Trout are not indigenous to most high mountain lakes in the west (Bahls, 1992). Trout were stocked in NOCA lakes during this century to provide recreational angling opportunities (Jarvis, 1987). The principle species that have been stocked in NOCA are cutthroat (Oncorhynchus clarki) and rainbow trout (O. mykiss). In some lakes, trout have established reproducing populations which can reach high densities. In other lakes, trout do not reproduce and fry are periodically stocked at relatively low densities.

A central hypothesis of this research is that effects of fish on native biota vary along a gradient of increasing fish predation intensity. Predation intensity is related to density and size structure of fish populations. Thus, it was necessary to obtain at least a relative estimate fish density in study lakes.

Population Estimates

Population estimates were conducted on 13 lakes between 1990 and 1993. Eight of these lakes support reproducing populations of trout. Fish in the other five lakes do not successfully reproduce (Table 1) and must be periodically stocked. Lack of reproduction was indicated by absence of suitable spawning habitat, failure to observe fry or smaller fish, and little variation in age and size structure. Abundance was estimated using two basic methods; fishing success and mark-and-recapture.

*CTT= cutthroat trout; RBT= rainbow trout

Fishing success method

In 1990, as part of the lake manipulations, an attempt was made to remove all fish from LS1, MR13-2, Lower Panther, and Upper Panther lakes using gill nets. These are all relatively small lakes with low fish density, and none supported reproducing populations of fish. From one to three gill nets were set in each lake for varying periods of time. When nets were retrieved, all fish were removed, and the gill nets were reset immediately. The period of time that each net was fished was recorded. This procedure continued until fish were no longer captured in the nets. Following removal, fry were stocked in Lower Panther, LS1, MR13-2, and MR11 (a lake with a history of stocking that was fishless in 1990 and probably for several prior years). Fry were not restocked in Upper Panther.

Catch per unit effort estimates were combined for all nets fished during a given time period. Regression of catch per unit effort on cumulative catch from gill nets was used to estimate total populations for these lakes according to the Leslie method (Ricker, 1975). Ninety-five percent confidence limits were calculated for each estimate (Ricker, 1975).

Mark-recapture estimates

Population density was estimated in a total of 11 lakes by mark-and-recapture. Fish density was estimated in two lakes (LS1, Lower Panther) by fishing success in 1990 and by mark-recapture in 1992. Lakes where mark-recapture was attempted were small (6.3 ha), and angling was the best available method for capturing fish. Fish in the mark sample were collected by angling, fin-clipped, and released. We were able to mark and release fish by angling from the shore along the entire perimeter of most lakes. Fish also were captured and marked from an inflatable boat. Injured fish (e.g., bleeding from gills or tongue) were either released unmarked or sacrificed. Two groups, each with 12 fish, were held overnight after capture and marking to estimate post-release mortality due to handling and marking.

Fish were recaptured with monofilament gill nets. The nets were usually set the day after marking was completed. To minimize reduction in capture efficiency, nets were checked frequently to insure that fish did not accumulate. Overnight sets were checked the following morning. In the course of recapturing, net locations were changed often.

In 1991, mark-recapture was conducted twice at McAlester, Rainbow, Lower Triplet and Upper Triplet lakes, and three times at LS2. Three mark-recapture estimation procedures were compared at LS2. Initially, marked (M), total fish captured by gillnetting (C), and marked fish recaptured by gillnetting (R) from each of the three sample periods at LS2 were combined, and population abundance was estimated with a single-census Petersen estimator (N =(M+1)(C+1)/(R+1); Ricker, 1975). A repeated census procedure used the mean of three single-census estimates (one for each sample period) to provide a second assessment of the population in LS2. The Schnabel multiple-census procedure was similar to the single census using all fish, except the product of marked and sampled fish was calculated separately for each sample period (N = (C_tM_t)/R; Ricker, 1975). Single-census estimates were used for all other lakes sampled in 1991. In 1992 and 1993, mark-recapture was conducted only one time at each lake, and density was estimated by a single-census Petersen procedure.

Assumptions associated with the single-census Petersen estimate include equal natural mortality and vulnerability to capture for marked and unmarked fish (Ricker, 1975). Marks must be retained and remain recognizable throughout the period. Distribution of marked and unmarked fish is assumed to be random, or if not random, subsequent sampling is proportional to the number of fish in different areas. Finally, recruitment should be negligible during the sample period (Ricker, 1975). Requirements for the Schnabel estimator are similar, except no mortality or recruitment can occur (Ricker, 1975).

Differences in gear selectivity may bias the outcome of mark-recapture estimates (Ricker, 1975). Size selectivity may be exacerbated in this study because angling may be selective for a different size range of fish than gill nets. Detection of unequal vulnerability can be accomplished by comparing capture and recapture rate for fish of different sizes, and the effects of selectivity can be minimized by dividing estimates among two or more size groups or excluding fish near the limits of a given fishing gear.

Fish caught by angling and marked were not measured prior to release; however, fish in the recapture sample provided an estimate of the size distribution of fish caught by angling. To evaluate the magnitude of possible differences in size selectivity of angling and gillnetting, mean size of marked fish in the recapture sample (an estimate of size of fish captured by angling) was compared to the mean size of all fish captured by gillnetting (ANOVA). Additionally, a direct comparison (ANOVA) of length-samples of angler-captured trout (measured to the nearest 25 mm) and those captured in gill nets was conducted on samples from Lower Triplet and Upper Triplet lakes in 1991. Because abundance estimates of larger fish may be more precise (Ricker 1975), abundances of fish > 170 mm and > 200 mm (TL) also were estimated.

Population estimates were converted to densities by dividing estimated abundance by lake surface area (ha). For each lake, mean individual weight for fish was estimated from a weight-length equation using the mean length of fish captured as the predictor variable. Density for each lake was multiplied by the mean individual weight of the sample to approximate biomass per ha (kg·ha^-1). Abundance of fish in lakes with reproducing populations of fish (a single mean value was used for lakes with more than one abundance estimate) was compared to fish abundance in lakes where reproduction did not occur (all estimates included) using one-way ANOVA.

In 1992 and 1993, we used gill nets with somewhat smaller mesh size than the nets used in 1991. To test for possible differences in net efficiencies, samples collected in 1991 and 1992 from LS2, McAlester, and Rainbow lakes were compared (ANOVA).

Age, Length. Weight

Fish captured in gill nets were measured to the nearest 1 mm (total length TL and fork length) and weighed to the nearest 10 g. Otoliths were removed from a subsample of fish from each lake for age determination. Mean length (TL), weight, and Fulton's condition factor (K; Ricker, 1975) were calculated for fish captured at each lake. A weight-length relationship was estimated for each population using a GM functional regression (Ricker, 1973; 1975). Age was estimated for individual fish by counting annuli occurring on otoliths (Secor et al., 1991).

Only a small number of fish from each lake have been aged so far, but age distribution of the these fish was expanded to the entire number of fish captured in each lake using an age-length key (Kimura, 1977; Westerheim and Ricker, 1978). Initially, age composition was determined for 10-mm length intervals in the subsample of fish that were aged at each lake. Subsequently, the proportion of fish of a particular age in each 10-mm size group (aged-fish only) was multiplied by the total number of fish captured in the respective size interval. Mean age for each sample was estimated from the expanded distribution at each lake. Because age information was limited to age-at capture, growth comparisons were based on variables obtained by the Brody-Bertalanffy procedure (Ricker, 1975). Mean length was estimated for each age group in the subsample of fish that were aged. The Brody Growth Coefficient (K) and the theoretical asymptotic length (L) were estimated by fitting mean length for each age group to a specialized form of the von Bertalanffy growth formula (surface factor (D) = 1; Pauly, 1984). Age groups with less than two fish were not included in the growth analysis. Age and growth information for a limited number of lakes is presented here. Analysis of age and growth for all lakes is continuing.

Trout Diet

In general, taxa from stomach contents were identified to the family level. Identified taxa were placed into food categories according to their probable location in the lake (benthic, water column, lake surface) when ingested.

Age, Length, and Growth

Mean length, weight, and condition factor of fish captured in gill nets from North Cascades lakes (Table 1) differed significantly among lakes (P <0.01). Largest mean lengths were recorded in MR9 in 1993 (314 mm) and Lower Panther Lake in 1990 and 1992 (278 and 263 mm, respectively), and the smallest fish were captured in Kettling Lake in 1993 (183 mm) and Lower Triplet Lake in 1991 (187 mm). Mean length, weight, and condition factors for fish from lakes with reproducing populations (Table 1) were significantly smaller (P <0.01) than estimates from lakes without reproduction.

Mean age of fish in lakes with reproducing populations (Table 2) ranged from 3.3 years in LS2 (1991) to 4.9 years in Upper Triplet Lake (1991). Although observed longevity was greatest in Upper Triplet Lake, distribution of fish among age groups suggests substantial variation in reproductive success in this lake.

Table 2. Age distribution (%) estimated from an age-length key for fish captured in gill nets from lakes in North Cascades National Park Complex, 1991-1992.

Fish sampled in LS1 and Lower Panther Lake in 1992 were from the cohort of fry (1091·kg^-1) stocked in 1990; thus all fish from these lakes were 2 years old (Table 2). Significant differences in the mean length of fish from the two lakes suggests that growth rate of fish is much greater in Lower Panther than in LS1; however, it is apparent that fish in both of these lakes are growing at a much faster rate than those in lakes with natural reproduction (Table 3).

Brody growth coefficients were quite similar for LS2, McAlester, and Lower Triplet lakes (Table 3). Results from this technique suggest that fish in Upper Triplet Lake are growing at a much faster rate than fish in the other three lakes. Sample size is small, however, and among lake comparisons are tentative. Analysis of age and growth is continuing.

Table 3. Mean length and number of otolith samples (n) of each age group of fish captured in gill nets in North Cascades National Park Service Complex, 1991-1992.

Lake Name	Lakes with Natural Reproduction
	Year	Age							K*	Length ()**
		1	2	3	4	5	6	7
LS2	1991	125 (2)	153 (4)	196 (15)	247 (7)	281 (1)	317 (1)		0.07	885
McAlester Lake	1992	123 (2)		179 (6)	194 (33)	216 (7)			0.09	429
Lower Triplet Lake	1991		139 (1)	188 (3)	202 (5)				0.10	333
Upper Triplet Lake	1991	122 (2)	187 (1)	216 (3)	243 (2)		276 (2)	285 (1)	0.35	308
Lake Name	Lakes without Natural Reproduction
	Year	Age							K*	Length ()**
		1	2	3	4	5	6	7
LS1	1992		203 (24)						--	--
Lower Panther Lake	1992		263 (20)						--	--

* Brody Growth Coefficient (Ricker, 1975)
** Asymptotic length

Population Estimates

Fishing success method

Fish did not reproduce in LS1, MR13-2, Upper Panther, and Lower Panther lakes. In 1990, population estimates were very low for these lakes (Table 4), but snorkeling observations and total fish removals suggest that these estimates were realistic. Six fish were removed from Upper Panther Lake by both gillnetting and angling; this was believed to represent the total number present in the lake. Upper Panther Lake is small (0.1 ha) and shallow (2.0 m maximum depth), and fish can be easily observed.

** Based on combined-samples single-census estimate.

After fish were removed from the lakes, fry were stocked in September 1990 (except in Upper Panther) at the normal stocking rate for each lake. Although stocking rate was highly variable, with the exception of Lower Panther, stocking density was 313 fish/ha (Table 5). Lower Panther, a relatively low elevation lake, is apparently stocked more heavily than other lakes, perhaps because fishing pressure is relatively high.

Table 5. Trout density in lakes that do not support reproducing populations, North Cascades National Park Service Complex, 1990-1992.

* Estimates of fish removed prior to fry stocking in September 1990.
** Represent fish surviving from fry stocked in 1990, estimated by mark-recapture (Table 6).

Mark-recapture estimates

Post-release mortality in the two overnight experiments was low. Only one of 24 fish died, and that fish was bleeding from the gills when placed in the holding area. According to normal procedures, injured fish would not have been included in the marked sample.

Population estimates based on single-census Petersen estimators ranged from 5086 (Rainbow Lake, 1992) to 25 (MR9, 1993; Table 4). Single-census, repeated single-census, and multiple-census estimates at LS2 were 722, 469, and 472 fish, respectively. Although the latter two estimates are substantially lower than the single-census calculation, confidence limits for the three estimates overlapped. Furthermore, the single-census figure in 1991 is most similar to the estimate obtained in 1992 (Table 4).

Substantial differences were apparent in the length-frequency distributions of all fish captured by gillnetting and recaptured-fish in the gill net catch (Figure 1a, 1b). Mean lengths were 209 and 235 mm for all fish captured and recaptured-fish, respectively, and the differences were statistically significant (P< 0.01). These results suggest that angling was selecting for larger fish than gillnetting; however, neither method sampled smaller fish in proportion to their relative abundance in the lake.

Differences between fish captured in gill nets and by angling in Upper Triplet Lake (187 and 213 mm, respectively) and Lower Triplet Lake (235 and 260 mm, respectively) in 1991 were not statistically significant (P> 0.60 and P> 0.13, respectively). The observed differences at both lakes (approximately 30 mm) were similar to those noted between unmarked and recaptured fish in gill-net samples. These results also suggest that angling selects for larger fish in the population. Lack of statistical significance may reflect high sample variance or difference in measurement techniques. Fish captured by angling were measured to the nearest 25 mm, but measurements of gill-net captured fish were to the nearest 1 mm.

In LS2, Rainbow, and McAlester lakes, comparison of mean length of fish captured in nets used in 1991 and nets with a smaller mesh used in 1992 suggests that differences in the size of fish were greater than would be expected by chance alone (210 and 198 mm, respectively; P< 0.01). It is difficult to assess the effects of these differences on population estimates because abundance estimates obtained in the two sample years for each of these lakes are within respective confidence intervals (Table 4).

The difference in mean population densities (all fish; Table 6) for lakes with reproducing fish populations (524 fish·ha^-1; n = 8) and lakes that did not support reproducing fish (106 fish·ha^-1; n = 7) was statistically significant (P< 0.01). The difference in mean biomasses of fish in lakes with and without reproducing fish (39 and 15 kg·ha^-1, respectively) was also statistically significant (P< 0.01). However, confidence intervals of the estimates are rather wide. The confidence intervals of the 1992 estimates for LS1 and Lower Panther, both with non-reproducing fish, overlap the confidence intervals of most reproducing populations. The fish in LS1 and Lower Panther are the remaining members of the cohorts of fry stocked in 1990.

Given the possible bias in population estimates due to size selection for larger fish, comparison among lakes may be most meaningful when limited to larger fish in the population. Within-lake similarity between estimates for all fish in the population and those limited to fish > 170 mm (Table 6) suggest that fish smaller than 170 mm may not be represented in the all-fish estimates in proportion to their abundance in the lake. Estimates of biomass within a given lake also were similar when all-fish and fish > 170 mm estimates were compared. Unfortunately, problems arise when comparisons were restricted by size because in some lakes the low number of recaptures caused statistical bias (Robson and Regier, 1964) in estimates (e.g., MR9). Limiting comparisons to fish > 200 mm substantially reduced abundance, density, and biomass estimates when compared to estimates for all fish and for fish > 170 mm, and furthermore, statistical bias was more common in estimates of fish > 200 mm (Table 6).

Table 6. Trout density and biomass (mean weight per fish estimated from weight-length equation) estimates for lakes in North Cascades National Park Service Complex, 1990-1993.

In 6 of 8 lakes with reproducing populations, estimates of density of fish > 170 mm were between 450 and 650 fish·ha^-1 (Table 6). MR16 was the only lake with natural reproduction where fish density and biomass was low. In contrast, all estimates of the density of fish > 170 mm in lakes without natural reproduction (n = 3, LS1 and Lower Panther in 1992, MR9 in 1993) were 320 fish·ha^-1. In 1990, estimates of fish density by the Leslie method were all <60 fish·ha^-1 (n = 4). If the mark-recapture estimates are compared statistically, differences are greater than would be expected by random variation (P <0.01). Average density of fry stocked in NOCA from 1976 to 1993 was 179 fish/ha (range 59.8 - 375 fish/ha; N = 37) except for Thunder, Upper Panther, and Lower Panther Lakes which are relatively heavily stocked. The average interval between stocking was > 5 years (Reed Glesne, North Cascades National Park Service Complex, 2105 Highway 20, Sedro Woolley, WA, 98284).

Trout Diet

Taxa found in trout stomach are given in Table 7. Diets reported in Table 8 are simply quantities of food types consumed, and do not reflect food preference or electivity. Food preference would entail evaluation of the amount of food consumed of a given type relative to the availability of that food type in the lake (Ivlev, 1961).

In forested and subalpine lakes trout fed on benthic, water column, and surface organisms (Table BA). In both vegetation zones, benthic prey comprised the largest proportion of the diet. They tended to be the most common food type early in the season in the subalpine. Benthic organism were more common in the diet of fish in forested lakes than in subalpine lakes.

Benthic larvae and pupae consumed by fish were primarily chironomids (Table 8B and C). In forested lakes, odonates (dragonflies and damselflies) and leeches, which are large prey items, were found occasionally in trout stomachs.

Table 7. Stomach contents of fish in 1990.

1. Benthic (Larva)

Coleoptera

Diptera
Dytiscidae
Hydroporidae
Gyrinidae

Gyrinus

Diptera


Chironomidae
Ceratopogonidae
Simuliidae
Tipulidae

Ormosia

Ephemeroptera


Baetidae

Callibaetis
SiphloNeuridae

Ameletus

Odonata


Anisoptera

Aeshnidae

Aeshna
Corduliidae

Somatochlora
Zygoptera

Coenagrionidae


Pelecypoda

Plecoptera

Chloroperlidae
Perlidae

Tricoptera

Limnephilidae

Apatania
Asynarchus
Desmono mono
Halesochila taylori
Hydropsychidae

2. Benthic (pupa)

Diptera
Chironomidae

Tricoptera

Limnephilidae

3. Water Column

Acari

Acarina

Diptera


Chaoboridae
Chaoborus

Hemiptera


Notoniectidae
Notonecta (adult)

Cladocera

Copepoda

4. Surface (Aquatic Adults)

Diptera

Ceratopogonidae
Empididae

Ephemeroptera


Baetidae

Callibaetis

Neuroptera


Sisyridae
Sisyra

Plecoptera

Chloroperlidae
Elateridae

5. Surface (Terrestrial, primarily adults)

Arachnidae

Chilopoda

Coleoptera

Curculionidae
Elateridae

Homoptera

Aphidae
Cercopidae
Cicadellidae

Hymenoptera

Formicidae

Hemiptera

Thysanoptera

6. Surface (Unknown)

Brachycera (adult)
Tipulidae (adult)

7. Benthic and water column

Amphipoda
Gammaridae
Gammarus

Coleoptera

Dytiscidae
Agabus
Potomonectes griseostriatus

Hirudinea

Nematoda

Turbellaria

8. Other

Salamander
Ambystoma sp.

Table 8a. Percentage of major food groups in trout diet determined from numbers of organisms found in stomachs.

Table 8b. Macroinvertebrate and vertebrate organisms recovered from trout-stomach samples (NOCA lakes, Summer 1990). Average number of prey organisms per fish for each food grouping arrange by vegetation zone.

Table 8c. Percent of prey organisms per fish for each food group arranged by vegetation zone.

Zooplankton occurring in fish stomachs were principally copepods, although Daphnia were found in one fish stomach in Trapper Lake. In the subalpine, prey from the water column were mostly zooplankton, but in forested lakes water column organisms other than zooplankton (Table 7) were most numerous in the diet. While benthic and surface organisms were found in the stomachs of fish from nearly all lakes, zooplankton were completely absent from fish stomachs in many lakes. In many of these lakes, particularly in the subalpine, the density of the largest common zooplankter, either D. kenai or D. arcticus, was very low (Table 8B and C). In some lakes this large zooplankter was not present (Upper and Lower Triplet and Thunder). Zooplankton, primarily Daphnia, also occurred in salamander diets, but contents of salamander stomachs have not yet been processed.

The majority of prey taken from the lake surface were terrestrial in origin (Table 8B and C). Terrestrial prey were largely coleopterans and formicids. Adults of aquatic insects did not occur very often in the diet. Fish from a few forested lakes had consumed empidids and callibaetids. A salamander was found in one stomach from Lower Panther Lake. A few other fish stomachs from this lake, as yet unprocessed, also contained salamander larvae.

Overall it appeared that fish were opportunistic feeders. Many fish appeared to be selective feeders. For example, in Upper Triplet on the first sample date, two fish concentrated on ants, another on terrestrial beetles and hemipterans, and another on chironomid pupae. At LS 1, one fish consumed chironomid pupae and Chaoborus, another selected chironomid larvae, a third concentrated on a Limnophilid, Asynarchus, and a fourth had consumed about 450 copepods.

1. The data analyzed to date suggest that reproducing fish populations have higher densities, lower individual growth rates, and are in poorer condition than fish from lakes where reproduction does not occur. In lakes with non- reproducing fish, fry are periodically stocked at relatively low densities and cohort density declines after stocking.

2. The mark-recapture procedure does not provide a reliable estimate of all fish in the population. Smaller fish were undersampled by both angling and gillnetting. Thus the estimates are, at best, indices of population density.

3. Angling and gillnetting tend to sample somewhat different segments of the population. Angling tends to select for a larger fish than gillnets. Thus, abundance estimates may be more meaningful when limited to larger fish (e.g., > 170 mm).

4. The mark-recapture technique seems to work best in relatively small lakes (<5-6 ha). Success also depends on the species and race of trout. Cutthroat and rainbow trout initially stocked several decades ago (Jarvis, 1987) were more vulnerable to both angling and gillnetting than Mt. Whitney rainbow, which have been recently stocked in NOCA lakes. In 1992 we were unable to obtain estimates of density in two manipulated lakes (MR13-2 and MR11) which had been stocked with the Mt. Whitney strain in 1990, because we were unable to capture by angling and mark a sufficient number of fish.

5. Trout in NOCA are opportunistic feeders. They preyed on benthic, water column, and surface organisms. Diets were highly variable between lakes.