Discussion of Observational Biases of Some Aircraft Types at NCEP Dr. Bradley Ballish NCEP/NCO/PMB 7 September 2006

1
Discussion of Observational Biases of Some
Aircraft Types at NCEPDr. Bradley Ballish
NCEP/NCO/PMB7 September 2006

• Where Americas Climate and Weather Services
  Begin

2
Overview

• Introduction
• Sonde/Aircraft temperature biases
• Monthly average temperature bias time series
  plots

3
Overview (Continued)

• Aircraft bias factors
• Aircraft biases by aircraft types
• Monthly average temperature increment plots
• Collocation results
• Monthly average plots of analysis minus guess
• Summary

4
Introduction

• Observational data biases are serious in part as
  they can cause errors in the analysis
• Biases can be due to errors in the data or our
  use that we would like to correct
• Biases can be due to forecast model bias that is
  best corrected in the model
• It is helpful to know if the bias is due to
  problems in the data or the guess
• Bias correction looks encouraging but has issues

5
Sonde/Aircraft Temperature Biases

• Data monitoring shows that aircraft temperatures
  as a whole are warmer than the NCEP guess
  especially around 250 hPa while radiosondes are
  colder there
• Aircraft and radiosonde data are very important
  for NWP model analyses and forecasts
• One objective of this study was to investigate
  the key reasons for the bias discrepancies and
  its potential impacts on model analyses and
  forecasts

6
Monthly Average Temperature Bias Time Series
Plots

• Biases are global for all data, passing QC from
  300 to 200 hPa for GDAS runs
• Note that on average, sondes are colder than the
  guess, while all aircraft types are warmer than
  guess
• We investigated biases for ACARS, AMDAR, AIREPS
  SONDES
• For more details, see our paper from the AMS
  annual meeting

7
Monthly Average Temperature Biases 300 to 200
hPa 00Z
8
Monthly Average Temperature Biases 300 to 200
hPa 12Z
9
Aircraft Bias Factors

• Many factors affect aircraft biases
• These include aircraft type, influence of past
  data on the guess, airlines, pressure level,
  software, temperature sensors and Phase of Flight
  (POF)
• Specific aircraft type seems to be most important
  such as 767-432 versus 767-322

10
Aircraft Temperature Biases by Aircraft Types 300
hPa and up all Times of Day
11
Aircraft Temperature Biases by Aircraft Types 300
hPa and up all Times of Day
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
Aircraft Temperature Biases 250 /- 25 hPa 00Zon
2.5 by 2.5 degree grid January 2005
17
Radiosonde Temperature Biases 250 /- 25 hPa
00ZJanuary 2005
18
Average Analysis minus Guess Temperature 250 hPa
January 2005
19
Aircraft Temperature Biases 250 /- 25 hPa 00Zon
2.5 by 2.5 degree grid July 2005
20
Radiosonde Temperature Biases 250 /- 25 hPa
00ZJuly 2005
21
Average Analysis minus Guess Temperature 250 hPa
July 2005
22
Discussion of Temperature Bias Impact

• The aircraft bias maps show mostly red dots
  (warm) while the sonde plots show mostly blue
  dots (cold) but not always
• The analysis minus guess plots often show
  patterns explainable by the data increments
• For 00Z January 2005, huge warming over NE Canada
  mixed changes over the CONUS, pattern bears
  comparison with data increments
• For 00Z July 2005, both data types show red dots
  in the Southern US resulting in a large warming
• Wherever both data types show blue dots, there is
  often cooling in the analysis

23
(No Transcript)
24
(No Transcript)
25
Summary

• The warm aircraft bias versus the cold sonde bias
  can be explained in part by RADCOR and large
  variance in aircraft biases for different types
• There is evidence of systematic impact on NCEP
  analyses due to these temperature biases

26
Summary (Continued)

• RADCOR needs fundamental improvement and more
  frequent updates
• Bias correction for aircraft biases needs to be
  performed
• Similar studies are planned for AMDAR data