Previous Page Table of Contents Next Page


2.1 Data: origin and treatment

Effort data are based on two frame surveys: one conducted in 1992 (Aarnink, Kapasa and Zwieten, 1993; Zwieten, Aarnik and Kapasa, 1995; Zwieten et al., 1996) and a second one in 1997 (Goudswaard, 1999). Data from earlier surveys conducted between 1955 and 1986 presented here are discussed and referenced in Zwieten, 1995. All surveys only covered the Zambian part of the Mweru-Luapula fishery, which encompasses around 51 percent of the lake and river area. Little is known about developments in fishing effort on the Congolese side of the lake, but there are indications from local Congolese fisheries authorities that the level of effort in terms of numbers of fishermen and boats is the same or even higher compared to Zambia (Anon., 1996; Goudswaard, 1999). Both frame surveys in the 1990’s were not limited to counting of fishermen, boats and fishing activity needed for an estimate of total catch, as part of the Zambian Catch and Effort Data Recording System (CEDRS) (Bazigos, 1974; Bazigos, 1975a, 1975b). They were designed as well to give additional information on gears, spatial and temporal activity patterns, migration within and outside the lake area and demographic patterns in the fishery: age distribution, ethnic origin, ownership of boats and gear, etc. The surveys distinguished between fishermen who are the owners of boats and gears and assistants (crew). The frame-survey of 1997 contained questions regarding the most important occupation before investing in fishing gear and the year the fisherman started (i.e. owning his own gear). From these data we can infer what economic sectors fishermen came from before taking up fishing, and how long they have been fishing. Subsequently, the proportion of the total number of fishermen entering into the fishery in a certain year could be derived from this. By assessing these numbers against the net increase in numbers between 1992 and 1997, we obtained an indication of the number of fishermen moving in and out of the fishery. In the surveys further questions were asked on the birthplace and the home village of the fisherman. From this information indications on both migration into the fishery and migration within the fishery could be deduced.

Catch data were obtained from the Department of Fisheries, Statistical Section in Chilanga, and are calculated from data obtained through the CEDRS (Bazigos, Grant and Williams, 1975a) of Lake Mweru. This CEDRS is designed as a boat based stratified random sampling system. The Mweru-Luapula fishery is divided into four major strata (Figure 2) each of which is subdivided into four minor strata. In each minor stratum three landing sites are chosen at random and sampled for three consecutive days each, according to a strict protocol. This procedure, called Catch Assessment Survey (CAS), is repeated between one to three times each year depending on the funds available. Average catch per boat is then calculated by major stratum. These figures are subsequently multiplied by the total effort expressed as number of boats and by the activity levels, both obtained through the Frame Survey, and added to obtain a total catch for the lake. In case a CAS is only carried out once in a year or in case of missing strata, the total catch figure of that year usually is weighed with data from the previous year. Though the CAS contains information on separate species or species groups, these data are not analysed. Furthermore, it would be possible to derive the error structure of the data sampled from individual boats, but this analysis is not done as well. This shows that the data are underutilized, with only limited evaluation with regard to their quality (Zwieten, Njaya and Weyh, 2003). However, the error in the total catch estimate is deemed to be around 15-20 percent (Lupikisha, pers. com.).

Water level data were obtained from the Department of Hydrology, Lusaka and our own measurements, both taken from the gauge at Nchelenge. Missing years could be interpolated by a correlation with data from Lake Bangweulu and from rainfall data. The latter two data sets were obtained from the Department of Hydrology in Lusaka.

Experimental gillnet surveys and the construction of biomass-size distributions

Experimental surveys with a fleet of multifilament gillnets ranging from 25 mm to 178 mm stretched mesh with 13 mm increments were carried out by the Department of Fisheries from 1970 to 1972, 1982 to 1985 and from 1993 to 1999. Sampling sites during the latter two periods were identical while during the first period the number of sampling sites was both larger and overlapping with later surveys. All fish or subsamples of fish in case high numbers were caught were measured as standard length (SL)[4], and weighed and sexed. The data set was digitized in PASGEAR[5] (Kolding 1999). Numbers of fish were corrected for the amount of effort of each net as 100 m2 of gillnet of a mesh size set during three years at all sampling sites. Numbers of fish caught were further corrected for the selectivity of the fleet of gillnets and for the proportional difference in number of settings at the sampling stations in the three periods. Selectivity curves were estimated with the aid of the selectivity module in PASGEAR that implements methods of estimation based on Millar, 1992; Millar and Holst, 1997 and Millar and Fryer, 1999.

Biomass-size distributions were constructed as follows. Length-frequencies of 1 cm classes were made with the corrected numbers of all species and for each length class the relative biomass per species was calculated. All weights by length class were subsequently added. The resultant biomass-size distributions thus are in fact catch rates of 100 m2 of gillnets of all fish caught by one centimetre length class, hence are an index of the actual biomass by length class. As the methods of sampling and sampling sites have not changed over the three periods examined, the distributions can be compared directly with each other. Changes in the distribution thus reflect changes in the fish community that are independent of the fishery and as viewed through the selective window of experimental gillnets.

Changes in the aggregated biomass-size distribution were assessed in two ways. Aggregated biomass per 1 cm length-class was 10log-transformed. Aregression of biomass over length from 14 cm onwards was made per period and the significance of the difference in slope and intercept of these regressions were examined by stepwise examining the reduction in variability of the different terms of the following model:

10 Log (B)ij = µij * Fj + bi * L(D:j)

in which,
µij = overall mean
B = biomass = CpUE per length class
Fj = factor by which the intercepts of the three regression lines are determined
L= length
D:j = decade from 1970-1972, 1982-1985, 1994-1997

Both length and biomass axes were centred by subtracting the mean before submitting the model.

Next, splines with l constrained to explain at least 90 percent of the variability in the aggregated and 10log-transformed biomass-size distributions were drawn. The shape of the expected values of biomass over length given by the spline for the three periods was interpreted.

[4] Standard Length is a measure of the length of the fish, more or less excluding the tail fin (for an exact definition see Bagenal (1978)).
[5] PASGEAR is a database package with a number of descriptive data analysis tools developed for artisanal and experimental fisheries with PASsive GEARs, i.e. gillnets, hooks, traps etc. ©Jeppe Kolding, Dep. of Fisheries and Marine Biology, University of Bergen, Norway. E-mail:

Previous Page Top of Page Next Page