Analysis%20and%20Impact

1
Analysis and Impact of Moisture from GPS
Slant-path Water Vapor Observations using
3DVAR with Flow-dependent Background Error EMC
seminar04/17/2007

• Haixia Liu1,2 and Ming Xue2
• 1 NCEP/EMC
• 2 SoM and CAPS, University of Oklahoma

2
Introduction

• Accurate characterization of 3D water vapor is
  important.
• for the forecast of CI and subsequent storm
  evolution
• for QPF
• Water vapor is under-sampled for convection
  processes.
• GPS can potentially provide water vapor
  measurements with possible high resolution under
  virtually all weather conditions.
• One form of GPS measurements is the slant-path
  water vapor (SWV).
• This study develops a 3DVAR system for analyzing
  SWV data.
• Examine the impact of SWV data on CI and
  precipitation (preliminary results)

3
Outline

• 3DVAR system with flow-dependent background error
• Moisture retrieval from SWV and surface data with
  3DVAR
• Numerical simulation of 12 June, 2002 IHOP case
• Impact of SWV data on predicting CI and
  precipitation within OSSE framework

4
GPS Observation System
Control segment
Ground-based receiver
Space segment
5
.
Ground-based GPS Network
The GPS-Met network consists of 386 sites.
http//gpsmet.fsl.noaa.gov/jsp/index.jsp

6
Ground-based GPS Data
Ionospheric delay Estimate from dual frequency
observations
Total atmospheric delay
Hydrostatic delay Estimate from surface pressure
measurements
Neutral delay
Wet Delay (SWD)
pw precipitable water in a column in vertical
direction ZWD zenith wet delay
7
3DVAR System with EF
Here,
so as to exclude the inverse of B in the
definition of J and to use explicit filter to
replace B.
8
3DVAR System with RF
9
Background Error Covariance B

• B is crucial to the successful analysis because
• variances determine the relative weights for the
  background and observations
• spatial covariance determine the spatial
  spreading or smoothing of observational
  information
• for multivariate analysis, cross-covariances
  reflect balance properties among fields.

10
Flow-dependent Anisotropic B
The flow-dependent B is formulated directly in
terms of the flow given a physically
meaningful correlation function form.
An important difference is the analysis
background field is used as the f in his case. In
our case, f, is defined as the error field.
11
Question How to obtain f which
represents the background error
pattern? Solution f isotropic
analysis increment
12
3D Moisture Retrieval from Simulated SWV and
Surface Data with 3DVAR using Anisotropic Spatial
Filters
13
Observation System Simulation
Hypothetical GPS Network
truth
receiver reso 150 km
14
Analysis background
B
A
15

• Explicit filter
• Anisotropic B based on true background error

Solid truth Dashed ana
truth-background
analysis increment
A
B
16
Explicit Filter

• ISO
• Isotropic B

analysis v.s. truth (solid) analysis increment

• UB
• (Updated B)
• Anisotropic B
• But the f field is the ISO analysis increment
• This is a two-step iterative procedure

17
Sensitivity to Lr Lf
RMSE (g/kg) w.r.t. Lr
18
Summary 1

• Our 3DVAR system incorporating background error
  through an isotropic Gaussian filter properly
  recovers 3D meso-scale moisture structure in a
  dryline case.
• The use of flow-dependent background error
  covariances realized through an anisotropic
  spatial filter improves the analysis.
• The two-step iterative procedure to estimate B
  proposed (covariance-updated) improves upon the
  result of isotropic analysis.
• Compared to EF, the biggest advantage of RF is
  the computational efficiency.
• The quality of analyses using RF is in general
  comparable to or better than those obtained with
  EF in terms of CC.
• Isotropic analysis is more sensitive to geometric
  de-correlation scale, Lr , than anisotropic
  analysis.

19
Numerical Prediction of Convective Initiation of
June 12, 2002 IHOP Case

• To obtain a realistic high-resolution simulation
• To understand the convective initiation processes
  (not reported here)
• To perform OSSEs on this case with 3DVAR system
  we developed in order to investigate the impact
  of SWV data on CI and subsequent storm evolution.

20
Case Overview
12Z, 12 June 03Z, 13 June, 2002
21
2045 UTC, 12 June 2002
22
Model Configurations
1800 UTC
1200 UTC
0000 UTC
0300 UTC
ADAS
CI 2100UTC
ADAS

• ARPS model with full physics, including ice
  microphysics soil model PBL and TKE-SGS
  turbulence

dx 3km
23
timing and location of CIs cells split, regroup
and intensify
3
2
1c
1b
1a
24
_at_2130, 12 June 2002
radar observation model forecast
25
_at_0000, 13 June, 2002
radar observation model forecast
26
_at_0100, 13 June, 2002
radar observation model forecast
Initiation of new cells by colliding boundaries 7
hours into the forecast
27
Summary 2

• The 12 June, 2002 case involved initiations of
  many convective cells. It is a complicated case
  with many mesoscale features playing roles in CI.
• ARPS model with 3km horizontal resolution
  produces rather realistic simulation of this
  case. It successfully reproduces most of the
  observed convective cells with remarkable
  accuracy in initiation timing and location
  compared to the simulation from Bastin et al.
  (2005).
• Predicts well the general evolution of the
  precipitation within the first 7-hour forecast
  period, including cell splitting, merging,
  regrouping and the triggering of secondary
  convective cells by outflow boundaries colliding
  with each other.
• The main deficiency of the prediction is the lack
  of organization of cells into a squall line and
  its inaccurate propagation during the last 2-hour
  of 9-hour forecast.

28
Impact on CI and Precipitation via OSSEs

• Experiment design
• SWV and surface data are generated from the
  numerical simulation.
• Analyze data using isotropic and flow-dependent
  B respectively.
• Analysis results are used as ICs to initialize
  ARPS model.
• ETA analysis, serving as the analysis first
  guess, is used as IC to make a forecast.
• Forecast results are verified against the
  truth to investigate the impact of SWV data and
  the use of flow-dependent B on the forecast.

29
Moisture field _at_ 1800
guess
truth
ISO
UB
30
(No Transcript)
31
EW vertical cross-section along AB
guess
truth
ISO
UB
32
Impact of analysis on Precipitation
ETS (equitable threat score) for composite
reflectivity
33
2045 UTC, 12 June 2002
34
List of forecast experiments
Forecast experiments Experiment description Experiment description CI timing (UTC) CI timing (UTC)
Forecast experiments Experiment description Experiment description CI2 CI4
Without GPS Moisture initial condition from ETA analysis Moisture initial condition from ETA analysis 2040 2000
withGPS_ANISO Moisture initial condition from 3DVAR analysis using GPS SWV and surface moisture data Anisotropic B based on true background error 2040 2030
withGPS_ISO Moisture initial condition from 3DVAR analysis using GPS SWV and surface moisture data Isotropic B 2040 2150
withGPS_UB Moisture initial condition from 3DVAR analysis using GPS SWV and surface moisture data Anisotropic B based on isotropic analysis 2040 2040
Truth Truth Truth 2040 2130
35
Moisture Analysis
guess
truth
ISO
UB
36
Analysis within Zoomed-in Region
guess
Truth
15g/kg
14g/kg
15g/kg
UB
ISO
c
c
d
15g/kg
15g/kg
14g/kg
14g/kg
37
2 hr Forecast
fcst_noSWV
Truth
15.5g/kg
13.5g/kg
CI _at_ 2000
CI _at_ 2130
fcst_UB
fcst_ISO
c
d
CI _at_ 2040
CI _at_ 2150
14g/kg
38
2040 UTC
Truth
With GPS
Without GPS
39
2200 UTC
Truth
With GPS
Without GPS
40
_at_0100, 13 June, 2002
radar observation model forecast
Initiation of new cells by colliding boundaries
5-7 hours into the forecast
41
fcst_UB
S-Pol obs
fcst_noSWV
42
_at_ 2300, 12 June, 2002
fcst_UB
S-Pol obs
fcst_noSWV
43
_at_ 0000, 13 June, 2002
fcst_UB
S-Pol obs
fcst_noSWV
44
_at_ 0030, 13 June, 2002
fcst_UB
S-Pol obs
fcst_noSWV
45
_at_ 0100, 13 June, 2002
fcst_UB
S-Pol obs
fcst_noSWV
46
Preliminary Conclusions

• The timing of the CI in the region without any
  significant low-level mesoscale forcing are very
  sensitive to the fine-scale structures in the
  moisture IC.
• GPS SWV data combined with surface moisture data
  can be effectively analyzed using our 3DVAR
  system to retrieve 3D moisture distribution, thus
  improves the prediction of the CI and subsequent
  evolution of convection.
• The anisotropic background error covariance helps
  better characterize detailed moisture structures
  dramatically. Thus the predicted reflectivity
  threat scores are higher.
• No significant positive impact on the exact
  timing and location of CI is found, due to the
  still insufficient data resolution to resolve
  structures at the 10 km scale.

47
Thank you Comments/Questions
48
Sensitivity Studies
Effect of vertical filtering
With/without surface data
Correlation coefficient w.r.t. vertical layer
49
Analysis increments from a single sfc observation
Obs14.72 g kg-1 Bg 0 g kg-1 Ana14.69 g kg-1
Dryline
single sfc ob.
Isotropic B
Anisotropic B
Lr 4 grid intervals Lf 2 g/kg
50
List of Analysis Experiments
Experiment anisotropic Filter RMSE (g kg-1) CC CC with EF
ISO_RF No 0.35 0.84 0.83
ANISO_RF Yes 0.28 0.91 0.93
UB_RF Yes 0.34 0.86 0.83
Lr 4, in unit of grid point, which is optimal,
for ISO_RF experiment while Lr 3 is optimal for
ISO experiment using explicit filters.
51
Recursive Filter
ISO_RF Isotropic B with RF
CC0.84 RMSE0.35 g kg-1
analysis increment truth-background
52
ANISO_RF Anisotropic B based on truth
CC0.91 RMSE0.28 g kg-1
analysis increment xz cross-section along AB
UB_RF (covariance-updated) Anisotropic B based
on the analysis with isotropic B
CC0.86 RMSE0.34 g kg-1
53
guess
truth
ISO
ANISO
UB
54
Moisture analysis
ANISO
truth
ISO
guess
UB