Gap filling of urban temperature time series by debiasing ERA5

Urban meteorological time series often contain gaps, which pose a challenge for data analysis and further application of the data. To address this problem, numerous gap-filling techniques have been developed, including the debiasing of ERA5 reanalysis data. However, these ERA5 debiasing methods are often evaluated individually and only on rural datasets. Since the ERA5 bias is more pronounced for urban locations, the knowledge of the ERA5 debiasing techniques is not directly applicable from rural to urban datasets. To achieve accurate gap-filling of urban time series, it is essential to understand the performance of these ERA5 debiasing techniques in urban contexts.

We evaluated a total of five gap-filling techniques, including three ERA5 debiasing approaches that incorporate a learning period and time window to account for the seasonal and diurnal variations in the ERA5 temperature bias. Our analysis is performed by filling artificially created gaps in urban temperature time series, and reveals a good performance of linear interpolation for small gaps, while large gaps are more effectively filled using the ERA5 debiasing techniques. Our results highlight the importance of applying an ERA5 bias correction when dealing with urban datasets.

In addition, we examined the optimal length and placement of the learning period and time window. Our results suggest that these parameters have minimal influence on the overall gap-filling performance. Based on these findings, we developed a gap-filling algorithm tailored to urban temperature time series. This algorithm selects the most suitable gap-filling method for each gap and is able to reconstruct the urban heat island effect, although minor over- or underestimations may occur.