G3.01    Incomplete knowledge of spatiotemporal atmospheric variability at the scale of the measurements and their co-location

Gap detailed description

Spatiotemporal variability of the atmosphere at the scale of the airmass being measured or - in the case of a measurement intercomparison - at the scale of the co-location, leads to additional uncertainties, not accounted for by the uncertainty budget reported with an individual measurement. To quantify these additional uncertainties (cfr. gaps G3.04 and G3.06), or to ensure that they remain negligible through the use of appropriate co-location criteria (cfr. G3.03), a prerequisite is a proper understanding of atmospheric variability of the targeted ECV on those scales.

While scales above approx. 100km/1h are relatively well captured for several GAIA-CLIM target ECVs in model or satellite gridded data (e.g. Verhoelst et al., 2015, for total ozone), information on smaller scales is most often restricted to results from dedicated campaigns or specific case studies, e.g. Sparling et al. (2006) for ozone, Hewison (2013) for meteorological variables, and Pappalardo et al. (2010) for aerosols. Due to the exploratory nature of these studies, neither global nor complete vertical coverage is achieved. For instance, information on small-scale variability in the ozone field is limited to altitudes and regions probed with dedicated aircraft campaigns. The validation of satellite data records with pseudo global networks of ground-based reference instruments on the other hand requires an appropriate quantification of atmospheric variability in very diverse conditions, covering all latitudes, altitudes, dynamical conditions, degrees of pollution etc..

This gap therefore concerns the need for a better, more comprehensive, quantification of the spatiotemporal variability of the ECVs targeted by GAIA-CLIM. Note that this gap is also closely related to WP4 gap G4.06, which deals with the impact of natural variability on measurement-model comparisons, and with WP1 gap G1.07, dealing with the assessment of gaps in the existing networks.

Activities within GAIA-CLIM related to this gap

Resolving this gap is in general beyond the scope of the work planned within GAIA-CLIM. Nevertheless, some specific case studies in T1.4 (D1.9) targeting T, q, and aerosol load, and using measurements from polar orbiting sensors such as IASI and MODIS, will allow further insight and reveal in more detail to what extent this gap is a major hurdle in validation studies of these ECVs. These case studies can be considered to belong to the remedies based on (satellite) measurements (as detailed below).

Gap remedy(s)

Several approaches can aid in reducing and mitigating this gap. These can be either model or measurement based.

Remedy #1

Specific remedy proposed

High-resolution modelling studies at the global scale, resulting in comprehensive data sets of atmospheric fields.

Measurable outcome of success

The quality of the model output at its finest resolution can be estimated by comparison with high-resolution measurement data sets, preferably those with limited horizontal, vertical, and temporal smoothing effects, e.g. from balloon-borne sondes. Ideally, an agreement is found within the combined model and measurement uncertainty, and the model uncertainty itself is below that of the measurement (to maximize Relevance, see below). 

 Achievable outcomes

Improved spatio-temporal resolution in atmosphere models is a much broader scientific goal, with great computational and theoretical (e.g. convection and turbulence treatment) challenges. As such, this remedy probably requires a level of effort and resources beyond what can be driven by the need for satellite data validation. The technological/ organizational viability is therefore considered medium and the cost estimate high.


If successful, this remedy would largely close the gap, and it would facilitate remedies for most other gaps identified here through the use of OSSEs (Observing System Simulation Experiments) based on these model fields.


The development and computation of higher resolution (re-)analysis data sets at the major meteorological centres (e.g. ECMWF) typically requires several years of sustained effort.

Remedy #2

Specific remedy proposed

Statistical analysis of existing and future satellite and non-satellite high-resolution data sets, which allows to separate the contribution of atmospheric variability from the total uncertainty budget of a data comparison, e.g. using so-called ‘structure functions’ or heteroskedastic functional regression. Within the geographical and temporal coverage of the data set, these methods produce an estimate of the variability (or auto-correlation) of the field.  Note that, as for Remedy #1, the scientific interest for higher resolution in the data sets is much broader than only the validation needs, e.g. for the identification of emission sources in an urban environment.

Measurable outcome of success

Publications describing for the different ECVs and various atmospheric regimes, locations and altitude ranges the atmospheric variability at scales ranging from those of in-situ measurements (e.g. 10s of meters for balloon sonde measurements)  to that of a satellite pixel (several 10s to 100s of kilometres).  These can be based either on exisiting data sets, or represent an exploitation of newly designed campaigns and missions.

Achievable outcomes

The technological and organizational effort required to make step changes in the spatiotemporal resolution of the observational data sets is in general very large, and comes with a large financial cost (more than 5 M euro), in particular if global coverage is aimed for.  Hence, such developments need a much larger user base and the use proposed here should be considered secondary to the scientific objectives of such new missions. Nevertheless, smaller dedicated campaigns with for instance aircraft or Unmanned Aerial Vehicles (UAVs)  can offer great insight at particularly interesting sites, and this at medium cost (between 1 and 5 M euro).


This remedy directly addresses the gap, as already illustrated for instance with aircraft data for ozone by Sparling et al. (2006).


The statistical analysis itself requires only a moderate amount of time (of the order of several months), but the design of dedicated field campaigns (up to a few years) or new satellite missions takes much longer. 

Gap risks to non-resolution

Identified future risk / impact

Probability of occurrence if gap not remedied

Downstream impacts on ability to deliver high quality services to science / industry / society

Unknown impact of co-location mismatch on comparisons performed for validation purposes.

Very high

Interpretation of satellite data validation results severely hampered. This impacts negatively the reliability of the data sets, the reported uncertainties, and the products and services derived from these.


Work package: