Marcus Sundbom, Department of Limnology, Evolutionary Biology Centre, Uppsala University, Norbyvägen 20, SE-752 36 Uppsala, Sweden (fax +46 18531134; e-mail marcus.sundbom@ebc.uu.se).
Summary
1. Freshwater fish are a potentially important link in the transfer of radionuclides from polluted ecosystems to people. A pulsed contamination event such as the Chernobyl fallout in 1986 is a challenge to the prediction of radioactivity in biota, because activity concentrations of radionuclides can change dynamically among populations during an initial equilibration phase. This was demonstrated from time-series of 137caesium (Cs) in fish from three Swedish lakes (1986–2000, eight species, > 7600 individuals). In addition, we used these data to test hypotheses about the influence of fish size and trophic level on the temporal patterns of 137Cs.
2. In order to facilitate comparisons, a pulse-response model was developed to extract key parameters from field data: the timing (tmax) and level (Csmax) of the peak concentrations, the near steady-state level (Csbase) and the long-term decline rate (λ).
3. Peak concentrations in different fish were attained 56–806 days after the fallout. This delay (tmax) increased with body size and trophic level. Csmax increased with fish size but was highest at intermediate trophic levels. Csbase increased by a factor of 1·8 per trophic level, but was not influenced by size across species. The 137Cs-size relationship within species, however, shifted from negative to positive during the first 2 years.
4. The apparent ecological half-life (T1/2) for 137Cs increased after tmax and initially differed among fish and lakes. About 10 years after the fallout T1/2 was no longer significantly different among fish or lakes, suggesting that steady-state among 137Cs levels in fish had been reached. The mean T1/2 during 1996–2000 was 46 years, implying that the future recovery from present 137Cs levels (more than 10-fold higher than pre-Chernobyl levels) will be dominated by the physical decay of 137Cs (30 years).
5. The influence of fish size and trophic level on 137Cs dynamics has been consistent among the three lakes. The duration of the 137Cs pulse in fish appeared to be regulated by fish ecophysiology, whereas the amplitude also appeared to be regulated by lake characteristics.
6.Synthesis and applications. These data have implications for the predictive modelling of pulsed contaminants in lake ecosystems and show that (i) complex temporal concentration trends in fish can be adequately modelled by a combination of simple first-order equations; (ii) a one-component decline function is insufficient to describe the observed increase in ecological half-life over time; (iii) fish weight and trophic level can and should be treated as independent variables because their influence differs and changes over time. For management purposes, a pulse-response model for fish can be scaled to different systems by using established relationships to account for differences in fallout and catchment biogeochemistry.
Introduction
The accumulation of contaminants in aquatic biota is a well-documented phenomenon, with potential hazards for both ecosystems and human health. Contaminants differ in bioavailability, biological turnover times and their transfer to higher trophic levels (Thomann 1981; Chen et al. 2000). Concentrations in organisms are traditionally modelled using bioaccumulation factors, bioenergetic mass balance and empirical relationships (Thomann 1981; Håkanson 1994; Rowan & Rasmussen 1994; Trudel & Rasmussen 2001), approaches that often assume steady-state conditions. However, anthropogenic contaminants often enter the environment episodically or as a single pulse. Both the absolute and relative concentrations in and among populations will then change over time due to varying ambient concentrations and different contaminant turnover rates in different ecosystem compartments. The type of contaminant, the biogeochemistry of the recipient system, and food web structure and dynamics may all influence the temporal pattern and the time before background levels are re-established or new steady-state attained.
Radiocaesium (Cs) is ideal for studying the principles of bioaccumulation of pulsed contaminants because (i) 137Cs is an anthropogenic nuclide and background levels are therefore very low (cf. heavy metals); (ii) Cs is readily bioavailable but has a simple chemistry and is not metabolized (cf. Hg, many organic pollutants); (iii) Cs is a non-essential element; (iv) 137Cs measurement is simple and samples can be stored for decades (physical half-life = 30·17 years). The large-scale fallout of 137Cs from the Chernobyl nuclear accident in late April 1986 was a typical pulse contamination event. Within a few days 137Cs reached surface waters in amounts that in many regions far exceeded the background levels remaining from the 1950–60s nuclear bomb tests (Anspaugh, Catlin & Goldman 1988). The maximum activity concentrations in lake water were attained immediately after fallout; the subsequent removal of 137Cs from the water column was first rapid and then gradually slowed (Saxén, Jaakkola & Rantavaara 1996; Smith et al. 2000b). The fast removal was largely governed by lake dilution and particle settling rates (Smith, Comans & Elder 1999a). Later on, as the concentrations levelled off, secondary input from the catchment and remobilization from sediments by diffusion (Comans et al. 1989; Smith & Comans 1996) or resuspension (Broberg, Malmgren & Jansson 1995; Meili, Braf & Konitzer 1997) acted as secondary sources of 137Cs to lake water, slowing down the net removal of 137Cs. By assuming an exponential decline, the 137Cs decline rate can be expressed as the effective half-life. In this study, we frequently refer to the ecological half-life, which is defined as the effective half-life excluding physical decay.
Activity concentrations in fish typically lagged behind the 137Cs pulse in the water, maximum concentrations in fish occurring months to years after the fallout (Kryshev, Ryabov & Sazykina 1993; Ugedal et al. 1995; Saxén & Koskelainen 2001). Such a time lag partly results from the relatively long biological half-life of 137Cs in fish, which increases with fish size (Rowan & Rasmussen 1995). Because 137Cs is predominantly taken up from food (Morgan, Tytler & Bell 1994), a slow transfer of 137Cs up the food chain may also explain delayed accumulation. The time lag is thus expected to increase with fish size and trophic level. For non-pulsed radiocaesium, activity concentrations in fish often increase with size (Hannerz 1966; Kolehmainen, Häsänen & Miettinen 1968; Carlsson & Lidén 1978) or trophic level (Carlsson & Lidén 1977; Rowan & Rasmussen 1994), although there are exceptions (Rowan, Chant & Rasmussen 1998). In lakes affected by the Chernobyl fallout, the relationships between 137Cs and body size (Elliott et al. 1992; Ugedal et al. 1995), as well as between 137Cs and trophic level (Saxén, Jaakkola & Rantavaara 1996), have varied among species, lakes and years. Still, more than 15 years after the Chernobyl fallout, there is no consensus on the effects of size and trophic level on either the time scale or the magnitude of 137Cs contamination and depuration in fish. The reasons are that (i) most published studies cover only the first years after fallout, include few species, use few data, or had limited information on individual fish size or diet; (ii) the non-steady-state situation of 137Cs in the food web was not considered in early post-Chernobyl studies; and (iii) fish size and trophic level have not been treated separately so that the effect of trophic level might have been mistaken for a size effect.
In this study we investigated how the Chernobyl 137Cs pulse has developed among fish populations in three Swedish lakes during 1986–2000. In particular we were interested in quantifying the influence of fish size and trophic level on the time scale, maximum level, new steady-state levels and long-term decline rate of 137Cs. We used an analytical solution of a dynamic model to parameterize the long-term pattern of 137Cs in different fish categories. Similarities and differences among the three lakes were also evaluated with the aim of investigating whether our simple model can be used as a predictive tool. The study is based on time-series of 137Cs in fish that were collected regularly, usually several times per year, using consistent methods. The data include > 7600 fish from eight species, covering different size classes and feeding habits. As far as we know, an equivalent amount of 137Cs data, with size and diet information on individual fish, and covering 14 years after a major fallout, has not previously been available for scientific analysis.
Methods
Fish were collected from three small oligotrophic forest lakes, about 100–150 km north-west of Stockholm, Sweden (Table 1). Lake Ekholmssjön (59°52′N, 17°09′E) is a dimictic lake with two basins; Lake Flatsjön (60°29′N, 17°10′E) is shallow and polymictic, while Lake Siggeforasjön (59°52′N, 17°03′E) is deep and dimictic. The lakes are located in an area that received large amounts of rain-borne radionuclides a few days after the Chernobyl accident. Based on fish catch statistics, roach Rutilus rutilus L. and perch Perca fluviatilis L. were the most abundant fish species in all the lakes, followed by bream Abramis brama L. and pike Esox lucius L. Other species included in the study were ruffe Acerina cernua L., rudd Scardinius erytrophtalmus L., bleak Alburnus lucidus L. and crucian carp Carassius carassius L. Fish were collected up to six times yearly during the period from May 1986 to June 2000 using gill nets with randomized sections of different mesh size (10–75 mm). The total length, weight, sex and often the gut content of the fish were determined. The volumetric proportion of different food organisms in individual guts was estimated by visual inspection on a flat surface.
Table 1. Initial 137Cs deposition, catchment area, land cover and lake morphometric and chemical characteristics of the three studied lakes
Dorsal muscle tissue was sampled for 137Cs analysis. Usually, the activity measurement was performed on samples from single fish. However, samples from very small fish of the same species and size (within 1 cm of length) were sometimes pooled in order to improve detection. The activity of 137Cs was quantified using hyperpure Ge-detectors (ORTEC, PGT, www.ortec-online.com) or a Gamma Counting System (Intertechnique Model CG4000, www.eurisysmesures.com) equipped with a 3′′ NaI well detector. The equipment was calibrated regularly using blanks and dry standards of highly contaminated lichens collected soon after the Chernobyl fallout. All measurements were corrected for background radiation and sample geometry. Counting times were adjusted to ensure a coefficient of variation below 5% (usually ≈1%). In addition, all 137Cs measurements were corrected for physical decay back to 1 May 1986. All activity concentrations refer to muscle tissue (wet weight). The fish 137Cs data were categorized by species and for abundant species also by fish size (Table 2). The body length range of each size category was based on empirical and statistical criteria, such as maturation, diet shifts and available number of samples.
Table 2. Size and trophic level (TL) of the fish categories and the sample size, R2 and parameter estimates (± SE, except where parameters were fixed) of the pulse-response model fitted to time-series of 137Cs in fish. Italic type indicates ecological half-life corresponding to a non-significant decline rate
Steady-state levels (Csbase) and long-term ecological half-lives (T1/2) based on fish caught 1996–2000.
†
Early half-lives based on the first 3 years after tmax.
Ekholmssjön
Bream
8·8–14·4
12·0
2·56
7
0·90
90 ± 301
450 ± 94
114 ± 425
0·07 ± 1·13
2·37 ± 135
78 ± 1
3·3
146
14·9–46·8
216·3
2·63
86
0·69
145 ± 4
229 ± 13
93 ± 32
0·61 ± 0·12
13·88 ± 1·75
65 ± 6
8·3
8
Cr. carp
33·7–36·6
1008·4
3·00
6
1·00
148 ± 3
789 ± 37
266 ± 19
0·15 ± 0·05
3·65
117
3·9
−
Perch
6·0–9·6
5·9
3·01
30
0·68
184 ± 23
2110 ± 128
56 ± 61
0·05 ± 0·09
12·05 ± 5·84
95 ± 20
2·4
14
9·8–14·7
16·8
3·07
68
0·84
141 ± 13
1482 ± 40
87 ± 13
0·05 ± 0·02
3·62 ± 4·38
111 ± 10
1·8
−88
15·3–23·9
79·6
3·54
148
0·85
452 ± 53
2903 ± 108
193 ± 22
0·27 ± 0·28
12·78 ± 4·75
187 ± 30
2·7
12
24·0–41·0
322·1
3·96
103
0·86
744 ± 42
2389 ± 51
425 ± 28
0·90
13·51 ± 2·12
264 ± 16
6·6
22
Pike
29·5–109
742·5
4·40
30
0·61
153 ± 17
1798 ± 181
723 ± 76
0·95 ± 0·01
0·37
294 ± 13
7·1
−
Roach
6·8–11·0
7·3
2·81
40
0·24
118 ± 17
1079 ± 84
107 ± 20
0·17 ± 0·14
13·88 ± 7·31
55 ± 6
1·5
183
11·1–17·3
24·2
2·61
126
0·85
141 ± 12
787 ± 20
140 ± 13
0·17 ± 0·09
14·61 ± 4·02
62 ± 6
3·3
25
17·5–28·2
85·9
2·46
81
0·71
82 ± 3
547 ± 20
135 ± 26
0·05 ± 0·02
3·65
76 ± 9
2·6
41
Ruffe
5·6–13·9
8·5
3·00
24
0·64
97 ± 7
787 ± 106
150 ± 29
0·46 ± 0·55
8·40 ± 4·38
63 ± 11
2·8
15
Flatsjön
Bream
6·0–14·4
9·9
2·86
67
0·80
2127 ± 53
7074 ± 726
166 ± 22
0·90
13·88 ± 1·02
734 ± 44
4·4
88
14·8–50·0
199·5
2·89
208
0·74
1145 ± 18
3775 ± 96
235 ± 34
0·10 ± 0·03
3·65
869 ± 56
4·8
−258
Perch
5·2–9·7
5·2
3·01
199
0·50
1626 ± 41
14038 ± 930
66 ± 6
0·02 ± 0·00
0·00
1702 ± 152
5·5
−34
9·8–15·2
14·5
3·09
254
0·67
3389 ± 433
13209 ± 679
97 ± 44
0·03 ± 0·04
8·40 ± 3·00
1547 ± 162
7·2
137
15·3–23·9
91·6
3·53
444
0·55
6739 ± 266
22387 ± 448
262 ± 18
0·95
7·31 ± 1·02
3238 ± 556
3·7
15
24·0–39·1
278·6
3·86
227
0·71
7362 ± 86
18084 ± 457
499 ± 38
0·80 ± 0·01
3·65
4892 ± 404
7·6
23
Pike
17·5–99·5
434·4
4·10
269
0·80
2811 ± 151
10483 ± 167
690 ± 26
0·90 ± 0·00
1·13 ± 1·39
2654 ± 286
5·6
31
Roach
7·0–11·0
7·8
2·55
166
0·78
1684 ± 50
7295 ± 168
112 ± 12
0·20 ± 0·08
8·04 ± 0·91
860 ± 75
2·7
−488
11·1–17·4
23·8
2·47
304
0·83
2402 ± 45
6354 ± 119
166 ± 8
0·99 ± 0·00
10·23 ± 0·58
972 ± 69
4·5
71
17·5–30·2
90·2
2·57
197
0·79
2181 ± 100
8303 ± 336
156 ± 18
0·10
7·67 ± 1·28
1115 ± 84
4·2
30
Rudd
7·6–37·7
80·6
2·60
73
0·51
2243 ± 191
5901 ± 1253
154 ± 58
0·84 ± 0·28
14·98 ± 3·10
675 ± 147
5·0
−10
Ruffe
5·2–12·3
6·6
3·00
53
0·88
1091 ± 26
9160 ± 558
132 ± 54
0·04 ± 0·03
0·37
1155 ± 154
3·5
63
Siggeforasjön
Bleak
8·1–17·3
12·5
2·95
50
0·89
562 ± 103
8039 ± 366
164 ± 22
0·09 ± 0·05
5·84 ± 6·21
380 ± 32
2·1
73
Bream
5·3–14·5
6·4
2·98
23
0·83
424 ± 22
5074 ± 448
228 ± 37
0·15 ± 0·09
0·37
407 ± 26
2·2
879
14·8–14·8
141·3
2·93
295
0·86
370 ± 8
4294 ± 96
193 ± 8
0·05 ± 0·00
0·37
407 ± 32
3·5
20
Perch
5·5–9·8
5·8
3·01
61
0·91
1489 ± 178
15302 ± 560
169 ± 13
0·11 ± 0·05
12·42 ± 3·25
520 ± 72
1·9
22
9·8–15·2
17·9
3·06
130
0·79
1514 ± 103
11819 ± 383
227 ± 15
0·28 ± 0·13
8·40 ± 2·01
715 ± 92
2·1
14
15·3–23·6
73·0
3·42
249
0·80
2489 ± 119
12229 ± 384
258 ± 14
0·25
10·23 ± 1·35
937 ± 155
2·9
−239
24·0–44·8
305·3
3·92
84
0·59
2738 ± 89
13971 ± 1135
547 ± 78
0·95
3·65
2185 ± 392
20·1
−48
Pike
17·2–97·8
523·6
4·25
81
0·74
1040 ± 116
6634 ± 248
806 ± 63
0·82 ± 0·07
4·38 ± 3·58
787 ± 123
11·7
29
Roach
6·8–11·0
6·0
2·74
40
0·51
1292 ± 88
8969 ± 593
152 ± 12
0·92 ± 0·02
23·38 ± 4·38
261 ± 48
3·1
407
11·1–17·3
24·5
2·65
138
0·91
742 ± 37
6785 ± 279
158 ± 9
0·10
11·69 ± 1·75
296 ± 24
3·2
52
17·5–29·2
84·8
2·57
171
0·80
431 ± 72
4156 ± 125
224 ± 30
0·08 ± 0·04
3·07 ± 4·75
358 ± 31
2·3
29
Rudd
8·0–30·3
71·2
2·38
85
0·61
1003 ± 46
5012 ± 501
156 ± 9
0·80 ± 0·01
9·86 ± 2·34
523 ± 133
3·7
–12
Ruffe
5·7–15·7
8·4
3·00
52
0·76
620 ± 54
6291 ± 335
115 ± 23
0·02 ± 0·01
7·31
416 ± 65
3·7
17
In order to compare formally the time-series of different fish categories, the 137Cs temporal pattern was parameterized using a model (see Supplementary material) with the following basic formulation:
(eqn 1)
where Csfish is the 137Cs activity concentration in fish (Bq kg−1 fw) and t the time elapsed since fallout (days). On a log scale, equation 1 depicts an asymmetric pulse function slowly approaching an asymptote with an intercept p4 and slope p5, given that all five parameters are positive and p5 << p2, p3. Equation 1 is an analytical solution of a dynamic bioaccumulation model, assuming a two-component decline function of 137Cs in the lake water (see Supplementary material).
We identified four curve shape features or parameters of radioecological relevance: the maximum activity concentration (Csmax), the time to reach maximum activity concentration (tmax), the new steady-state level (Csbase) and the long-term decline rate (λ), corresponding to the long-term ecological half-life (decay-corrected) of 137Cs (Fig. 1). The basic model (equation 1) can be expressed directly as a function of these key parameters by reformulation (see Supplementary material):
Illustration of the key parameters of the pulse-response model (equation 2) fitted to decay-corrected 137Cs activity concentrations in pike from Lake Flatsjön, Sweden, after the Chernobyl fallout. Csmax is the peak concentration that was attained tmaxdays after the Chernobyl fallout in late April 1986; Csbase is the intercept of the slow decline associated with the long-term ecological half-life T1/2.
( eqn 2)
where:
The parameter R, defined as the ratio of the ‘incline’ and ‘decline’ rates (p2 and p3 in equation 1), determines the shape of the curve, in particular the decline rate after the peak. R ranges from 0 to 1, the estimation being influenced by the other parameters, especially by tmax. For a given tmax, a high R yields a slow initial decline.
Equation 2, henceforth referred to as the pulse-response model, was log-transformed and fitted to time-series of log-transformed 137Cs activity concentrations in the different lakes and fish categories. Parameters and their approximate (asymptotic) standard errors were estimated by non-linear regression, i.e. by minimizing the residual sum of squares using an iterative method (Newton-Raphson). Least-squares were weighted by the number of fish in each sample to account for the assumed smaller variance of pooled samples. The pulse-response model could be fitted to the time-series of 137Cs for most fish categories. For some fish categories with few or variable data, however, parameters R or λ were fixed at an adequate value (after visual fitting of the model to data) to facilitate iteration convergence. As the function of slow decline represents a nearly horizontal line (on a log scale), small changes in λ only marginally affect the residual sum of squares. However, small changes in λ can cause considerable changes in the Csbaseparameter, the intercept of the long-term decline. Instead of Csbase from the pulse-response model we therefore used a more conservative estimate of the near steady-state levels (denoted Csbase hereafter), defined as the mean 137Cs activity concentrations observed during the period 1996–2000. This period had the highest sampling frequency and steady-state among fish appeared to have been reached. Ecological half-lives (T1/2) were not only calculated from λ but also from the regression slope between log 137Cs and time during different periods. Early T1/2 was based on the first 3 years after tmax and long-term T1/2 on the period 1996–2000. In order to illustrate changes over time, T1/2was estimated for running 4-year intervals from May 1986 to May 2000. These half-lives were calculated from the mean decline rate of all fish categories of each lake, assuming exponential decline during each interval.
The diet, based on semi-quantitative gut analyses of a subsample of fish from all three lakes (total n= 2109), was grouped into four categories: plants, zooplankton, zoobenthos and fish. We used a multi-category logistic regression model (Agresti 1996; SAS Institute Inc. 2000) to determine the volumetric proportion of the different dietary components as a function of fish length for each species and lake. The trophic level (TL) for each fish was crudely calculated as:
(eqn 3)
where P(i) is the proportion of dietary component i, and D(i) the trophic level of i. For plants D was given the value 1, for zooplankton and zoobenthos 2, and for fish diet Dwas 3 or more, depending on species and size of the prey. The trophic level of each fish category and each lake was then estimated as the mean of the trophic level of all individual fish. Hence, the calculated trophic levels are regarded as a mean for the whole period, neglecting seasonal and interannual changes.
The key parameter estimates of each fish group were compared among the lakes using analysis of covariance (ancova) with either trophic level, log weight or the dietary proportion of zooplankton as covariates. All statistical analyses were aided by JMP statistical software (SAS Institute Inc. 2000). Variation in the relationship between fish 137Cs and body size over time was studied for bream, roach, perch and pike. The linear regression slope between log 137Cs and total length was calculated for each sampling date.
Results
The pulse-response model (equation 2) provided a good fit to the temporal trend in 137Cs activity concentrations for most fish categories, as exemplified for pike in Lake Flatsjön (Fig. 1). R2-values typically ranged between 0·5 and 0·9 with an average of 0·74 (Table 2). For some fish categories, data were scarce before the peak, resulting in large standard errors of the parameter estimates, especially of tmax. The standard errors for the fitted parameters were typically below 10% for Csmax and Csbase, around 20% for tmax, but often higher for R and λ (Table 2).
trophic level
Maximum 137Cs levels were reached between 56 and 806 days after fallout (Table 2) and delayed significantly by fish trophic level (Fig. 2a; ancova, R2 = 0·64, F5,31 = 11·1, P< 0·0001), corresponding to a 2·5-fold increase in tmax per trophic level. There were no differences in tmax among the three lakes, either regarding the mean levels (P = 0·095) or the interaction between lake and trophic level, the effect testing for homogeneity of the covariate slopes (P = 0·97).
Properties of the 137Cs trend in different fish categories in three Swedish lakes after the Chernobyl fallout in 1986: (a–d) relationships between pulse-response model parameters (see Fig. 1) and mean trophic level (TL); (e,f) relationship between relative pulse amplitude (Csmax:Csbase) and the proportion of zoobenthos or zooplankton (P(Z)) in the diet. In (b) and (c) perch size classes are highlighted by larger symbols to indicate the tendency for higher 137Cs accumulation in this species. The parameters in the regression equations were estimated by ancova; the arbitrary coefficient a applies where intercepts differ significantly among lakes; NS = slopes not significant.
Trophic level had a small but significant effect on Csmax (Fig. 2b). Peak concentrations were highest at intermediate trophic levels, which was confirmed by ancova with trophic level as a second-degree polynomial covariate (R2 = 0·90, F8,28 = 30, P < 0·0001). Both the linear and quadratic effects were significant (P < 0·0001 and P = 0·0086) while the interaction between lake and trophic level was not (linear: P= 0·15; quadratic: P = 0·76). Lake Ekholmssjön had significantly lower Csmax than the other lakes (F1,28 = 92, P< 0·0001), which did not differ in Csmax (F1,28 = 0·65, P= 0·43).
The steady-state activity concentrations (Csbase) increased significantly by a factor of 1·8 per trophic level (Fig. 2c; ancova, R2 = 0·94, F5,31 = 92, P < 0·0001). Mean Csbase levels (P < 0·001) but not the slopes (P = 0·68) differed significantly among the lakes. Csbasewas highest for Lake Flatsjön, followed by Lake Siggeforasjön and Lake Ekholmssjön. Some of the within-lake variation resulted from species-specific differences. In general, both peak and base 137Cs levels were significantly higher for perch than for the other species (indicated in Fig. 2b,c). Small non-piscivorous perch had considerably higher levels than other fish on comparable trophic positions. Similarly, piscivorous perch had higher base levels than pike, except in Lake Ekholmssjön where the average pike were much larger than in the other lakes.
separating effects of trophic level and body size
As expected, fish weight and trophic level were correlated (r = 0·53, n= 37), making it difficult to separate their individual influences on 137Cs temporal patterns. However, the correlation was only pronounced for piscivorous fish, i.e. pike and large perch (r = 0·98, n= 9). Considering only the non-piscivorous fish, trophic level and weight were less correlated (r = −0·32, n= 28). For this reason, only non-piscivorous fish were used for a series of multivariate statistical analyses. The P-values below refer to F-tests in ancovas, with log weight and trophic level of non-piscivorous fish as linear covariates.
The parameter tmax increased significantly with weight (Fig. 3a; F1,23 = 11·2, P= 0·007) but not with trophic level (Fig. 3b; F1,23 = 0·02, P= 0·89). The relationship between tmaxand weight did not differ among lakes (F2,21 = 0·06, P= 0·44). Peak concentrations Csmaxdecreased with weight (Fig. 3c; F1,23 = 8·60, P= 0·0075) but increased with trophic level (Fig. 3d; F1,23 = 6·61, P= 0·017); there was no difference in slope among lakes for either weight (F2,19 = 0·08, P= 0·95) or trophic level (F2,19 = 1·83, P= 0·39). There was a weak positive effect of trophic level on Csbase (Fig. 3f; F1,23 = 7·37, P= 0·012) but no across-species effect of body weight (Fig. 3e; F1,23 = 0·93, P= 0·34). However, the 137Cs levels increased with size within some species (perch, roach and pike), although not during the first years after fallout (Fig. 4). The relative change in 137Cs concentration per cm of body length, i.e. the regression slope between log 137Cs and length, was initially negative for perch, then peaked during the first years, and finally levelled off as positive. The slope value of around 0·25 implies that typical mean levels in the largest perch were more than six times higher than in the smallest perch. The same pattern, but less pronounced, appeared for roach and pike, while bream had a slope near zero during the entire period, i.e. similar 137Cs activity concentrations irrespective of body size (Fig. 4).
Properties of the 137Cs trend in different fish categories in three Swedish lakes after the Chernobyl fallout in 1986: comparison of the effects of mean body weight (W) and trophic level (TL) of non-piscivorous fish on the pulse-response model parameters. See also te legend to Fig. 2.
Temporal variation (1986–2000) in the relationship between log 137Cs activity concentration and total length in four species of fish in three Swedish lakes. Linear regression slopes were calculated for each sampling occasion in each of the lakes. The curves are fitted spline functions illustrating the general trends.
pulse relative amplitude
Lake-specific differences in the temporal pattern, independent of fallout deposition density, can be analysed by comparing the ratio between Csmax and Csbase, i.e. the relative amplitude of the 137Cs pulse response in fish. For non-piscivorous fish Csmax:Csbase decreased with body size (Fig. 3g; ancova, R2 = 0·67, F4,23 = 11·1, P< 0·0001) but was not significantly related with trophic level (Fig. 3h; P= 0·51). The relationship with size did not differ among the lakes (P = 0·49). However, the lakes differed significantly in Csmax:Csbase (P = 0·0004). Lake Flatsjön had the lowest and least variable ratios: peak levels were on average 6·7 times higher than steady-state levels, compared with Ekholmssjön and Siggeforasjön, with mean ratios of 10 and 14, respectively. The pulse amplitude increased with the degree of zooplanktivory, measured as the mean proportion of zooplankton in the diet of the different fish categories (Fig. 2f; F1,33 = 16·5, P= 0·0005). There was no significant relationship between the degree of zoobenthivory and Csmax:Csbase (Fig. 2e; F1,33 = 0·05, P= 0·83). The calculations were performed on arcsine-transformed proportions including all fish categories. Zoobenthivory had no significant effect on the steady-state levels (F1,33 = 0·05, P= 0·83).
ecological half-life
Apparent ecological half-lives varied widely over the period 1986–2000. Early T1/2, 3 years after tmax, ranged between 1·5 and 7·5 years (significant T1/2 in Table 2) among fish categories, increasing with fish weight (ancova, R2 = 0·56, F4,31 = 9·78, P < 0·0001) but not with trophic level (P = 0·18). There was no significant difference in slope among lakes (P = 0·12). The least-squares mean estimate of early T1/2 among fish in Lake Flatsjön was 4·5 years, which is significantly (P = 0·0007) longer than the mean T1/2 of the other lakes (2·9 years). Short-term T1/2 based on 4-year intervals increased over the studied period (Fig. 5).
Changes over time in apparent ecological half-lives of 137Cs in fish from three Swedish lakes. The half-lives were calculated from the mean decline rates of all fish categories of each lake, assuming exponential decline during each interval (typically 4 years).
In contrast to the early T1/2, the long-term T1/2 (during 1996–2000) tended to decrease with trophic level (Fig. 2d) and weight in all lakes, although the slopes were not significant (ancova, R2 = 0·24, F4,22 = 1·77, P = 0·17). Median long-term T1/2 was shortest in Lake Ekholmssjön (24 years), followed by Lake Siggeforasjön (52 years) and Lake Flatsjön (71 years). In comparison, median T1/2 calculated from estimates of λ ranked similarly among the lakes, but were generally longer: 54, 70 and 85 years. However, neither the ancova above nor a median-test (n = 36, P = 0·23) revealed any significant differences in long-term T1/2 among lakes. Estimates of T1/2 varied widely due to the low decline rate, large individual variation and, for some species, few observations. For some fish categories the 137Cs levels even tended to increase after 1996 (the negative ‘half-lives’ in Table 2).
In order to get a robust estimate of the long-term T1/2, given our data, we pooled all perch and roach (the most abundant and most frequently caught species) in all lakes from the period 1996–2000. We used ancova (R2 = 0·85, F7,1478 = 1145, P < 0·0001) to estimate the rate of 137Cs decline in these fish, which resulted in an ecological half-life of 46 years, corresponding to an effective half-life of 18 years (including physical decay).
Discussion
The pulse-response model (equation 2) adequately described the contamination of fish over time. In other words, most of the variation in 137Cs among years, species and size classes could be accounted for by simple analytical functions. Koulikov & Ryabov (1992) and Smith et al. (2002) successfully fitted dynamic models to time-series of Chernobyl 137Cs in fish. Their models have similar mathematical properties to the model presented here. The successful fit of such a model to the much longer time-series available in this study indicates that three exponential functions and five parameters are sufficient for describing pulsed 137Cs in different fish, at least during the first two decades after a fallout event. The comparatively low R2-values (mean R2 = 0·74), as also observed in Smith et al.'s (2002) study (mean R2 = 0·53), reflect a large variation among individual fish rather than a poor fit. Individual measurements were typically within a factor of two on either side of fitted model values (90% confidence limits).
differences among fish
Fish trophic level had a profound effect on the 137Cs dynamics. Fish at higher trophic levels reached peak levels later and levelled off at higher steady-state levels. A delayed accumulation at higher trophic levels was expected, as trophic uptake dominates in freshwater fish (Morgan, Tytler & Bell 1994). However, we found no effect of trophic level on tmax for non-piscivorous fish. This may partly be explained by the low resolution in assigning trophic levels to the invertebrate food categories. A more plausible explanation is that low trophic levels in freshwater are occupied by small organisms (periphyton and invertebrates) that rapidly equilibrate with ambient 137Cs concentrations (Hübel & Sänger 1993), irrespective of trophic level. The impact on the tmax of their consumers is thus expected to be very small. In contrast to trophic level, increased body size delayed tmax for non-piscivorous fish, whereas Csmax decreased with size. Such a pattern was predicted by Koulikov & Ryabov (1992). They attributed this to slower 137Cs excretion rates in larger fish and the rapid decrease in dissolved 137Cs during the first years.
The pulse amplitude (Csmax:Csbase) decreased with fish size, which is partly a result of the negative relationship between Csmax and body size. The pattern was probably accentuated by zooplanktivores, as the degree of zooplanktivory showed a positive correlation with Csmax:Csbase. Crustacean zooplankton would have been among the dietary organisms most rapidly affected by the Chernobyl pulse, before substantial dilution and immobilization of 137Cs had occurred in the water column. Potentially high Csmax in zooplanktivores would result in bleak, small roach and small perch, as we often observed (Table 2). It has been hypothesized that 137Cs activity concentrations can be higher in benthivorous fish than in planktivorous fish due to their interaction with the highly contaminated surface sediments. In contrast to Rowan, Chant & Rasmussen (1998), we found no relationship between the steady-state levels and benthivory. Most sediment 137Cs is bound to indigestible clay minerals with high affinity for caesium. If particle-bound 137Cs is in equilibrium with the dissolved phase, an increased accumulation by fish or their benthic prey due to sediment ingestion would not be expected. Hypothetically, temporary anoxic periods could disturb the equilibrium and increase the bioavailable fraction. However, lakes in this study rarely develop anoxia in the upper sediment layer. Perch of all size categories had higher 137Cs levels than their counterparts at similar trophic levels, which has also been found in previous studies (Elliott et al. 1992; Särkkä, Jämsä & Luukko 1995; Koulikov 1996). The mechanism is not yet understood, but Elliott et al. (1992) suggested that the low gastric evacuation rate in perch might lead to a higher assimilation efficiency of radiocaesium.
Higher steady-state 137Cs levels at higher trophic levels suggest the importance of biomagnification. Direct uptake of 137Cs, compared with trophic uptake, is less important in fish than in most invertebrates (King 1964), which may explain why the biomagnification was weaker among non-piscivorous fish than in piscivores (cf. slopes in Figs 2c and 3f). Biomagnification in aquatic food chains is rare among trace elements, and whether Cs, like mercury (Power et al. 2002), is an exception has been disputed in the past (Rowan & Rasmussen 1994). However, our data and other recent studies (Kryshev 1995; Koulikov 1996; Särkkä, Keskitalo & Luukko 1996; Rowan, Chant & Rasmussen 1998) provide further evidence that 137Cs undergoes biomagnification, at least within fish communities. Biomagnification could not be observed for the peak levels, probably because the 137Cs accumulation in fish lagged behind the rapidly declining 137Cs concentrations in the water and prey during that period.
The intraspecific relationship between 137Cs and fish size changed markedly over time (Fig. 4). The shift from negative to positive regression slopes would be expected from the increasing delay of peak concentrations with increasing size and trophic level. Ugedal et al. (1995) found a similar pattern during 1986–89 in brown trout Salmo trutta and were later able to predict this shift using a mechanistic model where growth, 137Cs intake and biological half-life were related to fish weight by empirical power functions (Ugedal, Forseth & Jonsson 1997). However, the variable slope-shift patterns observed here indicate that weight-specific bioenergetic relationships differ among species. The slope at steady-state depends on how the ratio between 137Cs assimilation rate and the sum of 137Cs elimination and growth rates changes with age (Rowan, Chant & Rasmussen 1998). A positive slope is expected within species as growth and elimination rates generally decrease with fish size. However, bioenergetic changes due to ageing or maturation can occur within different size ranges in different species and locations (Rowan, Chant & Rasmussen 1998), which may explain why there was no effect of fish size on the steady-state levels when comparing across species (Fig. 3e). Ontogenetic shifts in trophic level would add to the bioenergetic size effect, which may explain the steeper 137Cs–size slope in perch than in the other species (Fig. 4). Apparently, accurate modelling of the commonly observed within-species ‘size effect’ requires species-specific information on ontogenetic changes in bioenergetics and diet.
differences among lakes
The near-steady-state level of 137Cs, as well as the relative pulse amplitude, differed among our study lakes (Fig. 2c,e). The most obvious explanation is that the lakes received different densities of radioactive fallout. However, fallout patterns alone cannot explain the differences. The aggregated transfer factors for the near steady-state levels in roach (Csbase divided by the fallout, m2 kg−1 fw) were 0·0025 in Lake Ekholmssjön, 0·009 in Siggeforasjön and 0·014 in Flatsjön, suggesting that the transfer of 137Cs to fish in Siggeforasjön, and particularly in Flatsjön, was more efficient than in Ekholmssjön. Part of this difference may be attributed to uncertainties in the estimates of initial fallout density, which is based here on an early air-borne survey with low spatial resolution. Other possible explanations are related to environmental factors that differ among the lakes. The dissolved potassium concentration is often negatively correlated with the 137Cs bioaccumulation factor in fish (Rowan & Rasmussen 1994; Smith et al. 2000c). Given the similar catchment properties of the three lakes (Table 1), we would also expect the transfer factors to be negatively correlated with potassium concentration. However, potassium does not explain the difference among the lakes, as its concentration was highest in Lake Flatsjön (Table 1), the lake with the highest transfer factor. Furthermore, 137Cs in fish can increase with water residence time, at least during the first year after a fallout (Håkanson, Andersson & Nilsson 1992). However, residence times are similar for the three lakes (Table 1). We suggest that lake morphometry and associated sediment dynamics explain some of the differences in the 137Cs transfer factor. This conclusion is supported by the observed differences in relative pulse amplitude: the pulse amplitude in roach was lower in Lake Flatsjön (seven- to ninefold Csbase) than in Ekholmssjön (seven- to 20-fold Csbase), which in turn was below the amplitude of Siggeforasjön (12- to 35-fold Csbase). The amplitudes correspond best with differences in lake depth (Table 1). However, the trend is opposite to expectations based on the initial dilution of 137Cs deposited on the lake, i.e. higher pulse amplitude in shallow lakes. An alternative explanation is linked to the abundance of particles controlling the fate of 137Cs. Sediment surface per water volume as well as sediment dynamics (resuspension frequency) are highest in shallow lakes. Resuspended sediments may reduce both the incline and the decline of 137Cs in fish (Broberg, Malmgren & Jansson 1995; Meili, Braf & Konitzer 1997), thus acting as an ecological buffer: early after fallout, sorption of 137Cs to suspended particles and subsequent removal by sedimentation competes with bioaccumulation by controlling the initial decline of dissolved 137Cs, whereas at a later stage the dynamics of 137Cs transfer from the sediment surface back into the water column by resuspension and diffusion may contribute to the elevated long-term 137Cs levels in biota. This would result in less pronounced but more extended pulse responses of 137Cs in shallow lake ecosystems, as was observed in Lake Flatsjön. The variability in relative pulse amplitude among fish categories within lakes was about three- to sevenfold. The pulse amplitude varied most for pelagic fish (see also Fig. 2e,f), which supports the buffering role of sediments rather than an influence of water renewal. On the other hand, it can be noted that the overall relative pulse amplitude of biotic 137Cs levels only varied about twofold among lakes, suggesting that the observed pulse patterns may be a fairly general phenomenon, and that the pulse-response model developed here may be generally applicable with limited adjustment of parameters.
ecological half-life
Apparent ecological half-lives were not constant but increased gradually after tmax(Fig. 5). Thus, a single exponential decline function (one-compartment approach) was inadequate to describe the long-term decline of 137Cs in fish, which may explain why early post-Chernobyl studies often over-predicted the subsequent recovery rate of contaminated lakes (Håkanson & Andersson 1992). Two- or three-component decline functions can serve as more adequate approximations (the model presented here; Saxén, Jaakkola & Rantavaara 1996; Jonsson, Forseth & Ugedal 1999; Smith et al. 2002). An alternative modelling approach for the period after tmax would be to use a one-component function but allow the decline rate to vary with time. By assuming a monotonous decline and fitting a trend line to the data in Fig. 5, we estimated an annual increment of T1/2 by 23% (T1/2 = T0e0·205t, R2 = 0·77, here T0 is 4·6 but can represent the ‘initial’ half-life at any reference time t). The observed increase in T1/2 indicated substantial changes over time in processes controlling the recovery of aquatic ecosystems: early half-lives varied among fish and lakes, whereas after about one decade no significant differences were found. Apparently, decline rates were initially controlled by in-lake processes but gradually became uncoupled from food web and water–sediment interactions, as steady-state was approached. Eventually the decline rates would be governed by catchment runoff, which is ultimately controlled by ion exchange processes in soil (Hilton et al. 1993; Kudelsky et al. 1996; Smith et al. 1999b). The temporal trends in apparent ecological half-lives were similar in the different lakes, as were the lakes’ catchment size and land cover (Table 1). An obvious question is whether this pattern can be generalized to lakes with different environmental characteristics, e.g. catchment with clay-rich soils. This would be consistent with a hypothesis stating that even very small amounts of illite clay in organic soils will dominate the adsorption kinetics and retention capacity of radiocaesium (Hird, Rimmer & Livens 1995; Staunton & Levacic 1999). In accordance, Smith, Comans & Elder (1999a) and Smith, Clarke & Saxén (2000a) found that the 137Cs input from the catchment had only little influence on the decline rates of dissolved 137Cs in lakes and rivers.
The parameter λ in the pulse-response model represents the future decline rate predicted by using data from the entire 14-year monitoring period. An overall estimate of T1/2 based on λ was 69 or 111 years (median or mean of all fish and lakes in Table 2), which is longer than the overall T1/2 based on pooled data from the period 1996–2000 alone (46 years). This indicates that half-lives may continue to increase, in which case physical decay (T1/2 of 30 years) will dominate the recovery rate even more beyond the first decade after fallout. Other recent studies in various types of lakes (Jonsson, Forseth & Ugedal 1999; Smith, Clarke & Saxén 2000a; Håkanson & Sazykina 2001) also report a long-term ecological half-life of the same order of magnitude as the physical decay rate. On the other hand, Hessen et al. (2000) found an average half-life of 2·5-years in Norwegian brown trout, and in North America the half-lives of contaminated streams and ponds (non-Chernobyl 137Cs) ranged between 3·5 and 16·7 years (Paller, Littrell & Peters 1999; Peles et al. 2000). This variation may be explained by short or otherwise limited data sets in combination with assumptions of simple exponential decline. The slow decline in combination with high individual variation of 137Cs in fish necessitates the use of long time-series and large sample sizes in robust model development for long-term prediction. This is illustrated by our study, where confident estimates of the long-term ecological half-life were difficult to obtain (Table 2), despite this being one of the longest and most intensive monitoring records of fish 137Cs available.
applications
We used a pulse-response model (equation 2) as a tool for analysing large amounts of field data in a formalized way to facilitate comparisons in contaminant behaviour across and within lakes. By adequate parameterization, it was possible to quantify and differentiate fundamental influences of ecological structure and dynamics on contaminant levels in lakes, including changes over time. Further, our results indicate that simple analytical equations without mass balance considerations or mechanistic assumptions can facilitate strictly predictive modelling of pulsed contaminants, even when addressing a diversity of target organisms. For this purpose, levels and dynamics of 137Cs concentrations in fish can be described as simple functions of body size and trophic level. Three aspects deserve particular consideration. First, the influence of size and trophic level differed among the individual model parameters. For optimal predictive modelling, size and trophic level should thus be treated as independent variables. Secondly, the influence of size and trophic level was not uniform over time: Csmaxdecreased with fish size while Csbase was indifferent or increased with size for some fish categories, and trophic level had a stronger effect on Csbase than on Csmax. Smith et al. (2002) related fish weight to 137Cs by a power function (Wn), which improved the fit of their original model to observed data. However, an assumed constancy of size influence over time appears inadequate, in particular during the initial dynamic period. Thirdly, the similarity of slopes suggested that the influence of fish size and trophic level on various model parameters was consistent among the lakes. The time scale and general shape of a radiocaesium pulse in a fish community is thus a predictable function of the structure and dynamics of populations and food webs, and may in principle be assessed without information on dissolved 137Cs concentrations, which are rarely available. This simplifies the development of a general model for the bioaccumulation of fallout 137Cs and probably of pulsed contaminants in general. For example, our pulse-response model can be modified to include fish trophic level and size by substituting some model parameters for the regression formulae in Figs 2 and 3.
Whereas the species and trophic level of the fish influenced the timing of both the pulse maximum and the early decline (i.e. the pulse duration), lake-specific abiotic factors appeared to be important for the maximum levels by influencing the overall pulse amplitude. Based on the limited number of lakes, we were not able to quantify the influence of different abiotic lake characteristics on Csmax or Csbase, which may include aspects of water chemistry, lake morphometry, sediment geochemistry, catchment properties and climatic conditions. However, empirical relationships have been established from cross-lake comparisons of 137Cs levels for dynamic (Håkanson, Andersson & Nilsson 1992) as well as steady-state conditions (Rowan & Rasmussen 1994). By using such relationships to substitute the two scaling parameters Csmax or Csbase, our one-equation pulse-response model can be adopted for research or management purposes to provide dynamic predictions based on limited ecological knowledge and without need of any advanced numerical simulation tools. This is particularly useful when addressing the fate of a contaminant, nutrient or tracer in a variety of aquatic organisms and habitats simultaneously.
Acknowledgements
We thank Åsa Sjöblom, Sebastian Sobek and Susanna Vesterberg for field and laboratory assistance, and Peter Eklöv and Jim Smith for comments on the manuscript. This study was supported financially by the Swedish Radiation Protection Authority (SSI) and Nordic Nuclear Safety Research (NKS).
(eqn 1)
( eqn 2)
(eqn 3)
No comments:
Post a Comment