This exercise clearly shows the difference in performance and the limitations of the three hydroacoustic systems when working in shallow lakes.
The conventional system with a single-piston transducer had problems in getting sufficient single-fish echoes on the transects to come up with reliable target-strength distributions. This clearly shows the limitations of the indirect method in shallow lakes where the small sample volume produced very few single-fish echoes. With only 4781 single-fish echoes on all the 11 transects together the system was used as a combined echocounter and an echointegration system. However, it came up with the mean fish density along each transect for fish ranging in target strength from - 56dB to - 37dB.
The split-beam system managed to resolve 4892 echoes from single fish on 7 transects. The target-strength distribution showed no distinct peaks, but had a maximum at - 45dB. The problem with the split-beam system was the absence of a working echo integrator. As delivered from Simrad, an echo-integrator system was not included. The system only produced target-strength distributions on resolved single fish along transects.
The dual-beam system collected 4210 echoes at 420 kHz over 23 transects. These echoes produced a target-strength distribution very much like that of the split beam.
The 95 % confidence intervals about all of the population estimates were overlapping due to the relatively high variance. The primary source of this variance was the patchy spatial distribution of fish densities, not the distribution of target strengths. This high sampling variability, which was qualitatively evident on the echograms, indicates that the spacing between transects was too wide. Future acoustic surveys for estimating the fish population in Lake Tegel should utilize more transects spaced closer together.
One important practical point is that fish stock assessment using echo systems still needs supplementary work with fishing gear, that is more time-consuming and labour-intensive than usually suggested (Dahm, 1986).
In the main experiment there were two aims of fishing: (1) collecting previously marked fish (cf. 2.3), (2) identification of fish species for the analysis of the acoustic survey data. The individual catches differed according to the type of gear as well as to depth and region of sampling (cf. 3.1), but the data provided some potential basis for calibrating the echo sounders. There were two main peaks in the length-frequency histograms which could be attributed to A. cernua and B. björkna respectively. Unfortunately the size-distributed patterns of the A. brama that were caught are rather flat, lacking clear and consistent peaks. Additionally, there occurs much overlap between species in the length-frequency histograms. Moreover, experience from the comparable Lake Constance experiment (Dahm et al., 1985) suggests that the catch from Lake Tegel was also biased towards young (but older than young-of-the year) and/or midsized fishes.
The specific fish biomass in Lake Tegel was in the same order of magnitude as in other hypertrophic lakes (Børgstrom and Eie, 1981; Doering, 1986; Lammens et al., 1985; Völzke, 1984).
The conventional population estimate from Lake Tegel indicated a much higher fish population than derived from echosounding and in this followed the pattern from Lake Constance (cf. 6). This result should not come as a surprise (Mulligan and Kieser, 1986). The hydroacoustic estimate was made from registrations of fish from 2 meter depth down to the bottom. The bottom echo will always cause problems of detection of fish close to the bottom. In the conventional single-beam system the registration of fish was stopped 1 meter above the bottom. In this shallow lake this means that some 25 % of the water volume was not sampled. Catch data show that fish were present both at the surface and at the bottom. If the shore area is added which was not surveyed, the hydroacoustic estimates should probably be increased by about 30 – 40 %. Under these conditions the difference in the population estimates made with the different techniques would not be that great. Indeed the estimates should lie within each others confidence limits.
