Factors associated with extirpation of the last Northern Sunfish (Lepomis peltastes Cope, 1870) population in western New York State, USA

Abstract The Northern Sunfish (Lepomis peltastes Cope, 1870) is threatened in New York state, USA, but this was not the case before 1940 when the NY Biological Survey documented the species at scattered, specialized habitats in six watersheds in the central and western parts of the state. After 1940 the historic populations could not be detected, but a new population was discovered in 1974 in lower Tonawanda Creek and the nearby Erie Canal. Northern Sunfish, and a few of their hybrids with other Lepomis species, were caught at these locations during irregular sampling through 2009, but no Northern Sunfish were caught after 2009. The objectives of our study were to: (1) Determine the extent of Northern Sunfish hybridization with other Lepomis species, and (2) Evaluate how well identifications of Lepomis species and their hybrids agreed among field keys, morphometric measurements and meristic counts, and genetic methods. In 2013, we collected Northern Sunfish (descended from fish captured in lower Tonawanda Creek from 2006-2009) from NY State Department of Environmental Conservation rearing ponds, plus wild Green Sunfish (L. cyanellus Rafinesque, 1819), Pumpkinseed (L. gibbosus Linnaeus, 1758), Bluegill (L. macrochirus Rafinesque, 1819), and suspected Lepomis hybrids from lower Tonawanda Creek. Ultimately, 91 fish were identified using field keys, morphometric-meristic analysis, and mtDNA and nuclear DNA analysis. Assuming genetic analysis provided accurate identification, we found 7 Bluegill × Northern Sunfish, 8 Bluegill × Pumpkinseed, 13 Bluegill × Green Sunfish, and 3 Green Sunfish × Pumpkinseed hybrids in our sample (female parent listed second in these crosses). Keyed and morphometric-meristic identifications did not differ in accuracy and averaged 81% of genetic identification accuracy. After Northern Sunfish stocking (not in our study area) and sampling from 2008 to 2018 in several watersheds with appropriate habitat and no recaptures after 2014, we conclude that the Northern Sunfish is extirpated in western New York state. Highlights While populations of Northern Sunfish (Lepomis peltastes) existed in several New York state watersheds before 1940, only one, discovered in 1974, persisted in small areas of lower Tonawanda Creek and the nearby Erie Canal. Despite high effort, no Northern Sunfish were captured after 2009 in the places they had occupied since 1974. Because many Northern Sunfish, all descended from wild fish in lower Tonawanda Creek and the Erie Canal, exist in two New York state hatchery ponds, we recommend attempting restoration in their former habitat by stocking.


Introduction
At least 57 species and subspecies of freshwater fishes in North America (Canada, the United States, and Mexico) have gone extinct since 1900.The extinction rate accelerated after 1989, a trend that without changes in human practices is projected to continue through 2050 and beyond (Dudgeon et al. 2006;Burkhead 2012).The primary causes of freshwater fish extinctions are habitat loss (e.g.desert pupfishes, F. Cyprinodontidae, in the western United States and Mexico) and introduction of non-indigenous species (e.g.Sea Lamprey, Petromyzon marinus Linnaeus, 1758 and Alewife, Alosa pseudoharengus Wilson, 1811 in the Great Lakes), with localized endemics and species with narrow habitat preferences being most vulnerable (Jelks et al. 2008;Burkhead 2012).Proximity to major urban areas, transportation corridors, and energy development and resource extraction sites (all cause pollution and sedimentation), as well as range restrictions and hydraulic alterations caused by dams, also contribute to extirpations (all individuals in a population have been lost) and extinctions (all individuals in a species have been permanently lost) (Jelks et al. 2008;Burkhead 2012).Also, small populations, often at the edges of their species' range, are vulnerable to Allee Effects where there is a positive association between per capita population growth rate and population size.When Allee Effects are strong they may give rise to a critical population size below which a population cannot persist (Drake and Kramer 2011).
The 33 species of basses and sunfishes in North America (F.Centrarchidae) are mostly widespread hardy freshwater fishes, with no extinct, endangered, or threatened species listed nationally (Page et al. 2013;NANFA, 2021).The Northern Sunfish (Lepomis peltastes Cope, 1870) was formerly recognized as one of two subspecies of the Longear Sunfish (L.megalotis Rafinesque, 1820): L. m. peltastes, the Northern Longear Sunfish, and L. m. megalotis, the Central Longear Sunfish (Jennings 1991;Bailey et al. 2004;Page et al. 2013;Kim et al. 2022).Here we report on factors that may have contributed to the extirpation of the Northern Sunfish (NS) in a western New York (NY) stream.
The distribution of NS is patchy across the Great Lakes-St.Lawrence River and upper Mississippi River drainages of North America (Scott and Crossman 1973;COSEWIC 2016).In Canada, it is found in northwestern, southern, and eastern Ontario, and southwestern Québec (Figure 1).In the United States, it is distributed across northern Ohio and Indiana, northeastern Illinois, the lower peninsula of Michigan, and eastern Wisconsin.A disjunct portion of the range occurs in north-central Minnesota, and it extends into northwestern Ontario (Figure 1).Several additional disjunct, and likely relict, populations are present in southern Minnesota, central/western Wisconsin, southern Illinois, and Iowa (probably extirpated) (COSEWIC 2016).The NS is presumed Extinct in Pennsylvania; its status is Secure in Michigan, and Special Concern and Vulnerable in Ontario and Quebec, Canada, respectively.It has no status rank in Illinois, Indiana, Wisconsin, and Ohio (COSEWIC 2016; NatureServe 2022).In NY, at the eastern end of its North American range (Figure 1), its status is threatened (NYNHP, 2022).Notably, in 2016 a new NS population was discovered in the Great Chazy River (Maxwell and Carlson 2018;Carlson 2019) in far northeastern NY (Figure 1).
The NS is small, ornate and occupies warm water streams, rivers, ponds, and small lakes with low turbidity.Spawning occurs in the early summer over bowl-shaped nests constructed by males in gravel substrate (Keenleyside 1972;Scott and Crossman 1973).Favored habitats are stream (and pond) margins with slack water with depths <1 m, calm water near currents or eddies, and shelter, including submerged aquatic vegetation and woody debris (fallen logs and brush) (Wells and Haynes 2007;COSEWIC 2016).
Biological surveys in NY from 1926 to 1940 indicated that NS were native in a handful of watersheds in central NY and the lower reaches of three southwestern Lake Ontario watersheds (Carlson et al. 2016): West Creek/Braddock Bay, Jeddo Creek/Johnson Creek, and Marsh Creek/Oak Orchard Creek (Figure 2).No NS were found at these locations after 1940.In 1974, another population was discovered in lower Tonawanda Creek (LTC) and the nearby NY Barge Canal (Erie Canal) that is tributary to the upper Niagara River (Figure 2).Between the LTC-canal confluence and Niagara River, the canal is the dredged Tonawanda Creek channel which has lateral wetlands (wide waters) fed by small tributaries.As of 2009, the last known wild NS in NY were found in the last 6 km of LTC before it joins the canal and a canal wide water 6 km downstream from the confluence with LTC (Figure 2; Carlson et al. 2016), but none were found after 2009.
The primary objective of our 2013-2014 study was to determine whether any NS remained in LTC and the nearby Erie Canal.However, because no NS had been caught in LTC since 2009 and numbers of other Lepomis species were high in 2005 (Wells and Haynes 2007;Wells 2009), we were concerned that hybridization or competition with other Lepomis species might be inhibiting the ability of NS to maintain a viable population in our study area.To address the concern about hybridization required genetic analysis of Lepomis species and their potential hybrids, so our secondary objective was to compare agreement among keyed, morphometric-meristic, and genetic identifications.This comparison may be useful for those who study similar problems in the future using only keyed or morphometric-meristic identifications.

Study area
Tonawanda Creek is a lowland stream with particularly turbid water due to clay banks and nearby loamy soils.From our study area, it flows via the Erie Canal through the city of Buffalo, NY into the upper Niagara River near its source, Lake Erie (Figure 2).The highly modified lower 18 km of Tonawanda Creek is channelized for navigation as part of the Erie Canal, which has served as a conduit for numerous invasive species that have spread across NY.Several Lake Erie-resident species migrate to spawn at a riffle in the natural stream bed of LTC at the upstream end of our study area (Figure 2), where NS were observed spawning in the past.There are records for 68 fish species in LTC (see database connected to Carlson et al. 2016).

Field sampling and specimen handling
Intensive boat electrofishing (Smith-Root GPP 5.0 system) caught thousands of fish in LTC during our study.(The same system was used in 2005; see Discussion: Historical catches of NS in the study area.).Sampling occurred during 8 h of 15-min power-on shocking runs, after which captured fish were identified and counted.Catches were converted to catch-per-unit-effort by dividing shock-run total counts by the seconds of power-on shock time, then Log10 (n + 1)-transformed to minimize skewing of the data by a few highly abundant species.
Lepomis species and their apparent hybrids collected in LTC (n = 127), plus 20 NS collected from NY State Department of Environmental Conservation (NYSDEC) rearing ponds, were identified using a key derived from two standard fish keys (Table 1; Sanderson-Kilchenstein 2015).Specimens that received both genetic and morphometric-meristic (Mo-Me) analysis included 32 apparent hybrids, 15 Pumpkinseed (L.gibbosus Linneaus, 1758; PK), 20 Bluegill (L.macrochirus Rafinesque, 1819; BG), 15 Green Sunfish (L.cyanellus Rafinesque, 1819; GS), and 19 NS.These 91 fish received a left pelvic fin clip then each fin was stored for genetic analysis in a separate labeled vial with 95% ethanol and delivered to Cornell University.All sunfish were frozen and delivered to the NY State Museum for morphometric measurements (n = 29) and meristic counts (n = 14) (Table 2).

Genetic identification
We extracted DNA (MasterPure TM DNA purification kit; Epicentre, Madison, WI, USA) and performed mitochondrial (mtDNA) and nuclear DNA analyses on Lepomis individuals.Mitochondrial DNA is maternally inherited and often used to distinguish closely related species.In hybridization studies, mtDNA genetic markers have been used to identify the maternal parent of hybrid specimens (cf.Seixas et al. 2018).However, the genetic characterization of putative hybrids requires the identification of fixed or nearly fixed allelic differences between parental species in the nuclear genome (cf.He et al. 2018).Here we focused on a single mitochondrial gene (COI) and 8 single-copy nuclear genes (Enc1, Glyt, Plagl2, ptr, sidkey, Tbr1, Cal, LmegA, and rag1) that, based on NCBI publicly available datasets (https://www.ncbi.nlm.nih.gov/),showed differences among the study species.To assess whether the reported genetic differences were indeed nearly fixed among the parental populations studied, we first selected eight individuals from the four putative species to screen each locus for robust PCR using genomic DNA as the template.After testing for co-amplification, all loci were used to genotype specimens in single multiplex PCR groups using the QIAGEN Multiplex PCR Kit (www.qiagen.com/us/).Individual specimens and loci were barcoded using Illumina TM S5 (www.qiagen.com/us/) and N7 Nextera TM (www.illumina.com)primers following the manufacturer's specifications.Amplicons were paired end sequenced (2 × 250 bp) on an Illumina MiSeq (www.illumina.com)at Cornell University's BioResource Center.
Post-sequencing data processing required extracting raw reads from the Illumina MiSeq sequencing run and assigning them to the appropriate specimen and allele locus with a custom Perl script that discarded low quality reads, trimmed adapters, overlapped paired-reads, identified reads corresponding to each locus, collapsed identical reads for each fish, and identified the two most frequent haplotypes for each fish at all loci (quality score: Q12, minimum overlap: 20, mismatch rate: 0.05; D' Aloia et al. 2017).At any given locus, an individual was scored as heterozygous if the minor allele frequency was ≥ 20%.Individuals that failed to amplify in more than 50% of the loci were excluded from analyses, and any fish with fewer than 10 total reads per allele was considered missing data (D' Aloia et al. 2017;Posso-Terranova and Andrés 2018).For each locus, individual haplotypes (i.e.alleles) were aligned to a reference using the ClustalW algorithm (Thompson et al. 1994) as implemented in DNA* (www.dnastar.com/t-dnastar-lasergene.aspx).The resulting multi-locus genotypes were used to calculate an admixture proportion (hybrid index) for each sampled individual as estimated by Structure (Pritchard et al. 2000).This index measures the relative probability that a composite genotype is the result of random  29) and meristic counts ( 14) used to identify the four Lepomis species and their hybrids in this study (l = lower; u = upper).

Morphometric characteristics
mating within each of the two parental species.The index ranges from 0 to 1 and will be close to one of these values when an individual genotype has a high probability of belonging to either parental species.The expectation is that genotypically or phenotypically intermediate individuals (hybrids) would have an index score intermediate to the parental species, for example, ~0.5 for first generation (F1) hybrids.

Morphometrics and meristics
To standardize fish proportions we divided each fish's morphometric characteristics by its standard length and excluded standard and total lengths from our analysis.Meristic counts (Table 2) for each fish were used directly without transformation.Compared to Cluster Analysis and PCA, Canonical Analysis of Principal Components (CAPC, Clarke and Gorley 2006) best discriminated the four Lepomis parental species and their four hybrid combinations.Similarity Percentages (SIMPER, Clarke and Gorley 2006) calculated similarities between the four species and their four hybrids revealed by CAPC, as well as dissimilarities between the 28 paired combinations of species and hybrids.Assuming that the genetic results were the correct identifications, we used Sign tests (Statistix 10 2013) to determine the accuracy of Mo-Me and keyed identifications.

Genetics
Based on nuclear genes, the number of genetic clusters (i.e.groups of individuals more related to each other than to any other sampled individual) was visualized using Discriminatory Analysis of Principal Components (DAPC; Jombart et al. 2010).DAPC transforms data using a Principal Component Analysis (PCA) before summarizing genetic variance between and within groups (i.e. a discriminant analysis [DA]; Jombart et al. 2010).The optimal number of demes (K), as described in Jombart and Collins (2015), was the one showing the lowest Bayesian Information Criterion (BIC).Then we used DAPC to assign individuals into parental and hybrid groups, retaining the number of principal components using 85% of the cumulative deviance.

Fish community comparisons
Because NS were caught in 2005, and not in 2013, we explored whether the fish communities had changed between sampling years using a Non-metric Multidimensional Scaling analysis (NMDS; Clarke and Gorley, 2006); ANOSIM was used to distinguish the 2005 and 2013 fish communities.Simpson's Index of Diversity for the 2005 fish community was compared with the 2013 fish community using an adapted Student's t-test (Brower and Zar 1984).

Results
We found no NS among 11,335 Lepomis sampled and identified in 2013; we did collect ~35 suspected hybrid Lepomis from this large sample.(Supplemental Appendix 1 contains Lepomis identification results (keyed, Mo-Me, genetic).

Morphometric and meristic identification
Four distinct clusters of Lepomis species and four hybrids (female parent listed second) were revealed by CAPC, with good separation among species and hybrids except for PK mixed with both BG and NS (Figure 3).Cluster 1 was predominately NS (n = 19) and BG × NS (n = 6), along with one BG and one BG × PK.Cluster 2 was predominately BG (n = 16) and BG × PK (n = 7), along with one BG × NS.PK were spread between clusters 1 (n = 5) and 2 (n = 6).Cluster 3 contained 11 BG × GS and three GS × PK, and cluster 4 contained 13 GS and two BG × GS (Table 3).(Supplemental Appendix 2 contains raw, transformed, and tabulated data for morphometric measurements and meristic counts).
SIMPER showed that five meristic characteristics accounted for 56.8-61.1% of the similarity within the species and hybrids (5: lateral line scales, 8: circumbody scales at the dorsal fin origin, 10: circumpeduncle scales, 11: predorsal scale rows, and 13: post anal scale rows; Table 2).These five characteristics plus seven other meristic characteristics (1, 2, 3, 4, 7, 12, and 14; Table 2) accounted for 91.9-92.5% of the similarity within the eight Lepomis groups (Table 4).Two meristic (6 and 9) and the 27 morphometric characteristics used for analysis (Table 2; standard and total lengths were not used) accounted for only 7.5-8.1% of the similarities within taxa and were judged non-essential for distinguishing our Lepomis species and their hybrids.

Genetic identification
The mtDNA results identified the female parent of Lepomis hybrids, and nuclear DNA revealed that at the selected loci putative "pure" individuals showed fixed genetic differences, with the exceptions of four allele introgressions: BG Cal1 allele in one NS, PK Plagl2 allele in one BG, BG Plagl2 allele in one GS, and NS LmegA allele in one GS (Table 6).These introgressions are consistent with either ancestral polymorphism or introgression from hybridization in previous generations.Among the four hybrid combinations we identified, their mean hybrid indexes ranged from 0.43 to 0.50, indicating that the hybrid individuals were first generation (F1) crosses of the parental species (Table 6).BG alleles (male parent) were detected in three of the four hybrid genotypes: BG × GS, BG × NS and BG × PK.Male GS alleles were detected in the fourth hybrid genotype: GS × PK (Table 6).All NS hybrids had a female NS parent.(Supplemental Appendix 3 contains genetic results).Consistently, our DAPC analyses showed that the BIC plateaus at K = 7-10 (Supplemental Figure 1).This strongly suggests that the genetic subdivision of our dataset into seven main genetic groups is correct: four parental population clusters (1: NS, 3: BG, 5: PK, and 7: GS) and three mixed ancestry clusters (2: BG × NS, 4: BG × PK, and 6: BG × GS) (Figure 4).Although three GS × PK were identified by mt DNA, only one had sufficient nuclear DNA reads for DAPC (#8; Figure 4).After the 7 groups were assigned, we determined the optimal number of PCs to retain using the accuracy-score approach (Jombart and Collins 2015;Pramoditha 2022) to avoid over-sacrificing information or over-fitting in the subsequent DAPC.The first two principal components of the DAPC explained 40% of this variation and were sufficient to capture the genetic structure in our dataset (Figure 4).The mean cluster membership probabilities based on the retained discriminant functions were ≥96%, which clearly showed discrete genetic clustering among parentals and hybrids.

Genetic, morphometric-meristic, and keyed identification comparisons
Assuming genetic identifications of 91 individual Lepomis and their hybrids were 100% accurate, we compared them with Mo-Me and keyed identifications and, also, compared Mo-Me with keyed identifications.For the genetic-Mo-Me comparison, 82.4% (75/91) of the identifications were the same, 13.2% (12) of the Mo-Me results misidentified one hybrid parent, and 4.4% (4) did not identify either parent correctly (p < 0.0001, Sign Test).For the genetic-keyed comparison, 80.2% (73/91) of identifications were the same, 13.2% (12) of the keyed results misidentified one hybrid parent, and 6.6% (6) did not identify either parent correctly (p < 0.0001).For the Mo-Me-keyed comparison, 87.9% (80) of identifications were the same; Mo-Me (4.4%, 4) and keyed (3.3%, 3) misidentified one hybrid parent (p = 1).MoMe and the key also failed to identify the same four single species individuals vs. their genetic identity.Compared with genetic identification of the four Lepomis species and four hybrids we collected, Mo-Me and keyed identifications averaged 81.3% correct.(Sign test data and results are in Supplemental Appendix 1).

Fish community comparisons: 2013 versus 2005
Total species richness at the same locations where NS were caught in 2005 (n = 25), but not caught in 2013 (n = 24), was the same, while average species richness per sample was higher in 2005 (n = 13.4) than in 2013 (n = 10.9). Simpson's Diversity was significantly higher (t = 3.05, 47 df, p = 0.004; α = 0.05) at the locations where NS were caught in 2005 (0.790) than it was at the same locations in 2013 (0.715).Despite similar species richness in 2005 and 2013, significant changes in the relative abundance of fish community members in LTC occurred between 2005 and 2013 (Figure 5; ANOSIM Global R = 0.806, p = 0.001; 2D stress level = 0.2, which is at the lower end of the range needing cautious interpretation).

Lepomis hybridization
Keyed, morphometric-meristic, and genetic methods all identified hybrid Lepomis individuals.Mo-Me analysis revealed distinct clusters of species and their hybrids, except for PK that were evenly distributed between the BG and NS clusters.Five meristic characteristics accounted for most of the similarity among the species and hybrids: lateral line scales, circumbody scales at the dorsal fin origin, circumpeduncle scales, predorsal scale rows, and post anal scale rows.Dissimilarity among the species and hybrids ranged from 3% to 12%, not a surprising result for four closely related members of the same genus.Among the 28 Lepomis species and hybrid pairs, five meristic characteristics accounted for a majority of their dissimilarity: circumbody scales at the dorsal fin origin, lateral line scales, circumpeduncle scales, predorsal scale rows, and scale rows below lateral line.The 27 morphometric characteristics we examined accounted for less than 10% of similarities and dissimilarities among species and hybrids, so only meristic data are needed to identify the taxa we studied.The same four hybrids were identified by meristic and genetic methods, while keyed identifications could only determine that a specimen appeared to be a hybrid.The Mo-Me and keyed methods correctly identified species and hybrids more than 80% of the time.Three of the four hybrids involved male BG which had mated with female NS, PK, and GS; the fourth hybrid was GS x PK.All crosses (importantly, BG male x NS female) were F1 progeny.Given that no NS, with a 4-8-year life span (Scott and Crossman 1973) were found in 2013, this indicates that after 2009 (when the last NS was caught in LTC) no NS pairs were able to breed.
Agreement among species identifications using keys, body measurements and counts, and genetic signatures from mtDNA and nuclear DNA was reasonably good (>80%).The low cost of modern genetic discrimination of taxa using fin tissue, which does not require sacrificing fish, may make Mo-Me studies cost-ineffective given the labor required to measure and count many characteristics (43 in our study; Table 2) of whole fish with no better identification accuracy than a species-specific key (Table 1) constructed from published keys, an approach that could be used for any group of closely related fish like Lepomis.
Natural and artificial hybridization have been well studied among Lepomis species (cf.Hubbs 1920;Hubbs and Hubbs 1932;Childers 1967;Keenleyside et al. 1973;Bolnick and Near 2005;Bolnick 2009).Bouton (1994) proposed that hybridization was a potential cause for the decline of NS in NY state.Hybridization can be common in habitats that are degraded or when a new species is introduced (COSEWIC 2016), and when hybrids are better adapted to or forced to occupy a different niche than their parental species.An example of the latter was reported by Hubbs (1955) for a stream in Michigan where the Lepomis community was 95% GS × BG hybrids vs. a downstream pond occupied by balanced parental populations and few hybrids.Like all hybrid Lepomis in our study, most hybrids that Hubbs observed were intermediate between the parental species (F1 generation), but he also saw hybrids that were intermediate between the F1 hybrids and the parental species, indicating that some of the normally sterile, predominately male hybrids in his stream were fertile.Hubbs (1955) also demonstrated through laboratory breeding experiments that previously described Lepomis species were hybrids: L. euryorus Mckay, 1881 was GS × PK, and L. ischyrus Jordan & Nelson, 1877 was GS × BG.We found the same hybrids in LTC; all were F1 based on their hybrid index scores.
Mechanisms for hybridization among Lepomis species in nature also stem from complex breeding behavior in the genus.NS and their conspecifics are colonial nesters, and interspecific colonies of Lepomis are common, especially when gravelly nesting substrate is limited like in LTC.Jennings and Philipp (2002) observed frequent interspecific nest intrusions in a nesting colony of Longear Sunfish and Redear Sunfish (L.microlophus, Pallas, 1814).Both small sneaker males, which mimic females with dull coloration to sneak past male nest guarders, and larger nesting males were observed intruding on adjacent nests of the other species.Similar to our results, Garner and Neff (2013) found unidirectional hybridization caused by BG sneaker males intruding on PK nests in a lake in southern Ontario, Canada.
The conditions that permitted BG x NS hybridization in LTC could be that female NS resemble small female BG and are accepted by spawning male BG, or nesting pairs of NS were intruded upon by larger nesting or smaller sneaker male BG.Any of these situations would explain why all NS F1 hybrids in our study had BG fathers and NS mothers.Clearly, female NS were present in the previous generation of the BG × NS we captured.Also, NS males likely had problems finding conspecific mates due their inability to stay on spawning habitat due to high competition for it with abundant and larger BG and GS males.Allee effect factors that cause reductions in the growth rates of very small populations as they decline also may have contributed to the observed hybridization of NS in LTC (Drake and Kramer 2011).
The GS is native to North America, was not detected in NY until 1935, and was not found in LTC until 1998, although it entered other Lake Erie tributaries starting in 1970 (Carlson et al. 2016).When introduced it has caused problems for fish communities, including local extirpations of native centrarchids (Moyle 1976) and cyprinids (Lemly 1985), and declines of threatened (Karp and Tyus 1990) and other fish species (Dudley and Matter 2000;Meador 2020).Initially, we suspected that the high abundance of GS in LTC (Wells and Haynes 2007;Sanderson-Kilchenstein 2015), coupled with its ability to disrupt fish communities, was likely a major driver in the decline of NS in LTC.However, the NS hybrids we analyzed genetically (n = 9) all had a native male BG parent, suggesting that NS have not hybridized with GS in LTC.From 1974From -2004From , 2006From -2012From and 2015From -2016, beach seines (6.25 mm mesh, with or without bags; various lengths and heights) and backpack electrofishing units (various models) were used to sample (unstandardized effort) accessible areas of the last 6 km in LTC before it enters the Erie Canal and nearby canal wide waters (Table 7).From 1974 to 2004, 13 NS were caught in the lower 6-km reach of LTC and five were caught in two wide waters along the 13-km reach of the Erie Canal downstream from its confluence with LTC (Table 7).Eight putatively hybrid specimens caught from 1999 to 2004 received Mo-Me analysis; two were identified as NS x PK (female and male parental species unknown), while the others had uncertain species parentage.

Historical catches of NS in the study area
In July 2005, 22 NS were caught in the lower 6-km reach of LTC (Figure 2), and one was caught in the Erie Canal wide water 6 km downstream from the confluence with LTC (3 h total power-on effort) using the same electrofishing system as this study (8 h total power-on effort) that caught no NS (Sanderson-Kilchenstein 2015).Given that all sampling in 2005 took place in July, the peak spawning season for NS (Scott and Crossman 1973), it is likely that some NS were caught more than once at the same spawning sites, which would have suggested a larger NS population than was in LTC.Also in 2005, several NS spawned while briefly in captivity at SUNY Brockport then were returned alive to LTC.
From 2006 to 2009, 18 NS were collected in LTC and nine in the Erie Canal wide water 6 km downstream from the confluence with LTC (Table 7).Among the 27 captured NS, seven and one were returned alive to LTC and the Erie Canal, respectively, and 18 were collected.Two were sacrificed for disease testing and 16 were used to develop an LTC brood stock in a NYSDEC rearing pond.No NS were caught during four efforts in LTC from 2010 to 2012.In 2013 we sampled an abundant (>40,000 fish) and diverse (25 species) community in LTC, not including minnow (F.Cyprinidae), redhorse (Moxostoma) and darter (Percina and Etheostoma) species, but no NS.GS catches were consistently high in LTC and low in the Erie Canal wide water 6 km downstream from the confluence with LTC, suggesting that GS may not be a primary cause of NS extirpation in our study area.(2), 2007 (11), 2008 (3), 2009 (2) tonawanda creek, 2 km upstream from ltc study area 1990 (1) erie canal, 6 km downstream from the ltc confluence 1988 (1), 1999 (1), 2008 ( 9) erie canal, 13 km downstream from the ltc confluence 2001 (2) a high-effort electrofishing (boat and backpack) and beach seining during the peak spawning season in July with repeat captures likely.

Additional factors that might account for NS extirpation
COSEWIC (2016) suggested three important limiting factors for NS that also characterized its situation in our study area.
1. Restricted spawning and juvenile habitat: Historically, NS were found only in a 6-km reach of LTC and one small wide water off the Erie Canal.2. Low dispersal capacity, which slows recovery following depopulation and diminishes potential for population rescue: There are no potential NS source populations elsewhere in the Tonawanda Creek watershed and Erie Canal.3. Low tolerance of turbidity: Tonawanda Creek and the Erie Canal are quite turbid.
These and other factors are addressed below.

Fish community changes
Fish communities, using the same gear and standardized methods, were different in 2005 (NS caught; Wells 2009) and 2013(Sanderson-Kilchenstein 2015;no NS caught).Community differences were driven by large changes in GS abundance (+941% CPUE; 1.2 fish per meter of LTC shoreline) and BG abundance (+260% CPUE) in 2013 vs. 2005, both indicating that the Lepomis community overall increased greatly in only 8 years.Similar fish community disturbance by high GS abundance was reported for urban streams in the southeastern U. S. (Meador 2020).
From 2005 to 2013, Northern Pike (Esox lucius Linneaus, 1758) increased in abundance by +220% CPUE and Round Goby (RG; Neogobius melanostomus Pallas, 1814) by +200% (Sanderson-Kilchenstein 2015).NS were the smallest Lepomis species, in size and number, occupying LTC and the nearby Erie Canal, and it is likely that they did not fare well due to increased interspecific competition and higher abundance of Northern Pike, a major centrarchid predator (Scott and Crossman 1973).Decreases in Smallmouth Bass (Micropterus dolomieu LaCepede, 1802) and redhorses, −78% and −48%, respectively, and the near disappearances of darters (−90%) and the Brindled Madtom (Noturus miurus D. S. Jordan, 1877; −96%) also were found in 2013 (Sanderson-Kilchenstein 2015).The decrease in Simpson's Index of Diversity from 2005 to 2013 in LTC also supports the idea that the proportions of species in the LTC fish community changed over 8 years.
The 200% increase in CPUE of the non-native RG from 2005 to 2013 in LTC could have contributed to the NS decline in LTC.This goby has had multiple negative impacts in tributaries of the Great Lakes, including predation on fish eggs and increasing competition for food (Pennuto et al. 2010;Poos et al. 2010;Kornis et al. 2012;COSEWIC 2016), but there is no evidence that the RG adversely affects Lepomis species.It is small, benthic, and has no swim bladder, so boat electrofishing in sometimes deep, always turbid water like LTC is not an efficient method for capturing them (Kornis et al. 2012).Undoubtedly, the true population size of RG was larger than our sampling suggested.RG were first caught in LTC in 2002, and subsequent seining by the NYSDEC showed that frequency of occurrence (# seines containing RG ÷ total number of hauls) had increased more than 80% by 2011 (Carlson, unpublished data).

Watershed disturbances
Important threats to NS in the Great Lakes-Upper St. Lawrence region of Canada are elevated levels of turbidity and contaminants from agriculture and development, especially elevated chloride (COSEWIC 2016).The Tonawanda Creek watershed has a long history of human disturbance (Wells 2009), including deforestation beginning in the late 1700s followed by agriculture (grain, row crops and orchards with some combination of manure, inorganic fertilizer, pesticide, and herbicide applications) from the early 1800s to mid-1900s.From the mid-1900s until today, dairy farms, cattle feed lots, and residential development with septic systems increased while traditional agriculture declined.Eutrophication and elevated levels of chloride from winter road salt application in the Tonawanda Creek watershed is pronounced.Human disturbance of land around Tonawanda Creek also makes it easy for the highly erodible Niagara loam soils of western New York to wash into it after heavy rain and rapid snow melt.Accordingly, LTC is very turbid with Secchi depths mostly <0.3 m, a habitat situation not favorable for NS whose habitat and water quality preferences were reviewed by Wells (2009) andCOSEWIC (2016).
Surveys for suitable NS habitat in the watersheds of Tonawanda, Johnson, and Oak Orchard creeks (Figure 2) were conducted in 2004 (Wells and Haynes 2007;Wells 2009).Based on literature review, habitat suitability for NS at 364 sites was judged by four criteria: predominant substrate composition (most suitable: silt, sand, small gravel), presence (suitable) or absence (unsuitable) of emergent/submerged aquatic vegetation, in-stream wood (more is better), and bank cover (more is better).Sampled sites were ranked from 1 (low quality: only one high quality criterion met) to 4 (all four high quality criteria met).Sites meeting all four high quality criteria were found in Tonawanda Creek (4), Johnson Creek ( 16), and Oak Orchard Creek (10).However, no NS were caught at these "high quality" sites; they were caught only in the last 6 km of LTC (n = 22) and one nearby Erie Canal wide water (n = 1) in 2005.

General decline of NS in North America and specific changes in LTC
Except for the recently discovered NS population in the Chazy River of northeastern NY state (Maxwell and Carlson 2018;Carlson 2019; Figure 1), there are no known populations in NY today (Carlson et al. 2016) Despite long-term water quality problems, fish community changes due to changing proportions of naturalized (e.g.BG, northern pike) and invasive (e.g.GS and RG) species, and hybridization with other Lepomis, the NS was caught consistently in LTC since their discovery in 1974, but two things happened after 2005.First, 67% of the NS caught from 2006-2009 were removed from LTC.None were caught there after 2009, except for a few BG x NS F1 hybrids in 2013.Second, in 2005, NS were collected and observed spawning on a small patch of gravelly substrate at the confluence of Mud Creek and LTC (Figure 2; Wells 2009).In 2012, a residential property on the steep north bank of Mud Creek 60 m upstream from LTC had a landslide that blocked the flow of Mud Creek for several weeks.This event delivered much sediment to the backwater eddy at the confluence and buried the small gravelly area suitable for NS spawning.By 2013, the silt deposits at this location had been washed away but we caught no NS there; seining in 2016 and 2018 also caught no NS there.
The two factors described above may have caused NS in LTC to fall below their minimum viable population size, or the number of individuals required for a high probability of a population surviving over a long period of time (Soule and Wilcox 1980).Although controversial (cf.Frankham et al. 2014;Franklin et al. 2014), the "50/500 rule" states that at least 50 adults of a species (or population) are required to avoid damaging effects of inbreeding, and at least 500 individuals are needed to avoid extinctions due to the species' (or a population's) inability to evolve to cope with environmental change (deterministic or stochastic).The rule is a commonly used metric in conservation biology to assess whether a population's size is viable long-term.Given the very small catches of NS in LTC and the nearby Erie Canal from 1974 to 2009 (Table 7), it is likely that the population never came close to meeting the 50-500 rule, yet a small NS population persisted for 35 years after its discovery.

Northern Sunfish stocking
The NYSDEC initiated a recovery stocking program in 2005.Three strains of NS, genetically evaluated by Near (2008), were established in separate production ponds at the Perch River Wildlife Management Area to prevent genetic exchange among the strains: NS from our LTC study area (Lake Erie basin; n = 16 from 2006 to 2009), Huron River near Detroit, Michigan (Lake Erie basin; n = 80 in 2005), and Moira River near Tweed, Ontario, Canada (Lake Ontario basin; n = 140 in 2006) (Carlson 2022, unpublished data).Subsequently, a separate rearing pond at a NYSDEC hatchery near eastern Lake Erie was established with NS from the LTC-strain pond at Perch River.Fourteen streams in central and western NY with historical records of NS, as well as others identified by Wells (2009) and NYSDEC as having potentially suitable habitat for NS in NY state, were stocked with more than 15,000 NS fingerlings from 2007 to 2015 (Carlson 2022, unpublished data).Of those, 5,100 fingerlings from the LTC brood pond were stocked from 2008-2015 in Tonawanda Creek, wide waters in the adjacent Erie Canal, and other streams in western NY (Marsh, Johnson, Murder, Cayuga, and Ellicott; Figure 2).None were stocked at our LTC study site to avoid potential harm to any potentially remaining wild NS (e.g.genetic introgression from the pond-raised NS).
Irregular follow-up sampling from 2008 to 2014 caught 7 stocked NS, some in spawning condition, in Oak Orchard Creek (2 sampling trips, one NS caught in 2014), Cayuga Creek (17 sampling trips, four NS caught from 2008-2012), Murder Creek (15 sampling trips, one NS caught in 2010), and Ellicott Creek (12 sampling trips, one NS caught in 2012) (Figure 2; Carlson 2022, unpublished data).Stocking in Ellicott and Cayuga Creeks, ~14 and ~28 km downstream from LTC's junction with the Erie Canal, respectively (Figure 2), occurred from 2007 to 2009.There were no records of NS in these creeks before stocking.In 2012 one, 2-year-old NS, that could not have been stocked, was caught in Cayuga Creek (Carlson 2022, unpublished data).No captures of NS adults or juveniles occurred in any stocked streams after 2014.Sodhi et al. (2009) summarized drivers leading to the extinction of rare species (or extirpations of small populations): changes in land use (habitat loss, degradation, or fragmentation), over exploitation, invasive species, disease, climate change, pollution levels, and catastrophic events.Any or a combination of the italicized factors in the previous sentence occurred in LTC and could have contributed to the extirpation of NS there.

Conclusion
We will never know how long NS might have survived in LTC without (1) the landslide in 2012 that smothered a good spawning area near the mouth of Mud Creek (Figure 2); (2) removal of 67% of the NS caught from 2006 to 2009 in LTC and the nearby Erie Canal for brood stock to support stocking efforts in western NY streams with apparently suitable NS habitat; and 3) substantial increases and decreases of some species in the LTC fish community from 2005 to 2013.We do know that NS in LTC had coexisted with GS, RG, and a community of native and naturalized fishes, albeit in changing proportions, since 1974.Given that intensive boat electrofishing conducted in 2013-2014 by Sanderson-Kilchenstein (2015) found no NS, it is likely that the NS is extirpated in LTC as well as in the entire Tonawanda Creek watershed surveyed in 2005 (Wells 2009).
Our findings offer a cautionary tale for others who are trying to conserve or restore a rare species or population.First, a natural-or human (like the one in 2012 that buried an important NS spawning habitat in LTC for at least a year)-caused event might send a small population into a downward spiral from which it cannot recover.Second, just because a species is barely persisting in a historical habitat like LTC that is degraded or changing in multiple ways does not mean that it can be removed, bred in captivity, and returned successfully.But this approach has not been tried in LTC and is the only hope left to restore NS in western NY.

Management recommendation
A year after the 2012 landslide in Mud Creek buried an adjacent NS spawning habitat in LTC, fluvial processes had restored the small area of gravel where NS were observed spawning in 2005.NS originally from LTC still reside in two ponds in NY state.In line with a draft management plan (Wells and Carlson 2006, unpublished), descendants from wild LTC NS should be stocked there.If removing 67% of the NS caught in LTC in 2006-2009 to establish a rearing pond population was the primary reason for their extirpation, the only way to find out whether NS can be restored in LTC is to stock their pond-reared descendants there, and in the canal wide water where they were last found, then determine if they survive and produce recruits.This would be done after fin clipping and stocking in May then sampling using standardized effort in May, July, and September of the same year to determine survival rates, then again in July of the following year.If NS are caught the following year, sampling should continue intermittently in future years so that CPUE and recruitment can be compared across years, things not done in the past, to quantitatively follow NS population size and stability in the two stocking areas.

Figure 1 .
Figure 1.Map of northern sunfish distribution in the united states and canada modified by Daemin Kim (Department of ecology and evolutionary Biology, yale university, new haven, connecticut 06520, usa) from figure 1 in Kim et al. (2022).fish specimen capture locations used by Mr. Kim for this figure were accessed through the fishnet2 Portal (http://www.fishnet2.net),which contained museum records for northern sunfish from 25 institutions dating from 1899-2020.the great chazy river (▲) and lower tonawanda creek (■) sites for northern sunfish in ny state were added by Mr. Kim in september 2022.

Figure 2 .
Figure 2. Map of northern sunfish distribution (1974-2009) in lower tonawanda creek (ltc) and the adjacent erie canal (hash marks), streams in western new york tributary to lake ontario and the niagara river with historical populations of ns in 1939 (*), and locations (o) with apparently suitable habitat where ns were stocked from 2006 to 2015 and surveyed from 2007 to 2018.

Figure 3 .
Figure 3. Morphometric proportions and meristic counts were used to distinguish parental Lepomis species and their hybrids in caPc (canonical analysis of Principal component) space.Polygons represent ≥ 96% similarity among the individual fish within them.symbols are genetic identifications of the same individual fish.

Figure 4 .
Figure 4. nuclear Dna was to distinguish parental Lepomis species and their hybrids in DaPc (Discriminatory analysis of Principal components) space.Polygons represent ≥96% similarity among the individual fish within them.symbols are genetic identifications of individual fish.
. The number of NS populations in the Provinces of Ontario and Quebec, Canada, also have declined, especially in northwestern Ontario near the state of Minnesota (from ~31 before 1985 to ~10 from 1985-2004 to three from 2005-2015; COSEWIC 2016) and southwestern Quebec in the Ottawa-Montreal region near the St. Lawrence River (from ~19 before 1985 to ~6 from 1985 to 2004 to one from 2005 to 2015; COSEWIC 2016).The NS is Secure in Michigan (NatureServe 2022), and appears stable in the drainages of lakes Ontario, Huron, and Erie in southern Ontario: before 1985, ~48 NS populations were known; between 1985 and 2004, ~43 populations were known; and from 2005 to 2015 ~ 40 populations were known (COSEWIC 2016).

Table 1 .
Key used for field and laboratory identification of the four co-occurring sunfish species in lower tonawanda creek developed from smith(1985)and scott and crossman(1973), with input from coauthor carlson.

Table 4 .
Percent similarity within four Lepomis species and their hybrids based on 14 meristic attributes.

Table 5 .
Percent dissimilarity between 28 pairs of four Lepomis species and their hybrids based on 14 meristic attributes.

Table 6 .
Maternal parents (mtDna) and nuclear Dna introgressions among four Lepomis species and their hybrids in lower tonawanda creek (f = female, m = male).

Table 7 .
northern sunfish records from tonawanda creek and the erie canal, 1974-2009, with number caught in parentheses.