16. February 2007

Investigators:

Marcus Paulat, Heini Wernli – Institute for Atmospheric Physics, University of Mainz

Christoph Frei – Bundesamt für Meteorologie und Klimatologie, MeteoSwiss Zürich

Martin Hagen - Institut für Physik der Atmosphäre, DLR Oberpfaffenhofen

Name of method:

SAL - Structure, Ampliutde, Location

Institute for Atmospheric Physics – University of Mainz

Literature / References: none- presentations at workshops in Boulder (2006) and ECMWF (2007)

- manuscript in preparation (Wernli et al., MWR)

• fields: so far only precipitation, in principle also applicable to other fields with well-structures signatures (e.g. wind gusts)

• presently used FC data: ECMWF and COSMO-LM precipitation forecasts (accumulated over 1 or 24 h) over Germany on rotated lon/lat grid with 7 km horizontal resolution

• presently used OBS data: gridded rain gauge data set on same grid as models, based upon ~4000 stations with 24 h values; 1-h disaggregated fields through combination of rain gauge and radar data

What kind of data?

• only gridded data

Marcus Paulat Christoph Frei Martin Hagen Heini Wernli

• SAL consists of three independent components

• components address quality of structure (S), amplitude (A) and location (L) of QPF in that area

• according to SAL a forecast is perfect if S = A = L = 0

• S requires the definition of precipitation objects, using a threshold value

• but: no attribution between precipitation objects in forecast and observations!

Basic strategy (I): General concept

• consider precipitation in pre-specified area (e.g. river catchment)

Marcus Paulat Christoph Frei Martin Hagen Heini Wernli

A = (D(Rmod) - D(Robs)) / 0.5*(D(Rmod) + D(Robs))

D(…) denotes the area-mean value (e.g. catchment)normalized amplitude error in considered areaA [-2, …, 0, …, +2]

L = |r(Rmod) - r(Robs)| / distmax

r(…) denotes the centre of gravity of the precipitation field in the areanormalized location error in considered areaL [0, …, 1]

S = (V(Rmod*) - V(Robs*)) / 0.5*(V(Rmod*) + V(Robs*))

V(…) denotes the weighted volume average of all scaled precipitation objects in considered areanormalized structure error in considered areaS [-2, …, 0, …, +2]

Marcus Paulat Christoph Frei Martin Hagen Heini Wernli

Basic strategy (II): Definition of the components

scaling for every object: R* = R / Rmax; R* [Rthresh/Rmax, …, 1]

x

R

x

R*Rmax 1

V(R*)circular precipitation

object

Rthresh Rthresh/Rmax

V(R)

Marcus Paulat Christoph Frei Martin Hagen Heini Wernli

Basic strategy (III): The S component

x

R

x

R*

V(R*)

x

R

x

R*Rmax 1

V(R*)

A < 0 S = 0

OBS

MOD Rmax

1

Marcus Paulat Christoph Frei Martin Hagen Heini Wernli

Basic strategy (IV): Example 1

x

R

x

R*

V(R*)

x

R

x

R*Rmax 1

V(R*)

A = 0 S > 0

OBS

MOD Rmax

1

Marcus Paulat Christoph Frei Martin Hagen Heini Wernli

Basic strategy (V): Example 2

S A L - statistics: 24h accumulated A

-co

mp

on

en

t

1

0

-1

-2

2

S-component0

S-component1 2-1-2 0 1 2-1-2

summer seasons 2001-2004 for catchment Rhine

aLMo ECMWF

Marcus Paulat Christoph Frei Martin Hagen Heini Wernli

Marcus Paulat Christoph Frei Martin Hagen Heini Wernli

Kind of verification information / Verification questions

• 3 independent components to quantify quality of structure, amplitude and location of QPF

• information is valid for a pre-specified area (e.g. river catchment, state, field campaign area, …)

• questions:

- is the domain averaged precipitation amount correctly forecast? -> A

- is the centre of the precipitation distribution in the domain correctly forecast? -> L

- does the forecast capture the typical structure of the precipitation field

(e.g. large broad vs. small peaked objects)? -> S

Marcus Paulat Christoph Frei Martin Hagen Heini Wernli

Strength

• physically meaningful information about amplitude, location and structure of QPF

• relatively intuitive

• SAL approach is close to “subjective human judgement” (claim)

• no attribution between objects is required (difficult for small objects)

• computationally simple

Marcus Paulat Christoph Frei Martin Hagen Heini Wernli

Weaknesses and limitations

• non-perfect QPFs can yield S = A = L = 0 (slight modification of L component is underway)

• no consideration of orientation of objects

• currently very simple definition of objects (threshold value)

• S value is sensitive to choice of threshold used for object definition (tests are underway to quantify sensitivity)

• so far only preliminary results for Germany have been computed