This paper explores new techniques based on sparse representations for estimating the spectral response functions of high-resolution spectrometers. The method is highly competitive, with commonly used parametric models yielding more accurate estimates while accounting for wavelength dependence. The resulting normalized estimation errors of the spectrometer spectral responses are less than 1 %, which will allow for better quantification of trace gas concentrations at the Earth surface.

Ice-nucleating particle (INP) observations often differ by 1 order of magnitude, when multiple instruments measure in parallel. As INP measurements are sensitive to the activation conditions, inaccuracies in the temperature measurement may induce significant deviations in the INP concentration. It was found that typical temperature uncertainties reported in intercomparison studies may explain divergences of a factor of 2 (± 0.5 °C) or 5 to 10 (± 1 °C) between two INP counters with opposite bias.