Evaluation of WINDSAT Data and its impact on Ocean Surface Analysis and Numerical Weather Prediction

1
Evaluation of WINDSAT Data and its
impact on Ocean Surface Analysis and
Numerical Weather Prediction
Robert Atlas Atlantic Oceanographic and
Meteorological Laboratory NOAA Office of
Oceanic and Atmospheric Research J. Ardizzone,
E. Brin, J. C. Jusem, J. Terry, D.
Bungato NASA/GSFC/SAIC
2
(No Transcript)
3

• SOURCES OF REMOTELY SENSED
• OCEAN SURFACE WIND DATA
• active measurements
• original scatterometer on Seasat
  1978
• AMI scatterometer on (ERS-1)/ERS-2
  1991-present
• NSCAT on ADEOS-1
  1996-1997
• SeaWinds on Quikscat
  1999-present
• SeaWinds on ADEOS-2
  2002-2003
• passive measurements
• SSM/I
  1987-present
• TMI on TRMM
  current
• WindSat (Polarimeter)
  launched 12/15/02

4
GEOPHYSICAL VALIDATION OF WINDSAT DATA

• Two-level approach to validating WINDSAT data
• COLOCATION COMPARISONS colocation statistics
  comparing WINDSAT data to other sources of
  surface wind information.
• - ships and buoys, low-level aircraft,
  cloud-tracked winds
• - SSM/I and QuikSCAT
• - analyses from NCEP, ECMWF and GEOS
• LIMITED DATA IMPACT EXPERIMENTS following
  methodology used to evaluate impact of SSM/I,
  NSCAT and QuikSCAT data.
• - using VAM, NCEP and GEOS data assimilation
  systems
• - objective measures of analysis and forecast
  accuracy
• - subjective comparisons of analyses, model
  fluxes and forecast fields
• - case studies examining specific effects of
  WINDSAT data

5
KEY WINDSAT VALIDATION CRITERIA

• Direction and speed differences between WINDSAT
  and other data types.
• Objective quality control of WINDSAT data.
• - Number of reports rejected.
• - Patterns of rejected reports.
• Synoptic plausibility of WINDSAT wind patterns.
• - Coherent, dynamically consistent patterns in
  space and time?
• - Agreement with other information (eg. visible
  imagery)?
• Effect of WINDSAT data on analyses and
  forecasts.
• - On average are analyses and forecasts
  improved?
• - Do significant negative effects occur?
• - What is the magnitude of the WINDSAT impact
  relative to QuikSCAT and SSM/I?

6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
Collocation Methodology

• Maximum Temporal Displacement 90 min
• Maximum Spatial Displacement 50 km
• Conventional Data All conventional ship and buoy
  wind reports were adjusted to an instrument
  height of 10 meters assuming neutral stability. A
  variational analysis method (VAM) was used to
  quality check the data as follows
• 1. gross outliers eliminated using NCEP
  analysis.
• 2. variational analysis of data performed using
  NCEP analysis (as background) with weak
  constraints.
• 3) QC performed relative to the new analysis
  with tighter constraints.
• Windsat Data All relevant quality flags were
  used including the rain flag, warm load anomaly
  flag and the wind speed check (lt 5m/s OR gt 20 m/s
  not used).
• Quikscat Data JPL Science Data Product. All
  relevant quality flags were checked including the
  rain flag and low wind speed check (lt 3m/s not
  used).
• Perfect collocations were performed by choosing
  the alias closest in direction to the collocated
  conventional data.
• Collocations are binned in 5m/s intervals
  according to the Windsat retrieved wind speeds.

12
RMS Speed Differences versus All Conv
13
Speed Bias versus All Conv
14
RMS Direction Differences For Best Aliasversus
All Conv
15
RMS Direction Differences For Selected
Aliasversus All Conv
16
Percent Correct versus All Conv
17
Directional Bias for Best Alias versus All Conv
18
Directional Bias for Selected Alias versus All
Conv
19
Impact of WindSAT using Variational Analysis
In the following slides, the variational analysis
method (VAM Atlas et al.,1996) is used to
evaluate the potential impact of WINDSAT
relative to the NCEP operational analysis, which
includes Quikscat and SSM/I surface wind
observations. The variational analysis generates
a gridded surface wind analysis which minimizes
an objective function F measuring the misfit of
the analysis to the background, the data and
certain a priori constraints. The following
expression is used F ?1SC ?2SB ?3SS
?4SVEL ?5SDIV ?6SVOR ?7SDYN where the
?'s are weights controlling the amount of
influence each constraint. Analyses are
performed at .25, .5, and 1 degree resolution,
with the NCEP operational analysis as the
background for the VAM, and only WINDSAT data
added. The VAM analyses with WINDSAT show
significant modifications to fronts, cyclones,
anticylones and other meteorological phenomena.
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
Example of WINDSAT data in N. Atlantic
26
Impact of WINDSAT on Cyclone in N.Atlantic
27
NCEP analysis 6 h later
28
WINDSAT FORECAST IMPACT EXPERIMENTS Using GEOS 4
Operational DAS
EXPERIMENT 1 CONTROL All Conventional Data
TOVS CTW SSM/I TPW CONTROL WINDSAT
version CONTROL QuikScat (To be
generated) FORECASTS 26 5-day forecasts from
each EXPERIMENT 2 CONTROL All Conventional
Data TOVS CTW SSM/I TPW CONTROL WINDSAT
version CONTROL QuikScat (To be
generated)
29
(No Transcript)
30
(No Transcript)
31
SUMMARY

• A detailed geophysical validation of WINDSAT data
  is being performed.
• This includes collocations with in situ data,
  satellite observations, and model analyses
  synoptic evaluations by highly skilled
  meteorological analysts objective quality
  control and impact experiments using both global
  and regional data assimilation systems.
• All of the measures thus far indicate potential
  for WINDSAT to improve ocean surface wind
  analyses and weather prediction, although there
  are significant limitations relative to
  scatterometry.