Evaluation of data time series in order to get information about water systems is one of the routinely needed tasks. The results are always associated with uncertainties, of which one arises from data scarcity. Traditional methods, such as regression analyses etc. become rapidly useless with decreasing number of data available. A method based on fuzzy set theory was applied to get more reliable information about the system from scarce databases. Monitored daily flow and water quality data of the medium size Zala River in Hungary were considered as elements of fuzzy sets. Fuzzy rules were generated form data pairs (flow, suspended solids concentration, water temperature and phosphorus load as inputs and output, respectively) from which combined rule bases were set up. These rule bases can be considered as a tool of mapping from the input space to the output space using defuzzification procedure. The method is trainable: it can learn from observations. It is demonstrated that the method is capable to generate daily phosphorus loads and annual balance with acceptable accuracy when it is trained only by weekly, biweekly or monthly data pairs. In comparison to other approaches the tool is well suited to utilize better the information content of scarce observations. Furthermore, monitoring costs can be considerably decreased without substantial information loss since sampling of expensive and labour intensive parameters can be reduced.

This content is only available as a PDF.