The influence of biomanipulations (fish removal) on the structure of lake ... The work is based on a comprehensive dynamic lake ecosystem model, LakeWeb, ...
87
Aquatic Ecology 37: 87–99, 2003. © 2003 Kluwer Academic Publishers. Printed in the Netherlands.
The influence of biomanipulations (fish removal) on the structure of lake foodwebs, case studies using the LakeWeb-model Lars Håkanson 1,*, Viktor V. Boulion 2 and Alexander P. Ostapenia 3 1
Department of Earth Sciences, Uppsala University, Villav. 16, Uppsala, 752 36, Sweden; 2Zoological Institute of RAS, Universitskaja emb., 1, St. Petersburg, 199034, Russia; 3Laboratory of Hydroecology, Belarus State University, F. Skorina av. 4, Minsk, 220050, Belarus; *Author for correspondence Received 30 October 2001; accepted in revised form 20 September 2002
Key words: Benthos, Biomanipulation, Environmental factors, Fish removal, Fish, Functional organisms, Lakes, Macrophytes, Models, Phytoplankton, Zooplankton Abstract This paper presents results on how intensive fishing (fish removal) is likely to influence the structure of lake foodwebs. The work is based on a comprehensive dynamic lake ecosystem model, LakeWeb, which accounts for production, biomasses, predation and abiotic/biotic interactions of nine key functional groups of organisms: phytoplankton, bacterioplankton, two types of zooplankton (herbivorous and predatory), two types of fish (prey and predatory), zoobenthos, macrophytes and benthic algae. The model uses ordinary differential equations, the ecosystem scale and gives seasonal variations (the calculation time is 1 week). It is designed to account for all fundamental abiotic/biotic interactions and feedbacks for lakes in general for the nine target groups. The LakeWeb-model has been calibrated and critically tested using empirical data and regressions based on data from many lakes. It has been shown that the model can closely capture typical functional and structural patterns in lakes, which should give credibility to the results presented in this work. Obtaining such results using traditional methods, i.e., extensive field studies in one or a few lakes, would be very demanding (in terms of money, persons involved and time). In this paper, results are presented for two lakes, one Swedish and one Belarussian. The intensive fishing operations carried out in Lake Blacksåstjärn, Sweden, to reduce Hg-concentrations in fish did not succeed. A typical cost of an intensive fishing is about 10,000–30,000 USD per lake of this size ( ⭐ 0.25 km 2). The costs to remove fish would be about 40–120 USD per kg ww fish removed! Intensive fishing simulated for Lake Batorino, Belarus, to reduce the fish biomass will likely increase the prey fish biomass as long as the predation pressure on prey fish is lower than during the prefishing stage. The biomass of predatory fish will recover only slowly. However, this operation is not likely to succeed in lowering the algal volume in lakes with a high biomass of predatory zooplankton. This is easy to state qualitatively and the LakeWeb-model offers a practically useful tool to quantify such changes and identify lakes where biomanipulations are likely to fail or succeed. Introduction and aim of the work The literature on lake biomanipulation is extensive and the focus has generally been on manipulating fish or zooplankton [Hrbácek (1958) (the founder of biomanipulation concept); Shapiro and Wright (1984) and McQueen et al. (1986), Benndorf (1990, 1992), Kitchell and Carpenter (1988), Shapiro (1990), Theiss et al. (1990), Moss et al. (1991), Hanson and Butler
(1994)]. The basic idea with many biomanipulation operations (here called the “classical approach”) is to change the structure of the lake foodweb to restore the lake from eutrophic conditions, with a high primary production, to more oligotrophic conditions so that, e.g., the water clarity increases and more desirable fish species can increase in number and biomass. This is schematically illustrated in Figure 1a. Moreover, Hrbácek (1958) has tested to introduce predatory fish
88
Figure 1. A. “Classical” biomanipulation scenario. Its primary goal is to reduced phytoplankton biomass (hence also increase Secchi depth). The action is to remove the overall fish biomass (both prey and predatory fish). The main requirement for success is to increase the biomass of herbivorous zooplankton, and the grazing pressure on phytoplankton. The success of this action depends on several factors related to the lake foodweb, the role of bacterioplankton (the grazing of Zooherb. on Bact. pl. should be low), predatory zooplankton (the grazing of Zoopred. on Zooherb. should be low) and zoobenthos (the grazing of prey fish. on Zoobent. should be low), as indicated in the figure.; B. “Biomanipulation – dilution” scenario. Its primary aim is to reduced the concentration of toxic substances (e.g., mercury) in fish consumed by man. The action is to selectively remove predatory fish by means of intensive fishing. The main requirement for success is to increase the biomass of prey fish so that the concentration of X (defined by the ratio between the amount of X in prey fish and the biomss of prey fish) is reduced. If the biomass of predatory fish is reduced, the grazing pressure on prey fish will be reduced, and the biomass of prey fish will increase. This means lower concentrations of X in prey fish, and hence also in due course lower concentrations of X in predatory fish consuming prey fish. Also the success of this action depends on several factors related to the lake foodweb, as indicated in the figure.
into systems to reduce risks of algal blooming in drinking reservoirs of Czechoslovakia (see also Kitchell and Carpenter (1988)). Removal of prey fish should lead to an increased biomass of herbivorous zooplankton via a lower predation pressure, and hence an increased grazing pressure on phytoplankton. Therefore, the algal volume will be reduced, preferably well below operational target limits, such as the “critical” limit of 5 mm 3/l (see Håkanson and Boulion (2002)) or the “alarm” limit of 10 mm 3/l. This theory will only work under certain conditions, e.g.: (1) predation pressure on Zooherb. from Zoopred. is small, (2) Zooherb. actually feeds on phytoplankton and not too much on bacterioplankton, and (3) prey fish biomass is kept low; this can only happen if (4) the biomass of predatory fish is high; however, predatory fish are generally also removed in the fishing operations. Consequently, there are many potential problems with this
operation. The lake foodweb model (LakeWeb; see Håkanson and Boulion (2002)) used here is intended to be a tool to address such problems, and to identify whether certain scenarios and foodweb combinations are likely to succeed or fail. This is especially important since the interactions within the aquatic community are very complex. Biomanipulations can also be used, and have been used (see Håkanson and Peters (1995)), for the opposite reason, i.e., to increase primary production, so that the concentration of toxic substances in fish eaten by man is reduced (here called the “biomanipulationdilution” approach). This is illustrated in Figure 1b and will be discussed in an another scenario. Basically, the intention is to reduce predatory fish, so that the predation pressure on prey fish is lowered and the biomass of prey fish can increase, leading to more predation on Zooherb., a reduction in Zooherb. biomass and more phytoplankton, so that a given con-
89 taminant load is diluted in a greater biomass. This is often referred to as “biological” dilution (see, e.g., Jansson et al. (1981) and Håkanson and Peters (1995)). However, this theory will only work under certain conditions (see Figure 1b) and the LakeWebmodel can be used to test when this operation is likely to work as intended. The LakeWeb-model has been tested along many types of limnological gradients and against empirical data and models and those tests have demonstrated that the model can capture fundamental lake foodweb interactions and abiotic/biotic interactions very accurately (see Håkanson and Boulion (2002)). The fact that the LakeWeb-model has been critically tested and produced good results should lend credibility and realism to the scenarios discussed in this work. One must note that it would be very demanding indeed (in terms of money, persons and time) to do this kind of study in the traditional way by extensive studies in one or a few lakes. The aim here is to use data from two lakes and demonstrate using these as case studies how the LakeWeb-model can be applied to address important management issues as a tool to get realistic expectations (both positive and negative) of the outcome.
Materials and methods The structure of the LakeWeb-model has been presented elsewhere (Håkanson and Boulion 2002) and will not be repeated here. Figure 2 shows the general outline of the model. The primary aim of the model is not to give good predictions for certain species or specific lakes, but rather to quantitatively describe typical, characteristic lake foodweb interactions. This way production, biomasses and predation can be determined for the nine functional groups of organisms included in the model, the three primary producers: phytoplankton, benthic algae and macrophytes, the five secondary producers (consumers of different orders), herbivorous zooplankton, predatory zooplankton, zoobenthos, prey fish and predatory fish, and one decomposer (saprophyte or reducer), bacterioplankton. These are key functional groups in the sense that predatory fish do the “job” of eating prey fish. Other groups of organisms, such as benthic bacteria and fungi, are not treated but are accounted for in the sense that they are included in the flux to zoobenthos called “zoobenthos production from other sediment sources” (see Figure 2). Prey fish is the most com-
plex group of all in the LakeWeb-model. Prey fish feed on three other groups, zoobenthos, herbivorous zooplankton and predatory zooplankton. So, prey fish includes benthivores and planktivores. To gain simplicity, neither prey fish nor predatory fish are “permitted” to display cannibalism in the LakeWebmodel, i.e., feeding on their own group (like roe and/or young-of-the-year). Cannibalism exists in natural aquatic systems among fish (see Menshutkin (1971)), but in the LakeWeb-model one calculates net production and net biomasses of these two groups of organisms. This also means that, e.g., small planktivorous perch are categorised as prey fish whereas large (> 100 g) piscivorous perch belong to predatory fish. Likewise, the herbivorous zooplankton compartment, which plays a key role in lake biomanipulations, has been simplified in the LakeWeb-model and treated as one group in spite of the fact that this group consists of very heterogeneous organisms, such as flagellates, ciliates, rotifers, copepods and cladocerans. The size range of the organisms is from 0.003 to 2 mm. Therefore, within this group there are also trophic interactions (Sheldon et al. 1986; Stockner and Porter 1988; Stone and Weisburd 1992). There are also facultative predators in this group, such as the large rotifer Asplanchna priodonta, which feeds on different species of zooplankton and on phytoplankton (Boulion et al. 1999). Simplifications like these mentioned for fish and zooplankton were necessary for several reasons, (1) to keep the LakeWeb-model as small as possible (it is still quite extensive), (2) to keep the driving variable as few and as accessible as possible (this is important for the practical use of the model), and (3) to be able to critically test the model using existing empirical data or empirical regressions. The idea was not to include “everything” but to focus on the key functional groups of organisms and on key abiotic/biotic relationships. The LakeWeb-model is primarily intended to cover a very wide domain of lakes: • temperate, boreal and cold zones (at latitudes from 40 °N to 75 °N and altitudes from the sea level to 2000 meter above sea level), • large and small lakes (from 0.01 to 300 km 2; larger lakes would generally have to be divided into sub-areas where different functional organisms prevail), • shallow and deep lakes (lakes with mean depths of 1 m to 100 m),
90
Figure 2. Outline of the LakeWeb-model illustrating the nine functional groups of organisms. The model also includes a lake temperature sub-model that predicts epilimnetic temperatures from latitude, altitude and continentality (= distance from the Ocean), a sub-model for the depth of the photic layer ( ⬇ Secchi depth), a sub-model for suspended particulate matter (SPM = seston) and a mass-balance model for phosphorus. The obligatory driving variables are, total-P-concentrations, lake colour, lake pH, lake mean depth, maximum depth, lake area and epilimnetic temperature.
• oligotrophic to hypertrophic lakes (characteristic total-P concentrations from 2 to 300 g/l), • oligohumic to polyhumic lakes (typical lake colour values from 2 to 300 mg Pt/l), and • acid to basic lakes (mean pH-values from 3 to 11). Since there are only few obligatory driving variables in the LakeWeb-model, it is evident that the aim is to model general conditions. It should also be emphasised that the LakeWeb-model is primarily intended to handle feedbacks among the nine functional groups. Because biotic/abiotic feedbacks are also of great interest, three main such interrelationships are included in the model. One concerns the influence of material produced in the lake itself (autochthonous materials) on the depth of the photic zone, the second reductions in lake total phosphorus concentrations related to biouptake of phosphorus, and the third sus-
pended particulate matter (= SPM = seston), a key factor for bacterial production, sedimentation and other important functions. The field data for the two lakes will be given in tables in connection with the following scenarios.
Results “Biomanipulation-dilution” scenario – intensive fishing to reduce mercury in fish in Lake Blacksåstjärn, Sweden Lake conditions Intensive fishing was carried out in the “Liming-mercury-cesium” project (see Håkanson et al. (1990) and Håkanson and Peters (1995)) as a measure to reduce levels of mercury and radiocesium in fish eaten by
91 Table 1. Compilation of data for fish removal in five Swedish lakes where intensive fishing was used as a method to reduce mercury levels in fish (data from Håkanson et al. (1990)) Lake
Fish removed Period (REM; kg ww)
1818, Bergtjärn 24 2121, 227 Blacksåstjärn 2216, 313 Lill-Bandsjön 2217, 733 Hamstasjön 2218, 656 Väster-Långedsjön Lake Removal (%)
1818, Bergtjärn 18.5 2121, 53.4 Blacksåstjärn 2216, 58.2 Lill-Bandsjön 2217, 69.6 Hamstasjön 2218, 60.3 Väster-Långedsjön Mean value 52.0 (MV):
TP (g/l)
Area (km 2)
Normal fish bio-mass (BM; kg ww)
Normal fish production (PR; kg ww/yr)
87-05 to 87-06 87-05 to 88-06
11 12.5
0.04 0.12
130 425
78 89
87-05 to 88-06
11.6
0.16
538
82
87-05 to 88-06
25.3
0.18
1053
180
87-05 to 88-06
16.7
0.25
1089
119
Effect (= 100·REM/(BM + PR)) year 1 year 2
year 3
year 4
year 5
year 6
11.6 44.1
8.4 37.6
6.6 32.8
5.4 29.1
4.6 26.1
4.0 23.7
50.5
44.5
39.9
36.1
33.0
30.3
59.5
51.9
46.1
41.4
37.6
34.4
54.3
49.5
45.4
42.0
39.0
36.4
44.0
38.4
34.2
30.8
28.1
25.8
man according to the theory illustrated in Figure 1b. Various quantities of fish were removed by intensive net fishing in five lakes where no other measures, such as lake liming, fertilisation or potash treatment, were carried out. Table 1 shows for each lake the amount of fish removed, the period during which the operations were carried out, lake characteristics (mean lake TP-concentration and lake area), the normal fish biomass (as calculated according to an empirical regression given by Peters (1986)), the normal fish production (from an empirical model given by Håkanson and Boulion (2002)), and calculated changes from year 1 to year 6 after the operations. From this table, one can note that about 20 to 70% of the estimated fish biomasses were removed. Note that all these lakes are small (< 0.25 km 2) and have relatively low productivity (TP < 26 g/l). During these fishing operations, it became evident that it is very demanding to carry out intensive fishing operations and remove much of the fish biomass in relatively large lakes. It is also evident that if only
a small amount of the fish biomass is removed, one cannot expect any major success from the operation. Therefore, a prerequisite for success is that a large amount of the fish biomass is actually removed (say more than 50%). This means that one needs good methods to measure and/or predict the biomass of fish in any given lake. Table 2 gives a further differentiation of the species removed (pike, perch, roach and other). The theories illustrated in Figure 1 demonstrate the importance of knowing how much predatory and prey fish are removed, respectively, because they work in opposite directions in the lake foodweb. One should also note that: • It is practically impossible to remove just predatory fish by intensive net fishing. All types of fish larger than a given mesh size are caught in the net, not just predatory fish. • Large perch are mainly piscivorous. Therefore, if one assumes that 50% of the perch could be categorised as “predatory” fish, one can estimate
92 Table 2. Removal of different species from the five lakes subject to intensive fishing in the “Liming-mercury-cesium” project. Data from Håkanson et al. (1990) Lake 1818, 2121, 2121, 2216, 2217, 2218,
Bergtjärn Blacksåstjärn Blacksåstjärn Lill-Bandsjön Hamstasjön Väster-Långedsjön
Period
Pike
Perch
Roach
Other
Total (kg ww)
Pred. fish (kg ww)
Pred. fish (%)
87-05-12 to 87-06-09 87-05-04 to 87-06-04 88-05-10 to 88-06-02 87-05 to 88-06 87-05 to 88-06 87-05 to 88-06
11 70 45 237 291 463
10 36 48 39 88 127
3 6 22 36 42 66
0 0 0 1 312 0
24 112 115 313 733 656
16 88 69 257 335 527
66.7 78.6 60.0 81.9 45.7 80.3
Table 3. Data for Lake Blacksåstjärn (2121), Sweden, subject to intensive fishing in 1987 and 1988 (data from Håkanson et al. (1990)) General parameters: Latitude (°N) Altitude (m.a.s.l.) Continentality (= distance from the Sea; km) Mean annual precipitation (mm) Catchment area (ADA, km 2) Theoretical water retention time (T w, yr) Size parameters: Max. depth (D max, m) Total area (km 2) Form parameters: Mean depth (D m, m) Volume development (Vd, dim.less) Water chemistry (mean annual values): pH, 86, 87, 88, 89 TP (g/l), 86, 87, 88, 89 Colour (mg Pt/l), 86, 87, 88, 89
60.0 164 250 745 1.9 0.47 6.8 0.12 2.4 1.06 6.3, 6.6, 6.5, 6.4 12.5, 8.8, 7.5, 9.8 126, 89, 83, 71
the percentage of predatory fish caught in these operations. Table 2 then indicates that 46 to 82% of the fish caught in these five lakes could be classified as predatory. The rest would be prey fish. From this background, we will randomly focus on the conditions in one of the Swedish lakes, Lake Blacksåstjärn (2121), see Table 3 for lake-specific data. Lake Blacksåstjärn Figure 3 gives a compilation of the empirical changes in mercury concentrations in fish after the fishing operations, which generally were carried out in 1987. Therefore, t 0 in Figure 3 refers to the conditions in 1986, one year before the fishing operation. The data are not given for Hg pi (i.e., the Hg-concentration in 1 -kg pike), which has an ecological halflife of about 3
Figure 3. Compilation of empirical data from intensive fishing to reduce Hg-levels in fish (changes in concentrations in per fry = 1 + perch) by means of intensive fishing. This method was tested in five lakes (n = 5). t 0 refers to the conditions in 1986 before the fishing operations (which were carried out in 1987 in these lakes).
years (see Håkanson (1999)), but for Hg pe, i.e., the Hg-concentration in small perch (perch fry = 1 + perch = young-of-the-year perch), which, by definition, has a much shorter ecological halflife, and can be used in this context to get meaningful results for the fishing operations. However, one should note that small perch is an important food for pike, so if the Hg-concentration in perch fry decreases so will the Hg-concentration in fish that eats perch fry, but it will take a longer time. Two results should be noted from Figure 3: 1. The Hg-concentration in perch fry actually decreases, but not very much ( ⬇ 20% after 2 years). 2. There is a large variability in results among the five lakes. Table 4 gives the empirical data for the five lakes, and in the following, we will focus on the conditions in Lake 2121 (Blacksåstjärn). Note specifically that in this lake, and in two other lakes, the Hg-concentration is higher the year after the intensive fishing, and then in the second year after the intensive fishing the Hg-concentration in perch fry is lower than the initial
93 Table 4. Lake data from five lakes subject to intensive fishing. ADA = area of catchment, D max = maximum depth, D m = mean depth, TP = total phosphorus, Hg T = total Hg in lake water, Hg S = Hg-concentration in surface sediments, Hg pe = Hg-concentration in perch fry (1 + perch). Data from Håkanson et al. (1990) Lake
1818, Bergtjärn
ADA (km 2)
D max (m)
Area (km 2)
D m (m)
pH
Colour TP (mg Pt/l) (g/l)
Hg T (ng/l)
Hg S (mg/kg dw)
Hgpe 86 —
Hgpe 88 (mg/kg ww)
Hgpe 89 —
0.56
13.5
0.04
5.3
5.5
107
11
2.74
305
0.43
0.21
0.17
6.8
0.12
2.4
6.4
115
11.6
7.42
180
0.15
0.18
0.17
4.9
0.16
2.8
5.8
110
10
5.28
178.3
0.3
0.25
0.16
11.2
0.16
4
6.5
35
11.1
4.22
306.7
0.17
0.18
0.15
7.8
0.18
3.8
6.6
46
22.9
5.33
193.3
0.13
0.13
0.13
4.3
0.25
2.1
6.5
69
13.8
5.14
156.2
0.16
0.18
0.09
2121, 1.9 Blacksåstjärn 2121, 3.6 Blacksåstjärn 2216, 6.04 Lill-Bandsjön 2217, 11.57 Hamstasjön 2218, 16.83 Väster-Långedsjön
Figure 4. Temporal development of pH, colour, Secchi depth, TP-concentrations in Lake 2121 (Blacksåstjärn), where intensive fishing were performed in 1987 and 1988 to reduce Hg-concentrations in fish.
value. Can this pattern be explained by the way in which the biomanipulation was carried out? Even though there has been no other direct actions in Lake 2121 except for the intensive fishing, the conditions in the lake depend on variations in precipitation, temperature, etc., and on possible alterations in land-use activities, which could also influence directly or indirectly the Hg-concentration in small perch. Figure 4 gives on overview of the temporal variations in pH and lake TP-concentrations, which influence Hg-concentrations in fish (see Håkanson (1999)) plus variations in lake colour and Secchi depth, which influence several processes in the lake
foodweb (see Håkanson and Boulion (2001)). One can see that there are no major trends in these data and in the following simulations we will simply assume that the mean characteristic pH is 6.4, the mean lake TP-concentration 11.6 and the mean colour 115 mg Pt/l. Foodweb simulations From these presuppositions, the results of the casestudy are summarised in Figure 6. Note that the target variables in this scenario are the biomasses of prey and predatory fish. If the biomass of prey fish is increased, one can assume that the mean concentra-
94
Figure 5. Schematic outline of a food choice panel for a secondary unit with two first order food choices, four second order and two third order food choices. PU = primary unit (the food); SU = secondary unit (the consumer); DC = distribution coefficient (dimensionless unit regulating food choices).
tion of Hg in prey fish will decrease, since this concentrations is basically defined by the ratio between the amount of Hg (in mg) bound in prey fish and the biomass of prey fish (in kg ww). The simulated intensive fishing is intended to describe the actual fishing in the lake. Totally about 50–55% of the fish biomass was reduced during two consecutive fishing events of almost equal size, one in 1987, the other in 1988. To describe these reductions, we have set the fishing rate to 0.3 week 389 and to 0.3 again week 441. A fishing rate of 0.3 means that 30% of the fish biomass is removed a given week. To understand the following results, it is important to mention that if there are more than two food choices for any secondary producer, the LakeWebmodel uses a simple general system to assign weights on such food choices and adjust the consumption rates for the number of food choices. In the LakeWebmodel, there are 3 food choices for zoobenthos (i.e., benthic algae, macrophytes and sediments) and also prey fish has a menu of 3 food alternatives (predatory zooplankton, herbivorous zooplankton and zoobenthos). If there are more than 2 food choices, they are first differentiated by a (dimensionless) distribution coefficient (DC 1) into two first order food choices and then by a second distribution coefficient
(DC 2) into second order food choices, etc. (see Figure 5 for illustration). Four different lake foodweb characteristics have been tested in this scenario: 1. Under the defined default conditions. This means that the first order distribution coefficient (DC PY1) regulating the prey fish consumption of either zooplankton or “the rest” is set to 0.5, and that the second order distribution coefficient (DC PY2) regulating the prey fish consumption of either Zoopred. or Zooherb. is set to 0.75, and that 10% of the total fish biomass is lost every year by fishing (of birds, animals and man). We will change these values in the following scenario. 2. The results given by curve 2 have been derived when the first order DC is set to 0.5/2 (and all else is kept constant). This means that prey fish consumes more zoobenthos, and that the role of zooplankton for the fish production is smaller. 3. Curve 3 gives the values when the second order DC is set to 0.75/2 (otherwise, we use the default conditions, as defined by Håkanson and Boulion (2002)). This means that prey fish consumes more Zooherb. and less Zoopred. 4. Curve 4, finally, gives the results when there is no default fishing. This is realistic when the fish bio-
95 mass is small. No other changes have been applied to the LakeWebmodel, except those for the lake-specific driving variables (area, mean depth, colour, pH, etc.; see Table 3). So, there has been no “tuning” of the model. This is also a sensitivity analysis where different foodweb characteristics are tested and all other conditions have been unchanged. These tests address the following questions: Will there be any major changes in the target variable, the prey fish biomass? What will happen to lake TP-concentrations and algal volumes? Since the prey fish biomass depends on the amount of zooplankton for prey fish consumption – will there be any changes in Zooherb. and Zoopred. biomasses? Figure 6 gives the results and one can note: 1. There are relatively small differences between the four curves for all six variables. So, the model predictions are not very sensitive to the data used for these DC-values in this particular lake. The largest departure from the default conditions concerns curve 4 in Figure 6A. If there is no default fishing, the biomass of predatory fish recovers faster, which is logical. Generally, it takes several years for the predatory fish biomass to recover after the intensive fishing week 389. 2. The patterns of the curves are interesting. Just after the fish removal, there is a very clear reduction in predatory fish biomass (Figure 6A), and also in the prey fish biomass (Figure 6B). This means an increased predation pressure especially on Zoopred. (since the default value for the second order DC is 0.75). The pattern for the prey fish biomass is mirrored by an inverse pattern for the Zoopred. biomass (figures 6B and 6F). Then, it is interesting to notice that the reduced prey fish biomass during the period just after the intensive fishing should cause an increased Hg-content in prey fish, and the following higher prey fish biomass should lower the Hg-content in prey fish. This is consistent with the empirical results for the Hgconcentration in small perch (see Table 4) in this lake, and two other lakes. 3. There are small differences among the four curves for TP-concentrations (Figure 6C), algal volumes (D) and biomass of Zooherb. (E).
“Classical biomanipulation” scenario – intensive fishing to reduce algal volume in Lake Batorino, Belarus Lake Batorino The aim of this scenario is to illustrate how the LakeWeb-model may be used as a tool in lake management to plan and predict the outcome of an action (here a fish removal), to get realistic expectations so that actions likely to fail can be avoided and time and money spent on more optimal measures. We have used the LakeWeb-model in eutrophic Lake Batorino, Belarus (see Table 5 for lake data), to test if it is possible to influence fundamental lake management variables in a desired direction by biomanipulation (here intensive fishing) so that there is a reduction in lake TP-concentrations and algal volume (and increased Secchi depth and macrophyte cover). This operation has not been carried out in the lake, but it could be a feasible option in lakes such as Lake Batorino with high TP-concentrations. The objective is to determine whether intensive fishing would reduce algal volumes (= AV) from an initial situation with AV-values around the critical limit of 5 mm 3/l. Foodweb simulations The set-up of the simulation scenario is given by: 1. An hypothetical fishing rate was set to 0.5 (week −1) on week 389 in the summer of 1987 (the first week of January 1980 is week 1). 2. Three different lake foodweb scenarios have been tested: 1. Under the defined default conditions (as defined by Håkanson and Boulion (2002)). 2. We have decreased the consumption rate for Zoopred. eating Zooherb. by a factor of 3 (i.e., 3 times lower) and at the same time increased the consumption rate of prey fish by a factor of 3 to optimise the results according to the theory given in Figure 1. 3. We have also decreased the consumption rate for Zoopred. eating Zooherb. by a factor of 6 (by 1/6), and also increased the consumption rate of prey fish by a factor of 6 to improve the results even more, according to the theory. No other changes have been done, except that the model is driven by the lake-specific data (area, mean depth, colour, pH, etc.; see Table 5). There are sev-
96
Figure 6. Simulation results of biomanipulation (intensive fishing) in oligotrophic Lake Blacksåstjärn, Sweden. The aim was to lower the Hg-content in fish eaten by man. There are no other changes to the default conditions for the LakeWeb-model, except that (1) the fishing rate was set to 0.3 (week −1) on week 389 (summer 1987) and again to 0.3 on week 441 (summer 1988), and (2) lake-specific values were used for all driving variables. Curve 1 gives the predictions under default conditions, curve 2 when the distribution coefficient (DC) regulating prey fish consumption either of zooplankton or zoobenthos has been changed from the default value of 0.5 to 0.25, curve 3 when the distribution coefficient for prey fish consumption of either predatory zooplankton or herbivorous zooplankton was changed from the default value of 0.75 to 0.75/2, curve 4 there is no default fishing (by birds, animals or man, except the intensive fishing). Figure A gives the results for predatory fish biomass, B for prey fish biomass, C for lake TP-concentrations, D for algal volume, E for biomass of herbivorous zooplankton and F for predatory zooplankton.
eral questions related to this scenario: Will there be major or small differences in the target variables when these consumption rates are changed? This is important since there are always uncertainties in the values for the consumption rates. Are the model predictions sensitive to such uncertainties? Is it realistic
to expect that the target variable for this biomanipulation, the algal volume, will go well below the critical limit? And for how long time will the algal biomasses (= volumes) likely stay low? The simulation results are summarised in Figure 7 and one can note:
97 Table 5. Driving variables for Lake Batorino, Belarus. The continentality (needed to calculate lake temperatures in the LakeWeb-model) is set to 500 km for the lake. C in = tributary concentration of total phosphorus (g/l), Lat = latitude, Alt = altitude and Prec = mean annual precipitation Catchment km 2
Area km 2
Mean depth Max. depth D max, m D m, m
pH
Colour mg Pt/l
C in g, P/l
Lat. °N
Alt. m.a.s.l.
Prec. mm/yr
92.5
6.3
3.0
8.0
54
120
54.5
165
650
5.5
• There are significant changes in predatory fish biomass (Figure 7A), more than a factor of 2 between curves 1 and 3. This is a result of the action to improve the conditions in the lake. It takes several years (about 10 years under default conditions) for the predatory fish biomass to recover after the intensive fishing on week 389. If the consumption rate for prey fish is increased (curves 2 and 3), the prey fish biomass will increase (as long as there is sufficient food for the prey fish), and hence also the biomass of predatory fish. • There are clear differences between the three curves for prey fish (Figure 7B), and the pattern of these curves is very interesting. Just after the fishing operation week 389, there is a reduction in the biomass of the prey fish. But the turnover time for the prey fish is shorter than for predatory fish, and since the biomass of the predatory fish remains low for a relatively long time, the predation pressure on the prey fish is reduced and the biomass of the prey fish will soon be significantly higher than before the intensive fishing. This means an increased predation pressure on Zooherb, but more so on Zoopred., since the default value of this distribution coefficient is 0.75, which means that the flux to prey fish from Zoopred. is multiplied by a factor of 0.75 and the flux to prey fish from Zooherb. by a factor of 0.25. So, the pattern for the prey fish biomass is inversely reflected in the pattern for the Zoopred. biomass (figures 7B and 7F). • There are almost no differences between the three curves for the TP-concentrations (Figure 7C). • The key issue in this scenario concerns the algal volume (Figure 7D) and the biomass of Zooherb. regulating the algal volume (Figure 7E). The principle way to reduce algal volume in this scenario, when there are no other changes except the fishing on week 389, is to increase the biomass of Zooherb. The best way to do this is
to reduce the predation pressure on Zooherb. from Zoopred., or to do this action in lakes with relatively small biomasses of Zoopred. From Figure 7E, one can see that there are significant changes in the biomasses of Zooherb, for the three scenarios with different consumption rates for Zoopred. eating Zooherb. This also means that the higher the biomass of Zooherb., the higher the grazing pressure on phytoplankton and the lower the algal volume.
Discussions and conclusions The “biomanipulation-dilution” scenario The intensive fishing operations carried out in Lake 2121 (Blacksåstjärn, Sweden) to reduce Hg-concentrations by increasing the prey fish biomass are not likely to succeed and field data demonstrate that these operations did not succeed. It is interesting to note that the seasonal variations in Hg-content in perch fry with higher values the first years after the fishing operation and the following lower values can be explained by variations in the biomass of prey fish. It should be stressed that the empirical data from this lake also show that the changes in the Hg-concentrations in small perch are small. This method of biomanipulation cannot be recommended for lakes of this type. Methods to reduce Hg-concentrations in fish are, unfortunately, needed for lakes of this type, not just in Sweden, but in many countries. A typical cost of an intensive fishing is about 10,000–30,000 USD per lake (of this size; ⭐ 0.25 km 2; see Håkanson et al. (1990)). So, the costs to remove fish would be about 40–120 USD per kg ww fish removed! However, if the fishing operation is carried out in the form of a fishing rally, there may be recreational benefits for the anglers, a monetary profit for the owner of the lake, and very small possible improvements of short duration also concerning the Hg-concentrations in the lake fish.
98
Figure 7. Simulation results of “classical” biomanipulation (intensive fishing) in eutrophic Lake Batorino, Belarus. The aim is to lower the algal volume clearly below the critical limit of 5 mm 3/l (see Fig. D). There are no changes in the default conditions for the LakeWeb-model, except that (1) the fishing rate was set to 0.5 (week −1) on week 389 and (2) lake-specific values have been used for all driving variables. The normal consumption rate for predatory zooplankton has been lowered in two steps (by 3 and 6 times) compared to the default conditions and the consumption rate for prey fish increased in two steps from the default conditions (by 3 and 6; the respective curves are marked 1, 2 and 3).
Had the LakeWeb-model been available when these actions were planned, they would probably never have been done. This scenario illustrates, we hope, the great benefit of the LakeWeb-model as a tool to plan lake management actions to get realistic expectations so that the best possible actions are taken and other actions avoided.
“Classical biomanipulation” scenario Intensive fishing simulated in this way for Lake Batorino by reducing the fish biomass by a given factor (here 0.5 week 389), will in this lake likely increase the prey fish biomass after the operation. The biomass of prey fish will be higher than the initial value as long as the predation pressure on prey fish is low. The biomass of predatory fish will recover only slowly.
99 This operation is not likely to succeed in lowering the algal volume in lakes with “normal” populations of predatory zooplankton, such as Lake Batorino.
Acknowledgements The studies presented here have partly been carried out with support of the Russian Foundation for Basic Research (Project 00-15-97825).
References Benndorf J. 1990. Conditions for effective biomanipulation: Conclusions derived from Whole-Lake experiments in Europe. Hydrobiologia 200/201: 187–203. Benndorf J. 1992. The Control of the Indirect Effects of Biomanipulation. Eutrophication: Research and Application to Water Supply. Freshwater Biological Association, London, pp. 82–93. Boulion V.V., Nikulina V.N., Paveljeva E.B. and Stepanova L.A. 1999. Microbial loop in trophic web of lake plankton. Journal of General Biology 60: 431–444. Håkanson L. 1999. Water Pollution – Methods and Criteria to Rank, Model and Remediate Chemical Threats to Aquatic Ecosystems. Backhuys Publishers, Leiden, 299 p. Håkanson L., Andersson P., Andersson T., Bengtsson Å., Grahn P., Johansson J.-Å. et al. 1990. Measures to reduce mercury in lake fish. Final report from the Liming-Mercury-Cesium project. Swedish Environmental Protection Agency, Stockholm, PM 3818, 189 p. Håkanson L. and Boulion V.V. 2001. Regularities in primary production, Secchi depth and fish yield and a new system to define trophic and humic state indices for aquatic ecosystems. Int. Rev. Hydrobiol. 86: 23–62. Håkanson L. and Boulion V. 2002. The Lake Foodweb – Modelling Predation and Abiotic/biotic Interactions. Backhuys Publishers, Leiden, 344 p.
Håkanson L. and Peters R.H. 1995. Predictive Limnology. Methods for Predictive Modelling. SPB Academic Publishing, Amsterdam, 464 p. Hanson M. and Butler M.G. 1994. Responses of plankton, turbidity and macrophytes to biomanipulation in a shallow Prairie Lake. Can. J. Fish. Aquat. Sci. 51: 1180–1188. Hrbácek J. 1958. Typologie und Produktivität der teichartigen Gewässer. Verh. Int. Ver. Limnol. 13: 394–399. Jansson M., Hayman U. and Forsberg C. 1981. Acid lakes and “biological buffering” (in Swedish). Vatten 37: 241–251. Kitchell J.F. and Carpenter S.R. 1988. Food web manipulation in experimental lakes. Verh. Int. Ver. Limnol. 23: 351–358. McQueen D.J., Post J.R. and Mills E.L. 1986. Trophic relationships in fresh-water pelagic ecosystems. Can. J. Fish. Aquat. Sci. 43: 1571–1581. Menshutkin V.V. 1971. Mathematical modelling of populations and communities of aquatic animals., Leningrad, (in Russian). Moss B., Stensfield J. and Irvine K. 1991. Development of daphnid communities in diatom and cyanophyte dominated lakes and their relevance to lake restoration by biomanipulation. J. Appl. Ecol. 28: 586–602. Peters R.H. 1986. The role of prediction in limnology. Limnol. Oceanogr. 31: 1143–1159. Shapiro J. 1990. Biomanipulation: The next phase making it stable. Hydrobiologia 200/201: 13–27. Shapiro J. and Wright D.I. 1984. Lake restoration by biomanipulation: Round Lake, Minnesota, the first two years. Freshwater Biol. 14: 371–383. Sheldon R.W., Nival P. and Rassoulzadegan F. 1986. An experimental ivestigation of flagellate – ciliate – copepod food chain with some observations relevant to the linear biomass hypothesis. Limnol. Oceanogr. 31: 184–188. Stockner J.G. and Porter K.G. 1988. Microbial Food Webs in Freshwater Planktonic Ecosystems. Complex Interactions in Lake Ecosystems., New York, pp. 69–83. Stone L. and Weisburd R.S.J. 1992. Positive feedback in aquatic ecosystems. Trends in ecology and evolution 7: 263–267. Theiss J., Zielinski K. and Lang H. 1990. Biomanipulation by introduction of herbivorous zooplankton. A helpful shock for eutrophic lakes? Hydrobiologia 200/201: 59–68.
