Temporal and spatial patterns in model evaluation – a comparison of PALM simulations for German cities with crowdsourced air temperature data

Rapid development of microscale urban climate models in the previous years requires ongoing model evaluation under different scenarios and conditions. The data for these can be acquired in intensive field campaigns, yet those are expensive and alternative observational data is required. In a previous study, we utilised crowdsourced air temperature data from Netatmo citizen weather stations for the evaluation of the PALM model during a hot summer day in the city of Bochum, Germany. The data proved valuable due to their high spatial resolution and thus a temporal pattern in model performance with an underestimation of air temperatures at nighttime was detected. However, uncertainties remained regarding the causes of the discrepancies between the measured and modelled data.

In this paper, PALM simulations for several cities in Germany (Dortmund, Cologne, and Berlin) are compared with respective crowdsourcing data to test the transferability and robustness of this approach. The cities were selected due to their varying sizes, densities, and geographical settings to identify factors which could influence model results. The simulations each cover a period of three days within a heat wave in August 2020, with meteorological conditions that are optimal for the formation of the canopy layer urban heat island (CUHI) to analyse the representation of microscale air temperature variations of the CUHI within the model.

First results show an overall high performance of the model with RMSE values close to 2 K. However temporal patterns in model accuracy can be detected and the model underestimates nighttime air temperatures. A special focus will be given to the potential causes for the underestimation of nighttime air temperatures. Furthermore, spatiotemporal patterns are investigated to identify geographical factors which influence model results. Finally, we will review whether the model evaluation with crowdsourced air temperature data can be transferred, and which restrictions exist.