We introduce a new approach to hydrologic and water quality modelling on a watershed scale, which
automatically induces suitable models from domain modelling knowledge and measured data. A
central component of the proposed methodology is the domain knowledge library, which was
developed within this dissertation. The knowledge encoded in the library comprises hydrological and
nutrient loading processes on a watershed scale. For each conceptual process, the library contains
various alternative formulations. The library is written in a formalism compliant with the equation
discovery tool ProBMoT. Given a user specification of the modelling task, ProBMoT searches the
space of alternative candidate models encoded in the library. The generated models are optimized
against provided measured data. The best fitted model is proposed to the user as the best model of the
observed system. The methodology was applied to the Quarteira river catchment (Algarve, Portugal).
Analysis of the average annual water balance, sediment yields and nutrient loadings to the Quarteira
river supports the findings of the previous research work, justifying the use of the proposed
methodology for hydrological and water quality modelling at the watershed scale. In terms of
accuracy, ProBMoT results are comparable or better than the results obtained with SWAT.
Consequently, the proposed methodology presents a worthy alternative to the existing conceptual
modelling approaches. By testing two different scenarios (climate change scenario and scenario of the
increased wastewater treatment plant utilization), we confirmed that the generated hydrological and
water quality models are logically responding to the modified parameter values and input data.
Therefore, such models could be incorporated into decision support systems in the field of integrated
water resources management.