Remedy 1: Comprehensive modelling studies of measurement process.

Detailed modelling of the measurement process, including multi-dimensional radiative transfer if applicable, to quantify the 4-D measurement sensitivity. An example are multi-D averaging kernels for retrieval-type measurements. This work requires a significant effort from the instrument teams, for which dedicated, though still relatively low (per instrument), resources are required, in particular for code modifications and additions. If appropriate, the results from these detailed calculations can be parametrized for easy and efficient use when calculating the resulting errors and uncertainties for large amounts of data. This uncertainty calculation is done by combining the quantification of the measurement sensitivity with knowledge on the spatiotemporal variability of the atmospheric field (cf. G3.01). When these detailed modelling studies are out of reach, a similar estimate of the multi-D measurement sensitivity can be made in a more pragmatic way based on the measurement principle and physical considerations (e.g. Lambert et al. 2011), or it can in some cases be estimated with empirical methods by comparing data sets with differing resolution. Note that an essential prerequisite is the availability of all required metadata with the measurements, such as viewing angles or GPS trajectories.