Remedy 1: Comprehensive modelling studies of measurement process.
Remedy 2: Empirical determination of true resolution by comparison with high-resolution data.
Remotely sensed data are often considered as column-like or point-like samples of an atmospheric variable, associated for instance with the location of a ground-based instrument. This is also the general assumption for satellite data, which are assumed to represent the column or profile at the vertical of the satellite field-of-view footprint in case of nadir sounders, and atmospheric concentrations along a vertical set of successive tangent points in the case of limb and occultation sounders. In practice, the quantities retrieved from a remote sensing measurement integrate atmospheric information over a tri-dimensional airmass and also over time. E.g., ground-based zenith-sky measurements of the scattered light at twilight integrate stratospheric UV-visible absorptions (by ozone, NO2, BrO etc.) over several hundreds of kilometres in the direction of the rising or setting Sun (Lambert et al., 1997). A satellite limb measurement will actually be sensitive to the atmosphere along the entire line-of-sight towards the photon source, depending on the specific emission, absorption, and scattering processes at play (e.g. von Clarmann et al., 2009). Similarly, in-situ measurements of atmospheric profiles cannot be associated with a single geo-location and time stamp, due for instance to balloon drift for ozone- and radiosondes. In a variable and inhomogeneous atmosphere, this leads to additional uncertainties not covered in the 1-dimensional uncertainties reported with the data (e.g. Lambert et al., 2011). A prerequisite for quantifying these additional uncertainties of multi-dimensional nature is not only a quantification of the atmospheric variability at the scale of the measurement (cfr. G3.01), but also a detailed understanding of the smoothing and sampling properties of the remote sensing system and associated retrieval scheme. Pioneering work on multi-dimensional characterization of smoothing and sampling properties of remote sensing systems and associated uncertainties was initiated during the last decade (e.g. in the EC FP6 GEOmon project and in the current EC H2020 GAIA-CLIM project), but in the context of integrated systems like Copernicus and GCOS, appropriate knowledge of smoothing and sampling uncertainties, still missing for several ECVs and remote sensing measurement types, has to be further developed and harmonized.
Addressing this gap was a major objective of GAIA-CLIM, within which specific tasks were dedicated to the characterisation of smoothing and sampling properties of selected instruments and for selected ECVs. Results have been obtained for total ozone columns, for ozone, temperature, and humidity profiles, and for aerosol columns and profiles from a diverse set of ground-based instruments. Regarding satellite data, only a selection of current missions were explored. Results were made available in technical notes, namely D3.4 (“Report on measurement mismatch studies and their impact on data comparisons”) and D3.6 (“Library of (1) smoothing/sampling error estimates for key atmospheric composition measurement systems and (2) smoothing/sampling error estimates for key data comparisons”), and through the ‘Virtual Observatory’. In the long term, this gap will require continued efforts to fully characterize the spatiotemporal smoothing and sampling properties of both new ground-based instruments and upcoming satellite sensors. Hence the gap requires constant re-evaluation as technology and observing programs evolve.
G3.01. To be addressed before G3.04
Argument: A quantification of the uncertainties that result from the specific sampling and smoothing properties of an instrument requires information on the spatiotemporal variability of the atmospheric field.
G3.06. To be addressed after G3.04
Argument: Error/uncertainty budget decomposition of a comparison requires a proper understanding of the smoothing and sampling properties of the instruments involved, i.e. requires G3.04 to be remedied.
G6.03. To be addressed after G3.02
Argument: Deciding on the best time and location for targeted reference observations should be informed by information on the actual sampling and smoothing properties of the measurement systems.
- Lambert, J.-C., et al., Comparison of the GOME ozone and NO2 total amounts at mid-latitude with ground-based zenith-sky measurements, in Atmospheric Ozone - 18th Quad. Ozone Symp., L’Aquila, Italy, 1996, R. Bojkov and G. Visconti (Eds.), Vol. I, pp. 301-304, 1997.
- Lambert et al., “Multi-dimensional characterisation of remotely sensed data”, EC FP6 GEOmon Technical Notes, 2011
Lambert, J.-C., et al., “Comparing and merging water vapour observations: A multi-dimensional perspective on smoothing and sampling issues”, in “Monitoring Atmospheric Water Vapour: Ground-Based Remote Sensing and In-situ Methods”, N. Kämpfer (Ed.), ISSI Scientific Report Series, Vol. 10, Edition 1, 326 pp., ISBN: 978-1-4614-3908-0, DOI 10.1007/978-1-4614-3909-7_2, © Springer New York 2012
- Von Clarmann et al., “The horizontal resolution of MIPAS”, AMT v2, 2009
- Seidel et al., “Global radiosonde balloon drift statistics”, J.G.R. v116, 2011
- Sihler, H., Lübcke, P., Lang, R., Beirle, S., de Graaf, M., Hörmann, C., Lampel, J., Penning de Vries, M., Remmers, J., Trollope, E., Wang, Y., and Wagner, T.: In-operation field-of-view retrieval (IFR) for satellite and ground-based DOAS-type instruments applying coincident high-resolution imager data, Atmos. Meas. Tech., 10, 881-903, https://doi.org/10.5194/amt-10-881-2017, 2017.
This gap concerns the need for a more detailed characterisation of the actual spatiotemporal smoothing and sampling properties of both satellite-based EO measurements and ground-based in-situ or remote-sensing measurements. Indeed, EO measurements are most often associated with single locations, or at best pixel footprints, while in fact the actual measurement sensitivity covers a larger spatiotemporal extent, due for instance to the radiative transfer determining the measured quantities, or the actual measurement geometry (choice of line-of-sight, trajectory of a weather balloon, etc.). In an inhomogeneous and variable atmosphere, this leads to additional errors and uncertainties that are not part of the reported measurement uncertainties, but still need to be quantified, in particular when performing comparisons with other types of measurements, with different smoothing and sampling characteristics. For several ECVs and measurement techniques, significant work is needed to (1) determine/model the actual spatiotemporal smoothing and sampling properties, and (2) quantify the resulting uncertainties on the measurements of the variable atmosphere.