An Integrated Rapid Multi-Scale Analysis and Min Chen Ming-Xuan Chen Cong-lan Cheng Feng Gao Lin-ye Song Prediction System (RMAPS-IN) in Beijing Area Ins nstitut ute of of Urba ban Meteor orol olog ogy, Beij ijin ing, CM CMA and its Preliminary Performance Evaluation 2016 2016-07 07-27 27
CONTENTS PART 01 WHAT’S RMAPS PART 02 RMAPS-IN FRAMEWORK PART 03 RMAPS-IN QPE+QPF PART 04 PRELIMINARY EVALUATION PART 05 CONCLUSION
WHAT’S RMAPS and RMAPS-IN RMAPS: R apid updated M ulti-scale What to INTEGRATE: A nalysis and P rediction S ystem DATA RADAR QPE: RMAPS-NOW Analysis Background: RMAPS-ST 4 Components: AWS OBSERVATIONS ST( S hort- T erm): WRF+WRFDA (0-24h) TECHNIQUES NOW( NOW casting): AutoNowcasting+VDRAS (0- Wind analysis background: RMAPS-NOW 2h) Motion Vector: RMAPS-NOW Urban Blending Weight: RMAPS-NOW IN( IN tegration): IN CA (Beijing Version, 0-12h) To provide 10-min updated unified 0-12h output >12h 2-12h 0-2h RMAPS-Urban RMAPS-NOW RMAPS-ST RMAPS-IN
CONTENTS PART 01 WHAT’S RMAPS PART 02 RMAPS-IN FRAMEWORK PART 03 RMAPS-IN QPE+QPF PART 04 PRELIMINARY EVALUATION PART 05 CONCLUSION
Doma main Con Configuration and and Downsc scal aling to 1 1km Interpolated Downscaled 3km Grid points: 511*581 from 30”USGS to 1km Topography in Global Terrain RMAPS-ST Height Grid distance: 1km Dataset 3223 AWS stations 6 Doppler Radars
RM RMAPS PS-INv Nv1.0 Fe Features 2-D Convective 2-D Analysis Horizontal Vertical 3-D Analysis Parameters and Forecasts 80 ％ Temperature CAPE, CIN, LCL, LFC Lambert 2-m Temperature True z-coordinate 0-4000m Humidity Instability Indices (LI, 2-m Relative projection TQ: dz=200m, Wind 1x1 km Showalter, ..) Humidity 20 layers Trigger-Temperature- 10-m Wind UV: dz=125m, Precipitation Deficit 32 layers Equivalent Potential Precipitation type Snowfall line Temperature Moisture Icing potential convergence Wind chill Mass convergence Visibility
RMAPS-IN STRATEGY PRECIPITATION T/Q/WIND RADAR QPE NWP (RMAPS-ST) = = + ∆ P ( i , j ) P ( i , j ) X ( i , j , m ) X ( i , j , m ) X ( i , j , m ) ANALYSIS ANA STAT ANA ST + BACKGROUND [ ] − ∆ = − * * * * OBS ST - v P ( i , j ) P ( i , j ) X X X RADAR RADSTAT k k EXTRAPOLATION PERSISTENCE+NWP FORECASTED TENDENCY Jim Wilson TT-DGNT 2016 NOWCASING [ ] P ( t ) = + − X (t ) X (t ) f X (t ) X (t ) EXTRAP i − − IN i IN i 1 T ST i ST i 1 12 hours 12hours FORECAST = + − = + − LENGTH AND * 1 1 - P ( t ) gP ( t ) ( g ) P ( t ) X (t ) g X (t ) ( g ) X (t ) - IN i EXTRAP i ST i IN i IN i ST i BLENDING STRETEGY
AWS c S cut ut-off t time and R and Runn unning Time T Tabl able Timetable 7*24 Updated time interval: 10-min RMAPS-IN products delay time: 8+2+2 minutes 8-min: AWS observation cut-off time 2-min: RMAPS-IN running 2-min: Products distribution Strategies to accelerate the running Compilation optimization OPENMP
CONTENTS PART 01 WHAT’S RMAPS PART 02 RMAPS-IN FRAMEWORK PART 03 RMAPS-IN QPE+QPF PART 04 PRELIMINARY EVALUATION PART 05 CONCLUSION
QPE of RMAPS-IN Radar QPE Rain-Gau Gauge Obse servat ations Downscal aling Distanc ance-weigh ghted d interpol polation on; Elevati vation n Depend ndent nt Bi-line Bi near ar Interpolati ation; n; Param Par ameterized QPE QPE scal aled; 2-D smoo mooth Interpolated AWS precipitation Precipitation Analysis Blended QPE Topography corrected AWS Scaled Radar QPE precipitation 22:00BJT, 28 th August 2014
Parameterization of elevation effects on precipitation in Beijing area complicated variability of precipitation- elevation gradients RMSE of 12-hour an intensity-dependent parameterization accumulated algorithm of elevation effects applied on precipitation minimum hourly precipitation in Beijing area the mountain precipitation is derived as a function of valley precipitation the physics of the orographic precipitation 参数 Pc 值 弱降水 临 界 Pcc 值 站点 对 a 值 b 值 process called the seeder-feeder 最小 RMSE(mm) # (mm) (mm) mechanism 妙峰山 - 菩 萨 鹿 1 1.59 0.5 0.59 0.5 31.3212 云蒙山 - 琉璃 庙 2 2.09 0.8 0.68 0.5 110.931 DATA: 玻璃台 - 镇罗营 3 1.60 0.4 0.76 0.5 109.181 AWS rain-gauge observations 松山 - 野 鸭 湖 4 1.79 0.7 0.56 0.5 33.6355 2008.8.1-2015.5.31 5 百花山 - 清水 2.65 1 0.82 0.5 47.195 12-hour accumulated rainfall observation 禾子 涧 - 古将 6 1.86 0.8 0.54 0.5 31.1186 (00-12UTC, 12-00UTC) mountain-valley station pairs are required Elevation difference < 500m Empirically Horizontal distance < 6km Good historical archived consistency Zmax=2000m Six representative station pairs Δ Z=500m
Climatological scaling of radar data Since the radar field is strongly range- RFC k RFC l dependent and contains biases due to topographic shielding it must be scaled before used in the precipitation analysis. 2787 AWS stations 1-hr accumulated AWS precipitation from 2 warm seasons (2014-2015) Totally 2157 time samples Station Scaling: FINAL RFC Grid-point Scaling: Scaled Radar QPE ：
RADAR QPE AWS PRCP SCALED QPE ELEVATION EFFECT QPE ANA
QPF by moving vector extrapolation 10-min accumulated radar 10-min accumulated AWS 10-min accumulated blending QPE with climatological precipitation interpolated with precipitation analysis scaling correction elevation correction STEP I RMAPS-In Gridded 500/700hPa Wind Moving vector defined with two extrapolated moving vectors Constraint from NWP consecutive 10-min precipitation analysis STEP II t-10min t t+10min Extrapolation Forecast STEP III
CONTENTS PART 01 WHAT’S RMAPS PART 02 RMAPS-IN FRAMEWORK PART 03 RMAPS-IN QPE+QPF PART 04 PRELIMINARY EVALUATION PART 05 CONCLUSION
Absolute error of the analysis and forecasts during 10 July – 10 Oct 2015 of RMAPS-IN Absolute Error of RMAPS-IN 10-min updataed analysis: Temperature: <0.35 o C daytime <0.20 o C nighttime U/V:<0.6m/s Relative Humidiry:<2.3% Absolute Error of RMAPS-IN 10-min updataed forecasts: The blending effect of NWP+AWS may last longer than 6 hours Temperature:<2.5 o C U/V:<1.3m/s Relative Humidity:<14%
RMAPS-IN ANALYSIS AND FORECAST Absolute Error
10-minute updated 2-m Temperature Analysis (19:00-20:50BJT, 7 th August 2015) RMAPS-IN VDRAS 1900BJT 7 AUG 2015 2000BJT 7 AUG 2015 A heavy convective storm case in Beijing Area The boundary of cold pool incurred by convection easily identified Well matched with the analysis from VDRAS but with more detailed structures
2015062521+04h 2015062521+05h 2015062521+03h 2015062521+07h 2015062521+08h 2015062521+06h RMAPS-ST Forecast 2015062600+04h 2015062600+01h 2015062600+02h 2015062600+03h 2015062600+05h 2015062600+00h RMAPS-INv1.0 Forecast 2015062600+01h 2015062600+02h 2015062600+00h 2015062600+04h 2015062600+03h 2015062600+05h RMAPSv1.0 ANA
09BJT 2 nd - 00BJT 3 rd May, 2016 ANALYSIS 0-1hr 1-2hr 09 时 -5 月 3 日 00 00 时 (BJ IN 系 统 在 2016 年 5 月 2 日 08 时 -5 月 2 日 23 时 (BJT) IN 系 统 在 2016 年 5 月 2 日 07 时 -5 月 2 日 22 时 (BJT) 2016 年 5 月 2 日 09 RMAPS APS-IN 2016 BJT) RMAPS APS-IN 逐小 时 累 积 QP 起始的 0-1hr 定量降水 预报 起始的 1-2hr 定量降水 预报 QPE
23BJT 19 th July – 21BJT 20 th July, 2016 ANALYSIS 0-1hr 1-2hr
INTERCOMPARISON of RMAPS-IN with BLENDING * RMAPS-IN BLENDING Data AWS Rain Gauge Observation/Radar QPE/NWP QPF Radar QPF/Radar QPE/NWP QPF Gridded Precipitation Analysis/0-6h Nowcasting/0-12h Products 0-6h Blended QPF Blended QPF Precipitation Y N Analysis Nowcasting Y N Blending Forecast Y Y Resolution 1km 1km Updated Interval 10-min 10-min Nowcasting Method Motion Vector Extrapolation Nowcasting products dependent Blending Time 0-6h 0-6h Length Blending Forecast Per 10-min Per 1 hour Output Interval Blending Forecast Time Weighted Blending of Extrapolation with Time Weighted Blending of Extrapolation with NWP Method NWP Blending Weight Time Weighted Hyperbolic tangent curve NWP Treatment Averaged into per 10-min Phase Correct and Intensity Calibration * PROTOTYPE IS FROM RAPIDS
CSI/BIAS of the operational forecasting systems during the warm season (20150716-20150905) 1-2h: RMAPS-IN>BLENDING~RMAPS-NOW>RMAPS-ST >3h: RMAPS-IN~BLENDING~RMAPS-ST>RMAPS-NOW
CASE CASE1: Precipitation Analysi sis and 0-1h 1h For orecas asts (2015071622U 2015071622UTC+1h 1h, Va Valid at at 2015071623U 2015071623UTC) RMAPS-IN ANA BLENDING ANA AWS RAIN-GAUGE Structures well matched RMAPS-IN 0-1hr FORECAST BLENDING 0-1 hr FORECAST RMAPS-NOW 0-1hr FORECAST RMPAS-ST FORECAST
Recommend
More recommend
