Radar Quality Control and Quantitative Precipitation Estimation Intercomparison Project Status Paul...

Radar Quality Control and Quantitative Precipitation Estimation

Intercomparison ProjectStatus

Paul JoeEnvironment Canada

Commission of Instruments, Methods and Observations (CIMO)Upper Air and Remote Sensing Technologies (UA&RST)

Outline

• Project Concept

• The Problem

• Overview of Data Quality Techniques

• Pre-RQQI Results

• Status

External Factors

Segmenting the DQ Process for Quantitative Precipitation Estimation

Remove Artifacts- Cleaned Up 3D volume

Estimating Surface/3D Reflectivity

Estimating Surface/3DPrecipitation

MosaicingSpace-Time Estimation

Focus on Reflectivity

NowcastingClear Air Echo as Information

Segmenting the DQ Process

Remove Artifacts- Cleaned Up 3D volume

Estimating Surface/3D Radar Moments

Estimating Surface3D Precipitation (Classification)

MosaicingSpace-Time Estimation in 3D

Reflectivity

Radial Velocity

Dual-Polarization

Every radar has clutter due to environment!

Sea Clutter and Ducting

Electromagnetic Interference

Techniques

CAPPI is a classic technique to overcome ground clutter

VVO

5o

4

3

2

1

0

Lines are elevation angles at 1o spacing, orange is every 5o.

Canada Australia

U.S./China VCP21 Whistler Valley Radar

3.0 CAPPI

1.5 CAPPI

There are a variety of Scan Strategies(CAPPI Profiles)

Make better or drop

The elevation angles but nature of weather important for CAPPI

2.5o

1.5o

0.5o

PPI’s1.5km CAPPI

Doppler Zero Velocity Notch

1. Doppler Velocity Spectrum• Pulse pair (time domain)• FFT (frequency domain)

2. Reflectivity statistics

Before

After

Doppler Filtering

SNOWRAIN

Too much echo removed! However, better than without filtering?

Data Processing plus Signal Processing

Texture + Fuzzy Logic + Spectral

Dixon, Kessinger, Hubbert

Data Processing plus Signal Processing

FUZZY LOGIC

Removal of Anomalous Propagation

NONQC QC

Liping Liu, CMA

The Metric of Success

Iso-range “Variance” as an intercomparison Metric

Daniel Michelson, SMHI

Accumulation – a winter season log (Raingauge-Radar Difference)

No blockageRings of decreasing value

Difference increases range!

almost

Convection

Stratiform

Snow

Vertical Profile of Reflectivity is smoothed as the beam spreads in range

Due to Earth curvature and beam propagating above the weather.

Variance Metric

Similar to before except area of partial blockage contributes to lots of scatterAlgorithms that are able to infill data should reduce the variance in the scatter!

Michelson

Proposed Metric

Alternate MetricsAccumulation of Radial Velocity should produce the mean wind for the site.

Both look believable, maybe difference is due to different data set length

nonQC QC

Modality

• Need a variety of techniques

• Need a variety of scan strategies

• Need a variety of data sets that integrate to a uniform pattern

• Need weather with a variety of artifacts

Pilot Study

Purpose is to test the assumptions of the project modality-Short data sets for uniformity-Check the interpretation of the metric-Variety of scan strategies, algorithms, etc-Evaluate feasibility

Uniform Fields

Sample Cases

• Uniform with local clutter (XLA)• Uniform with partial blocking (WVY)• Urban Clutter/Niagara Escarpment (WKR)• Strong Anomalous Propagation Echo (TJ 2006)• Strong AP with Weather (TJ 2007)• Sea Clutter (Sydney AU, Kurnell)• Sea Clutter / Multi-path AP (Saudi 2002)• Convective Weather with Airplane Tracks - One

season (TJ Radar 2007)

XLA

The data accumulates to uniform pattern. Widespread snow. A baseline case. IRIS formatted data. 24 elevation angles. Doppler (dBZT, dBZc, Vr, SPW) at low levels. Range res = 1km or 0.5 km. Az res = 1 or 0.5 degrees.

WVY

The data accumulates to uniform pattern with an area of blockage. Widespread snow. A baseline case. IRIS formatted data. 24 elevation angles. Doppler (dBZT, dBZc, Vr, SPW) at low levels. Range res = 1km or 0.5 km. Az res = 1 or 0.5 degrees.

WKR

The data accumulates to uniform pattern with an area of blockage. Widespread snow. Urban (skyscrapers) and small terrain clutter. IRIS formatted data. 24 elevation angles. Doppler (dBZT, dBZc, Vr, SPW) at low levels. Range res = 1km or 0.5 km. Az res = 1 or 0.5 degrees.

BSCAN of Z accumulation with no filtering, Doppler and CAPPI

CAPPI

Doppler

No Filtering

Azimuth

Ran

ge [

km]

0

1

00

Probability Density Function of Reflectivity as a function of range

Raw Doppler CAPPI

What length of data sets are needed?

Highly Variable More uniform, smoother, more continuous

The Techniques

• Doppler Notching

• CAPPI 1.5km

• CAPPI 3.0km

• Mixed of Doppler Notching and CAPPI

• Radar Echo Classifier (REC)– Anomalous Propagation– Sea Clutter

• REC-CMA

The Statistic

Spread of PDF (at constant range) for various cases and techniques…

Status

Status and Acknowledgements

•Kimata, Japan•Liu, China•Seed, Australia•Michelson, Sweden•Sempere-Torres, Spain•Howard, USA•Hubbert, USA•Calhieros, Brazil•Levizzani, Italy/IPWG•Gaussiat, UK/OPERA HUB•Donaldson, Canada

Data Providers

Algorithm Providers

Evaluation Team

Reviewers

Summary

• On-going• Data Providers, Processors

identified• ODIM_H5 format identified• BOM will host and convert

data for Data Processors• Initial Metric identified• Review

– Variety of techniques– Variety of scan strategies– Variety of data sets – Weather (e.g. convective, snow)

with a variety of artifacts

Alternate radial velocity metric