Anvendelse af vejrradar -Plan A Teknologisk Institut d. 3 September 2009 Lektor Michael R. Rasmussen Institut for Byggeri og Anlæg mr@civil.aau.dk 1
Data availability DMI radar data (C-band): - Rømø - Sindal - Virring - Stevns - Bornholm DHI radar data (X-band): - Aalborg - Århus - Vejle - Odense - Egedal - Hørsholm - Hvidovre 2
Statisk kalibrering 3
Statisk kalibrering Hvidovre LAWR 37 rain gauges within 60 km 35 rain gauges within 30 km 29 rain gauges within 15 km 4
Rain intensity (mm/min) Time series example I 0.25 0.2 rain gauge radar 500 x 500 m 0.15 0.1 0.05 0 12-Aug-2008 13:12:00 12-Aug-2008 15:36:00 12-Aug-2008 18:00:00 12-Aug-2008 20:24:00 Event Aug. 2, 2008 Accumulated rain depth, rain gauge 14.8 mm Accumulated rain depth, radar 12.7 mm R 2 0.26 5
Rain intensity (mm/min) Time series example II 0.45 0.4 rain gauge radar 500 x 500 m 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 02-Aug-2008 00:00:00 02-Aug-2008 02:24:00 02-Aug-2008 04:48:00 Event Aug. 2, 2008 Accumulated rain depth, rain gauge 21.8 mm Accumulated rain depth, radar 20.3 mm R 2 0.49 6
Statisk kalibrering af DMI radar DMI calibration Z=200 R 1.6 7
L L L L L On-going research: Stochastic calibration approach - The concept is to randomize the static and dynamic parameters which are used to calibrate the radars by appling a number of monte carlo runs - By comparing the model with measured rain data it is possible using the GLUEconcept to find the set of parameters which gives the best fit 1 total 1 1 0.5 0.5 0.5 0 1 0.5 0 5 10 no of timesteps 0 0.115 0.12 0.125 c2 0 0 5 10 number of gauges 1 0.5 0-5 0 5 Time displacement (min) 0 0 2 4 c1 x 10-4 red dots: simulations with a poorer fit compared to the static calibration green dots: simulations with a poorer fit compared to the static calibration 8
Stochastic calibration approach preliminiary results It is then possible to calculate the find the best set of parameters conditional on the observed rain and to calculate confidence bands 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 03:00 06:00 09:00 12:00 15:00 18:00 21:00 00:00
regn prognose Regn prognose servere (AAU) Aalborg Aalborg Afløbsprognose server (Krüger) Århus Århus Odense Odense Åben web: Animeret radar prognoser Holstebro DMI radar Holstebro Åben web: Flow/niveau prognoser Regn prognose middel for deloplandet Hvidovre Hvidovre Morten Grum/Krüger
direct runoff indirect runoff slow runoff diurnal variation saturation storage precipitation regn prognose regn prognosen Regn prognose servere (AAU) Aalborg Aalborg Afløbsprognose server (Krüger) Århus Århus Odense daily temperature Odense Åben web: Animeret radar prognoser afløbsmodel snow storage normal potential Holstebro evaporation DMI radar Holstebro Åben web: Flow/niveau prognoser Hvidovre Hvidovre flow measurement flow/niveau prognose Morten Grum/Krüger
regn prognose Regn prognose servere (AAU) Aalborg Aalborg Afløbsprognose server (Krüger) Århus Århus Odense Odense Åben web: Animeret radar prognoser Holstebro DMI radar Holstebro Åben web: Flow/niveau prognoser Hvidovre Hvidovre SRO målinger flow/niveau prognose Morten Grum/Krüger
C-Band Weather Radar Weather forecast model (HIRLAM) X-Band Weather Radar Storm and Wastewater Informatics (SWI)
Forecasting using radar data Global vector: Correlation between two radar images TREC Correlation between subsets of two radar images CO-TREC Correlation between subsets of two radar images with constraints on direction and speed as well as temporal averaging 14
Forecast example, July 7 2008, 14:25 Aalborg radar 0 min. 15 min. 30 min. 60 min Obs. CO-TREC Global vector
Forecast example, July 7 2008, 14:25 CO-TREC Global vector 16
Forecast example, July 25 2008, 16:00 Rømø radar 0 min. 30 min. 60 min. 120 min Obs. CO-TREC Global vector
www.vejrradar.dk