On%20re-stating%20the%20boundary%20layer%20characteristics%20in%20dispersion%20models

1
On re-stating the boundary layer characteristics
in dispersion models

• M.Sofiev, Finnish Meteorological Institute
• E.Genikhovich, Main Geophysical Observatory

2
Content

• Motivation for the study
• Problem statement
• Solution based on basic meteorological variables,
  non-iterative
• Accuracy of the solution, ways of verification
• comparison of the core method with observations
• comparison of re-stated fields with HIRLAM for
  2000
• Call for future studies

3
Motivation meteorological pre-processor

• Off-line dispersion modelling needs it for
  preparing the meteorological data to dispersion
  simulations (meteorological fields are NEVER in a
  complete agreement with formulations of
  dispersion model)
• extra variables, non-existing in the input files
• checking/restating the governing equations as
  they are in the dispersion model
• enhanced resolution in time and/or space
• Varying levels of complexity
• min simple interpolation range-checking
• max own assimilation of meteorological
  observations and recomputation of dynamic
  equations (MM5 / WRF)
• compromise completing the variable list for
  dispersion, plus re-stating those basic equations
  that are used explicitly

4
Motivation boundary layer parameters

• Numerous approaches to parameterization
• Specific variables and equations vary from model
  to model and even from run to run
• Most of ABL parameters are not explicitly
  validated in NWP models and not available in the
  output files
• Result practically all dispersion models include
  re-stating the ABL basic parameters in their
  meteorological pre-processor

5
Problem statement

• Available profiles of basic meteorological
  variables wind , temperature T, humidity q
• To find basic ABL parameters temperature,
  velocity and humidity scales T, u, q,
  Monin-Obukhov length L, profile of some
  characteristic of turbulence, e.g. KZ if
  K-theory is used
• Verification possibility consistency checking
  via comparison of sensible and latent heat fluxes
  HS, Hl.
• These fluxes are NOT validated inside NWP and
  thus should not be used as the input variables
  for the ABL re-stating.
• Deviation between NWP fluxes and dispersion
  models ones does not mean that one of the models
  is wrong but rather points to differences in the
  governing equations representation

6
Problem solution
Here all derivatives are NOT computed
numerically but rather taken from the analytical
approximations of profiles. Since zk1m, these
profiles can be taken purely logarithmic.
Non-logarithmic corrections start to play a
strong role at z/L0.5 Assuming the
logarithmic shape, it is enough to have 2 values
at the screening and the 1st model levels to
determine the profile. All fluctuating and not
well-defined parameters are inside the integral,
thus their effect is smoothed out
7
Method verification measurements
Eddy-correlation measurements, Tsimlyansk, 1976
Profile measurements, Cabauw, 1987
Groisman Genikhovich (1997), using the lower
available measurement level and ground surface
the temperature jump is estimated after
Zilitinkevich (1970)
8
Comparison with NWP(HIRLAM, ECMWF)

• Intuitively, there must be almost 11 agreement
• theoretical basis is more or less the same,
  variations in the formulations should not lead to
  excessive quantitative discrepancies
• within HIRLAM u,q,T profiles and heat fluxes are
  computed together, thus being highly correlated
• However, certain differences are inevitable
• latent heat flux depends on surface moisture a
  highly uncertain parameter and thus used as a
  tunable variable to meet overall temperature
  profile (obs the known moisture problem!)
• still, there are differences in the computational
  algorithms
• HIRLAM ECMWF provide accumulated fluxes e.g.
  for 3 hours, while u,q,T are instant, thus
  re-stated fluxes will be instant too

9
Verification statisticsHIRLAM, Jan-March 2000,
night
10
Verification statisticsHIRLAM, May-Sep 2000, day
11
Verification statistics time correlation,
quantile charts
12
Comparison oftime series (latent flux)
13
Comparison oftime series (sensible flux)
14
Discussion of comparison

• Fluxes should not be the same (above reasons)
• Given this, the re-stated and original NWP fluxes
  are close, often surprisingly close
• Near-neutral and stable cases are re-stated
  practically 11
• Strongly unstable cases in re-stated fields are
  somewhat less strong for terrestrial areas and
  more strong for marine ones
• The current methodology has reproduced the
  behavior of HIRLAM ABL module quite well.
  However, both of them lack the non-classical
  non-local elements
• Stable cases
• imposed free-flow static stability
• long-living stable BL
• capping inversions
• Unstable cases
• asymmetric vertical velocity spectrum in strong
  convection

15
Call for future studies

• Available methodology
• universal approach for re-stating the main ABL
  characteristics from the basic meteorological
  variables
• verification against observations showed good
  results
• comparison with HIRLAM showed quite nice
  correspondence
• Existing limitation
• the method has no treatment of strong
  stable/unstable cases when the local similarity
  theory cannot be used. Coherence with HIRLAM does
  not mean much because that model misses these
  constructions either
• certain deviations from HIRLAM are seen for
  unstable cases (not necessarily bad thing but
  reason is yet unknown)
• Research needed
• comparison with independent datasets
• ECMWF model fields
• mast data
• fine-tuning of the methodology (in particular,
  treatment of non-classical cases) and/or its
  implementation.