occurrence sites and the resultant (a) mean 2km-IAO calculated using 13 gird layers at varying positions (all 20 at-risk freshwater fish species in Ontario, P < ); (b) Stream BioAO in stream length multiplied by stream width (17 of the 20 species, P =0.097); (c) suitable habitat area within circular buffer (17 of the 20 species, P < 0.001) Figure 12. Point distribution records of a freshwater mussel species the Wavy-rayed Lampmussel, Lampsilis fasciola, in the Thames River basin (black circle), and the gridbased IAO of 1km2(shaded grids) and 2km2 (open grids) scales considered by COSEWIC.48 ix

10 List of Appendices Appendix 1. COSEWIC designated conservation status and threat factors identified for freshwater fish species in Canada Appendix 2. COSEWIC designated conservation status and reported threat factors for freshwater mollusc species in Canada...81 Appendix 3. Type of habitats used by freshwater species at risk in Canada. a) fishes; b) molluscs Appendix 4. Proportion of the two components of BioAO: stream occupancy and lake/wetland occupancy for 20 fish species at risk in Ontario Appendix 5. Breakdown of stream BioAO for 17 species inhabited stream environment: area for raw length of occupied stream segment, buffer segment area and area for single locations x

11 1. Introduction Anthropogenic activities have accelerated species extinction rates to an historically high level, leading to an increased demand for conservation actions (Butchart et al. 2010). Prioritizing conservation efforts has become one of the most crucial steps in preserving biodiversity (Hoffmann et al. 2010; Keith et al. 2004; Mace et al. 2008; Miller et al. 2007). One way of achieving such a goal is through assessing and ranking species by the likelihood of extinction (Keith et al. 2004; Mace and Lande 1991; Mace 1994). Over the past four decades, a number of risk-ranking protocols and subsequent lists have been developed across various geographical scales, ranging from global to regional and local scales (Gärdenfors et al. 2001; Mace 1994; Miller et al. 2007; Millsap et al. 1990; Milner-Gulland et al. 2006). These lists provide guidance for developing recovery strategies, designating protected habitat area, and informing policy makers (Lamoureux et al. 2003). The International Union for Conservation of Nature (IUCN) Red List Categories and Criteria is the most recognized conservation status assessment framework (Martín-López 2011). The IUCN Red List status assignment incorporates a set of quantitative criteria to predict extinction probabilities (IUCN Standards and Petitions Working Group 2010). Predictors, such as population size, fragmentation, geographic distribution, and rate of decline, are determined and compared against a series of numerical thresholds to differentiate species into various risk levels. Such systematic procedures have been developed to facilitate objective evaluations and to standardize assessments across taxonomic groups (Mace 1994). IUCN Criteria consist of five major rules, two of which involves spatial analyses of the species geographic distribution (Table 1; IUCN Standards and Petitions Working Group 2010). 1

12 Two spatial indices are used: extent of occurrence (EO); and, area of occupancy (AO). EO measures the total range of a species geographic distribution that encompasses all occurrence records, whereas AO is defined as the actual habitat area occupied by individuals of the species. The two parameters differ in that AO recognizes that not all areas within the geographic range are suitable habitat. These spatial indices are incorporated into conservation assessments to provide insight on population size and its trends, particularly when adequate abundance data are lacking (Gaston and Fuller 2009, Mace et al. 2008; Pritt and Frimpong 2009). Geographic distribution is one of the fundamental ecological and evolutionary characteristics of a species that is both directly and indirectly linked to population health (Hengeveld and Haeck 1982; Gaston and Lawton 1990b; Gaston et al. 2000). It is often regarded as a surrogate for population abundance in conservation risk assessments, based on the notion that a population with a greater number of individuals tends to be more widespread (Cardoso et al. 2011). Numerous studies have demonstrated this positive correlation between range size and population abundance (Bock and Ricklefs 1983; Gaston 1994; Gaston et al. 2000; Lacy and Bock 1986; Schoener 1987). Both spatial parameters are assessed under IUCN Criterion B, which classifies a species into atrisk categories if the geographic range distribution is very limited (IUCN Standards and Petitions Working Group 2010). AO is also considered under subcriterion D2, which qualifies species with extremely restricted population distribution as threatened to reflect the extinction risk associated with demographic stochasticity and the elevated susceptibility to single threat event (Lande 1993; IUCN Standards and Petitions Working Group 2010). These two spatial criteria are the most frequently used reasons for qualifying a species into threatened categories under IUCN Red List (Abeli et al. 2009; Gaston and Fuller 2009). Criterion B alone was identified as the measure influencing the listing of more than 40% of all at-risk species (Gaston and Fuller 2009). 2

13 Table 1. Summary of the five quantitative criteria used by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC) for assessing conservation status of wildlife species, which is adapted from the IUCN Red List Categories and Criteria. Criterion Reasons for designation Indices analyzed A B C D Decline in total number of mature individuals Small distribution range and decline or fluctuation Small and declining number of mature individuals Very small or restricted total population Number of mature individuals EO, AO, and number of locations Number of mature individuals Number of mature individuals, or, AO and number of locations E Quantitative analysis The probability of extinction The current approaches used for calculating the spatial criteria have received numerous criticisms on their effectiveness and applicability (Abeli et al. 2009; Cardoso et al. 2011; Hartley and Kunin 2003; Keith et al. 2000; Mace et al. 2008). EO, measuring the overall range of the species, is usually estimated by a minimal convex polygon (MCP) that encloses all distributional records (IUCN Standards and Petitions Working Group 2010). Inclusion of inappropriate habitats due to discontinuities and disjunctions in the range distribution leads to biases in EO calculation (Burgman and Fox 2003). Although IUCN advices area of unsuitable environments to be excluded from EO estimates, explicit instruction on how to do so is not provided (Willis et al. 2003). 3

14 In the case of AO, a grid-based method is recommended. It is performed by overlaying uniform-sized grids on the range of a species and summing the area of the cells in which the species occurred (IUCN Standards and Petitions Working Group 2010). This approach is frequently opposed because it is extremely sensitive to the resolution of grid cells (Hengeveld and Heack 1982; Joseph and Possingham 2008; Keith et al. 2000; Whittaker et al. 2005). The scale at which AO is measured varies greatly across taxonomic groups, where cell sizes ranged from 0.1 to 1000 km 2 in a review of 47 species (Gaston and Fuller 2009). Consequently, standardization to a reference scale (currently 2km by 2km grid size as recommended by IUCN) based on scale-area relationship is required (Mace et al. 2008). The ability to detect population decline through grid-ao is also questionable. Joseph and Possingham (2008) showed the AO measured at the traditionally recommended scale was insufficient to accurately detect declines in population abundance. Freshwater ecosystems support diverse groups of species and, yet, are under the most substantial extinction crisis (Bruton 1995; Duncan and Lockwood 2001). Great proportions of freshwater fauna are imperiled, including freshwater fishes, molluscs, crayfishes, and amphibians (IUCN 2012; Jelks et al. 2008; Régnier et al. 2009; Ricciardi and Rasmussen 1999). The projected global extinction rate is strikingly higher in these taxa compared to their terrestrial counterparts (Ricciardi and Rasmussen 1999). In North America, 61 species and subspecies of freshwater fishes and 21 species of freshwater molluscs have become extinct (Jelks et al. 2008). The leading factors contributing to the endangerment of aquatic organisms have been identified as habitat alteration and degradation, followed by overexploitation, pollution, and introduction of alien species (Dextrase and Mandrak 2006; Jelks et al. 2008). However, freshwater fauna have received a disproportionately small amount of conservation attention (Duncan and Lockwood 2001; Maitland 1995). Only 33% of all described freshwater fish species and 3% of 4

15 molluscs have been assessed under the IUCN classification regime, in comparison to 100% coverage in mammals and avian species, and 94% in amphibians (IUCN 2012). The suitability of current spatial criteria for these freshwater species has been debated. In addition to the shortcomings described above, the measurement for AO is thought to be especially problematic for freshwater organisms. During the early development of IUCN Red List Criteria, grid-measured AO was derived from an array of studies on abundance distribution relationship measured by number of occupied grids (Mace 1994). However, most of these studies were conducted for land birds over broad geographic ranges (Bock 1984; Brown 1984; Ford 1990; Schoener 1987) and this relationship was rarely shown in aquatic species (Gaston and Lawton 1990a). The fundamental problem is that the grid-adjacency measure of AO does not account for the confined geographic characteristics of aquatic habitats (Gaston and Lawton 1990a). Unlike those in terrestrial and marine environments, where movements are relatively free over large areas, freshwater organisms are restricted to limited space within clear boundaries and dispersal across space is directional and hierarchical within watercourse networks (Hitt and Angermeier 2008). Freshwater organisms are also often associated with linear habitat ranges, such as streams and lakeshore areas (Burgman and Fox 2003; Joseph and Possingham 2003; Mace et al. 2008) or confined to small waterbodies. As a result, grid-based occupancy measurement may fail to capture the dimensionality of such habitats and to reflect the underlying geographic patterns in species distribution and abundance, leading to subsequent misinterpretation in extinction risk. The Committee on the Status of Endangered Wildlife in Canada (COSEWIC) is responsible for identification of threatened wildlife species at the national level in Canada, and adopts the IUCN Red List protocol (COSEWIC 2010). The risk categories assigned by 5

16 COSEWIC follows that of IUCN, including Extinct (EX), Extirpated (ET), Endangered (EN), Threatened (TH), Special Concern (SC), Not at Risk (NR), and Data Deficient (DD). EN and TH are considered the at-risk groups, where critical thresholds in conservation risk criteria are evaluated. SC also represents an at-risk category, which is composed of species with known declines and/or existing threats, but do not meet thresholds for designation in higher risk categories. COSEWIC conservation assessments provide the basis for listing under the federal Species at Risk Act (SARA), and for development and implementation of species recovery strategies and recovery action plans. COSEWIC separates species into designatable units (DUs), where appropriate, and assesses each DU separately. For simplicity, heretofore all species, subspecies, designatable units, and populations are referred to as species. Like the IUCN framework, COSEWIC incorporates an occupancy parameter, the Index of Area of Occupancy (IAO), measured following the grid concept. IAO is required to be calculated at a scale of a 2km-grid (i.e. 2km by 2km). An alternative scale of 1km-grid size is recommended in cases where biological relevance can be argued (COSEWIC 2010). However, given that most streams in Canada average less than 100m in width, IAO values measured at these resolutions are likely to include more terrestrial habitat than actual occupied aquatic habitats, grossly overestimating AO for most freshwater taxa. In this study, we used COSEWIC assessments for freshwater organisms in Canada as a case study to: 1) understand the relative importance of spatial criteria for status designation; 2) further explore the potential biases and shortfalls associated with the grid IAO approach; and, 3) develop a biologically relevant measure of AO that reflects the risk of extinction and can be universally applied with relative ease. The first objective was achieved by compiling all existing status assessments available for freshwater species, including fishes and molluscs. The latter two objectives involved measuring 6

17 IAO and a proposed biological AO for 20 imperiled freshwater fish species native to southwestern Ontario. 7

18 2. Methods COSEWIC Assessments for Freshwater Species I compiled all available COSEWIC status assessment reports on Canadian freshwater species, totaling 131 fishes and 22 molluscs (Appendix 1, Appendix 2). Assessment reports were obtained from the SARA Registry ( as of January For each species evaluated, the following data were gathered: 1) current conservation status; 2) criteria for designation; 3) reported measures and trends (decline and/or fluctuation) in spatial indices, including EO, AO, number of locations and habitat quality, and method used to calculate indices when applicable; 4) threat factors suspected to have led to observed or projected population decline; and, 5) preferred habitat. Species from all risk categories were included in this compilation. A species could be classified into an at-risk category by meeting the threshold for one or more criteria. The reason(s) for designation was derived from the technical summary section of the assessment report or inferred from the report when not explicitly stated. Threat factors were classified into eight categories, adapted from IUCN (2012): habitat loss and degradation; alien species invasion; over-harvesting; pollution; natural disaster; change in native species dynamic; persecution; and, other human disturbances. Threats were further identified as either primary or secondary causes of endangerment for each given species, similar to the approach taken by Dextrase and Mandrak (2006). A primary threat was defined as a major factor known to cause risk of extinction, while secondary threat referred to factors having a minor role or projected effect, or of unknown significance. 8

19 Evaluating IAO Calculations I used distribution data for 20 imperiled freshwater fish species in Ontario to reproduce IAO calculations, with the goal of investigating issues surrounding the grid adjacency approach. Occurrence data of these species were obtained from the Department of Fisheries and Oceans (DFO) (Mandrak, unpublished data). The distribution data were originally compiled for conducting COSEWIC assessments, which included both historical and recent records collected from all available sources, ranging from community surveys and targeted sampling conducted by various organizations including DFO and Ontario Ministry of Natural Resources (OMNR), independent surveys conducted for research purposes, and museum records from Canadian Museum of Nature and Royal Ontario Museum (Doolittle et al. 2007). The area covered by the distribution localities was restricted to southern Ontario (Figure 1). This area is known to have undergone urbanization, dam construction, water extraction, pollution, and invasion by several alien species (Dextrase and Mandrak 2006) that have decreased the ecosystem health and imperiled the local freshwater fauna. IAO measurements were performed using a Geographic Information System (GIS), following the IUCN, and hence COSEWIC, guidelines. The distribution data were georeferenced latitude-longitude point records and were projected in ArcGIS v10. Vector layers of uniform grid cells were created using E.T. GeoWizard and overlaid onto species distribution maps. Using the Spatial Analyst extension, cells were then spatially joined to species occurrence records, which permitted objective tallying of occupied cells. The E.T. GeoWizard tool pack was advantageous in this calculation because it created grid cells as vector features, with the ability to customize grid sizes and location of cell boundaries. 9

20 Figure 1. Study area. Three major issues associated with the grid-adjacency method were examined: 1) scaledependence; 2) influence of position and orientation; and, 3) reproducibility. To address the scale-dependency issue, I calculated IAO at three spatial scales: 2km, 1km, and 0.5km. These grid sizes reflected the currently recommended scale, the frequently utilized alternative scale for freshwater organisms, and a conservative scale matching the average width of streams, respectively (see Figure 2 for an example). For each species, 2km-IAO was compared to IAO values resulted from the finer resolutions. IAO values at all three resolutions were compared against the critical thresholds. The second issue, grid location and orientation, was approached by producing grid layers of the same resolution with shifted border positions. This was achieved by specifying coordinates of the grid layers range in the E.T. GeoWizard. Thirteen 2km-grid 10

21 layers were generated in this fashion, eleven of which were north-south oriented, whereas, the remaining two were tilted at an angle, which was sometimes used in COSEWIC assessments (Mandrak, unpublished data). For each species, mean value, maximum, minimum, and standard deviation of the resultant IAO values were recorded for comparison against the critical thresholds. Lastly, I evaluated the reproducibility of the grid-based approach by identifying discrepancies between the calculated IAOs and the COSEWIC-reported values. The average values of all 2km-IAO calculated for each species were compared to that reported by COSEWIC. Figure 2. Example of Index of Area of Occupancy (IAO) measurements for Black Redhorse in the Grand River, Ontario, at three spatial scales: 0.5km, 1km and 2km. 11

22 Biological AO As an alternative to IAO, I propose a biologically relevant measure of Area of Occupancy for aquatic species living in confined habitats, heretofore referred to as the BioAO. I categorized species occurrences into four major groups of freshwater habitats: stream, lakeshore, wetland, and open lake. Separate approaches were taken for occurrences in streams, and in lakes (lakeshore and open lake) and wetlands. Such separation of habitat type recognized the fundamental differences in geographical shapes of the available habitat and the differences in constraints on fish movements. Total BioAO for a specific species was the sum of the two components. Biological Area of Occupancy in Streams BioAO for stream occupancy was calculated as the product of occupied stream segment length (m) and estimated stream width (m). Occupied stream segment length was measured as distances between species distribution points through a stream network following a least-cost routing application (ArcGIS, Network Analyst). Distribution records of the 20 fish species were projected onto a stream network layer of the Great Lakes drainage basin. This stream network was created by converting line features of stream segments developed by Aquatic Landscape Inventory Software (ALIS) into an interconnecting network. ALIS is a stream segmentation application that classifies stream segments based on a set of variables including hydrography, surficial geology, connectivity, flow barriers, and thermal regime (Valley Segment Committee 2001). The stream network covered in our study area was developed with hydrographic maps at a fine scale (1:10000), which was crucial for recognizing small headwater and accurate stream 12

23 order classifications (Hughs et al. 2011). Each stream segment in ALIS was described by watershed code, segment length, and Strahler and Shreve stream orders. As point distribution data typically underestimate actual distribution, a buffered approach was used to determine occupied stream segment length. I introduced a term, buffer distance, to quantify the length of the stream segment I assumed an individual could travel through. This method utilized species-specific spatial scales by considering variations in mobility among species. Assuming each occurrence location represented at least one individual, the capacity by which the individual might move between these interconnected locations through the streams was considered to determine the size of its occupancy. If two occurrence sites of a species were separated by a distance greater than twice that of the buffer length, the stream segment in between was considered unoccupied and excluded from being part of BioAO. The buffer distance was also introduced to each end of an occupied segment to account for potential habitat usage. Localities with single occurrence record were considered as single locations, for which the size of occupied stream reach was assigned as one buffer (see Figure 3 for an example). Buffer distance used in BioAO analysis was determined based on the predicted home ranges. This approach is consistent with the notion that area of occupancy is essentially the sum of home range areas of all individuals within a population (Gaston 1991). Home range is defined as the area over which an individual travels or lives in and has been found to correlate positively with body size in freshwater fish species (Minns 1995, Woolnough et al. 2009). I used the following allometric relationship constructed for lotic species (Woolnough et al. 2009) to predict home range as a function of body size: log home range (m) = x log body size (cm 3 ). 13

24 Body size was calculated following method described by Woolnough et al. (2009) using average fish body length (mm) reported in Ontario by Holm et al. (2010). Back-transformation of logtransformed home range sizes were corrected for bias (Sprugel 1983). A set of biological AO was calculated using the predicted home range as buffer length, heretofore, referred as HR- BioAO. The final proposed BioAO was an adjusted buffer length based on home range sizes in the following manner: the buffer range for a given species was arbitrarily set to 1km or 2km if home range was less than 60m, or between 60m and 2000m, respectively. For species with home range estimated greater than 2000m, home range size was used directly as the buffer distance. The 1- and 2-km river lengths corresponded to those scales currently being applied under COSEWIC and IUCN assessments for grid measurements, which provide rationale for comparing against the current thresholds. These scales are also similar to those referred to as local segment scales used in other studies, such as linear stream buffer of 5km used by Fagan et al. (2005) and 3.25km used by Groce et al. (2012). I included a layer of hydrological structures (OMNR 2012) to reflect barriers to fish movements and dispersal in stream habitats. Point locations of dams were used as barriers within the stream network when conducting the least-cost calculation to indicate terminals of occupied stream segments. 14

25 Figure 3. Example of BioAO measurement for stream occupancy of Black Redhorse in the Grand River, Ontario. The highlighted stream stretch indicates the segments considered occupied habitat. Localities distant from others were considered single locations. Estimates of stream width for the occupied segments were the other component of BioAO measure. A dataset of stream width information for over 100 Ontario streams was attained (N.E. Mandrak, DFO, unpublished data). It contained information collected from a total of 2431 sampling events, with records of waterbody names, date of the sampling, and latitudelongitude coordinates. Two approaches were taken to estimate occupied stream segment width under different scenarios. If an occupied stream was sampled in the width dataset, then the average width for that particular stream was used. If the information was not immediately available, I predicted stream width as a function of Strahler stream order. 15

26 A linear regression analysis between Strahler order and the stream width was constructed (ln-transformed; Hugh et al. 2011) for streams in Ontario. I projected the sampling records as point vectors onto ALIS stream network and performed spatial join to associate width data with the Strahler order of the stream reach in which it was measured. I manually relocated points that were, based on the waterbody name, snapped onto incorrect waterbodies during the spatial join. Stream widths estimated through this relationship were corrected using the anti-logarithmic bias corrector (Sprugel 1983). Lake and Wetland Occupancy I combined all lakeshore, wetland, and open-lake occurrence sites and measured occupancy using a circular buffer approach (see Figure 4 for an example). For each occurrence locality, a circular buffer was created with a species-specific range. I determined the radius of the buffer size based on the species movement size calculated the same way for the stream occupancy. HR-BioAO was calculated based on a radius of the predicted home range size of the species, and BioAO was calculated based on a radius size of the adjusted buffer length. To eliminate terrestrial area from the AO measure, the buffered zone was overlaid onto the map of major and minor waterbodies from Canada Water Maps (2012). Only areas of the waterbodies intersected within circular buffer were considered as occupied habitat. 16

27 Figure 4. Example of BioAO calculation for lakeshore occupancy for Grass Pickerel in Long Point Bay, Lake Erie. Applicability of BioAO I compared the risk of extinction and conservation status according to the COSEWIC guidelines based on IAO and my proposed BioAO. Note that COSEWIC analyses for six of the 20 species (Bridle Shiner, Channel Darter, Grass Pickerel, Northern Brook Lamprey, River Redhorse, Silver Lamprey) included DUs covering areas for which I did not have point distribution data. For this reason, for the other fourteen species with complete distribution data, the calculated IAO values, BioAO values were compared to the COSEWIC-reported AOs and their conservation status were reassessed. 17

28 In addition, I investigated the extent to which the different methods were influenced by the number of distribution localities available. Linear regression of the 2km-IAO, stream BioAO, and lake BioAO against the number of geographically distinct occurrence localities were performed. 18

29 3. Results Summary of COSEWIC Assessments for Freshwater Species The review of COSEWIC reports for freshwater fishes and molluscs revealed the high degree of endangerment of these aquatic species (Figure 5a, b). Among 131 freshwater fish species assessed under COSEWIC, 12 species were extinct or extirpated, and 79 existing species were classified in the imperiled categories (EN, TH, or SC). Out of the 22 freshwater mollusc species assessed under COSEWIC, one species was considered extirpated, and 20 species were classified in the imperiled categories (EN, TH, or SC), among which 16 species were Endangered. Compilation of these assessments confirmed that the two spatial criteria (Criterion B and D2) were used most frequently as the factor for designation of at-risk aquatic organisms (Figure 5c, d). Both Criterion B and D2 were used alone, or in conjunction with other criteria, to determine the imperilment of species; 54% of the endangered and 48% of the threatened fish species were assessed using Criterion B, and 57% of the threatened fish species were assessed using Criterion D2. Criterion B was also used for assessing 88% of the endangered, and the only threatened, freshwater molluscs. AO measures reported in the COSEWIC assessments for the freshwater taxa varied in the methods used. Ninety of the 131 freshwater fishes and 20 of the 22 freshwater molluscs were reported with at least one measure of AO. The most common AO measure reported in COSEWIC assessments was IAO measured at 2km-scale (38 fish species; 10 mollusc species), whereas 1km-IAO (25 fish species; 5 mollusc species), biological AO in the form of estimated 19

30 stream area (21 fish species; 13 mollusc species), and preferred habitat area (25 fish species; 1 molluscs species) were also presented. Ten freshwater fish species had IAO reported to be greater than their EO. A number of threat factors were identified to have contributed to the imperilment of Canadian freshwater fauna (Figure 6). This review only considered species in categories at Special Concern or higher because information was not always available for species that were Not at Risk or Data Deficient. Multiple threats were listed for a majority of the species (71 of the 88 freshwater fishes, 20 of the 21 freshwater molluscs). Habitat loss and degradation were the most significant threat to imperilment (45 fish species; 17 mollusc species; Figure 6). Introduced species was cited as the primary threat for 19 freshwater fish and 13 mollusc species; Pollution was identified as the primary threat for 15 freshwater fish and 9 mollusc species; and freshwater fishes also suffered from over-exploitation. 20

31 Designating Criteria Designating Criteria Extinct Extirpated Endangered Threatened Special Concern Not at Risk A B C D E EN TH (a) Data Defficient Number of Species (c) Number of Species Extinct Extirpated Endangered Threatened Special Concern Not at Risk A B C D E EN TH Data Defficient Number of Species Number of Species (b) (d) Figure 5. Number of freshwater fish (a) and mollusc (b) species listed under COSEWIC in each threat category, and the number of Endangered (EN) and Threatened (TH) freshwater fish (c) and mollusc (d) species designated by criterion. 21

32 Habitat loss/degradation Alien invasive species Harvest Pollution Natural disaster Change in native species Persecution Human disturbances Primary Secondary (a) Number of Species Habitat loss/degradation Alien invasive species Harvest Pollution Natural disaster Change in native species Persecution Human disturbances Primary Secondary (b) Number of Species Figure 6. Primary and secondary threats identified responsible for species decline in freshwater fish (a) and mollusc (b) species of Canada that are classified as Extinct, Extirpated, Endangered, Threatened and Special Concern. 22

33 IAO calculations The IAO measures increased with increasing grid scale (Figure 7, Table 2). All 2km- IAO values were below the TH threshold under subcriterion B2 (2000km 2 ) and above that for D2 (20km 2 ), and ranged from 64km 2 (Warmouth) to 1196km 2 (Redside Dace). Based on the 2km-IAO, 16 species met the criteria for Endangered under B2 (500km 2 ). Based on the 1km- IAO, all species met the threshold for being Endangered (B2), ranging from 25km 2 to 409km 2. Based on the 0.5km-IAO, almost half of the species (8 out of 20) qualified for being geographically restricted (8.5km 2 to km 2 ). On average, 2km-IAO values were 325% of the corresponding 1km-IAO values, and 1120% of the corresponding 0.5km-IAO values. Variation in grid placement yielded insignificant amounts of variation in 2km-IAO measures for majority of the species (Table 3; Figure 8). Maximum difference among measures was greatest for Redside Dace (80km 2 ) and Grass Pickerel (60km 2 ). Blackstripe Topminnow, Eastern Sand Darter, and Silver Chub also showed variation in IAO results to a certain degree (greater than 40 km 2 maximum differences). For Silver Shiner, maximum and minimum values were 476km 2 and 508km 2, respectively, which were below or above the critical threshold for EN. Relatively poor reproducibility of the grid-approach was demonstrated by the discrepancies between the mean calculated 2km-IAO measures and COSEWIC-reported values (Table 3). The mean 2km-IAO values that I calculated strongly deviated from the IAOs reported by COSEWIC for five species: Blackstripe Topminnow, Bridle Shiner, Lake Chubsucker, Pugnose Minnow, and Silver Shiner, where the COSEWIC reported IAO values exceeded my calculations by more than 50%. 23

34 IAO (km2) TH: B EN: B2 100 TH: D2 10 2X2 IAO (km2) 1X1 IAO (km2) 1 0.5X0.5 IAO (km2) Figure 7. Grid-based IAO calculations for 20 freshwater fish species measured at three spatial scales: 2km X 2km (circles); 1km X 1km (triangles); 0.5km X 0.5km (squares), with species arranged on the x-axis in the order of decreasing 2km-IAO size. Dashed lines indicate threshold values for IUCN criteria. 24

35 Table 2. Total number of occurrence record, number of geographically distinct localities, 2km-IAO, 1km-IAO, and 0.5km-IAO calculated using grid cells of the same position and orientation, ratio of 2km-IAO to 1km-IAO, and 2km-IAO to 0.5km-IAO in percentage for 20 at-risk freshwater fish species in Ontario. Common name Scientific name COSEWIC status Total number of occurrences Number of geographically unique localities 2km-IAO (km 2 ) 1km-IAO (km 2 ) 0.5km- IAO (km 2 ) 2km-IAO to 1km-IAO (%) 2km-IAO to 0.5km- IAO (%) Black Redhorse Blackstripe Topminnow Bridle Shiner 1 Channel Darter 1 Eastern Sand Darter Grass Pickerel 1 Lake Chubsucker Moxostoma duquesnei Fundulus notatus Notropis bifrenatus Percina copelandi Ammocrypta pellucida Esox americanus vermiculatus Erimyzon sucetta TH % % SC % % SC % % TH % % TH % % SC % % EN % % 25

36 Lake Sturgeon (Great Lakes - Upper St. Lawrence populations) Northern Brook Lamprey (Great Lakes - Upper St. Lawrence populations) 1 Northern Madtom Pugnose Minnow Pugnose Shiner Redside Dace River Redhorse 1 Silver Chub Acipenser fulvescens Ichthyomyzon fossor Noturus stigmosus Opsopoeodus emiliae Notropis anogenus Clinostomus elongatus Moxostoma carinatum Macrhybopsis storeriana TH % % SC % % EN % % TH % % EN % % EN % % SC % % EN % % 26

37 Silver Lamprey (Great Lakes - Upper St. Lawrence populations) 1 Silver Shiner Spotted Gar Spotted Sucker Warmouth Ichthyomyzon unicuspis Notropis photogenis Lepisosteus oculatus Minytrema melanops Lepomis gulosus SC % % TH % % TH % % SC % % SC % % 1 COSEWIC species assessments included distribution records outside of Ontario which were not available to be incorporated in this analysis. 2 COSEWIC reported 2km-IAO included historical data. 3 IAO in 1km-scale, calculated with single grid layer or reported by COSEWIC. * 2km-IAO value for Ontario occurrence only (Mandrak, unpublished data). 27

38 Table 3. Mean, standard deviation, maximum, minimum values of 2km-IAO calculated using 13 layers of grid cells at varying positions for 20 at-risk freshwater fish species in Ontario. Also shown are COSEWIC reported 2km-IAO, and ratio of COSEWICreported values to the calculated mean IAO values in percentage. Common name COSEWIC Standard Max 2km-AO Min 2km-IAO COSEWIC Mean 2km- COSEWIC- status IAO deviation in (km 2 ) (km 2 ) 2km-IAO reported 2km- (km 2 ) 2km-IAO (km 2 ) IAO to mean (km 2 ) 2km-IAO (%) Black Redhorse TH Blackstripe Topminnow SC % Bridle Shiner 1 SC * 224% Channel Darter 1 TH Eastern Sand Darter TH 540 (172 3 ) (304 3 ) 103% (177% 3 ) Grass Pickerel 1 SC Lake Chubsucker EN 243 (80 3 ) (243 3 ) 165% (304% 3 ) 28

39 Lake Sturgeon (Great Lakes - Upper St. Lawrence populations) Northern Brook Lamprey (Great Lakes - Upper St. Lawrence populations) 1 TH SC Northern Madtom EN % Pugnose Minnow TH % Pugnose Shiner EN % Redside Dace EN 1207 (409 3 ) (441 3 ) (108% 3 ) River Redhorse 1 SC Silver Chub EN Silver Lamprey (Great Lakes - Upper St. Lawrence populations) 1 SC Silver Shiner TH % 29

40 Spotted Gar TH Spotted Sucker SC Warmouth SC COSEWIC species assessments included distribution records outside of Ontario which were not available to be incorporated in this analysis. 2 COSEWIC reported 2km-IAO included historical data. 3 IAO in 1km-scale calculated with single grid layer or reported by COSEWIC. * 2km-IAO value for Ontario occurrence only (Mandrak, unpublished data). 30

41 Area of Occupancy (km2) 1000 TH: B2 EN: B2 100 TH: D Mean 2km-IAO COSEWIC reported 2km-IAO BioAO COSEWIC reported biological AO HR-BioAO Figure 8. Mean values of calculated 2km-IAO (triangles) based on 13 placements of grids, COSEWIC-reported 2km-IAO (squares), proposed BioAO (circles), COSEWIC-reported biological AO (diamonds), and HR-BioAO (crosses) for 20 at-risk freshwater fish species in Ontario, with species arranged on the x-axis in the order of decreasing mean 2km-IAO size. Dashed lines indicate threshold values for IUCN criteria. 31

42 Proposed Biological AO calculations Seventeen of the 20 species had stream occupancies, of which Black Redhorse and Blackstripe Topminnow were strictly stream species (Table 4), and 18 species occurred in lakes and wetland habitats, of which Spotted Gar and Warmouth were restricted to wetlands. Linear individual home range calculations based on body length ranged from 29.3m to m (Table 5). Four cyprinid species and two percid species had linear home range estimates of less than 60m and, therefore, were assigned with a minimum buffer distance of 1km. Four largebodied species had home ranges predicted to be greater than 2km, and the remaining species had adjusted buffer distance set to 2km. The total biological AO calculated as the sum of occupancies in two habitat types yielded AO measures of small values (Table 8 and Appendix 4). All BioAO calculated with the adjusted buffer length were below the TH threshold under subcriterion B2 (2000km 2 ). All species except for Silver Chub (1643km 2 ), had BioAO below the EN threshold under subcriterion B2 (500km 2 ). Three species that met the TH threshold under subcriterion D2 (20km 2 ) were Blackstripe Topminnow (3.28km 2 ), Northern Brook Lamprey (18.01km 2 ), and Pugnose Minnow (12.31km 2 ). Another three species also demonstrated relatively limited occupancy: Black Redhorse (25.68km 2 ), Redside Dace (30.06km 2 ), and Silver Shiner (27.92km 2 ). Frequency distribution showed the majority of the assessed species (13 out of 20) had BioAO sizes between 20km 2 and 200km 2 (Figure 10).When home range predicted from the allometric relationship was used directly as buffer distance, all HR-BioAO were smaller than the EN threshold under subcriterion B2, eleven of which fell below the TH threshold of subcriterion D2. 32

43 Table 4. Habitat classification for the 20 at-risk freshwater fish species included in this study. Common Name Habitat Type Stream Nearshore lake Wetland/marsh Open lake Black Redhorse X Blackstripe Topminnow X Bridle Shiner X X Channel Darter X X Eastern Sand Darter X X X Grass Pickerel X X X Lake Chubsucker X X X Lake Sturgeon X X X Northern Brook Lamprey X X X Northern Madtom X X Pugnose Minnow X X Pugnose Shiner X X X Redside Dace X River Redhorse X X Silver Chub X Silver Lamprey X X X Silver Shiner X Spotted Gar X Spotted Sucker X X X Warmouth X 33

44 Table 5. Home range estimates based on average body length reported in Ontario (Holm et al. 2010) and the equation of Woolnough et al. (2009), and the adjusted buffer scales used for BioAO calculation for the 20 at-risk freshwater fish species included in this study. Common name Average body length in Ontario (mm) Raw home range predicted (m) Home range after bias correction (m) Adjusted buffer distance (m) Black Redhorse Blackstripe Topminnow Bridle Shiner Channel Darter Eastern Sand Darter Grass Pickerel Lake Chubsucker Lake Sturgeon Northern Brook Lamprey Northern Madtom Pugnose Minnow Pugnose Shiner Redside Dace River Redhorse Silver Chub Silver Lamprey Silver Shiner Spotted Gar Spotted Sucker Warmouth

45 loge-transformed Width (m) Stream-width records were associated with streams, with Strahler stream order ranging from 1 to 9 on the 1: scale ALIS map. Regression analysis showed that Strahler order was a significant predictor for stream width (r 2 = 0.62, P < ) (Figure 9). The predicted width estimates ranged from 3.66m for 1 st order streams to m for 9 th order streams (Table 6) y = x r² = Strahler Order Figure 9. Linear regression (solid line) between Strahler stream order and log e -transformed stream width (m) based on 2431 DFO survey records in Ontario (open circles), P < Predicted stream width (see Table 6) at each Strahler order is estimated by the regression. 35