During the course of the experiments, the results improved with experience. Results are presented in roughly chronological order with much of the poorer results evident in the early data, Descriptions and discussion of the poorer results and failures is given in sufficient detail to avoid unnecessary problems elsewhere.
The system comprising 3 plastic basins, mattress pad filters, and various lengths of garden hose works we!! as a small scale hatchery for Clarias. It is simple, understandable, inexpensive, easy to construct using materials locally available in western Kenya, and light weight (no small consideration when transporting to a distant pond).
The smaller, 1/2" interior dia syphon hose gave an adequate water supply from the pond to the filtration system and is recommended, although this is not crucial. The larger 3/4" hose gives a higher flow which did not prove necessary, and if a small (100 m2) pond is the source, the pond water level can be significantly lower overnight. The filters also need more frequent cleaning. The larger hose should still be used between the basins to insure unrestricted flows
One area that needs further work is the attachment of the garden hose to the bottom of the containers and the net barrier in the spawning and incubation container. The PVC glue used becomes brittle after a short time and easily breaks off creating leaks. A better quality glue would be preferable.
The actual choice of containers is still open to question. Any non-toxic material will do. De Graaf (pers.comm.) mentions that farmers in South East Asia use clay pots for incubation of fish eggs, and on one occasion we successfully induced a pair of Clarias to spawn in a 60 i clay pot. it may prove more beneficial to permanently construct beside the pond a series of small concrete or baked day basins for filtration and spawning, This would avoid the risk of theft and would probably be cheaper.
Finally, if one requires larger numbers of larvae, it is possible to use the same system for incubating eggs obtained from artificial reproduction. Although not reported here, this was done by project staff on several occasions with excellent results, The point is that it is possible to spawn Clarias gariepinus on a small scale using very minimal equipment and facilities.
There were a total of 65 spawning trials. A total of 168 778 larvae were produced, in 8 trials one or both of the females died before spawning occurred. There was only 1 occasion where the female was alive but failed to spawn following injection.
3.2.1 Sand and mattress filters
The first 13 trials were done using sand fitters. The results were not encouraging (see appendix i), A considerable effort was made with these filters and several configurations were tried. The surface area. was increased or decreased, the thickness of the sand layer modified, flow rates were either adjusted to the minimum requirement or to allow excess water for self cleaning, both percolation and steady volume methods were tried, and so on. Problems still occurred. During 5 days of filtration, microscopic algae still managed to pass through the filter. In the covered incubation chamber, this algae died and settled to the bottom smothering the eggs. A heavy organic load is retained within the filter and apparently biochemical changes occur during the 5 days, resulting in either de-oxygenated water or perhaps formation of hydrogen sulfide or a similar noxious substance.
The fish spawned in all trials save 1 where the female died, but either the eggs died or they hatched successfully only for the larvae to die the following day. Only 3 738 larvae were produced in all 13 attempts, and most of these in only 2 trials. Sand filters are not recommended, and inclusion of this data in calculations on mean yield, expected yield, and comparison of 1 or 2 females per container would skew the data on the lower side. All further calculations in this paper exclude the data from those trials using sand filters
Results were far better when using the mattress pads as filters, Table I summarizes the results of 51 spawning trials using pond water and the foam mattress filtration system. The table compares the results obtained when using 1 or 2 couples per container (see following page).
There is no advantage gained by using 2 spawning pairs instead of 1. In fact the opposite is true; the larval yield was significantly higher (P<0.05) with 1 couple, These results were probably influenced to some extent by the smaller size of the breeders used in the trials with 2 couples, but this would not explain the nearly 3 fold increase in larvae obtained from only 1 pair. This data would indicate that there is more likely an inhibitory response when multiple breeders are confined in a small container.
3.2.2 Number of larvae per spawn
In practical terms, one can expect between 3 000 to 5 000 larvae /per spawn (p<0,05) using a single spawning pair (see Table 1). Results were more variable at the beginning of the trials but became more consistent with time and experience (see appendix I). It would be useful if one could predict the number of larvae from the weight of the female when using semi-artificial reproduction. This was not possible from these trials; no correlation was found between female weight and the number of larvae produced (r2 =0.095), Smaller fish often gave far more larvae than larger fish. The actual number of eggs laid, hatching percentage, and larval mortality was not determined. These factors obviously influence the number of viable larvae.
The staff conducting the trials preferred using female fish between 200 and 400 g, This is subjective. Females less than 200 g appeared more susceptible to serious injury Fish of 200 to 400 g are easier to handle and inject with a syringe than smaller fish and assumably produce more eggs. Fish above 450 g may be too large for the containers or may release too many eggs for the system, but this was not explored.
| One couple | Two couples | |
| Number of trials | 35 | 16 |
| Average weight of each female (g) | 340 | 145 |
| Standard error of the mean (S.E.M.) | 35.9 | 23.3 |
| Average total weight of females per spawning container | 340 | 291 |
| Number of trials where all females died; no spawning | 4 | 3 |
| Percent trails with female mortality and no spawning | 11.4 % | 18% |
| Average number of larvae per trial (excludes no spawning) | 3 985 | 1 366 |
| Standard error of the mean (S.E.M.) | 1 090 | 618 |
| Total number of larvae produced | 139 452 | 21 866 |
3.2.3 Introduction of a barrier
The 12 % mortality rate (ail trials) among the breeders is high. This led to the introduction of the net barrier to separate the adult fish, In the 7 trials conducted with the net barrier no broodstock mortalities occurred. Spawning and fertilization did not appear to be inhibited by the net or separation of the breeders. The highest single yield of larvae (13 422) was produced with this method. This small number of trials is insufficient to warrant a comparison without the net barrier, but there appears to be no adverse effect and it may well serve to lower the female mortality rate,
The initial results of 28 trials of rearing larva! Clarias using very small ponds completely covered with a 4 mm mesh nylon net are presented in Table II (see following page). Pond rearing of fish larvae is notorious for variable results (Culver et al, 1993, de Graaf et al, in press), This data is no exception; there is a considerable variation although all trials had fairly similar treatments. These are preliminary results. It is prudent at this point to only present a simple analyses of mean values, standard deviation, and percentages. However, the results are very promising for small scale rearing of Clarias larvae.
3.3.1 Harvested weight
Most of the fish harvested from the small ponds were 10 to 20 mm total length with corresponding weights varying from 100 to 1 000 mg. The smaller fish were those harvested after 14 or 15 days, the larger sizes when the rearing period was extended to 23 to 28 days.
It may be stretching the definition of a fingerling to consider Clarias of 10 mm weighing less than 1 g as such. It is, however, entirely possible to manually count and transport Clarias of 100 to 1 000 mg in field conditions as their branchial organs are now developed. Demoulin (pers.comm.) reports very good results with stocking this size Clarias directly into rearing ponds in Rwanda. This has not yet been explored in western Kenya where the preference has been to stock larger, 3 to 5 g Clarias in rearing ponds.
Table II
Preliminary results form rearing larval Clarias gariepinus in small ponds protected by nets
| Pond size m 2 | No. of days. | No. larvae stocked | Stocking density (/m2) | No fingerlings (/m2) | Density at harvest (/m2) | Survival | Mean weight (g) | Comments |
| 50 | 14 | 11 000 | 220 | 5 110 | 102 | 46 % | 0.20 | |
| 50 | 19 | 5 000 | 100 | 291 | 6 | 6 % | no data | 8 600 tadpoles |
| 50 | 26 | 5 000 | 100 | 621 | 12 | 12 % | no data | 5 000 tadpoles |
| 50 | 14 | 5 000 | 100 | 1 234 | 25 | 25 % | 0.11 | < 500 tadpoles |
| 10 | 26 | 3 200 | 320 | 2 188 | 218 | 68 % | 1.00 | |
| 10 | 27 | 1 900 | 190 | 800 | 80 | 42 % | 0.50 | |
| 10 | 23 | 3 000 | 300 | 1 360 | 138 | 46 % | 0.50 | |
| 10 | 28 | 5 110 | 511 | 678 | 68 | 13 % | 0.93 | Few fish at 4 g |
| 10 | 27 | 5 000 | 500 | 279 | 28i | 6 % | 0.96 | Few fish at 4 g |
| 10 | 27 | 3 000 | 300 | 366 | 37 | 12 % | 0.52 | Few fish at 4 g |
| 10 | 15 | 2 340 | 234 | 614 | 61 | 26 % | 0.15 | |
| 10 | 15 | 2 340 | 234 | 589 | 59 | 25 % | 0.13 | |
| 10 | 15 | 2 340 | 234 | 421 | 42 | 18 % | 0.14 | |
| 10 | 15 | 2 340 | 234 | 433 | 43 | 19 % | 0.18 | |
| 10 | 15 | 2 340 | 234 | 394 | 33 | 17 % | 0.50 | |
| 10 | 15 | 2 340 | 234 | 695 | 70 | 30 % | 0.3S | |
| 10 | 14 | 2 400 | 240 | 1 003 | 100 | 42 % | 0.05 | < 5C tadpoles |
| 10 | 14 | 2 400 | 240 | 851 | 65 | 35 % | 0.07 | |
| 10 | 14 | 2 400 | 240 | S08 | 61 | 25 % | 0.14 | |
| 10 | 14 | 2 400 | 240 | 551 | 55 | 23 % | 0.13 | < 50 tadpoles |
| 10 | 14 | 2 400 | 240 | 715 | 72 | 30 % | 0,13 | < 50 tadpoles |
| 10 | 14 | a 400 | 240 | 1 437 | 144 | 60 % | 0 14 | |
| 10 | 22 | 3 000 | 300 | 2 269 | 227 | 76 % | 0.07 | |
| 10 | 22 | 3 000 | 300 | 841 | 64 | 28 % | 0.07 | |
| 10 | 14 | 2 400 | 240 | 672 | 67 | 28 % | 0.11, | |
| 10 | 14 | 2 400 | 240 | 941 | 94 | 39 % | 0.13 | |
| Totals: | 86 450 | 25 983 | ||||||
| Averages | 18 | 3 438 | 252.5 | 965 | 77.61 | 31 % | 0.28 | |
| Std. Dev. | 5.3 | 1 844 | 91.9 | 962 | 53.23 | 18 % | 0.29 |
It remains to be seen if it is realty necessary to raise Clarias fingerlings to 3 g size before stocking, and if necessary, can the fish attain this size in the small 10 m2 ponds by prolonging the rearing time to 30 to 45 days. Indications are that this would be possible as growth is apparently not restricted by the small pond environment (see below),
3.3,2 Yield of fish per m2
It is common practice to present data for larval pond production in terms of fish per m/2, With all trials, the average number of fish harvested was 77/m2, with a maximum of 226 /m2 and a minimum of 6 /m2 (the later pond was Invaded with tadpoles). This mean value increases to 85/m2 if only data from the 10 m2 ponds is used.
Previous work in Kenya using large ponds (1 340 m2) for raising Clarias larvae gave an average yield of 5.8 fish/m2, std 4 9 (Obuya et all, 1995). The highest value obtained was 21.6 fish/m2. De Graaf et a! (in press) found a mean value of 5.0 /m2 in unprotected, 100 m2 ponds, and a mean value of 32.3 fingerlings /m2 for ponds protected by surrounding them with roofing sheets. Their highest recorded value was 61.9 /m2. Our mean value of 55 fish /m2 for 10 m2 ponds compares very favorably.
One should hasten to add that there are 2 major factors that inhibit an in-depth comparison of this data with that in the literature. A pond of 10 m2 is very small; caution should be taken with direct comparison to results of larger ponds.
The other consideration is the short duration of the current experiments; most trials reported in the literature last 30 to 45 days. The 9 initial trials lasted 22 to 26 days and the 17 final trials lasted 14 to 15 days. These later trials were intentionally curtailed due to the need to obtain as much data as possible in the short time available, This was done on the basis that the survival rates were comparable between the shorter and longer trials, indicating that most of the larval mortality occurs very early after stocking and does not increase substantially as the trials continue. Additionally, the mean weight of harvested fish held at least 3 weeks in the 10 m2 ponds was 5 to 10 times that of those harvested after 2 weeks, indicating that the small pond environment is not limiting to growth, or at least for the first 22 days.
The increased ratio of dike area to pond surface appears to have a positive influence. From this perspective, it is possible to make a preliminary comparison between these very small ponds and those of 100 m2 sited in the literature (de Graaf et al, in press). An adjacent series of 6 ponds of 10 m2 each has a total surface area, including the separating dikes, of 100 m2. The average yield from each of these 6 ponds is 850 fish, or a total of 5 100 fingerlings per 100 m2. This compares favorably with data from protected ponds of 100 m2 water surface area, where the yield was 3 230 fish per 100 m2.
3.3.3 Survival rates
Survival varied from 6 to 76 % within the individual trials, The lower values were found where frogs and tadpoles were present and in 3 trials where a few large (4 g) fingerlings were found and cannibalism occurred (see table II). With these later 3 trials, it is strongly suspected that these few large individuals were fish left accidently in the pond from the previous experiment as the ponds had been emptied and refilled the next day. In no other trials did we find such a large size differentiation in the harvested fish, although this is common if Clarias fingerlings are left for over 30 days in a pond. AN the large fish in these 3 trials were of a uniform size as were all the smaller fish; there were no individuals with sizes in between. After this experience, ponds were more carefully emptied and dried for at least 1 week before restocking, and this did not occur again. With all data, the mean survival rate from individual trial values was 31 %. If one excludes those ponds where tadpoles and the left-over fish were present, the mean survival rate increases to 36 %,
3.3.4 Amphibian control
With the 10 m2 ponds, tadpoles were only present in 3 out of 18 trials (16 %) but at negligible numbers (less than 50 per pond). Kibos has a flock of 20 sheep (used to keep the grass short), and for some reason, they were very much attracted to the grass protruding through the mesh of the nets. The sheep occasionally tore the nets with either their feet or teeth, allowing a very few frogs to enter in some of the trials. It was not possible to eliminate frogs and tadpoles in the 50 m2 pond in any of the 4 trials done, and they out-numbered and out-weighed the Clarias at harvest on 3 occasions. This would indicate that it is easier to control frogs in the smaller, 10 m2 ponds.
3,3.5 Total fish yields
The most important figure is the total number of fish recovered during each trial. With ail data, the average harvest was 966 fish per pond. Using only the 10 m2 ponds, the average fails to 693, with a maximum of 2 269 and a minimum of 279 fingerlings harvested in each pond, This level of production should prove adequate for the small scale farmer.
One of the most important considerations is the economic return available to the farmer. This system appears to offer an adequate return. The current cost (September 1995) of the syphons, filtration system, spawning and incubation container, and the syringes, salt, and other husbandry equipment is approximately Ksh 1 200 (1 US$ = Ksh 55). The estimated life-span is 1 year. A 10 m2 pond costs about Ksh 500 for labour, Ksh 500 for a 2" PVC drain pipe and elbow, and a 11 × 2 m piece of 4 mm mesh net costs Ksh 360, for a total of Ksh 1 360. The life of the pond can be also estimated at 1 year, although with maintenance it would be far longer than this. At this time, we would recommend 2 ponds of 10 m2 each for each spawn to divide the risk.
The total cost of each installation with spawning equipment and 2 ponds would be Ksh 3 920, rounded to Ksh 4 000. Feeding and manuring costs are negligible, and one can assume that the cost of the breeders and pituitary donor is compensated by sale or consumption of the fish following the exercise. With an average yield of about 700 fish per pond, this would give about 1 400 fish available for sale or stocking at the end of the cycle. The current price for Clarias fingerlings varies between Ksh 1 and 5 apiece. The potential revenue is Ksh 1 400 to 7 000 per cycle. Using these price extremes, it is possible to recoup the investment in 1 to 3 cycles or within 6 months. This would indicate an annual rate of return of at least 100 %, which gives a reasonable economic basis for future expansion. All this must be verified in actual practice before a conclusive economic analysis can be made.
