Parameter Sensitivity of the LETKFWRF System for Assimilation of - - PowerPoint PPT Presentation

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Parameter Sensitivity of the LETKF–WRF System for Assimilation of Radar Observations in a Case of Deep Convection in Argentina

Paula Maldonado, Juan Ruiz, Celeste Saulo

Centro de Investigaciones del Mar y la Atmósfera (CIMA-CONICET/UBA) Departamento de Ciencias de la Atmósfera y los Océanos (DCAO-FCEN-UBA) UMI-IFAECI/CNRS - Servicio Meteorológico Nacional

12th EnKF Workshop

June 13th, 2017

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 1 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

High-impact Weather Events

Days per Year with Favorable Severe Parameters (Brooks et. al, 2003)

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 2 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

High-impact Weather Events

Days per Year with Favorable Severe Parameters (Brooks et. al, 2003)

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 2 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Convective-scale Forecasts

High-resolution NWP Models (> 4 km) ✔ Eliminate uncertainty associated to cumulus parameterization ✪ Significant errors in location and timing of convective systems Remote Sensing Observations ✔ Describe the state of the atmosphere in the convective scale Advantages of Using Radar Data High-resolution, 3D observations Temporal frequency necessary to retain the storm’s structure SINARAME Project

FIGURE – Weather radar network nowadays (blue

line) and SINARAME radars (red line). Simple-pol (dash line) and dual-pol (so- lid line) radars.

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 3 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Objectives

MAIN GOAL Develop, implement and evaluate a radar data assimilation system based on the Local Ensemble Transform Kalman Filter (LETKF) for very short-term weather forecast

• f high-impact weather events in South America

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 4 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Objectives

MAIN GOAL Develop, implement and evaluate a radar data assimilation system based on the Local Ensemble Transform Kalman Filter (LETKF) for very short-term weather forecast

• f high-impact weather events in South America

TALK’S GOALS Using Observing System Simulation Experiments (OSSEs) :

1 Evaluate the performance of the LETKF–WRF system 2 Asses the sensitivity of the LETKF–WRF system to :

The type and magnitude of the multiplicative inflation The specification of initial and boundary perturbations The localization scale

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 4 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Observing System Simulation Experiments (OSSEs)

Nature Run (NR) Synthetic Radar Observations Data Assimilation Verification

High-resolution Model Domain

Longitude (°) Latitude (°)

−80 −75 −70 −65 −60 −55 −50 −45 −40 −35 −60 −55 −50 −45 −40 −35 −30 −25 −20

Height (m)

200 400 600 800 1000 1200 1400 1600 1800

Nature Run WRF Model Configuration Domain : 500 x 500 km, 60 vertical levels Horizontal resolution : 500 m BC-IC : Downscaling (GFS) NR - Initial Assimilation Time

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 5 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Observing System Simulation Experiments (OSSEs)

Nature Run (NR) Synthetic Radar Observations Data Assimilation Verification

High-resolution Model Domain

Longitude (°) Latitude (°)

−80 −75 −70 −65 −60 −55 −50 −45 −40 −35 −60 −55 −50 −45 −40 −35 −30 −25 −20

Height (m)

200 400 600 800 1000 1200 1400 1600 1800

Nature Run WRF Model Configuration Domain : 500 x 500 km, 60 vertical levels Horizontal resolution : 500 m BC-IC : Downscaling (GFS) NR - Initial Assimilation Time NOT a Perfect Model Experiment Differences between NR and experiments : Horizontal resolution Initial and boundary conditions Microphysics parameterization

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 5 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Observing System Simulation Experiments (OSSEs)

Nature Run Synthetic Radar Observations Data Assimilation Verification

Synthetic Radar Observations Reflectivity and Doppler velocity Realistic radar geometry : 240 km range, 14 antenna elevations Uncorrelated observational errors Gaussian distribution

Longitude (°) Latitude (°)

−66 −65.5 −65 −64.5 −64 −63.5 −63 −62.5 −62 −61.5 −61 −36.5 −36 −35.5 −35 −34.5 −34 −33.5 −33 −32.5

Reflectivity (dBZ)

10 20 30 40 50 60 70

Longitude (°) Latitude (°)

−66 −65.5 −65 −64.5 −64 −63.5 −63 −62.5 −62 −61.5 −61 −36.5 −36 −35.5 −35 −34.5 −34 −33.5 −33 −32.5

Doppler Velocity (ms

−1)

−30 −20 −10 10 20 30

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 6 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Observing System Simulation Experiments (OSSEs)

Nature Run Synthetic Radar Observations Data Assimilation Verification

Synthetic Radar Observations Reflectivity and Doppler velocity Realistic radar geometry : 240 km range, 14 antenna elevations Uncorrelated observational errors Gaussian distribution

Longitude (°) Latitude (°)

−66 −65.5 −65 −64.5 −64 −63.5 −63 −62.5 −62 −61.5 −61 −36.5 −36 −35.5 −35 −34.5 −34 −33.5 −33 −32.5

Reflectivity (dBZ)

10 20 30 40 50 60 70

Longitude (°) Latitude (°)

−66 −65.5 −65 −64.5 −64 −63.5 −63 −62.5 −62 −61.5 −61 −36.5 −36 −35.5 −35 −34.5 −34 −33.5 −33 −32.5

Doppler Velocity (ms

−1)

−30 −20 −10 10 20 30

LETKF coupled with WRF model Horizontal resolution : 2 km Ensemble members : 60 Assimilation frequency : 5 min Assimilation period : 140 min Multiplicative inflation factor : 1.1 Localization scale : 4 km Initial Ensemble Random Gaussian perturbations to velocity and temperature fields Perturbation amplitude : 0.5

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 6 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Qualitative Evaluation

Analysis mean after 20 assimilation cycles (100 min)

Nature Run No Data Assimilation With Data Assimilation

Reflectivity field (shaded; dBZ), temperature anomaly 2 K (black contour) in 1 km

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 7 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Qualitative Evaluation

1-hr Ensemble Forecast initialized after 4 assimilation cycles (20 min)

Nature Run No Data Assimilation With Data Assimilation

10-min accumulated precipitation (shaded; mm), wind speed over 15 m s−1 (red contour)

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 8 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Quantitative Evaluation

Analysis mean 20 40 60 80 100 120 140 2 4 6 8 10 12 14 Assimilation Time (min) U (m/s) RMSE NODA RMSE SPREAD Ensemble Forecast 20 40 60 80 100120140160180 2 4 6 8 10 12 14 Forecast valid time (min) U (m/s) Errors increase with time in the analysis mean Ensemble spread collapses after 40 min and maintains very low values

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 9 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Sensitivity study

1 Covariance Inflation : Relaxation to Prior Spread (RTPS)

The analysis ensemble standard deviation is relaxed back to the background values at each grid point The multiplicative inflation is proportional to the amount of the ensemble spread being reduced by the assimilation of observations

2 Initial and Boundary Perturbations : Balanced Perturbations

Represent the large-scale flow (i.e. synoptic scale) Perturbation amplitude : 0.05 → 5% of climatology

3 Covariance Localization : Same as before

The state estimate is updated only by using observations within a local region defined by the localization scale radius

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 10 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Experimental Settings

Sensitivity

• Exp. Name

Configuration Multiplicative Inflation RTPS0.0

• Balanced and random perturbations with

RTPS0.7 0.05 and 0.5 amplitude, respectively RTPS0.8

• Localization scale : 2 km

RTPS0.9 Perturbations B&RP

• Relaxation to prior spread inflation : 0.9

RP

• Localization scale : 2 km

BP Localization Scale LOC1

• Relaxation to prior spread inflation : 0.8

LOC2

• Balanced and random perturbations with

LOC4 0.05 and 0.5 amplitude, respectively

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 11 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Sensitivity to the Type and Magnitude of Multiplicative Inflation

Root Mean Square Error Assimilation Time (min) U (m/s) 20 40 60 80 100 120 140 160 4 6 8 10 12 14 RTPS0.0 RTPS0.7 RTPS0.8 RTPS0.9 Consistency Ratio 20 40 60 80 100 120 140 160 0.2 0.4 0.6 0.8 Assimilation Time (min) U (m/s) Applying a RTPS scheme improves the performance of the filter Lower RMSE correspond to bigger inflation parameter

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 12 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Sensitivity to the Specification of Initial and Boundary Perturbations

Root Mean Square Error Assimilation Time (min) U (m/s) 20 40 60 80 100 120 140 160 4 6 8 10 12 14 B&RP RP BP Consistency Ratio 20 40 60 80 100 120 140 160 0.2 0.4 0.6 0.8 Assimilation Time (min) U (m/s) Biggest error when using only balanced perturbations Using both types of perturbations simultaneously helps maintain the spread high during the entire assimilation period

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 13 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Sensitivity to the Localization Scale

Root Mean Square Error Assimilation Time (min) U (m/s) 20 40 60 80 100 120 140 160 4 6 8 10 12 14 LOC1 LOC2 LOC4 Consistency Ratio 20 40 60 80 100 120 140 160 0.2 0.4 0.6 0.8 Assimilation Time (min) U (m/s) Assimilating a greater number of observations during the first 40 min helps reduce the errors while a smaller localization scale shows better results after 80 min Overall, the best results are achieved with a 2 km localization scale

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 14 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Conclusions

Assimilation of radar observations has a positive impact in both the analysis mean and the very short-term weather forecasts Using a constant multiplicative inflation factor produces the ensemble spread to collapse rapidly Adding a RTPS inflation scheme and balanced perturbations helps to keep the spread up The best configuration for the LETKF–WRF system is achieved, so far, by using a RTPS inflation parameter of with 0.9, initialization with both balanced and random perturbations and a 2 km localization scale Future Work Keep improving the assimilation system (e.g. test bigger inflation parameter and different amplitudes for perturbations) Test the assimilation system with real radar observations Generate experiments for different types of convection organization

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 15 / 15

Introduction Performance of the LETKF–WRF system Sensitivity experiments Conclusion l

Conclusions

Assimilation of radar observations has a positive impact in both the analysis mean and the very short-term weather forecasts Using a constant multiplicative inflation factor produces the ensemble spread to collapse rapidly Adding a RTPS inflation scheme and balanced perturbations helps to keep the spread up The best configuration for the LETKF–WRF system is achieved, so far, by using a RTPS inflation parameter of with 0.9, initialization with both balanced and random perturbations and a 2 km localization scale Future Work Keep improving the assimilation system (e.g. test bigger inflation parameter and different amplitudes for perturbations) Test the assimilation system with real radar observations Generate experiments for different types of convection organization Thank you!

P . Maldonado, J. Ruiz, C. Saulo CIMA-CONICET/UBA 12th EnKF Workshop June 12-14th, 2017 15 / 15