Introduction to Radar Based Nowcasting WOO Wang-chun Forecast Development Division, Hong Kong Observatory E-mail: wcwoo@hko.gov.hk 26 July 2016 WMO WWRP 4th International Symposium on Nowcasting and Very-short-range Forecast 2016 (WSN16)
What We Need... Actual Quantitative Precipitation Estimate (QPE) Forecast Quantitative Precipitation Forecast (QPF) Severe Weather Lightning, Gust, Hail Services Forecasts & Warnings
Actual (QPE) Products Local Rainfall Map for Forecasters Regional Rainfall Map on GIS for Forecasters Local Rainfall Map for the Public
Forecast (QPF) Products For Public For Forecasters For Public
Severe Weather Products
Services For Internal Customer (Forecasters)
Services For Public
Scale matters …
WEATHER W WARNINGS GS
Radar-based vs NWP-based Rad Radar ar-base ased NWP-base ased Basically Image Processing Based on Primitive Equations • • – Correlation-based Not so good for first few hours • due to spin-up problem – Optical flow – Convolutional LSTM More skillful than Radar-based • afterwards Good for first few hours • Skills deteriorates rapidly • afterwards
Rain Gauge vs Radar vs Satellite Rain Gauge Satellite Radar Type In-situ Remote Sensing Remote Sensing Accuracy Best Moderate Worst Spatial Continuous, Continuous, Discrete Resolution up to 200 m up to 500 m Spatial Regional, effective Half the Globe At point only Coverage up to 256 km (radius) (geostationery) Cheap as single unit Expensive to launch Cost Expensive to operate Expensive as network Cheap to use
SWIRLS – HKO Rainstorm Nowcasting System
SWIRLS – Shor ort-range ge War arni ning o of Int ntense se Rai ainst nstorm i in n Local alized Systems ACTUAL :- Quant ntitative precip ipita tation e estim timatio ion ( (QPE) • rada dar-bas ased, d, rai ainga gauge ge-base sed and b blending w with th s satellite cloud images – TREND : :- Retrie ieval l of echo m motio tion • tracki cking by by max aximum cor orrelation (TR TREC) – tracki cking by by opt optica cal flow – obj object ct-or oriented tracking of s storm motion on – FOR OREC ECAST : :- semi-Lag agran angian an ad advecti tion t to extrapolate r rad adar ar r reflecti ctivity u ty up to o 6 6 / 9 9 hour ours • OUT UTPUTS : S :- computa tatio ion of gridde dded p d precipi pitation no nowcas cast (QPF) and and l loca cati tions of stor orm • objects cts o on n convectiv ctive w wind g gust, t, lightn tnin ing and and hail, suppor ort to to decision m on making ng UNCERTAINTY : UN :- probabilistic Q c QPF a and b blend nding ng wi with c h conv nvection on-pe permitting N g NWP WP model • PR PRODUCTS : S :- now owcasting pr prod oducts f for i internal us users a and p pub ublic •
QPE – Rainfall Calibration Module • Schematic diagram showing the calibration of radar reflectivity using real-time raingauge measurement. • Z-R relation for converting reflectivity to rainfall rate Z = b aR dBZ i =b dBG i + 10log(a) • Gridded rainfall analysis computed by Barnes successive correction or more advanced co- kriging algorithm
Barnes Analysis grid-point analysis by Barnes method • – interpolation with Gaussian weighting according to distance between data & estimation point – consider correction using residuals and grouping of rainguages G B h G G B : barnes estimation (mm) L : radius of influence G N 0 : number of gauge report G G G i : i-th gauge report (mm) w i : weight of i-th gauge h i : distance between gauge and estimation point
Co-kriging Analysis R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j G i G i G G i i G i G i G G i i R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j K(x K(x 0 K(x K(x 0 0 ) 0 ) ) ) R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j N M ∑ ∑ 0 0 = λ + λ h h h h G i G i G G co-Kriging estimate: ( K x ) ( x G ) ( x R ) i i 0 i 0 i j 0 j G i G i G G i i i =1 j =1 { } [ ] R R R R R R R R R R R R R R R R σ = − 2 R R R R R R R R R R R R R R R R 2 seek to minimize: E K x ( ) G x ( ) j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j 0 0 N M ∑ ∑ 0 0 λ = λ = subject to constraints: ( x ) 1 & ( x ) 0 i 0 j 0 i =1 j =1 Solution: N M ∑ ∑ 0 0 λ γ + λ γ + µ = γ = ( x ) ( x , x ) ( x ) ( x , x ) ( x ) ( x , x ), for n 1, , N i 0 GG n i j 0 GR n j G 0 GG n 0 0 i =1 j =1 N M ∑ ∑ 0 0 λ γ + λ γ + µ = γ = ( x ) ( x , x ) ( x ) ( x , x ) ( x ) ( x , x ), for m 1, , M i 0 RG m i j 0 RR m j R 0 RG m 0 0 i =1 j =1
Rad adar ar ec echo ho trac acking i in n SWIRLS
Com omparison of of e ech cho t o track cking TREC Optical flow
1-hr Quantitative P Preci cipitation on For orecast ( (QPF PF) TREC Optical flow Actual Rainfall
Rainfall nowcast from SWIRLS
Multi-Sensor QPE/QPF 512 km Satellite Channels 512 km 1728 km TMS Radar composite through collaboration with Guangdong 904 km 1804 km 1158 km
9-hour Nowcast of Radar Reflectivity Base Time : 2013-04-05 02: 12 HKT ( 6 hour forecast ) Actual Extrapolate with only Extrapolate with Multi- Hong Kong Radars Sensors
Decision Support - SWIRLS Integrated Panel Sample Time : 2014-07-22 15:24 SWIRLS SWIRLS Severe Ensemble Weather Rainstorm Viewer Nowcast (SERN) Amber expected in 36 minutes with criteria Forecast M and S met Rainfall (QPF) Thunder- storm with high gusts expected during this period Actual SWIRLS Rainfall Rainstorm (QPE) Viewer
Location-based Nowcasting Service • Available on “MyObservatory” mobile app • rainfall nowcast for the next 2 hours at your location – data from SWIRLS QPF • personalized automatic alerting service based on user location and expected rainfall
Location-specific Nowcasting Service • Personalized & customizable: – update frequency – notification intervals – range of detection – forecast location • as set on a map
Integration with GIS • Internet website ： http://www.weather.gov.hk/nowcast/prd/api/ • forecast rainfall maps over the Pearl River Delta region in the next 2 hours • updated every 12 min • downloadable as KML files
Rainfall Nowcast on Automatic Regional Weather Forecast website http://maps.weather.gov.hk/ocf/
Rainfall Nowcast Rainfall nowcast Provide image and animation sequence of rainfall forecast Click to display time series map over HK and Pearl River Delta for the next 2 hours 網址 : : http://maps.weather.gov.hk/ocf/index_uc.html
Cell Tracking in SWIRLS Group echo identification 39 dBZ 39 dBZ y a (major axis) V (TREC speed and direction) θ (orientation) x b (minor axis) = π A ab area ε = − 2 2 a b / a eccentricity max dBZ <--> max rainfall = π ε P 4 aE ( / 2 , ) perimeter ave dBZ <--> ave rainfall − θ = 1 tan ( b / a ) orientation ave (max50% rainfall) ∑ = × ∆ total intensity I dbZ A ave (max25% rainfall) ellipse
Tracking Capabilities Based on moving speed, size, overlapping area Searching radius translation merging splitting
Application to squall-line Two systems merging and moving steadily towards SE T+18min T+0min T+6min T+12min
Lightning C Con once ceptual M Mod odel • +/ − ve charges carried by ice and graupel respectively • charges separated vertically by updraft • Important distribution in the mixed layer from 0 ° C to - 20 ° C: + + + - - + + 0 to -20 ° C - - - -
Isothermal Reflecti tivity ty • 3D temp & height fields from hourly-updating model analysis • interpolate to radar grid (cartesian) • interpolate reflectivity to isothermal levels D 1 3-D D 0 ℃ D 2 A 1 temperature A 0 R 0 C 1 ℃ ℃ A 2 C 0 ℃ from NWP B 1 C 2 D B 0 ℃ C model B 2 R A B
Downburst Con once ceptual M Mod odel
Other her Sever ere e Wea eathe her N Nowcas ast Al Algorithm hms • Hail – 60-dBZ TOPS > 3 km – 0-2km VIL < 5 mm • probability of precipitation – Time-lagged ensemble of blending QPF • probability of lightning threat – time lagged ensemble of extrapolated sub-zero reflectivity fields based on optical flow
PoP by Time-lagged Ensemble Aggregate latest 10 RAPIDS QPF according to exponential decreasing weights
Probabilistic nowcast of precipitation SWIRLS Ensemble Rainstorm Nowcast (SERN) Spread of radar rainfall nowcast via selecting various parameters in echo motion retrieval
Recommend
More recommend
