Data assimilation for urban air quality simulation Vivien Mallet 1 , 2 1 Inria 2 CEREA, joint ENPC - EDF R&D laboratory, université Paris-Est With contributions by Raphaël Périllat 1 , 2 , Anne Tilloy 1 , 2 , SME Numtech, association Airparif BIS 2014, CNAM, June 2014 V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 1 / 20
Road network for trafic modeling Example of "Votre Air" (part of Paris), in collaboration with Airparif and Numtech 7000 +2.426e6 6000 5000 4000 3000 2000 1000 0 594000 596000 598000 600000 602000 604000 V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 2 / 20
Output points for the air quality model Example of "Votre Air" (part of Paris), in collaboration with Airparif and Numtech 7000 +2.426e6 1000 +2.427e6 6000 5000 800 4000 600 3000 400 2000 200 1000 0 594600 594800 595000 595200 595400 595600 0 594000 596000 598000 600000 602000 604000 V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 3 / 20
Simulated air quality Example of "Votre Air" (part of Paris), in collaboration with Airparif and Numtech Map of [NO 2 ] (µg m −3 ) simulated for 26 June 2011 at 07:00 V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 4 / 20
Air quality after data assimilation Example of "Votre Air" (part of Paris), in collaboration with Airparif and Numtech Map of [NO 2 ] (µg m −3 ) analyzed for 26 June 2011 at 07:00 V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 5 / 20
Data assimilation over Paris Example of "Votre Air" (part of Paris), in collaboration with Airparif and Numtech Map of [NO 2 ] (µg m −3 ), simulated Map of [NO 2 ] (µg m −3 ), analyzed 26 June 2011 at 07:00 V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 6 / 20
Data assimilation over Paris Example of "Votre Air" (part of Paris), in collaboration with Airparif and Numtech Map of [NO 2 ] (µg m −3 ), simulated Map of [NO 2 ] (µg m −3 ), analyzed 24 June 2011 at 09:00 V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 7 / 20
Data assimilation over Paris Example of "Votre Air" (part of Paris), in collaboration with Airparif and Numtech Map of [NO 2 ] (µg m −3 ), simulated Map of [NO 2 ] (µg m −3 ), analyzed 26 June 2011 at 15:00 V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 8 / 20
Data assimilation over Paris Example of "Votre Air" (part of Paris), in collaboration with Airparif and Numtech Map of [NO 2 ] (µg m −3 ), simulated Map of [NO 2 ] (µg m −3 ), analyzed 29 June 2011 at 19:00 V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 9 / 20
Data assimilation over Paris Example of "Votre Air" (part of Paris), in collaboration with Airparif and Numtech Map of [NO 2 ] (µg m −3 ), simulated Map of [NO 2 ] (µg m −3 ), analyzed 30 June 2011 at 08:00 Currently installed or being installed in at least 12 cities in France, by the SME Numtech. V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 10 / 20
Mobile application: air quality index Example of "Votre Air" (part of Paris), in collaboration with Airparif and Numtech V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 11 / 20
Underlying computations: quick introduction Best linear unbiased estimator, the so-called BLUE The model computes the vector c whose error is assumed to have Zero mean Variance B The observation vector o has an error with Zero mean Variance R No correlation with the error on c The observation operator H is introduced so that o is comparable with H c The linear estimator without bias and with minimum error variance trace is c ∗ = c + K ( o − H c ) � − 1 HBH ⊤ + R K = BH ⊤ � Note: this is the correction stage of a Kalman filter. Cf. the data assimilation library Verdandi, http://verdandi.gforge.inria.fr/ V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 12 / 20
Over the eastern part of Paris and for particulate matter Map of [PM 10 ] (µg m −3 ), simulated Map of [PM 10 ] (µg m −3 ), analyzed 1st September 2012 at 18:00 This approach was applied to NO 2 , PM 2.5 , PM 10 and black carbon. It was also used for Île-de-France (i.e., Paris region). V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 13 / 20
Toward network design Variance (µg 2 m −6 ) of the a posteriori error on [NO 2 ] V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 14 / 20
Toward network design Variance (µg 2 m −6 ) of the a posteriori error on [NO 2 ] V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 15 / 20
Toward network design Variance (µg 2 m −6 ) of the a posteriori error on [NO 2 ] V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 16 / 20
Toward network design Variance (µg 2 m −6 ) of the a posteriori error on [NO 2 ] V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 17 / 20
Toward an application to noise pollution Daily average noise map, computed by the city of San Francisco Cf. Sara Hachem’s talk— A Middleware Solution for Democratizing Urban Data V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 18 / 20
Management system for a smart city: available tools Modeling High-dimensional and complex numerical models for analysis, real-time simulation and forecasting Fast surrogate models derived from dimension reduction and statistical emulation Observations Fixed networks with few high-quality sensors Toward larger networks with lower-quality sensors Toward crowd-sourced, possibly mobile, sensing Mathematical methods Data assimilation, machine learning, statistics, . . . V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 19 / 20
Management system for a smart city: objectives and tools Objectives Tools Best evaluation of past, current Data assimilation and future states of the city Best estimate of forcings Inverse modeling Screening scenarios, impact studies Model reduction Risk management Uncertainty quantification Probabilistic forecasts V. Mallet (Inria & CEREA) Data assimilation at urban scale June 2014 20 / 20
Recommend
More recommend
