Relationship between stream temperature, thermal refugia and rainbow trout Oncorhynchus mykiss abundance in arid-land streams in the northwestern United States

Ecology of Freshwater Fish 2001: 10: 1–10 Copyright C Munksgaard 2001 Printed in Denmark ¡ All rights reserved ISSN 0906-6691 Relationship between stream temperature, thermal refugia and rainbow trout Oncorhynchus mykiss abundance in arid-land streams in the northwestern United States Ebersole JL, Liss WJ, Frissell CA. Relationship between stream J. L. Ebersole1, W. J. Liss1, temperature, thermal refugia and rainbow trout Oncorhynchus mykiss C. A. Frissell2 abundance in arid-land streams in the northwestern United States. 1Department of Fisheries and Wildlife, Oregon Ecology of Freshwater Fish 2001: 10: 1–10. C Munksgaard, 2001 State University, Corvallis, 2Flathead Lake Biological Station, University of Montana, Abstract – Warm stream temperatures may effectively limit the distribution Polson, USA and abundance of Pacific salmon Oncorhynchus spp. in streams. The role of cold thermal refugia created by upwelling groundwater in mediating this effect has been hypothesized but not quantitatively described. Between June 21 and September 15, 1994, rainbow trout O. mykiss abundance within 12 northeast Oregon (USA) stream reaches was inversely correlated with mean ambient maximum stream temperatures (rΩª0.7, P∞0.05). Some rainbow trout used thermal refugia (1–10 m2 surface area) that were on aver- age 3–8æC colder than ambient stream temperatures. Within the warmest reaches, high ambient stream temperatures (±22æC) persisted from mid- June through August, and on average 10–40% of rainbow trout were ob- served within thermal refugia during periods of midday maximum stream Key words: stream temperature; refugia; rainbow temperatures. Frequency of cold-water patches within reaches was not sig- trout nificantly associated with rainbow trout density after accounting for the J. L. Ebersole, Department of Fisheries and influence of ambient stream temperature (PΩ0.06; extra sum of squares F- Wildlife, Nash 104, Oregon State University, test). Given prolonged high ambient stream temperatures in some reaches, Corvallis OR, 97331-3803 USA; e-mail: ebersolj/ucs.orst.edu,the thermal refugia available in the streams we examined may be too small William.Liss/orst.edu,and too infrequent to sustain high densities of rainbow trout. However, frissell/selway.umt.eduthese refugia could allow some rainbow trout to persist, although at low Accepted for publication November 25, 1999densities, in warm stream reaches. Un resumen en espan˜ol se incluye detra´s del texto principal de este artı´culo. Introduction Thermal conditions in temperate arid land streams can be extreme, varying from near 0æC in winter to over 29æC in midsummer. Periods of high stream temperatures during midsummer can elevate meta- bolic demands and induce mortality in fish, par- ticularly if these temperatures persist for several days (Brett 1979). High water temperatures can ef- fectively limit the longitudinal distribution of fish within streams (Meisner 1990), restrict seasonal migration patterns (Berman & Quinn 1991), and fragment populations within a watershed by isolat- ing suitable thermal habitats (Matthews & Zimm- erman 1990). 1 In the lower Snake River basin in the United States, streams and rivers typically warm in a downstream direction, and midsummer water tem- peratures in lower-elevation rivers and streams fre- quently exceed upper incipient lethal temperatures for salmonids (22–26æC; Brett 1952). As a result, the primary spawning and rearing habitats for salmonids Onchorhynchus and Salvelinus spp. in the interior Pacific Northwest are presently limited to headwater portions of most riverine networks where water temperatures seldom exceed 17æC. However, in recent investigations we have observed juvenile rainbow trout O. mykiss in northeast Ore- gon streams where maximum temperatures ex- ceeded 26æC for several hours a day over periods Ebersole et al. of more than 7 days (Frissell et al. 1996). Salmonids in these warm stream reaches may per- sist by using thermal refugia to reduce their ex- posure during these periods (e.g., Gibson 1966; Kaya et al. 1977; Berman & Quinn 1991; Li et al. 1994). Although the use of thermal refugia such as coldwater patches created by tributaries, ground- water seeps, springs, and thermal stratification within stream channels has been described for salmonids from various regions of the western United States (e.g., Li et al. 1993; Nielsen et al. 1994; Matthews & Berg 1997), little is known re- garding the influence of these coldwater patches on the abundance and distribution of fish at the reach or watershed scale (stream lengths ±100 m). Given the possibility of stream warming with human land use and climate change, thermal refugia provided by groundwater-fed habitats may be increasingly important for persistence of coldwater fishes (Me- isner 1990; Nielsen et al. 1994). While previous studies have examined use of thermal refugia within sections of warm streams, we are aware of no systematic attempts to examine thermal refuge use and salmonid relative abundance across longi- tudinal gradients of elevation, stream size, ambient stream temperature and other environmental con- ditions within a watershed. Knowledge of the con- tribution of thermal refugia to the persistence of coldwater fishes could contribute to watershed management and restoration strategies in regions Fig. 1. Map of the study areas. The numbered locations refer to study reaches described in Table 1. 2 where warm summer stream temperatures are a concern. Protection and restoration of coldwater habitats is particularly urgent in regions such as the interior Columbia River basin, where several salmonid stocks have been recently listed or pro- posed for listing under the federal Endangered Species Act. Our objective in this study was to determine how water temperature patterns at two spatial scales may be related to the distribution and abundance of rainbow trout in four northeast Oregon streams. Specifically, we posed two questions: How are rain- bow trout densities related to ambient summer water temperatures and thermal refuge occurrence across reaches in four watersheds? Secondly, what are the seasonal and diurnal patterns of thermal refuge use by rainbow trout? Methods Twelve study reaches were sampled from June 21 to September 15, 1994. Study reaches were located in Joseph Creek, Cottonwood Creek, Crooked Creek, and North Pine Creek (Fig. 1). These streams lie within the lower Grande Ronde River and Pine Creek basins of northeast Oregon. The region is underlain by the Grande Ronde flood ba- salts into which streams have incised steep-walled valleys and canyons. Mixed-conifer forests and bunchgrass steppe are the primary vegetation types, with deciduous trees and shrubs present in riparian areas. Land uses are primarily livestock grazing and timber harvesting. Along each of the four streams, three study reaches were established consisting of a down- stream location near the stream mouth, a middle location, and an upstream location. These reaches were positioned to encompass an array of stream habitat conditions from warmer, lower elevation reaches to colder, higher elevation reaches within each of the four study streams. Reaches were located in alluvial valleys or alluviated canyons where thermal heterogeneity of surface waters created by upwelling of groundwater or hyporheic flow was deemed likely to occur, based upon pre- viously observed associations between alluviated stream channels and groundwater inflow (Frissell et al. 1996). Within each study reach a 100- to 500-m length of stream was inventoried using a modified stream survey method (Hankin & Reeves 1988). Reach length varied according to stream size and was ad- justed to include at least three riffle-pool se- quences. In each reach, we measured mean width, mean length, and maximum depth of individual pools, riffles, glides and cascades (Bisson et al. 1982). Channel gradient was measured with a cli- Stream temperatures, refugia and rainbow trout nometer. Mean reach elevation and mean distance to the Snake River (a relative measure of within- basin location) were determined from USGS topo- graphic maps. Stream discharge was estimated using a pygmy current meter at established cross sections (Corbett 1955). We assessed availability of coldwater patches within a reach by conducting intensive thermal surveys. Temperature was measured in detail throughout each survey reach using a digital thermometer (Atkins model 39658-K, accuracy∫0.1æC) with a sounding probe taped to a 3.5-m telescoping fiberglass rod. These surveys were conducted during periods of base flow (0.3– 0.6 m3/s) when ambient stream water temperatures exceeded 16æC, allowing detection of inflowing groundwater averaging 12æC (Frissell et al. 1996). This allowed for a minimum detectable difference in temperature of at least 3æC from the ambient streamflow (Ozaki 1988). Coldwater patches were defined as relatively small (∞1 channel width maxi- mum length) and relatively cold (at least 3æC col- der than ambient at time of detection) features within stream channels. Thermal refugia, in con- trast, are those coldwater patches actually used by fish during periods of presumed heat stress. Ther- mal surveys consisted of sweeping the ther- mometer probe across the stream channel bed while wading upstream. A minimum sweep fre- quency of one sweep per meter of longitudinal stream channel was maintained to maximize the probability of detection of small coldwater patches. When temperatures more than 3æC cooler than the ambient stream temperature were encoun- tered, intensive probing was used to delineate the spatial bounds of the coldwater patch. Identified coldwater patches were measured to determine sur- face area (mean width¿mean length) and mean depth. More detailed discussion of coldwater patch characteristics and landscape associations for the study region is provided in Frissell et al. (1996). For each coldwater patch, the percentage of sur- face area influenced by cover elements (large wood, overhanging banks, overhanging vegetation, inter- stitial space, surface turbulence, water depth) for juvenile salmonids was visually estimated. Sub- mersible temperature data recorders measured and recorded temperatures at thirty minute intervals for the duration of the study period within 12 of the coldwater patches (Optic StowAway, Onset Computer; accuracy∫0.2æC). Temperature data re- corders were also placed within well-mixed por- tions of the main channel within each survey reach. Each surveyed reach was visited four times dur- ing summer 1994 with one exception: reaches at middle and lower Crooked Creek were surveyed only three times due to extremely turbid water con- 3 ditions in July following a large landslide in the upper basin. During each visit, a team of snorke- lers (2–3) moved slowly upstream through the reach and all fish were identified to species and classified into 5-cm length categories. Reliance on snorkel observations minimized possible disturb- ance or injury to endangered Snake River spring- summer chinook salmon potentially present in the study reaches. Fish species and lengths were re- corded for each pool, riffle, glide, cascade or cold- water patch. Snorkel surveys were timed to include periods of daily maximum stream temperatures (generally 1400–1800 hours) to maximize the likeli- hood of observing coldwater refuge use. Ad- ditional mid-summer snorkel surveys were con- ducted in eight coldwater patches in the lower North Pine and upper, middle and lower Joseph reaches to assess diurnal shifts in thermal refuge use. During these surveys, divers determined fish abundance in each coldwater patch and the sur- rounding sampling reach in the lower North Pine and Joseph Creek reaches during portions of the day that were relatively cool (early morning or late evening) and relatively warm (mid-afternoon). At one location, we were able to establish a blind on a bridge above a thermal refuge that allowed direct observation of undisturbed rainbow trout behavior and movements. Photographs taken from this location were subsequently digitized to illustrate locations of fish in relation to the mapped thermal gradient within the coldwater patch. We report observations of rainbow trout only because they were the most widely distributed spe- cies that consistently used thermal refugia. Data for other fish species were provided in Frissell et al. (l996). Rainbow trout abundance was standard- ized as reach density (number of rainbow trout ob- served per square meter of stream surface area) for each reach. Refuge use by rainbow trout was stan- dardized as refuge density (number of fish in a ref- uge/area of refuge) and proportional refuge use (number of individuals in a refuge/total number of individuals in surveyed reach). We tested for significant trends in rainbow trout density across sampling dates for each reach using repeated measures ANOVA (Type III sum of squares; generalized linear model procedure; SAS Institute 1989). We used Pearson’s correlation analysis to identify relationships between physical characteristics of reaches and mean rainbow trout densities (nΩ12). Reach characteristics included in correlation analysis are listed in Table 1. Associ- ations between coldwater patch mean depth, mean ambient-cold patch temperature difference, total cover and rainbow trout densities within coldwater patches were examined similarly (nΩ22). Because coldwater patch surface area and coldwater patch Ebersole et al. Table 1. Physical characteristics and rainbow trout densities in study reaches. Map code locations are shown in Fig. 1. Mean daily maximum Mean Distance Mean Relative area Rainbow ambient summer Channel to Mean maximum Coldwater of coldwater trout Map temperature discharge gradient Snake R. elevation pool patches patches density Reaches code (æC) (m3/s) (%) (km) (m) depth (m) per 100 m (%) (number per m2) Lower Joseph Cr. 1 25.06 0.43 1 14 354 0.99 0.35 0.22 0.01 Middle Joseph Cr. 2 23.94 0.41 1.5 51 719 1.13 0.98 0.11 0.01 Upper Joseph Cr. 3 23.75 0.33 1 72 902 1.09 1.46 3.80 0.07 Lower N Pine Cr. 4 22.40 0.36 2 14 732 0.76 1.10 3.06 0.11 Middle N Pine Cr. 5 20.79 0.35 2 21 829 0.79 0.77 0.11 0.13 Upper N Pine Cr. 6 19.20 0.08 2 29 1012 0.67 2.25 3.74 0.19 Lower Cottonwood Cr. 7 20.19 0.36 2.5 14 354 0.45 0 0 0.13 Middle Cottonwood Cr. 8 20.27 0.35 2.5 18 427 0.52 0 0 0.51 Broady Cr. 9 15.68 0.06 2.5 26 927 0.43 1.70 6.61 0.24 Lower Crooked Cr. 10 20.32 0.66 2 85 597 0.99 0.48 1.60 0.35 Middle Crooked Cr. 11 18.36 0.59 2 93 628 0.69 0.6S 0.22 0.36 First Cr. 12 16.06 0.24 2.5 98 914 0.54 0 0 0.50 rainbow trout density are by definition associated, we also examined relationships between pro- portional refuge use (number of individuals in a refuge/total number of individuals in surveyed reach) and physical characteristics of coldwater patches listed in Table 2, including coldwater patch surface area. Differences in physical characteristics (Table 2) between coldwater patches in which fish were observed at least once and coldwater patches in which fish were never observed were compared using two-sample t-tests. Multiple linear regression Table 2. Physical characteristics and rainbow trout use of coldwater patches Surface Mean Mean ambient-cold Mean O. mykiss Mean O. mykiss area depth patch temp. differencea proportional density Total cover Code (m2) (m) (æC) refuge use (number per m2) (% area) LJ1 48.41 0.08 0.20 0.124 0.01 15 LJ2 4.38 0.15 1.43 0.167 0.08 2 LJ3 14.00 0.16 0.97 0.000 0.00 17 LJ4 1.00 0.07 1.47 0.000 0.00 13 MJI 1.43 0.14 5.15 0.133 3.61 6 MJ2 0.50 0.09 5.80 0.030 2.00 6 MJ3 0.84 0.10 3.00 0.000 0.00 10 UJI 6.67 0.50 3.93 0.202 3.72 30 UJ2 4.00 0.20 2.40 0.034 1.50 25 UJ3 16.69 0.19 6.78 0.054 2.64 35 UJ4 2.00 0.37 2.75 0.126 17.50 6 UJ5 25.03 0.74 1.80 0.141 4.31 40 UJ6 38.12 0.41 3.47 0.113 4.16 45 LPI 2.00 0.39 3.30 0.246 6.50 56 LP2 11.02 0.08 4.30 0.025 3.32 5 LP3 32.13 0.26 1.97 0.019 4.33 27 MPI 1.20 0.08 3.50 0.000 0.00 15 UPI 15.56 0.30 5.33 0.084 1.03 45 UP2 1.50 0.18 4.23 0.018 0.72 81 BRI 2.61 0.10 8.00 0.000 0.00 46 LC1 21.25 0.25 4.73 0.002 0.03 30 MCI 1.76 0.09 3.00 0.011 2.99 35 a Mean difference in temperature between ambient streamflow and coldest location within coldwater patch for all sampling visits. 4 (SAS Institute 1989) was used to examine additive effects of ambient stream temperature and cold- water patch availability on rainbow trout density between the 12 reaches. Results Thermal characteristics of reaches Mean daily maximum ambient stream tempera- tures generally increased downstream between reaches (Table 1), except for lower Cottonwood Stream temperatures, refugia and rainbow trout Creek, which was cooled by a tributary entering upstream. Ambient stream temperatures peaked in late July or early August for all reaches (Fig. 2). We identified 22 coldwater patches accessible to fish in nine of the twelve study reaches examined (Table 2). Coldwater habitats deemed inaccessible to fish (isolated pools or seeps) were not included in this analysis. Coldwater patches identified in- cluded lateral seeps, cold side-channels, floodplain tail seeps, floodplain seeps and stratified pools. Descriptions of coldwater patch types can be found in Frissell et al. (1996). We did not attempt to distinguish differences in physical characteristics or rainbow trout use between these functional cate- gories of coldwater patch because of low sample sizes and high variability within coldwater patch classes. Coldwater patches were 3–10æC colder than the ambient streamflow at time of detection. For the coldwater patches monitored with recording ther- mometers, mean daily temperatures were generally 1–8æC colder than mean ambient temperatures (Table 2). Two seasonal temperature patterns were apparent within the 12 coldwater patches moni- tored throughout the summer. Three coldwater patches maintained relatively cold, stable tempera- tures (rangeΩ12–17æC) that slowly increased through the summer and into the fall (Fig. 2A, B). The remaining coldwater patches displayed daily maximum temperature fluctuations that corre- sponded closely with daily and seasonal fluctu- ations in ambient streamflow temperature. Within coldwater patches temperatures were seldom spa- tially uniform, and coldwater patches in back- waters or alcoves were observed to occasionally stratify. Coldest temperatures occurred along the stream bottom and near the source of upwelling groundwater. Surface waters and shallow (∞5 cm deep) margins of several coldwater patches ex- posed to solar radiation heated rapidly during warm afternoons, and temperatures in these areas often exceeded ambient stream temperatures. Distribution and abundance of rainbow trout Rainbow trout density did not differ through time among the twelve reaches (repeated measures ANOVA, model effects for time and (time¿loc- ation) effects not significant; P±0.05). Density ob- servations were subsequently averaged over the four sampling periods for comparisons among reaches. Of the suite of reach characteristics that were measured (Table 1), only mean daily maxi- mum ambient stream temperature was significantly correlated with mean density of rainbow trout (correlation analysis, rΩª0.70, PΩ0.01; Fig. 3). To examine the hypothesis that coldwater patch 5 Fig. 2. Seasonal temperatures and refuge use by rainbow trout for reaches where refugia were present. Solid lines denote daily maximum temperatures of ambient stream flow at upper, middle and lower reaches within each stream. Dotted lines de- note daily maximum temperatures of coldwater patches in which continuously recording thermographs were placed. Solid bars show mean proportion of rainbow trout in refugia for specified sampling dates. frequency as well as ambient maximum daily stream temperature influenced rainbow trout den- sity, we forced coldwater patch frequency into a multiple regression model with mean daily maxi- mum temperature. Coldwater patch frequency was marginally associated with rainbow trout density (PΩ0.06; extra sum of squares F-test) after ac- Ebersole et al. Fig. 3. Relationship between mean rainbow trout density and mean daily maximum water temperatures of 12 study reaches. The numbers refer to map codes for study reaches listed in Table 1. Fig. 4. Proportion of rainbow trout (solid circles) within refugia at different times of day. Shown are all observations of all refug- ia used at least once during the course of the study. Also shown is a mean diurnal ambient stream temperature curve (solid line) compiled from all twelve reaches for August 1, 1994. counting for ambient stream temperature (multiple regression model: densityΩ1.175–0.0428 max daily temperature ª0.098 coldwater patch frequency; R2Ω0.67). Use of thermal refugia Timing of refuge use by rainbow trout varied daily and seasonally. Directional movements of rainbow trout into and out of coldwater refugia were ap- parent for the eight refugia in Joseph and North Pine Creeks that were sampled multiple times per day at intervals of 2–5 hours during warm, mid- summer sampling dates. Movement of rainbow trout into refugia occurred when ambient tempera- tures ranged from 18 to 25æC. We observed no con- sistent threshold temperature that triggered entry into refugia. Use of refugia tended to peak mid- 6 late afternoon, when temperatures in many streams were near the daily maxima, and declined with cooling stream temperatures by evening (Fig. 4). Within refugia, rainbow trout were frequently observed to aggregate at densities much higher than those observed elsewhere within a stream, presumably to fit within the often small confines of available refuge space (Fig. 5). Where differ- ences in fish size existed within a refuge, we oc- casionally observed the largest fish dominating the coldest portion of the refuge and apparently ex- cluding other rainbow trout from the coldest avail- able temperatures (Fig. 5). Because refuge use was highly time and tempera- ture-specific throughout the summer, summaries of seasonal refuge use were compiled only from ob- servations made during mid-day (1400–1800 hours) when refuge use was at, or near, maximum. Seasonally, proportional refuge use peaked in mid- summer (mid-July to mid-August) at several reaches, coinciding with seasonal temperature maxima. Refuge use subsequently declined with cooler temperatures in September (Fig. 2). Cold- water refuge use was not observed in the Broady Creek reach, where ambient stream temperatures never exceeded 20æC. Although maximum daily stream temperatures exceeded 22æC for several weeks in many reaches, overall proportional refuge use did not exceed 0.55 on any sample date in any reach (Fig. 2). Fig. 5. Distribution of rainbow trout in relation to water tem- perature gradients in coldwater refuge UJ4, upper Joseph Creek, July 13, 1994 Stream temperatures, refugia and rainbow trout Fig. 6. Relationship of mean refuge use (mean proportion of rainbow trout in a site that were utilizing refugia over all sam- pling dates) and mean daily maximum water temperatures for reaches with refugia. The numbers refer to map codes for study reaches listed in Table 1. Seasonal mean proportions of rainbow trout utilizing coldwater refugia were significantly higher in reaches with warmer mean ambient stream tem- peratures (rΩ0.71, PΩ0.04; Fig. 6). A few cold- water patches that we monitored were never used (LJ3, LJ4, MJ3, MP1 and BR1; Table 2). BR1 was located in a cold reach where ambient stream tem- peratures rarely warmed to stressful levels. The re- maining unused coldwater patches were vacant even on warm days when rainbow trout were pres- ent nearby. We found no significant differences in physical characteristics (Table 2) between cold- water patches used at least once and those cold- water patches never used (two-sample t-test, two- sided P±0.05, d.f.Ω20 for all pair-wise compari- sons of factors). Among coldwater refugia used by rainbow trout, mean mid-day rainbow trout density within refugia was positively associated with mean refuge depth (Pearson’s correlation coefficientΩ0.46; PΩ 0.03). The proportion of rainbow trout within a sampling reach using individual thermal refugia was also positively associated with the mean refuge depth (Pearson’s correlation coefficientΩ0.60; PΩ 0.003). Mean ambient-cold patch temperature dif- ference and total coldwater patch cover were not significantly correlated with either mean mid-day rainbow trout density or proportional use (P±0.4 for all factors). Coldwater patch surface area was not significantly correlated with proportional use (P±0.5). Discussion We found no clear effect of coldwater refuge fre- quency on rainbow trout density at the stream reach scale. Rather, ambient stream temperatures 7 explained most of the difference in overall rainbow trout densities between the study reaches. Within individual warm stream reaches, from 10% to 40% of individuals regularly congregated in small, rela- tively cold refugia during daily and seasonal periods of presumed thermal stress. Total numbers of rainbow trout utilizing individual refugia were low and seldom exceeded 10 individuals at any given time. Coldwater patches were seldom used by salmonids in reaches with cool ambient tem- peratures. While several researchers have observed pro- nounced movements of salmonids into refugia at threshold temperatures near 22–23æC (Li et al. 1993; Nielsen et al. 1994), we observed rainbow trout moving into coldwater refugia as streams warmed across a range of temperatures from 18 to 25æC. Differences in temperatures at which fish seek out thermal refugia may reflect acclimation to temperatures within different streams or even differences in temperatures at different times of the day or season (e.g., Konecki et al. 1995). Variation in reach specific thermal tolerances, and subsequent refuge use patterns, could con- ceivably arise from differences in physiological tolerance or behavioral habituation (Taylor 1991) between local populations within a region. Esti- mates of upper thermal tolerance or avoidance limits for juvenile rainbow trout (at unlimited ra- tion) range from 22 to 26.6æC, and final preferen- dum or maximum growth temperatures range from 18 to 19æC (Coutant 1977; Eaton et al. 1995). However, Needham (1937) reported finding Nelson’s trout O. mykiss nelsoni residing in streams in Baja California, Mexico, where water temperatures exceeded 27æC. Our failure to detect a strongly significant influ- ence of coldwater patch frequency on reach level differences in rainbow trout abundance could have several sources. Refugia within our reaches com- prised very small portions (∞7%) of total available space (Table 1). Like small habitat islands in a fragmented landscape, small, infrequent refugia could presumably be difficult for stressed individ- uals to locate (Sedell et al. 1990). Competition for desired thermal environments within refugia may also exist (Magnuson et al. 1979). Our obser- vations of large trout dominating the coldest por- tions within certain refugia suggest this possibility. In experimental thermal gradients, large bluegill Lepomis macrochirus excluded sub-dominant fish from preferred thermal hahitats (Beitinger & Mag- nuson 1975). While we observed no significant as- sociation between coldwater patch area and pro- portional use by rainbow trout, coldwater patch mean depth was positively associated with rainbow trout density and proportional use. The mean Ebersole et al. depth of coldwater patches used by rainbow trout ranged from 0.08 to 0.74 m. Coldwater refuges located at channel margins may be significantly shallower than adjacent thalweg habitats, but the range of depths we observed is not outside the range of depths occupied by juvenile O. mykiss under more optimal thermal conditions in thalweg habitats (Everest & Chapman 1972; Bugert et al. 1991). However, if subject to increased mammalian or avian predation risk while concentrated at high densities in shallow refugia, fish may face a trade- off between optimal thermal microhabitats in shal- low refugia and decreased predation risks in deeper main-stem habitats. Only a small proportion of rainbow trout in a reach were observed to use thermal refugia at any given time, even at very warm temperatures. Given the small available area within refugia, perhaps some fish were excluded from refugia, as proposed by Biro (1998) for brook trout Salvelinus fontinalis utilizing spring seeps in lakes. Alternatively, some individuals may have a higher tolerance for warm temperatures or may have used refugia that we did not detect. We frequently observed young-of-the- year rainbow trout nosing into crevices between cobbles during periods of stream warming but were rarely able to detect substantial (±3 cm3) vol- umes of cooler water in these areas. Tiny seeps that mix rapidly with surface waters may have provided abundant, small refugia that we were unable to de- tect. Jobling (1981) proposed that fish may sample thermal environments intermittently. If this were the case, individual fish may have moved in and out of refugia repeatedly between sampling obser- vations, so that the total number of fish utilizing a refuge may have been much higher than the num- ber observed within the refuge at any one time, producing underestimates of refuge use rates. We did not measure the distance rainbow trout moved to use thermal refugia, but Li et al. (1993) and Nielsen et al. (1994) observed salmonids mov- ing more than 25 m to use cold seeps and sug- gested that actual distances may be much higher. If demand for suitable thermal refuge space exceeds availability within warm stream reaches, resident fish unable to establish positions within refugia may be forced to emigrate or may perish. Limited suitable habitat for adult striped bass Morone sax- atilis seeking refuge from anoxic or warm waters has been proposed as an explanation for summer die-offs and poor fish condition in reservoirs and estuaries (Coutant 1985). It is possible that potential refugia defined ac- cording to one physiological parameter (thermal requirements) may fail to meet other physiological or ecological requirements of the organism. Fish using patches that are distinct from the surround- 8 ing matrix in one factor (e.g., temperature) may face potential trade-offs in other, unmeasured en- vironmental conditions (Matthews, 1998, p. 375). Potential trade-offs for using groundwater-fed thermal refugia, especially along stream margins, include decreased dissolved oxygen (Matthews & Berg 1997), decreased food availability and in- creased predation risk. At present, the few thermal refugia available in the streams we examined may be too shallow, too small, too infrequent or otherwise insufficient to alone sustain high densities of rainbow trout in a reach given prolonged high ambient stream tem- peratures. However, the few thermal refugia pres- ently available could be allowing some coldwater fishes to persist, although at low densities, in warm stream reaches. The strategy of coldwater refuge use may effectively broaden the regional range of suitable rearing area in arid-land streams and allow use of resources available in highly pro- ductive lower gradient stream reaches otherwise in- accessible to salmonids during summer due to high water temperatures. The northeast Oregon streams we studied, like many throughout the western United States, have experienced rapid changes in channel morphology, riparian vegetation, and hydrological regime over the past century (McIntosh et al. 1994). Human land uses such as logging, grazing, urbanization, groundwater pumping, or stream regulation may have contributed to altered thermal patterns in streams, not only by allowing warming of ambient stream temperatures but also by altering the fre- quency, size, accessibility, or other physical charac- teristics of thermal refugia (Torgersen et al. 1999). Such alterations could be induced by reducing op- portunity for thermal stratification within stream channels (Nielsen et al. 1994), eliminating sources of cold groundwater or tributary flow (Berman & Quinn 1986), warming shallow groundwater (Hew- lett & Fortson 1982), or simplifying stream chan- nels and floodplains (Frissell et al. 1996). For these reasons, we do not know whether the patterns of thermal heterogeneity observed today reflect con- ditions under which native fishes evolved, nor do we know how rainbow trout might have histori- cally used habitats that today are marginally suit- able. Thermal refugia are of obvious value to cold- water fishes in many circumstances (e.g. Nielsen et al. 1994; Biro 1998). However, our study sug- gests that at least for interior rainbow trout and steelhead, it may be unreasonable to expect pro- found density or production responses from ac- tions that simply maintain or restore thermal refu- gia and fail to ameliorate ambient water column temperatures in altered arid-land streams. Stream temperatures, refugia and rainbow trout Resumen 1. Altas temperaturas en rı´os pueden limitar la distribucio´n y abundancia de Oncorhynchus spp. Se han planteado hipotesis sobre el papel de refugios te´rmicos frı´os originados por surgen- cias de aguas subterra´neas, aunque nunca se han descrito de forma cuantitativa. 2. En 12 localidades de varios rı´os situados en el noreste de Orego´n (USA), las abundancias de O. mykiss, observadas entre el 21 de junio y el 15 de septiembre del an˜o 1994, estuvieron inversamente correlacionadas con la media de las temperaturas ma´ximas de los rı´os (rΩª0.7, P∞0.05). Algunos individuos uti- lizaron refugios te´rmicos (1–10 m2 de a´rea superficial) que de promedio tuvieron 3–8æC por debajo de la temperatura de los rı´os. En los sectores ma´s ca´lidos, temperaturas ±22æC prevale- cieron desde mediados de junio hasta agosto, encontra´ndose un 10–40% de los individuos dentro los refugios te´rmicos durante el mediodı´a que es el momento de mayor temperatura en los rı´os. 3. Tras considerar el efecto de la temperatura, comprobamos que la frecuencia de puntos frı´os dentro de una misma localidad no estuvo significativamente relacionada con la densidad de las truchas (PΩ0.06, F-test). Debido al prolongado mantenimiento de las altas temperaturas en algunas localidades, el nu´mero de refugios te´rmicos disponibles en los rı´os debe ser demasiado pequen˜o e infrecuente como para mantener densidades elevadas de O. mykiss. Sin embargo, estos refugios permiten que la espe- cie persista, au´n en bajas densidades, en las zonas de agus ca´li- das de estos rı´os. Acknowledgments We thank S. Alvord, T. Rohweder, R. Frank and K. Loso for outstanding field assistance. B. Vondracek, Carl R. Ruetz, C. Torgersen, H. Li, R Gresswell, C. Coutant, and T. Pearsons provided helpful advice. S. Wulffprovided statistical assistance. We thank the landowners along Joseph Creek, particularly L. Scott and J. Kootch, for access and assistance to the field reaches. Funding for this study was provided by the U. S. En- vironmental Protection Agency, ERL Duluth. References Beitinger, T.L. & Magnuson, J.L. 1975. Influence of social rank and size on the thermoselection behavior of bluegill (Lepomis macrochirus). Journal of the Fisheries Research Board of Canada 32: 2133–2136. Berman, C.H. & Quinn, T.P. l99l. Behavioural thermoregul- ation and homing by spring chinook salmon, Oncorhynchus tshawystcha (Walbaum), in the Yakima River. Journal of Fish Biology 39: 301–312. Biro, P.A. 1998. Staying cool: behavioral thermoregulation dur- ing summer by young-of-year brook trout in a lake. Trans- actions of the American Fisheries Society 127: 212–222. Bisson, P.A., Nielsen, J.L., Palmason, R.A & Grove, L.E. 1982. A system of naming habitat types in small streams, with examples of habitat utilization by salmonids during low stream flow. In: Armantrout, N.A., ed. Acquisition and util- ization of aquatic habitat inventory information. Proceedings of a symposium, 28–30 October 1981, Portland, Oregon. American Fisheries Society. Brett J.R. 1952. Temperature tolerance in young Pactfic salmon, genus Onchorhynchus sp. Journal of the Fisheries Research Board of Canada 9: 265–323. Brett J.R. 1979. Energetic factors and growth. In: Hoar, W.S., Randall, D.J. & Brett, J.R., eds. Fish physiology, volume 8. New York: Academic Press. 599–675. Bugert, R.M., Bjornn, T.C. & Meehan, W.R. 1991. Summer 9 habitat use by young salmonids and their responses to cover and predators in a small southeast Alaska stream. Trans- actions of the American Fisheries Society 120: 474–485. Corbett, D.M. 1955. Stream-gaging procedure. A manual de- scribing methods and practices of the Geological Survey. United States Geological Survey Water Supply Paper 888: 1– 245. Coutant, C.C. 1977. Compilation of temperature preference data. Journal of the Fisheries Research Board of Canada 34: 739–745. Coutant, C.C. 1985. Striped bass, temperature, and dissolved oxygen: a speculative hypothesis for environmental risk. Transactions of the American Fisheries Society 114: 31–61. Eaton J.G., McCormick, J.H., Goodno, B.E., O’Brien, D.G., Stefan, H.G., Hondzo, M. & Scheller, R.M. 1995. A field information-based system for estimating fish temperature tolerances. Fisheries 20(4): 10–18. Everest, F.H. & Chapman, D.W. 1972. Habitat selection and spatial interaction by juvenile chinook salmon and steelhead trout in two Idaho streams. Journal of the Fisheries Research Board of Canada 29: 91–100. Frissell, C.A., Ebersole, J.L., Liss, W.J., Cavallo, B.J., Poole, G.C. & Stanford J.A. 1996. Potential effects of climate change on thermal complexity and biotic integrity of streams: seasonal intrusion of non-native fishes. Environmen- tal Protection Agency Final Report CR-822019-01-0. Duluth, Minnesota. Gibson, R.J. 1966. Some factors influencing the distribution of brook trout and young Atlantic salmon. Journal of the Fish- eries Research Board of Canada 23: l977–1979. Hankin, D.G. & Reeves, G.H. 1988. Estimating total fish abun- dance and total habitat area in small streams based on visual estimation methods. Canadian Journal of Fisheries and Aquatic Sciences 45: 834–844. Hewlett, J.D. & Fortson, J.C. 1982. Stream temperature under an inadequate buffer strip in the Southeast Piedmont. Water Resources Bulletin 18(6): 983–988. Jobling, M. 1981. Temperature tolerance and the final preferen- dum-rapid methods for the assessment of optimum growth temperatures. Journal of Fish Biology 19: 439–455. Kaya, C.M., Caddying, L.R. & Burkhalter, D.E. 1977. Use of a cold-water refuge by rainbow and brown trout in a geother- mally heated stream. Progressive Fish Culturist 39: 37–39. Konecki, J.T., Woody, C.A. & Quinn, T.P. 1995. Critical ther- mal maxima of coho salmon (Oncorhynchus kisutch) fry under field and laboratory acclimation regimes. Canadian Journal of Zoology 73: 993–996. Li, H.W., Pearsons, T.N., Tait, C.K., Li, J.L. & Gaither, R. 1993. Approaches to evaluate habitat improvement programs in streams of the John Day Basin. Completion Report, Ore- gon Cooperative Fishery Research Unit, Oregon State Uni- versity, Corvallis, Oregon. Li, H.W., Pearsons, T.N., Tait, C.K., Li, J.L. & Gaither, R. 1994. Cumulative effects of riparian disturbances along high desert trout streams of the John Day basin, Oregon. Trans- actions of the American Fisheries Society 123: 627–640. Magnuson, J.J., Crowder, L.B. & Medvick, P.A. 1979. Tempera- ture as an ecological resource. American Zoologist 19: 331– 343. Matthews, K.R. & Berg, N.H. 1997. Rainbow trout responses to water temperature and dissolved oxygen stress in two southern California stream pools. Journal of Fish Biology 50: 50–67. Matthews, W.J. & Zimmerman, E.G. 1990. Potential effects of climate change on native fishes of the southern Great Plains and the Southwest. Fisheries 15: 26–32. Matthews, W.J. 1998. Patterns in freshwater fish ecology. New York: Chapman and Hall. McIntosh, B.A., Sedell, J.R., Smith, J.E., Wissmar, R.C., Clar- Ebersole et al. ke, S.E., Reeves, G.R. & Brown, L.A. 1994. Historical changes in fish habitat for select river basins of eastern Ore- gon and Washington. Northwest Science 68: 36–52. Meisner, J.D. 1990. Effect of climatic warming on the southern margins of the native range of brook trout, Salvelinus fontin- alis. Canadian Journal of Fisheries and Aquatic Sciences 47: 1065–1070. Needham, P.R. 1937. Notes on the introduction of Salmon nel- soni Evemmann into California from Mexico. Transactions of the American Fisheries Society 67: 139–146. Nielsen, J.L., Lisle, T.E. & Ozaki, V. 1994. Thermally stratified pools and their use by steelhead in northern Califonia streams. Transactions of the American Fisheries Society 123: 613–626. Ozaki, V.L. 1988. Geomorphic and hydrologic conditions for cold pool formation on Redwood Creek, Califomia. Red- 10 wood National Park Research and Development Technical Report 24. 57 pp. SAS Institute Inc. 1989. SAS/STAT user’s guide, version 6. 4th edn. Volume 1. Cary, NC: SAS Institute Inc. Sedell, J.R., Reeves, G.H., Hauer, F.H., Stanford, J.A. & Hawk- ins, C.P. 1990. Role of refugia in recovery from disturbance: modern fragmented and disconnected river systems. Environ- mental Management 14: 711–724. Taylor, E.B. 1991. A review of local adaptation in Salmonidae, with particular reference to Pacific and Atlantic salmon. Aquaculture 98: 185–207. Torgersen, C.E., Price, D.M., Li, H.W. & McIntoch, B.A. 1999. Multiscale thermal refugia and stream habitat associations of chinook salmon in northeastern Oregon. Ecological Ap- plications 9: 301–319.