Precipitation forecast

Question 1

Predictability

Answer 1

 With the quantitative precipitation forecast (QPF) intensity, type (e.g. rain or snow) and the space-time distribution of events (occurrence, start, end) are given.

 The predictability quantifies the performance of the forecast depending on the lead time

 The prediction performance is usually deacreasing with increasing lead

 Diverse forecasting approaches have special performances/ are differently suitable depending on the lead time (see Fig. 8.3).

Question 2

Predictability, Eulerian persistence:

Answer 2

Precipitation is not changing with respect to a fixed coordiante system (constant for x-y coordinate).

Question 3

Predictability Lagrangian persistence:

Answer 3

Precipitation is not changing for a moving coordinate system (constant within storm)

Question 4

Methods

Answer 4

• Numerical Weather Prediction Models (NWP): Deterministic modelling of atmospheric processes
• Statistical models & softcomputing: Time series models (e.g. ARMA), Regression, k- Nearest Neighbor, Neural Networks, Fuzzy Rule Based Modelling
• Radar tracking methods: Based on conceptual, statistical and/or soft computing methods

Question 5

Prinzipal Radar Tracking

Answer 5

1. Identification e.g. A > Athr , PI > PIthr

2. Tracking: shape matching, distance, velocity, -> “global” (window) with cross-correlation
or local (storm) with v = s/t

3. Linear Forecats: extrapolation w/o change of A, v, PI
4. Non-linear Forecast: extrapolation with change of A, v, PI

Question 6

Weather prediction system of the DWD

Answer 6

-Global ICON
13 km raster width, forecast horizon 180h,
(ICON-EU nested with Europe, 6.5 km)

-Regional model COSMO-D2
2,2 km raster width, Germany, Swiss,
Austria and border regions, forecast horizon 27h

-Ensembles ICON-EPS & COSMO-D2-EPS
With 40 or 20 ensemble members

Question 7

Radar based precipitation forecast of the DWD

Answer 7

 Approach RADVOR-OP for whole Germany
 Based on linear storm tracking with 2h forecast horizon
 Maximum temporal resolution is dt=5min
 Most frequent update-interval is every 15 min
 5-Min radar based forecasast without fitting to rain gauges
 Aggregated 1h forecasts will be fitted to rain gauges