Assimilation of 3D Radar Data and Derived Objects on the Convective - - PowerPoint PPT Presentation

▶
assimilation of 3d radar data and derived
SMART_READER_LITE
LIVE PREVIEW

Assimilation of 3D Radar Data and Derived Objects on the Convective - - PowerPoint PPT Presentation

Jan 16, 2023 357 likes •572 views

Assimilation of 3D Radar Data and Derived Objects on the Convective Scale with an Ensemble-based Data Assimilation System 7 th International Symposium on Data Assimilation 24th of January 2019 RIKEN Center for Computational Science, Kobe, Japan

slide-1
SLIDE 1

Assimilation of 3D Radar Data and Derived Objects on the Convective Scale with an Ensemble-based Data Assimilation System

Christian A. Welzbacher

  • E. Bauernschubert, U. Blahak, R. Feger, A. de Lozar, R. Potthast,
  • C. Schraff, K. Stephan, M. Werner

7th International Symposium on Data Assimilation

24th of January 2019 RIKEN Center for Computational Science, Kobe, Japan

slide-2
SLIDE 2

Observations from Radar Network

➢ Systems:

17 Doppler C-Band

➢ Observables:

Reflectivity, radial wind, polarimetric moments

➢ Temporal resolution:

Volume scan + terrain-following precipitation-scan every 5 minutes

➢ Spatial resolution:

1° azimuthal angular 10 elevations (between 0.5° and 25°) 1 km radial (up to 180 km)

2 Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019

slide-3
SLIDE 3

Assimilation of (radar-) objects direct

Overview: Assimilation of Radar Data

Object generation: Thresholding of composites

Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019 3

  • I. Direct

assimilation of 3D radar data Radar data

  • II. Texture

indirect Object generation: Nowcasting Object ID

  • III. Features
slide-4
SLIDE 4

KENDA#: 4D-LETKF† (conventional data + 3D radar data / derived objects) [+ LHN*]

1 hour 1 hour

4

K: Kalman gain for ensemble mean

Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019

DA-System: KENDA

x(t): state vector in model space

Forward operator (e.g. EMVORADO

[Zeng2016]) H(x,t): model equivalent in observation space

#[Schraff2016], † [Hunt2007], * [Stephan2008]

slide-5
SLIDE 5

5 Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019

  • I. Direct

assimilation of 3D radar data

5

  • Full 3D radar data similar [Bick2016]
  • Height & elevation dependent
  • bservation error by statistics

[Desroziers2005]

  • Trigger for missing convection (warm

bubbles, LHN)

slide-6
SLIDE 6

Assimilated Observations

Observables: Reflectivity and radial wind (Volume scan only) Temporal resolution: Every 60 minutes (analysis time) Spatial resolution: 10 km (superobbing) Trigger: Warm bubbles if missing cells

6 Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019

Superobbing grid

slide-7
SLIDE 7

7 Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019

Threshold on radar reflectivity composites

Assimilation of (radar-) objects

  • II. Texture

indirect Object generation: Thresholding of composites

slide-8
SLIDE 8

8 Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019

Threshold on radar reflectivity composites

Assimilation of (radar-) objects

  • II. Texture

indirect Object generation: Thresholding of composites Poster p3-16 by Shigenori Otsuka et al.: promising results using this approach applied to precipitation with SPEEDY-LETKF

slide-9
SLIDE 9

Normalized counts (observation) Normalized counts (model equivalent)

Texture Assimilation - Idea

Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019 9

= fulfills condition (e.g. reflectivity > threshold) → Observation: normalized number of in box = p( )

slide-10
SLIDE 10

Texture Assimilation - Idea

Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019 10

Normalized counts (observation) Normalized counts (model equivalent)

= fulfills condition (e.g. reflectivity > threshold) → Observation: normalized number of in box = p( )

slide-11
SLIDE 11

11 Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019

  • Spatial distribution of features
  • Number of features in area
  • Mean distance of features in

area

Assimilation of (radar-) objects

  • III. Features

indirect Object generation: Nowcasting Object ID

  • Adaptive thresholding
  • 3D Cells with attributes
  • Tracks and forecasts available
slide-12
SLIDE 12

Feature Assimilation - Idea

Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019 12

Counts (observation) Counts (model equivalent)

= identified objects (with collective attributes)

/

slide-13
SLIDE 13

Feature Assimilation - Idea

Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019 13

= identified objects (with collective attributes)

/

Counts (observation) Counts (model equivalent)

slide-14
SLIDE 14

14 Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019

  • I. Direct

assimilation of 3D radar data

14

Results vs.

  • II. Texture
slide-15
SLIDE 15

Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019 15

  • 14 days with

hourly cycle

  • 40 members

+ deterministic

  • verified obs. same

in all experiments

DA-cycle: FG- and ANA-RMSE

3D Radar Data + bubbles Texture (T=30 dBZ, b= 7x7 gp) No Radar Data RH T WIND p

RMSE:

: first guess : analysis

slide-16
SLIDE 16

Lead time [h]

Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019 16

FSS 27.05.-10.06.2016 6h forecasts, hourly 10-18 UTC, deterministic run FSS-Threshold: 20 dbZ FSS-Threshold: 30 dbZ

Forecast Verification (6h, Reflectivity)

Fraction Skill Score [Roberts & Lean, 2008] 3D Radar Data + warm bubbles Texture (T=30 dBZ, b= 7x7 gp) No Radar Data FSS-Scale: 21x21 gp

→ Similar results for precipitation verification

slide-17
SLIDE 17

Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019 17

Forecast Verification (24h, Surface/Temp)

3D Radar Data + warm bubbles better No Radar Data better Improvement in RMSE up to 5 % ▪ Wind direction ▪ Wind speed ▪ Relative humidity ▪ Temperature

slide-18
SLIDE 18

18 Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019

Conclusions

  • 1. Direct assimilation of 3D Radar Data (Reflectivity and Radial Wind)

+ Positive in FG (RH, wind)* + Positive in reflectivity/precipitation (not shown) up to 0.2 in FSS over 6h* + Consistent improvement in upper air and surface verification*

+ Neutral/positive results in DA-Cycle, reflectivity, upper air and surface verification, worse in precipitation verification (compared to operational system, not shown)

  • 2. Texture

+ Neutral (RH, T), positive (Wind) in FG + Almost equal skill in FSS for same threshold ? Vertical localization, proper metric

  • Loss of information (smaller threshold)
  • Double counting of observations
  • 3. Features (future)

+ No matching, reducing double penalty ? Vertical localization, proper metric

  • Loss of information
  • Double counting of observations

* compared to no radar data

slide-19
SLIDE 19

19

Thank you for your attention Gracias por tu atención Bedankt voor uw aandacht Grazie per l'attenzione Merci de votre attention مكمامتهلب اركش Danke für Ihre Aufmerksamkeit

Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019

ご清聴ありがとうございました Obrigado pela sua atenção Takk for din oppmerksomhet Kiitos huomiostasi 谢谢你的关注 Спасибо за внимание

slide-20
SLIDE 20

20

Literature & Contact

[Bick2015]: doi 10.1002/qj.2751 [Desroziers2005]: doi 10.1256/qj.05.108 [Hunt2007]: doi j.physd.2006.11.008 [Roberts & Lean, 2008]: doi 10.1175/2007MWR2123.1 [Schraff2016]: doi 10.1002/qj.2748 [Stephan2008]: doi 10.1002/qj.269 [Zeng2016]: doi 10.1002/qj.2904

Christian A. Welzbacher - Radar Data Assimilation – ISDA 2019

  • Dr. Christian A. Welzbacher

Data Assimilation Unit (FE12) Deutscher Wetterdienst E-Mail: christian.welzbacher@dwd.de