Empirical Models and Data Assimilation

1
Empirical Models and Data Assimilation

• John Horel
• Department of Meteorology
• University of Utah
• jhorel_at_met.utah.edu

2
Meteo 5120/6120 Applied Math and Statistics for
Environmental Scientistshttp//www.met.utah.edu/
class/jhorel/5120/ Draft Course Notes Chapter
6 5120.Notes.6.pdf

• Resource Kalnay, E., 2003 Atmospheric Modeling,
  Data Assimilation and Predictability
• model tool for simulating or predicting the
  behavior of a dynamical system such as the
  atmosphere
• Types of models include
• heuristic rule of thumb based on experience or
  common sense
• conceptual framework for understanding physical
  processes based on physical reasoning
• empirical prediction based on past behavior
• analytic exact solution to simplified
  equations that describe the dynamical system
• numerical integration of governing equations by
  numerical methods subject to specified initial
  and boundary conditions

3
Monitoring Current Conditions
September 6 20GMT
A D A S
4
Potential Classroom Discussion Points

• Why are analyses needed?
• Application driven data assimilation for NWP
  (forecasting) vs. objective analysis (specifying
  the present, or past)
• What are the goals of the analysis?
• Define microclimates?
• Requires attention to details of geospatial
  information (e.g., limit terrain smoothing)
• Resolve mesoscale/synoptic-scale weather
  features?
• Requires good prediction from previous analysis
• Whats the current state of the art and whats
  likely to be available in the future?
• Deterministic analyses relative to ensembles of
  analyses (see www.image.ucar.edu/DAReS/DART/tut1.p
  df Jeff Anderson)
• How is analysis quality determined? What is
  truth?
• Why not rely on observations alone to verify
  model guidance?

5
Observations vs. Truth

• Truth? You cant handle the truth!
• Truth is unknown and depends on application
  expected value for 5 x 5 km2 area
• Assumption average of many unbiased observations
  should be same as expected value of truth
• However, accurate observations may be biased or
  unrepresentative due to siting or other factors

6
How representative is a single observation site
for specifying an analysis value?

• Adequate instrumentation
• Good local siting
• Persistent ridging can lead to
• cold pools in gaps and high
• temperature on ridges
• Myrick and Horel
• (WAF 2006)

7
Sub-5km Variability in Terrain Height
Dark 200m
Myrick and Horel (WAF 2006)
8
Motivating issue Whats an appropriate analysis
given the inequitable distribution of
observations?
9
Motivating issue whats an appropriate analysis
given the variety of weather phenomena?
Elevated Valley Inversions
Front
O
?
O
?
?
O
O
O
O
z
T
10
Analyses vs. Truth
Analysis value Background value observation
Correction

• An analysis is more than spatial interpolation
• A good analysis requires
• a good background field supplied by a model
  forecast
• observations with sufficient density to resolve
  critical weather and climate features
• information on the error characteristics of the
  observations and background field
• good techniques (forward observation operators)
  to transform the background gridded values into
  pseudo observations
• Analysis error relative to unknown truth should
  be smaller than errors of observations and
  background field
• Ensemble average of analyses should be closer to
  truth than single deterministic approach IF the
  analyses are unbiased

11
Truth Continuum vs. Discrete
Truth is unknown Truth depends on application
Temperature
Truth
West
East
12
Discrete Analysis ErrorGoal of objective
analysis minimize error relative to Truth not
Truth!
Temperature
Truth
West
East
13
Observational Errors
Temperature
Truth
West
East
14
Assume Analysis Value 4C Truth
Relative to observations, analysis quality would
be judged to be low Relative to analysis,
observation quality would be judged to be low
9
-1
15
Broader Context

• Admiral Lautenbacher, NOAA Director
• Observations alone are often meaningless without
  the actions that provide economic and societal
  benefit
• GEOSS effort
• Comprehensive weather and climate observing
  system requires the integration and synthesis of
  observations into gridded analyses of current and
  past states of atmosphere

16
Proposed Analysis of Record program has several
components

• Real-time Mesoscale Analysis (RTMA)
• Hourly within 30 minutes of nominal observation
  time
• Initially a prototype, or proof-of-concept, for
  AOR
• NCEP/EMC, GSD/ESRL, and NESDIS building first
  phase
• Matures into quality real-time analysis component
• Undergoing testing grids available via ftp from
  NCEP
• http//www.emc.ncep.noaa.gov/mmb/rtma
• Analysis of Record (AOR)
• State-of-the-science analysis (best possible)
• Delayed for late arriving data assets
• Methodology to be determined (likely community
  effort)
• Accepted truth for use in studies and
  verification
• Climate Reanalysis on Mesoscale
• Apply mature AOR methodology retrospectively
• 30 year time history of AORs

Horel and Colman (BAMS 2005)
17
Exercise Objective

• Utilize MesoWest/MADIS observations as a resource
  to compare observations to analyses
• Netcdf files used for this exercise courtesy of
  ESRL/MADIS
• Introduce NCEP/EMC Real Time Mesoscale Analysis
  (RTMA)
• 2D variational adaptation of 3Dvar gridded
  statistical interpolation (GSI) used by NCEP
• Wu et al. (2002 MWR)
• http//www.emc.ncep.noaa.gov/mmb/rtma
• Download grib2 grids from ftp.emc.ncep.noaa.gov/mm
  b/mmbpll/rtma
• Demonstrate techniques to improve student
  understanding of
• observational errors
• characteristics of an analysis system
• Use conceptual models/physical reasoning to
  deduce analysis errors/successes of RTMA

18
MesoWest

• A cooperative program to collect, archive,
  distribute, and integrate into gridded analyses
  provisional environmental observations across the
  nation with emphasis on the western United States
  (BAMS Horel et al. 2002)
• Goal leverage existing environmental
  observations for a variety of uses with a focus
  on real-time weather applications
• Participation and coordination
• with MADIS/NSWOS
• Considerable effort placed on
• managing metadata
• Integration of environmental
• and GIS information

19
MesoWest Mix of surface observing assets
http//www.met.utah.edu/mesowest
20
Google API User Interface
21
ADAS

• Near-real time surface
• analysis of T, RH, V
• (Lazarus et al. 2002 WAF
• Myrick et al. 2005 WAF
• Myrick Horel 2006 WAF)
• Analyses on NWS GFE
• grid at 5 km spacing
• Background field RUC
• Horizontal, vertical anisotropic weighting

22
Real-Time Mesoscale Analysis

• The RTMA is a fast-track, proof-of-concept effort
  intended to
• leverage and enhance existing analysis
  capabilities in order to generate experimental
  CONUS-scale hourly 5-km analyses
• OPERATIONAL at NCEP and on AWIPS by end of FY2006
• Hourly temp, wind, moisture plus precip sky on
  5 km NDFD grid
• also provide estimates of analysis uncertainty
• establish benchmark for future AOR efforts
• build constituency for subsequent AOR development
  activities

G. DiMego, EMC/NCEP
23
RTMA Procedure

• Temperature dew point at 2 m wind at 10 m
• RUC forecast/analysis (13 km) is downscaled by
  FSL to 5 km NDFD grid (terrain is smoothed)
• Downscaled RUC used as first-guess in NCEPs
  2DVar analysis of ALL surface observations
• Wind analysis done using streamfunction and
  velocity potential
• Estimate of analysis error/uncertainty
• Precipitation NCEP Stage II analysis
• Sky cover NESDIS GOES sounder effective cloud
  amount

G. DiMego, EMC/NCEP
24
Why not another solution?

• Overarching Constraints
• No funding available must rely on volunteered
  resources
• Need to generate centrally as part of NCEP
  operations
• RSAS, OI, SCM or Barnes legacy techniques with
  known issues where obs density changes
• ADAS, LAPS or STMAS built around non-NCEP
  environments, too long costly to adapt, no
  expertise at NCEP (for OM and evolution)
• Ensemble Kalman Filter brand new technique
  (risk), does not scale to large number of obs,
  too long costly to adapt, no expertise at NCEP

G. DiMego, EMC/NCEP
25
Why 2DVar solution?

• 2DVar is subset of NCEPs more general 3DVar
  Grid-point Statistical Interpolation (GSI)
• Connected to state-of-the-art unified GSI
  development at NCEP / JCSDA
• 2DVar is already running in NAM (low risk)
• Anisotropy built into 2DVar provides way to
  restrict influence of obs based on
• Elevation (terrain height NAM ADAS in WR)
• Flow (wind)
• Air mass (potential temperature)
• 2DVar is more than fast enough to run overtop of
  hourly RUC in tight NCEP Production suite
• Can provide estimate of analysis error as well as
  accuracy via built-in cross-validation

G. DiMego, EMC/NCEP
26
ALL Surface Obs 89126 total
27
Step 1 Using MesoWest from browser to view
current weather

• http//www.met.utah.edu/mesowest
• Click Florida on map
• Toggle Settings to All Networks
• Select Metric Units
• Be sure to click Map It!
• Look at other features
• Zoom
• Station popups
• Show Tables
• Accumulate stations
• Overlay other variables
• Limited QC info

28
Step 2 Using MesoWest from browser to view past
weather

• http//www.met.utah.edu/mesowest
• Click Florida on map
• Toggle Settings to All Networks
• Be sure to click Map It!
• Click Select Date
• Uncheck Auto Current Time
• Change to 1730 UTC July 2 2006
• click Map It!
• Examine sea-breeze penetration on both coasts
• Consider measurement uncertainty, siting, network
  issues associated with observations

29
Step 3 Using IDV to Examine RTMA in Florida

• Keep browser up with observations for July 2,
  1730 UTC
• Bring up IDV with as much memory as possible
• From the File menu select
• 2006070217_flcase.xidv
• Be patient!
• MADIS netcdf file of mesonet observations takes a
  long time to load and requires about 800 mb of
  memory
• Discussion points
• Handling of sea breeze circulation
• Temperature features
• Inland water mask issues

30
Step 4 Using IDV to Examine RTMA in Utah

• Bring browser up with observations for northern
  Utah at 2330 UTC 9 July 2006
• Bring up IDV with as much memory as possible
• From the File menu select
• 2006070923_utcase.xidv
• Be patient!
• MADIS netcdf file of mesonet observations takes a
  long time to load and requires about 800 mb of
  memory
• Discussion points
• Wind relative to terrain, Lake, and deserts
• Temperature relative to (actual) and analysis
  terrain
• Great Salt Lake issues

31
RUC
32
RTMA
33
ADAS
34
RTMA Surface Pressure
35
ADAS Surface Pressure
36
(No Transcript)
37
Step 5 Explore Other Regions/Parameters of
Interest

• Modify mesowest station model change
  temperature/wind font to something other than
  black
• Use IDV tools to select other RTMA variable or
  region

38
Summary

• Encourage student learning about
• the basics of data assimilation and objective
  analysis techniques
• limitations of observations as well as analyses
• constructive evaluation of model and analysis
  errors using heuristic, conceptual and empirical
  models
• MesoWest available for many applications in the
  classroom
• Experimental RTMA Grib-2 files available now from
  NCEP
• AOR-type effort requires community participation
  to be effective