Comparison of HIRDLS data against sondes and ECMWF analyses

1
Comparison of HIRDLS data against sondes and
ECMWF analyses

• Alison Waterfall,
• J. Reburn, B. Kerridge
• and the HIRDLS team
• Rutherford Appleton Laboratory

2
Data Sources

• UKMO high resolution radiosondes
• for 9 stations (ranging from Shetland Falkland
  Islands).
• Coincidence criteria 300km, 3hrs (4 hrs for St
  Helena)
• Mean differences calculated using the closest
  profile in distance (within the time criteria)
• ECMWF operational analysis data obtained via BADC
• 1.125x1.125 degree grid, 6 hourly data
• 60 model levels up to 0.1 hPa prior to 1st Feb
  2006
• 91 model levels up to 0.01 hPa after 1st Feb 2006
• Data interpolated to location of the HIRDLS
  profiles
• HIRDLS data versions
• V2.04.09 (latest released version)
• V2.04.19

3
Sonde comparison (v2.04.09)

• HIRDLS often within 1K of sondes (biased high)
• Generally larger biases around 100hPa,
  particularly at the tropical tropopause
• Larger differences and variability below 100 hPa

4
Sonde comparison (V2.04.19)

• Better agreement than v2.04.09
• HIRDLS data is now generally within 0.5K of
  sondes, 100-10 hPa
• Still a warm bias around 100 hPa for St Helena
  (only tropical sonde)

5
V2.04.09

• St Helena
• 15.9S, 5.66W
• Generally very good agreement, with HIRDLS
  following many of the sonde features
• Differences around the tropopause

6
GIBRALTAR (36.2N, 5.3W) V2.04.09
7
HIRDLS/ECMWF comparison

• ECMWF operational analysis data was interpolated
  to the location of the HIRDLS points
• Differences (HIRDLS ECMWF) were averaged, and
  binned in latitude for the whole data set, or
  shorter periods of time
• ECMWF has been upgraded several times over 3
  years of HIRDLS data. (Differences are
  particularly apparent in O3)
• Only removing HIRDLS points with negative
  precision (50 apriori influence)

8
Global Mean Differences (HIRDLS ECMWF), v2.04.09

• 100 - 5 hPa HIRDLS is 1-2 K higher than ECMWF
• below 100 hPa HIRDLS is on average cooler than
  ECMWF.
• Above 1 hPa HIRDLS data becomes much cooler than
  ECMWF, with unrealistically low values observed
  above 0.1 hPa.

9
Global Mean Difference (HIRDLS ECMWF), v2.04.19
Differences are reduced with version 2.04.19
10
Daily mean differences V2.04.09

• Differences exhibit similar features over time
• Agreement generally improves over time
• HIRDLS is slightly warmer than ECMWF in most
  regions, but much lower above 1 hPa.

11
Latitudinal variation V2.04.09
August 2005

• HIRDLS ECMWF temperature differences in 5
  degree latitude bins for August 2005,2006,2007
• See an improvement with time

August 2006
August 2007
12
HIRDLS ECMWF, August 2006
V2.04.09
V2.04.19

• v2.04.19 is closer to ECMWF than v2.04.09
• Many of the features of the difference are
  consistent between the two versions

13
Ozone Mean Differences (HIRDLS ECMWF), v2.04.09
60N-84N
30N-60N
14
HIRDLS ECMWF Ozone, v2.04.19
15
V2.04.09Ozone Differences
16
Conclusions

• HIRDLS temperatures are generally within 1K of
  sondes (high bias) in v2.04.09, with larger
  differences around the tropopause
• HIRDLS v2.04.09 temperatures are generally higher
  than ECMWF temperatures by 1-2 K between 100-5
  hPa. At the tropical tropopause HIRDLS is
  several K higher than ECMWF.
• More recent versions have improved agreement.
• HIRDLS and ECMWF temperature agreement improves
  with time (ECMWF improvements?)
• Larger changes in the ECMWF ozone data are
  observed, making it harder to draw conclusions on
  the quality of the ozone data.