Uncertainty quantification in multi-parameter and multi-physics ensemble simulations of urban climate and building energy demand using major UCM+BEMs

  Several tens of urban canopy models (UCMs) have been developed, and some of those have been coupled with building energy models (BEMs) for dynamic anthropogenic heat (Q_F) representation. While several of those UCM+BEMs have been evaluated in model intercomparison using meteorological outputs, BEMs’ outputs have not sufficiently due to the lack of observations. Considering potential of UCM+BEMs for design of climate change mitigation and adaptation strategies attributable to their modelling of the interaction between urban climate and building energy demand, UCM+BEMs should be compared including BEMs’ outputs to clarify uncertainties in their predictions.

  Hence, the authors conducted ensemble simulations applying 4 major UCM+BEMs with WRF to the summer Tokyo where high resolution electricity consumption (EC) data were obtained. The 4 UCM+BEMs adopt relatively complex UCMs (BEP-BEM and CM-BEM) and simple ones (SLUCM-BEM and TEB-BEM) with different complexity in BEMs. As for the input parameters related to canopy geometry and building properties, 3 datasets were used. Those are gridded parameter dataset with 1km resolution and its low-resolutionized ones containing parameters averaged into either the conventional urban three categories or the LCZ urban 10 categories. Then the simulations were carried out for the combinations of 4 UCM+BEMs and 3 parameter datasets, and those results were analyzed for 2-m temperature (T2), 10-m wind speed (U10), EC, EC for air conditioning (ECAC), and building Q_F (Q_Fb) using those observations as references. The resultant statistics indicated larger model-dependent uncertainties compared to parameter-dependent ones regarding T2, U10 and ECAC but with similar uncertainties for EC and Q_Fb suggesting relative importance of model physics and parameters in UCM+BEMs simulations. The magnitude of uncertainties among variables showed relationship ‘T2 < EC ≈ ECAC ≈ Q_Fb < U10’. The combinations of complex UCM+BEMs and gridded parameters produced most realistic simulations for EC and ECAC (so probably for Q_Fb).