stratospheric temperature trends from gps-ro and aqua amsu ... · cosmic-10, estes park, sep. 25,...
TRANSCRIPT
COSMIC-10, Estes Park, Sep. 25, 2017.
Stratospheric temperature trends fromGPS-RO and Aqua AMSU measurements
Hans Gleisner, Johannes K. Nielsen, Stig Syndergard, Kent B. Lauritsen
DMI & ROM SAF
COSMIC-10, Estes Park, Sep. 25, 2017.
Contents
 Trends in the 15-year ROM SAF dry-temperature data record:- construction of the time series, estimation of uncertainties- trends from linear regression- impact of serial correlations- impact of length of time series
 Results from study on stratospheric temperature trends in RO and Aqua AMSU. Done in collaboration with Sergey Khaykin, Beatriz Funatsu, and colleagues, at the LATMOS laboratory in France.
COSMIC-10, Estes Park, Sep. 25, 2017.
15-year RO climate data record ––
ROM SAF reprocessed data set basedon UCAR excess-phase data (atmPhs).
CHAMP, GRACE, COSMIC, and Metop. Metop-A available from Mar 2008, and Metop-B from Feb 2013.
Large step in data numbers from CHAMP-only to COSMIC.
COSMIC-10, Estes Park, Sep. 25, 2017.
Level 2aREFRACTIVITY
DRY TEMP, PRESS, GEOPOT
1D-VAR RETRIEVAL
Level 2bTEMPERATURE
SPECIFIC HUMIDITYGEOPOTENTIAL HEIGHT
Level 1bBENDING ANGLES (L1, L2, LC)
GEOMETRIC OPTICS > 25 kmWAVE OPTICS < 20 km
Level 1aPHASES, AMPLITUDES,SATELLITE POS & VEL
ROM SAF reprocessing chain– processing steps and algorithms –
BINNING & AVERAGINGSAMPLING ERROR CORRECTION
Level 3BENDING ANGLEREFRACTIVITY
DRY TEMP, PRES, GEOPOTTEMPERATURE
SPECIFIC HUMIDITYGEOPOTENTIAL HEIGHT
IONOSPHERIC CORRECTIONSTATISTICAL OPTIMIZATION
ABEL TRANSFORMHYDROSTATIC INTEGRAL
DATA PROCESSING STEPS ALGORITHMS
Canonical Transform (CT2) below 20 km.Transition to Geometric Optics above 25 km.
Using ERA-Interim short-term forecasts as a priori.
Optimal Linear Combination of L1 and L2 BAs.2-parameter fit of observed BA to a best fitting BAprofile found by global search in BAROCLIM.Downward integration of hydrostatic integral from150 km. Upper boundary conditions from d(log(N))/dh.
Weighted averaging into 5o latitude grids.Sampling error estimation based on 2.5ox2.5o
ERA-Interim fields.
COSMIC-10, Estes Park, Sep. 25, 2017.
Monthly means––
The MULTI data set combines all 4 missions.We use the sampling error corrected means.
COSMIC-10, Estes Park, Sep. 25, 2017.
Monthly mean anomalies––
– =
Monthly means for June 2008 Means over all June months 2007-2016 Monthly anomalies for June 2008
We construct monthly mean temperature anomalies by subtracting a mean seasonal cyclefrom the observed monthly mean temperatures.
Temperature anomalies can be up to 5-10 K in the zonal monthly means.
We estimate trends from time series of anomaly data – directly in the zonal grid, or with theanomaly data averaged over latitudes.
COSMIC-10, Estes Park, Sep. 25, 2017.
The visual impression is a cooling in the stratosphere and a warming in the troposphere.QBO oscillations are clearly seen also in global data.
90S – 90N
Monthly mean anomalies averaged over latitudes––
COSMIC-10, Estes Park, Sep. 25, 2017.
5S – 5N
Monthly mean anomalies averaged over latitudes––
COSMIC-10, Estes Park, Sep. 25, 2017.
Zonal monthly uncertainties––
MEASUREMENT ERRORS:
RESIDUAL SAMPLING ERRORS:
SYSTEMATIC ERRORS:
TOTAL UNCERTAINTY:
from Scherllin-Pirscher et al. (2011).
COSMIC-10, Estes Park, Sep. 25, 2017.
Measurement errors––
COSMIC+METOP PERIODCHAMP ONLY PERIOD
The measurement errors for the individual profiles are assumed purely random.Uncertainties of the monthly means scale as the square root of the data numbers.
COSMIC-10, Estes Park, Sep. 25, 2017.
Residual sampling errors––
CHAMP ONLY PERIOD COSMIC+METOP PERIOD
The residual sampling error of the monthly means are assumed purely random.Uncertainties scale with the data numbers, but in a more complicated way.
COSMIC-10, Estes Park, Sep. 25, 2017.
Trends from linear regression––
𝑇 = 𝐴 + 𝐵𝑡 + 𝜖()* + 𝜖+,
: atmospheric variability in the monthly mean anomaly time series𝜖()*𝜖+, : RO observational errors in the monthly mean anomaly time series
The eatm time series is obviously serially correlated.
One way to handle eatm is through multi-variate regression.
If using ordinary linear regression, the standard estimated regression parameters may still be unbiased estimators. But the uncertainties of the estimated parameters(A, B) must be corrected for the reduced number of degrees of freedom. A consequence of serial correlation is that it is more difficult get significant results.
COSMIC-10, Estes Park, Sep. 25, 2017.
Linear trends in monthly mean TDRY
Trends from standard linear regression.Global trends (90N-90S) in 1-km vertical layers to the left.
COSMIC-10, Estes Park, Sep. 25, 2017.
Significances assuming no serial correlations
Sigma values from standard error of the regression, assuming no auto-correlation in time series.p values from two-sided t-test.
COSMIC-10, Estes Park, Sep. 25, 2017.
Significances assuming serial correlations
Here accounting for auto-correlation in time series by decreasing the number of degrees of freedom.p values from two-sided t-test.
COSMIC-10, Estes Park, Sep. 25, 2017.
Significances assuming serial correlations
Here accounting for auto-correlation in time series by decreasing the number of degrees of freedom.p values from two-sided t-test.
COSMIC-10, Estes Park, Sep. 25, 2017.
Global trends in 5 km height layers
15 – 20 km
COSMIC-10, Estes Park, Sep. 25, 2017.
20 – 25 km
Global trends in 5 km height layers
COSMIC-10, Estes Park, Sep. 25, 2017.
25 – 30 km
Global trends in 5 km height layers
COSMIC-10, Estes Park, Sep. 25, 2017.
30 – 35 km
Global trends in 5 km height layers
COSMIC-10, Estes Park, Sep. 25, 2017.
Impact of length of time series
COSMIC-10, Estes Park, Sep. 25, 2017.
Monthly averaged GPS-RO data:- RO dry-temperature profiles from the 15-year ROM SAF CDR;- Vertical weighted averaging of dry-temperature profiles, using fixed AMSU weighting functions;
- Monthly averaging of AMSU-averaged temperatures into 5-degreelatitude bands + sampling error correction;
- Anomaly time series by subtraction of mean seasonal cycle.
Study on GPS-RO and Aqua AMSU
Monthly averaged AMSU data:- Aqua AMSU data from NOAA STAR;- Using the inner 10 fields (swath width 500 km) to avoid limb effects;- Monthly averaging of near-nadir data into 5 degree latitude bands;- Anomaly time series by subtraction of mean seasonal cycle.
Computation of trends- Robust linear regression on the anomaly time series, providingstatistical uncertainties of the trends;
Study results published in Khaykin et al., GRL, 2017.
Purpose of study- To compare trends in GPS-RO and Aqua AMSU;- To detect stratospheric temperature trends over a 15-year period.
COSMIC-10, Estes Park, Sep. 25, 2017.
Global temperature anomalies: RO and AMSU
channel 985S – 85N
COSMIC-10, Estes Park, Sep. 25, 2017.
channel 1085S – 85N
Global temperature anomalies: RO and AMSU
COSMIC-10, Estes Park, Sep. 25, 2017.
channel 1185S – 85N
Global temperature anomalies: RO and AMSU
COSMIC-10, Estes Park, Sep. 25, 2017.
channel 1285S – 85N
Global temperature anomalies: RO and AMSU
• Excellent agreement between RO and Aqua AMSU global temperature anomalies• Somewhat larger differences during CHAMP-only period • Stronger cooling in first part of period present in both AMSU and RO
COSMIC-10, Estes Park, Sep. 25, 2017.
Main findings in Khaykin et al. [GRL, 2017]
Temperature trends over 14 years for GS-RO and AMSU channels 9 to 12. GPS-RO data have been verticallyaveraged to símulate AMSU temperatures.
Temperature trends by month for Aqua AMSU channel 10, and the corresponding RO data.a) high northern latitudes (60-90 N)b,c) low latitudes (0-30 S, 0-30 N)d) high southern latitudes (60-90 S)
Conclusions: - excellent agreement between RO and AMSU trends- global cooling trends statistically significant in channels 11 and 12; - cooling trends in 30 degree latitude bands are not significant;
COSMIC-10, Estes Park, Sep. 25, 2017.
stop