Correction of temperature and relative humidity biases in ERA5 by bivariate quantile mapping: Implications for contrail classification

Abstract. The skill of the atmospheric reanalysis ERA5 from the European Centre for Medium-Range Weather Forecasts (ECMWF) at simulating upper atmospheric temperature and relative humidity is assessed by using five years of In-service Aircraft for a Global Observing System (IAGOS) observations. IAGOS flight trajectories are used to extract co-located meteorological conditions – temperature, relative humidity, and wind speed – and are compared with the IAGOS measurements. This assessment is particularly relevant to the study of contrail formation, so focuses on the highly frequented air space that spans the Eastern United States over the North Atlantic and to central Europe. The comparison is performed in terms of mean, median, probability density functions, and a confusion matrix. For temperature a good agreement is identified with a maximum bias of −0.4 K at the 200 hPa level. Larger biases are found for relative humidity with up to −5.5 % at the 250 hPa level. To remove the systematic biases, which mostly tend towards too dry and cold, a bias correction method, based on a multivariate quantile technique, is proposed and applied. After the correction the bias in temperature is reduced to below 0.1 K and in relative humidity to below −1.5 %. To estimate the representation of contrail occurrence in ERA5, data points from IAGOS as well as corrected and uncorrected data points from ERA5 of temperature and relative humidity are flagged for contrail formation using the Schmidt-Appleman–criterion. In the IAGOS data set 39.2 and 16.9 % of the samples represent conditions for non-persistent contrails and persistent contrails, respectively. The corresponding numbers for original ERA5 analyses are 40.8 and 17.5 %, respectively, indicating good agreement overall. Applying a proposed quantile mapping correction method and removing the biases in temperature and relative humidity has only a small effect on the distributions but leads to an overestimation of non-persistent contrail occurrence (44.0 %) and underestimation of persistent contrails (16.8 %). Differences in contrail occurrence that remain after the bias correction are traced back to the underling biases in temperature and relative humidity, indicating that ERA5 is either too dry and warm or cold and moist with largest differences at 250 hPa and decreasing with increasing altitude.