A nudging scheme for the assimilation of rainfall data: application to the 2001 Algerian Flood

1
25th EWGLAM 10th SRNWP meetings Lisbon,
Portugal, 6-9 October 2003 Joint session with
COST-717 WG3
A nudging scheme for theassimilation of rainfall
data application to the 2001 Algerian Flood
S. Davolio and A. Buzzi ISAC - Institute for
Atmospheric Sciences and Climate CNR - National
Research Council s.davolio_at_isac.cnr.it
INSTITUTE OF ATMOSPHERIC SCIENCES AND CLIMATE ,
ISAC-CNR
2
Summary

• Description of the assimilation scheme
• Idealized experiments (OSSE, Lagged Forecast)
• Case study Algeria flood 2001
• Results of the simulations and scores
• Sensitivity tests
• Conclusions

INSTITUTE OF ATMOSPHERIC SCIENCES AND CLIMATE ,
ISAC-CNR
3
THE NUDGING SCHEME

• k model ?-level (for each grid point)
• q(k) specific humidity before nudging
• q(k) saturation humidity profile (from model)
• typical relaxation time scale
• over/under saturation coefficient
• ?(k) vertical modulation profile O(1)

• If RRf lt RRt
• q(k) is forced gradually toward a (slightly)
  super-saturation profile
• If RRf gt RRt
• q(k) is forced toward an under-saturation profile

INSTITUTE OF ATMOSPHERIC SCIENCES AND CLIMATE ,
ISAC-CNR
4
Remarks (1)

• What are RRt and RRf ?
• Observed rainfall is accumulated over 1-3 hours
  interval.
• RRt mean constant rain rate within the
  accumulation interval.
• RRf forecast rain rate up to the current time
  step, updated every
• ? 20min (once the model precipitation is
  available convective
• step).
• Once available RRf is compared with RRt
• Therefore, the scheme does not instantaneously
  adjust the rain rate at each time step, but
  rather adjusts the rain accumulated up until the
  current time step, seeking to recover the
  observed precipitation at the end of the
  accumulation interval.

5
Remarks (2)

• The forcing is a function of the precipitation
  type (as estimated by the model)
• Stratiform precipitation
• ?s(k) is such that q is changed only in the
  middle-lower troposphere where large scale
• condensation takes place.
• RRf lt RRt q(k)
• RRf gt RRt q(k)
• Convective precipitation
• ?s(k) is such that q is changed only in the
  boundary layer.
• RRf lt RRt q(k)
• RRf gt RRt q(k)

• If RRf 0 and RRt gt 0 both types of
  precipitation are provisionally considered,
  unless the surrounding grid points are
  exclusively experiencing one type of rainfall.
• As for the convective (and all physical)
  tendency, the nudging adjustment is distributed
  over all time steps in the interval between two
  times at which rain rates are compared.

INSTITUTE OF ATMOSPHERIC SCIENCES AND CLIMATE ,
ISAC-CNR
6
Nudging vertical profiles

• In the presence of both types of precipitation,
  the profile for large scale precipitation is
  slightly modified in the lower part in order to
  have ?conv (k) ?ls (k) ?
  1

7
BOLAM MODEL

• Limited area, hydrostatic, PE model,
  ?-coordinate.
• u, v, ?, q, ps dependent variables.
• Horizontal resolution ? 16 km Vertical
  resolution 38 levels (highest resol. in the
  PBL).
• Lat-Lon rotated grid, horizontal discretization
  ? Arakawa C-grid.
• Stratiform precipitation described by means of 5
  prognostic variables (cloud ice, cloud water,
  rain, snow, graupel). Simplified approach
  (Schultz, 1995).
• Deep convection ? Kain-Fritsch convective
  scheme.
• Initial and boundary conditions ECMWF analyses
  0.5 x 0.5 res.

8
(No Transcript)
9
Idealized Experiments

• METHOD Lagged Forecast scheme
• Two different simulations from initial condition
  12 hours apart
• Control Run represents the reference state and
  provides the target rain rate.
• Forecast Run represents the real forecast to
  be improved.
• Nudging procedure applied for 12 hours to a
  simulation starting from the same initial
  condition of the Forecast Run (Nudging
  Run).

10
Results
11
Results
Improved! Rain band slightly shifted eastward but
correct in intensity Rain band in phase but
intensity too low
12
Results at the end of the nudging stage Hit Rate
and False Alarm Rate - 6h precipitation

• X axis
• precipitation thresholds (mm/6h)
• ( ) n. points where obs. rain rate gt threshold

13
RESULTS after the nudging stage Equitable Threat
Score vs simulation time
end of nudging
14
Cross section
15
(No Transcript)
16
Impact on cyclone development and evolution
F
C
12 hours after the end of the assimilation
17
Impact on cyclone development and evolution
N
C
12 hours after the end of the assimilation
18
(No Transcript)
19
Sensitivity to rainfall accumulation interval ETS
vs simulation time
Threshold 2mm/6h
Threshold 5mm/6h
nudging (3h)
nudging (1h)
forecast
nudging (6h)
nudging (2h)
20
Sensitivity to rainfall data errors ETS vs
simulation time
Threshold 2mm/6h
Threshold 5mm/6h
nudging (half prec)
nudging (shift prec)
forecast
nudging (double prec)
nudging
21
Conclusions

• The proposed nudging technique allows the
  assimilation of precipitation also when the rain
  is not purely convective, an advantage in
  midlatitudes with respect to reverse scheme.
• Encouraging results from the experiments the
  scheme seems able both to reduce and increase the
  precipitation patterns.
• Improvements in precipitation forecasts are
  associated to a better reproduction of vertical
  motion in the rainy area.
• The rainfall forecast improvements is observed
  during the assimilation phase and persists for
  several hours of free forecast (18-24 hours).
• Improvements on the dynamics the modification
  of the 3-dimensional humidity field (and
  consequently of the latent heat and temperature
  profiles through the model precipitation scheme)
  due to the nudging has a positive impact on the
  development and evolution of the cyclone.
• Assimilation of real data seems feasible, even
  if it is necessary to account for the statistical
  weight of the background field (model) and
  observation.

22
Particular ECMWF analysis and BOLAM solutions for
the event of Nov. 2001