stratospheric temperature variations and trends: recent radiosonde results dian seidel, melissa free...
Post on 21-Dec-2015
215 views
TRANSCRIPT
Stratospheric Temperature Variations and Trends:
Recent Radiosonde Results
Dian Seidel, Melissa FreeNOAA Air Resources Laboratory
Silver Spring, MD
SPARC Temperature Trends Workshop
University of Reading, 3-4 March 2005
Topics
• New Radiosonde Datasets
• Comparison of Stratospheric Trends
• Evaluation of Sources of Differences
• Linear Trends and Other Models
• Upper-Air Measurement Requirements for Climate (time permitting)
Topics
• New Radiosonde Datasets
• Comparison of Stratospheric Trends
• Evaluation of Sources of Differences
• Linear Trends and Other Models
• Upper-Air Measurement Requirements for Climate
NOAA Datasets
• CARDS became IGRA (Imke Durre, Russ Vose, NCDC)– Integrated Global Radiosonde Archive of quality controlled
(not homogeneity-adjusted) soundings– Metadata update is ongoing
• Angell (2003) reduced network from 63 to 54 stations
• Lanzante-Klein-Seidel (2003a,b) adjusted for inhomogeneities– Adjustments based on station history metadata, statistical
change-point identification, and evaluation of real abrupt changes
– 87 stations, 1948-1997
NOAA Datasets (cont.)
• LKS updated for RATPAC– Radiosonde Atmospheric Temperature Products for
Assessing Climate– Joint ARL/GFDL/NCDC effort– 16 levels, surface -10 hPa– Climate monitoring data product– Two basic datasets:
• Large-scale anomaly time series based on LKS adjustments through 1979 and first-difference method and metadata
• Station data with no adjustments post-1979
– General distribution after peer-review
Met Office Datasets
• HadRT (Parker et al. 1997) – Based on monthly-mean CLIMAT reports– 444 stations used to create gridded product– Referenced to MSU in stratosphere
• HadAT (Peter Thorne et al.)– >600 stations using to create several gridded
products– Homogenized using LKS results and neighborhood
checks, with focus on troposphere– Includes analysis of uncertainty– 9 levels, 850-30 hPa
• GCOS Upper-Air Network Monitoring Center (Mark McCarthy)
Rest of the World
• All-Russian Research Institute of Hydrometeorological Information (Alex Sterin)– Ongoing analysis of global and regional data
• Other efforts ???
Topics
• New Radiosonde Datasets
• Comparison of Stratospheric Trends
• Evaluation of Sources of Differences
• Linear Trends and Other Models
• Upper-Air Measurement Requirements for Climate
Signals of large-scale, short-lived stratospheric variations in different datasets
are in good agreementQ
BO
0.00
0.01
0.02
0.03100-50 hPa MSU 4
Ang
ell-6
3
Ang
ell-5
4
Had
RT
RIH
MI
LKS
UA
H
RS
S
Had
RT
LKS
Pin
atu
bo
0.00.51.01.52.0
Ang
ell-6
3
Ang
ell-5
4
Had
RT
RIH
MI
LKS
UA
H
RS
S
Had
RT
LKS
EN
SO
0.000.040.080.120.16
19
76
-77
0.00.40.81.21.6
100-50 hPa
850-300 hPa
850-300 hPa
MSU 4
MSU 2
Trend sensitivity to LKS adjustments (solid)
HadRT and LKS agreement deteriorates with adjustments
Trends from Sondes, MSU & Reanalyses
• Large confidence intervals, but these do not address all sources of trend uncertainty
• More disparity among datasets than for shorter-term signals
• Reanalyses are outliers and are not reliable for trends
• Stratospheric cooling is stronger than tropospheric warming, but not more consistently estimated
• Sondes show more cooling than MSU
• Conventional wisdom is that sonde trends are too strong, but this is not firmly established.
Global Temperature Trends
1979-2003 Trend (K/decade)
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0
Surface
MSU LT
Fu-LT
850-300 hPa
500 hPa
300-100 hPa
MSU MT (2)
200 hPa
0100-50 hPa
MSU LS (4)
30 hPa
NCEP reanalysisERA40 reanalysisHadCRU surfaceNASA GISS surfaceNOAA GHCN surfaceVG MSURSS MSUUAH MSUHadAT raobLKS-RATPAC raob
RATPAC 1979-2003 Trends
RATPAC 1960-2003 Trends
Zonal Stratospheric Trends
RATPAC Zonal T Trends (K/decade)
Latitude
-80 -60 -40 -20 0 20 40 60 80-2.0
-1.5
-1.0
-0.5
0.0
0.5
1960-150 1960-100 1960-70 1960-50 1960-30 1979-150 1979-100 1979-70 1979-50 1979-30
Zonal Stratospheric Trends
RATPAC Zonal T Trends (K/decade)
Latitude
-80 -60 -40 -20 0 20 40 60 80-2.0
-1.5
-1.0
-0.5
0.0
0.5
1960-150 1960-100 1960-70 1960-50 1960-30
Zonal Stratospheric Trends
RATPAC Zonal T Trends (K/decade)
Latitude
-80 -60 -40 -20 0 20 40 60 80-2.0
-1.5
-1.0
-0.5
0.0
0.5
1979-150 1979-100 1979-70 1979-50 1979-30
Topics
• New Radiosonde Datasets
• Comparison of Stratospheric Trends
• Evaluation of Sources of Differences
• Linear Trends and Other Models
• Upper-Air Measurement Requirements for Climate
Comparing Effects of 3 Factors to Evaluate HadRT / LKS Trend Differences
• Data Adjustments – compare trends from adjusted and unadjusted data (see above)
• Spatial and temporal sampling – compare trends from subsampled datasets and complete datasets, using MSU and reanalysis as complete datasets
• Source radiosonde data – compare trends from 71 common stations
Spatial sampling differences71 stations in common
HadRT and LKS trends agree better at 71 common stations than for full networks
(time sampling at month-to-month level only)
(ADJ)
Subsampling MSU makes little difference to MSU/sonde discrepancy
3 factors have comparable effects on global trend differences, with adjustments dominating in the stratosphere
Roles of Factors Vary Regionally
Topics
• New Radiosonde Datasets
• Comparison of Stratospheric Trends
• Evaluation of Sources of Differences
• Linear Trends and Other Models
• Upper-Air Measurement Requirements for Climate
Alternative models may provide better fits (Seidel and Lanzante, 2004)
• Models evaluated using Bayesian Information Criterion
• Net stratospheric temperature change depends on model selected– Linear -1.13 K– Sloped steps -0.88 K– Censored, flat steps -
0.83– Censored, linear -0.99
• Different models suggest different physical interpretations
MSU4
1980 1985 1990 1995 2000
-1.0
1.0
1.0
1.0
1.0
-1.0
-1.0
1.0
-1.0
1.0
3.0
0.0
0.0
0.0
0.0
0.0
0.0
2.0
0.0
2.0
Observations
Censored: Linear+AR(1)
Censored: Flat Steps+AR(1)
model
model
residuals
residuals
Sloped Steps+AR(2)
model
residuals
Topics
• New Radiosonde Datasets
• Comparison of Stratospheric Trends
• Evaluation of Sources of Differences
• Linear Trends and Other Models
• Upper-Air Measurement Requirements for Climate
Issues Affecting Climate Statistics and Trends
• Measurement Precision
• Sampling Frequency– Number of observations/day– Number of observations/month
• Long-term measurement stability
• Network size
• Locations of network stations
Tests with NCEP Reanalysis Data
• Start with 6-hourly data, at 6 pressure levels, at 15 locations, for 1948-2003
• Subsample, or introduce artificial changes
• Compare with unaltered data– Monthly means and variances– Multi-decadal trends
Precision Effects on Monthly MeansMeans are within 0.05K for precision <0.5 K
Estimated Minus Actual Monthly Mean Temperature (K)
Precision of Temperature Measurement (K)
0.01 0.10 0.50 1.00-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
Ratio of Estimated to Actual Monthly Standard Deviation
Precision of Temperature Measurement (K)
0.01 0.10 0.50 1.000.9
1.0
1.1
1.2
1.3
Effects of Reduced Measurement Precision onMonthly Means and Standard Deviations of Temperature
Full Temporal Sampling (4/day, every day)n=60480 (15 locations, 6 pressure levels, 672 months)
MAX
MIN
25%
75%50%
Reduced Diurnal Sampling:
• Effect varies with size and shape of diurnal cycle
• Going from 4 to 2 obs/day makes significant change in monthly means in only 2.3% of cases (mainly near surface and in stratosphere)
• For only 1/day, means change in 13-17% of cases
Conclusion: 2 obs/day may be enough
Number of Observations Per MonthEstimated Minus Actual Monthly Mean Temperature (K)
Sampling Frequency
1/2d 1/3d 1/7d-12
-8
-4
0
4
8
Ratio of Estimated to Actual Monthly Standard Deviation
Sampling Frequency
1/2d 1/3d 1/7d0.0
0.5
1.0
1.5
2.0
2.5
Effects of Subsampling the Month onMonthly Means and Standard Deviations of Temperature
Comparisons based on 2/day, every dayn=60840 (15 locations, 6 pressure levels, 672 months)
Observations every other day give • monthly means accurate to within 2 K or better • trends that are not statistically different in 90% of cases
Effects of Data Stability on Trends - 1 event
Reliable trend estimates require measurement stability within 0.5 K over 20-50 years.
Long-Term Data Stability
Effects of Random Interventions on Trends:Percent of Statistically Significantly Different Trends
Twice-Daily Sampling, Every Day, Full Precision
Segment Length (yrs)
20 25 30 35 40 45 50
Err
or R
ate
(%)
0
10
20
30
40
50
0.10K0.25K0.50K0.75K1.00K1.50K2.00K
Maximum Intervention
Effects of Multiple Random Interventions on Trends
Percent Statistically Significantly Different 25-Year Trends
Number of Interventions
0 1 2 3 4 5
Err
or R
ate
(%)
0
10
20
30
400.1K0.5K1.0K2.0K
Maximum Intervention
Percent Statistically Significantly Different 50-Year Trends
Number of Interventions
0 1 2 3 4 5
Err
or R
ate
(%)
0
10
20
30
40 0.1K
0.5K
1.0K
2.0K
Maximum Intervention
Effects of multiple changes = More errors
But the first event causes most of the error
Spatial sampling errors in trends from hypothetical networks from reanalysis– decrease with increasing size
Error = trend in subsampled minus trend in complete network
50
200
500 850
Spatial sampling errors in trends from actual radiosonde networksusing reanalysis- no decrease with increasing size
50
200
500 850
Take-Away Points• New Radiosonde Datasets – several,
unpublished• Comparison of Stratospheric Trends –
increasing cooling with height, with large uncertainties
• Evaluation of Sources of Differences – everything matters, but especially homogeneity adjustments in the stratosphere
• Linear Trends and Other Models – reasonable to consider nonmonotonic changes
• Upper-Air Measurement Requirements for Climate – input from this group would be very welcome!
Take-Away Points• New Radiosonde Datasets – several,
unpublished• Comparison of Stratospheric Trends –
increasing cooling with height, with large uncertainties
• Evaluation of Sources of Differences – everything matters, but especially homogeneity adjustments in the stratosphere
• Linear Trends and Other Models – reasonable to consider nonmonotonic changes
• Upper-Air Measurement Requirements for Climate – input from this group would be very welcome!
Take-Away Points• New Radiosonde Datasets – several,
unpublished• Comparison of Stratospheric Trends –
increasing cooling with height, with large uncertainties
• Evaluation of Sources of Differences – everything matters, but especially homogeneity adjustments in the stratosphere
• Linear Trends and Other Models – reasonable to consider nonmonotonic changes
• Upper-Air Measurement Requirements for Climate – input from this group would be very welcome!
Take-Away Points• New Radiosonde Datasets – several,
unpublished• Comparison of Stratospheric Trends –
increasing cooling with height, with large uncertainties
• Evaluation of Sources of Differences – everything matters, but especially homogeneity adjustments in the stratosphere
• Linear Trends and Other Models – reasonable to consider nonmonotonic changes
• Upper-Air Measurement Requirements for Climate – input from this group would be very welcome!
Take-Away Points• New Radiosonde Datasets – several,
unpublished• Comparison of Stratospheric Trends –
increasing cooling with height, with large uncertainties
• Evaluation of Sources of Differences – everything matters, but especially homogeneity adjustments in the stratosphere
• Linear Trends and Other Models – reasonable to consider nonmonotonic changes
• Upper-Air Measurement Requirements for Climate – input from this group would be very welcome!