GRAS SAF Workshop on Applications of GPS Radio Occultation Measurements. ECMWF Reading, UK; June 2008

Transcription

1 GRAS SAF Workshop on Applications of GPS Radio Occultation Measurements ECMWF Reading, UK; June 08 Climate Signal Detection with GPS Radio Occultation Measurements (focus trend signal detection, mainly based on GPS/Met and CHAMP radio occultation records and 01-08) G. Kirchengast 1, A.K. Steiner 1, U. Foelsche 2,1, B.C. Lackner 3,1, B. Pirscher 1, and M. Borsche 1 (1) WegCenter & IGAM, Univ. of Graz, Graz, Austria (2) UCAR/COSMIC, Boulder, CO, USA (3) Univ. of Edinburgh, Edinburgh, UK gottfried.kirchengast@uni-graz.at Thanks for funds to 1

2 Outline Radio Occultation (RO) method and properties (just to have an intro included, essentially skipped in the talk) RO data set and climatologies On uncertainties and consistency of climatologies Climate trend signal detection study design Trend, error, and uncertainty calculations Results and discussion Conclusions and outlook 2

3 Radio Occultation Principle GNSS LEO satellite constellations Radio Signals (λ cm) Transmitter GPS GALILEO allow radio occultation observations in an active limb sounding mode exploiting the atmospheric refraction of GNSS signals providing measurements of phase path delay for the retrieval of Receiver key atmospheric/climate parameters Satellite in Low Earth Orbit: GPS/MET, CHAMP, SAC-C, GRACE, COSMIC, MetOp, Oceansat-2 Courtesy: U. Foelsche e.g., refractivity N, geopotential height Z, temperature T, humidity q 3

4 Retrieval of Atmospheric Variables Phase Delay Initialization of bending angle via statistical optimization with background data Orbit Information, Ionospheric Correction Bending Angle Abel Transform ρ(h) = p(h) = h T(h) = M R p(h) T(h) = g(h )ρ(h )dh M R ph ( ) ρ( h) Refractivity Clausius-Mossotti Relation Water Vapor M N(h) R 77,6 Density Temperature Hydrostatic Integral Pressure Temperature Ideal Gas Law Water Vapor (Dry) Temperature Stratosphere Upper Troposphere Lower/Middle Troposphere 4

5 RO Data Characteristics Properties Global coverage ~0 RO events/day atmospheric profiles/day Essentially all weather capability High accuracy and high vertical resolution in the Upper Troposphere and Lower Stratosphere (UTLS) (~8 km) Long-term stability due to intrinsic self-calibration, precise timing with atomic clocks Homogeneity and consistency despite very different orbits, different instruments & raw processing chains Atmospheric temperature profile 1

6 RO Mission GPS/Met Proof of concept on MicroLab-I 199 Intermittent Measurements Prime time data used from Oct 9, Feb 97 1 Temperature profiles near England At about 9-4-:00:00 UTC Pressure, mbar 0 Occultation at 2.6N. 3 E. Radiosonde at 4. N E. Radiosonde at 3. N. 37 E Temperature, K Courtesy: UCAR 6

7 RO Mission CHAMP CHAMP in orbit since July, 00 Continuous RO measurements since Sep 01 Occultation antenna GPS Receiver onboard CHAMP Courtesy: J. Wickert (GFZ) 7

8 CHAMP RO data record 01 T. Schmidt, GFZ 06 GRACE data ~1 180 RO atm.profiles/day 09 CHAMP: First opportunity for multi-year RO climatologies 8

9 RO based Climatologies Event Distribution and Dry Temperature Climatologies Example Months February/October 04 Feb 04 : CHAMP Event Distribution 60 Oct 04 : CHAMP Event Distribution No. of Events: No. of Events: Feb04 : CHAMP Dry Temp [K] Oct04 : CHAMP Dry Temp [K] WegCenter Retrieval CCRv2.3, OPSv.2/v.3 Phase & orbit data from GFZ Potsdam Monthly mean fields Zonal mean fields Binning and Averaging 36 latitude bands latitudinal width Background bending angle data for statistical optimization: ECMWF analyses for CHAMP (T42L60, T42L91) ERA-40 for GPS/Met (T42L60) 9

10 (on uncertainties and consistency of climatologies ) Systematic Difference Summer JJA 02: Systematic Dry Temp Difference ECMWF - CHAMP JJA 03: Systematic Dry Temp Difference ECMWF - CHAMP [K] [K] JJA 04: Systematic Dry Temp Difference ECMWF - CHAMP JJA 0: Systematic Dry Temp Difference ECMWF - CHAMP Wave-like bias structure Antarctica [K] [K] Gobiet et al., GRL 0, Foelsche et al., ClimDyn 07

11 Systematic Difference JJA06 February 06: Major change at ECMWF Horizontal Resolution: T11 T799 Vertical Resolution: L60 L91 Top model level: hpa 1 hpa 3 JJA 0: Systematic Dry Temp Difference ECMWF - CHAMP JJA06: JJA06: Systematic Systematic Dry Temp Dry Temp Difference Difference ECMWF-CHAMP_GRACE [K] CHAMP CHAMP + GRACE only [K] Tropical Tropopause differences disappeared Wave-like bias structure now over the Arctic Foelsche et al., ClimDyn 07

12 3 Systematic Difference JJA 07 as of December 06: Assimilation of RO data at ECMWF JJA 07: Systematic Dry Temp Diff ECMWF-CHAMP CHAMP JJA 03: Systematic Dry Temp Difference ECMWF - CHAMP Latitude 7 [deg] Wave-like bias disappeared [K] [K] Foelsche et al., OPAC 08 12

13 3 Systematic Difference JJA 07 as of December 06: Assimilation of RO data at ECMWF JJA 07: Systematic Dry Temp Diff ECMWF-Formosat-3/COSMIC COSMIC - All 6 satellites JJA 03: Systematic Dry Temp Difference ECMWF - CHAMP Latitude 7 [deg] Wave-like bias disappeared [K] [K] Foelsche et al., OPAC 08 13

14 Consistency of RO Data 3 SON 06: (FM1-Sampl COSMIC-FM1 Err) minus - (FM3-Sampl COSMIC-FM3 Err) Sampling errors subtracted, Mean(ΔT) = K Note different contour spacing! [K] Foelsche et al., TAO 08 14

15 3 Consistency of RO Data SON 06: (FM3-Sampl COSMIC-FM3 Err) minus - (FM-Sampl COSMIC-FM Err) Sampling errors subtracted, Mean(ΔT) = 18 K Note different contour spacing! [K] Foelsche et al., TAO 08

16 Consistency of RO Data 3 JJA 07: (FM2-Sampl COSMIC-FM2 Err) minus - (FM4-Sampl COSMIC-FM4 Err) Sampling errors subtracted, Mean(ΔT) = 14 K Note different contour spacing! [K] Foelsche et al., OPAC 08 16

17 Consistency of RO Data JJA JJA 07 07: COSMIC-FM2 (FM2-Sampl Err) minus - (FM6-Sampl COSMIC-FM6 Err) Sampling errors subtracted, Mean(ΔT) = 03 K Note different contour spacing! [K] Foelsche et al., OPAC 08 17

18 Difference RO Climatologies 3 JJA 07: (COSMIC-Sampl Err) - (CHAMP-GFZ-Sampl Err) [K] JJA 07: (CHAMP-UCAR-Sampl Err) - (CHAMP-GFZ-Sampl Err) Mean(ΔT) = 0.23 K Small systematic difference between GFZ and UCAR phase delay& orbital data JJA 07: (COSMIC-Sampl Err) - (CHAMP-UCAR-Sampl Err) [K] Mean (ΔT) = +6 K [K] Mean(ΔT) = 0.29 K Reason currently studied; very similar for SON 06 and April 02 Foelsche et al., OPAC 08 18

19 Still consistency amongst GFZ, JPL, UCAR, WEGC Absolute CHAMP RO refractivity differences within ~% (~0.2 K) Further improvements at each center will make differences even smaller Ho et al., RO N Trends Intercomp Study, 08 19

20 Still consistency amongst GFZ, JPL, UCAR, WEGC CHAMP RO refrac anomalies fit very closely (trends within ~%/yrs (~K/yrs) In the tropics refrac trends even consistent within %/decade (K/decade) Ho et al., RO N Trends Intercomp Study, 08

21 0 Lapse Rate Tropopause Temperature (Tropics) Consistency Example Derived Parameters Tropical Lapse Rate Tropopause 18 Lapse Rate Tropopause Altitude (Tropics) Temperature [K] SAC-C NCEP SAC-C (a) CHAMP ECMWF GRACE F3C (b) CHAMP ECMWF Year GRACE F3C Year Excellent agreement between data from different RO mission (CHAMP and COSMIC). Anchor points from SAC-C and GRACE. Borsche et al., GRL 07, Foelsche et al., TAO 08 21

22 2 0 Consistency Example Derived Parameters RO-based Synthetic MSU/AMSU Records 3 Jul03 : CHAMP Dry Temp (Zonal Mean) Steiner et al., JGR [K]

23 Signal Detection Setting the Scene: Atmospheric Variability Internal climate variability Solar variability (based on Foelsche et al., JGR, 08) Within data-covered time periods: No relevant volcanic eruption No major El Niño/La Niña events (Steiner et al., JGR, 07; after Karl et al., CCSP Report, 06) No high solar activity; F.7 indices ~60 160, wrt ionospheric variability 23

24 Climate Trend Signal Detection Study Design Questions: Is a significant trend detectable in available RO data? Does the observed trend exceed inter-annual variability? Does the observed trend exceed natural trend variability? GPS/Met and CHAMP dry temperature climatologies: Tdry(p) zonal monthly means for February and October Timeframe: October 199 and (7 yrs with data, 12 yrs in total) February 1997 and (7 yrs with data, 11 yrs in total) Regions: Tropics ( S N) NH Extratropics ( N 0 N) SH Extratropics ( S 0 S) Pressure/Altitude Layers: UTLS within 0 hpa (~9 km) Lower Stratosphere LS (0 hpa) Tropopause TP (0 0 hpa) Currently record Upper Troposphere UT (0 0 hpa) extended to 08 Additional analysis for finer resolution: regions and pressure levels (0, 0, 0, 0, hpa) 24

25 RO measurement errors: Sampling error: e samp (estimated using ECMWF analyses True profiles at the RO locations minus True mean field) Observation error: e obs = 1K/sqrt(N), N Number of RO profiles, 1K for individual profile Systematic error: e sys = 0.2 K 0 hpa, K >0 hpa (Gobiet et al., 07) Total error: Calculations of Measurement Errors, Trends, and Uncertainties e ROtotal ( e ) + ( e ) 2 ( e ) 2 2 sys obs e = + ROtotal samp Linear Trend: δt least-squares fit considering s ROtotal for each individual month Uncertainties: Uncertainty of the trend: s δt Inter-annual variability of inspected period de-trended StDev: s 97-07,, s 9-06 N Natural climate trend variability: s 1 gcm NatVar s estimated based on NatVar = s i multi-century pre-industrial control runs N i= 1 of 3 representative global climate models for IPCC AR4 (ECHAM+HadCM+CCSM3) Signal-to-Noise Ratio SNR: Signal/ Interannual Variability: Signal/ Natural Trend Variability: Significance: Students t-test SNR SNR ( sδt + s ( sδt + s NatVar ) ) = = s s 2 δt 2 δt δt + s δt + s NatVar

26 Trend Differences T phys minus T dry, Feb (estimated based on ECMWF analyses) NHE-0 Tropics SHE-0 Use of T dry instead of T phys induces trend differences < K/decade (with T dry trends slightly conservative) 26

27 RO Temperature Trends and RO Data Errors February NHE 0 Temperature anomalies (left) wrt mean with individual total RO error Max Total RO error: 0.98K Feb97 Tropics SHE 0 Component errors (right): Sampling error: max 0.97 K Feb97 dominant for GPS/Met and in NHE Observation error: ~ K Systematic error: 0.2 K in LS K in UT, TP dominant error source in tropics 27

28 RO Temperature Trends and RO Data Errors October NHE 0 Temperature anomalies (left) wrt mean with individual total RO error Max Total RO error: ~0.8K Oct01 Tropics SHE 0 Component errors (right): Sampling error: max ~0.8 K Oct01 dominant for GPS/Met in NHE/SHE and for CHAMP in SHE Observation error: ~ K Systematic error: 0.2 K in LS K in UT, TP dominant error in tropics, NHE LS 28

29 RO Temperature Trends vs Key Uncertainties δtrend, interann.var, nat.var Feb Oct SHE-0 Tropics NHE-0 SHE-0 Tropics NHE-0 Significant 11-yr trend in Feb in UT & LS over natural and inter-annual variability orange: > 9% significance level yellow: > 90% significance level 29

30 Discussion: Comparison to Trends from GCM runs and Radiosonde Data RO: 199/97 GPS/Met, 01/02-06/07 CHAMP; HadAT2: Raobs climate dataset/ same years; GCM runs: IPCC AR4 A2 and B1 scenario runs of ECHAM+HadCM+CCSM3, 11-yr trends sampled from 01- period Agreement within uncertainty estimates in February between RO, Raobs, GCMs Agreement within uncertainty estimates in October in UTLS

31 Discussion: QBO Quasi-Biennial Oscillation QBO pattern in tropical lower stratosphere with a period of ~28 months seasonal changes in radiative heating due to dissipation of vertically prop. waves downward propagating wind/temperature anomalies constrained to N S and to vertical layer of ~16 22 km T-anomalies of up to ±6 K at ~16 km, at tropopause decrease to < ± K GPS/Met T-anomalies CHAMP/GRACE GPS/Met at 24 km (Randel et al., JGR, 03, Fig. 16) (Schmidt et al., ACP, 0, updated) (Randel et al., JGR, 03, Fig. 18) Estimated contribution from QBO mean over tropical region ( N S) and over vertical layer of ~16 km: < K in standard deviation (LS Trend still significant) 31

32 Conclusions and Outlook Given available monthly GPS/Met and CHAMP RO records within October 199 and / February 1997 and we addressed the following signal detection questions: Is a trend detectable exceeding inter-annual variability in Oct9-06 or Feb97-07? Is a trend detectable exceeding natural trend variability estimated from preind.ctrl.? A significant warming/cooling trend, relative to noise and to natural variability, was found in February in the tropical UT/LS. Comparison with trends from HadAT2 RAOBs climate data showed agreement within uncertainty estimates in February; GCMs agree in envelope as well (though the bulk shows less warming/cooling contrast across the tropical tropopause) Discussion of QBO contribution estimate suggests minor influence to trend significance in LS Ongoing work includes the quantitative estimation of the QBO contribution from RO data, the extension of the trend estimates to Oct07 and Feb08, respectively, the re-check of GPS/Met data processing, the cross-check of use of either GFZ or UCAR phase delay and orbital data, the consistency with COSMIC, and more 32

33 Conclusions and Outlook Given available monthly GPS/Met and CHAMP RO records within October 199 and / February 1997 and we addressed the following signal detection questions: Is a trend detectable exceeding inter-annual variability in Oct9-06 or Feb97-07? Is a trend detectable exceeding natural trend variability estimated from preind.ctrl.? A significant warming/cooling With more than trend, a decade relative of to time noise now and spanned to natural by variability, was found in February RO in data the it tropical is exciting UT/LS. times that have recently started for climate signal detection based on this unique Comparison with trends from HadAT2 RAOBs climate data showed agreement data source in terms of accuracy, consistency, and within uncertainty estimates in February; GCMs agree in envelope as well (though long-term stability; promises of around twenty years the bulk shows less warming/cooling contrast across the tropical tropopause) ago are more and more confirmed by real data Discussion of QBO contribution estimate suggests minor influence to trend significance in LS Thank You! Ongoing work includes the quantitative estimation of the QBO contribution from RO data, the extension of the trend estimates to Oct07 and Feb08, respectively, the re-check of GPS/Met data processing, the cross-check of use of either GFZ or UCAR phase delay and orbital data, the consistency with COSMIC, and more 33