Abstract. Atmospheric trace gas measurements of greenhouse gases are critical in their precision and accuracy. In the past 5 years, atmospheric measurement and gas metrology communities have turned their attention to possible surface effects due to pressure and temperature variations during a standard cylinder's lifetime. This study concentrates on this issue by introducing newly built small-volume aluminum and steel cylinders which enable the investigation of trace gases and their affinity for adsorption and desorption on various surfaces over a set of temperature and pressure ranges. The presented experiments are designed to test the filling pressure dependencies up to 30 bar and temperature dependencies from −10 ∘C up to 180 ∘C for these prototype cylinders. We present measurements of CO2, CH4, CO and H2O using a cavity ring-down spectroscopy analyzer under these conditions. Moreover, we investigated CO2 amount fractions using a novel quantum cascade laser spectrometer system enabling measurements at pressures as a low as 5 mbar. This extensive dataset revealed that for absolute pressures down to 150 mbar the enhancement in the amount fraction of CO2 relative to its initial value (at 1200 mbar absolute) is limited to 0.12 µmol mol−1 for the prototype aluminum cylinder. Up to 80 ∘C, the aluminum cylinder showed superior results and less response to varying temperature compared to the steel cylinder. For CO2, these changes were insignificant at 80 ∘C for the aluminum cylinder, whereas a 0.11 µmol mol−1 enhancement for the steel cylinder was observed. High-temperature experiments showed that for both cylinders irreversible temperature effects occur especially above 130 ∘C.

21Keywords 22 ammonia in ambient air, traceability, reference gas standards, optical transfer standard, validation and testing 23 infrastructure 24 Abstract 25The environmental impacts of ammonia (NH 3 ) in ambient air have become more evident in the recent decades, 26 leading to intensifying research in this field. A number of novel analytical techniques and monitoring 27 instruments have been developed, and the quality and availability of reference gas mixtures used for the 28 calibration of measuring instruments has also increased significantly. However, recent inter-comparison 29 measurements show significant discrepancies, indicating that the majority of the newly developed devices and 30 reference materials require further thorough validation. There is a clear need for more intensive metrological 31 research focusing on quality assurance, intercomparability and validations. MetNH3 (Metrology for ammonia in 32 ambient air) is a three-year project within the framework of the European Metrology Research Programme 33 (EMRP), which aims to bring metrological traceability to ambient ammonia measurements in the 0.5 -34 500 nmol/mol amount fraction range. This is addressed by working in three areas: 1) improving accuracy and 35 2 stability of static and dynamic reference gas mixtures, 2) developing an optical transfer standard and 3) 36 establishing the link between high-accuracy metrological standards and field measurements. In this article we 37 describe the concept, aims and first results of the project. 38

Abstract. For many years, the comparability of measurements obtained with various instruments within a global-scale air quality monitoring network has been ensured by anchoring all results to a unique suite of reference gas mixtures, also called a “primary calibration scale”. Such suites of reference gas mixtures are usually prepared and then stored over decades in pressurised cylinders by a designated laboratory. For the halogenated gases which have been measured over the last 40 years, this anchoring method is highly relevant as measurement reproducibility is currently much better (< 1 %, k = 2 or 95 % confidence interval) than the expanded uncertainty of a reference gas mixture (usually > 2 %). Meanwhile, newly emitted halogenated gases are already measured in the atmosphere at pmol mol−1 levels, while still lacking an established reference standard. For compounds prone to adsorption on material surfaces, it is difficult to evaluate mixture stability and thus variations in the molar fractions over time in cylinders at pmol mol−1 levels. To support atmospheric monitoring of halogenated gases, we create new primary calibration scales for SF6 (sulfur hexafluoride), HFC-125 (pentafluoroethane), HFO-1234yf (or HFC-1234yf, 2,3,3,3-tetrafluoroprop-1-ene), HCFC-132b (1,2-dichloro-1,1-difluoroethane) and CFC-13 (chlorotrifluoromethane). The preparation method, newly applied to halocarbons, is dynamic and gravimetric: it is based on the permeation principle followed by dynamic dilution and cryo-filling of the mixture in cylinders. The obtained METAS-2017 primary calibration scales are made of 11 cylinders containing these five substances at near-ambient and slightly varying molar fractions. Each prepared molar fraction is traceable to the realisation of SI units (International System of Units) and is assigned an uncertainty estimate following international guidelines (JCGM, 2008), ranging from 0.6 % for SF6 to 1.3 % (k = 2) for all other substances. The smallest uncertainty obtained for SF6 is mostly explained by the high substance purity level in the permeator and the low SF6 contamination of the matrix gas. The measured internal consistency of the suite ranges from 0.23 % for SF6 to 1.1 % for HFO-1234yf (k=1). The expanded uncertainty after verification (i.e. measurement of the cylinders vs. each others) ranges from 1 to 2 % (k = 2). This work combines the advantages of SI-traceable reference gas mixture preparation with a calibration scale system for its use as anchor by a monitoring network. Such a combined system supports maximising compatibility within the network while linking all reference values to the SI and assigning carefully estimated uncertainties. For SF6, comparison of the METAS-2017 calibration scale with the scale prepared by SIO (Scripps Institution of Oceanography, SIO-05) shows excellent concordance, the ratio METAS-2017 / SIO-05 being 1.002. For HFC-125, the METAS-2017 calibration scale is measured as 7 % lower than SIO-14; for HFO-1234yf, it is 9 % lower than Empa-2013. No other scale for HCFC-132b was available for comparison. Finally, for CFC-13 the METAS-2017 primary calibration scale is 5 % higher than the interim calibration scale (Interim-98) that was in use within the Advanced Global Atmospheric Gases Experiment (AGAGE) network before adopting the scale established in the present work.

This publication revises the deteriorated performance of field calibrated low-cost sensor systems after spatial and temporal relocation, which is often reported for air quality monitoring devices that use machine learning models as part of their software to compensate for cross-sensitivities or interferences with environmental parameters. The cause of this relocation problem and its relationship to the chosen algorithm is elucidated using published experimental data in combination with techniques from data science. Thus, the origin is traced back to insufficient sampling of data that is used for calibration followed by the incorporation of bias into models. Biases often stem from non-representative data and are a common problem in machine learning, and more generally in artificial intelligence, and as such a rising concern. Finally, bias is believed to be partly reducible in this specific application by using balanced data sets generated in well-controlled laboratory experiments, although not trivial due to the need for infrastructure and professional competence.

To answer the needs of air quality and climate monitoring networks, two new gas generators were developed and manufactured at METAS in order to dynamically generate SI-traceable reference gas mixtures for reactive compounds at atmospheric concentrations. The technical features of the transportable generators allow for the realization of such gas standards for reactive compounds (e.g. NO2, volatile organic compounds) in the nmol · mol−1 range (ReGaS2), and fluorinated gases in the pmol ⋅ mol−1 range (ReGaS3). The generation method is based on permeation and dynamic dilution. The transportable generators have multiple individual permeation chambers allowing for the generation of mixtures containing up to five different compounds. This mixture is then diluted using mass flow controllers, thus making the production process adaptable to generate the required amount of substance fraction. All parts of ReGaS2 in contact with the gas mixture are coated to reduce adsorption/desorption processes. Each input parameter required to calculate the generated amount of substance fraction is calibrated with SI-primary standards. The stability and reproducibility of the generated amount of substance fractions were tested with NO2 for ReGaS2 and HFC-125 for ReGaS3. They demonstrate stability over 1–4 d better than 0.4% and 0.8%, respectively, and reproducibility better than 0.7% and 1%, respectively. Finally, the relative expanded uncertainty of the generated amount of substance fraction is smaller than 3% with the major contributions coming from the uncertainty of the permeation rate and/or of the purity of the matrix gas. These relative expanded uncertainties meet then the needs of the data quality objectives fixed by the World Meteorological Organization.