1. Tanja Winterrath Tanja.Winterrath@dwd.de
A new Method for the Nowcasting of Precipitation using Radar and NWP Data
Tanja Winterrath

2. Overview Overview of Radar Nowcasting at DWD Method Clutter Filter
Divergence of the Wind Field
Examples
Summary and Outlook
WSN05 - Toulouse

3. Radar Nowcasting at DWD
- Convective: KONRAD = recognition and linear extrapolation of cells (max. +1h) - Stratiform: Rosenow = pattern recognition and linear extrapolation of image structures (max. +2h) - Blending of the results
- Latent Heat Nudging in NWP (LMK)
Forecast lead time
Information content
after Golding (1998)
WSN05 - Toulouse

4. The new Approach...
Aim Extending the forecast beyond 2 hours
Basic Approach Displacement of radar precipitation data with horizontal LM wind fields  non-linear and temporally variable displacement of precipitation fields  external wind fields in contrast to directly derived displacement vectors
WSN05 - Toulouse

5. Input data
Radar product - Germany composite - Precipitation amount, derived with improved Z-R relationship (so-called RZ product)
NWP product - Horizontal wind components taken from LM (Lokalmodell, 7 km) or LM-K (2.8 km)
WSN05 - Toulouse

6. Method
Horizontal LM wind field - Choice of a suitable pressure level - Transformation and spatial interpolation onto radar-,grid‘ (1 km x 1 km) - Temporal interpolation in each time step of the advection scheme
Displacement of the precipitation field - 2d Eulerian advection scheme (Bott, 1993) - Time step approx seconds (CFL criterion)
WSN05 - Toulouse

7. Problems and Solutions
Numerical diffusion leads to a smoothing of clutter pixels  Clutter filter
Unrealistic stretching and compression of precipitation patterns (i.e., creation of new extrema)  Elimination of divergences of the wind field
WSN05 - Toulouse

8. Clutter Filter
Development of a procedure to completely eliminate clutter pixels from input data
Basic approach: - Marking of all pixels, the 31x surrounding square of which contains more than 85% of pixels with a value of less than 45% of the value of the centre pixel < 0.01 mm/h. - Replacement of the centre pixel‘s value by the mean of the data pixels within the surrounding square zero.
WSN05 - Toulouse

9. Clutter Filter - Results
Original
Radar Data
Radar Data after Clutter Elimination
WSN05 - Toulouse

10. Divergence-free Wind Field
Minimisation of the functional
Iterative Solution: - Gauß-Seidel - Successive Over-Relaxation (SOR) - Chebyshev Acceleration
(after Sherman, 1978)
WSN05 - Toulouse

11. Shown are the absolute values of the wind velocities;
Horizontal
LM wind field
original data
Horizontal
LM wind field
divergence-free
Difference
Shown are the absolute values of the wind velocities;
the mean relative change is approx. 10%.
WSN05 - Toulouse

12. Example 18.08.04, 500 hPa Start time = 20:00 Hourly successive...
... radar measurements:
... model results:
WSN05 - Toulouse

13. 1. Row: Measurements; 2. Row: 500 hPa; 3. Row: 700 hPa
+1h h h h
+2h, 700 hPa
+3h, 700 hPa 1 2 3 4
+1h, 700 hPa
+4h, 700 hPa
WSN05 - Toulouse

14. Summary
Aim: Extension of the precipitation nowcast beyond 2 hours to close the gap between linear extrapolation and model nowcast
Status: - Effective clutter filter realised - Area-preserving advection due to divergence-free wind field Work in progress...
WSN05 - Toulouse

15. Outlook Poster No. 2.39 on display now! Validation and Verification
Determination of optimal pressure level(s) for wind field extraction
Determination of cross-over times to linear extrapolation and NWP
Introduction into operational process chain
Poster No on display now!
WSN05 - Toulouse