Consolidated ground-based assessment of MIPAS V5/V6/V7 full mission : temperature, ozone & altitude

Transcription

1 MIPAS QWG #41, ESRIN, 6-8 Apr 2016 Consolidated ground-based assessment of MIPAS V5/V6/V7 full mission : temperature, ozone & altitude D. Hubert, J. Granville and J.-C. Lambert Royal Belgian Institute for Space Aeronomy (BIRA-IASB) daan.hubert@aeronomie.be Acknowledgements Ozonesonde & lidar PIs ESA, BELSPO

2 Validation methodology Data sets & selection MIPAS V5 : IPF 5.05/R (29502 orbits) /R (4935 orbits) V6 : ML2PP 6.00/U (35034 orbits) V7 : ML2PP 7.03/W (34929 orbits) Ground-based data Ozonesonde : ground 33km, 82N 90S (85 sites) Ozone lidar : 15-47km, 80N-67S (12 sites) Temperature lidar : km, 80N 20S (9 sites) Data selection MIPAS screening : V5 = V5 Readme file V6 & V7 = V6 Readme file Ground screening : as recommended in the literature Co-location criteria : (within 500 km and ±6h) & (closest SAT-GND pair)

3 Validation methodology Preprocessing Baseline profile representation Unit T / O3_vmr / altitude Vertical coordinate pressure Grid levels SPARC-DI levels Conversion correlative data Ozonesonde (pressure, O3_vmr, T) (altitude, O3_ndens) using measured T profile Ozone lidar (altitude, O3_ndens) (pressure, O3_vmr) using ERA-I T profile Temperature lidar (altitude, T) (pressure, T) Vertical smoothing Correlative data smoothed with MIPAS vertical AK in (pressure, T) or (pressure, O3_vmr) system SPARC-DI grid selection T time series selection O3 time series sonde range O3 lidar range T lidar range

4 Validation methodology Statistics Time series of absolute or relative difference to ground-based Δ = X MIPAS X ground [K, m] or Δ = 100*(X MIPAS X ground )/X ground [%] Time series of differences between MIPAS versions δ VX = (Δ VX Δ V6 ) [K, m] or δ VX = 100*(Δ VX Δ V6 ) / (1 + Δ V6 ) [%] Nearly identical to MIPAS VX MIPAS V6 or 100*(MIPAS VX /MIPAS V6 1), subsampled to SAT-GND pairs Aggregated over latitude bands Median ( median bias ) and 68% inter-percentile ( IP68 comp ) of Δ-distribution

5 temperature* + regression analysis & annual cycle * Taken from ppt at WebEx meeting, 12 Feb 2016

6 Median bias : separated FR & OR

7 Comparison spread : separated FR & OR

8 Changes in comparison time series hpa (sonde) 12-month running median δ V5-V6 V7-V6

9 Changes in comparison time series hpa (lidar) 12-month running median δ V5-V6 V7-V6

10 Comparison time series hpa (sonde) 12-month running median Δ V5 V6 V7

11 Comparison time series hpa (lidar) 12-month running median Δ V5 V6 V7

12 Vertical drift profile Linear regression model OR only : Weighted average over lidar & sonde station sample 30S 60S 60S 90S

13 Annual cycle in bias (lidar)

14 Annual cycle in bias (sonde)

15 Annual cycle in bias hpa (sonde) 2-week running median Δ V5 V6 V7

16 Annual cycle in bias hpa (lidar) 2-week running median Δ V5 V6 V7

17 Conclusions temperature MIPAS V5 & V6 resemble each other more than V7 Full Resolution : worse quality V7 bias is generally worse (more negative & less consistent with OR) than that of V6 Comparison spreads are similar, and somewhat better for V7 in the UT/LS Optimized Resolution : similar quality Overall, bias and comparison spread are very similar for V5, V6 and V7 V7 clearly drifts to higher temperature relative to V6 (or V5) by ~ K / decade Relative to ground p 10 hpa : V5/V6 trend close to zero, V7 trend too positive p < 10 hpa : V5/V6 trend too negative, V7 trend close to zero Annual cycle in bias Seen for all MIPAS versions At mid-latitudes and Antarctica, much less clear in Arctic and Tropics MIPAS bias changes by ~1-2K between June-July and November-December MIPAS issue or caused by different sampling of a changing (semi-)diurnal cycle?

18 ozone* + regression analysis & annual cycle * Taken from ppt at WebEx meeting, 12 Feb 2016

19 Median bias : separated FR & OR

20 Comparison spread : separated FR & OR

21 Changes in comparison time series hpa (sonde) 12-month running median δ V5-V6 V7-V6

22 Changes in comparison time series hpa (lidar) 12-month running median δ V5-V6 V7-V6

23 Comparison time series hpa (sonde) 12-month running median Δ V5 V6 V7

24 Comparison time series hpa (lidar) 12-month running median Δ V5 V6 V7

25 Vertical drift profile 60S 90S Linear regression model OR only : Weighted average over lidar & sonde station sample

26 Some month dependence of bias : lidar

27 Some month dependence of bias : sonde

28 Annual cycle in bias hpa (sonde)? 2-week running median Δ V5 V6 V7

29 Annual cycle in bias hpa (lidar)? 2-week running median Δ V5 V6 V7

30 Dependence of bias on profile representation : MIPAS ML2PP 7.03 *MIPAS altitude = corrected altitude (which is inferior to ECMWF corrected altitude, see next part) **When AK are not used for smoothing, then all four systems behave very very different

31 Dependence of bias on profile representation : MIPAS ML2PP 6.0 *MIPAS altitude = corrected altitude (which is inferior to ECMWF corrected altitude, see next part) **When AK are not used for smoothing, then all four systems behave very very different

32 Dependence of bias on profile representation : MIPAS IPF 5.05 / 5.06 *MIPAS altitude = corrected altitude (which is inferior to ECMWF corrected altitude, see next part) **When AK are not used for smoothing, then all four systems behave very very different

33 Conclusions ozone MIPAS V5 & V6 resemble each other more than V7 Full Resolution : similar quality Bias and comparison spreads are similar for all versions Largest differences in UT/LS, where V7 has smaller bias in Tropics & mid N, and larger elsewhere Optimized Resolution : similar quality No clear signs of drift of V7 relative to V6 (or V5) V7 ozone is 1-2% larger than V5/V6 (which increases bias to correlative data) Overall, bias and comparison spread are very similar for V5, V6 and V7 Dependences of bias Perhaps annual cycle, certainly not as clear as for temperature Bias O3_vmr differs from O3_ndens by ~2-4%, no difference altitude/pressure Both dependences seen for all MIPAS versions

34 altitude* * Taken from _MIPAS_V7_altitude MultiTASTE.pdf, circulated around QWG on 25 Nov 2015

35 MIPAS corrected altitude Ozonesonde Difference time series hpa 3-month moving median ± IP68% MIPAS: V5 V6 V7

36 MIPAS corrected altitude Ozonesonde Difference time series hpa 3-month moving median ± IP68% MIPAS: V5 V6 V7

37 MIPAS ECMWF corrected altitude Ozonesonde Difference time series hpa 3-month moving median ± IP68% raw data MIPAS: V7

38 MIPAS ECMWF corrected altitude Ozonesonde Difference time series hpa 3-month moving median ± IP68% raw data MIPAS: V7

39 Conclusions altitude (From _MIPAS_V7_altitude MultiTASTE.pdf, circulated around QWG on 25 Nov 2015) Corrected altitude Bias : generally negative, V7 is less biased than V5/V6 in FR, but not in OR Stability : negative drift for all versions, sometimes more pronounced for V7 than for V5/V6 Very clear annual cycle, for all versions Agreement with correlative data degrades at higher latitudes pt retrieval always beneficial for bias for V5/V6, but not necessarily for V7 (e.g. Arctic, Tropics) ECMWF corrected altitude (V7 only) FR/OR : almost no difference in data quality Bias : generally <100m; and <200m around 10hPa and 30-90S Spread : generally <100m, amplitude annual cycle at most 50m Stability : <10m/year No clear latitude dependence Recommendation ECMWF corrected altitude scale is clearly of much higher quality and should be recommended in the Readme file, but it is not available for all Level-2 profiles

40 backup

41 MIPAS altitude scales in Level-2 data z 4 Dz 4 Dz 4 Dz 4 z 3 Dz 3 Dz 3 Dz 3 z i = z a + S k=a i Dz k z 2 z i = z 0 +S k=1 i Dz k Dz 2 Dz 2 = z 2 -z 1 Dz 1 z 1 Dz 2 Dz 1 Dz 1 z a Engineering altitude scan_information_mds.tangent_altitude_los this is the engineering pointing information from L1b available for V5, V6, V7 z 0 Corrected altitude scan_information_mds.tangent_altitude_los + [pt_retrieval_mds.h_corr 0]*1d-3 anchor point is lowest scan, altitude taken from engineering pointing information altitude of higher scans reconstructed using retrieved p&t and hydrostatic equation (any issues with pt will propagate to Dz i ) available for V5, V6, V7 z 0 = z 0 ECMWF corrected altitude pt_retrieval_mds.ecmwf_corr_alt anchor point altitude taken from co-located ECMWF z-p profile (I am not sure which scan is anchor) altitude of higher scans reconstructed using retrieved p&t and hydrostatic equation (hence, by construction, Dz i =Dz i ) only available for V7

42 MIPAS engineering altitude Ozonesonde Difference time series hpa 12-month moving median ± IP68% MIPAS: V5 V6 V7

43 MIPAS corrected altitude Ozonesonde Difference time series hpa 12-month moving median ± IP68% MIPAS: V5 V6 V7

44 MIPAS ECMWF corrected altitude Ozonesonde Difference time series hpa 12-month moving median ± IP68% raw data MIPAS: V7

45 Validation methodology Preprocessing MIPAS vertical scale pressure retrieved, native scale engineering alt scan_information_mds.tangent_altitude_los (V5, V6, V7) corrected alt scan_information_mds.tangent_altitude_los (V5, V6, V7) + [pt_retrieval_mds.h_corr 0]*1d-3 ECMWF corrected alt pt_retrieval_mds.ecmwf_corr_alt (V7) Sonde vertical scale pressure measured, native scale reconstructed altitude altitude from measured p&t and hydrostatic equation (less accurate above 20hPa) GPS altitude altitude from GPS no differentiation made here between GPS & reconstructed Vertical smoothing & comparison grid ground-based & MIPAS profiles regridded to fixed pressure grid (spacing between pressure levels equivalent to ~1km) no further specific smoothing of ground-based profiles Statistics Δ = X MIPAS X ground [m] Median ( median bias ) and 68% inter-percentile ( IP68 comp ) of Δ-distribution

46 Bias between FR and OR : temperature 60N - 90N 30N - 60N 30N 30S 30S 60S 60S 90S

47 Bias between FR and OR : ozone 60N - 90N 30N - 60N 30N 30S 30S 60S 60S 90S