Effects of inter-annual variability, seasonality and persistence on the perception of long-term trends in catch rates of the industrial pelagic purse-seine fishery of northern Lake Tanganyika (Burundi)

Abstract

Industrial and artisanal fisheries exploit the pelagic fish community of northern Lake Tanganyika with total annual landings fluctuating between 16,000 and 30,000 t since 1974. The pelagic catch consists of the commercial categories ‘small’ (<20 cm) ( Stolothrissa tanganicae, Limnothrissa miodon and juvenile Lates stappersii), ‘medium’ (20–40 cm) ( L. stappersii) and ‘large’ (>40 cm) (three Lates spp.) sized fish. Thirty-seven years (1956–1992) of monthly catch rates by the industrial purse-seine fishery were evaluated for the capacity of fisheries management authorities to detect possible long-term trends given large variability around such a trend due to inter-annual variability, seasonality and persistence. Variability in 10log-transformed total catch rates (kg/vessel/month) was high ( s=0.22), but considerably lower than the variability for the commercial categories ‘small’ ( s=0.32), ‘medium’ ( s=0.59) and ‘large’ ( s=0.69). Long-term downward trends in catch rates were significant for catch rates as total catch (1.6% per year, r 2=0.21) and for the categories ‘small’ (1.4% per year, r 2=0.08) and ‘large’ (2.7% per year, r 2=0.21). Seasonality, confounding the perception of long-term trends in monthly catch rates, although with temporal predictability, explained between 3 and 10% of the total variance in monthly catch rates. Significant short-term persistence in the residuals of series, adjusted for between year and seasonal differences, and as described by ARMA models, must be due to fleet–stock interactions. Further obscuring trends, this persistence explained 5, 11, 13 and 3% of the variance in total catch rates and of the commercial categories ‘small’ and ‘medium’ and ‘large’. Basic uncertainty, as variance not explained by trend or seasonality, comprises the largest part of the variance in catch rates (‘total’ (75%), ‘small’ (83%) and ‘medium’ (89%)). Trend-to-noise ratios ( b/ s) were calculated with the slope ( b) in 10log-transformed monthly catch rates and the standard deviation in the residuals around the regression line ( s). These ratios were high, necessitating long series of data to detect significant long-term trends. Faced with such large uncertainties and compounding effects of the various types of variance around long-term trends, the Burundi management authorities of the Lake Tanganyika fisheries have few possibilities to evaluate the effectiveness of, for instance, controlling fishing effort as a direct and basic management measure.