Differences in trends and anomalies of upper-air observations from GPS RO, AMSU, and radiosondes Florian Ladstädter Hallgeir Wilhelmsen Barbara Angerer Andrea K. Steiner Wegener Center and IGAM/Institute of Physics, University of Graz, Austria florian.ladstaedter@uni-graz.at
Overview 1. Comparisons of GPS RO temperature time series in its core region (8 km to ~25 km) to layer-averaged MSU brightness temperatures (lower-stratospheric TLS channel 4) 2. Vertical trends from the GPS RO temperature time series, compared to radiosondes and models Sep 26, 2017 F. Ladstädter Page 2 of 33
Data Use WEGC RO OPSv5.6 temperatures, all available missions Sep 2001 to Feb 2017 For radiosondes, use Vaisala RS80/90/92/41 from the ERA-I archive, and also calibrated, gridded monthly data: RAOBCORE, RICH For AMSU, use calibrated, gridded monthly TLS data from RSS, STAR, UAH Sep 26, 2017 F. Ladstädter Page 3 of 33
Methods Climatologies, Sampling Error, MSUequivalent temperatures Use radiative transfer model (RTTOV) on single profiles to retrieve MSU-equivalent brightness temperatures Do this for RO and RS, only if profile sufficiently covers range for channel 4 weighting function (~ 8 30 km) Calculate gridded RO climatologies for multisatellites, and correct for sampling error Do the same for RS, and also correct for sampling error Do the same for ECMWF analysis and ERA- Interim reanalysis MSU weighting function (from RSS) Sep 26, 2017 F. Ladstädter Page 4 of 33
GPS RO sampling Jan 2002 Jan 1 2010 Good spatial coverage of RO. But need to take care when doing trend analysis, because of CHAMP period. Sep 26, 2017 F. Ladstädter Page 5 of 33
Radiosonde sampling only Vaisala Sparse spatial sampling; constant number of measurements during this time period, but changes in instrumentation for Vaisala sondes. Sep 26, 2017 F. Ladstädter Page 6 of 33
MSU TLS Anomalies RSS STAR UAH RO RS ECMWF ERAI All datasets look very similar, but looking at the differences... Sep 26, 2017 F. Ladstädter Page 7 of 33
MSU TLS Anomaly Differences to GPS RO 30r1 38r2 RSS RO STAR RO UAH RO RS RO ECMWF RO ERAI RO Jumps coincide with changes in ECMWF model: 30r1 and 38r2, both correspond to increases in vertical resolution; also significant change in sampling in mid-2006. Sep 26, 2017 F. Ladstädter Page 8 of 33
MSU TLS Anomaly Differences to GPS RO RSS RO WEGC CHAMP STAR RO UAH RO RS RO ECMWF RO ERAI RO CHAMP-only looks very similar, no sampling issue! Sep 26, 2017 F. Ladstädter Page 9 of 33
MSU TLS Anomaly Differences to GPS RO RSS RO DMI CHAMP (*) STAR RO UAH RO RS RO ECMWF RO ERAI RO DMI is not using ECMWF for initialization (*) using an unofficial beta-version of the next DMI ROMSAF RO reprocessing Sep 26, 2017 F. Ladstädter Page 10 of 33
MSU TLS Anomaly Differences to GPS RO RSS RO WEGC FM1 STAR RO UAH RO RS RO ECMWF RO ERAI RO Investigate second jump in 2013: look at F3C FM1 only... Sep 26, 2017 F. Ladstädter Page 11 of 33
MSU TLS Anomaly Differences to GPS RO RSS RO WEGC METOP-2 STAR RO UAH RO RS RO ECMWF RO ERAI RO and METOP-2 only. METOP-2 is less influenced by the high-altitude initialization than F3C/COSMIC. Sep 26, 2017 F. Ladstädter Page 12 of 33
Trends in MSU TLS Anomalies TROPICS 30 N 70 N GLOBAL ERA-INTERIM ECMWF RS RO UAH STAR RSS Very consistent trends for the MSU datasets, slightly negative globally; small positive trends for RO, RS, ERA-I, reasons for differences partly known as shown in previous slides Sep 26, 2017 F. Ladstädter Page 13 of 33
Vertical Trends Method Use multiple linear regression Time period September 2001 to February 2017 Either linear, or using new, vertically resolved atmospheric variability indices to describe QBO and ENSO in the regression Solar flux F10.7 cm also included Sep 26, 2017 F. Ladstädter Page 14 of 33
Vertical Trends Tropics linear 32 km @17 km 0 km 09/2001 02/2017 Positive trends in lowermost stratosphere; effects of background visible below 8 km. Sep 26, 2017 F. Ladstädter Page 15 of 33
Vertical Trends Tropics linear robust? 32 km @17 km 0 km 09/2001 02/2017 01/2005 02/2017 Sep 26, 2017 F. Ladstädter Page 16 of 33
Vertical Trends Tropics linear robust? 32 km @17 km 0 km 09/2001 02/2017 01/2005 02/2017 01/2007 02/2017 Sep 26, 2017 F. Ladstädter Page 17 of 33
Vertical Trends Tropics linear robust? 32 km @17 km 0 km 09/2001 02/2017 01/2004 12/2014 01/2005 02/2017 01/2007 02/2017 Sep 26, 2017 F. Ladstädter Page 18 of 33
Vertical Trends Tropics linear robust? 32 km @17 km 0 km 09/2001 02/2017 01/2005 02/2017 01/2007 02/2017 01/2004 12/2014 09/2001 12/2011 Sep 26, 2017 F. Ladstädter Page 19 of 33
Vertical Trends Tropics linear robust? 32 km @17 km 0 km 09/2001 02/2017 01/2005 02/2017 01/2007 02/2017 01/2004 12/2014 09/2001 12/2011 08/2006 10/2016 Sep 26, 2017 F. Ladstädter Page 20 of 33
Vertical Trends Variability Indices (M2) Need to reduce natural variability in the regression Use PCA on RO tropical temperature anomalies on each altitude level to describe the most important variability Use the PCs as variability indices in the regression, similar to often-used ENSO SST and QBO wind indices See Hallgeir's talk after lunch, and Wilhelmsen et al., AMTD, doi: 10.5194/amt-2017-226 Tropical temperature anomalies PC1 and PC2 for each altitude level Sep 26, 2017 F. Ladstädter Page 21 of 33
Vertical Trends Tropics M2 32 km @17 km 0 km 09/2001 02/2017 Sep 26, 2017 F. Ladstädter Page 22 of 33
Vertical Trends Tropics M2 32 km @17 km 0 km 09/2001 02/2017 01/2005 02/2017 Sep 26, 2017 F. Ladstädter Page 23 of 33
Vertical Trends Tropics M2 32 km @17 km 0 km 09/2001 02/2017 01/2005 02/2017 01/2007 02/2017 Sep 26, 2017 F. Ladstädter Page 24 of 33
Vertical Trends Tropics M2 32 km @17 km 0 km 09/2001 02/2017 01/2004 12/2014 01/2005 02/2017 01/2007 02/2017 Sep 26, 2017 F. Ladstädter Page 25 of 33
Vertical Trends Tropics M2 32 km @17 km 0 km 09/2001 02/2017 01/2005 02/2017 01/2007 02/2017 01/2004 12/2014 09/2001 12/2011 Sep 26, 2017 F. Ladstädter Page 26 of 33
Vertical Trends Tropics M2 32 km @17 km 0 km 09/2001 02/2017 01/2005 02/2017 01/2007 02/2017 01/2004 12/2014 09/2001 12/2011 08/2006 10/2016 Sep 26, 2017 F. Ladstädter Page 27 of 33
Vertical Trends Linear vs. M2 regression Linear trend Vertically resolved regression indices (M2) Sep 26, 2017 F. Ladstädter Page 28 of 33
Vertical Trends Linear vs. M2 regression Linear trend Vertically resolved regression indices (M2) Trend ± conf. int. Variance Trend ± conf. int. Variance Total PC1 PC2 Residual 01/2007 02/2017 Much reduced variance for M2 regression. Sep 26, 2017 F. Ladstädter Page 29 of 33
Vertical Trends Compare RO, RS, ERA-I TROPICS 30 N 70 N EUROPE ERA-I RO RS RS shows positive trend in lower troposphere in tropics, in UTLS consistent with RO; for mid-latitudes RO and RS highly consistent. Sep 26, 2017 F. Ladstädter Page 30 of 33
Vertical Trends Compare RO, RS, ERA-I RO ERA-I ECMWF RS RAOBCORE RICH RO: warming above tropopause, trend structure remarkably consistent with all RS datasets except mid-troposphere; ECMWF too large warming, due to model changes; ERA-I and ECMWF wave-like structures, ERA-I missing some warming in upper troposphere, but shows more warming in the TRP Sep 26, 2017 F. Ladstädter Page 31 of 33
Conclusions RO and RS consistent in upper-troposphere/lower stratosphere Differences of AMSU and RO for the TLS channel partly due to high-altitude initialization with ECMWF, and not due to jumps in number of occultations (sampling), but further issues remain High-altitude initialization matters! RO time series is short for trend analysis, but using vertically resolved variability indices, robustness can be increased a lot. Do not miss Hallgeir's talk after lunch! Sep 26, 2017 F. Ladstädter Page 32 of 33
Sampling Error MSU equivalent TLS RS RO Transition from CHAMP to COSMIC period is not a big issue in the tropics. Sep 26, 2017 F. Ladstädter Page 33 of 33