Larval Fish Transport and Retention and the ... - Wiley Online Library

Transactions of the American Fisheries Society 138:328–347, 2009 Ó Copyright by the American Fisheries Society 2009 DOI: 10.1577/T07-274.1

[Article]

Larval Fish Transport and Retention and the Importance of Location for Juvenile Fish Recruitment in Upper Klamath Lake, Oregon DOUGLAS F. MARKLE*

AND

SUSAN A. REITHEL

Department of Fisheries and Wildlife, Oregon State University, 104 Nash Hall, Corvallis, Oregon 97331-3803, USA

JOHN CRANDALL The Nature Conservancy in Oregon, 226 Pine Street, Klamath Falls, Oregon 97601, USA

TAMMY WOOD U.S. Geological Survey, Oregon Water Science Center, 10615 Southeast Cherry Blossom Drive, Portland, Oregon 97216, USA

TORREY J. TYLER U.S. Bureau of Reclamation, 6600 Washburn Way, Klamath Falls, Oregon 97603, USA

MARK TERWILLIGER

AND

DAVID C. SIMON

Department of Fisheries and Wildlife, Oregon State University, 104 Nash Hall, Corvallis, Oregon 97331-3803, USA Abstract.—In the Upper Klamath Lake basin of Oregon, we applied the member–vagrant hypothesis to larval lake fishes, documented marsh and hydrographic retention for larval shortnose suckers Chasmistes brevirostris and Lost River suckers Deltistes luxatus, and examined the effect of larval retention on the shoreline abundance of the juveniles of five species of fish in August. Emigration of larval suckers was highest from natal rivers, and immigration was highest in nonnatal areas downstream of the lake. Lake retention was facilitated by a wind-generated gyre and advection by an eastern boundary current. Marshes and other shoreline irregularities acted as traps that slowed river and eastern boundary current advection and retained larvae, especially for shortnose suckers, for up to 3 weeks. The interaction of behavior, seasonal spawning, and seasonal system dynamics appears to drive retention patterns. For both species of suckers and three minnow species (fathead minnow Pimephales promelas, blue chub Gila coerulea, and tui chub G. bicolor), the abundance of juveniles along the shoreline increased with distance from the lake outlet and was weakly related to lake elevation, a surrogate for available habitat. For these five species, 45.5% of the variation in August juvenile abundance was determined by distance from the lake outlet. This work suggests that the nonbehavioral component of freshwater retention is complex and depends on hydrographic features, the boundary conditions of hydrographic features, and shoreline habitat.

When there is larval dispersal from spawning grounds, fish life histories and the local hydrographic regime must be compatible so that larvae are dispersed to areas where they can grow and recruit back to the parent population rather than to areas where they are lost to the parent population. According to the member–vagrant hypothesis (Sinclair 1988), members gain access to areas that promote recruitment to the spawning stock, whereas vagrants are not retained, * Corresponding author: [email protected] Received December 18, 2007; accepted September 12, 2008 Published online March 23, 2009

either perishing or contributing to other populations. The hypothesis recognizes links between spawning areas and times, larval retention areas that ensure continuation of the population, and the role of vagrants in starting new populations or contributing to sink populations (Sinclair 1988). Vagrancy, mediated by anomalous surface transport, has been suggested as a major factor in interannual differences in year-class success (Parrish et al. 1981). The original example of a retention mechanism was an oceanographic gyre retaining larval Atlantic herring Clupea harengus in restricted areas (Iles and Sinclair 1982), but tidal currents and two-level circulation

328

329

UPPER KLAMATH LAKE FISH RECRUITMENT

patterns with countercurrents have also been suggested for marine species (Fortier and Gagne´ 1990; Hare et al. 1999). Although these aggregating mechanisms exist in many habitats, downstream transport is dominant for both marine (Hare et al. 2002) and freshwater fish larvae (Brown and Armstrong 1985). When there is downstream transport, behavior or ontogenetic changes in anatomy are hypothesized to moderate advection to avoid vagrancy (Manteifel et al. 1978; Taradina et al. 1997; Hare et al. 2002). Nocturnal drift, which is common in freshwater fishes (Brown and Armstrong 1985), may be one such behavior, but to date, there are few examples of retention mechanisms in freshwater fishes. One analytical problem is distinguishing mortality and migration, primarily the consequences of advection (Taggart and Leggett 1987a). Over small spatial and short temporal scales, advection can account for 32– 66% of apparent mortality (Taggart and Leggett 1987b) and can greatly distort mortality estimates (Helbig and Pepin 1998a). The fishes in Upper Klamath Lake, Oregon, constitute a freshwater system with which to test the member–vagrant hypothesis. Five resident fish species—Lost River sucker Deltistes luxatus, shortnose sucker Chasmistes brevirostris, blue chub Gila coerulea, tui chub G. bicolor, and fathead minnow Pimephales promelas—have protracted spawning from spring to early summer and differ in abundance. The suckers tend to spawn earlier at known lakeshore springs and lake tributaries (Lost River suckers initiate spawning before shortnose suckers), followed by the chubs and fathead minnows at largely unknown sites. Markle and Dunsmoor (2007) showed that the interannual variability (1995–2005) in larval sucker survival from 10 to 15 mm (both species combined) ranged from 4.4% to 48%, and survival appeared to be constrained such that low fathead minnow abundance and high lake elevation (a surrogate for the volume of emergent vegetation) were associated with higher and more variable survival. M. S. Cooperman (NOAA Fisheries, Orono, Maine) and D. F. Markle (unpublished data) examined survival in 29 larval sucker cohorts (10–15 mm) during 1995–2001 in relation to habitat and climate and found it depended on volume of emergent vegetation and wind events. The negative effect of high wind speed and the positive effect of habitat volume could be related to several factors, including feeding success and advection, but Markle and Clauson (2006) found no evidence that larval prey were limiting in this hypereutrophic system. They

suggested that shoreline configuration and winds helped to accumulate larvae in emergent vegetation, which is a predation refuge. Because sucker spawning sites are well known, we are able to examine larval transport in Upper Klamath Lake in the context of the member–vagrant hypothesis. We use patterns of larval sucker mortality to infer siteor area-specific immigration and emigration processes for the 2004 year-classes of larval suckers. We linked these patterns of mortality and movement of larval suckers in 2004 to a hydrographic model of the lake’s circulation and confirmed the general results via limited sampling in 2005. If advection dynamics of larvae are a controlling aspect of year-class formation, we would expect the average individual’s location to be closer to the lake outlet when vagrancy was greater and juvenile abundance to be positively associated with distance from the lake outlet. We evaluated the generality of this expectation by examining 11 years (1996–2006) of standardized August juvenile abundance and distribution data for five abundant shoreline fishes, including both sucker species. Study Site Upper Klamath Lake is the largest of the remnant lakes of Pleistocene Lake Modoc (Orr et al. 1992) and covers approximately 34,400 ha at a water surface elevation 1,262.9 m (U.S. Bureau of Reclamation data). Upper Klamath Lake is connected to Agency Lake to the north and receives most of its inflow from the Williamson River and its largest tributary, the Sprague River (Figure 1A). The lake is generally shallow (average depth, ,2.5 m), especially along the eastern margin, where depths 1 km from shore are often less than 1.5 m. Several shoreline springs found on the eastern shore (e.g., Sucker Springs) are important sucker spawning areas (Markle and Cooperman 2002). The eastern shoreline is also bounded by faulted topography that is important in channeling wind southward. At the southern terminus of the lake, the Link River connects Upper Klamath Lake with Lake Ewauna and Keno Reservoir, a 32.2-km-long body of water that includes the uppermost section of the Klamath River downstream to an irrigation dam at Keno, Oregon (Figure 1A). Throughout, we use the term ‘‘Ewauna–Keno’’ when combining data from these two areas. Link River Dam at the terminus of Upper Klamath Lake is used to store and divert water for the Klamath Irrigation Project and to regulate water flow downstream. It was built on a natural basalt reef that was

330

MARKLE ET AL.

about 16,000 ha of wetlands have been diked and drained (USFWS 2001). Recently, there has been interest in restoring the lake’s deltaic and lakeshore marshes, both to improve water quality and to provide larval habitat for endangered suckers. In 2000 and 2003, The Nature Conservancy (TNC) and its partners completed several small-scale (,75 ha) pilot wetland restoration projects at the Williamson River Delta Preserve. Restoration at the 11-ha Riverbend site occurred in the fall of 2000 on the north bank of the Williamson River, 5 km upstream from the mouth (Figure 1A). The South Marsh site, located near the southern extent of the historical Williamson River Delta and adjacent to Upper Klamath Lake, was allowed to flood via seepage through the levees in 1997. In the fall of 2003, the South Marsh levees were breached in two locations reconnecting 70 ha to Upper Klamath Lake.

FIGURE 1.—Panel (A) shows the locations of catostomid and cyprinid larval sampling sites and areas in the Upper Klamath basin in 2004. Sites C (Chiloquin) and M (Modoc Point Road) were sampled by the U.S. Geological Survey; sites RB (Riverbend), WR (Williamson River), and SM (South Marsh) were sampled by the Nature Conservancy (sites RB and WR are adjacent to one another and overlap on this map); and the sites indicated only by small circles were sampled by Oregon State University. The five areas into which we divided the basin are indicated by dashed lines. Panel (B) shows an example of the distribution of stratified, random cast-net sites (dots) around the shoreline of Upper Klamath Lake from the 2005 juvenile survey.

modified to allow the lake to be lowered below the natural sill elevation (USFWS 2001). Historical (predam, 1904–1921) lake elevations fluctuated between 1,262.2 and 1,263.1 m. Postdam (1921 to the present) lake elevations have varied as much as 2.0 m in a year (1,261.1–1,263.3 m; Boyle 1976; USFWS 2001). Outflow varies seasonally and interannually. During the period of relevance for larval advection in this study, 1 May to 30 June 2004, the average daily flow rate from the lake to Link River was 27.4 m3/s and was 18.7 m3/s from the lake to the Klamath Irrigation Project through the A-Canal. Around the perimeter of Upper Klamath Lake,

Methods Hydrography.—Based on the UnTRIM numerical model of Casulli and Zanolli (2002), we developed a 3-dimensional model of the hydrodynamics of Upper Klamath Lake on an unstructured grid. Two Acoustic Doppler Current Profilers were deployed in the lake during summer (June-September) in 2003 and 2004, which we used to calibrate an early version of the model. During these months, wind data were collected at one or two sites from rafts in the lake (Wood et al. 2006) and data from one was used to force the model with a spatially invariant wind field. This model (Wood et al. 2008) demonstrated that currents in the lake were determined primarily by wind forcing at the surface of the lake and improvement in the model’s performance would be achieved by resolving the spatially variable wind field over the lake. In 2005 two wind anemometers were installed on the lake and four were installed around the lake shoreline to better define spatial variability in the wind field. Five of the six sites were collecting accurate data by July 26, and all were collecting accurate data by August 18 (Wood et al. 2008). An atmospheric boundary layer model that interprets the surface winds between sites was incorporated into the UnTRIM model (Wood et al. 2008). A comparison of simulated velocities to velocities measured with the profilers deployed on the lake bottom during the calibration process showed that the use of a spatially variable wind field resulted in lower mean and root mean square errors between the simulated and measured results (Wood et al. 2008). Thus, the most accurate wind forcing function for the model was for the period 26 July to 12 October 2005, which did not


331

FIGURE 2.—Panel (A) shows the residual wind-driven circulation pattern in Upper Klamath Lake, the wide arrows indicating the gyre, the narrow arrows the eastern boundary current, and the diamond the representative wind site (MDL). The other three panels depict the modeled 20-d tracks of particles released into the flow at three different tributary inputs under strong and weak prevailing winds (distances between dots ¼ 1 d); panel (B) depicts the track of particles released at the mouth of the Williamson River and retained in Upper Klamath Lake, panel (C) that of particles released at an eastern shore spring (Sucker Springs) and lost from the lake, and panel (D) that of particles released at the mouth of the Williamson River and lost from the lake. All simulations were based on lake elevation and air temperature measurements during April and June 2004. The strong and weak wind-forcing functions were deriveded from wind data collected after 26 July 2005.

overlap the March–July 2004 period of larval collections. There are no wind data from the lake from March to June 2004 because this preceded the summer installation of the rafts. To better understand the seasonal patterns, we

examined wind data collected by the Bureau of Reclamation for an Agrimet site at the northern end of Agency Lake. Because all of the other boundary conditions and forcing functions required to run the model during the March to June 2004 period were

332

MARKLE ET AL.

available (tributary inputs, discharge at the outlet of the lake, air temperature, relative humidity, solar insolation, and lake elevation), we used an idealized wind forcing to provide insight into larval transport, even though it is not a direct simulation of observed winds. We constructed two surface wind-forcing scenarios, the first from characteristically strong prevailing winds and the second from characteristically weak prevailing winds (see Wood and Cheng 2006). Both strong and weak prevailing winds are north to northwesterly (coming from 3008 to 3508 from true north at the representative wind site (MDL in Figure 2). The strong winds were characterized by median and 90thpercentile speeds of 6.5 and 9.4 m/s, respectively, whereas the weak winds were characterized by speeds of 3.1 and 5.7 m/s. Model simulations were run for 20 d using both wind forcing scenarios. The use of strong and weak prevailing winds was intended to bracket the expected winds and gain insight into the relation between wind strength and transport of a passive particle. Groups of numerical particles were released into the model on a weekly basis (starting on 1 April and ending on 23 June) from an eastern shoreline spawning location, Sucker Springs, and from the mouth of the Williamson River. The particles were used as a first-order estimate of travel time and retention of larvae. Larval samples.—Both sucker species spawn at lakeshore springs and in the lower reaches of the Williamson and Sprague rivers over gravel and coarse substrates (Scoppettone and Vinyard 1991). Larvae hatch at about 7–10 mm notochord length (Hoff et al. 1997) and generally emerge from the gravel when they are longer than 10 mm and older than 10 d. In 2004, we conducted three coordinated larval sucker surveys, as follows: The Nature Conservancy (TNC) evaluated daytime larval sucker use of restoration marshes, the U.S. Geological Survey (USGS) evaluated nighttime patterns of larval sucker production in the Williamson and Sprague rivers before a planned removal of a dam at Chiloquin, Oregon, and Oregon State University (OSU) evaluated daytime larval sucker production and distribution in Upper Klamath Lake, Lake Ewauna, and Keno Reservoir. The Nature Conservancy sites were located at the restoration marsh at Riverbend, an area adjacent to the Williamson River and the restoration marsh at South Marsh (Figure 1A). Sucker larvae were collected with fine-meshed (1,000-lm bar mesh) dip nets in daylight (0800–1600 hours) using five to eight 15-min line transects under calm and clear water conditions from late April to late July (see Crandall et al. 2008 for details). All sites were sampled once

every 7–12 d, all sites and transects being sampled within a 3-d period. The two USGS sites were located at bridge crossings on the Sprague River at Chiloquin, (site C; river kilometer [rkm] 0.75), which was about 200 m downstream of the largest known sucker spawning area, and on the Williamson River at Modoc Point Road (site M; rkm 7.4; Figure 1A). Because riverine drift occurs at night (Cooperman and Markle 2003), samples were collected from 1732 to 0631 hours, 80.8% being collected between 2100 and 0200 hours. Sampling was three times per week from 6 April to 25 June (or 1 July 2004 at site C). Each night, 6–24 total net samples were collected, usually within a 2-h interval, which provided a total of 339 samples at site C and 375 at site M. Samples were collected with a 0.3m-diameter-opening plankton net that was 2.4 m long (800-lm Nitex mesh and 500-lm Nitex mesh collection cup). A deployment–retrieval rope and a 3.5-kg pancake weight were attached to opposite sides of the stainless steel collar at the opening of the net to ensure nets fished perpendicular to currents. At each site, two nets were deployed simultaneously at the surface and midwater column depth for approximately 10 min. Larvae were fixed in 5–15% solutions of formalin, sorted within 24 h, and stored in 95% solutions of ethanol. The OSU sites were located along the shorelines of Upper Klamath Lake, Lake Ewauna, and Keno Reservoir (Figure 1A) and were selected based on availability and access, the intent being broad coverage and a range of habitats. For some comparisons, sites were grouped into five areas from north to south, with approximately equal distances from spawning sites: area 1 ¼ the northern and western shoreline; area 2 ¼ the eastern shoreline, including the mouth of the Williamson River and shoreline springs; area 3 ¼ the southern shoreline; area 4 ¼ Lake Ewauna; and area 5 ¼ Keno Reservoir (Figure 1A). The OSU sampling used a larval trawl with a 0.8-m 3 1.5-m opening and a 2.5-m, 1,000-lm bar mesh Nitex net mounted on an aluminum frame with runners (LaBolle et al. 1985). The trawl was set 3–12 m offshore in water up to 1 m deep (range, 0.2–1.0 m), allowed to soak for 10 min, and then pulled to shore with ropes. Sampling began the first full week of April and continued through late July; samples were collected every third week, for a total of six sampling surveys. A total of 120 samples were collected from 10 sites that were sampled twice per survey in Upper Klamath Lake, and 180 samples were collected as part of long-term monitoring (Reithel 2006) from 30 sites sampled once per survey in Ewauna–Keno. During the 2-week interval between


these surveys, we sampled 12 additional sites at the southern end of Upper Klamath Lake (area 3) and 12 in Lake Ewauna (area 4). Nine of these surveys were conducted from late March through June with two samples per site per survey (N ¼ 432). During eight of the 2-week intervals, we also sampled 10 sites at or near eastern shore springs (area 2) with one or two samples per site (N ¼ 82). Larvae were preserved in 5% formalin for 24–36 h and stored in 70% ethanol. Two additional sources of data were used to compare temporal trends. In 2005, long-term larval trawl sampling in Upper Klamath Lake and Keno was augmented with sampling in Riverbend and South Marsh and the age distributions patterns compared with the more intensive sampling in 2004. Differences in larval extrusion between gears were considered minimal. We used the diagonal of the mesh as the limit of absolute retention (Smith and Richardson 1977), and body depth at the cleithrum (BD) as the limiting structure. The relationship to notochord or standard length (SL) was: BD ¼ 2.63 þ 0.322 SL (Remple and Markle, unpublished data; N ¼ 576, R2 ¼ 92.8%), suggesting absolute retention in all gears was less than 12.5 mm. Because the plankton net is deployed in a current for a longer duration than the other gears, extrusion should be greater but the absolute retention of the 500-lm mesh was even smaller, 10.4 mm. Differences in avoidance between the larval trawl and 1-, 3-, and 5-min dip-net transects were evaluated from 17 May to 15 June, 1994, at the mouth of the Williamson River. Over the 12.5–15.5mm size range available during these dates (about 20– 35 d), there were no differences in the slopes (P ¼ 0.91) or intercepts (P ¼ 0.51) of the declining limbs of the catch curves. Nighttime USGS sampling could not be compared directly with the daytime sampling. However, Cooperman and Markle’s (2003) data suggest that 0.3-m plankton net sampling in the river is representative of the river population sampled with daytime shoreline pop nets, with similar proportions of preflexion (84–74%) and postflexion (0.002–0.02%) sucker larvae. Because comparisons between sites must be tempered by potential gear differences, we used one gear for most direct comparisons below. Larval sucker identification and aging.—Shortnose suckers have 41–44 post-Weberian vertebrae and Lost River suckers have 44–48 (Markle et al. 2005). A third co-occurring species, the Klamath largescale sucker Catostomus snyderi, has vertebral counts similar to the shortnose sucker and cannot be distinguished in the larval stage, but juveniles can be distinguished. In Upper Klamath Lake, juvenile Klamath largescale suckers are rare, representing less than 2% of all juvenile suckers (OSU and USGS,

333

unpublished data). However, they would be expected in the river, so some misidentifications are probable in those collections. All three species hybridize (Markle et al. 2005). We counted myomeres, which have a 1:1 relationship with vertebral counts, for 574 larvae (10.2–24.9 mm SL) and found that pigmentation patterns correlated with myomere counts. Interspecific differences between shortnose and Lost River suckers were found in number of melanophores along the lateral line, dorsal pigmentation patterns, and number of snout and head melanophores (Remple and Markle, unpublished data). Some larvae had intermediate patterns and were classified as unidentified. For the OSU data set, these larvae were cleared and stained and their vertebrae were counted. Those larvae with 44 post-Weberian vertebrae remained unidentified. For all data sets, we assigned unidentified larvae to the most numerous species within that age-class on that sample day. We did this because Lost River suckers generally have earlier hatch dates, so that there were species-specific age differences in most samples. Age determination followed the methods of Terwilliger et al. (2003), a procedure with a relatively high reading precision (average error, 2.2%). We used lapilli ground by hand along the sagittal plane on both sides to create a thin section with visible increments along the entire surface. Daily increments, verified by Hoff et al. (1997), were counted along a consistent transect from core to edge along the leading growth axis using a digital imaging system. Three independent counts were made, and all counts were made without knowledge of fish length or catch date. The median count was chosen. Size-at-age regressions were compared between species and months and resulted in monthly age–length relationships. Comparison of regressions used a general linear model (GLM). To compare species within a month, species was treated as a categorical variable scored as 0 or 1; the model was thereby reduced to a simple regression for each species, so that the slopes and intercepts could be compared (Tabachnick and Fidell 2001). Final regressions were based on month and similarly scored, which reduced the model to a simple regression for each month. All fish, including the aging subset, were assigned ages based on monthly age–length regression from the GLM. Larval analyses.—We used a loge-transformed abundance catch curve to estimate total mortality, Z (Quinn and Deriso 1999) for a site. Thus, we assume mortality is constant over time. For the OSU collections, the numbers of larvae were pooled in 5-d age-class bins for analyses (e.g., age-class 17 pooled ages 15–19). For each analysis, the youngest or

334

MARKLE ET AL.

FIGURE 3.—Hypothetical age-frequency distribution for a species with upstream spawning and downstream transport. Panel (A) shows that the peak abundance at the upstream site (T1) precedes that at the downstream site (T2) by the average number of days required for transport from upstream to downstream. The downstream curve will be flat from T2 to T3 when immigration to the downstream site balances natural mortality. Panel (B) shows natural mortality (M) in a closed system, which is reflected in the declining limb of the pooled data. Emigration and immigration (G) are deviations from M, emigration (G , 0) being the dominant process upstream and immigration (G . 0) the dominant process downstream.

threshold age was determined from the age-class with maximum abundance and the oldest age was conservatively truncated when the abundance of an age-class was less than 10 (Seber [1982] recommends truncation with an abundance less than 5). For collections with large numbers and few age-classes, we used the abundance in a daily age-class and truncated when abundance was less than 5. The expectations for larval catch curves showing dispersal from an upstream site to a downstream site are summarized in Figure 3. The youngest ages should be more abundant near the spawning grounds (Smith 1972), and peak abundance should occur when all individuals have swum up and recruited to the gear (at time T1 in Figure 3A). Thus, the ascending limb at a site would be expected to reflect variation in the age of swim-up, advection–retention dynamics of the site, and cumulative mortality at age. In our data, small numbers of pre-swim-up yolk sac larvae (,10 d old) are included at near-natal sites (C, M, and WR) and are

probably prereleased from turbulence at the base of Chiloquin dam. The difference in age between peak abundance at the downstream and upstream sites (e.g., T2 T1 in Figure 3A) is an approximation of the travel time between sites. A flat top, or uniform size distribution at the downstream site, implies that immigration at that site is balancing mortality over the age range encompassed by the flat region (e.g., T3 to T2 in Figure 3A). The slope of the declining limb of the catch curve (Z, usually interpreted as natural mortality or M; Quinn and Deriso 1999) is confounded by migration, G, such that Z ¼ M þ G where G can be positive (immigration) and reduce the slope or negative (emigration) and increase it. If the system is closed, pooled data from the upstream and downstream sites provide the best estimate of M. The upstream site will be expected to have a steeper slope because there is net emigration (G , 0) and the downstream site will be expected to have a shallow slope because there is net immigration (G . 0; Figure 3). Following Pepin et al. (1995), who argued that advective losses are inversely proportional to area sampled, we combined OSU Upper Klamath Lake plus Ewauna–Keno data as the best estimate of each species’ natural mortality, M, because emigration and immigration should cancel and G ¼ 0. Migration at a site, S, was therefore estimated as GS ¼ ZS Z(combined), where Z(combined) ¼ M for Upper Klamath Lake and Ewauna–Keno combined. Because recruitment to a site can create positive slopes for an age-class (the appearance of no mortality), we deviate from convention and include the sign for all slopes of Z, M, and G. We also calculated life table statistics (Quinn and Deriso 1999) from catch-at-age data pooled in 5-d ageclass bins for all collections. We used Ca, the catch at age a, to calculate survival, Va, as Caþ1/Ca, where Caþ1 is the catch in the next older age category. Total mortality, Za, was estimated as loge(Va). Again, we assumed that all local emigrants were sampled in the combined Upper Klamath Lake and Ewauna–Keno data and estimated natural mortality for each species, Ma, from loge(Va) from this combined data set. We therefore estimated Ga at a site, S, as GaS ¼ ZaS Za(combined), where Ma ¼ Za(combined). Juvenile samples.—We determined shoreline fish distribution and abundance in August 1996–2006 via sampling with a 5-m-diameter multifilament cast net with 6.3-mm-bar mesh within 10 m of the shoreline at depths less than 1.5 m; sampling followed a stratified random design (Figure 1B). The catch efficiency of enclosure devices is generally thought to be influenced by few variables and not to vary significantly in typical shallow-water habitats, at least in estuaries (Rozas and Minello 1997). The cast-net catch efficiency from

335


enclosure sampling showed no differences in median density between the cast-net estimate and the initial enclosure density (P . 0.18) and no size bias in a signed rank test for 5-mm size-groups (10–80 mm TL) of three abundant species from (P . 0.29; OSU, unpublished data). We stratified shoreline sampling based on substrates. In fall 1994, low lake levels allowed us to construct a linear map of the shoreline substrates around the perimeter of Upper Klamath Lake. We created 2,595 shoreline sites located at 50-m intervals and assigned each site to one of eight substrate categories: fines (,0.06-mm particles; 1,244 segments), sand (0.06–2 mm; 215), gravel (2–64 mm; 81), cobble (64–250 mm; 265), boulder (.250 mm; 66), small mixed substrates (.75% of particles ,64 mm; 116), intermediate mixed substrates (no preponderance of any substrate; 210), and large mixed substrates (.75% of particles .64 mm; 398). Along vegetated shorelines, the substrate along the margin was described. We chose 50-m intervals because, at the beginning of the study, differential corrections of GPS introduced uncertainty in real time coordinates. Based on preliminary work from 1991 to 1994, effort was stratified by substrate to maximize sampling of substrates that had higher sucker catch rates as well as ensure that a minimum of five sites were sampled in each substrate. Sites were selected randomly with replacement, and if a site was selected twice, the second sample was taken in an undisturbed area near the first site. A hand-held GPS unit was used to locate sites. Actual substrate type at each site was verified, and bottom water temperature, dissolved oxygen, pH, and conductivity were measured using a Hydrolab Reporter Multiprobe and Surveyor 3 Display Logger. The approximate percent effort per actual substrate was: fines (22%), sand (9%), gravel (17%), cobble (16%), boulder (3%), small mixed substrates (18%), intermediate mixed substrates (8%), and large mixed substrates (6%). Surveys were conducted in late August of 1996 to 2006; 140 sites were sampled in each survey (Figure 1B). Cast-net samples were collected from nonvegetated or sparsely vegetated locations or along the edge of vegetation wherever the shoreline was vegetated. Age-0 suckers were preserved in 95% ethanol; the remaining catch was subsampled if needed, identified, measured, and counted. Juvenile analyses.—The mean number of individuals of a particular species per cast net and survey yst was calculated as ! L X yst ¼ Nh yh =N; h¼1

L ¼ the number of strata (here 8); Nh ¼ the number of available units in stratum h; yh ¼ the mean number of fish in stratum h; N ¼ the total number of sampling units available (here 2,595). Variable yh was calculated as yh ¼

nh X

! yhi =nh ;

i¼1

nh ¼ the number of units sampled in stratum h; yhi ¼ the number of fish in the ith unit of stratum h. The mean number of individuals per cast net per survey was converted to density using the average net coverage area (5 m2); this was expanded to shoreline abundance within 10 m of the shore by multiplying density by the total area within 10 m of the shore (i.e., shoreline circumference 3 10). Logarithmic transformations of these annual abundance estimates were used to describe species location patterns. For each species in each cast-net survey, we calculated a center of distribution as the average latitude and longitude of each positive catch for a species and a center of abundance as the average latitude and longitude of each specimen of a species (i.e., an abundance-weighted center of distribution). Because the distance from the lake outlet is best approximated by latitude, we examined the relationship between the latitude of the August center of distribution and overall retention. Log of shoreline abundance in August was evaluated using a GLM analysis of variance (ANOVA) based on predictors of species, lake elevation on July 15, and latitude of the center of distribution. Lost River suckers were not captured in August 2003, so a center of distribution could not be calculated. We used statistical programs SYSTAT version 8.0 and Statgraphics Centurion XV version 15.0.00. Results Particle Transport The topography surrounding Upper Klamath Lake channels wind over the lake along a NW to SE axis, especially in the southern two-thirds of the lake, but a W to slightly WNW axis prevails over the northern third. Under prevailing NW winds, the residual flow in Upper Klamath Lake is a clockwise gyre extending north near the shoreline near Agency Lake and south near Buck Island (Figure 2). East of Bare Island in the north there is a broad, shallow flow to the southeast that returns as a narrow, deep northwest flow through a

336

MARKLE ET AL.

trench west of the island (Wood et al. 2008). The relatively small fraction of the flow exiting the lake is confined to a narrow boundary current along the eastern shoreline that passes east of Buck Island in the south. Particles leave the lake if they never get out of the eastern boundary current, but they will be retained in the lake if they are entrained in the clockwise flow that skirts the north side of Buck Island. Strong prevailing winds drive a stronger clockwise circulation than weak prevailing winds, and particles are more likely to be entrained in the gyre and stay in Upper Klamath Lake under strong wind conditions. Particles released at Sucker Springs on the eastern shore always left the lake under weak prevailing winds but showed variable retention (0–60%) under strong wind conditions (see depictions in Figure 2). Representative particles retained in the lake illustrate that the gyre pushes farther south and east under strong wind conditions, making it is easier for a particle to become entrained; this explains the greater retention under strong wind conditions. For example, during April to June under the weak wind scenario, average retention for particles released from the Williamson River was 5.7% and from Sucker Springs was 0%. During April to June under the strong wind scenario, average retention for particles released from the Williamson River was 36.2% and from Sucker Springs was 20.1%. Overall, retention was greater for particles released from the Williamson River than for particles released from Sucker Springs. Wind data collected from the Agency Lake Agrimet site indicate that April and May winds tend to be stronger than summer (June–August) winds, although the interannual variation is large. In 2004, for example, the 90th percentile of hourly wind speed distribution was 6.1 m/s in April and 2.8 m/s in July. In addition, spring winds may be characterized by more frequent wind reversals (usually 1–2 d) from the prevailing direction. In April 2004, 36% of hourly wind vectors originated in the SE or SW quadrants (91–270 degrees), whereas in July 2004, 25% percent of hourly wind vectors originated in the SE or SW quadrants. Our particle-tracking model suggests that spring retention in the lake is more closely represented by the strong wind scenario and summer retention by the weak wind scenario. The model has shown that wind reversals during the summer have the effect of stalling the prevailing circulation pattern, although a reversal lasting beyond 1 d may reverse the circulation direction. The effect of wind reversals is most likely to slow advection through a combination of slowing

travel and increasing the chances of particles being entrained into the gyre. Although strong winds increase the proportion of retained particles, those particles not retained also reach the lake outlet faster under strong wind conditions. Representative travel times to the lake outlet are 5–6 d (Sucker Springs) for strong prevailing winds and 8–17 d (Williamson River mouth) for weak winds (Figure 2). Other variables, in particular discharge at Link River Dam, also affect travel time out of the lake. Thus, interannual differences in advection potential would probably be due to a complex of variables in addition to wind. Larval Sucker Identification and Aging The percentage of initially unidentified sucker larvae assigned to a species was 28% for TNC, 11.5% for USGS, and 13.4% for OSU. In general, identifications were easier with younger larvae, such as those in the USGS night drift samples. Both the TNC and OSU samples had older larvae but clearing and staining increased the number of larvae that could be identified in the OSU samples. We aged 624 larvae (10.4–20.0 mm SL); length-atage regressions between species for each month produced similar results. For the relatively large sample sizes in May (N ¼ 187, R2 ¼ 68.0%), length-at-age relationships from the GLM reduced to: Lost River suckers (length ¼ 10.3716 þ 0.121694 3 age; N ¼ 54) and shortnose suckers (length ¼ 9.9966 þ 0.147677 3 age; N ¼ 133). Slopes were not different (P ¼ 0.11) but intercepts were (P ¼ 0.02), suggesting that growth is similar but that Lost River suckers are 0.4 mm larger at hatch. Hatchery data also indicate Lost River suckers are longer at hatch (Reithel 2006). Terwilliger et al. (2003) likewise found no differences in growth rates of shortnose suckers and Lost River sucker juveniles. Because growth rates were not different, we ignored the small difference in hatch size and pooled species for subsequent age analyses. The effects of length and month on age were compared assuming equal intercepts in the GLM. Slopes were significantly different among months (P , 0.0001), and a single model was constructed (N ¼ 624, R2 ¼ 76.9%) that was reduced for each month: April (age ¼ 33.1186 þ 4.38291 3 length, where N ¼ 71), May (age ¼ 33.1186 þ 4.34396 3 length, where N ¼ 231), June (age ¼ 33.1186 þ 4.15556 3 length, where N ¼ 192), and July (age ¼ 33.1186 þ 4.04082 3 length, where N ¼ 127). Thus, for a 15-mm larva, the age difference

337


TABLE 1.—Catch-curve analysis for Lost River sucker and shortnose sucker larvae from Upper Klamath basin in 2004. Sampling sites and abbreviations are described in Figure 1A, except E–K (Lake Ewauna and Keno combined) and UKL (Upper Klamath Lake). Mortality (Z) was estimated from the declining limb of the log abundance catch curve over the range of ageclasses indicated; adjusted R2 and P values are from the catch-curve regression; migration (G) was estimated as Z Z(combined), where ‘‘combined’’ ¼ data for Upper Klamath Lake plus Ewauna-Keno. Asterisks indicate the use of unpooled daily catch data. Age range (d)

Site or region

*18–29 *19–25 *15–20 *17–27 *17–27 17–37 22–37 17–37

Site C Site M Site WR Site RB Site SM UKL E–K Combined

*18–26 *18–24 *15–31 *17–38 *20–39 17–37 22–37 17–37

Site C Site M Site WR Site RB Site SM UKL E–K Combined

a

Z Lost River suckers 0.54 0.33 0.30 0.26

R2

P

G

94.9 75.6 69.1 85.8

,0.0001 0.0031 0.0252 0.0001

0.51 0.30 0.27 0.23

78.5 38.8 75.8

0.0288 0.7270 0.0348

0.05 þ0.02

90.9 65.8 87.4 93.8 74.5 94.3 59.9 86.4

,0.0001 0.0090 ,0.0001 ,0.0001 ,0.0001 0.0038 0.1442 0.0143

0.60 0.26 0.19 0.11 0.02 0.05 þ0.02

a

0.08 0.01 0.03 Shortnose suckers 0.68 0.35 0.28 0.19 0.10 0.14 0.06 0.09

Insufficient data.

between a slow grower in April (32.6 d) and a faster grower in July (27.5 d) is 5.1 d. Larval Sucker Catch-Curve Analysis All catch-curve slopes were significantly less than zero, except those from Ewauna–Keno (Table 1). For both species, Z increased in the direction of larval drift from 0.54 to 0.68 at site C in the Sprague River to 0.01 to 0.06 in Ewauna–Keno. Assuming that the combined Upper Klamath Lake and Ewauna–Keno data account for emigration and immigration, its estimate of Z is the best estimate of each species’ natural mortality (0.0262/d for Lost River suckers and 0.0857/d for shortnose suckers; Table 1). These values provide the best estimates of overall mortality in each species, and the difference between the mortality estimate from an individual location and the overall value of Z provides a measure of immigration and emigration, G, occurring at the individual site. A relatively high and negative G suggests rapid emigration; G was highest in the Williamson (sites M and WR) and Sprague (site C) rivers (Table 1). The average G at these three sites was 0.36 for Lost River suckers and 0.35 for shortnose suckers. Ignoring mortality, this suggests that about 95% of a larval sucker cohort should be out of the river within 8–9 d after swim-up. The age-frequency distribution for both species at site M also showed abundance approached zero about 9 d

after peak abundance (Figure 4A). Relative to larvae in the adjacent river (site WR), the Riverbend restoration wetland in the Williamson River Delta (site RB) reduced emigration slightly for Lost River suckers (0.23 versus 0.27) and by about half for shortnose sucker (0.11 versus 0.19). For shortnose suckers at site RB, we estimate that 95% leave in about 28 d because their age-frequency distribution showed abundance had declined 97% over 21 d from age 17 d (exponent ¼ 5.91) to age 38 d (exponent ¼ 2.28). Similarly, the lakeshore restoration wetland at South Marsh (site SM) reduced shortnose sucker emigration by about half compared to Upper Klamath Lake emigration (0.02 versus 0.05). After 28 d, about 76% of shortnose sucker larvae had emigrated from the lake, whereas only 42% of shortnose sucker larvae in South Marsh had emigrated from the marsh back into the lake. Lost River sucker larvae were not caught in sufficient quantity at South Marsh for analysis. Estimates of G were negative at all sites except Ewauna–Keno, where Z was also not different from zero (Table 1). For Lost River suckers 20–39-d old (age-class bins 22–37 d), immigration exceeded the effects of emigration in Ewauna–Keno by 1.5%/d. For 20–39-d-old shortnose suckers in Ewauna–Keno, immigration exceeded emigration by 2.2%/d. At these rates the numbers of Lost River suckers in Ewauna– Keno would increase by 52% in 28 d and the number

338

MARKLE ET AL.

suckers differed from Lost River sucker in having a very low but constant abundance (Figure 5D). In 2005, age patterns were similar. Mean ages of Lost River sucker larvae in the lake (19.9 d) and marshes (20.6 d) were similar and about a week younger than those in Ewauna–Keno (29.8 d; ANOVA, P ¼ 0.0001). The age of shortnose sucker larvae differed depending on whether they were occupants of the lake (20.8 d), the marshes (22.4 d), or Ewauna– Keno (34.4 d; ANOVA, P , 0.00001), and again there was a downstream increase in age. Overall, the larval suckers in marshes (27.5 d) were about 4 d older than the lakewide mean (23.3 d) and those in Ewauna–Keno (31.1 d) were about 8 d older. Larval Sucker Life Table

FIGURE 4.—Larval catch curves for shortnose suckers from (A) U.S. Geological Survey samples from site C and downstream site M, in which the steep curves indicate rapid emigration, and (B) Nature Conservancy samples from site WR and the adjacent restored marsh (site RB), in which the relatively flat curves indicate that immigration at ages 18–24 d is sufficient to compensate for natural mortality.

of shortnose suckers in Ewauna–Keno would increase by 86%. At a finer scale, within Upper Klamath Lake and Ewauna–Keno the age structure patterns also showed downstream dispersal. For Lost River sucker larvae, the maximum abundance occurred on the eastern (area 2) and southern (area 3) shorelines, where the 17-d ageclass (Figure 5A) predominated and abundance declined steadily at older ages. In Lake Ewauna (area 4), the maximum abundance of Lost River suckers was also attained by the 17-d age-class, but abundance stayed relatively constant through the 42-d age-class. In Keno Reservoir, maximum abundance was reached slightly later at the 22-d age-class but also had little or no decline through the 42-d age-class (Figure 5C). Shortnose suckers had a very similar pattern, the maximum abundance in areas 2, 3, and 4 being reached by the 17-d age-class; the abundance of older ageclasses declined in areas 2 and 3 but remained constant or declined only slightly in area 4 until the 37-d ageclass (Figure 5B). In Keno Reservoir, shortnose

For both species at all sites, there were age-specific differences in Ga. For 17–22-d-old Lost River suckers, emigration occurred at all sites, including Ewauna– Keno (Table 2). Lost River suckers 27 d or older were essentially absent from river and restoration marsh sites, were somewhat more important components of shoreline samples from Upper Klamath Lake, but were especially abundant in Ewauna–Keno. Immigration exceeded emigration for the 27-d-old and 32-d-old Lost River suckers in Ewauna–Keno (Table 2). In contrast, the relative effects of emigration and immigration were variable for 17–22-d-old shortnose suckers, and sites M, RB, SM and Ewauna–Keno had relatively more immigration than emigration for this age-group (Table 2). Older (27 and 32 d) shortnose suckers were important components of river and restoration marsh sites. A comparison of shortnose sucker age structure at RB and the adjacent WR sites illustrates the system dynamics. Young larvae move through the river quickly at WR (Figure 4B), emigration exceeding immigration at 17 d (Ga ¼ 1.82), equivalent to an emigration rate of 99% in 5 d (Table 2). At the same age range, there was a net movement (Ga ¼ 0.24) of shortnose sucker larvae into RB that was equivalent to a doubling of immigrants (2.053) in 3 d (Table 2; Figure 4B). From ages 18–22 d there was no apparent mortality in RB (Figure 4B) suggesting shortnose sucker larvae were continuing to immigrate to RB at a rate that compensated for natural mortality (13–15%). After about age 22 d, emigration exceeded immigration and larvae left RB at rates slightly lower than those for the river site, a reduction of 86% after 3 d at RB versus 92% at WR (Table 2). The comparison suggests RB was a net retainer of shortnose sucker larvae up to age


339

FIGURE 5.—Larval catch curves for samples taken in Upper Klamath Lake (UKL) and Lake Ewauna–Keno Reservoir in 2004 (see Figure 1). Panel (A) pertains to Lost River suckers and shows the increase in older age-classes from area 2 on the eastern shore to area 3 on the southern shore to area 4 in Lake Ewauna; panel (B) pertains to shortnose suckers and shows a similar pattern; panel (C) pertains to Lost River suckers and shows the approximately 5-d delay in peak abundance in Ewauna–Keno and the flatness of the Ewauna–Keno curve, which indicates that the relatively large emigration rate from Upper Klamath Lake compensates for natural mortality; and panel (D) pertains to shortnose suckers and shows the approximately 10-d delay in peak abundance in Ewauna–Keno and the lower relative emigration rate from Upper Klamath Lake.

22 d and slowed emigration for ages 22–32 d; all of these age groups emigrated from the river quickly (Figure 4). For both species, age-specific survival for all age groups appeared greater at RB than in the adjacent WR site (Table 2), but this is, of course, confounded by emigration and immigration. Age-specific survival for ages older than 12 d tended to be higher in Ewauna–Keno than other areas (Table 2). This occurred because both species accumulated in Ewauna–Keno, immigration outpacing emigration in 27–32-d age-groups (Table 2). Shortnose suckers accumulated at younger ages as well, but Lost River suckers showed evidence of additional emigration at younger and older ages. The overall pattern for both species was similar to the accumulation pattern at RB, but at older ages. Shortnose suckers accumulated up to about age 28 d (Figure 5D), after which their Z approached the natural mortality rate. Lost River suckers were essentially at constant abundance in Ewauna–Keno from ages 17–42 d (Figure 5C), suggesting that immigration was matching natural mortality. Despite this, the Ga for Lost River suckers

suggested emigration from Ewauna–Keno by several age-groups (Table 2). Lost River suckers with early hatch dates, especially the last week of April and first week of May, were associated with greater transport to Ewauna–Keno, whereas those born the second week of June were less likely to be transported (Figure 6). Because our sampling continued through late July, the latter was not a sampling artifact. Shortnose suckers were more likely to be transported the second week of May and less likely in June. August Juvenile Sampling Sampled substrates in August were broadly distributed in the lake but small mixed and intermediate mixed substrates tended to have more southerly distributions (data not shown). There were no differences in standard deviations (Levene’s test for equality of variance) across substrates among the five species (P ¼ 0.82–0.86). The frequency of occurrence of species on substrates differed consistently with small substrates (fines and sand) having lower frequencies for all

340

MARKLE ET AL.

TABLE 2.—Life table data for Lost River and shortnose sucker larvae from Upper Klamath basin in 2004 (see Figure 1A for site locations). All variables pertain to age-a fish and are defined as follows: Ca ¼ catch; Sa ¼ survival (Caþ1/Ca); Za ¼ total mortality (loge[Sa]); Ma ¼ natural mortality (estimated from the combined values of Za for Upper Klamath Lake and Ewauna– Keno); and Ga ¼ migration (Za Ma). Site or area

C

M

WR

RB

SM UKL

E–K

C

M

WR

RB

SM

UKL

E–K

a

Insufficient data.

Age (d)

12 17 22 27 12 17 22 27 12 17 22 12 17 22 27

Ca

Sa

Lost River suckers 217.89 24.91 5,426.92 0.54 2,917.65 0.03 99.78 27.60 11.68 322.24 0.86 277.57 0.05 14 85.66 2.33 199.18 0.08 16.34 19.51 11.90 232.24 0.52 121.44 0.24 28.75

Za

Ma

Ga

þ3.22 0.62 3.38

þ1.19 þ0.08

1.81 3.46

þ2.46 0.15 2.99

þ1.19 þ0.08

1.34 3.07

þ0.84 2.50 þ2.48 0.65 1.44

þ1.19 þ0.08 þ1.19 þ0.08

þ0.84 3.69 þ2.48 1.84 1.52

a

12 17 22 27 32 37 17 22 27 32 37 42 12 17 22 27 12 17 22 27 12 17 22 27 32 12 17 22 27 32 37 17 22 27 32 37 12 17 22 27 32 37 17 22 27 32 37

20.99 46.17 27.21 30.48 10.01 10.80 22.84 47.73 25.03 35.39 35.31 26.60

2.20 0.59 1.12 0.33 1.08 0.78 2.09 0.52 1.41 1.00 0.75 0.16

Shortnose suckers 563.11 7.28 4,102.08 0.41 1,700.35 0.01 24.22 17.4 16.54 287.76 1.00 287.43 0.05 14 814.34 3.09 2,514.82 0.14 351.66 0.38 134.00 0.18 24.00 92.49 16.73 1,547.76 1.10 1,700.56 0.46 779.25 0.23 178.00 0.38 67.75 62.16 3.52 218.64 0.90 197.74 0.45 88.50 0.61 54.00 35.9 5.86 210.7 0.66 139.7 0.67 94.1 0.37 34.9 0.40 14.1 0.60 43.0 1.83 78.8 1.23 97.0 0.43 41.7 0.87 36.2 0.23

a

a

a

þ0.79 0.53 þ0.11 1.11 þ0.08 þ0.74 0.65 þ0.35 0.00 0.28 1.83

þ1.19 þ0.08 0.30 0.20 þ0.02 þ1.19 þ0.08 0.30 0.20 þ0.02 0.28

0.40 0.61 þ0.41 0.91 þ0.06 0.45 0.73 þ0.65 þ0.20 0.30 1.55

þ1.99 0.88 4.25

þ1.95 0.15 0.13

þ0.03 0.73 4.12

þ2.81 0.00 3.02

þ1.95 0.15 0.13

þ0.85 þ0.15 2.89

þ1.13 1.97 0.96 1.72

þ1.95 0.15 0.13 0.91 0.42 þ1.95 0.15 0.13 0.91 0.42 1.10 0.15 0.13 0.91 0.42 1.10 þ1.95 0.15 0.13 0.91 0.42 1.10 0.15 0.13 0.91 0.42 1.10

0.83 1.82 0.83 0.80

þ2.82 þ0.09 0.78 1.48 0.97 þ1.26 0.10 0.80 0.49 þ1.77 0.41 0.40 0.99 0.91 0.50 þ0.61 þ0.21 0.84 0.14 1.48

þ0.86 þ0.24 0.65 0.56 0.55 þ1.41 þ0.03 þ0.11 0.07 0.18 0.26 0.26 0.08 0.49 þ0.60 þ0.76 þ0.34 þ0.07 þ0.28 0.38


341

7B, C). In a two-factor ANOVA, the August latitude of the center of distribution was dependent on species (P , 0.0001) and year (P ¼ 0.003). Sample size was too small to include an interaction term, but residual plots suggested that interannual patterns were generally similar across species. August log shoreline abundance was weakly related to July 15 lake elevation (P ¼ 0.08) and strongly related to species (P ¼ 0.0003) and latitude of the center of distribution (P ¼ 0.0053). Discussion Sucker Larvae Advection

FIGURE 6.—Proportional differences in the numbers of Lost River suckers (LRS; N ¼ 322) and shortnose suckers (SNS; N ¼ 759) collected in Lake Ewauna–Keno Reservoir [E–K] and Upper Klamath Lake, by week of hatch (consecutive week of the year). For example, for fish that hatched during week 17, the proportion of Lost River suckers was 18% greater in E–K than in Upper Klamath Lake whereas the proportion of shortnose suckers was 2% greater in E–K.

species. The pooled abundance of the five species followed a similar pattern, mean abundance being less than 15 fish/sample on fines and sand and more than 39 fish/sample on larger substrates. Frequency of occurrence also differed by species, the most frequent (blue chub) occurring in 41.5% of all August cast-net samples from 1996 to 2006 (N ¼ 1,560) and the least frequent (Lost River suckers) occurring in only 9.9%. The latitude of the center of distribution in August was usually north (92.6%) and west (88.9%) of the center of abundance (N ¼ 54; no Lost River suckers collected in 2003), indicating a general downstream bias in abundance (Figure 7A). August abundance was strongly related to both centers. A slope of the regression of log shoreline abundance on latitude of the center of distribution was positive (adjusted r2 ¼ 45.5%, P , 0.0001), suggesting that a more northerly center was associated with higher abundance (Figure 8A). Among species, those with higher abundance had more northerly average distributions in August (Figure 8B) and less variance in the latitude of the center of distribution. The positive relationship between abundance pattern and latitude also held within species for Lost River suckers (adjusted r2 ¼ 34.4%, P ¼ 0.04), shortnose suckers (adjusted r2 ¼ 73.0%, P ¼ 0.0005), and tui chub (adjusted r2 ¼ 31.1%, P ¼ 0.05). These three species also had greater interannual variation in the center of distribution compared with the consistently abundant fathead minnow (examples in Figure

The age-class of peak abundance increased downstream from 17 d in the river to 22 d in Ewauna–Keno for Lost River suckers or 27 d for shortnose suckers. The abundance of 12-d-old Lost River suckers decreased from the eastern shore to Ewauna–Keno, where none were collected in Ewauna–Keno, further corroborating the species differences and transport direction. The average transport time from the river to the lake outlet could therefore be estimated at about 5 d for Lost River suckers and about 10 d for shortnose suckers. The hydrographic model also suggested that under strong or weak wind conditions, transport from eastern shore springs, where more Lost River suckers spawn, to the lake outlet would take about 5–6 d and from the Williamson River mouth about 8–17 d. Information below corroborates this faster average transport of Lost River sucker larvae than shortnose sucker larvae. There was also a downstream reduction in Z in both species. We interpreted this reduction as reflecting patterns of emigration and immigration. Helbig and Pepin (1998b) estimated that advective transport loss represented about 50% of overall larval mortality in their field data. Our data agree with this generalization but also show that immigration can be an equally important process in some locations. Taggart and Leggett (1987b), for example, noted that immigration explains 32–66% of apparent mortality in their data. Further, particular age-classes, such as 17-d-old Lost River sucker larvae in river sites C and M, may have substantially greater transport losses than other agegroups. Following Helbig and Pepin (1998a), we assumed that natural mortality was constant in space and time. We used the combined Upper Klamath Lake plus Ewauna–Keno OSU data as our best estimate of natural mortality over the approximate 3-week period between ages 15 and 39 d. If our assumption about natural mortality is reasonable, we would expect Z from the

342

MARKLE ET AL.

FIGURE 8.—Relationships between log-transformed shoreline abundance and latitude of the center of distribution for the five most abundant shoreline fish species in Upper Klamath Lake, 1996–2006. Panel (A) shows the regression line for all species (y ¼ 2627.58 þ 62.375x), panel (B) the means and 95% confidence intervals for each species individually (species abbreviations are given in Figure 7).

FIGURE 7.—Maps of the southern end of Upper Klamath Lake showing cast-net sampling results for 1996–2006. Panel (A) shows the average center of distribution (open symbols) and average center of abundance (closed symbols) for the five most abundant shoreline fish species—blue chub (BCH), fathead minnow (FHM), Lost River sucker (LRS), shortnose sucker (SNS), and tui chub (TCH). Panel (B) shows the interannual variation in the center of distribution for fathead minnow, and panel (C) the interannual variation in the center of distribution for shortnose suckers.

catch curve to approach zero and Ga from the life table to be positive for most age-classes in Ewauna–Keno because the loss from Upper Klamath Lake should tend to reduce apparent mortality. This prediction was correct for Z for both species and for Ga for all agegroups of shortnose suckers and two age-groups (27 and 32 d) of Lost River suckers. The alternative explanation, that larval sucker mortality was zero in Ewauna–Keno, is untenable. Our assumption also included no site-specific differences in mortality. Cooperman and Markle (2004) found larger and better-fed sucker larvae in emergent vegetation. However, Markle and Dunsmoor (2007) found average annual larval sucker survival from 10 to 15 mm ranged from 4.4% to 48.0% over an 11-year period when available vegetated habitat, as expressed by average June lake elevation, had been in a narrow range (1,262.49–1,262.84 m). They noted that fluctuations in abundance were probably due to a combination of larger-scale processes, such as climate and biotic interactions, affecting the whole system. We have seen little or no differences in the potential


predator–competitor field in the OSU long-term larval trawl sites in Upper Klamath Lake, Ewauna–Keno, and the TNC marshes from 2005 to 2006 (unpublished data). In fact, the Ewauna–Keno area had the lowest Z and usually had the highest densities of fathead minnow and exotic centrachids and the lowest availability of emergent vegetation. Marshes and emergent vegetation probably have counterbalancing effects, attracting species like fathead minnow (for spawning substrate) while providing cover for sucker and other larvae (Markle and Dunsmoor 2007). At the scale of this study, we believe that site-specific differences in mortality were minor relative to the flux of emigration and immigration. Partitioning the roles of predation, emigration, and immigration and testing site-specific mortality would be difficult (Taggart and Leggett 1987c; Pepin et al. 2002), but it would be necessary to confirm the validity of our assumption. Sample sites differed in their ability to retain larvae. Both catch-curve and life table analyses suggest that larval suckers were transported downstream and that restoration marshes, especially for shortnose suckers, greatly reduced transport compared with adjacent areas. Cooperman and Markle (2003) showed that sucker larvae in the Williamson River aggregated along the shoreline during daylight and released into adjacent currents at night. At its simplest, this system is a hydrological sluiceway, larvae produced upstream and future members of the population being composed of those individuals that are retained in Upper Klamath Lake. Emigration was highest in the natal Williamson and Sprague rivers, but a restored wetland (Riverbend) in the Williamson River delta accumulated 17–22-d-old shortnose sucker larvae and retained them up to 19 d longer than did the river. Even with high transit rates, some older juveniles can still be found in restored marshes (Crandall et al. 2008) and the lower river, though a proportion may be Klamath largescale suckers (Cooperman and Markle 2003). A restored wetland adjacent to Upper Klamath Lake (South Marsh) also retained shortnose larvae and had a 28-d emigration rate back into the lake of 42% compared with a lakewide 28-d 76% emigration rate out of the lake. In both marshes, there was less retention of Lost River sucker larvae. In 2004, Lost River suckers dispersed more quickly and over a greater age range (17–42 d) than shortnose suckers (17–27 d). Their retention in the lake appeared to be primarily a function of distance from the spawning site to the lake outlet because a

343

greater proportion of Lost River sucker spawn at shoreline springs. They may also be more dependent on gyre retention, which is not well described by the shoreline sampling. Retention of shortnose suckers appeared to be a function of both distance and access to shoreline retention areas such as marshes. At the downstream catchment, Ewauna–Keno, immigration was large enough to offset daily mortality. An important, unexplored aspect of this system is the effect of larval behavior on advection and retention. Advection of a passive particle, where transport is directly related to water velocity, is a frequent starting point in describing larval dispersal (Taggart and Leggett 1987a), but it is almost certainly an incorrect generalization, even for poorly developed larvae (Leis et al. 2007). Ontogenetic changes in swimming speed (Fisher et al. 2005), behavioral choices to swim or not swim (Hogan and Mora 2005), nonrandom choices in swimming direction or depth (Leis et al. 2006, 2007), ontogenetic and diel changes in geotaxis (Bradbury et al. 2006), and habitat selection through olfaction (Gerlach et al. 2007) can all interact to limit advection. Despite extensive pelagic periods, ontogenetic changes in behavior can limit transport and vagrancy (Bradbury et al. 2006), even allowing panda clownfish Amphiprion polymnos to settle within 100 m of their natal site after a 9–12 d pelagic larval phase (Jones et al. 2005). Species differences in volitional movement and behavior, even in swim-up larvae, can lead to different outcomes under the same or similar hydrographic regimes. Klamath largescale suckers, which also spawn in the Williamson River but whose juveniles are seldom found in Upper Klamath Lake, probably have early behaviors that limit advection. For example, Kennedy and Vinyard (1997) found that larval Warner suckers Catostomus warnerensis have behaviors that limit advection and keep them in streams. For a species that spawns in tributaries or lakeshore springs and rears in a lake, such as the two suckers, some individual variability in timing and duration of drift and timing of retention behaviors would seem advantageous given different distances from spawning grounds and interannual variability in river discharge, wind speed and direction, and lake elevation. Thus, although there is reasonable agreement between our advection model and our empirical data, it should only be considered a first-order approximation. August Juvenile Patterns We tested whether differences in larval advection might explain interannual differences in the shoreline

344

MARKLE ET AL.

abundance of juveniles in August for the two sucker and three other abundant shoreline species. The latitude of the center of distribution was almost always north of the latitude of the center of abundance, suggesting that abundance was usually skewed towards the south, a pattern consistent with downstream advection during the early life history. August abundance was also related to the center of distribution. Among the five species and 11 years of data, the latitude of the center of distribution explained 45.5% of the variation in log shoreline abundance, increased abundance being associated with more northerly distributions. The most abundant species in the lake, fathead minnow and blue chub, were consistently the most northerly and least variable in centers, whereas the two sucker species had the most southerly and variable centers. Within species with more variable centers (both suckers and tui chub), interannual differences in abundance were strongly and positively related to latitude of the center of distribution, accounting for 73% of the variation in shortnose sucker abundance and 31–34% in the others. Interannual differences in log shoreline abundance in August were not related to water quality variables or lake elevation, whereas interannual and species differences were strong. There was no tendency for preferred substrates to have a northerly distribution within the lake, and thus the patterns are most likely due to advection and not strictly to habitat. Recruitment The combination of factors responsible for advection dynamics (hatch date, weather, and behavior) seems to determine the spatial distribution of a year-class within Upper Klamath Lake and to be a strong predictor of year-class abundance in August. Among the five abundant shoreline species in the lake, August abundance was positively related to a northern center of distribution, and centers were highly dependent on species and year. Species differences in timing of spawning and behavior probably control details of retention. For the suckers, M. S. Cooperman and D. F. Markle (unpublished data) found that survival from 10 to 15 mm (29 larval sucker cohorts from 1995 to 2001) was dependent on lake elevation (habitat volume) and a low proportion of high-wind events. Larvae of early spawning species, like Lost River suckers, will experience stronger winds and will be more likely to be retained by the gyre, but they will be more quickly advected by the eastern boundary current. Lost River sucker retention may be weighted more to hydrographic than shoreline retention, a type of retention that

would presumably not include the predation refuge advantage of emergent vegetation (Markle and Dunsmoor 2007). Further work is needed on the interrelationship of open lake and shoreline nursery areas. Larvae of later-spawning species will experience weaker winds and be more dependent on shoreline retention. In support, we found sucker larvae hatched in the first 2–3 weeks of May made up a greater proportion of fish in Ewauna–Keno than those hatched in the first 2–3 weeks of June. The larval sucker retention data in 2004 showed shortnose suckers more dependent on shoreline retention than Lost River suckers. Shoreline retention seems to be mostly a function of shoreline irregularity and the temporary trap created when a current passes (i.e., ‘‘hydrological roughness’’; Okubo 1973). This may contribute to relatively long-term (days to weeks) retention in shoreline marshes and bays and at least short-term reduction in transport to the south in the case of other irregularities, such as promontories. Among the five species, the interannual effect of reduced southern transport was most obvious in shortnose sucker juveniles, 73% of the log-transformed variation in shoreline abundance in August being attributable to the latitude of the center of abundance. The potential benefits of retaining recruits through the association of the hydrodynamic regime with diverse shoreline structures are at direct risk from marsh, shoreline, and channel alterations. The diking of marshes in the lake and along the lower Williamson River and its delta has presumably reduced this roughness and decreased retention compared with historical patterns. Restoration efforts in progress to restore marshes and a complex delta should improve larval retention. The three minnow species face different seasonal conditions because they spawn later than suckers (unpublished data). Because later-spawned fish experience lower average lake elevations, we might expect them to encounter less retention habitat, but these species had the highest abundance and most northward retention. Larval cyprinids tend to drift less each night as they grow (Manteifel et al. 1978). For larval roach Rutilus rutilus the light intensity threshold for rheotaxis decreased until the juvenile stage, when the behavior ceased (Pavlov 1994). If resistance to vagrancy is based on development, the seasonal increase in temperature, reduced wind, and more established vegetation could reduce the duration and vulnerability to vagrancy, mediate reduced total retention habitat, and lead to more northward retention and greater abundance.


This study is the first application of the member– vagrant hypothesis in larval freshwater fishes adapted to a lake and one of the few to try to quantify a larval retention mechanism. It suggests that the retention mechanisms for freshwater fish rely more on the interaction of hydrography and shoreline topography and habitat than those for marine fish larvae. The parallels we see with marine and estuarine systems are the interactions of ontogeny and behavior with hydrography to influence advection trajectories (Leis et al. 2006; Bradbury et al. 2006). We also show that, regardless of larval behavior and advection dynamics, the abundance of juveniles of five species in August is related to a northerly center of distribution away from the lake outlet. Acknowledgments

Financial support was provided by the U.S. Bureau of Reclamation (Klamath Project), the U.S. Fish and Wildlife Service (USFWS; cooperative agreement 11450-I-J514 with the Nature Conservancy), and the Oregon State University (OSU) Agricultural Experiment Station. Specimen collections were authorized under USFWS scientific taking permits TE 006333-10 (OSU), TE 007907-4 (U.S. Geological Survey), TE 041204-0 (the Nature Conservancy), and OSU institutional animal care and use permit LAR-2162. Field assistance was provided by J. Van Noord, K. Crowley, A. Hill, and C. Stevens; L. Dunsmoor, H. Li, J. Power, and R. Shively provided helpful comments on an early draft. References Boyle, J. C. 1976. Fifty years on the Klamath. Klocker Printery, Medford, Oregon. Bradbury, I. R., K. Gardiner, P. V. R. Snelgrove, S. E. Campana, P. Bentzen, and L. Guan. 2006. Larval transport, vertical distribution, and localized recruitment in anadromous rainbow smelt (Osmerus mordax). Canadian Journal of Fisheries and Aquatic Sciences 63:2822–2836. Brown, A. V., and M. L. Armstrong. 1985. Propensity to drift downstream among various species of fish. Journal of Freshwater Ecology 3:3–17. Casulli, V., and P. Zanolli. 2002. Semi-implicit numerical modeling of nonhydrostatic free-surface flows for environmental problems. Mathematical and Computer Modelling 36:1131–1149. Cooperman, M. S., and D. F. Markle. 2003. Rapid outmigration of Lost River and shortnose sucker larvae from in-river spawning beds to in-lake rearing grounds. Transactions of the American Fisheries Society 132:1138–1153. Cooperman, M. S., and D. F. Markle. 2004. Abundance, size, and feeding success of larval shortnose suckers and Lost

345

River suckers from different habitats of the littoral zone of Upper Klamath Lake. Environmental Biology of Fishes 71:365–377. Crandall, J. D., L. B. Bach, N. Rudd, M. Stern, and M. Barry. 2008. Response of larval Lost River and shortnose suckers to wetland restoration at the Williamson River delta, Oregon. Transactions of the American Fisheries Society 137:402–416. Fisher, R., J. M. Leis, D. L. Clark, and S. K. Wilson. 2005. Critical swimming speeds of late-stage coral reef fish larvae: variation within species, among species, and between locations. Marine Biology 147:1201–1212. Fortier, L., and J. A. Gagne´. 1990. Larval herring (Clupea harengus) dispersion, growth, and survival in the St. Lawrence estuary: match/mismatch or membership/ vagrancy? Canadian Journal of Fisheries and Aquatic Sciences 47:1898–1912. Gerlach, G., J. Atema, M. J. Kingsford, K. P. Black, and V. Miller-Sims. 2007. Smelling home can prevent dispersal of reef fish larvae. Proceedings of the National Academy of Sciences of the USA 104:858–863. Hare, J. A., J. A. Quinlan, F. E. Werner, B. O. Blanton, J. J. Govoni, R. B. Forward, L. R. Settle, and D. E. Hoss. 1999. Larval transport during winter in the SABRE study area: results of a coupled vertical larval behaviour–threedimensional circulation model. Fisheries Oceanography 8(Supplement 2):57–76. Hare, J. A., J. H. Churchill, R. K. Cowen, T. Berger, P. Cornillon, P. Dragos, S. Glenn, J. J. Govoni, and T. N. Lee. 2002. Routes and rates of larvae fish transport from the southwestern to the mid-Atlantic North American continental shelf. Limnology and Oceanography 47:1774–1789. Helbig, J. A., and P. Pepin. 1998a. Partitioning the influence of physical processes on the estimation of ichthyoplankton mortality rates, I. Theory. Canadian Journal of Fisheries and Aquatic Sciences 55:2189–2205. Helbig, J. A., and P. Pepin. 1998b. Partitioning the influence of physical processes on the estimation of ichthyoplankton mortality rates, II. Application to simulated and field data. Canadian Journal of Fisheries and Aquatic Sciences 55:2206–2220. Hoff, G. R., D. J. Logan, and D. F. Markle. 1997. Otolith morphology and increment variation in young Lost River and shortnose suckers. Transactions of the American Fisheries Society 126:488–494. Hogan, J. D., and C. Mora. 2005. Experimental analysis of the contribution of swimming and drifting to the displacement of reef fish larvae. Marine Biology 147:1213–1220. Iles, T. D., and M. Sinclair. 1982. Atlantic herring: stock discreteness and abundance. Science 215:627–633. Jones, G. P., S. Planes, and S. R. Thorrold. 2005. Coral reef fish larvae settle close to home. Current Biology 15:1314–1318. Kennedy, T. B., and G. L. Vinyard. 1997. Drift ecology of western catostomid larvae with emphasis on Warner suckers, Catostomus warnerensis (Teleostei). Environmental Biology of Fishes 49:187–195. LaBolle, L. D., Jr., H. W. Li, and B. C. Mundy. 1985. Comparison of two samplers for quantitatively collecting larval fishes in upper littoral habitats. Journal of Fish Biology 26:139–146.

346

MARKLE ET AL.

Leis, J. M., A. C. Hay, and T. Trnski. 2006. In situ ontogeny of behaviour in pelagic larvae of three temperate, marine, demersal fishes. Marine Biology 148:655–669. Leis, J. M., K. J. Wright, and R. N. Johnson. 2007. Behaviour that influences dispersal and connectivity in the small, young larvae of a reef fish. Marine Biology 153:103– 117. Manteifel, B. P., I. I. Girsa, and D. S. Pavlov. 1978. On rhythms of fish behaviour. Pages 215–224 in J. E. Thorpe, editor. Rhythmic activity of fishes. Academic Press, New York. Markle, D. F., and M. Cooperman. 2002. Relationships between Lost River and shortnose sucker biology and management of Upper Klamath Lake. Pages 93–117 in W. S. Braunworth, Jr., T. Welch, and R. Hathaway, editors. Water allocation in the Klamath Reclamation Project, 2001: an assessment of natural resource, economic, social, and institutional issues in the Upper Klamath basin. Oregon State University Extension Service, Corvallis. Available http://extension.oregonstate. edu/catalog/html/. (December 2007). Markle, D. F., M. R. Cavalluzzi, and D. C. Simon. 2005. Morphology and taxonomy of Klamath basin suckers (Catostomidae). Western North American Naturalist 65:473–489. Markle, D. F., and K. Clauson. 2006. Ontogenetic and spatial changes in diet of late larval and juvenile suckers (Catostomidae) in Upper Klamath Lake, Oregon. Western North American Naturalist 66:492–501. Markle, D. F., and L. K. Dunsmoor. 2007. Effects of habitat volume and fathead minnow introduction on larval survival of two endangered sucker species in Upper Klamath Lake, Oregon. Transactions of the American Fisheries Society 136:567–579. Okubo, A. 1973. Effect of shoreline irregularities on streamwise dispersion in estuaries and other embayments. Netherlands Journal of Sea Research 6:213–224. Orr, E. L., W. N. Orr, and E. M. Baldwin. 1992. Geology of Oregon. Kendall/Hunt, Dubuque, Iowa. Parrish, R. H., C. S. Nelson, and A. Bakun. 1981. Transport mechanisms and reproductive success of fishes in the California Current. Biological Oceanography 1:175–203. Pavlov, D. S. 1994. The downstream migration of young fishes in rivers: mechanisms and distribution. Folia Zoologica 43:193–208. Pepin, P., J. A. Helbig, R. Laprise, E. Colbourne, and T. H. Shears. 1995. Variations in the contribution of transport to changes in planktonic animal abundance: a study of the flux of fish larvae in Conception Bay, Newfoundland. Canadian Journal of Fisheries and Aquatic Sciences 52:1475–1486. Pepin, P., J. F. Dower, J. A. Helbig, and W. C. Leggett. 2002. Estimating the relative roles of dispersion and predation in generating regional differences in mortality rates of larval radiated shanny (Ulvaria subbifurcata). Canadian Journal of Fisheries and Aquatic Sciences 59: 105–114. Quinn, T. J., III, and R. B. Deriso. 1999. Quantitative fish dynamics. Oxford University Press, New York. Reithel, S. A. 2006. Patterns of retention and vagrancy in larval Lost River and shortnose suckers from Upper

Klamath Lake, Oregon. Master’s thesis. Oregon State University, Corvallis. Rozas, L. P., and T. J. Minello. 1997. Estimating densities of small fishes and decapod crustaceans in shallow estuarine habitats: a review of sampling design with focus on gear selection. Estuaries 20:199–213. Scoppettone, G. G., and G. Vinyard. 1991. Life history and management of four endangered lacustrine suckers. Pages 359–377 in W. L. Minckley and J. E. Deacon, editors. Battle against extinction: native fish management in the American West. University of Arizona Press, Tucson. Seber, G. A. F. 1982. The estimation of animal abundance. Macmillan, New York. Smith, P. E. 1972. The increase in spawning biomass of northern anchovy, Engraulis mordax. U.S. National Marine Fisheries Service Fishery Bulletin 70:849–874. Smith, P. E., and S. L. Richardson. 1977. Standard techniques for pelagic fish egg and larval surveys. FAO (Food and Agriculture Organization of the United Nations) Fisheries Technical Paper 175:1–100. Sinclair, M. 1988. Marine populations: an essay on population regulation and speciation. University of Washington, Seattle. Tabachnick, B. G., and L. S. Fidell. 2001. Using multivariate statistics. Harper Collins, New York. Taggart, C. T., and W. C. Leggett. 1987a. Wind-forced hydrodynamics and their interaction with larval fish and plankton abundance: a time series analysis of physical– biological data. Canadian Journal of Fisheries and Aquatic Sciences 44:438–451. Taggart, C. T., and W. C. Leggett. 1987b. Short-term mortality in postemergent larval capelin Mallotus villosus, I. Analysis of multiple in situ estimates. Marine Ecology Progress Series 41:205–217. Taggart, C. T., and W. C. Leggett. 1987c. Short-term mortality in postemergent larval capelin Mallotus villosus, II. Importance of food and predator density and density dependence. Marine Ecology Progress Series 41:219–229. Taradina, T. G., D. S. Pavlov, and A. I. Lupandin. 1997. The connection of vertical distribution of fish juveniles in downstream migrations with their floatability and flow turbulence. Journal of Ichthyology 37:523–528. Terwilliger, M., D. F. Markle, and J. Kann. 2003. Associations between water quality and daily growth of juvenile shortnose and Lost River suckers in Upper Klamath Lake, Oregon. Transactions of the American Fisheries Society 132:691–708. USFWS (U.S. Fish and Wildlife Service). 2001. Biological conference regarding the effects of operation of the Bureau of Reclamation’s Klamath Project on the endangered Lost River sucker (Deltistes luxatus), endangered shortnose sucker (Chasmistes brevirostris), threatened bald eagle (Haliaeetus leucocephalus), and proposed critical habitat for the Lost River/shortnose suckers, Klamath Falls, Oregon. Available: http://www. usbr.gov/. (December 2007). Wood, T. M., and R. T. Cheng. 2006. Use of UnTRIM to investigate dissolved oxygen transport in Upper Klamath Lake, Oregon. Pages 1–12 in M. Piaseck, editor. Proceedings of the 7th international conference on


hydroscience and engineering. College of Engineering, Drexel University, Philadelphia. Available: hdl.handle. net/1860/1424. (February 2009). Wood, T. M., G. Hoilman, and M. Lindenberg. 2006. Water quality conditions in Upper Klamath Lake, Oregon, 2002–2004. U.S. Geological Survey Scientific Investigations Report 2006–5209.

347

Wood, T. M., R. T. Cheng, J. W. Gartner, G. R. Hoilman, M. K. Lindenberg, and R. E. Wellman. 2008. Modeling hydrodynamics and heat transport in Upper Klamath Lake, Oregon, and implications for water quality. U.S. Geological Survey Scientific Investigations Report 20085076. Available http://pubs.usgs.gov/sir/2008/5076/. (June 2008).