viai theh lilagrangian gg assimilationiili algorithmlihg...

1
i il i f d if j i i l i fi ld f l d d dl A i il ti f d ift ’t j t i i l it fi ld f tl d d dl Assimilation of drifters’ trajectories in velocity fields from coastal radar and model Assimilation of drifters trajectories in velocity fields from coastal radar and model Assimilation of drifters trajectories in velocity fields from coastal radar and model Assimilation of drifters trajectories in velocity fields from coastal radar and model i h L i i il i l ih LAVA i t ll b t@ i it via the Lagrangian assimilation algorithm LAVA maristella berta@sp ismar cnr it via the Lagrangian assimilation algorithm LAVA maristella berta@sp ismar cnr it via the Lagrangian assimilation algorithm LAVA. [email protected] via the Lagrangian assimilation algorithm LAVA. () ll () ff ( ) ld ( ) () l d( ) ll d () M Berta (1) L Bellomo (2) A Griffa (1 3) M G Magaldi (1 4) J Marmain (2) A Molcard (2) V Taillandier (5) M Berta (1) L Bellomo (2) A Griffa (1 3) M G Magaldi (1 4) J Marmain (2) A Molcard (2) V Taillandier (5) M. Berta (1), L. Bellomo (2), A. Griffa (1,3), M.G. Magaldi (1,4), J. Marmain (2), A. Molcard (2), V. Taillandier (5) M. Berta (1), L. Bellomo (2), A. Griffa (1,3), M.G. Magaldi (1,4), J. Marmain (2), A. Molcard (2), V. Taillandier (5) (1) ISMAR CNR It l (2) MIO U i ité d S dT l V F (3) RSMAS/MPO) U i it f Mi i USA (4) JHU B lti USA (5) LOV CNRS Vill f h F (1) ISMARCNR Italy (2) MIO Université du Sud ToulonVar France (3) RSMAS/MPO) University of Miami USA (4) JHU Baltimore USA (5) LOVCNRS Villefranche France (1) ISMARCNR, Italy (2) MIO, Université du Sud ToulonVar, France (3) RSMAS/MPO) University of Miami, USA (4) JHU, Baltimore, USA (5) LOVCNRS, Villefranche, France LAVA (LA i V i ti l A l i) i LAVA (LAgrangian Variational Analysis) is a LAVA (LAgrangian Variational Analysis) is a LAVA (LAgrangian Variational Analysis) is a i ti l l ith th t i il t d ift variational algorithm that assimilates drifters variational algorithm that assimilates drifters variational algorithm that assimilates drifters t j t i i th l it fi ld (f d trajectories in the velocity fields (from radar or trajectories in the velocity fields (from radar or trajectories in the velocity fields (from radar or dl) I h fi ld b models) It corrects the current fields by models) It corrects the current fields by models). It corrects the current fields by h d b df minimizing the distance between drifter minimizing the distance between drifter minimizing the distance between drifter observed positions and numerical trajectories observed positions and numerical trajectories observed positions and numerical trajectories. Th f it ll t ti i t t t Therefore it allows to optimize tracer transport Therefore it allows to optimize tracer transport Therefore it allows to optimize tracer transport di ti prediction Fi 1 Fi 3 prediction. Figure 1 Figure 3 prediction. Figure 1 Figure 3 I thi td LAVA i li d t TOSCA In this study LAVA is applied to a TOSCA In this study LAVA is applied to a TOSCA In this study, LAVA is applied to a TOSCA After LAVA correction the prediction of the model After LAVA correction the prediction of the model j i i h l i f After LAVA correction, the prediction of the model project experiment in the coastal area in front project experiment in the coastal area in front project experiment in the coastal area in front t j t i i i ifi tl i d (Fi 1 d ift trajectories is significantly improved (Fig 1: drifter f l ( ) f trajectories is significantly improved (Fig.1: drifter of Toulon (France) Surface currents are trajectories is significantly improved (Fig.1: drifter of Toulon (France) Surface currents are of Toulon (France). Surface currents are t j t i i bl k d l trajectories are in black; green and purple l bl f k ( k trajectories are in black; green and purple available from a WERA radar network (2km trajectories are in black; green and purple available from a WERA radar network (2km available from a WERA radar network (2km t th ti dl t j t i bf d represent synthetic model trajectories before and represent synthetic model trajectories before and ti l l ti 20 i t ) d f represent synthetic model trajectories before and spatial resolution every 20 minutes) and from spatial resolution, every 20 minutes) and from f LAVA) spatial resolution, every 20 minutes) and from after LAVA) after LAVA) th GLAZUR dl (1/64° ti l l ti after LAVA). the GLAZUR model (1/64° spatial resolution the GLAZUR model (1/64 spatial resolution, the GLAZUR model (1/64 spatial resolution, h ) 7 d ift h b d f th every hour) 7 drifters have been used for the every hour). 7 drifters have been used for the Rd t j t i it ll ith d ift every hour). 7 drifters have been used for the Radar trajectories compare quite well with drifters Radar trajectories compare quite well with drifters, i il ti d th lt lid t d i Radar trajectories compare quite well with drifters, assimilation and the results are validated using assimilation and the results are validated using b h f h i d b LAVA (Fi 2) assimilation and the results are validated using but they are further improved by LAVA (Fig 2) but they are further improved by LAVA (Fig 2) h 13 d if but they are further improved by LAVA (Fig.2). other 13 drifters other 13 drifters other 13 drifters. df l h l f ld f l Using drifters only the velocity field is satisfactorily Using drifters only the velocity field is satisfactorily Three assimilation cases are considered Using drifters only, the velocity field is satisfactorily Three assimilation cases are considered: Three assimilation cases are considered: reconstr cted and similar to the radar one bt reconstructed and similar to the radar one but i) ti f th d l it fi ld reconstructed and similar to the radar one, but i) correction of the radar velocity field reconstructed and similar to the radar one, but Fi 2 i) correction of the radar velocity field, Figure 2 i) correction of the radar velocity field, li it d t th i hb f th d ift th (Fi 3) Figure 2 limited to the neighbor of the drifter paths (Fig 3) ii) ti f th dl l it fi ld limited to the neighbor of the drifter paths (Fig.3). ii) correction of the model velocity field limited to the neighbor of the drifter paths (Fig.3). ii) correction of the model velocity field, ii) correction of the model velocity field, iii) t ti f th l it fi ld f d ift l iii) reconstruction of the velocity field from drifter only iii) reconstruction of the velocity field from drifter only . iii) reconstruction of the velocity field from drifter only . Th Md TOSCA j i k ld d The Med TOSCA project is aknowledged The MedTOSCA project is aknowledged.

Upload: hoangdung

Post on 20-Feb-2019

216 views

Category:

Documents


0 download

TRANSCRIPT

i il i f d if ’ j i i l i fi ld f l d d d lA i il ti f d ift ’ t j t i i l it fi ld f t l d d d lAssimilation of drifters’ trajectories in velocity fields from coastal radar and modelAssimilation of drifters trajectories in velocity fields from coastal radar and modelAssimilation of drifters trajectories in velocity fields from coastal radar and modelAssimilation of drifters  trajectories in velocity fields from coastal radar and model j yi h L i i il i l i h LAVA i t ll b t @ i itvia the Lagrangian assimilation algorithm LAVA maristella berta@sp ismar cnr itvia the Lagrangian assimilation algorithm LAVA maristella berta@sp ismar cnr itvia the Lagrangian assimilation algorithm LAVA. [email protected] the Lagrangian assimilation algorithm LAVA. a ste a.be ta@sp. s a .c . tg g gg g g

( ) ll ( ) ff ( ) ld ( ) ( ) l d ( ) ll d ( )M Berta (1) L Bellomo (2) A Griffa (1 3) M G Magaldi (1 4) J Marmain (2) A Molcard (2) V Taillandier (5)M Berta (1) L Bellomo (2) A Griffa (1 3) M G Magaldi (1 4) J Marmain (2) A Molcard (2) V Taillandier (5)M. Berta (1), L. Bellomo (2), A. Griffa (1,3), M.G. Magaldi (1,4), J. Marmain (2), A. Molcard (2), V. Taillandier (5)M. Berta (1), L. Bellomo (2), A. Griffa (1,3), M.G. Magaldi (1,4), J. Marmain (2), A. Molcard (2), V. Taillandier (5)( ) ( ) ( ) g ( ) ( ) ( ) ( )(1) ISMAR CNR It l (2) MIO U i ité d S d T l V F (3) RSMAS/MPO) U i it f Mi i USA (4) JHU B lti USA (5) LOV CNRS Vill f h F(1) ISMAR‐CNR Italy (2) MIO Université du Sud Toulon‐Var France (3) RSMAS/MPO) University of Miami USA (4) JHU Baltimore USA (5) LOV‐CNRS Villefranche France(1) ISMAR‐CNR, Italy (2) MIO, Université du Sud Toulon‐Var, France (3) RSMAS/MPO) University of Miami, USA (4) JHU, Baltimore, USA (5) LOV‐CNRS, Villefranche, France( ) , y ( ) , , ( ) / ) y , ( ) , , ( ) , ,

LAVA (LA i V i ti l A l i ) iLAVA (LAgrangian Variational Analysis) is aLAVA (LAgrangian Variational Analysis) is aLAVA (LAgrangian Variational Analysis) is ai ti l l ith th t i il t d iftvariational algorithm that assimilates driftersvariational algorithm that assimilates driftersvariational algorithm that assimilates driftersg

t j t i i th l it fi ld (f dtrajectories in the velocity fields (from radar ortrajectories in the velocity fields (from radar ortrajectories in the velocity fields (from radar orj y (d l ) I h fi ld bmodels) It corrects the current fields bymodels) It corrects the current fields bymodels). It corrects the current fields by) y

h d b d fminimizing the distance between drifterminimizing the distance between drifterminimizing the distance between driftergobserved positions and numerical trajectoriesobserved positions and numerical trajectoriesobserved positions and numerical trajectories.p jTh f it ll t ti i t t tTherefore it allows to optimize tracer transportTherefore it allows to optimize tracer transportTherefore it allows to optimize tracer transport

di tiprediction Fi 1 Fi 3prediction. Figure 1 Figure 3prediction. Figure 1 Figure 3g g

I thi t d LAVA i li d t TOSCAIn this study LAVA is applied to a TOSCAIn this study LAVA is applied to a TOSCAIn this study, LAVA is applied to a TOSCAAfter LAVA correction the prediction of the model

y, ppAfter LAVA correction the prediction of the modelj i i h l i f After LAVA correction, the prediction of the modelproject experiment in the coastal area in front After LAVA correction, the prediction of the modelproject experiment in the coastal area in frontproject experiment in the coastal area in frontt j t i i i ifi tl i d (Fi 1 d ift

p j ptrajectories is significantly improved (Fig 1: drifterf l ( ) f trajectories is significantly improved (Fig.1: drifterof Toulon (France) Surface currents are trajectories is significantly improved (Fig.1: drifterof Toulon (France) Surface currents areof Toulon (France). Surface currents aret j t i i bl k d l

( )trajectories are in black; green and purplel bl f k ( k trajectories are in black; green and purpleavailable from a WERA radar network (2km trajectories are in black; green and purpleavailable from a WERA radar network (2km j g p pavailable from a WERA radar network (2km

t th ti d l t j t i b f d(

represent synthetic model trajectories before andrepresent synthetic model trajectories before andti l l ti 20 i t ) d f represent synthetic model trajectories before andspatial resolution every 20 minutes) and from p y jspatial resolution, every 20 minutes) and fromf LAVA)

spatial resolution, every 20 minutes) and fromafter LAVA)after LAVA)th GLAZUR d l (1/64° ti l l ti after LAVA).the GLAZUR model (1/64° spatial resolution )the GLAZUR model (1/64 spatial resolution,the GLAZUR model (1/64 spatial resolution,

h ) 7 d ift h b d f thevery hour) 7 drifters have been used for theevery hour). 7 drifters have been used for the R d t j t i it ll ith d iftevery hour). 7 drifters have been used for the Radar trajectories compare quite well with driftersy ) Radar trajectories compare quite well with drifters,i il ti d th lt lid t d i

Radar trajectories compare quite well with drifters,assimilation and the results are validated using

j p q ,assimilation and the results are validated using b h f h i d b LAVA (Fi 2)assimilation and the results are validated using but they are further improved by LAVA (Fig 2)g but they are further improved by LAVA (Fig 2)h 13 d if

but they are further improved by LAVA (Fig.2).other 13 drifters

y p y ( g )other 13 driftersother 13 drifters.

d f l h l f ld f lUsing drifters only the velocity field is satisfactorilyUsing drifters only the velocity field is satisfactorilyThree assimilation cases are considered Using drifters only, the velocity field is satisfactorilyThree assimilation cases are considered: g y, y yThree assimilation cases are considered:reconstr cted and similar to the radar one b treconstructed and similar to the radar one buti) ti f th d l it fi ld reconstructed and similar to the radar one, buti) correction of the radar velocity field reconstructed and similar to the radar one, but

Fi 2i) correction of the radar velocity field, Figure 2i) correction of the radar velocity field,li it d t th i hb f th d ift th (Fi 3)Figure 2 limited to the neighbor of the drifter paths (Fig 3)g

ii) ti f th d l l it fi ld limited to the neighbor of the drifter paths (Fig.3).ii) correction of the model velocity field limited to the neighbor of the drifter paths (Fig.3).ii) correction of the model velocity field,ii) correction of the model velocity field,iii) t ti f th l it fi ld f d ift liii) reconstruction of the velocity field from drifter onlyiii) reconstruction of the velocity field from drifter only.iii) reconstruction of the velocity field from drifter only.

Th M d TOSCA j i k l d d) y y

The Med TOSCA project is aknowledgedThe Med‐TOSCA project is aknowledged.p j g