An assessment of the climatological representativeness of IAGOS-CARIBIC trace gas measurements using EMAC model simulations

Abstract

Abstract. Measurement data from the long-term passenger aircraft project IAGOS-CARIBIC are often used to derive climatologies of trace gases in the upper troposphere and lower stratosphere (UTLS). We investigate to what extent such climatologies are representative of the true state of the atmosphere. Climatologies are considered relative to the tropopause in mid-latitudes (35 to 75° N) for trace gases with different atmospheric lifetimes. Using the chemistry–climate model EMAC, we sample the modeled trace gases along CARIBIC flight tracks. Representativeness is then assessed by comparing the CARIBIC sampled model data to the full climatological model state. Three statistical methods are applied for the investigation of representativeness: the Kolmogorov–Smirnov test and two scores based on the variability and relative differences. Two requirements for any score describing representativeness are essential: representativeness is expected to increase (i) with the number of samples and (ii) with decreasing variability of the species considered. Based on these two requirements, we investigate the suitability of the different statistical measures for investigating representativeness. The Kolmogorov–Smirnov test is very strict and does not identify any trace-gas climatology as representative – not even of long-lived trace gases. In contrast, the two scores based on either variability or relative differences show the expected behavior and thus appear applicable for investigating representativeness. For the final analysis of climatological representativeness, we use the relative difference score and calculate a representativeness uncertainty for each trace gas in percent. In order to justify the transfer of conclusions about representativeness of individual trace gases from the model to measurements, we compare the trace gas variability between model and measurements. We find that the model reaches 50–100 % of the measurement variability. The tendency of the model to underestimate the variability is caused by the relatively coarse spatial and temporal model resolution. In conclusion, we provide representativeness uncertainties for several species for tropopause-referenced climatologies. Long-lived species like CO2 have low uncertainties ( ≤ 0.4 %), while shorter-lived species like O3 have larger uncertainties (10–15 %). Finally, we translate the representativeness score into a number of flights that are necessary to achieve a certain degree of representativeness. For example, increasing the number of flights from 334 to 1000 would reduce the uncertainty in CO to a mere 1 %, while the uncertainty for shorter-lived species like NO would drop from 80 to 10 %.