Presently, limited availability of traceable uncertainty estimates for non-satellite measurement techniques propagates to other applications, such as satellite characterisation. Such applications would be significantly improved were traceable uncertainty estimates more broadly available on the comparator measurements. The development work of the GAIA-CLIM Virtual Observatory has been addressing the selection of reference data, provision of measurement and co-location uncertainty estimates, and the provision of match-ups with satellite data to be characterized. This work has highlighted the relative geographical paucity of reference quality qualified measurement systems and their concentration in certain regions, principally Northern Hemisphere mid-latitudes. It can be expected that for other ECVs in atmospheric, but also oceanic and terrestrial domains, similar issues exist.

The issue of uneven geographical distribution of high-quality observation sites pervades many observational networks. In earlier versions of the GAID, a number of gaps pertaining to weaknesses in individual networks were identified. On further reflection, these gaps are sufficiently similar that the underlying challenges, and therefore solutions, were better addressed collectively through a recognition that this uneven sampling is a generic cross-cutting issue requiring a holistic, rather than per network consideration from the perspective of end-users, such as satellite calibration and validation activities. Compounding that is a lack of work that extends that knowledge to enable utilisation of remaining observations with requisite confidence.

While a vast amount of data are potentially available, unfortunately, the uncertainty of these data is all too often - in a metrological sense - insufficiently specified, estimated or even unknown, which frequently limits the applicability of the measurements to uses such as satellite characterisation. In order to achieve progress, it is critical to have data records that are stable over time, metrologically traceable to the method of measurement, uniformly processed worldwide (and thus comparable), and based on traceable references. This will allow us to establish the robust scientific basis for using such data as a transfer standard in satellite-dataset characterization and other activities, and for assessing the cost-effectiveness of potential observing system enhancements.

Thorne et al. (2017) provide the rationale behind and defining characteristics of a system-of-systems approach of “reference”, “baseline” and “comprehensive” networks. In that work, it is recognised that datasets from baseline and comprehensive networks provide valuable spatiotemporal coverage, but lack the metrological characteristics needed to facilitate traceable uncertainty estimates. It is therefore essential to identify the scope for baseline and comprehensive networks to leverage expertise from reference networks, including adopting elements of best practice from reference networks, and/or facilitating reprocessing that iteratively improves dataset quality. Such work may increase their utility for a range of applications, including satellite characterisation.