Gaps in urban meteorological time series complicate the analysis and usage of datasets. Various gap-filling techniques exist, including the debiasing of ERA5 reanalysis data. Unfortunately, an extensive evaluation of these debiasing techniques is lacking for urban datasets. This research compares five gap-filling techniques for urban temperature time series, including three debiasing techniques that employ a learning period and time window to take into account the seasonal and diurnal ERA5 temperature bias. The evaluation, performed by filling manually constructed gaps, reveals that short gaps are best filled by linear interpolation, while longer gaps benefit from ERA5 debiasing. The bias correction is crucial for urban locations, with all debiasing techniques performing similarly. The exact length and placement of the learning period and time window have limited impact on the performance, however a symmetrical placement of the learning period with a minimum length of 10 days and a small time window provide the best outcome. Based on these results, a gap-filling algorithm is designed which efficiently fills all gaps in temperature time series by selecting the most optimal technique for each gap. The algorithm can reproduce the urban heat island effect, although a small over- or underestimation might occur.

中文翻译：

---

通过消除 ERA5 再分析数据的偏差来填补城市温度观测中的空白


城市气象时间序列中的差距使数据集的分析和使用变得复杂。存在各种间隙填充技术，包括 ERA5 再分析数据的去偏倚。不幸的是，对于城市数据集，缺乏对这些去偏差技术的广泛评估。本研究比较了城市温度时间序列的五种空白填充技术，包括三种采用学习周期和时间窗口来考虑季节性和昼夜 ERA5 温度偏差的去偏差技术。通过填充手动构建的间隙来执行的评估表明，短间隙最好通过线性插值来填充，而较长的间隙则受益于 ERA5 去偏置。偏差校正对于城市位置至关重要，所有去偏差技术的性能都类似。学习期和时间窗口的确切长度和位置对性能的影响有限，但是最短长度为 10 天且时间窗口的对称位置可提供最佳结果。基于这些结果，设计了一种间隙填充算法，该算法通过为每个间隙选择最佳技术来有效地填充温度时间序列中的所有间隙。该算法可以重现城市热岛效应，但可能会出现小幅高估或低估的情况。