ARTICLE

Metapopulation structure of a semi-anadromous fish in adynamic environmentFrederick Feyrer, James Hobbs, Shawn Acuna, Brian Mahardja, Lenny Grimaldo, Melinda Baerwald,Rachel C. Johnson, and Swee Teh

Abstract: The Sacramento splittail (Pogonichthys macrolepidotus) is a relatively large (400 mm), long-lived (8 years) demersalcyprinid of conservation importance endemic to the San Francisco Estuary (SFE), California, USA. It exhibits a semi-anadromouslife cycle spending adult life in low to moderate salinity (0–12) habitat with migrations into upstream freshwater rivers andfloodplains for spawning during winter–spring. The species persists as two genetically distinguishable populations — onedominant and one subordinate — separated by discrete spawning habitats that we suggest resemble an island–mainlandmetapopulation structure. The populations overlap in distribution in the SFE, yet segregation is maintained with individualstending to aggregate or school with others of similar population heritage and natal origin. The populations are spatiallyconnected via dispersal of the dominant population into the subordinate population’s spawning habitat when climate patternsproduce freshwater outflow sufficient to form a bridge of suitable low salinity habitat across the upper SFE. Habitat affinities ofthe two populations, hydrodynamic modeling studies, and historical outflow records together suggest such conditions occur inapproximately 1/3 of years overall with an irregular frequency. This dynamic pattern of spatial connectivity controlled by climatevariability may be an important driver of gene flow between the two populations.

Résumé : Pogonichthys macrolepidotus est un cyprinidé démersal longévif (8 ans) relativement grand (400 mm) d’importance pourla conservation et endémique de l’estuaire de San Francisco (SFE; Californie, États-Unis). Il présente un cycle biologique semi-anadrome, passant sa vie adulte dans des habitats de salinité faible a modérée (0–12) avec des migrations dans des rivières d’eaudouce et des plaines alluviales situées plus en amont pour frayer a l’hiver et au printemps. L’espèce persiste en deux populationsgénétiquement distinctes, une dominante et l’autre subordonnée, séparées par des habitats de frai distincts qui ressemblent,selon nous, a une structure de métapopulation de type île-continent. Si les aires de répartition des populations se chevauchentdans le SFE, une ségrégation est maintenue, les individus tendant a se regrouper ou former des bancs avec d’autres individusprovenant de la même population et de la même origine natale. Les populations sont connectées dans l’espace par la dispersionde la population dominante dans l’habitat de frai de la population subordonnée quand les aléas du climat produisent des débitssortants d’eau douce assez importants pour former un pont d’habitats d’assez faible salinité de part en part du SFE supérieur. Lacombinaison de l’affinité des habitats des deux populations, d’études de modélisation hydrodynamique et des registres histo-riques des débits sortants indiquerait que de telles conditions se produisent environ une année sur trois, a une fréquenceirrégulière. Ce motif dynamique de connectivité spatiale contrôlée par la variabilité du climat pourrait être une importantecause de flux génétique entre les deux populations. [Traduit par la Rédaction]

IntroductionThe extent to which discrete populations are connected by mi-

gration of individuals can substantially influence local and overallpopulation dynamics, as well as the evolution and risk of extinctionof a species. There is a great diversity of dispersal mechanismssupporting population connectivity across taxonomic groups andhabitats. For example, freshwater invertebrates disperse actively asadults or passively in other life stages by water or animal vectors(Bilton et al. 2001), and marine populations are frequently connected

by dispersal of planktonic larvae transported by ocean currents(Cowen and Sponaugle 2009). Understanding the mechanisms,scales, and outcomes of population connectivity under present andfuture climate conditions are crucial for guiding policy and manage-ment decisions, especially for at-risk species.

Metapopulation theory serves as a useful framework for eval-uating how spatial structures function to influence local andoverall population dynamics (Hanski and Simberloff 1997), es-pecially in aquatic environments (Dunham and Rieman 1999;Kritzer and Sale 2004). Yet, it is challenging to collect the necessary

Received 25 September 2014. Accepted 18 December 2014.

Paper handled by Associate Editor Aaron Fisk.

F. Feyrer.* Bay Delta Office, Bureau of Reclamation, 801 I Street, Sacramento, CA 95814, USA.J. Hobbs. University of California, Davis, Department of Wildlife, Fish, and Conservation Biology, One Shields Avenue, Davis, CA 95616, USA.S. Acuna. Metropolitan Water District of Southern California, 1121 L Street, Suite 900 Sacramento, CA 95814, USA.B. Mahardja† and M. Baerwald. University of California, Davis, Department of Animal Science, One Shields Avenue, Davis, CA 95616, USA.L. Grimaldo. ICF International, 620 Folsom Street, Suite 200, San Francisco, CA 94107, USA.R.C. Johnson. Fisheries Ecology Division, Southwest Fisheries Science Center, National Marine Fisheries Service, NOAA, 110 Shaffer Road, Santa Cruz,CA 95060, USA; and University of California Davis, Department of Animal Science, One Shields Avenue, Davis, CA 95616, USA.S. Teh. Aquatic Health Program, Department of Anatomy, Physiology and Cell Biology, School of Veterinary Medicine at the University of CaliforniaDavis, Davis, CA 95616, USA.Corresponding author: Frederick Feyrer (e-mail: ffeyrer@usgs.gov).*Present address: US Geological Survey, California Water Science Center, 6000 J Street, Placer Hall, Sacramento, CA 95819-6129, USA.†Present address: Aquatic Ecology Section, California Department of Water Resources, 3500 Industrial Blvd., West Sacramento, CA 95691, USA.

Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)

1

Can. J. Fish. Aquat. Sci. 72: 1–13 (2015) dx.doi.org/10.1139/cjfas-2014-0433 Published at www.nrcresearchpress.com/cjfas on 8 January 2015.

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

.S. G

eolo

gica

l Sur

vey

on 0

3/12

/15

For

pers

onal

use

onl

y.

demographic data to elucidate that populations are in factfunctioning as metapopulations. The majority of empirical stud-ies have been demonstrated in terrestrial systems, though somestudies on aquatic systems have emerged in recent years (Hanksiand Gaggiotti 2004). Historically, a substantial impediment tounderstanding and quantifying population connectivity and dis-persal has been the inability to track the position, movements,and origins of individuals, especially in aquatic environments.A number of technologically advanced tools are now availableto researchers for studying population connectivity. These in-clude actively observing animal movements by monitoring markedor tagged individuals and passively reconstructing animal move-ments through the use of physical, chemical, or molecularmarkers.

In aquatic habitats, there has been substantial development ofthe application of genetic and otolith microchemical markers astandem tools for understanding movements and dynamic spatialstructures in fish populations (Milton and Chenery 2001; Milleret al. 2005; Barnett-Johnson 2007). Genetic markers are useful forunderstanding population structure on an evolutionary timescale as indicated by gene flow and also to address questions at acontemporary scale by providing real-time estimates of individualmovements (Paetkau et al. 2004; Manel et al. 2005; Lowe andAllendorf 2010). Otolith microchemical markers are useful at eco-logical time scales because of how otolith elemental compositionreflects that of the water in which fish grow (Campana 1999). Inthis study, we combine the application of genetic and otolithmarkers in a complementary fashion to discern the temporal andspatial scales of population connectivity (spatial overlap) of a semi-anadromous migratory fish species of concern, the Sacramento split-tail (Pogonichthys macrolepidotus; hereinafter referred to as splittail) inthe San Francisco Estuary (SFE), California, USA (Fig. 1).

The splittail and SFE are a model species and system to study therole of connectivity on population dynamics, evolution, and riskof extinction. The species is endemic to the SFE, which offers anexceptional opportunity to exhaustively study the populations

that contribute to the status of the entire species. Splittail haspreviously been listed as threatened under the US EndangeredSpecies Act and is presently listed as a Species of Concern by theState of California (Sommer et al. 2007). There is an urgent need tobetter understand splittail population connectivity and dynamicsbecause (i) it is the only extant member of its genus following theextinction of the Clear Lake splittail (Pogonichthys ciscoides) (Moyle2002), and (ii) the information is needed to guide species manage-ment associated with contentious water development and habitatrestoration activities in the SFE (Sommer et al. 2007).

The splittail is a relatively large demersal cyprinid (maximumfork length �400 mm), which exemplifies the periodic strategistlife history type under the Winemiller and Rose (1992) triangularlife history model (Sommer et al. 1997; Moyle et al. 2004; Feyreret al. 2005). The species exhibits a semi-anadromous life cycle inthat it spends the majority of its adult life in low to moderatesalinity (0–12, all salinity values reported as practical salinityunits) habitats of the SFE and migrates into upstream freshwaterrivers and inundated floodplains for spawning during late winterand spring. Some splittail may remain in the spawning groundsyear-round. This is especially true in the Napa and Petaluma riversbecause elevated salinity in these rivers can constrain suitablehabitat to the upstream regions where spawning also occurs. Pro-duction is strongly tied to hydrology as wet winters trigger massupstream migrations and spawning on inundated floodplains(Sommer et al. 1997; Moyle et al. 2004). Offspring move from natalfreshwater habitats downstream to the brackish SFE on the de-scending limb of the seasonal hydrograph during late spring andsummer (Feyrer et al. 2006, 2007b). Juvenile splittail rear in the SFEfor approximately 2 years until reaching sexual maturity and thenbegin spawning migrations (Daniels and Moyle 1983). Splittail areiteroparous with a life-span of approximately 8 years (Daniels andMoyle 1983).

There are two genetically distinguishable populations of split-tail (Baerwald et al. 2006), one that spawns in western tributaries(the Petaluma and Napa rivers; hereinafter referred to as the PN

Fig. 1. Map of the upper San Francisco Estuary showing the study area. The Sacramento–San Joaquin Delta is not shown and is locatedimmediately upstream (to the east) of the area depicted on the map. SR = Sacramento River; SJR = San Joaquin River.

Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)

2 Can. J. Fish. Aquat. Sci. Vol. 72, 2015

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

.S. G

eolo

gica

l Sur

vey

on 0

3/12

/15

For

pers

onal

use

onl

y.

population) and another that spawns in eastern tributaries lo-cated within California’s Central Valley (hereinafter referred to asthe CV population). Although the absolute sizes are uncertain,genetic evidence suggests that the CV population is several ordersof magnitude larger than the PN population (Mahardja et al. 2014).Although the very existence of two genetically distinguishablepopulations suggests at least some degree of isolation, the dynam-ics of how the two populations are connected is unknown and hasimportant implications for how the species as a whole is managed.Given that the two populations (i) exhibit genetic differentiation,(ii) are substantially different in size, and (iii) exhibit spatial isolationwith discrete spawning habitats, we hypothesized that there wasminimal connectivity and that the two populations represented sep-arate evolutionarily significant units (ESUs) or demes (Waples 1991;Waples and Gaggiotti 2006). The alternative hypothesis is that thereis nontrivial connectivity and the two groups function as some typeof metapopulation. The case for metapopulation dynamics in split-tail is analogous to that of anadromous fishes such as salmonids(Schtickzelle and Quinn 2007), though on a different spatial scale.Both splittail and anadromous salmonids spawn in distinct breedinghabitats, for which they appear locally adapted (salmonids: e.g.,Quinn 2004, Carlson and Seamons 2008; splittail: N. Fangue, De-partment of Fish, Wildlife, and Conservation Biology, Universityof California, Davis, unpublished data). For splittail, high salinityregions of the SFE physically separate the two populations duringthe breeding season and could be considered the functional equiv-alent of the ocean for anadromous salmonids during the non-breeding season where the populations co-occur (Baerwald et al.2008).

At least three basic conditions are required to suggest the oc-currence of metapopulation dynamics: (1) populations inhabit dis-crete breeding habitat patches separated by unsuitable habitat,(2) populations are connected through dispersal, and (3) popula-tion dynamics and traits are not perfectly synchronous (Hanskiet al. 1995). Although extinction has historically been an impor-tant aspect of metapopulation dynamics (Levins 1969; Hanski andSimberloff 1997; Hanski 1999), we agree with others (Kritzer andSale 2004) that presence–absence data alone is sometimes toocoarse for modern metapopulation dynamic applications in fisheriesand other ecological considerations, especially in cases involvingthe conservation of threatened and endangered species. The firstof the key metapopulation dynamic criteria is fulfilled in splittail,as the saline region of SFE is not suitable breeding habitat and sepa-rates the two populations during breeding. However, the extent towhich the populations are connected through the exchange of indi-viduals and the level of demographic or trait independences remainlargely untested and represent the primary foci of our study.

In this study, we took an interdisciplinary approach using em-pirical data collected from both the field and laboratory to answerthe question of whether the two splittail populations representdistinct ESUs or function as metapopulations. These data wereused to quantify the extent of spatial overlap between the twopopulations, characterize differences in demographic traits, andassess the health and condition of splittail at multiple scales. Toexamine spatial dynamics, we intensively sampled adult splittailin the SFE when they reside on foraging grounds during the non-breeding season and also in the PN population’s breeding habitatand applied previously developed genetic (Baerwald et al. 2006)and otolith microchemistry (Feyrer et al. 2007a) tools to identifyfrom which genetic population and natal regions individual split-tail originated. We assessed demographic traits and health andcondition by examining a suite of size, growth, morphometric,and nutritional indicators. We addressed three specific studyquestions and integrate the results to address our hypothesis onsplittail population structure: (i) Are individuals from the two

populations found disproportionately in specific geographic for-aging regions or habitats of the estuary? (ii) Do individuals withinaggregations or schools in their foraging grounds originate fromthe same population or natal region? (iii) Do individuals within apopulation or within a geographic region exhibit differences indemographic or health metrics relative to others?

Study areaThe SFE is located on the Pacific Coast of the United States in

central California (Fig. 1). It has an open water surface area ofapproximately 1235 km2, a mean depth of 4.6 m, and a volume ofapproximately 5.8 × 109 m3. From its relatively narrow connectionto the Pacific Ocean at the Golden Gate, the estuary opens up intoseveral relatively large embayments, including South Bay, CentralBay, and San Pablo Bay. South Bay is generally considered to be amarine lagoon environment, while the rest of the system is muchmore affected by seasonal and annual variations in outflow fromthe watershed. Within San Pablo Bay and upstream into SuisunBay, Grizzly Bay, and Honker Bay, the system typically exhibits arange of conditions intermediate between marine and freshwaterenvironments. The system ultimately becomes dominantly a fresh-water environment in the Sacramento–San Joaquin Delta, an expan-sive maze of tidal channels encapsulating dry and inundated tracts ofland, which is formed at the confluence of the Sacramento andSan Joaquin rivers, the two longest rivers in California.

Physical and biological properties of the SFE are controlled bycomplex interactions of processes that operate at varying frequen-cies and scales. Key primary external factors include climatic driv-ers that manifest through conditions in the watershed and thenortheastern Pacific Ocean. The prevailing Mediterranean climateis generally characterized by a warm and dry summer–fall periodand a cool and wet winter–spring period, which controls theamount outflow entering the estuary from the watershed.

Methods

Field samplingField work was conducted in splittail foraging grounds in the

SFE during November and December in 2010 and 2011 under astratified random sampling design targeting full coverage of split-tail estuarine distribution. The upper SFE from San Pablo Bayupstream to the confluence of the Sacramento and San Joaquinrivers was divided into six approximately equal area strata, andrandom sampling sites within them were generated using Arc-MAP GIS software (ESRI, Redlands, California, USA). During 2011,random sampling sites were stratified equally between offshoreand littoral zones, whereas in 2010 there was no such stratifica-tion between these two habitat zones. Random sampling was sup-plemented with nonrandom targeted collections of splittail in theSFE within the strata framework in an attempt to increase thenumber of individual splittail available for assessing individualhealth and fitness.

Targeted sampling of splittail in their spawning habitat withinthe Petaluma and Napa rivers was also conducted to collect indi-viduals for health and fitness, assess the overall spatial distribu-tion of splittail within these two rivers, and to test the extent towhich adults originated from each population. In the Petalumaand Napa rivers, sampling was initiated in downstream locationsin habitats considered too saline for splittail occupancy and pro-gressed to the upstream-most accessible freshwater spawninghabitat, covering all representative offshore and littoral zones.The relatively small size of the channels in the Petaluma River andNapa River systems resulted in efficient sampling and accuratecharacterization of the spatial distribution of splittail. A total of178 random samples and 209 targeted collection samples werecompleted.

Sampling was conducted with multimesh gill nets that mea-sured 1.8 m in height × 45.7 m in length with five equal length

Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)

Feyrer et al. 3

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

.S. G

eolo

gica

l Sur

vey

on 0

3/12

/15

For

pers

onal

use

onl

y.

panels of 38, 51, 64, 76, and 89 mm mesh. Mesh sizes were selectedto effectively sample all age groups of adult splittail and werechosen based upon the results of previous unpublished studieswith experimental gill nets (R. Baxter, California Department ofFish and Wildlife, Stockton, California). We used the “gillnetfunc-tions” package for R (https://www.stat.auckland.ac.nz/�millar/selectware/R/, accessed on 13 August 2013) to fit relative retentioncurves to evaluate the selectivity of our gill nets (Millar and Holst1997). Counts of splittail fork lengths (FL; grouped into 10 mmincrements) captured in each mesh were used in the analysis. Alog-linear model that assumed equal sampling effort for eachmesh was used to estimate the selectivity curves based on662 individual splittail that were collected in 387 gill net sets. Theresulting selectivity curves (available upon request) indicated thatwhile individual mesh sizes were selective for particular sizeranges of splittail, together they effectively sampled splittailranging in size from approximately 150 to 400 mm FL, whichcorresponds to ages 1–8. Splittail become sexually mature atapproximately 180 mm FL as they near age 2, and few records ofsplittail >400 mm FL exist; thus, our collections generated agood representation of the adult splittail population.

Because the splittail is a demersal fish, the gill nets featured aheavy lead line and a floating top line to maintain position on thebottom of the water column and to ensure they were stretchedvertically. The gill nets were stretched horizontally with the aid ofanchors attached to the ends of the lead line. The gill nets were setduring daylight hours for a duration targeting 60 min; set timeswere based upon the effective collection of splittail in a previousstudy (Nobriga et al. 2005). Relatively short 60 min set times wererequired to ensure that collected splittail remain alive immedi-ately prior to laboratory processing (described below) and to min-imize the incidental catch of other state and federally listedendangered fish species. In practice, under variable field condi-tions, set time averaged 58 min and ranged from 25 to 127 min.Catches were standardized for effort (set time) for data analyses asdescribed below.

Individual splittail collected in each set were carefully extractedfrom the gill net and kept alive in large containers filled withwater taken at the site of collection. Gill net mesh size and splittailFL were recorded for each individual collected. Individuals weremarked (unique fin clips) corresponding to the specific net set inwhich they were collected and transported to the laboratory at theend of the day with aeration and salt added to the water to mini-mize stress. Travel time from the field to the laboratory wasapproximately 60 min, and splittail were processed immediatelyupon arrival.

Habitat characteristics recorded in the field included watertemperature (°C), salinity, turbidity (NTU), and chlorophyll con-centration (�g·L−1), all of which were measured for each samplingevent at the depth of the gill net. Secchi depth (cm) and total waterdepth were also recorded. Offshore distance (km) of each samplewas obtained by plotting sample coordinates (latitude and longi-tude) in Google Earth and using its ruler tool to estimate theshortest direct distance to land.

Genetic population assignmentsDNA was extracted from caudal fin tissue using the QIAGEN

DNEasy 96 kit (QIAGEN Inc.). Nineteen microsatellite markers wereused for the study: CypG3, CypG4, CypG23, CypG25, CypG28, CypG35,CypG39, CypG40, CypG43, CypG45, CypG48, CypG52, CypG53, Pmac1,Pmac4, Pmac19, Pmac24, Pmac25, and Pmac35 (Baerwald and May2004; Mahardja et al. 2012). Methods for PCR and allele genotypingfollowed Mahardja et al. (2014).

Genetic assignment of individuals to their respective popula-tion was conducted using STRUCTURE 2.3.3 (Pritchard et al. 2000).We selected q value of 0.8 as the threshold for distinguishingbetween purebred individuals and potential hybrids or unas-signed based on the efficiency and accuracy scores found in Vähä

and Primmer (2006). Age-0 splittail collected in Baerwald et al.(2006) and Mahardja et al. (2014) were used as references for thepopulations. Ten iterations of K = 2 were performed with all indi-viduals and references included, no prior location information,500 000 burn-in period, and 1 000 000 Markov chain Monte Carlorepetitions under the assumption of admixture and correlatedallele frequencies. Replicate runs were averaged in CLUMPP 1.1.2(Jakobsson and Rosenberg 2007) with the FullSearch algorithm.

Otolith microchemistry natal origin assignmentsStrontium isotopic composition in otoliths (molar ratios of 87Sr/

86Sr) has previously been used to determine the natal origins ofindividual splittail within the CV population (Feyrer et al. 2010).Specifically, individuals born within the upper Sacramento Riverregion above the city of Sacramento can be differentiated fromthose born downstream in the Yolo Bypass and other regions,including the Delta and the San Joaquin River. For simplicity, werefer to this differentiation as the upper Sacramento River versusthe Delta. Otolith 87Sr/86Sr values ≤ 0.706 indicate individualswith natal origins in the upper Sacramento River, while highervalues indicate individuals with natal origins in the Delta at a97% accuracy rate (Feyrer et al. 2010). Existing otolith chemistrymarkers are unable to differentiate whether individuals in the PNwere born in the Petaluma or Napa river with any reasonabledegree of accuracy nor can it differentiate PN from individualsborn in the San Joaquin River in the Central Valley (Feyrer et al.2010). Methods used for otolith preparation and chemistry analy-ses followed Feyrer et al. (2010).

Demographic and health metricsWe used a suite of metrics to characterize key aspects of splittail

demographics and health: size structure (length of individuals),mass at length, length at age, sex composition, Fulton’s conditionfactor (K), hepatosomatic index (HSI), gonadosomatic index (GSI),and muscle protein, lipid, and ash content.

Splittail age was obtained from interpreting information con-tained within otolith microstructures. Thin sections of otolithwere digitally imaged using an American Optical (AO) dissec-ting scope fitted with an ocular digital camera (AM Scope: http://www.amscope.com) at 10× magnification. Annual age incrementswere quantified and measured using Image-J (NIH Imaging: http://rsb.info.nih.gov/ij/). Each otolith was aged by two readers. Otolithswith discrepancies in annual age determination between the tworeaders were read a third time by each reader. Otoliths that haddiscrepancies between multiple age readings greater than 1 yearwere discarded from the study.

We calculated K as 100 × (total body mass/standard length3), HSIas 100 × (liver mass/total body mass), and GSI as 100 × (gonadmass/total somatic mass). Muscle lipid, protein, and ash contentwere obtained from splittail muscle samples that were frozen andstored in a −80 °C freezer. All muscle samples were freeze-driedand ground into a fine powder. The percent crude protein wascalculated by determining the nitrogen content of the muscle.The ground muscle was weighed and combusted as detailed byAOAC Method 990.03 (AOAC 2006a) using a TruSpec CN Analyzer,which quantifies the amount of nitrogen in the muscle. Theamount of nitrogen was converted to protein by the standardconversion factor for fish of 6.25 (AOAC 2006a). The percent crudeprotein was reported as percentage of protein (converted fromnitrogen) in the dried sample. The percent crude lipid was calcu-lated by determining the total lipid content of the muscle. Theground muscle was weighed, and crude fat was extracted by theSoxhlet extraction method as detailed by AOAC Method 2003.05(AOAC 2006b). Lipids were extracted from the sample using thesolvent ethyl ether. The solvent was evaporated by heat and recov-ered by condensation. The amount of crude lipid residue wasdetermined gravimetrically after drying. The percent crude fatwas reported as the percentage of crude fat residue in the dried

Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)

4 Can. J. Fish. Aquat. Sci. Vol. 72, 2015

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

.S. G

eolo

gica

l Sur

vey

on 0

3/12

/15

For

pers

onal

use

onl

y.

sample. Crude percent ash was calculated by determining theinorganic content of the muscle. The ground muscle was weighedand combusted as detailed by AOAC Method 942.05 (AOAC 2006c)using a muffle furnace at 550 °C for at least 3 h. The amount of ashremaining was determined gravimetrically after drying. The per-cent ash was reported as percentage of ash determined gravimet-rically in the dried sample.

Data analyses

Study question (i): Are individuals from the two populationsfound disproportionately in specific geographic foraging regionsor habitats of the estuary?

We used a combination of mapping spatial distributions andgeneralized additive modeling of habitat associations to addressthis study question using the random sampling data. Splittail spa-tial distributions were mapped with ArcMAP GIS software (ESRI,Redlands, California, USA) to determine if individuals from eachpopulation were found disproportionately in specific geographicregions of the estuary. Population-structured habitat associationsof splittail were developed using generalized additive models (GAMs)implemented with the mgcv package in the R statistical program-ming language (Wood 2006). GAMs are nonparametric extensions ofgeneralized linear models useful for describing nonlinear relation-ships between variables. They are data-driven and do not presupposea particular functional relationship between variables; smootherscharacterize the empirical relationships between explanatory andresponse variables. Response variables were counts of splittail fromeach population in each sample standardized to 100 min of samplingeffort (gill net set time). We first constructed single term models withthe independent variables offshore distance, salinity, turbidity, anddepth because existing knowledge (Moyle et al. 2004), and explor-atory data analyses indicated the potential importance of these hab-itat variables. We excluded two measured habitat variables frommodel construction, Secchi depth and chlorophyll concentration,because they were both correlated to turbidity (Secchi depth, r =−0.57, P < 0.001; chlorophyll, r = 0.37, P < 0.001), and temperature wasdeemed unimportant because all sampling was conducted within asingle season and did not meaningfully vary spatially. Models werefit using a Poisson distribution with cubic regression spline smooth-ing functions. Gamma was set to 1.4 to help avoid the models fromover-fitting the data (Wood 2006). We chose habitat variables thatproduced statistically significant single-term models to build fullmultiterm models. In this case, habitat variables that produced sta-tistically significant single-term models proved to also be statisticallysignificant in the multiterm models; thus, no habitat variables wereremoved from the multiterm models in a backwards stepwise fash-ion. Habitat associations of the two populations were compared byassessing independent variables included in each final model andvisually inspecting the form of their relationships to the responsevariable.

Study question (ii): Do individuals within aggregations or schoolsin their foraging grounds originate from the same population ornatal region?

The extent to which mixing of populations or natal regionsoccurred within individual aggregations or schools was deter-mined by testing the null hypothesis that the fractional composi-tion (relative abundance) of individuals from each population ornatal region in an individual sample (gill net set) was as expectedby chance. This hypothesis was tested for each aggregation orschool size observed in which there were at least two individuals.We used a resampling procedure to establish empirical samplingdistributions by randomly resampling the available popula-tions 10 000 times. Generating empirical sampling distribu-tions allowed us to account for uneven population sizes, samplesize variability, and non-normal distributions in the observeddata. P values were calculated as the number of simulations forwhich the statistic was greater than or equal to the observed

statistic divided by the total number of simulations, the statisticin this case being the fractional composition (relative abundance)of the numerically dominant population or natal region in a sam-ple. We interpreted P values < 0.50 (which indicate that the sam-pling statistic was >50% of the 10 000 simulations) as evidence thatthe fractional composition of the dominant population or natalregion was greater than what was expected by chance. This anal-ysis explicitly assumes that an individual gill net sample providesa suitable representation of an individual aggregation or school ofsplittail. The analysis of natal origins was limited to individualsthat were genetically assigned to the CV population because of theotolith microchemistry assignment limitations described above.

Study question (iii): Do individuals within a population or withina geographic region exhibit differences in demographic or healthmetrics relative to others?

Size structure was compared by visual examination of length-frequency plots across populations and at spatial scales corre-sponding to three regions: the brackish nonbreeding feedinggrounds in the estuary, Petaluma River, and Napa River. Mass atlength was evaluated by examining variability in the slope of thelength–mass relationship across the same population and spatialscales. The length–mass relationship was expressed as a linearfunction: log(mass) = � × log(length) + �. Variability in slopes wasinferred from comparing 95% confidence limits for �. Variabilityin K, HSI, GSI, muscle protein, lipid, and ash was assessed sepa-rately by sex across the same population and spatial scales usinganalysis of covariance (ANCOVA). For these analyses we examineddata pooled across years from individuals collected in both therandom and targeted sampling.

Results

General resultsThe 178 random estuary samples collected (coincidentally)

178 individual splittail. Of this total, 116 (65%) were geneticallyassigned to the CV population, 56 (31%) were assigned to the PNpopulation, and 6 (3%) were unassigned (Table 1). Natal originswere determined using otolith 87Sr/86Sr for 105 of 116 individualsgenetically assigned to the CV population, the majority of which(59%) originated in the delta (Table 2).

Overall, splittail were observed in 30 of the 178 (17%) randomsamples, all of which occurred well east of San Pablo Bay and inlittoral habitat (Fig. 2). Overall mean (± standard deviation) CPUE(number of individuals per 100 min of sampling effort) was 1.6 ±5.5. Albeit with considerable variability, overall mean CPUE of theCV population (1.1 ± 3.6) was approximately double that of the PNpopulation (0.5 ± 2.2). Within the CV population, the CPUE ofindividuals with natal origins in the Delta (0.59 ± 1.9) was margin-ally higher than that of individuals with natal origins in the upperSacramento River (0.39 ± 1.8). Mean CPUE solely for samples inwhich splittail were present was 9.5 ± 10.3.

Within the targeted sampling conducted in the Petaluma andNapa rivers, individual splittail were only observed in the upstream-most reaches of these rivers in which sampling occurred (Fig. 2).The nature of these samples being targeted limits us to relativelysimple interpretations of these particular catch data. The mostnotable result was that splittail were only observed in relativelyconfined upstream regions where salinity was ≤12.6; for perspec-tive, salinity gradually increased downstream to >20 in San PabloBay.

Study question (i): Are individuals from the twopopulations found disproportionately in specificgeographic foraging regions or habitats of the estuary?

Individual splittail from the two populations overlapped inspatial distribution within the estuary and therefore were notexclusively found in specific geographic regions (Fig. 3; Table 1).

Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)

Feyrer et al. 5

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

.S. G

eolo

gica

l Sur

vey

on 0

3/12

/15

For

pers

onal

use

onl

y.

Individuals from the CV and PN populations were observed inapproximately the same relative proportions in the estuary’sembayments in both years of study. Individuals from the CV pop-ulation were not found in the Petaluma and Napa rivers in 2010, adry year, but were in 2011, a wet year (Table 1). The individuals inthe Petaluma and Napa rivers that were identified as originatingfrom the CV population had high Q values (all >0.80), suggestingthe origin of these individuals is strongly assigned using our ge-netic approach.

The full multiterm GAMs explained 80% and 61% of the abun-dance variation (deviance) in the CV and PN populations, respec-tively. The CV GAM included all four available habitat variables:offshore distance, salinity, turbidity, and depth as statistically signif-icant predictors of splittail abundance (P < 0.001). The PN GAM wassimilar except that offshore distance was not statistically significant(P > 0.05) and was therefore not included in the model. Devianceplots of partial responses showed that the two populations exhibitedsimilar responses to turbidity and depth but different responses tosalinity (Fig. 4). With regards to turbidity, the response of both pop-ulations increased up to values of approximately 7–8 NTU and re-mained essentially flat thereafter. The response of both populationsto depth dropped sharply at approximately 7–8 m. As mentioned, thetwo populations had different responses to salinity. Abundance ofindividuals from the CV population dropped when salinity reachedapproximately 9–10, while for the PN population it dropped whensalinity reached approximately 15. The response to offshore distancefor the CV population resembled a steep linear negative relationship;the response of the PN population was similar but, as mentioned, notstatistically significant.

Study question (ii): Do individuals within aggregations orschools in their foraging grounds originate from the samepopulation or natal region?

For the analysis on populations, there were 24 instances in theset of random estuary samples in which more than one individualsplittail was collected (Table 3). In 18 of the 24 samples (75%) thefractional composition (relative abundance) of the dominant pop-ulation was higher than what was expected by chance. Instancesof the fractional composition of the dominant population being

higher than expected by chance were insensitive to aggregationsize.

For the analysis on natal origins, there were 17 instances in theset of random estuary samples in which more than one individualsplittail was collected (Table 4). In 11 of the 17 samples (65%) thefractional composition (relative abundance) of the dominant na-tal origin was higher than what was expected by chance. Instancesof the fractional composition of the dominant natal origin beinghigher than expected by chance decreased as aggregation sizeincreased.

Study question (iii): Do individuals within a population orwithin a geographic region exhibit differences indemographic or health metrics relative to others?

Overall splittail size structure was generally consistent acrosspopulations and varied similarly by sex within each population(Fig. 5). Splittail mass at length expressed over all individualsobserved was described by the function: log(mass) = 3.20 ×log(length) − 5.22 (P < 0.001, R2 = 98.9). Slopes of the length–massrelationship (�) varied from 2.94 to 3.32 across all possible combi-nations of comparison; 95% confidence limits suggested consid-erable overlap in the slopes of the length–mass relationship,regardless of how the fish were grouped (Table 5).

Splittail ranged in age from 1 to 8 in our samples. Splittail fromboth populations and sexes exhibited indeterminate growth inlength with age (Fig. 6). ANCOVA results indicated that splittaillength varied significantly as a function of age (P < 0.001) and sex(P = 0.001), with females longer than males at all ages, but we didnot find evidence for variation in length at age across populations(P = 0.45). Regionally, however, there was a complete absence ofindividuals older than age 3 in the Petaluma River.

There was notable variability in K, HSI, GSI, muscle protein, lipid,and ash (Fig. 7). ANCOVA results indicated statistically significantpopulation differences in male GSI (PN > CV; P = 0.015) and spatialdifferences in female lipid (estuary > Napa; P = 0.006), female ash(Napa > estuary; P = 0.021), male K (estuary > Napa; P = 0.031), femaleGSI (estuary and Napa > Petaluma, P = 0.006), male GSI (Petaluma andNapa > estuary, P < 0.001), female HSI (Petaluma > Napa > estuary;P < 0.001), and male HSI (Petaluma > Napa > estuary; P < 0.001).

DiscussionSplittail population structure best matches our alternative

hypothesis of a metapopulation, consisting of a dominant and asubordinate population that we suggest resembles the mainland-island structure in metapopulation ecology (Harrison et al. 1988).All three key conditions for metapopulation dynamics appear tobe fulfilled in splittail: (1) populations inhabit discrete breedinghabitat patches separated by saline habitat unsuitable for spawn-ing, (2) populations are spatially connected through dispersal ofthe dominant population into the subordinate population’s spawn-ing grounds, and (3) population dynamics and traits are not perfectlysynchronous. No data presently exist, genetic or otherwise, to sug-gest finer-scale population structure may exist in splittail, i.e., each

Table 1. Counts (and rounded percentages in parentheses) of genetically derived populationassignments of individual splittail observed in 2010 and 2011 in the estuary, Petaluma River,and Napa River.

2010 2011

Region CV PN Unassigned CV PN Unassigned

Estuary 49 (67%) 21 (29%) 3 (4%) 67 (64%) 35 (33%) 3 (3%)Petaluma River 0 5 (100%) 0 1 (3%) 37 (95%) 1 (3%)Napa River 0 36 (97%) 1 (3%) 7 (14%) 39 (76%) 5 (10%)

Note: CV = Central Valley population; PN = Petaluma–Napa population.

Table 2. Counts (and rounded percentages in parentheses) of otolithchemistry derived natal areas of individual splittail observed in 2010and 2011 in the estuary, Petaluma River, and Napa River.

2010 2011

RegionSacramentoRiver Delta

SacramentoRiver Delta

Estuary 16 (43%) 21 (57%) 27 (40%) 41 (60%)Petaluma River — — — 1 (100%)Napa River — — 5 (63%) 3 (38%)

Note: The data pertain to individual splittail genetically assigned to the Cen-tral Valley population. Sacramento River refers to individuals born in upperSacramento River upstream of Yolo Bypass, while Delta refers to individuals born inthe Yolo Bypass and other downstream regions, including the Sacramento–San Joaquin Delta and the San Joaquin River. Dashes indicate that no CentralValley splittail were observed in the Petaluma and Napa rivers in 2010.

Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)

6 Can. J. Fish. Aquat. Sci. Vol. 72, 2015

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

.S. G

eolo

gica

l Sur

vey

on 0

3/12

/15

For

pers

onal

use

onl

y.

CV or PN population could potentially be composed of sets of popu-lations. For example, there is no evidence that splittail within eachpopulation home to specific natal sites for spawning. Thus, until newdata emerge to demonstrate otherwise, we feel it is appropriate atthis time to use the term population to refer to the two groups.

The CV population is numerically larger and occupies a largerhabitat area of likely better quality than the PN population, lend-ing credence to the mainland–island type of structure. Similar tothe mainland–island metapopulation model where dispersalfrom the larger mainland dominates the direction of connectivityto the smaller island population preventing extinction, here theCV population likely plays the mainland role to the smaller PNpopulation as evidenced by CV origin fish occurring in the Peta-luma and Napa rivers in the wet year (2011). Mahardja et al. (2014)estimated the effective size of the CV population to be orders of

magnitude greater than the PN population. Based on geographyalone, surface area of the occupied habitat in the Petaluma andNapa rivers is just a fraction of that in the Central Valley; therelative proportion of available habitat skews even further whenwater of suitable salinity is considered. Our observation that thehealth of individual splittail in terms of lipid stores and conditionwas less favorable in the Petaluma and Napa rivers (Fig. 7) providesindirect evidence of meaningful differences in habitat quality.Salinity, which is generally elevated in the Petaluma and Naparivers, can also be viewed as a driver of habitat quality. Splittailinhabit elevated salinity sooner in life in the Petaluma and Naparivers compared with the Central Valley (Feyrer et al. 2010), andemerging data suggest that splittail populations exhibit differen-tial physiological performance traits in response to elevated

Fig. 2. Spatial distribution of splittail catch per unit effort (CPUE; number of individuals per 100 min of sampling effort). The size of the CPUEdata points is scaled according to the legend. Data points within the Petaluma and Napa rivers, including zero catches indicated by +, aretargeted nonrandom sampling locations. All other data points outside of the Napa and Petaluma rivers, including zero catches indicated by ×,are random sampling locations within the nonbreeding rearing grounds.

Fig. 3. Spatial distribution of splittail population assignments in random samples located within the estuary nonbreeding rearing grounds.Pie chart sizes are scaled to the number of individuals in a sample (range = 1–34), and jitter was introduced to the sample locations to avoidpie chart superimposition.

Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)

Feyrer et al. 7

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

.S. G

eolo

gica

l Sur

vey

on 0

3/12

/15

For

pers

onal

use

onl

y.

Fig. 4. Plots showing habitat associations of individual splittail from each population. Plots are fitted smooths and 95% confidence intervals for partial responses from generalizedadditive models. Units on the y axis are centered on zero, and the number in the label is the estimated degrees of freedom of the smooth. Tick marks along the x axis representobservations.

Pagination

notfinal(cite

DO

I)/

Pagination

provisoire(citer

leD

OI)

8C

an.J.Fish

.Aqu

at.Sci.Vol.72,2015

Publish

edby

NR

CR

esearchPress

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

.S. G

eolo

gica

l Sur

vey

on 0

3/12

/15

For

pers

onal

use

onl

y.

salinity (N. Fangue, Department of Fish, Wildlife, and Conserva-tion Biology, University of California, Davis, unpublished data).

Splittail populations appear spatially connected by dispersal,yet segregation is maintained at multiple scales. In the SFE, indi-vidual splittail tend to aggregate or school with others of similarpopulation heritage and natal origin. Generally similar behaviorhas been observed at various scales in other migratory fishes (e.g.,McKinnel et al. 1997; Fraser et al. 2005; Barnett-Johnson 2007;Weitkamp 2010). Associating with kin in splittail could potentiallyfunction to maintain population structure by exerting pressureon mate selection. It is unclear, however, if individual splittailobtain fitness benefits from such a pattern of segregation. For ananalogous situation, genetic diversity benefits were shown to out-weigh the benefits of kin association in a study of Atlantic salmon(Salmo salar) (Griffiths and Armstrong 2001). With the exception ofmale GSI, none of the health and condition metrics we examinedwere statistically different between populations; the only appar-ent differences in health and condition were those mentionedabove that varied according to geographic location. The range ofpossibilities driving the observed level of segregation range fromsimple chance encounter with no resulting benefits to complexsocial behaviors that could result in individual or group benefitsnot addressed in our study. Kin identification is not unprecedented,as it has been demonstrated in other migratory fish species (Brownand Brown 1992). The two splittail populations further maintained

segregation at the habitat scale, with the PN population generallyoccupying a higher salinity than the CV population, which reducesthe likelihood that chance encounter is the driving force of segrega-tion at the individual aggregation or school level.

The concept of synchrony requires special consideration andfurther study in the case of splittail because of the existence ofdominant and subordinate populations. Further, some aspects ofsplittail life history traits suggest variable levels of synchrony,which may be unusual in classical metapopulation theory. Syn-chrony is typically assessed with time series abundance data,which, as with many ecological systems, is not available for thetwo splittail populations. Thus, we examined traits that could beindicators of demographic processes; results show that there wereno meaningful differences in the traits we measured. This sup-ports the idea that while there are two genetic populations, thereis sufficient dispersal to maintain these traits in the two popula-tions. The two splittail populations are clearly not synchronouswith respect to absolute size. However, the relative abundances ofage-0 production of the two populations may fluctuate synchro-nously in response to environmental forcing, a condition knownas the Moran effect (Moran 1953). This seems reasonable becauseage-0 production of the CV population is strongly tied to hy-drology (Sommer et al. 1997; Moyle et al. 2004), and available dataindicate that interannual fluctuations in outflow are similar inthe Petaluma–Napa and Central Valley basins, albeit with consid-erable differences in absolute flow volume. Increased synchronyis generally thought to correspond with increased extinction riskto the metapopulation (Earn et al. 2000). Modeling studies havesuggested that the CV population is resilient to environmentalconditions and has a low risk of extinction (Moyle et al. 2004).However, similar efforts should be made to model extinction risksfor the PN population, which may be particularly vulnerable tolocalized catastrophic events.

Table 3. Splittail population composition in random samples col-lected in the nonbreeding feeding grounds in the San Francisco estu-ary (see Fig. 3 for locations).

Size ofaggregationor school

Numericallydominantpopulation

Fractional compositionof numerically dominantpopulation P value

2 PN 1.00 0.12*2 PN 1.00 0.12*2 PN 1.00 0.12*2 CV 1.00 0.43*2 CV 1.00 0.43*2 CV 1.00 0.43*2 CV 1.00 0.43*3 CV 0.67 0.29*3 CV 1.00 0.29*4 Even 0.50 0.894 CV 0.75 0.594 CV 0.75 0.594 CV 0.75 0.594 CV 1.00 0.19*5 CV 0.80 0.45*5 CV 0.80 0.45*8 CV 0.88 0.04*8 CV 1.00 0.04*10 CV 0.60 0.7810 CV 1.00 0.01*11 CV 0.64 0.46*16 PN 0.56 0.14*19 PN 0.58 0.29*34 CV 0.62 0.66

Note: Data are provided for each sample in which at least two individualswere observed. Data shown include the size of the aggregation or school ob-served (total number of individuals observed), the numerically dominant popu-lation (PN = Petaluma–Napa, CV = Central Valley), the fractional composition(relative abundance expressed as a fraction) of the numerically dominant pop-ulation in the aggregation or school, and P values associated with a test of thenull hypothesis that the fractional composition was as expected by chance (seeMethods for details). Asterisks indicate samples in which there were more indi-viduals of the numerically dominant population observed than what was ex-pected by chance.

Table 4. Splittail natal regions in random samples collected in thenonbreeding feeding grounds in the San Francisco estuary (see Fig. 3for locations).

Size ofaggregationor school

Dominantnatalregion

Fractional composition ofdominant natal region insample P value

2 Delta 1.00 0.33*2 Delta 1.00 0.33*2 Delta 1.00 0.33*3 Delta 1.00 0.19*3 Delta 1.00 0.19*3 Delta 1.00 0.19*3 Delta 0.67 0.19*4 Delta 0.75 0.19*4 Sacramento 0.75 0.19*4 Even 0.50 0.807 Sacramento 0.71 0.12*7 Delta 0.57 0.687 Delta 0.57 0.688 Even 0.50 0.798 Sacramento 0.62 0.06*10 Delta 0.60 0.5921 Delta 0.52 0.78

Note: Data are provided for each sample in which at least two individualswere observed and for individuals that had been genetically assigned to theCentral Valley splittail population. Data shown include the size of the aggrega-tion or school observed (total number of individuals observed), the numericallydominant natal region, the fractional composition (relative abundance ex-pressed as a fraction) of the numerically dominant natal region in the aggrega-tion or school, and P values associated with a test of the null hypothesis that thefractional composition was as expected by chance (see Methods for details).Asterisks indicate samples in which there were more individuals of the numer-ically dominant natal region observed than what was expected by chance.

Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)

Feyrer et al. 9

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

.S. G

eolo

gica

l Sur

vey

on 0

3/12

/15

For

pers

onal

use

onl

y.

A poignant example of how population structure and dynamicsplay into extinction risks occurred in September 2010 when a spillreleased up to 2300 L of oil in the upper Petaluma River. This spillwas concurrent with our study and may explain the observedabsence of large adult females in this area during our study. Ifindeed these populations function as viable a metapopulation,then CV population adults could supplement or colonize the Peta-luma River (as could PN population individuals from the NapaRiver). Such a process would resemble a mainland–island metapo-pulation structure, where persistence of the PN populationdepends on the existence of the more extinction-resistant CV pop-ulation. However, given the genetic differences in the CV and PNpopulations, it is unclear whether the fitness of CV populationsspawning in the Petaluma River could in fact result in meaningfulcontribution to the PN population. This example illustrates thatin addition to dispersal–migration and connectivity, metapopula-tion models need to consider fitness differences in migrants, re-

gionally correlated environmental forcing, and local processesand events to assess extinction risks.

Climate-induced dynamic habitat fragmentation could be a keydriver of splittail metapopulation structure. Our results suggestthat the two populations are connected spatially via dispersalof the dominant population into the subordinate population’sspawning habitat when climate patterns produce enough rainfalland associated freshwater outflow to form a bridge of suitable lowsalinity habitat across the upper SFE. Specifically, no CV-originfish were observed in the Petaluma and Napa rivers in the dryyear, 2010. This is in contrast with the wetter year of 2011 whereCV individuals were found to disperse into the PN. Woods et al.(2010) similarly used otolith chemistry and genetic tools to deter-mine that catchment dispersal of Australian smelt (Retropinnasemoni) was mediated by changes in hydrological connectivity. Habi-tat affinities of the two splittail populations suggest that a sustainedsalinity of approximately ≤12 is needed to form such a bridge. Hydro-dynamic modeling studies (Gross et al. 2009) and historic outflowrecords together suggest such conditions occur in approximately 1/3of years overall with an irregular frequency. The degree to which thisdynamic pattern of spatial connectivity drives gene flow remains animportant question for understanding splittail population dynam-ics. One possibility is that this pattern of spatial connectivityprovides sufficient gene flow to maintain the existence of the twogenetically similar populations. Another possibility is that thedynamic spatial connectivity between the two populations is suf-ficiently low that genetic drift and (or) selection will eventuallydrive the two populations to become separate species once ade-quate evolutionary time has passed. In either case, the trajectoryof splittail evolution and population structure is likely to beshaped by future climate conditions. It is likely that a drier futureconstraining spatial connectivity could further diverge the twopopulations, while a wetter future increasing spatial connectivitycould homogenize the two populations. The same general con-cept holds for management actions that would produce suchconditions by altering the present physical configuration orhydrodynamics of the estuary, an important consideration giventhe strong interest in large-scale habitat restoration in the SFE. Theavailability of detailed future climate change scenarios downscaledfor the SFE (Cloern et al. 2011) enable the trajectory of splittail

Fig. 5. Length-frequency plots showing the size structure of each splittail population separately for males and females.

Table 5. Slopes (�) and 95% confidence limitsobtained from linear regressions of splittail log-(mass) at log(length) aggregated across species(all individuals), population (PN = Petaluma–Napa population, CV = Central Valley popula-tion), and spatial (estuary feeding grounds, NapaRiver, Petaluma River) scales.

GroupSlope±95%confidence limit

All individuals 3.20±0.04CV population female 3.32±0.10CV population male 3.15±0.09PN population female 3.14±0.07PN population male 3.13±0.10Estuary female 3.24±0.10Estuary male 3.13±0.08Napa River female 3.23±0.13Napa River male 3.17±0.17Petaluma River female 2.94±0.26Petaluma River male 3.14±0.01

Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)

10 Can. J. Fish. Aquat. Sci. Vol. 72, 2015

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

.S. G

eolo

gica

l Sur

vey

on 0

3/12

/15

For

pers

onal

use

onl

y.

evolution and population structure to be modeled. Such an effortwould require additional research to quantify gene flow betweenthe two populations. The ability to model climate-driven habitatfragmentation and associated evolutionary and population ecol-ogy trajectories of an aquatic species provides an unprecedentedopportunity for the testing and application of theoretical andapplied ecological science, as well as natural resource manage-ment.

AcknowledgementsWe thank M. Gingras, R. Soto, E. Santos, N. Sakata, N. van Ark,

C. Grant, D. Obegi, M. Nobriga, D. Riordon, A. Chandos, K. Gehrts,B. Coalter, B. Harrell, A. Mueller-Solger, M. Young, A. Bibian, L. Conrad,B. Schreier, R. Baxter, E. van Nieuwenhuyse, S. Waller, T. Foin,

N. Fangue, B. May, K. Reece, M. MacWilliams, L. Brown, R. Baxter,T. Sommer, and many others for assistance. S. Teh, S. Acuna, andJ. Hobbs were funded by an agreement (No. R10AC20095) betweenthe US Bureau of Reclamation and the University of California–Davis awarded to S. Teh. B. Mahardja and M. Baerwald were sup-ported by Delta Science Program Grant No. 2037.

ReferencesAOAC 2006a. AOAC Official Method 990.03. Protein (Crude) in animal feed,

combustion method. In Official Methods of Analysis of AOAC International.18th ed. Revision 1, 2006. Chapter 4. AOAC International, Gaithersburg, Md.pp. 30–31.

AOAC 2006b. AOAC Official Method 2003.05. Crude fat in feeds, cereal grains,and forages. In Official Methods of Analysis of AOAC International. 18th ed.(2006), Chapter 4. AOAC International, Arlington, Va. pp. 40–42.

Fig. 6. Box plot representations of the length of individual male and female splittail by age for each population. ANCOVA results indicatedthat there was no statistical difference in length at age between the two populations and that females were longer at a given age than males.

Fig. 7. Box plot representations of data on individual male and female splittail for Fulton’s condition factor (K), hepatosomatic index (HSI),gonadosomatic index (GSI), and muscle protein, lipid, and ash content. The data shown are aggregated according to the spatial locationwithin the system (Petaluma River, Napa River, estuary), with letters denoting statistically significant differences within boxes.

Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)

Feyrer et al. 11

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

.S. G

eolo

gica

l Sur

vey

on 0

3/12

/15

For

pers

onal

use

onl

y.

AOAC 2006c. AOAC Official Method 942.05. Ash of animal feed. In Official Meth-ods of Analysis of AOAC International. 18th ed. (2006), Chapter 4. AOACInternational, Gaithersburg, Md. p. 8.

Baerwald, M.R., and May, B. 2004. Characterization of microsatellite loci for fivemembers of the minnow family Cyprinidae found in the Sacramento–SanJoaquin Delta and its tributaries. Mol. Ecol. Notes, 4: 385–390. doi:10.1111/j.1471-8286.2004.00661.x.

Baerwald, M., Bien, V., Feyrer, F., and May, B. 2006. Genetic analysis reveals twodistinct Sacramento splittail (Pogonichthys macrolepidotus) populations. Con-serv. Genet. 8: 159–167. doi:10.1007/s10592-006-9157-2.

Baerwald, M., Feyrer, F., and May, B. 2008. Distribution of genetically differen-tiated splittail populations during the nonspawning season. Trans. Am. Fish.Soc. 137: 1335–1345. doi:10.1577/T07-097.1.

Barnett-Johnson, R. 2007. Spatial scales of mixing and natal source contributionsof salmon populations in the coastal ocean detected by otolith and geneticsignatures of origin. Ph.D. dissertation, Department of Ecology and Evolu-tionary Biology, University of California, Santa Cruz, Calif.

Bilton, D.T., Freeland, J.R., and Okamura, B. 2001. Dispersal in freshwater inver-tebrates. Annu. Rev. Ecol. Syst. 32: 159–181. doi:10.1146/annurev.ecolsys.32.081501.114016.

Brown, G.E., and Brown, J.A. 1992. Do rainbow trout and Atlantic salmon dis-criminate kin? Can. J. Zool. 70(8): 1636–1640. doi:10.1139/z92-227.

Campana, S.E. 1999. Chemistry and composition of fish otoliths: pathways,mechanisms and applications. Mar. Ecol. Prog. Ser. 188: 263–297. doi:10.3354/meps188263.

Carlson, S.M., and Seamons, T.R. 2008. A review of quantitative genetic compo-nents of fitness in salmonids: implications for adaptation to future change.Evol. Appl. 1: 222–238. doi:10.1111/j.1752-4571.2008.00025.x. PMID:25567628.

Cloern, J.E., Knowles, N., Brown, L.R., Cayan, D., Dettinger, M.D., Morgan, T.L.,Schoellhamer, D.H., et al. 2011. Projected evolution of California’s San Fran-cisco Bay–Delta–River System in a century of climate change. PLoS ONE, 6(9):e24465. doi:10.1371/journal.pone.0024465. PMID:21957451.

Cowen, R.K., and Sponaugle, S. 2009. Larval dispersal and marine populationconnectivity. Annu. Rev. Mar. Sci. 1: 443–466. doi:10.1146/annurev.marine.010908.163757.

Daniels, R.A., and Moyle, P.B. 1983. Life history of the splittail (Cyprinidae:Pogonichthys macrolepidotus) in the Sacramento–San Joaquin Estuary. Fish.Bull. 84: 105–117.

Dunham, J.B., and Rieman, B.E. 1999. Metapopulation structure of bull trout:influences of physical, biotic, and geometrical landscape characteristics.Ecol. Appl. 9: 642–655. doi:10.1890/1051-0761(1999)009v0642:MSOBTI]2.0.CO;2.

Earn, D.J., Levin, S.A., and Rohani, P. 2000. Coherence and conservation. Science,290(5495): 1360–1364. doi:10.1126/science.290.5495.1360. PMID:11082064.

Feyrer, F., Sommer, T., and Baxter, R. 2005. Spatial-temporal distribution andhabitat associations of age-0 splittail in the lower San Francisco EstuaryWatershed. Copeia, 2005: 159–168. doi:10.1643/CE-04-187R1.

Feyrer, F., Sommer, T., and Harrell, W. 2006. Managing floodplain inundationfor native fish: production dynamics of age-0 splittail in California’s YoloBypass. Hydrobiologia, 573: 213–226. doi:10.1007/s10750-006-0273-2.

Feyrer, F., Hobbs, J., Baerwald, M., Sommer, T., Yin, Q., Clark, K., May, B., andBennett, W. 2007a. Otolith microchemistry provides information comple-mentary to microsatellite DNA for a migratory fish. Trans. Am. Fish. Soc. 136:469–476. doi:10.1577/T06-044.1.

Feyrer, F., Sommer, T., and Hobbs, J. 2007b. Living in a dynamic environment:variability in life history traits of age-0 splittail in tributaries of San FranciscoBay. Trans. Am. Fish. Soc. 136: 1393–1405. doi:10.1577/T06-253.1.

Feyrer, F., Hobbs, J., and Sommer, T. 2010. Salinity inhabited by age-0 splittail(Pogonichthys macrolepidotus) as determined by direct field observation and retrospec-tive analyses with otolith chemistry [online]. San Francisco Estuary and Water-shed Science, 8(2). Available from http://escholarship.org/uc/item/14j5160h.

Fraser, D.J., Duchesne, P., and Bernatchez, L. 2005. Migratory charr schoolsexhibit population and kin associations beyond juvenile stages. Mol. Ecol. 14:3133–3146. doi:10.1111/j.1365-294X.2005.02657.x. PMID:16101779.

Griffiths, S.W., and Armstrong, J.D. 2001. The benefits of genetic diversity out-weigh those of kin association in a territorial animal. Proc. R. Soc. Ser. B Biol.Sci. 268:1293–1296. doi:10.1098/rspb.2001.1660.

Gross, E.S., MacWilliams, M.L., and Kimmerer, W.J. 2009. Three-dimensionalmodeling of tidal hydrodynamics in the San Francisco Estuary. San FranciscoEstuary and Watershed Science, 7(2).

Hanski, I. 1999. Metapopulation ecology. Oxford University Press, Oxford.Hanski, I. and Gaggiotti, O.E. (Editors). 2004. Ecology, genetics, and evolution of

metapopulations. Elsevier Academic Press.Hanski, I., and Simberloff, D. 1997. The metapopulation approach, its history,

conceptual domain, and application to conservation. In Metapopulationbiology: ecology, genetics, and evolution. Edited by I. Hanski and M.E. Gilpin.Academic Press, San Diego, Calif. pp. 5–26.

Hanski, I., Pakkala, T., Kuussaari, M., and Lei, G.C. 1995. Metapopulation persis-tence of an endangered butterfly in a fragmented landscape. Oikos, 72: 21–28.doi:10.2307/3546033.

Harrison, S., Murphy, D., and Ehrlich, P. 1988. Distribution of the bay checker-spot butterfly, Euphydryas editha bayensis: evidence for a metapopulationmodel. Am. Nat. 132: 360–382. doi:10.1086/284858.

Jakobsson, M., and Rosenberg, N.A. 2007. CLUMPP: a cluster matching and per-mutation program for dealing with label switching and multimodality inanalysis of population structure. Bioinformatics, 23: 1801–1806. doi:10.1093/bioinformatics/btm233. PMID:17485429.

Kritzer, J.P., and Sale, P.F. 2004. Metapopulation ecology in the sea: from Levins’model to marine ecology and fisheries science. Fish Fish. 5: 131–140. doi:10.1111/j.1467-2979.2004.00131.x.

Levins, R. 1969. Some demographic and genetic consequences of environmentalheterogeneity for biological control. Bull. Ent. Soc. Am. 15: 237–240. doi:10.1093/besa/15.3.237.

Lowe, W.H., and Allendorf, F. 2010. What can genetics tell us about populationconnectivity? Mol. Ecol. 19: 3038–3051. doi:10.1111/j.1365-294X.2010.04688.x.PMID:20618697.

Mahardja, B., May, B., and Baerwald, M.R. 2012. Characterization of 36 addi-tional microsatellite loci in splittail (Pogonichthys macrolepidotus) and cross-amplification in five other native Californian cyprinid species. Conserv.Genet. Resour. 4: 917–921. doi:10.1007/s12686-012-9673-y.

Mahardja, B., May, B., Feyrer, F., Coalter, R., Fangue, N., Foin, T., and Baerwald, M.R.2014. Interannual variation in connectivity and comparison of effective pop-ulation size between two splittail (Pogonichthys macrolepidotus) populations inthe San Francisco Estuary. Conserv. Genet. doi:10.1007/s10592-014-0665-1.

Manel, S., Gaggiotti, O., and Waples, R. 2005. Assignment methods: matchingbiological questions with appropriate techniques. Trends Ecol. Evol. 20: 136–142. doi:10.1016/j.tree.2004.12.004. PMID:16701357.

McKinnel, S., Pella, J.J., and Dahlberg, M.L. 1997. Population-specific aggrega-tions of steelhead trout (Oncorhynchus mykiss) in the North Pacific Ocean. Can.J. Fish. Aquat. Sci. 54(10): 2368–2376. doi:10.1139/f97-143.

Millar, R.B., and Holst, R. 1997. Estimation of gillnet and hook selectivity usinglog-linear models. ICES J. Mar. Sci. 54: 471–477. doi:10.1006/jmsc.1996.0196.

Miller, J.A., Banks, M.A., Gomez-Uchida, D., and Shanks, A.L. 2005. A comparisonof population structure in black rockfish (Sebastes melanops) as determinedwith otolith microchemistry and microsatellite DNA. Can. J. Fish. Aquat. Sci.62(10): 2189–2198. doi:10.1139/f05-133.

Milton, D.A., and Chenery, S.R. 2001. Can otolith chemistry detect the popula-tion structure of the shad hilsa Tenualosa ilisha? Comparison with the resultsof genetic and morphological studies. Mar. Ecol. Prog. Ser. 222: 239–251.doi:10.3354/meps222239.

Moran, P.A.P. 1953. The statistical analysis of the Canadian lynx cycle. II. Synchro-nization and meteorology. Aust. J. Zool. 1: 291–298. doi:10.1071/zo9530291.

Moyle, P.B. 2002. Inland fishes of California. Revised and expanded. University ofCalifornia Press, Berkeley, Calif.

Moyle, P.B., Baxter, R.D., Sommer, T., Foin, T.C., and Matern, S.A. 2004. Biologyand population dynamics of Sacramento splittail (Pogonichthys macrolepidotus)in the San Francisco Estuary: a review. San Francisco Estuary and WatershedScience 2:2(May 2004), Article 3.

Nobriga, M., Feyrer, F., Baxter, R., and Chotkowski, M. 2005. Fish community ecologyin an altered river delta: spatial patterns in species composition, life historystrategies, and biomass. Estuaries, 28: 776–785. doi:10.1007/BF02732915.

Paetkau, D., Slade, R., Burden, M., and Estoup, A. 2004. Genetic assignmentmethods for the direct, real-time estimation of migration rate: a simulation-based exploration of accuracy and power. Mol. Ecol. 13: 55–65. doi:10.1046/j.1365-294X.2004.02008.x. PMID:14653788.

Pritchard, J.K., Stephens, M., and Donnelly, P. 2000. Inference of population struc-ture using multilocus genotype data. Genetics, 155: 945–959. PMID:10835412.

Quinn, T.P. 2004. The behavior and ecology of Pacific salmon and trout. Ameri-can Fisheries Society in association with University of Washington Press.

Schtickzelle, N., and Quinn, T.P. 2007. A metapopulation perspective for salmonand other anadromous fish. Fish Fish. 8: 297–314. doi:10.1111/j.1467-2979.2007.00256.x.

Sommer, T., Baxter, R., and Herbold, B. 1997. Resilience of splittail in the Sacra-mento–San Joaquin Estuary. Trans. Am. Fish. Soc. 126: 961–976. doi:10.1577/1548-8659(1997)126<0961:ROSITS>2.3.CO;2.

Sommer, T.R., Baxter, R., and Feyrer, F. 2007. Splittail revisited: how recentpopulation trends and restoration activities led to the “delisting” of thisnative minnow. In Status, distribution, and conservation of freshwater fishesof western North America. Edited by M.J. Brouder and J.A. Scheuer. AmericanFisheries Society Symposium 53, Bethesda, Md. pp. 25–38.

Vähä, J.P., and Primmer, C.R. 2006. Efficiency of model-based Bayesian methodsfor detecting hybrid individuals under different hybridization scenarios andwith different numbers of loci. Mol. Ecol. 15: 63–72. PMID:16367830.

Waples, R. 1991. Pacific salmon, Oncorhynchus spp., and the definition of “speciesunder the Endangered Species Act. Mar. Fish. Rev. 3: 11–22.

Waples, R., and Gaggiotti, O. 2006. What is a population? An empirical evalua-tion of some genetic methods for identifying the number of gene pools andtheir degree of connectivity [online]. Publications, Agencies and Staff ofthe US Department of Commerce. Paper 463. Available from http://digitalcommons.unl.edu/usdeptcommercepub/463.

Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)

12 Can. J. Fish. Aquat. Sci. Vol. 72, 2015

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

.S. G

eolo

gica

l Sur

vey

on 0

3/12

/15

For

pers

onal

use

onl

y.

Weitkamp, L.A. 2010. Marine distributions of Chinook salmon from the westcoast of North America determined by coded wire tag recoveries. Trans. Am.Fish. Soc. 139: 147–170. doi:10.1577/T08-225.1.

Winemiller, K.O., and Rose, K.A. 1992. Patterns of life-history diversification inNorth American fishes: implications for population regulation. Can. J. Fish.Aquat. Sci. 49(10): 2196–2218. doi:10.1139/f92-242.

Wood, S.N. 2006. Generalized additive models: an introduction with R. Chapmanand Hall/CRC.

Woods, R.J., Macdonald, J.I., Crook, D.A., Schmidt, D.J., and Hughes, J.M. 2010.Contemporary and historical patterns of connectivity among populations ofan inland river fish species inferred from genetics and otolith chemistry.Can. J. Fish. Aquat. Sci. 67(7): 1098–1115. doi:10.1139/F10-043.

Pagination not final (cite DOI) / Pagination provisoire (citer le DOI)

Feyrer et al. 13

Published by NRC Research Press

Can

. J. F

ish.

Aqu

at. S

ci. D

ownl

oade

d fr

om w

ww

.nrc

rese

arch

pres

s.co

m b

y U

.S. G

eolo

gica

l Sur

vey

on 0

3/12

/15

For

pers

onal

use

onl

y.