8 th Atmospheric Composition Constellation Meeting ACC8 18

8 th Atmospheric Composition Constellation Meeting (ACC-8), 18 -19 April 2012, Columbia, MD Smoothing and Sampling Issues Affecting Data Comparisons Jean-Christopher Lambert (1) with contributions by C. De Clercq (1), Q. Errera (1), J. Granville (1), and T. von Clarmann (2) (1) Belgian Institute for Space Aeronomy, Brussels, Belgium Karlsruhe Institute of Technology, Karlsruhe, Germany (2) CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Smoothing and sampling issues affecting data comparisons 1. 2. Intro: The ideal observation operator? Illustrations 1. 2. 3. 4. 3. Assessment of smoothing errors Interpretation of comparisons Optimised co-location criteria Trace-tracer correlations, hydrogen budget… Conclusion Reported work funded by EC FP 4 ESMOS and SCUVS, EC ECRP 4 COSE, BELSPO/Pro. DEx SECPEA and EC FP 6 GEOmon. Continuation within BELSPO/Pro. DEx A 3 C and EC FP 7 NORS. CEOS ACC-8, Columbia, MD, April 18 -19, 2012

The ideal observation operator? For ideal co-location criteria and ideal data assimilation… CEOS ACC-8, Columbia, MD, April 18 -19, 2012

The ideal observation operator? Air masses probed by satellites and by NDACC ground-based instruments in the vertical AND horizontal dimensions CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Observation operators used in chemical data assimilation Errera et al. , Atmos. Chem. Phys. , 8, 2008 Underlying assumption: H(x(ti)) reproduces perfectly smoothing and sampling characteristics of the observation. CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Horizontal smoothing by MIPAS: O 3 2 -D horizontal Averaging Kernels for a 1 D profile retrieval MIPAS processor settings ESA IPF 4. 61/nominal mode von Clarmann et al, AMT, 2, 47 -54, 2009 CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Horizontal smoothing by MIPAS: T and H 2 O 2 -D horizontal Averaging Kernels for a 1 D profile retrieval MIPAS processor settings ESA IPF 4. 61/nominal mode von Clarmann et al, AMT, 2, 47 -54, 2009 CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Illustrations (1) Smoothing errors CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Smoothing error for O 3 column measurements – Ground-based zenith-sky observation at twilight Tarawa (Kiribati, 1°N / 173°E) Kerguelen (Indian Ocean, 49°S / 70°E) Dumont d’Urville (French Antarctica, 66°S / 140°E) CEOS ACC-8, Columbia, MD, April 18 -19, 2012 Lambert, ULB, 2006

Illustrations (2) Interpretation of comparisons CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Error budget of a data comparison Error budget of MIPAS validation vs. ozonesondes Method described in Section 4. 1 of Cortesi et al. , ACP 2007 CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Error budget of a data comparison Error budget of MIPAS validation vs. lidar Method described in Section 4. 1 of Cortesi et al. , ACP 2007 CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Error budget of a data comparison Sampling differences between MIPAS and radiosondes Lambert et al. , ISSI, 2012 CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Error budget of a data comparison Lambert et al. , ISSI, 2012 CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Illustrations (3) Co-location criteria CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Co-location for satellite NO 2 validation Selection within 500 km radius: pollution, meridian gradients, diurnal cycle… Lauder, New Zealand (45°S) – Pure stratospheric signal => meridian gradients, diurnal cycle Bauru, Brazil (22°S) – 500 km radius influenced by pollution from Sao Paulo area Lambert et al. , EUMETSAT O 3 M-SAF TN, 2007 CEOS ACC-8, Columbia, MD, April 18 -19, 2012 UVVIS data courtesy: CNRS / UNESP (Bauru) and NIWA (Lauder)

Co-location for satellite NO 2 validation Selection based on observation operators Lauder, New Zealand (45°S) – Pure stratospheric signal => meridian gradients, diurnal cycle Bauru, Brazil (22°S) – Zenith-sky air mass is over the Atlantic Lambert et al. , EUMETSAT O 3 M-SAF TN, 2007 CEOS ACC-8, Columbia, MD, April 18 -19, 2012 UVVIS data courtesy: CNRS / UNESP (Bauru) and NIWA (Lauder)

Illustrations (4) Tracer-tracer correlations, hydrogen budget… CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Tracer-tracer correlations and hydrogen budget Lambert et al. , ISSI, 2012 (with figures adapted from von Clarmann et al. , AMT 2009) CEOS ACC-8, Columbia, MD, April 18 -19, 2012

Conclusion q q q Bias and noise introduced by differences in smoothing and sampling can spoil the value of a data comparison. The problem is a combined effect of measurement properties (measurement + retrieval) and of atmospheric properties. The problem can be multi-dimensional. Observation operators have been/are being published for major remote sensing techniques and a few key molecules. Consideration of smoothing/sampling issues enables: q q q Optimization of co-location criteria Assessment of smoothing errors of individual observation systems Assessment of discrepancies due to differences in smoothing and sampling of atmospheric field More compact tracer-tracer correlations, hydrogen budgets… Propagation of smoothing/sampling properties in L 3/L 4/merged data sets ? CEOS ACC-8, Columbia, MD, April 18 -19, 2012

THANK YOU ! CEOS ACC-8, Columbia, MD, April 18 -19, 2012