Long-term Trends in Coral Reef Fish Yields and Exploitation Rates of Commercial Species from Coastal Kenya

Analysis of long-term (1978–2001) marine fisheries data showed that Kenyan coralreefs produced an estimated 2–4 metric t/km 2 year of demersal fish. A rapid overall decline in landings occurred during the 1990s. Yields (t/km /year) showed bimodal peaks in 1982 (2.98) and 1991 (2.90). The average total landings dropped by 55% during the last decade following peak landings in 1982. Landings of the commercially important families (e.g., Siganidae, Lethrinidae, Lutjanidae and Serranidae) declined by about 40% during the last decade, with the groupers (Serranidae) showing the steepest (72%) decline. Analysis of landings per administrative district showed a 78% decline in the densely populated Mombasa district between the periods 1983–1991 and 1992–2001. The less populated districts have registered stable (e.g., Kilifi) to increasing (e.g., Kwale) catches over time. An autoregressive moving average (ARIMA) model forecast of landings predicted a gradual decline in catches during the next decade (2002– 2011) with a trend slope of -0.01 t/km . Length-frequency analysis for the commercially important species indicated above optimum exploitation (E) and fishing mortality (per year) rates for the sky emperor, Lethrinus mahsena (E = 0.64; F = 2.48) and lower but strong rates for the emperor, L. sangueinus (E=0.51; F=0.93). The more abundant and commercially important whitespotted rabbitfish, Siganus sutor, showed equally strong rates (E= 0.56; F = 1.44/year). A precautionary approach in the management of Kenya’s coral-reef fisheries is recommended.


INTRODUCTION
Fishing is the dominant extractive activity in Oceania and an important source of income and sustenance in coastal communities worldwide.However, in the past decade many marine fisheries resources have declined (FAO, 1995).Although ocean climate variation has likely played an important role in many regional declines (Lauck et al., 1998;Drinkwater & Mountain, 2002), the most important factor has been overfishing (Pauly & Christensen, 1995;Hutchings, 2000;Rose et al., 2000).The effects of fishing have been the subject of recent reviews (Jennings & Polunin, 1996;Jennings & Kaiser, 1998).Growth overfishing reduces the size and yield of target species (Koslow et al., 1988;Russ, 1991;Munro, 1983), recruitment overfishing reduces the recruitment success of populations (Jennings & Lock, 1996) while ecosystem overfishing alters species interactions and habitat quality (McClanahan, 1995).Marine fishes have been thought to be resilient to these effects (Musick, 1999), but such resilience has likely been overstated, and will depend on the degree and frequency of impact and the life history traits of target species (Sadovy, 2001).In developing countries, the effects of fishing are exacerbated by inadequate resources available to management agencies and increasing human populations (McManus, 1997).The outdated notion that fisheries resources are unlimited, often reinforced by increasing annual catches, has led to overfishing of many tropical fisheries resources (Pauly et al., 2002).
In East Africa, coral reef fisheries have a long history (Brochman, 1984).In Kenya, reef fisheries are exploited by approximately 8000 artisanal fishermen.These fishermen mostly use traditional dugout canoes on grounds that include most of the lagoons between shore and the fringing reefs (Brochman, 1984).East African reefs are thought exploited at sustainable levels or to be somewhat overfished (McClanahan & Obura, 1995).However, there are no estimates of yields or exploitation rates for these fisheries, largely as a consequence of inadequate landing statistics (UNEP, 1998).In other tropical reef systems, few estimates of long-term yields of coral reef fisheries have been reported (Dalzel, 1996;Jiddawi & Stanley, 1999;Mapya et al., 2002), but crossregional comparisons may be problematic because of local variations in reef areas and types, depth and fishing effort (Russ, 1991).Recent increases in human population in East Africa (e.g., 4.2% per year in Kenya) have likely increased the demand for marine food fish.However, the effects of any such increased demand on reef fish resources have not been quantified.
Studies on long-term yields of coral reef fisheries are few (Mapya et al., 2002), but can provide important information to management on the state of stocks.In this paper, we document longterm trends in the yields of coral reef-associated fish families in Kenyan waters of the western Indian Ocean.We also provide estimates of exploitation and fishing mortality rates for the principal commercial species, and develop a time series model of landings with long-term projections.

MATERIALS AND METHODS
The Kenya Fisheries Department routinely collects fish landing statistics along Kenya's approximately 600-km-long coastline (Fig. 1).Fish scouts who are supervised by Fisheries Assistants record landed weight of fish by taxonomic families at designated landing beaches.The data from all landing sites within the administrative units or districts (e.g., Mombasa, Kilifi, Kwale and Lamu) are compiled into a national annual statistical bulletin.Landing data contained in the annual statistical bulletins from 1978 to 2001 were analysed in this study for reef-associated families (Lethrinidae, Lutjanidae, Siganidae, Scaridae, Acanthuridae, Serranidae, and an 'others' category).Catches of non-reef pelagic families (mostly Clupeidae and Carangidae) are highly variable and were excluded.
Exploitation levels and fishing mortality rates were derived from length-frequency data obtained from commercial trap landings.At bi-weekly intervals from May 2000 to April 2 trained assistants sampled trap fish at four active fish landing sites adjacent to Watamu and Malindi Marine National Parks (Fig. 1).Total length (to the nearest mm) of random samples of landed fish species, number of traps per fisherman, fishing ground and the fisherman's name were recorded.Length-frequency data of the commercially important fish species (the whitespotted rabbitfish, Siganus sutor, the sky emperor, Lethrinus mahsena and the trumpet emperor, L. miniatus) were used to estimate exploitation and fishing mortality rates and population growth parameters.

Data analyses
It is estimated that artisanal fishermen in coastal Kenya operate within an area of approximately 800 km 2 of lagoons that extend between the shore and continuous fringing reefs at depths less than 5 m at low tide (McClanahan & Obura, 1995;UNEP, 1998).This reef area was used to standardise catches (tonnes/km 2 ).Time series of landings (in tonnes/km 2 /year) were used to analyse temporal trends in catches of the major demersal families of coral reef fish (e.g., Siganidae, Lethrinidae, Lutjanidae, Scaridae, Acanthuridae, Serranidae and 'others').
A locally-weighted scatterplot smoother (LOWESS) (Cleveland, 1979) was used to fit smoothed trend lines to the full data series using the MINITAB package.The LOWESS is based on a weighted least squares algorithm that gives local weights the most influence while minimising the effects of outliers (Cleveland, 1979).A smoothness parameter (f) of 0.2 was found to adequately smooth the data without distorting the main temporal patterns.Additionally, an autoregressive integrated moving average (ARIMA) model (Box & Jenkins, 1976) was used to forecast landings for the next decade (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011).An ARIMA (111) model that integrates first order autoregressive (AR) and moving average (MA) model parameters with first differencing of the annual catches (Rothschild et al., 1996;O'Donovan, 1983) was used to forecast catches as: where, y t is the first difference of the catches at  The ARIMA model is considered parsimonious in analysing data whose underlying structure is unknown and with individual observations that are prone to error (Box & Jenkins, 1976).The longterm landing data were divided into 3 time intervals (1978-1982, 1983-1991 and, 1993-2001) based on a preliminary assessment of the temporal pattern of landings.A one-way ANOVA was used to test for significant differences in landings (metric tonnes) between the 3 time periods.All data were first tested for normality and homoscedasticity using Kolmogorov's test and Levene's test, respectively, (Zar, 1975).Where variances were significantly different within the families (e.g., Serranidae and Scaridae) and in the districts (e.g., Mombasa), the catch was first log e -transformed before ANOVA was performed.Where the mean catch between the time periods were found to be different (ANOVA, P < 0.05), a Bonferroni test was used for post hoc analysis.
Exploitation level (E) and fishing mortality rate (F) and growth parameters (e.g., instantaneous annual growth rate, K, and the asymptotic length, L ∞ ) were estimated for the three major commercial species using length-frequency analysis (LFA).LFA was carried out using length-based routines in the FiSAT package (Gayanilo et al., 1995).In order to increase the modal sizes, length-frequency data were pooled tri-monthly for the sky and the trumpet emperors for the period May 2000 to January 2002.Monthly length-frequency (June 2000 to March 2002) data were analysed for the more abundant whitespotted rabbitfish.Having obtained estimates of the growth parameters (K and L ∞ ) from ELEFAN I sub-package in FiSAT, ELEFAN II was used to estimate instantaneous total (Z) and natural mortality (M at 27 ºC) rates from linearised length-converted catch curve and Pauly's empirical formula (Pauly, 1980(Pauly, , 1984)): where T is the annual sea surface temperature (27 ºC).F was then obtained from the difference between Z and M. The exploitation rate (E) for each of the three species was derived from the ratio, F/ Z, (Gulland, 1971).The exploitation rate indicates whether the stock is lightly (E < 0.5) or strongly (E > 0.5) exploited, based on the assumption that the fish are optimally exploited when F = M or E = 0.5 (Gulland, 1971).

Total catch
The long-term total landings of demersal coral reef fishes averaged 2.11 ± 0.49 t/km 2 /year.Annual landings peaked in 1982 (2.98 t/km 2 /year) and 1991 (2.90 t/km 2 /year) (Fig. 2a).Following the peak in 1982 the catches dropped by 23% to 2.3 in 1984 and remained relatively stable during 1983-1991 when landings averaged 2.53 ± 0.20 t/km 2 /year (Table 1 and Fig. 2a).Following this period of relative stability, annual landings declined by 30% from 1991 to 1992 (2.03), with a further 25% decline in subsequent years to the lowest levels,

Forecast landings (2002-2011)
Based on ARIMA (111) model, the combined catches of all the families were used to forecast total landings for the next decade following the last recorded landings in 2001 (Fig. 4).The model generated for this forecast was: y t = -0.007-0.058y t-1 + a t + 0.950a t-1 (4) The model output indicated increased landings in 2002 (1.71.3 t/km 2 /year) following the recorded catches in 2001 (1.3 t/km 2 /year).Subsequent forecasts predicted a consistent decline in catches to the year 2011 with a shallow slope of -0.01 t/ km 2 (equation 2) (Fig. 4).The 95% prediction limits indicated greater confidence in the forecasts during the next 4 (2002)(2003)(2004)(2005), with decreasing confidence (increasing confidence interval) in subsequent years.

DISCUSSION
The data indicate that a rapid overall decline in landings occurred in coastal Kenya during the 1990s.The decline was most severe at Mombasa (78%) which contributes > 40% of coastal landings and was less evident in some districts (e.g., Kwale).Environmental conditions could play a role in declining catches, but there is little evidence of large-scale change in climate along the East African coast in recent times.Annual human population growth rate in Kenya is estimated at 4.2% and the rate is thought to be higher in coastal towns (UNEP, 1998).Population-driven demand for food and   3. b Fishing mortality and exploitation rates are above optimum levels for the commercially exploited species in these families.It is likely that sustainable yield levels have been exceeded.The lack of confidence limits of the FiSAT program parameter outputs (e.g.K and Z) makes it difficult to assess the precision of the estimated mortality and exploitation rates and is a major drawback of the program (J.L. Munro, pers. comm.).However, precise estimation of sustainable yields is likely to be difficult due to the many landing sites, many gears and inappropriate records, a problem that is common for most tropical stocks (Russ, 1991).
It is noteworthy that the landings of the 'others' category is showing an increasing trend in recent years (Fig. 2h), and this has likely been caused by increased marketing of low trophic level species formerly considered less valuable.In coastal Kenya, the fisheries have shifted from hand-line capture for the large groupers to drive net, seine and gill net capture of low trophic level groups (e.g.Labridae and Acanthuridae).Similar results showing changes in reef fish community structure attributed to fishing have been reported for fished reefs in Jamaica (Koslow et al., 1988) and the Seychelles (Jennings & Kaiser, 1998).
Amongst the families studied, the groupers (family Serranidae) showed an earlier and a steeper decline (~72%) in catches compared to the other groups.The groupers may have suffered high fishing mortality due to their sedentary life and the tendency to form spawning aggregations in some species (Thomson & Munro, 1983), which makes them highly vulnerable to exploitation.It is likely that some serranids have suffered spawning collapse because of reduced population levels.Furthermore, fishing may skew sex ratios in protogynous (maturing as females) groups such as the Serranidae (Thomson & Munro, 1983), thereby causing spawning failure.Elsewhere, the Nassau grouper, Epinephelus striatus, is commercially extinct in the Bermuda fishery (Luckhurst, 1996).The gonochoristic (separate sex) families (e.g., Lutjanidae, Lethrinidae, Siganidae) are likely to be more resilient under high fishing effort and the different reproductive ecology and behaviour may have allowed these groups to maintain stock levels without suffering dramatic declines as in the Serranidae.The Siganidae and the Lethrinidae form the bulk (~40%) of the artisanal landings in coastal Kenya, and it is noteworthy that their landings declined by 40% during the 1990s.The overall decline in landings is likely influenced greatly by the decline in the yield of the two groups.However, declines in all the major demersal fish families (e.g., Siganidae, Lethrinidae, Lutjanidae and Serranidae) have occurred over the last decade.
Earlier reviews of coral reef fish yields (Marshall, 1980) concluded that yields range from 0.8-5 metric t /km 2 /year.However, studies of reeffish yields from American Samoa and the Philippines (Wass, 1982;Dalzel, 1996;Mapya et al., 2002) have documented yields in the range of 8-27 t /km 2 /year.Our analysis of reef fish landings indicate that Kenyan reef-fish yields averaged 2.1 t/km 2 /year.However, this figure is likely conservative given that about 40% of the landings may not be reported (FAO, 1985).Therefore, potential yields may actually average 3-4 t/km 2 / year.Yield estimates from different regions may not be strictly comparable because of the variations in reef area and type, depths, fishing intensity (Russ, 1991) and fish assemblages.
The continued decline in marine fisheries production may have been downplayed by the relatively small role of the marine fishery in national fisheries production; of Kenyan total annual fishery production, only 7.4% comes from marine waters (UNEP, 1998).There are a number of physical, climatic and economic factors that combine to constrain the marine fishery in Kenya.Firstly, the area of the continental shelf, to a depth of 200 m, is only about 8500 km 2 , less than 10% of the fishable area of Lake Victoria (the largest freshwater lake in the country with landings >100,000 MT/year).Secondly, the South-East Monsoon which is prevalent from March to October is associated with very strong winds and rough currents.These conditions constrain the use of small dug-out canoes which are the main fishing craft.Thirdly, the East African coast does not have high productivity due to the oceanic origin of the coastal currents and lack of major upwelling areas (Hamilton & Brakel, 1984).Despite these constraints and the low contribution to national production, marine fishery resources are locally important sources of livelihood and protein for coastal communities.
The conventional methods of regulating a fishery are often difficult to enforce in developing countries (McManus, 1997).Recent interest has focused on the use of marine reserves to complement traditional fisheries management methods.Marine reserves are straightforward and cost effective tools for managing overutilised resources in developing countries.Kenya has four national marine parks, and although these comprise less than 5% of the total reef area, they have been found to effectively conserve biodiversity and local fish biomass (McClanahan & Kaunda-Arara, 1996).Reserves, appropriate traditional reef management methods (Johannes, 1978), improved fisheries data collection and effective gear regulation are complementary measures needed to manage overexploited stocks.
In conclusion, the trend analysis has shown that demersal coral-reef fish production has declined by over 50% for the total catches and 78% in the densely populated district (Mombasa) during the last decade.The principal commercial species are likely being exploited beyond optimal levels, and our forecast predicts a gradual decline in landings during the next decade.However, model forecasts are only instructive in terms of the direction of change while absolute forecast values are likely unreliable over a long-term (J.Wroblewski, pers.comm.).The causes of the declining trends cannot be determined from the present data, but human population driven increase in fishing effort as a result of increased demand for food may have played an important role.More detailed analysis will require improvement of the data collection system to include species records, catch per gear and the associated fishing effort.

Fig. 1 .
Fig. 1.Kenyan coastline showing the major fish landing sites and marine national parks

Fig. 4 .
Fig. 4. Forecast landings of demersal coral reef fish in coastal Kenya for the next ten years (2002-2011).The middle line represents the mean forecast values while the upper and lower dashed lines are the corresponding 95% confidence intervals.

Fig. 6 .
Fig. 6.Linearised length-converted catch curves for estimating annual instantaneous total (Z) and natural mortality rates (M at 27 ºC) for (a) Siganus sutor, (b) Lethrinus mahsena from exploited reefs in coastal Kenya.Data are fitted using ELEFAN II package in FiSAT.N is number of fish in length class i and dt is time needed for fish to grow through the length class, sample sizes as in Fig. 5. Closed circles (•) used in the regression, open circles (o) not used.

Table 3 . Annual exploitation (E) and mortality (Z,M,F) rates and population growth parameters (L ∞ , cm, and K/year) of three commercially important coral reef fish species from coastal Kenya derived from length-frequency analysis of fisheries landing data from May 2000 to January 2002. Parameters are explained in the text
Woodland (1990)),b Carpenter & Allen (1998)