Assessment of Sustainable Yield and Optimum Fishing Effort for the Tilapia (Oreochromis niloticus L. 1758) Stock of Lake Hawassa, Ethiopia

The tilapia (Oreochromis niloticus, L. 1758) stock of Lake Hawassa, Ethiopia, was assessed to estimate sustainable yield (MSY) and optimum fishing effort (fopt) using length-based analytical models (Jone's cohort analysis and Thompson and Bell). Pertinent data (length, weight, catch, effort, etc.) were collected on a daily basis for 514 days during 27-12-2003 to 24-05-2005 at the major landing site known as Fish market or AmoraGedel. The sampling days fell into two periods demarcated by the date (08-04-2004) when a management action was implemented, which reduced fishing effort by half (from1954 gillnets/day to below 800 nets/day). Thus, data obtained during 27-12-2003 to 08-04-2004 pertained to the period prior to effort reduction whereas the rest pertained to the period after the reduction. The two data sets were analyzed separately to get a basis to evaluate the effect of the management action of effort reduction. The assessment gave an estimate of current yield of 526.8 t/year for the period before effort reduction whereas 441.6 t/year afterward. The predicted MSY was 514.5 t/year and 441.6 t/year for the period before and after effort reduction, respectively. The respective F-factor is estimated to be 0.5 and 1.0.This suggested that the fishing effort before the reduction of effort (1954nets/day) was very high and, as already implemented, should have been reduced by half(i.e., an F-factor of 0.5).Therefore, the implemented management measure to reduce effort below 800 nets/day is appropriate. Likewise, since the estimated MSY of441.6 t/year for the period after the reduction would be obtained at an F-factor of 1.0, it was concluded that the current level of fishing effort of 696 gillnets/day can be maintained as fopt for sustainable exploitation of the stock.


Sampling regime and data collection
Data were collected from the cooperative fishermen landing site, locally known as 'Amora-Gedel Asa Gebeya', where fishermen retail their catch. Since all the fish landed were brought to Amora-Gedel for retail, sampling only from this site was considered as sufficient to obtain representative sample. The data mainly constituted information on the O.niloticus fishery of the lake that are useful to assess the stock and estimate maximum sustainable yield and biologically optimum level of fishing effort. Specifically the basic information collected included i) the length composition of O.niloticus caught by the fishery, ii) total O.niloticus yield, iii) fishing effort expanded, iv) number of fishermen in operation and v) fishing site. Contour lines are bathymetric lines at 4 meters depth interval. Inset: map of Ethiopia showing the northeastern section of the Great Rift (Elizabeth Kebede, 1996).
During each sampling day, random samples of 30 up to 50 O. niloticus were taken from the catch of each fisherman and their length was measured to the nearest mm. Also the total weight of the length measured fish was recorded (to the nearest gm) as well as the total catch of each fisherman was weighed. The latter data was then used to estimate the total number of O.niloticus caught by the respective fisherman.
The daily catch and yield data were collected from fishermen for 514 days. i.e., the lake was visited on a daily basis, from 27-12-2003 to 24-05-2005.The sampling days fell into two sampling periods. These were before reduction of fishing effort (between 27-12-2003 to 08-04-2004) and after reduction of fishing effort (between 10-04-2004 to 24-05 2005). Furthermore, the sampling days encompassed both fasting and non-fasting periods of the Ethiopian Orthodox Church and this enabled to get representative catch data during the time when the lake was lightly as well as intensively fished.

Data summarization and analysis
The catch data were summarized in a manner useful for stock assessment using Jones length based cohort analysis model (Jones, 1984) and length-based Thompson and Bell yield prediction model (Thompson and Bell, 1934;Sparre and Venema, 1992).The catch data collected before and after reduction of the fishing pressure were separately analyzed and interpreted. The summarization and analysis were done by using Microsoft Office Excel 2007 software.
Accordingly, the length composition catch data of O. niloticus were summarized to prepare a table of the length composition of total annual catch of fish and this was done as follows (Pauly, 1984;Sparre and Venema, 1992).

Preparing length frequency of the sample catch
Length measurements recorded daily were grouped into two cm length intervals to prepare a The length and weight of a total of 5,787 fish were measured during the 104 days of sampling before reduction of the fishing effort. Also, the length and weight of 20,509 fish were measured during the 410 days of sampling after reduction of fishing effort. Overall, 26,296 fish were measured during the 514 days of sampling and the length frequency produced using such a large sample size was considered as adequate to give a good picture (Sparre and Venema, 1992) of the length frequency of the catch of O.niloticus in the lake.

Estimating the total number of fish landed per day by each fisherman
This was estimated by multiplying the number of length measured fish by a conversion factor (W/w) where W= the total weight of the catch of respective fisherman and w = sample weight of the length measured fish. Thus fish that were simultaneously counted and weighed were used to determine appropriate raising factor to convert records of the daily weight of the catch of respective fishermen into numbers.

Estimating the length composition of the total daily catch
This was achieved by multiplying the total numbers caught per day by the relative frequency of each length group in the daily sample obtained under item '2.3.1' above. The total length frequency of fish landed during the sampled days was then determined by summing the frequencies of respective length groups.

Estimating total number of O.niloticus caught during theun-sampled days of the year
Since the catch and effort expanded differed during the fasting and non-fasting days of the Likewise, the average daily catch of the sampled major fasting days was used to estimate the catch of the un-sampled major fasting days. In a similar manner, the total weight of the catch (yield) and effort expanded were estimated for the un-sampled days of the year, categorizing the dates into three categories as explained above.

Estimating the annual total length composition of landed fish
This was done by multiplying the length frequency of the sampled days catch by an appropriate conversion factor which was equal to C/c, in which 'C' =the estimated total catch of fish during the whole year and 'c' = the total catch of fish during the sampled days.

Estimating population size and fishing mortalities using Jones length based cohort analysis
The Jones length based cohort analysis model (Jones, 1984) was used to estimate the population abundance and fishing mortality coefficient by length group of O.niloticus. To get started with the analysis, the total annual catch in each length group [C(L1,L2)]was used as the basic input data. This was done in three steps as follows: Yosef, T. G., Alemken, B and Elias, D (MEJS)

Estimating the population number of the largest length group in the catch
The following equation was employed (Jones, 1984;Sparre and Venema, 1992)

Estimating the population numbers of consecutively younger length groups in the catch
This was done using the following equation:

Estimating the fishing mortality rate for the respective length groups
Fishing mortality values for each length group was estimated using the following equation.
To use equations 2, 3, 4 and 5, the following input data and parameters were prepared in advance.
i) First a table of the total annual catch distributed by length group was prepared as described earlier.
iii) Thirdly, an estimate of the natural mortality coefficient (M) for O.niloticus stock of Lake Hawassa, which is equal to 0.35 yr -1 , was estimated using Pauly's empirical formula as follows (Pauly, 1984). Where values of L and K are as described above for O.niloticus stock and T is the mean annual surface water temperature of Lake Hawass arecorded during the study period, which was 21 o C.

Predicting sustainable fish yield and optimum fishing effort
The outputs of the above cohort analysis procedures were used as input data for the Thompson and Bell yield prediction model to predict sustainable fish yield at different levels of fishing mortalities (Thompson and Bell, 1934;Pauly and Morgan, 1987;Schnute, 1987;Sparre and Venema, 1992).
For the length based Thompson and Bell model, input data and sources comprised the following i.Length composition of the annual total number of fish landed by the fishery. This was obtained from field data collection (catch statistics data) as described earlier.
ii.Estimates of population numbers of fish and fishing mortality coefficient (F) by length group.
Source: results of the Jones length based cohort analysis described earlier.
iii.An average estimate of natural mortality coefficient (M) and the Von Bertalanffy growth parameters (L and K). The same values discussed earlier have been used.
iv.Mean weight of the landings per length group. This was estimated by using the mean length of each length group and the length-weight relationship formula expressed as follows:
To establish the above length-weight regression relationship, a random sample of 1000 O.niloticus that encompassed a wide range of length groups were length and weight measured.
In due regard, the total weight of fish landed per year in each length group was estimated by multiplying the average weights of each length group by the corresponding total annual catch of respective length group. The computation procedures of the Thompson and Bell (1934) yield model consisted of two main stages as described below.

Estimating fish yield under the current level of effort
The yield of fish under the level of fishing effort expanded on the stock was estimated using annual catch data of each length group and the average weight of fish of respective length group.
For this, first the yield in weight obtained per year from the respective length group of fish was calculated by multiplying the total annual catch in numbers of each length group by the mean weight of the respective length group. i.e.,

Yield predictions under different levels of fishing effort
The second step of the Thompson and Bell yield prediction procedure involved assessment of the effects of changes in the current level of fishing effort (and hence that of fishing mortalities) on fish yield. This was done by predicting fish yield at higher and/or lower levels of fishing mortality coefficients pertaining to the respective length groups (F-at-length-array). i.e., the current fishing mortality values of the respective length groups estimated following the Jones length based cohort analysis were used as reference and these were increased and/or decreased by certain raising factors (F-factor) to predict new values of yield corresponding to the changed fishing mortalities (Venema et al., 1988;Spare and Venema, 1992).Details of the procedure were as follows.

i) Estimating population abundance under the changed level of fishing mortality
Since a change in fishing mortality obviously results in a change in population number of fish in the water, new estimates of population numbers in each length group need to be predicted under the changed fishing mortality condition. Thus the population numbers under the changed fishing mortality were calculated from the following exponential decay relationship (Schnute, 1987;Sparre and Venema, 1992).

ii) Estimating the total death and catch in each length group under the changed fishing level
The total number of deaths expected while the fish grew from length L1 to length L2, i.e., D(L1, L2) under the changed fishing level is equal to N(L1) -N(L2). From this total death, the fraction died due to fishing make up the total catch. Accordingly, the catch per length interval corresponding to the changed fishing mortality[C(L1, L2)]was calculated from the following relationship (Wetherall et al., 1987).

C(L1, L2) = [N(L1) -N(L2)] * F(new)/Z(new) ----------------------------------Equation 11
Where, F (new) and Z (new) are the fishing and total mortality coefficients, respectively, under the changed level of fishing effort. Then, to estimate the expected yield obtained from respective length groups annually (Y(L1,L2)) under the changed fishing mortality, the expected catch in number under the changed fishing level was multiplied by the mean weight of each length group as illustrated by equation 8. The total annual yield to be expected under the new level of fishing effort was then predicted by summing up the contributions of each length group.
Such predictions were evaluated for different values of fishing mortalities so as to see the full spectrum of the effect of changing fishing effort on the stock. According to the above analysis, the level of fishing mortality that gave maximum sustainable yield was considered as the biologically optimum level of fishing mortality. Since there is a one to one correspondence between fishing mortality (F) and fishing effort (f), the value of F-factor chosen as optimum was used to recommend how much the current level of fishing effort need to be increased or decreased to get the maximum sustainable yield from the stock (Sparre and Venema, 1992).

Status of the tilapia fishery of Lake Hawassa
There were overall 72 and 66cooperative member fishermen operating daily on the lake during the period before and after reduction of fishing effort, respectively (Table 1). These fishermen set on average 1,954 and 696 nets daily on the lake before and after reduction of effort, respectively. Each fisherman on average owned 27 nets prior to reduction of efforts but after reduction of efforts each fisherman on average owned 10.6 nets. The nets were basically set to catch O.niloticus but these nets also caught some C.gariepinus and rarely L. intermedius. The nets used were similar during the two periods. Each net was on average 80 m long and 2.5 m deep and it had an average stretched mesh size of 6-8 cm. Overall, an estimated number of 713,063 nets were operated per year prior to reduction of the fishing effort and after reduction of the fishing efforts, the total number of nets set per year was 253,956.With these levels of fishing effort, an estimated total number of 2,658,906and 2,046, 077 O. niloticus were caught per year before and after reduction of fishing effort, respectively. Accordingly, the total yield before and after reduction of the fishing efforts were estimated at 526.76 and 441.6 t/year, respectively (Table 1).   Table 3 gives estimates of population number and fishing mortality coefficient by length group of O.niloticus that composed the fishery for the periods before and after reduction of the fishing pressure. Estimates of population numbers (N(L1)) and fishing mortality coeffecients (F(L1, L2)) are direct outputs of the Jones length based cohort analysis computed using equations 2 and 4, respectively.

Estimates of population abundance, fishing mortality coefficient and current yield by length group of O. niloticus that composed the fishery
During the period prior to reduction of the fishing pressure, a population of over 16.6 million O.niloticus has been estimated to exist in the fished part of the lake as obtained by summing the population numbers of the respective length groups that composed the fishery shown by column 2 (Table 3). This estimate for the period after reduction of the fishing pressure was over 14.6 million (Column 3) and it is failry comparable to the estimate obtained for the period before reduction of effort. Also both estimates pertain to the population of the whole area of the lake where the fishery was active as the fishermen were operating in the whole area of the lake.
As estimated by the model, over 4.7 million of O. niloticus measuring 14 to 16 cm were recruited to the fishery every year prior to reduction of the fishing pressure (Column 2, Table 3).
Similarly, based on data collected after reduction of the fishing pressure, the estimated number of O.niloticus recruited to the fishery attaining a length of 14 to 16 cm was close to 4 million fish (Column 3, Table 3). Accordingly, these estimates were fairly comparable.   (Table 4) are the annual catch of the respective length group of fish before and after reduction of the fishing effort and they are displayed here to illustrate the intermediary calculation steps. The current total yield per year pertaining to the respective length group (columns 5 and 6) were obtained by multiplying the total catch per year of the respective length group by the corresponding mean weight values as depicted by equation 8. The estimated annual total yield of O.niloticus during the sampled year prior to and after reduction of the fishing effort was 526.8 t/year and 441.6 t/year, respectively. There is a yield difference of about 85 t/year between the two periods but given that there was fishing pressure differences, the observed yield difference is not significant. Table 3. Population number and fishing mortalities by length group estimated based on data collected before and after reduction of fishing efforts.  Similarly, the analysis done on data collected after reduction of the fishing effort indicated that the predicted maximum sustainable yield of O.niloticus is 442.1 t/year and this is obtained at an F factor of 1.2 (Table 5). It indicates that the fishery can have small room for expansion and the level of fishing effort that was exerted at the time of data collection (i.e., 696 nets/day) can be increased by 20%. i.e., it can be elevated to 835 nets/day. However, since the safe level of exploitation is to reduce the F factor that gives the maximum sustainable yield by 20 % (Pauly, 1984;Sparre and Venema, 1992), it is advisable to recommend an F factor of 1 and keep the fishing effort as it is at 696 nets/day, which corresponds to a yield of 441.6 t/year.

Status of the tilapia fishery of Lake Hawassa
After measures were taken to reduce the fishing efforts (i.e., after 08-04-2004), the number of nets set per day were drastically reduced by 2.8 times, i.e., from close to 2000 nets per day to about 700 nets per day. However the number of fishermen was relatively the same as the period prior to reduction of the nets. Owing to reduction of efforts, the catch per net (catch per unit effort) had increased by more than two fold from an average catch of 3.7 fish per net (  to the fishery of Lake Hawassa attaining a total length of 16 cm. Given that recruitment considerably varies from year to year, the present finding was fairly comparable to the previous estimates. In the present study, a total yield of 526.8 t/year and 441.6 t/year of O. niloticus were estimated based on data collected before and after reduction of the fishing pressure, respectively, and these estimates were fairly comparable to the values reported earlier for O.niloticus stock of Lake Hawassa. For example, Reyintjens and Tesfaye Wudineh (1998) estimated a total annual yield of 520 tons of O.niloticus/year as harvested by the fishermen cooperatives of Lake Hawassa.
The estimate in 2002 was about 540 t/year (Yosef Tekle-Giorgis, 2002) and the estimate in 2012 for one-third of the lake was about 192 t/year (Sintayehu Adissu, 2012), which when triplicated becomes comparable to the current yield estimate.
The analysis done on data collected prior to reduction of fishing effort indicated that the fishing effort expanded before reduction of the fishing pressure (1954 nets/day) should be reduced by half in order to sustainably exploit the stock. Again the analysis done on data collected after reduction of fishing effort indicates that the biologically optimum level of fishing efforts to be expanded on the stock is around 700 nets/day. Thus, the results obtained from both analyses, although do not exactly match, they somehow corroborate each other. In due regard, the recommended safe level of efforts to be exerted on the stock should be between 700 to 800 nets per day and measures taken to reduce the fishing efforts was appropriate. The estimated optimum effort level in the present study for O.niloticus stock of Lake Hawassa is similar to the recommended safe level of fishing reported by Reyintjens and Tesfaye Wudineh (1998)