The optimal allocation of computational resources for efficient and high performance Grid applications is not a trivial task and mainly arises a challenge for calibration and execution of huge hydrological models. Hydrological models are used to assess the sustainability and vulnerability of different geographical regions. However, the calibration process that is necessary for the SWAT hydrological models demands high computational resources. The gSWAT application allows the calibration in a flexible manner, by a Web-based interface, and by taking advantage of the computational resources provided by the Grid infrastructures. This paper presents one component of the gSWAT platform that deals with resource allocation, and highlights the parameters that should be taken into account by this component. These parameters are related with the complexity of the SWAT model but also with the availability of Grid resources. The paper argues by experimental results the impact of such parameters on the execution time, and generally on the system performance.