M. Rezaul Hasan and Hans A.J. Middendorp
Oxbow Lakes Small Scale Fishermen Project, Danida Technical Assistance
P.O. Box 12, Jessore, Bangladesh
ABSTRACT
Oxbow lakes (baors) in western Bangladesh are being brought under culture-based fisheries management (CBFM) through Oxbow Lakes Project (OLP-II). Six species of Indian major carps (rohu, catla and mrigal) and Chinese carps (silver carp, grass carp and common carp) are regularly stocked and harvested. The average yield from 1994 to 1996 of stocked carps from 19 oxbow lakes under study increased from 121 kg/ha/year in 1991–92 to 520 kg/ha/year, varying between 146 and 1008 kg/ha/year. The present paper reports the relationships between water transparency, carp yield and stocking density and examines the species interaction and their interdependence.
In analysis of the average of two seasons (1994–95 and 1995–96) data of water transparency [Secchi depth (cm)] and stocking density (no./ha) from 19 oxbow lakes, a descriptive multivariate regression model shows that the yield (Y) (kg/ha/year) is the function of Secchi depth (X) and stocking density (Z): Y = 516.99 - 2.371X + 0.074Z (R = 0.732; n= 19; P<0.01). Water temperature, lake size, water volume and water depth were not significant in fish yield prediction.
By simplifying the above multivariate regression model, two operational models are developed. i) Yield (Y1) is the function of Secchi depth (X): Y1 = 814.5 - 3.13X (r= -0.53; n = 19; P<0.05) and ii) yield (Y2) is the function of stocking density (Z): Y2 = -264.5 + 0.41Z -0.00004Z2 (r = 0.834; n = 19; P<0.01). Based on a Secchi depth reading, stocking density can be estimated with assumption that Y1 = Y2. If Secchi depth data are available, yield can be predicted from the first equation and once yield is known, stocking density can be estimated from the second equation.
Path coefficient analysis was done to examine the interaction between yields of different fish species, Secchi depth and macrophyte density. A significant negative correlation was observed between Secchi depth and yield of silver carp and mrigal. Positive correlation was established between macrophyte density and yield of grass carp but macrophyte density was negatively correlated with silver carp, catla and mrigal. The path coefficient analysis also showed that silver carp as surface feeder interacts positively with mrigal whereas catla interacts positively with common carp, and rohu with grass carp.
1. INTRODUCTION
Oxbow lakes, locally known as “baor”, are segments of dead river created when the flowing river changed its course. Fish culture in oxbow lakes is a practice by which an open water fisheries is converted by screening the inlets and outlets into culture-based fisheries. Three species of Indian major carp (rohu, Labeo rohita; catla, Catla catla; and mrigal, Cirrhinus mrigala), two species of Chinese carp (silver carp, Hypophthalmichthys molitrix; grass carp, Ctenopharyngodon idella), and common carp (Cyprinus carpio) are regularly stocked and harvested. Twenty oxbow lakes located in four districts (Jessore, Jhenaidah, Chuadanga and Kushtia) of western Bangladesh are being brought under culture-based fisheries management (CBFM) through Oxbow Lakes Project (OLP-II). The project is implemented by the Department of Fisheries (DoF) through the Project Implementation Unit (PIU) based in Jessore. The International Fund for Agricultural Development (IFAD) is providing funds for the hire of project staff, infrastructure development and for beneficiary training and credit. The Danish International Development Agency (Danida) provides technical assistance to the project as a grant. This package consists of a Technical Assistance (DTA) Team, credit guarantee fund and additional training and extension support through a collaborating NGO (the Bangladesh Rural Advancement Committee, BRAC) (Middendorp, Hasan and Apu, 1996).
The development objectives of this project are to develop the fisheries sector particularly in western Bangladesh in order to ensure (a) participation of poor people in management of the inland aquatic resource to enhance their social and economic well being, and (b) increase in fish production from inland water bodies so as to maximise socio-economic benefits to poor people. The project attaches special emphasis on research and development work, in order to study the magnitude and dynamics of fish production in oxbow lakes. The results of this research are to assist in the formulation of scientific management approaches to the conservation and maintenance of these important fishery resources of the country.
The objectives of this study were:
To establish the relationships between water transparency measured as Secchi depth, carp yield and stocking density, in order to predict the fish yield in an oxbow lake, and on basis of this to adjust the stocking density to an optimum level.
To examine fish species interactions and their interdependence, in order to better understand the principles governing the best species combination for achieving optimal yields, with special attention to be paid to physical and ecological characteristics of the oxbow lakes.
2. MATERIALS AND METHODS
2.1 Data collection
The data were collected from July 1994 to June 1996, unless otherwise specified.
Water transparency was measured once a week by using Secchi disc at one location in each baor at the same time (morning) throughout the year. Water temperature (°C) and water depth fluctuation (cm) were also measured at the same location once a week.
Density (kg/m2) of submerged macrophytes was recorded once every three months by “Quadrant” method.
Fish stocking density (no/ha) was recorded as the number of carps (total and by species) stocked annually in each baor.
The water area used in this study is the measured water area (Standard Water Area, SWA) (ha) based on a field survey conducted in early 1995.
Fish yield (kg/ha) was estimated by dividing the total annual harvest by SWA.
Harvest (kg) was expressed as the total weight of fish harvested and total number of fish harvested.
Production variables of 19 lakes, which are used to develop the yield prediction model and to examine the species interaction, are presented in Table 1.
Table 1. Production variables for 19 oxbow lakes (baors) during 1994–96 (mean of 1994–95 and 1995–96).
| Sl. # | Lake | Water area (ha) | Secchi Depth (cm) | Yield (kg/ha) | Stocking Density (no/ha) | Average Harvest Size (kg) |
|---|---|---|---|---|---|---|
| 1 | Nasti | 54 | 50.9 | 1008 | 3500 | 0.71 |
| 2 | Saster | 140 | 68.9 | 232 | 1288 | 0.91 |
| 3 | Porapora | 58 | 79.6 | 514 | 2763 | 0.79 |
| 4 | Sarjad | 9 | 61.6 | 820 | 3189 | 0.90 |
| 5 | Kayetpara | 116 | 140.5 | 146 | 1056 | 0.66 |
| 6 | Saganna | 35 | 76.0 | 738 | 4895 | 0.77 |
| 7 | Benipur | 45 | 97.9 | 613 | 3075 | 0.86 |
| 8 | Marufdia | 25 | 61.4 | 875 | 4970 | 0.39 |
| 9 | Bhanderdah | 48 | 115.3 | 321 | 1824 | 1.02 |
| 10 | Ujjalpur | 34 | 102.5 | 354 | 2769 | 0.50 |
| 11 | Bukbhara | 138 | 139.9 | 417 | 2325 | 0.61 |
| 12 | Hamidpur | 13 | 46.1 | 334 | 1850 | 0.63 |
| 13 | Khatura | 69 | 150.7 | 213 | 1545 | 0.92 |
| 14 | Khedapara | 45 | 91.5 | 453 | 3198 | 0.74 |
| 15 | Hariharnagar | 30 | 37.9 | 648 | 2611 | 0.72 |
| 16 | Bakra | 20 | 83.0 | 620 | 7578 | 0.54 |
| 17 | Bahadurpur | 110 | 186.0 | 468 | 2721 | 0.97 |
| 18 | Kannadah | 23 | 109.4 | 617 | 5849 | 0.52 |
| 19 | Kaliganga | 28 | 112.3 | 483 | 1979 | 1.00 |
| Mean | 520 | 3104 | 0.75 | |||
2.2 Data analysis
Yield predictive model
A yield predictive model was developed by establishing i) multiple relationship between carp yield (kg/ha) as dependent variable, and Secchi depth (cm) and stocking density (no/ha) as independent variables; and ii) linear relationships between carp yield (dependent variable) and Secchi depth (independent variable), and between carp yield (dependent variable) and stocking density (independent variable).
Relationship between stocking density and mean harvest size
Simple linear regression equation was used to establish the relationship between stocking density (independent variable) and mean harvest size (kg) of all fish species (dependent variable).
The computations for the above relationships were based on data of two fishing years (July, 1994 – June, 1995 and July, 1995 – June, 1996) for 19 oxbow lakes for which a complete set of data was available. The means of two seasons data were used in order to minimise variation so that the models developed would be more valid.
Species interaction
Path coefficient analyses between yields of different fish species (kg/ha), mean Secchi depth (cm), water area (ha) and macrophyte density (% area covered) were done to find out the interaction between different species and other variables. Separate analysis was done for data of 1994–95 and 1995–96 to observe the trends between years.
All data were analysed using a computer package MS Excel 5.0". Statistical analyses (multiple regression and correlation matrix) were done using Microstat and MSTAT-C statistical package.
3. RESULTS
3.1 Yield predictive model
First, an overall multivariate regression analysis was done by including the two seasons mean data (1994–95 and 1995–96) of the following biophysical parameters: a) lake area, b) water depth, c) water temperature, d) Secchi depth/water transparency, e) stocking density and f) species composition. Of all the above biophysical parameters, only Secchi depth and stocking density were found significant in fish yield prediction. Therefore, the following yield prediction models were developed using Secchi depth and stocking density:
a. Descriptive multivariate regression model: multiple regression equation between yield (dependent variable) and Secchi depth and stocking density as independent variables:
Y = 516.99 - 2.371X + 0.074Z (R = 0.732; F = 9.257; n = 19; P<0.01)...................Equation 1
where, Y = yield (kg/ha), X = Secchi depth (cm) and Z = stocking density (no/ha).
Yield of a lake could be predicted from this descriptive model provided both Secchi depth and stocking density are known. But this may not be a useful management tool as it does not give an indication of an optimum stocking density appropriate for maximising the yield. Therefore, two simple predictive operational models were developed from the above multivariate regression model.
b. Operational models:
i) Simple linear regression equation between yield (dependent variable) and Secchi depth (independent variables) (Fig. 1):
Y1 = 814.48 - 3.134X (r = -0.53; n = 19; P<0.05)......................................Equation 2
where Y1 = yield (kg/ha) and X = Secchi depth (cm).
ii) Second order polynomial regression equation between yield (Y2) and stocking density (X) (Fig. 2):
Y2 = -264.45 + 0.407Z - 0.00004Z2 (r = 0.835; n = 19; P<0.01).........................Equation 3
Water transparency is inversely related to the yield (Equation 2 and Fig. 1). Similarly, stocking density has a positive significant influence on the yield (Equation 3 and Fig. 2). It is apparent that increasing stocking density beyond about 4000/ha does not result in increased yield.
Yield of a lake can be predicted from the first operational equation if Secchi depth data (mean of twelve readings collected over a one-year period) are available. Once yield is known, stocking density can be estimated from the second equation with the assumption that Y1 = Y2.
3.2 Relationship between stocking density and mean harvest size
Y = 0.928 - 0.00006Z (r = -0.54; n = 19; P<0.05).......................................Equation 4
where Y = harvest size (kg) and Z = stocking density (no/ha).
Stocking density is negatively correlated with mean harvest size (Equation 4 and Fig. 3) implying that catchable size of fish reduces with increasing stocking density.
3.3 Species interaction and interdependence of variables
The analysis of data of 1995–96 showed that there was a significant negative correlation between Secchi depth and yields of silver carp and mrigal (Table 3). Negative correlation existed between macrophyte density and yields of silver carp, catla and mrigal and positive correlation between macrophyte density and yield of grass carp (Table 3). Water area was negatively correlated (P<0.05) with the yield of silver carp, but it was not significantly correlated (P>0.05) with total fish yield. The path coefficient analysis between yields of different fish species, Secchi depth, water area and macrophyte density of 1994–95 was similar to the trends observed in 1995–96 (Tables 2 and 3).
Figure 1. Relationship between Secchi depth and carp yield in 19 baors (simple liner regression).
Figure 2. Relationship between stocking density and carp yields of 19 baors (second order polynomial regression)
Figure 3. Relationship between stocking density and mean harvested size of all carps (simple liner regression).
The path coefficient analysis also showed that there is a positive correlation between the yields of silver carp and mrigal for 1994–95 and 1995–96. Similarly, positive correlation existed between catla and common carp and between common carp and mrigal (Tables 4 and 5).
Table 2. Correlation matrix between fish yield (kg/ha) and other variables (1994–95)
| Variables | Water area (ha) | Mean Secchi depth (cm) | Submerged macrophytes (kg/m2) |
|---|---|---|---|
| Silver carp | -0.542* | -0.626* | -0.692* |
| Catla | 0.041 | -0.197 | 0.176 |
| Rohu | 0.090 | 0.319 | 0.454* |
| Common carp | 0.052 | -0.460* | -0.198 |
| Mrigal | -0.589* | -0.495* | -0.483* |
| Grass carp | 0.485* | 0.319 | 0.455* |
| All carps | -0.436* | -0.592* | -0.471* |
P<0.05 = ± 0.4426 at d.f. 18
Table 3. Correlation matrix between fish yield (kg/ha) and other variables (1995–96).
| Variables | Water area (ha) | Mean Secchi depth (cm) | Submerged macrophytes (kg/m2) |
|---|---|---|---|
| Silver carp | -0.490* | -0.704* | -0.541* |
| Catla | -0.169 | -0.363 | -0.453 |
| Rohu | -0.084 | 0.135 | 0.199 |
| Common carp | -0.048 | -0.201 | -0.108 |
| Mrigal | -0.431 | -0.524* | -0.457* |
| Grass carp | 0.090 | 0.159 | 0.553* |
| All carps | -0.381 | -0.529* | -0.332 |
P<0.05 = ± 0.4426 at d.f. 18
Table 4. Correlation matrix between fish yield (kg/ha) and other species (1994–95).
| Species | Silver carp | Catla | Rohu | Common carp | Mrigal |
|---|---|---|---|---|---|
| Catla | -0.341 | ||||
| Rohu | -0.164 | 0.174 | |||
| Common carp | 0.244 | 0.482* | 0.065 | ||
| Mrigal | 0.661* | -0.056 | -0.118 | 0.329 | |
| Grass carp | -0.478* | 0.408 | 0.243 | 0.136 | -0.429 |
P<0.05 = ± 0.4426 at d.f. 18
Table 5. Correlation matrix between fish yield (kg/ha) and other species (1995–96).
| Species | Silver carp | Catla | Rohu | Common carp | Mrigal |
|---|---|---|---|---|---|
| Catla | 0.431 | ||||
| Rohu | 0.076 | 0.343 | |||
| Common carp | 0.426 | 0.549* | 0.302 | ||
| Mrigal | 0.832* | 0.392 | 0.212 | 0.502* | |
| Grass carp | -0.053* | -0.036 | 0.492* | -0.019 | 0.028 |
P<0.05 = ± 0.4426 at d.f. 18
4. DISCUSSION
4.1 Yield predictive model and stocking
Yield predictive modelling in the development and management of culture-based fisheries is the basic tool for effective development and management of such fisheries. Morpho-edaphic index (MEI = Conductivity/ Mean Depth) has been used as the universally accepted predictive model for deep lakes in temperate regions (Ryder, 1965, 1982). Jenkins (1982) reported that fish productivity in American reservoirs was highly correlated with MEI (P=0.0001). However, yield prediction based on MEI has not proved to be useful for tropical and sub-tropical lakes. No significant relationship between fish productivity and conductivity was established in Chinese reservoirs (Li and Xu, 1995). Similarly, Hartmann and Aravindakshan (1995) reported that MEI was not an efficient tool for yield prediction in Indian reservoirs. In our study, it is apparent that amongst the four parameters tested (temperature, water depth, lake area and Secchi depth), only Secchi depth was significant. In relationship between Secchi depth and yield (Equation 2 and Fig. 1), r value is 0.53. It needs to be pointed out that although the r value appears to be less than desired for the lake fishery with all its inherent variability, potential discrepancies in data collation/reporting, etc., the obtained r value is very much in the range usually obtained for comparable fisheries.
Secchi depth is considered a relatively reliable indicator of the biological productivity due to the generally very low turbidity in oxbow lakes. The yield descriptive model may be further substantiated with a biologically more acceptable indicator, e.g. chlorophyll estimation. This will also further clarify the yield predictive model(s) by enabling independent estimations of the fish yields.
A yield predictive model by itself has little use, unless it can be used to determine the stocking level. In a stock and capture fishery, as that of the investigated oxbow lakes, one of the most important management tools available is the stocking strategy, i.e. the size, the species, the number stocked (total and species-wise) and the time of stocking. Manipulation of the stocking strategy enables an increase in the yield.
The available data provide very useful guidelines for implementing an effective stocking strategy. Equation 3 shows that yield is directly dependent on stocking density (Fig. 2). The above equation permits one to estimate the number of fingerlings to be stocked into a lake for which the potential fish productivity is known. Further analysis needs to be done on this relationship to determine the maximum and the most economical stocking density (based on the yield per numbers stocked). Eye estimation shows that the corresponding stocking numbers are approximately 4500 and 3700 fingerlings per ha, respectively.
Mean harvest size was inversely related to stocking density as shown in Equation 4 and Fig. 3. A similar relationship between stocking density (fish/ha) and fish yield (kg/ha) was established in Nanshahe reservoir in China where yield was directly proportional to stocking density during the period from 1966 to 1975. As the stocking density increased, yield rose, but catchable size is reduced as growth is density dependent (Li and Xu, 1995).
4.2 Species interaction and interdependence of variables
Silver carp, an efficient filter feeder, has a highly significant negative interaction (Table 3; r = -0.704; P<0.01) with Secchi depth. Catla, however, also a filter feeder, appears to have no significant relationship with Secchi depth. The yield of mrigal was negatively correlated (r = -0.524; P<0.05) with Secchi depth, although this relationship may not be due to the Secchi depth itself, but probably due to the positive interaction between silver carp and mrigal (Table 3; r = 0.832; P<0.01). Similar positive relationships also existed between catla and common carp, rohu and grass carp and common carp and mrigal. We observed similar positive interactions between silver carp and mrigal, catla and common carp, and grass and rohu when yield data of 1994–95 were analysed (Tables 2 and 4; Middendorp, Hasan and Apu, 1996). The occurrence of these positive interactions over a two-year study period shows the presence of two species complexes in oxbow lakes: silver carp-mrigal and common carp-catla but no definite explanation for the observed positive interactions between silver carp-mrigal and common carp-catla is available. It is noted that Li and Xu (1995) similarly categorised Chinese reservoirs into two distinct types based on their dominant species, i.e. silver carp and common carp.
5. CONCLUSIONS
This study indicated that of all physical and ecological parameters, Secchi depth had the strongest correlation with fish yield, and that macrophytes density had negative impact on the total fish yield. Secchi disc is a simple tool that fishermen can use for the prediction of fish yield and so determine stocking density of a lake provided Secchi depth data (a mean of 12 readings collected over a one-year period) are available. The use of Secchi disc depth for determining stocking density is one of the most important management tools for increasing fish production.
Based on the above findings we can conclude that in lakes with a large stock of silver carp, mrigal performs better as bottom feeder, whereas in lakes with a large stock of common carp, catla performs better.
ACKNOWLEDGMENTS
We gratefully acknowledge the assistance of Dr. Sena S. De Silva of the School of Aquatic Science and Natural Resources Management, Deakin University, Warrnambool, Australia, in the analysis of the data.
REFERENCES
Hartmann, W.D. and N. Aravindakshan. 1995. Strategy and Plans for Management of Reservoir Fisheries in Kerala, India. Indo-German Reservoir Fisheries Development Project, Department of Fisheries, Vikas Bhavan, Trivandrum 695 001, India. 123p.
Jenkins, R.M. 1982. The morphoedaphic index and reservoir fish production. Trans. Am. Fish. Soc.111: 133–140.
Middendorp, A.J., M.R. Hasan and N.A. Apu. 1996. Community fisheries management of freshwater lakes in Bangladesh. The ICLARM Quarterly Naga 19: 4–8.
Li, Sifa and S. Xu. 1995. Culture and Capture of Fish in Chinese Reservoirs. Southbound, Penang and International Development Research Centre, Ottawa. 128p.
Ryder, R.A. 1965. A method for estimating the potential fish production of north-temperate lakes. Trans. Am. Fish. Soc. 94: 214–218.
Ryder, R.A. 1982. The morphoedaphic index- use, abuse and fundamental concepts. Trans. Am. Fish. Soc. 111: 154–164.
