University of Groningen The biology and impacts of Oreochromis niloticus and Limnothrissa miodon introduced in Lake Kariba Chifamba, Chiyedza Portia

(1)

University of Groningen

The biology and impacts of Oreochromis niloticus and Limnothrissa miodon introduced in

Lake Kariba

Chifamba, Chiyedza Portia

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.

Document Version

Publisher's PDF, also known as Version of record

Publication date: 2019

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):

Chifamba, C. P. (2019). The biology and impacts of Oreochromis niloticus and Limnothrissa miodon introduced in Lake Kariba. Rijksuniversiteit Groningen.

Copyright

Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

(2)

528794-L-bw-Chifamba 528794-L-bw-Chifamba 528794-L-bw-Chifamba 528794-L-bw-Chifamba Processed on: 6-2-2019 Processed on: 6-2-2019 Processed on: 6-2-2019

Processed on: 6-2-2019 PDF page: 117PDF page: 117PDF page: 117PDF page: 117

_{miodon (Boulenger 1906) estimated from diurnal}

increments in otoliths

Portia C. Chifamba

Jan H. Wanink

Britas K. Eriksson

John J. Videler

7 Chapter

(3)

Abstract

A commercial fishery in Lake Kariba, on the introduced freshwater sardine

Lim-nothrissa miodon (Boulenger 1906) started in 1974, rose to a peak in 1990, and

crushed thereafter. Overfishing and population changes in the plankton commu-nity, driven by a rise in temperature, might have reduced the sardine population and have changed its life history parameters such as growth and age at maturity. We used otolith reading to test if the population collapse was due to:

1) stunted growth, as a result of impoverished food sources, by estimating changes in growth rate from 1993 to 2013, or to

2) fishing pressure, that in time affected the size and age at maturity.

Sagittae, the largest otoliths of L. miodon were examined under the Transmission Electron Microscope (TEM) to determine periodicity of growth increments. Diurnal increments were validated using edge-increment analysis and used to estimate age and growth parameters. Fish for increment validation were sampled over a 24-h cycle, at 2-h intervals. The increments at the margins of all otoliths from fish caught at night were light coloured under TEM, while 80% of those from fish caught during daytime were dark. Each pair of dark and light increments were taken as a diurnal increment and used for ageing.

The growth trajectories of L. miodon vary among years and display differen-ces in the growth of juveniles/small adults (< 10 months) and large adults (> 10 months). Juveniles and small adults grew fastest in 1996 and slowest in 2013, which may indicate particularly favourable conditions in1996. This supports the hypothesis of poorer plankton diets in recent years. The Gompertz model fitted the data better than the von Bertalanffy, logistic and power growth models. Esti-mates of asymptotic length (L∞) from the Gompertz model were 18.0, 9.6 and 15.2

cm total length in 1993 – 1994, 1996 and 2012 – 2013, respectively. These values are comparable to the asymptotic length of L. miodon reported for its lake of origin, Lake Tanganyika. Age at first maturity was 8.02 and 7.90 months for females and males, respectively. First maturity occurred at a much smaller size in this study (females: 3.43 cm; males: 3.63 cm), compared to reported values for 1970 – 1972 (females: 5.2 – 5.6 cm; males: 7.1 – 7.3 cm). Temporal differences in growth rate and length at first maturity reflect a flexible life-history strategy of L. miodon under the influence of fishing pressure and possibly changes in food availability.

(4)

119

7

Introduction

The freshwater sardine, Limnothrissa miodon (Boulenger 1906) is a pelagic fish, introduced from Lake Tanganyika into Lake Kariba in the late 1960s to increase fish production (Balon 1974; Bell-Cross & Bell-Cross 1971). Commercial fishing began in 1973 and rose to a maximum yearly catch of 20 112 metric tons landed in Zimbab-we in 1990 (Karenge & Mugwagwa unpublished report). The collapse of the fishery, thereafter, has been attributed to overfishing and increased water temperature (Magadza 2011; Marshall 2012a). Catches of L. miodon are negatively correlated to mean maximum air temperature (Chifamba 2000). Increase in temperature caused a change in phytoplankton species composition, favouring unpalatable cyanobacteria and resulting in a reduction of the zooplankton upon which L. miodon feeds (Magadza 2011). Inadequate food may have reduced growth of L. miodon, which we investigate here.

Growth and reproduction

Growth of L. miodon in Lake Kariba has been studied using body length and otolith-based methods (Cochrane 1984; Marshall 1987a; Chifamba 1992). Using length-otolith-based methods, Cochrane (1984) and Marshall (1987a) estimated an asymptotic length (L∞) of 8.1 and 7.4 cm total length (TL), respectively, and concluded that the

popu-lation of Lake Kariba was stunted. Based on reading otoliths, however, Chifamba (1992) found an asymptotic length (L∞ = 13.8 cm) comparable to the values reported

for Lake Tanganyika and Lake Kivu, a natural lake where L. miodon was introduced (Spliethoff et al. 1983; Moreau et al. 1991; Mulimbwa & Shirakihara 1994). In Lake Kariba, L. miodon up to 15.5 cm length are rarely caught, hence the low asymptotic length obtained in the length frequency analysis by Marshall (1987a). The mean size in the commercial fishery was 5.7 cm in 1993 and 5.1 cm in 1996 (Chifamba, un-published data).

Previous studies in Lake Kariba show that L. miodon reached sexual maturity at a small length compared to populations in Lakes Tanganyika and Kivu (Woodward 1974; Spliethoff et al. 1983; Moreau et al. 1991). The smallest size at first maturity in 1970 was 5.2 – 5.6 and 7.1 – 7.3 cm fork length (FL) for females and males respec-tively (Woodward 1974). More recent estimates of growth rate as well as length and age at first maturity are needed, considering the ecological changes that occurred as the lake matured, and the L. miodon fishery that has evolved. Here we present such data, based on studying diurnal increments in the otoliths.

Reading otoliths

Otoliths (lapilli, sagittae and asterisci; located in the inner ear and used for hearing and acceleration detection) can be used to age fish. Between 90 – 99% of the total

(5)

120

otolith mass is calcium carbonate. The remaining 1 – 10% is an organic matrix com-posed of proteins (Degens et al. 1969; Payan et al. 2004; Borelli et al. 2001). During otolith formation and growth, the protein matrix plays a major role in cyclic calcium deposition. Otolith matrix proteins are necessary for the deposition of otolin, a protein that anchors otoliths in the sensory maculae and stabilises the otolith matrix (Murayama et al. 2000, 2002, 2005).

Daily increments can be seen in the otoliths as alternating layers of organic materials and minerals, mostly calcium carbonate (CaCO3) (Watabe et al. 1982; Mugiya 1987). Borelli et al. (2003) proposed that organic matrix and calcium carbonate deposition in otoliths varies diurnally, with the organic matrix deposited during the day and CaCO3 at night. The alternate deposition of CaCO3-rich and protein-rich layers results in the formation of daily increments in the otoliths.Fish kept under constant dark or constant light conditions maintain the cyclic rhythm, indicating that the rhythm may be intrinsic or due to other factors (Roberts et al. 2004).

Daily increments of sagittae, the largest otoliths in most teleosts (Harder 1975), have traditionally been used for age determination (Pannella 1971; Zekeria et al. 2006; Nyamweya et al. 2010). Validation of increments can be done by reading otoliths of fish of known age and counting the number of rearing days. It can also be done by counting the number of increments during a known period following marking

(Campana & Neilson 1985; Geffen 1992). The latter method is recommended for larvae and fish species with a short life span (Brothers 1976). Counting daily increments for age determination is appropriate for L. miodon because the majority of the fish caught are less than one year old, and the life span does not exceed two years (Cochrane 1984). This technique has been applied to age adult and juvenile L.

miodon in Lake Kariba by Chifamba (1992) and Mtsambiwa (1993). At the time of

these studies, no validation of periodicity of increment deposition was done. Meisfjord et al. (2006) confirmed a daily deposition of increments in otoliths using chemical markers on reared juvenile L. miodon.

Marginal increments show the last deposition on the outer rim of the otolith.

Alternating organic and mineral layers are deposited discontinuously during the daily cycle. To visualise the separate margins with sufficient resolution, Tanaka et al. (1981) and Zhang & Runhau (1992) studied the formation of marginal increments in Oreochromis niloticus otoliths using scanning (SEM) and transmission electron

microscopy TEM). We used TEM to do this for L. miodon.

Objectives

The objectives of this study were to:

1) validate the periodic increments of the sagittae of L. miodon from Lake Kariba; 2) use these increments to age L. miodon caught in 1993 – 1994, 1996 and 2012 – 2013

in order to determine any temporal changes in the growth parameters of the fish; 3) establish the age and length at first maturity;

(6)

121

7

4) use this information to assess whether the fishery catches mature fish, and if

fishing affects size and age at maturity. This assessment is important because catching fish before they reproduce, undermines the capacity of the population to sustain itself. Furthermore, evolutionary changes of life-history parameters induced by fishing can affect the sustainability of the L. miodon fishery.

Materials and methods

Fish samples for the validation of otolith increments and those for the estimation of growth parameters were collected and analysed separately.

Collection of samples for validation of increments

Fish were collected from the Sanyati basin of Lake Kariba (Figure 1.1) over one 24-h period, using a trawl net by day and a lift net at night. During the night, samples were taken at 2-h intervals, starting at 18:00 h and ending at 04:00 h. Daylight samples were taken at 08:00, 10:00, 12:00 and 14:00 h. Nets were set for one hour in both cases. The fish were preserved on ice in the field, for transport to the laboratory. Capture time was recorded for each sample. In the laboratory, the two sagittae otoliths were removed and stored together in the primary fixative, cacodylate buffered glutaraldehyde.

Preparation of otoliths for Electron Microscope examination

The sagittae were fixed, stained, dehydrated and embedded following the procedure described by Glamart (1984) and used by Zhang & Runhau (1992), with the follow-ing modifications. Cacodylate buffer and dry acetone were used in place of saline buffer and propylene oxide. For primary fixation, otoliths were kept in cacodylate buffered glutaraldehyde, prepared using 50 ml 0.2-M cacodylate buffer, 12 ml 25% glutaraldehyde and 88 ml of distilled water.

Sectioning and Staining

Initially, 1-mm thick sections were cut on an OM-U2 Reichert Ultra microtome, using glass knives. The sections were stained with 1% Toluidine blue and examined under an Olympus EHS light microscope for opaque and translucent bands, which appeared dark and light respectively. When increments were found, ultra-thin sections were then cut using the same microtome, placed on copper grids and stained with 5% Uranyl acetate in 30% methanol for one hour, followed by Reynolds' lead citrate for 5 – 7 minutes (Reynolds 1963). Stained sections were examined using a Carl Zeiss EMIOC Transmission Electron Microscope at magnifications between 750 and 16 000 times.

Interpretation of increments

For each sample that showed a marginal increment, the type of the increment was noted, measured and matched with the time of capture. At each time period, the numbers of fish

(7)

122

with either a light or a dark increment were added up and used to assess the pattern of increment formation.

Collection of samples for age and growth determination

Fish samples were obtained from commercial catches and by experimental fishing in the Sanyati basin of Lake Kariba in 1993, 1994, 1996, 2012 and 2013, and in the Mlibizi basin in 1992 (Figure 1.1). Specimens from 1992 and 1996 were used to study the width of daily increments, and the relationship between the radial of an otolith and fish length. The total number of increments was counted in otoliths from all samples and used to establish the ages of fish. Additional data from 1988 were taken from Chifamba (1992).

Larval and juvenile L. miodon are found in shallow, inshore areas, and the adults in deep, open water (Cochrane 1984; Mtsambiwa 1989). Adult fish were collected from commercial landings, while larvae and juveniles (TL < 2.5 cm) were captured in shallow areas (depth < 10 m) using a lift net lined with a 1-mm mesh net. Total length of each fish was measured to the nearest mm. Fish collected in 2013, and those larger than 8 cm in other years (from samples preserved in ethanol), were sexed. The fish were examined under a dissecting microscope to determine the sex and the stage of gonadal development. The sex of immature fish could not be determined with certainty. In all analyses, this group is assumed to contain both sexes, a procedure adopted by Kimura (1995). Sagittae were removed and cleaned in water, then air dried. The pair of sagittae from an individual fish were wrapped in plastic (stretch wrap) and stored in labelled paper envelops until needed for further analysis.

Ageing using increments

The procedures for storage and mounting of otoliths used in this study were those recommended by Morales-Nin (1992). Sagittae were mounted on glass microscope slides using transparent nail varnish, except before 1996, when those larger than 3 cm were mounted in blocks of epoxy resin.

To reveal increments, the otoliths were grounded and polished sequentially on silicon carbide paper of decreasing grain size: 40.5 µm (320), 15.3 µm (1200), 6.5 µm (2400) and 2.5 µm (4000), and on 3-µm Imperial lapping film. Grinding and polishing was done with the otolith surface submerged in a film of soapy water. Throughout the grinding process, each otolith was checked periodically to avoid over grinding and to achieve adequate polishing.

After drying the slide, increments were read under oil emersion at 1000× and 2000× magnification, using a stereo compound light microscope. The increments were read using one of the following three methods, depending on availability of equipment and size of the fish:

1) direct reading on the microscope, using a calibrated eyepiece graticule and measuring the increments in groups of ten (suitable fish < 3 cm);

(8)

123

7

Figure 7.1 Sagittae otolith of a) a juvenile (1.6 cm TL) showing 35 daily increments and b) an adult

sardine (11.7 cm TL) with 552 increments, showing the focus and the shaded area within which increments were counted and the radial that was measured.

2) taking digital photographs from the microscope and read on a computer; 3) using a digitizer, a computer and a monitor connected to the microscope, to count

and measure increment width using a computer programme described by Andersen & Moksness (1988).

Increments were counted in the cauda of the otoliths, along a radial from the focus posteriorly to the margin in the region marked in Figure 7.1. The radial was selected on the basis of clarity of increments. A set of one dark and one light increment was considered to be a daily increment and used to age the fish. One sagitta per fish was counted and each otolith was counted at least twice, depending on the clarity of increments, and the counts averaged.

Estimation of growth parameters

Even though the von Bertalanffy growth model (VBGM) is a prime choice in fish studies including L. miodon, models such as the Gompertz and logistic models better fit data of some species (Cochrane 1984; Marshall 1987a; Kimura 1995; Katsanevakis 2006; Chapter 3 – Chifamba & Videler 2014). Therefore, to find the model that best describes our length at age data, we fitted the following four commonly-used growth models to all data sets and estimated the growth parameters using Sigma Plot 12: 1) VBGM L(t) = L∞ (1 – exp (– k1 (t – t0)))

where L(t) = total length at age t (cm), L∞ = asymptotic length (cm), t = age (months)

at capture, k1 = relative growth parameter (monthly) and t0 = age at which individuals

would have been at zero length;

(9)

124

where L(t) = total length at age t (cm), L∞_{= asymptotic length (cm), t = age (months)}

at capture, k2 = rate of exponential decrease in relative growth rate (monthly) and t1 = 1 / k2 ln λ, where λ = theoretical initial growth rate at zero age;

3) Logistic L(t) = L∞ / (1 + (t / t2) – k3)

at capture, k3 = relative growth rate (monthly) and t2 = inflection point of the sigmoid

curve;

4) Power L(t) = a1 tb

at capture, a1 = scaling factor or rate at which length increases with time, and b =

exponent describing the rate of change in the relationship between length and age. An evidence-based approach was used to select the model that best explains the data in order to avoid making prior assumption of a model. Models were compared using the Akaike Information Criterion (AICc) and the R2_{value (the coefficient of determination)} (Anderson et al. 2000; Burnham & Anderson 2002; Burnham et al. 2011).

The age in days was divided by the average number of days in a month (365.25 days and 12 months in a year) to convert days to months. The mean length at age was calculated and used to compare growth trajectories among years.

Age and length data from 2013 were used to estimate the age (A50) and length

(L50) at first maturity, as the value at which 50% of the fish in the population

(esti-mated from the sample) are mature (Trippel & Harvey 1991; Chen & Paleheimo 1994; Piñeiro & Saínza 2003; Karna & Panda 2011). In estimating A50, the fish in

the sample were allocated to two-month age classes, such that Class 0 contained fish from 0.0 to 1.9 months old, Class 1 fish from 2.0 to 3.9 months, and so on.

Length classes for estimating L50 were made at 1-cm intervals, such that they

contained fish lengths from 1.0 to 1.9 cm, 2.0 to 2.9 cm, and so on. The age and length at which 50% of the fish in the sample reached maturity, were estimated from sigmoidal curves fitted to plots of the percentage of mature fish in a class against the age and length class. After comparing several methods to estimate maturity in fish, Trippel & Harvey (1981) recommended the use of maximum likelihood fit of a sigmoidal function on data that display successive increase of mature fish with length or age. Similarly, Chen & Palohemo (1994) recommended the use of a two-para-meter logistic model, fitted using non-linear least squares method to the estimated 50% maturity. Since L. miodon data showed successive increase of maturity with age and length, L50 and A50 could be estimated from the sigmoid function given

(10)

125

7

M = 100 / (1 + exp (- (a1 - a2) / b))

where M is the percentage of mature fish, at the age or length a1, a2 is the age or

length at the inflection point or mid-point of the sigmoid curve, and b is the slope or steepness of the curve. The curve’s maximum value was set at 100 representing 100% maturity. The length at first maturity is where M equals 50% mature fish, which in this model is a2.

Results

Validation of increments

Sagittae of 32 fish, varying in size between 43 and 71 mm (TL), were examined. Twenty-seven showed a marginal increment and were used in the analysis. Magnifications of 4 000 – 16 000× revealed successions of increments consisting of dark and light material (Figure 7.2). Figure 7.3 shows the typical dark and light margins observed at various times of the day. During daytime, eight out of ten sagittae had a dark incre-ment at the margin (Figure 7.4). Of the fish caught at night, all sagittae investigated had a light increment at the margin. The dark increment starts to increase in the morning and peaks around noon. The deposition pattern of the light increment is not clear but appears to start forming in the afternoon (Figure 7.5).

Figure 7.2 Diurnal successions of light and dark increments showing the differences in the density

of the protein matrix in a decalcified section of an otolith, viewed at a magnification of 16 000×. The scale bar is 0.5 µm.

(11)

126

Figure 7.3 Dark and light increments (marked with bracket) at the edges of otoliths taken from L. miodon caught in daytime at a) 10.00 h and b) 14.00 h, and at night at c) 22.00 h and d) 04.00 h

(Scale bar = 1 µm).

Increment width and ageing

A total of 349 fish were aged and used to study growth of L. miodon. The largest fish analysed was 16.5 cm and the oldest had 815 daily increments. Ontogenetic change in the shape of the otolith has implications to the readability of the sagittae. Distortion of the increments causes difficulties in resolving the increments, especially in specimens > 8 cm. The shape of the sagittae changes with age from almost circular in larvae, to an elongate shape due to an enlarged anterior rostrum in adult fish (Figure 7.1). The antirostrum is another projection in the anterior of the otolith that is separated from the rostrum by a notch. In this study, increments along the radial from the focus to the anterior margin were most consistent among specimens. Therefore, this radial was used to study increments and age. The focus is the centre of the area known as the nucleus which is formed early in the development of the otolith, around which the increments form.

(12)

127

7

Figure 7.4 Frequency distribution of L. miodon specimens with dark (black bars) or light (white bars) increments at the edge of the otolith, at a given time of the day (sunrise at 06:26 h and sunset at 17:38 h on the day of sampling). There is no data for 16:00 h (n = 27). Figure 7.5 The width of dark (black dots) and light (white dots) increments at the edge of the otolith, at a given time of the day (sunrise at 06:26 h and sunset at 17:38 h on the day of sampling) (n = 27). Time of day (h) N um ber of fi sh (n ) 0 1 2 3 4 5 sunrise sunset 08:00 12:00 16:00 20:00 00:00 04:00 Time of day (h) E dg e i nc re m en t w id th (m) 0.0 0.5 1.0 1.5 2.0 sunset sunrise 08:00 12:00 16:00 20:00 00:00 04:00 R2_{= 0.81} p = 0.0006

(13)

128 Figure 7.6 a) Relationships between fish length and radial of sagittae, from fish caught in water < 10 m and > 20 m deep in 1996 (n = 154). A logarithmic curve is fitted through all data, b) change in increment width (± standard deviation) of daily increments against increment number starting from the focus to the margin of sagittae from fish caught in 1996 in shallow and deep water, c) change in increment width (± standard deviation) against increment number in 1992 (n = 30). c) increment width 1992

Increments number from focus

0 50 100 150 200 250 300 In cr em ent w idt h (µ m ) 0 2 4 6 8 10 b) increment width 1996

Increment number from focus

0 50 100 150 200 250 300 Incr em ent w idt h (µ m ) 0 2 4 6 8 10 <10 m > 20 m a) radial 1996 Total length (cm) 0 2 4 6 8 R adi al (µ m ) 0 100 200 300 400 500 600 < 10 m depth > 20 m depth logarithmic all

(14)

129

7

The length of the radial has a logarithmic relationship with fish length, signifying

the slowing down of growth with age (Figure 7.6a). For the deep-water (> 20 m) fish, the width of the daily increment increases with age until around the 100th in-crement, after which it decreases (Figure 7.6b, c). Increment width of otoliths from fish caught in shallow water (< 10 m) is smaller than that of fish caught in open water (> 20 m). Shallow water fish also show an earlier peak in increment width than their conspecifics from deeper water (Figure 7.6b). The average increment width was 3.8 ± 2.2 µm, 3.34 ± 0.15 µm and 1.74 ± 0.15 µm in fish caught in deep water in 1992 (n = 30) and 1996 (n = 57), and in shallow water in 1996 (n = 97), respectively.

Table 7.1 Model parameters and their standard error (SE), R2_{(computed in Sigma Plot) and}_AICc

values for the VBGM, Gompertz, logistic and power models. Highlighted are statistically significant (p < 0.05) values of L∞, k and t0, and the best R2 and AICc values (`Did not converge` is when the programme could not make an estimate because data does not fit the model well).

Model L∞ ± SE k / a ± SE t0 / b ± SE R2 _AICc

VBGM 2013 51.9 ± 43.2 0.01 ± 0.01 ‐0.25 ± 0.45 0.89 33 Females Did not converge Males 37.3 ± 44.0 0.02 ± 0.02 ‐0.50 ± 0.50 0.82 1996 12.4 ± 0.9 0.10 ± 0.01 0.21 ± 0.13 0.90 ‐47 1993 25.2 ± 7.3 0.04 ± 0.02 1.21 ± 0.78 0.87 32 All data 53.9 ± 23.9 0.01 ± 0.01 ‐0.65 ± 0.25 0.89 78 Gompertz 2013 15.2 ± 1.4 8.00 ± 0.99 8.53 ± 0.94 0.89 31 Females 17.5 ± 1.9 9.32 ± 1.11 10.42 ± 1.13 0.93 Males 10.6 ± 1.3 5.96 ± 0.95 6.12 ± 0.88 0.84 1996 9.6 ± 0.2 3.10 ± 0.23 3.47 ± 0.13 0.91 5 1993 18.0 ± 1.8 7.62 ± 1.39 8.94 ± 1.09 0.88 ‐13 All data 18.1 ± 1.1 9.37 ± 0.68 9.50 ± 0.72 0.89 80 Logistic 2013 61.0 ± 67.6 ‐1.02 ± 0.15 78.04 ± 118.44 0.89 33 Females 2.2×103_{± 1.0×10}5 _{‐0.97 ± 0.13} _4.3x103_{± 2.0x10}5 _0.92 Males 50.2 ± 90.2 ‐0.96 ± 0.18 78.53 ± 197.80 0.81 1996 12.6 ± 1.1 ‐1.42 ± 0.12 6.70 ± 1.07 0.90 5 1993 25.5 ± 7.8 ‐1.47 ± 0.31 17.40 ± 7.53 0.87 32 All data Did not converge

Power asymptotic a b 2013 0.80 ± 0.09 0.91 ± 0.04 0.89 31 Females 0.67 ± 0.07 0.97 ± 0.04 0.92 Males 0.81 ± 0.09 0.87 ± 0.05 0.82 1996 1.26 ± 0.05 0.76 ± 0.02 0.88 ‐1 1993 0.99 ± 0.22 0.88 ± 0.08 0.86 32 All data 1.01 ± 0.05 0.85 ± 0.02 0.89 77

(15)

130

Figure 7.7 The relationship between the number of increments and total length of L. miodon in a)

2013 juveniles, males & females, b) 1996, c) 1993 and d) 1988, fitted with Gompertz growth models.

Selection of growth models

All the four growth models used fit the data well and explain between 86 and 91% of the variation (p < 0.05; ANOVA). The Gompertz model has the highest R2_{values. This is} also the best model based on the AICc values for the 2013 and 1993 data (Table 7.1). VBGM is the best only for the 1996 data based on the AICc. All parameters estimated with the Gompertz and the power models are significant (t-test, p < 0.05). For VBGM and the logistic model, the estimates for 2013 are not significant, and the estimates of L∞ are high and therefore not reliable for data set comparisons. Results from the AICc criteria vary, but the power and the Gompertz model fit the 2013 data best (Table 7.1). The Gompertz model appears to be the best for making comparisons among years.

Comparison of growth between years using the Gompertz model

The growth trajectories of L. miodon vary among years, with L∞ being highest in

1993 and relatively low in 1988 and 1996 (Figures 7.7 and 7.8). In 1993, 1996 and 2013, L∞ was 18.0, 12.4 and 15.2 cm, respectively (Table 7.1). In 1993, the estimate

for L∞ was larger than the largest fish in the sample, which measured 16.5 cm TL.

There were annual differences in growth of different age groups. Even though L∞ was

lowest in 1996, fish below 10 months of age that year grew distinctly faster than in other

d)1988 Age (months) 0 5 10 15 20 25 30 To ta l l en gt h (c m ) 0 5 10 15 20 c)1993 To ta l l en gt h (c m ) 0 5 10 15 20 b)1996 Age (months) 0 5 10 15 20 25 30 Tot al le ng th (c m ) 0 5 10 15 20

a) 2013 juveniles, females & males

To ta l le ng th (c m ) 0 5 10 15 20 Females Males Females Males Juv eniles

(16)

131

7

Figure 7.8 The relationship between the mean length at age and the age of L. miodon in 1988, 1993, 1996 and 2013 fitted with Gompertz growth model. Table 7.2 Mean length‐at‐age (in cm), standard deviation (SD) and the number of fish (n) analysed for each age group of L. miodon in 1993, 1996 and 2013. Age (months) 1993 1996 2013

Length ± SD n Length ± SD n Length ± SD n

0 1.0 ± 0.1 15 1.4 ± 0.3 3 1 1.6 ± 0.4 30 1.4 ± 0.4 19 2 2.4 ± 0.8 3 2.2 ± 0.8 78 1.5 ± 0.2 13 3 2.3 ± 0.4 7 3.5 ± 0.9 31 2.2 ± 0.5 14 4 2.2 ± 0.2 4 4.0 ± 1.7 2 3.1 ± 1.1 10 5 4.4 1 6.4 1 3.8 ± 1.6 9 6 2.5 ± 0.6 3 7 4.2 ± 0.5 2 8 8.0 1 8.8 1 5.4 ± 2.2 2 9 5.7 1 7.9 1 7.7 ± 3.0 4 10 9.3 ± 4.4 3 6.9 ± 2.2 5 11 5.9 1 8.9 ± 0.2 3 7.5 ± 2.1 6 12 9.1 ± 0.6 5 8.1 ± 0.6 3 13 9.0 ± 4.1 3 9.1 ± 1.6 2 8.7 ± 2.2 5 14 11.8 ± 2.2 4 8.6 ± 0.1 2 15 11.1 ± 3.3 2 10.0 ± 0.9 4 16 11.3 ± 3.8 2 9.6 ± 0.0 2 9.1 ± 1.9 4 17 14.1 ± 1.5 2 11.1 ± 2.3 4 18 15.1 ± 0.1 2 9.0 1 11.9 ± 0.2 2 19 14.0 ± 1.4 2 12.6 ± 1.4 2 20 14.8 1 13.0 ± 0.7 4 23 14.7 ± 1.3 2 24 15.4 ± 1.1 2 26 16.3 ± 0.3 2 Age (month) 0 5 10 15 20 25 30 Le ng th (c m ) 0 4 8 12 16 20 1988 1993 1996 2013 1988 1993 1996 2013

(17)

132

years (Figure 7.8; Table 7.2). In comparison to 1993, L. miodon shows a generally slower growth to a lower L∞ in 2013.

Sex and individual growth comparison

Apart from the annual differences in growth, a difference between sexes was found. All large fish (> 13.6 cm) collected in 1993 and 1994 were females. Furthermore, there were no male fish above 8 cm TL in the samples from 2013. Generally, L.

miodon of the same age group may show a large variation in size (Figure 7.7). Age and length at first maturity

To estimate age at first maturity (A50), we used 104 fish of 0 – 17 months old (Table

7.3). The youngest mature fish observed was a three months old female, and age at first maturity was 8.02 and 7.90 months for females and males, respectively. All fish were mature at 12 months of age (Figure 7.9a). The relation between fish age (a) and the proportion of mature fish (M) is described by the function:

Females M = 100 / (1 + exp (- (a - 8.02) / 1.82)) (R2 = 0.99; p < 0.0001) Males M = 100 / (1 + exp (- (a - 7.90) / 2.64)) (R2_{= 0.79; p = 0.0031)} A total of 173 fish of 1.0 – 8.0 cm TL were used to estimate length at first maturity (L50)

(Table 7.3). The smallest mature fish was a female of 2.8 cm long, and all specimens of length classes 6 – 6.9 cm and above were mature. The estimated length at first maturity was 3.43 and 3.63 cm for females and males, respectively. Figure 7.9b shows the sigmoid relationship of length (L) and % maturity (M) described by the functions:

Females M = 100 / (1 + exp (- (L - 3.43) / 0.38)) (R2_{= 1.0; p < 0.0001)} Males M = 100 / (1 + exp (- (L - 3.63) / 0.44)) (R2_{= 1.0; p < 0.0001)} Discussion

Validation of increments

When examined under TEM, decalcified otoliths of L. miodon show alternating dark (opaque) and light (translucent) increments. We assume that the different increments represent the cyclic deposition of organic and mineral material, which agrees with the results of other studies. Demineralised trout otoliths are composed of proteins (48%), collagens (23%), and proteoglycans (29%), and some carbohydrates and lipids which remain after decalcification (Borelli et al. 2001; Payan et al. 2004; Guibbolini et al. 2006). In our study, the dark increments were deposited during the day, and most of the light ones at night. Similar results were obtained by Takagi et al. (2005) who suggest that formation of the dark increments in rainbow trout (Oncorhynchus mykiss) is caused by increased synthesis of the otolith matrix protein otolin-1 (Murayama et al. 2002) in the early hours of daylight.

(18)

133

7

Table 7.3 The numbers of immature and mature fish used to estimate the proportion of mature fish

in a sample of L. miodon caught in the Sanyati basin.

Size group Immature Mature Total % Mature

Age class (months) 0 – 1.9 22 0 22 0 2 – 3.9 26 1 27 4 4 – 5.9 12 7 19 37 6 – 7.9 3 2 5 40 8 – 9.9 1 4 5 80 10 – 11.9 4 6 10 60 12 – 13.9 0 6 6 100 14 – 15.9 0 5 5 100 16 – 17.9 0 5 5 100 Totals 68 36 104 Length class (cm) 1.0 – 1.9 39 0 39 0 2.0 – 2.9 26 1 27 4 3.0 – 3.9 17 10 27 37 4.0 – 4.9 4 36 40 90 5.0 – 5.9 1 15 16 94 6.0 – 6.9 0 6 6 100 7.0 – 7.9 0 7 7 100 8.0 – 8.9 0 11 11 100 Totals 87 86 173 The lighter increments of decalcified otoliths of L. miodon contain a lower density of materials than the darker ones, representing diurnal fluctuation in the deposition of matrix proteins. Our results suggest that more proteins were deposited during daytime, which agrees with the observed periodicity of protein concentrations found in the endolymph of other fish species (Edeyer et al. 2000); Borelli et al. 2003). Trout (Onchorhynchus mykiss) show increasing protein levels in the endolymph after the dark–light transition, peaking approximately 5ꞏh after the beginning of the light period (Borelli et al. 2003). Edeyer et al. (2000) reported anti-phasic fluctuation of protein in the endolymph of turbot (Psetta maxima) with higher protein concen-trations during daytime.

Our study shows that the increment at the edge of a decalcified otolith of L.

miodon, viewed under a TEM microscope, can be used to validate growth increments

of otoliths. The periodicity of protein and calcium deposition follows a diurnal cycle. Hence, a pair of light and dark increments can be taken as a daily band and be used to age the fish.

(19)

134

Figure 7.9 The percentage mature female and male L. miodon in each a) age class (n = 104) and b)

length class (n = 174) in the 2013 data set, with the age and length at first maturity indicated by arrows at the point where 50 % of the fish are mature.

(20)

135

7

Increment width and ageing

Ontogenetic change in otolith shape is a genetically programmed process that can be influenced by the environment (Gagliano & McCormick 2004; Hüssy 2008). This change in shape makes it difficult to count increments on every radial of the otolith, because of distortion and compression of increments. Studying L. miodon from Lake Tanganyika, Kimura (1995) observed that in some specimens the increments on the rostrum and the antirostrum become indistinct at the margin of the otolith. That diminishes the suitability of these radials for ageing. Increment width tends to decrease towards the margin of the otolith, making the ageing of old fish relatively difficult (Kimura 1995).

In Lake Kariba, recruitment of L. miodon into the fishing ground is based on size rather than on age. Fish of different length are spatially segregated, making it easy to control the size caught in the fishery, by restricting the minimum fishing depth for commercial fishing to 20 m (Cochrane 1984; Mtsambiwa 1989). However, some of the small fish caught in shallow water appeared to be up to five months old, the age at which larger larger conspecifics already recruited into the commercial fishery. This means that fast-growing L. miodon are recruited into the offshore fishery at a younger age than slow-growing individuals, that remain in the lake margin for a longer period. Such a process causes a relatively high mortality risk in the fast-growing part of the population. In an experiment using Atlantic silverside (Menidia

menidia), Conover & Munch (2002) demonstrated how selective fishing can cause

a genetic change from fast- to slow-growing individuals. This process may have contributed to the relatively slow growth of L. miodon in Lake Kariba in 2013, as compared to earlier years, which can be explored in future studies.

Selection of growth models

Based on the significance levels reached when fitted to our data, the von Bertalanffy (VBGM), the Gompertz, the logistic and the power growth model, can all satisfac-torily describe growth independently. Without considering other possible models, the VBGM has been used to describe the growth of L. miodon in Lakes Kariba and Tanganyika (Marshall 1987a; Chifamba 1992; Kimura 1995). When compared to other models, the VBGM is not necessarily the best model to describe the growth of multiple species, or intraspecific growth patterns in different time periods (Katsa-nevakis 2006; Chapter 3 – Chifamba & Videler 2014). The power and logistic models were mostly the best to describe the growth of the tilapiine cichlids Oreochromis

niloticus and O. mortimeri in Lake Kariba, using the AICc criteria (Chapter 3 – Chifamba & Videler 2014). This finding demonstrates the importance of selecting the model that best fits the available data, to improve confidence in the parameters. Even though all models used describe our data well, some of the parameter estimates have large standard errors and are statistically insignificant. The estimates of L∞ for

(21)

un-528794-L-bw-Chifamba 528794-L-bw-Chifamba 528794-L-bw-Chifamba 528794-L-bw-Chifamba Processed on: 6-2-2019 Processed on: 6-2-2019 Processed on: 6-2-2019

136

acceptable given the maximum size of the fish encountered in this study (16.5 cm). The Gompertz model is the best with respect to R2_,_{AICc and significance of the} parameter estimates. Our results show that applying symptotic growth models to un-suitable data (such as our 2013 dataset) that visually suggests an asymptotic model, can give unrealistic results.

Temporal and spatial variation in growth

Growth rate of L. miodon varies between years, between lakes (Tanganyika, Kivu and Kariba), and within Lake Tanganyika (Spliethoff et al. 1983; Kimura 1995) and Lake Kariba (this study). In Lake Kariba, VBGM estimates of growth rate for 1976 (k = 0.145) and 1981 (k = 0.254) were higher than in this study (Cochrane 1984; Marshall 1987a). Many studies have shown that temporal and spatial variation in the growth rate of fish is associated with environmental conditions such as temperature and food availability (Kimura 1995; Yasue & Takasuka 2009; Pörtner & Peck 2010; Itoh et al. 2011). For example, a positive relationship between sea temperature and growth of larval Japanese anchovy Engraulis japonicus in the Kii Channel was reported by Yasue & Takasuka (2009). A deviation from the optimum temperature for larval growth in E. japonicus and Japanese sardine (Sardinops melanostictus) caused a shift in their relative abundance. This resulted in an alternation between a warm anchovy and a cool sardine fishery in the western North Pacific (Takasuka et

al. 2007). Therefore, the observed temperature rise in Lake Kariba can affect the

growth of L. miodon either directly or indirectly, through its effect on plankton. Rising temperature in Lake Kariba is thought to have caused changes in the plankton community, productivity, timing of stratification and the depth of the epi-limnion, which may affect the quantity and quality of food for L. miodon (Chifamba 2000; Magadza 2011; Mahere et al. 2014). A similar consequence of warming was found in Lake Tanganyika, where the water circulation is reduced and hence, nutrients get locked up in the hypolimnion (O’Reilly et al. 2003; Verburg et al. 2003). Temperature, lake turnover and nutrient availability determine the composition of the phytoplankton, with cyanobacteria associated with higher temperatures (Magadza 1980; Ramberg 1987; Zhang & Prepas 1996; Cronberg 1997; Murrell & Lores 2004; Dalu et al. 2018). The ability of cyanobacteria to grow at higher temperatures than chlorophytes, was confirmed in the laboratory (Sibanda 2003; Butterwick et al. 2005). Cyanobacteria are a poor food source for zooplankton, compared to other groups of phytoplankton, because of their toxicity and morpho-logy (DeMott et al. 1991; Wilson et al. 2006). An experimental study on the growth of larval North Sea cod showed that diatoms can affect the growth rate significantly, due to the nutrition they provide in the form of essential fatty acids (St. John et al. 2001). Therefore, the decrease in phytoplankton quality and the rise in temperature in Lake Kariba may explain the faster growth of juvenile and subadult L. miodon in 1976, 1981, 1988 and 1996, in comparison to 1993 and 2013. There is a need to

(22)

137

7

directly test these linkages in Lake Kariba, in order to draw firm conclusions on the

relationships between growth, population size and environmental variables.

Fast growth may be accompanied by a low asymptotic length. Such a pattern was found for L. miodon caught in 1996. Fish younger than 10 months grew much faster than those caught in 1993 and 2013, yet to a much lower asymptotic length. A similar change in growth pattern was observed in another small pelagic species, the zoo-planktivorous cyprinid Rastrineobola argentea from Lake Victoria (Wanink et al. 1998). In 1988, R. argentea grew at a faster rate than in 1983 but to a smaller final size, showing a flexibility in growth pattern comparable to that observed for L.

miodon in this study.

Chifamba & Videler (2014 – Chapter 3) showed that different growth models tend to give either high or low parameters, and that length at age can be used to compare growth among years. Our length at age analysis indicates slow growth rates in 1993 and 2013, as compared to 1988 and 1996, which could have contributed to a lower fish production. In agreement with growth patterns reported for Lake Tanganyika (Kimura 1995), we found that female L. miodon from Lake Kariba grow to a larger size than males. In 2013, the largest males were 8.6 cm long, while females reached a maximum size of 13.9 cm. In the other years, samples with large fish (> 13.4 cm) did not contain males. Being large is beneficial to females, as fecundity increases with body size, resulting in increased fitness (Parker 1992; Magurran & Garcia 2000).

Size at first maturity

The size at first maturity of L. miodon in Lake Kariba decreased substantially, from 5.2 - 5.6 and 7.1 - 7.3 cm fork length in 1970 (Woodward 1974) to 3.43 and 3.63 cm total length in 2013 (this study) for females and males, respectively. The reported decline in the age at first maturity in Lake Kariba is confirmed by changes in the minimum length of mature fish (from years in which size at first maturity was not estimated). This decreased from 4.0 cm in 1975 - 1976 (Cochrane 1984) to 3.5 cm in 1981 - 1983 (Marshall 1993), then to 2.8 cm in this study. Comparable reductions in the smallest size at first maturity were observed in the Lake Victoria minnow (Rastrineobola argentea), Pacific salmon (Oncorhynchus spp), North Sea sole (Solea solea) and Atlantic cod (Gadus morhua). They have been attributed to high adult mortality, caused by fishing and predation (Wanink et al. 1998; Hutchings 2005; Olsen et al. 2005; Mollet et al. 2007; Morita & Fukuwaka 2007). By studying guppies (Poecilia reticulata) in either a high- or a low-predation environment, and manipulating mortality rates in nature, Reznick & Ghalambor (2005) proved that high adult mortality can reduce size and age at maturity. Therefore, the observed decrease in the size at maturity of L. miodon in Lake Kariba is consistent with fisheries-induced evolutionary change, as predicted by life history theory, where

(23)

138

fitness is optimized by increasing reproductive effort and decreasing size and age at maturity in response to high adult mortality (Stearns & Koella 1986).

The observed sizes at maturity in Lake Kariba in 2013, and before the start of the fishery in 1970, are smaller than those in Lake Tanganyika (7.5 and 6.4 cm fork length, for females and males, respectively) and Lake Kivu (6.24 and 6.06 cm fork length for females and males, respectively) (Ellis 1971; Spliethoff et al. 1983). Ellis (1971) found that maturity varied within Lake Tanganyika, not only for L. miodon, but also the lake’s other pelagic clupeid, Stolothrissa tangnicae. Ndebele-Murisa et

al. (2010) reported differences in plankton composition and a lower range of

phyto-plankton production in Lake Kariba (0.10 – 1.76 g C m2_h-1_{) than in Lakes Tanganyika} (0.16 – 4.30 g C m2_h-1_{) and Kivu (0.85 – 2.20 g C m}2_h-1_{), which may explain the} differences in the size at maturity found between these lakes.

We conclude that fishing pressure and environmental variation (temperature increase in particular) cause annual variation in the growth rate of Limnothrissa

miodon, and a decrease in the size at maturity, reflecting the species’ flexible

life-history strategy. Our study shows that growth of L. miodon varies yearly. To capture and understand these variations and their impact on the fishery, we recommend continuous monitoring of life-history and environmental parameters such as temperature, nutrient availability, plankton composition and productivity.

Acknowledgements

Fish were sampled using the fishing vessels belonging to Lake Kariba Fisheries Research Institute (LKFRI), University Lake Kariba Research Station and Mash Fishing Enterprise, for which we are thankful. The analysis of otoliths for validation was done at the Electron Microscope Unit of the University of Zimbabwe. We thank Mr Claudius Mutariswa and Mr Patrick Kurangwa for assisting in the preparation, reading and photographing of the otoliths. We are also grateful for the use of the otolith reading equipment at LKFRI. Mr Paul Daley of Zambezi Proteins provided the 1994 sample of large fish used in the study. We are indebted to the late Joel Chisaka, the boat crews, and others who assisted in many aspects of the project. Financial support from Nuffic - NFP-PhD.11/ 858 enabled data collection in 2012 and 2013, and the stay of Portia Chifamba in The Netherlands for data analysis and thesis writing.