Search

The relationship between fish yield, water quality and fishing effort was explored in 70 Finnish lakes. The total yield (kg ha−1 yr−1) was also split to species level (11 species). Fishing effort was estimated by the number of fishermen (fishermen km−2). The principal components of water quality variables were extracted. The first component comprised total nitrogen and phosphorus, water colour and oxygen saturation; the second was an association of pH, alkalinity and conductivity. Together they retained about 77% of the variation in the original data. The best single predictor for the total fish yield was the number of fishermen (R2 = 0.47), but also the second principal component entered as a variable in a stepwise regression model (R2 = 0.53). No regional differences were found, so the model is potentially valid for the whole country. However, the spread of the data around the model was considerable; close to 50% of the variation in yield remained unexplained. At the species level, fishing effort was needed as a variable in six models (in four of them as the first variable entering the model) and water quality in five models (four times as the first variable). The variation of yield explained by these models ranged from 21% to 49%. Generally, water quality was less important in determining the yield than fishing effort. This was confirmed by partitioning out the effect of effort variables from the correlation between yield and water quality. No obvious improvements in the coefficients were observed.

We examined the relationship between water quality and fish catch in 80 small Finnish lakes. Generally, fishing effort and gear may vary considerably among lakes; therefore, the detection of any relationship between water quality and fish catch may be hampered. To avoid this bias, fish in the present study were trapped with a standardised gill net (catch refers to the pooled biomass of trapped fish, catch per unit effort (CPUE). Water quality was assessed by measuring lake-specific values of pH, calcium, dissolved and total aluminium, conductivity, alkalinity and water colour. From these data, we extracted two principal components, together they retained 78% of the variation in the original data. Despite our efforts to minimise noise in the catch data, we could not find any predictive relationship between water quality and total biomass of fish. We conclude that in Finnish lakes it is difficult, if not impossible, to predict the lake-specific fish catch based on water quality. However, even with the present data there were differences in the total catch of fish originating from lakes with diverse values along the first principal component (a gradient composed of pH, calcium, conductivity and alkalinity). Lakes in the low end averaged a catch of 6.4 kg, while those at the high end averaged 9.2 kg. This was due to roach having increasing catches along the gradual transition from acid to neutral lakes.

One of the paradigms in the management of freshwater fish resources is that the annual yield is a function of water quality. We attempted to predict the yield of seven fish species (vendace, whitefish, perch, pike, burbot, roach and id) in 83 Northern Finnish lakes based on 16 variables of water quality. In the first step of our analysis, we demonstrate that the species have somewhat differing water quality requirements. This justifies the search for empirical models which could potentially be used to predict the yield of the species based on water quality of the lakes. In the pursuit of the predictive models, we used stepwise regression analysis in two differing ways. First, with a forward stepping method we attempted to find the best predictor combination out of the 16 water quality variables. The success was poor; only with perch did water conductivity fulfil the minimum criteria of a variable to be included in a regression model. However, even in this case the predictive power of the model remained poor ( R 2 = 0.06). Second, we began with a full regression model including the 16 water quality variables and also six variables characterising fishing effort (traps ha −1 ) in the study lakes. From the total of 22 variables, we removed the variables which were not needed in the regression model with a stepwise elimination method. The results were clear cut. In no case were water quality variables powerful enough to remain in the model. For six species, variables of fishing effort were accepted into the model (vendace gill nets for vendace, gill nets for whitefish and burbot, weirs for perch and id, and number of fishermen for pike). The coefficients of determination in these models were ∼ 50%. In roach, none of the 22 variables was powerful enough to be used in a predictive model.

The relationship between annual fish catch, fishing effort and water quality in 155 small (area 46–6500 ha) lakes originating from three different drainage areas in North-eastern Finland is studied. The three first principal components explain 70% of the variation in the original 12 water quality variables. Attempts are made to predict lake-specific yield with the help of the principal components. No common statistical model can be found for the three drainage areas. In all three drainage areas annual yield correlates positively with the fishing effort. Because one has to know beforehand the lake-specific effort this model has no practical applications in fisheries management. A common predictive model for the fish catch in these lakes was the positive relationship between the logarithm of lake area and the logarithm of catch per trap, and this holds even when lakes with the most effective traps are excluded. The most efficient way to predict the lake, specific value in fisheries terms is to regress (log) fish catch on (log) lake area. This model gives a good fit on all three drainage areas. Values of the predictor variable, lake area, are reliably estimated and there is also not much annual variation. This makes the empirical relationship between lake area and fish catch a rather simple but effective tool in lake management applications.