The short-time forecasting and nowcasting technology of severe convective weather for aviation meteorological services in China Xinhua Liu National Meteorological Center (NMC) of China Meteorological Administration (CMA)
Content Brief introduction of aeronautical meteorology products in NMC Probability-matching calibration method, neighborhood approach Multi-model integration method for precipitation and radar echo Time-lag precipitation correction method Precipitation forecasting method based on Titan and auto-station data Disscusion
Milestone 2007 2009 2015 2017 Established Asia Aeronautical Meteorological Service Website The aviation meteorological operational system began to be constructed Provide technical services and training to professional As the technical support center to participate in the preparation of WMO Asia Regional Hazardous Weather Advisory Centre Establishment of the aeronautical meteorological organization of the NMC Expand aviation meteorological business cooperation aeronautical meteorological users
Seamless monitoring and forecasting of severe convective weather Severe convective weather information extraction technology based on multi-source data, convective storm identification tracking and extrapolation technology SWPC was established in March 2009 Develop multiple key technologies Integrated monitoring technology Multi-model integrated forecast based high resolution model and upscaling forecast 2009 Extrapolation technology 2010 Analysis and diagnosis 2011 Probabilistic forecasting techniques 2015 Short time forecast techniques Ingredient-based method fuzzy logic and machine learning 2017 forecaster Object ive The objective prediction score is equivalent to the subjective prediction Complete operational technical support system
Numerical model Sand storm model Regional model 10 km, 3 km Global model 25 km Dust transportation model Typhone tracking model Sea wave model
Monitoring Lightning Satellite MCS identification tracking based on sattellite MCS identification tracking based on lightning
Composite monitoring platform lightning sattellite cloud classification pup AWS Quality control of thunderstrom wind Wind profile radar objective analysis of AWS
Extrapolation of of radar echo base on machine learning Extrapolation of radar echo base on optical flow TITAN extrapolation of radar echo forecast products for airport Time series of wind field and element forecast sounding profile forecast
Flash density Lightning Extrapolation Thunderstorm identification, tracking and extrapolation
3DVar 0.02 度分辨率 COTREC 3D wind field COTREC wind field MCS identification, tracking and CI Cloud Classification
Machine learning forecast thunderstorm thunderstorm gale short duration heavy rain Ingredient based forecast hail thunderstorm thunderstorm gale short duration heavy rain hail
Turbulence Icing Cb Precipition extrapolation base on TITAN Multi-model integration of radar echo Multi-model integration of short duration heavy rain
cross section
Low altitude flight forecast base on GRAPES-CR Vertical wind shear, convection, turbulence and cloud base height
Grid QPF Wind velocity on grid
Sand storm forecast Visibility forecast 5 km,2517 sites
Typhon forecast
Probability Probabilistic forecasts from a neighbourhood approach Method Introduction neighbourhood approach : When postprocessing the model forecast at a given location (x0, y0) of the model grid and for a given forecast lead time T0, a neighbourhood around this grid point is defined. It extends into both space (x, y) and time T. Figure 1 shows a schematic view of the neighbourhood in the (x, y) plane and in the (x, T) plane, where x and y denote the size of a grid box and T denotes the time step between successive model output times( Theis et al, 2005 ). Threshold Correction A technique is employed to correct model bias for precipitation fields based on a comparison of the frequency distributions of observed and forecast precipitation amounts. Hourly precipitation data are calculated for the observation and model prediction, respectively. Then, find the model prediction values corresponding to the observation thresholds(20 or 50mm); and the model prediction values are used as the threshold to compute the probability forecast product of Short-duration Heavy Rainfall. 0.1 0.08 0.06 0.04 0.02 8.87 20.76 OBS GRAPES-3km GRAPES-10km 0 0 10 20 30 40 50 Preciptation(mm)
Probabilistic forecasts from a neighbourhood approach Related probabilistic forecast products(based on Grapes_3km model): >35dBz radar echo Short-duration Heavy Rainfall(20mm&50mm) Radar Forecast & Obs Rain (20mm 50mm) Forecast & Obs
Probabilistic forecasts from a neighbourhood approach TS score: Heavy rain occurred in north of china between 19-20 th July 2016 Case: 02:00-07:00 on 20 th July 2016 0.9 0.8 0.7 0.6 0.5 0.4 0.3 TS FAR MISS 0.2 0.1 0 nearpro_5% Grapes_3km
GRAPES-RAFS Multi-model integration GRAPES-3KM Model of East China Model of Nanjing University Unified range Unified resolution Unified projection bilinear interpolation, 0.05*0.05 Hourly precipitation 5 Days Sliding correction Probability-matching Model of Guangdong Model of Beijing 短时预报客观指导产品 Rolling upgrade! High resolution! Multi-model integration! 6 Models 1 3 6h Precipitation corrected products Assign different weights 预警标准 1 3 6h precipitation 3h update Multi-model integration
Observation Issued by forecaster 融合 2 小时外推预报 08:00 forecast 17:00 forecast
0.3 0.2 0.1 0 TS of 15-22th July for short-duration heavy rain 华东 RAFS 集合 3h rolling update of 1 3 6h precipitation forecast
Short-range high-resolution model(grapes-rafs) precipitation forecast skill: Time-lagged ensembles Step1: Time-lagged ensembles spin-up and the uncertainty of the initial field of the model Step2: Frequency Matching Method for precipitation calibration Red point : monitor Blue point: forecast Blue line: fitting with the exponential function The model convective precipitation forecast is significantly weaker Step3:correlation coefficient method for displacement adjustment the location forecast offset for the main precipitation area
June-August 2015 threat score red:after calibration blue:model original 5-10mm monitor 10-15 15-20 20-25 25-30 model original after calibration
1 The multi-threshold reflectance factor extraction and time sliding accumulation Extraction of different of reflectance factor Accumulation
2 Multi-threshold of TITAN Low value High value Low value TITAN include high value TITAN No Yes Execute the algorithm from low to high Area of high value equal 30% of low value No Keep the low value area Yes The multiple threshold products are integrated and processed to obtain the extrapolation vector Keep the high value area Record identification and extrapolation information
3 QPF based TITAN and auto-station Multi-value blend CTREC wind field AWS Precipita tion Match and then keep the precipitation in a storm quality control Radar echo Extrapolation and accumulation every 10 minutes Output products
东北华北华东 华中华南西南
Nowcasting of precipitation based on TITAN and auto-station data Precipitation forecast for 10 minutes interval
Disscusion These methods have been used for short-time forecasting and nowcasting of thunderstorm, short duration heavy rain, thunderstorm wind and hail. These techniques have improved and promoted the short-time forecasting and nowcasting of the severe convective weather to some extent. The application of these technologies to aviation meteorological service will meet the urgent need of aviation weather for the short-time forecasting and nowcasting of the severe convective weather. The application of the above method shows good results to some extent. These technologies have a significant role to play in decisionmaking, whether for the weather forecasters at the airport or for the airport's controllers. It also improves the level and accuracy of the short-time forecasting and nowcasting for severe convective weather to a certain extent.
Thank you