On the use of multi-objective optimization for multi-site calibration of extensive green roofs

Conceptual hydrological models are practical tools for estimating the performance of green roofs. Such models require calibration to obtain parameter values, which limits their use when measured data are not available. One approach that has been thought to be useful is to transfer parameters from a gauged roof calibrated locally (single-site calibration) to a similar ungauged roof in a different location. This study tested this approach by transferring calibrated parameters of a conceptual hydrological model between sixteen extensive green roofs located in four Norwegian cities. The approach was compared with a multi-site calibration scheme that explores trade-offs of model performances between the sites. The results showed that single site calibration could yield optimal parameters for one site and perform poorly in other sites. In contrast, obtaining a common parameter set that yields satisfactory results (Kling Gupta Efficiency >0.5) for different sites, and roof properties could be achieved by multi-site calibration.