As information on fisheries under varying exploitation is restricted for a large number of African waters, the most useful approach to assess potential fish yields of these waters is through comparative studies of similar fisheries for which information is available. Models developed through these studies should have limited data requirements and the variables used in these models should be in the units in which they were originally measured. Existing models using ratios like yield per unit of area, mean depth and MEI were not included in this study because statistical results based on these ratios are prone to spurious self-correlations contributing to inflated correlations and incorrect interpretation.
Attention was focused on models relating total catch to morphometric and edaphic factors and fishing effort. Analysis of the information in the Source Book revealed limited availability of these data for most African inland waters.
As surface area for most African lakes and reservoirs is known, a first estimate of potential fish yield can be made using the Catch vs Area model. Using the average catch per fishermen of 2.3 tonnes per year based on the Catch vs Effort model, it is possible to estimate the fishing effort necessary to realize the potential fish production.
Potential yields of rivers can be estimated using models based on the length or the drainage area of the river. Potential fish production of floodplains can be estimated using the Catch vs Floodplain Area model.
The reliability of the yields predicted by these models, which are only based on a single morphometric parameter, is restricted. As long as data on catch and effort are unreliable for most African inland waters, and morphometric and limnological data are limited, improvement of these models will be difficult.
Collection of standardized data on fish biomass, chlorophyll, nutrient levels and morphometric and climatic factors for a selected number of African inland waters, as done by Quiros (1990) in Argentina, could be an approach for more reliable models.
