Segmentation and tracking of anticyclonic eddies during a submarine volcanic eruption using ocean colour imagery

Senso s 2015, 15, 8732-8748; doi:10.3390/s150408732 senso s ISSN 1424-8220 www.mdpi.com/jou nal/senso s A icle Segmen a ion and T acking o An icyclonic Eddies du ing a Subma ine Volcanic E up ion Using Ocean Colou Image y Ja ie Ma cello 1,*, F ancisco Eugenio 1, Sheila Es ada-Allis 2 and Pablo Sang à 1 1 Ins i u o de Oceanog a ía y Cambio Global, Uni e sidad de Las Palmas de G.C., Campus Uni e si a io de Ta i a, 35017 Las Palmas de G an Cana ia, Spain; E-Mails: [email p o ec ed] (F.E.); [email p o ec ed] (P.S.) 2 Depa amen o de Física, Uni e sidad de Las Palmas de G an Cana ia, 35017 Las Palmas de G an Cana ia, Spain; E-Mail: [email p o ec ed]lpgc.es * Au ho o whom co espondence should be add essed; E-Mail: ja ie [email p o ec ed]; Tel.: +34-928-457365; Fax: +34-928-451243. Academic Edi o : Fab izio Lambe i Recei ed: 15 Janua y 2015 / Accep ed: 7 Ap il 2015 / Published: 14 Ap il 2015 Abs ac : The e up i e phase o a subma ine olcano loca ed 2 km away om he sou he n coas o El Hie o Island s a ed on Oc obe 2011. This ex ao dina y e en p o oked a d ama ic pe u ba ion o he wa e column. In o de o unde s and and quan i y he en i onmen al impac s caused, a egula mul idisciplina y moni o ing was ca ied ou using emo e sensing senso s. In his con ex , we pe o med he sys ema ic p ocessing o e e y MODIS and MERIS and selec ed high esolu ion Wo ld iew-2 image y o p o ide in o ma ion on he concen a ion o a numbe o biological, physical and chemical pa ame e s. On he o he hand, he e up ion p o ided an excep ional sou ce o ace ha allowed he s udy a a ie y o oceanog aphic s uc u es. Speci ically, he Cana y Islands belong o a e y ac i e zone o long-li ed eddies. Such s uc u es a e usually moni o ed using sea le el anomaly ields. Howe e hese p oduc s ha e coa se spa ial esolu ion and hey a e no sui able o pe o m submesoscale s udies. Thanks o he olcanic ace , de ailed s udies we e unde aken wi h ocean colou image y allowing, using he di use a enua ion coe icien , o moni o he p ocess o ilamen a ion and axisymme iza ion p edic ed by heo e ical s udies and nume ical modelling. In ou wo k, a no el 2-s ep segmen a ion me hodology has been de eloped. The app oach inco po a es di e en segmen a ion algo i hms and egion g owing echniques. In pa icula , he i s s ep ob ains an ini ial eddy segmen a ion using h esholding o clus e ing me hods and, nex , he ine OPEN ACCESS Senso s 2015, 15 8733 de ail is achie ed by he i e a i e iden i ica ion o he poin s o g ow and he subsequen applica ion o wa e shed o h esholding s a egies. The me hodology has demons a ed an excellen pe o mance and obus ness and i has p o en o p ope ly cap u e he eddy and i s ilamen s. Keywo ds: olcano; eddies; segmen a ion; clus e ing; h esholding; wa e shed; MODIS; MERIS; Wo ld iew-2 1. In oduc ion Mesoscala oceanic eddies a e ubiqui ous s uc u es ha can be ound anywhe e in he wo ld’s oceans, wi h a p edominance o an icyclonic o e cyclonic o exes [1]. Se e al s udies [2–4] poin ou ha hese mesoscala s uc u es can exe an impo an ole on he global clima e and local oceanic dynamic due o i s in luence in he anspo o ocean hea and momen um, as well as, in he enhancemen o e ical shea mixing and bo h bio ic and abio ic ace s. In he No h Eas A lan ic Ocean, he p esence o he Cana y A chipelago ac s as an obs acle o he oceanic and a mosphe ic lows, o igina ing a ich mesoscale a iabili y mainly ep esen ed by an icyclonic and cyclonic eddies spun o om he islands’ lanks [5,6]. The mechanisms o such a iabili y a e in oked h ough he pe u ba ion o he Cana y Cu en and T ades Winds by he all opog aphy islands such as El Hie o [7]. Eddy adius a ies be ween he o igin island adius (~12–25 km) a hei ini ial s age and c.a. 75 km a hei dissipa i e s age [5]. They a e be ween 300 and 700 m dep h depending on hei in ensi y [8]. An icyclonic eddies a e ini ially mo e in ense wi h a o a ion a e o abou 3 days whe eas he ini ial o a ion a e o cyclones is ca. 5 days [6]. Cana y Island eddies a e long-li ed eddies (>4 mon hs) being a he o igin o he Cana y Eddy Co ido which is he main pa hway o long-li ed eddies in he sub opical No h A lan ic Ocean [9], he e o e hey ha e no only a local impac bu also a emo e impac . In his con ex , ac i e and passi e sa elli e senso s ha e been ope a ing o in es iga e he a iabili y o mesoscale ea u es in he ocean. In pa icula , sea le el anomaly ields om al ime e s [10] ha e sys ema ically been used o de ec and moni o eddies. Some ecen s udies can be ound in [11–15]. Un o una ely, hese al ime e p oduc s ha e spa ial esolu ion o ens o kilome es and, hus, hey a e sui able o global s udies bu a e no adequa e o pe o m de ailed analysis a submesoscale le el, o analyse hei ilamen s o o s udy eddies wi h ho izon al scales below 50 km, which is close o wo old he clima ological i s ba oclinic Rossby adius o de o ma ion o he Cana y Islands egion. This scale is he na u al scale o mesoscale eddies a hei ini ial s age and dec eases wi h inc easing la i ude. Howe e , hese ine s udies can be o e come om space by op ical/in a ed senso s [16–19]. In his way, sa elli e de i ed ull esolu ion da a o chlo ophyll-a concen a ion o sea su ace empe a u e can be used o ho oughly cha ac e ize eddies. Un o una ely, no all hese s uc u es can be clea ly seen in ocean colou o he mal image y, as his depends on which way hey o a e, apa om he clima ological incon enience. An i-cyclonic eddies in he no he n hemisphe e a e pa icula ly di icul o de ec and analyse wi h hese da a. Senso s 2015, 15 8734 The e up ion o a subma ine olcano a El Hie o Island p o ided a unique and ou s anding sou ce o ace ha enabled us o s udy a a ie y o s uc u es. El Hie o, an island o he A lan ic Ocean coas o No h A ica was ocked by housands o emo s and ea hquakes since July 2011. On Oc obe 10 o he same yea an unde wa e olcanic e up ion s a ed 300 me e s below sea le el. Figu e 1 displays he loca ion o he olcano and images du ing he i s days a e he e up ion. I has o be poin ed ou ha his na u al ace elease has p o ided an excep ional oppo uni y o pe o m a de ailed moni o ing o an an icyclonic eddy. Low and high- esolu ion sa elli e images ob ained om MODIS, MERIS and Wo ldView senso s we e sys ema ically p ocessed [20] o p o ide in o ma ion on he concen a ion o a numbe o ma ine pa ame e s: chlo ophyll-a concen a ion, suspended ma e , PIC, POC, e c. These oceanog aphic emo e sensing da a also played a undamen al ole du ing ield campaigns guiding he Spanish Go e nmen oceanog aphic essel o he app op ia e sampling a eas [21]. (a) (b) (c) Figu e 1. Subma ine olcano e up ion a El Hie o Island du ing Oc obe 2011: (a) loca ion o he unde wa e olcano; (b) pho og aphs du ing he e up ion; and (c) sa elli e image by NASA EO-1 ALI© senso . T adi ional eddy de ec ion me hods using al ime e da a can be ca ego ized in o h ee classes [22]: (i) hose based on a physical pa ame e , usually he Okubo–Weiss app oach; (ii) hose based on low geome y, wi h he winding-angle app oach being a ep esen a i e me hod ha iden i ies eddies by clus e ing closed o spi al s eamlines; and (iii) hose based on sea su ace heigh o le el anomaly in which a h eshold is always equi ed o delimi eddy dimensions. Se e al h esholds ha e been applied, ypically anging om 6 cm o 10 cm, depending on he egion o s udy and size o he s uc u es o conside . In ou wo k, as we deal wi h highe esolu ion op ical ocean colou da a, he de ec ion app oaches a e di e en , mainly add essing segmen a ion echniques. Segmen a ion is pe o med based on di e en a ibu es o an image such as size, colou , ex u e e c. Se e al image segmen a ion echniques ha e been de eloped o a a ie y o applica ions in a ious ields o image analysis [23–27], om ea u e-based app oaches, such as clus e ing and his og am h esholding, as well as spa ial domain-based me hods, such as egion g owing-spli ing-me ging, ene gy-d i en ac i e con ou s, Senso s 2015, 15 8735 g aph cu s, and wa e sheds. Ou me hodology o he au oma ic and p ecise de ec ion o oceanog aphic ea u es is based on an ini ial s uc u e de ec ion ollowed by a ine de ail g owing p ocess. This pape illus a es he capabili ies o mul ispec al sa elli e emo e sensing sys ems o imp o e he unde s anding and moni o ing o subma ine olcanic p ocesses and, mainly, i p esen s a no el image p ocessing me hodology o he segmen a ion o oceanog aphic s uc u es. To ha espec , au oma ed de ec ion me hods a e essen ial o unde s anding he complex mo emen s and he dynamic cha ac e is ic o mesoscale eddies. The o ganiza ion o con en s is s uc u ed as ollows: Sec ion 2 includes de sa elli e image y used and desc ibes he wo-s eps segmen a ion algo i hm. Sec ion 3 p esen s he main esul s. Finally, Sec ion 4 summa izes he main conclusions. 2. Me hodology 2.1. Ocean Colou Image y Sa elli e da a om AQUA/TERRA-MODIS and ENVISAT-MERIS senso s ha e been he main sou ce o emo e sensing image in o ma ion o moni o ing El Hie o subma ine olcanic p ocesses. The NASA Mode a e- esolu ion Imaging Spec o adiome e (MODIS) has a iewing swa h wid h o 2330 km, i images he Ea h in 36 spec al bands, anging om 0.405 µm o 14.385 µm, and i collec s da a a h ee spa ial esolu ions: 250, 500, and 1000 me e s. The ESA Medium Resolu ion Imaging Spec ome e (MERIS) has a swa h wid h o 1,150 km, measu ing he sola adia ion e lec ed by he Ea h in 15 spec al bands om 0.4125 µm o 0.9 µm and wi h a pixel ield o iew o 0.019°, gene a ing ull esolu ion da a a 300 m esolu ion. We ha e implemen ed a me hodology o p o iding mul i empo al and mul isenso image y in o ma ion, a 1000 m (MODIS) and 300 m (MERIS) esolu ion, on he concen a ion o speci ic oceanog aphic pa ame e s, mainly chlo ophyll-a (Chl-a) concen a ion and di use a enua ion coe icien (Kd). The di use a enua ion coe icien is an impo an wa e p ope y ela ed o ligh pene a ion in he blue-g een wa eleng hs o he spec um. I can p o ide aluable in o ma ion abou he anspa ency o he amoun o dissol ed and suspended ma e ials in he wa e column. Sa elli e obse a ions o his coe icien , a he wa eleng h 490 nm, a e an e ec i e me hod o p o ide u bidi y maps o ocean and coas al wa e s a high spa ial and empo al esolu ions. Se e al empi ical and semi-analy ical models ha e been de eloped o es ima e Kd; howe e , hey a e gene ally applicable o clea open ocean wa e s. Recen ly, imp o ed models ha e appea ed [28] mos ly applicable o u bid coas al wa e s. In ou case, sa elli e es ima ions o he di use a enua ion coe icien and chlo ophyll-a concen a ion unde hese abno mal condi ions we e assessed [20]. A e compa ing in-si u measu emen wi h Chl-a p oduc s ob ained using he MODIS and MERIS algo i hms, we concluded ha chlo ophyll es ima ions we e e oneous in such a eas wi h ex eme u bidi y. On he o he hand, es ima ions o Kd using common ope a ional algo i hms we e no ully accu a e due o he physical-chemical al e a ions p o oked by he unde wa e olcano. Fo his eason, o MODIS and MERIS, we ob ained and alida ed imp o ed Kd models o coas al u bid wa e s [20]. As a consequence, only Kd p oduc s ha e been conside ed app op ia e o he moni o ing o he oceanog aphic s uc u es du ing he e en . Senso s 2015, 15 8736 Using he men ioned algo i hms, we p oduced pe iodic K d maps du ing he ime ame o he olcanic e up ion o allow he moni o ing o he plume e olu ion and o he gene a ed s uc u es. As an example, Figu e 2 p esen s mul i empo al and mul isenso K d maps o he plume and he eddy p og ession du ing speci ic days o Oc obe and No embe 2011. Apa om he egula moni o ing o he olcano ac i i y wi h he p o ision o all cloud ee oceanog aphic p oduc s o he go e nmen al and esea ch use s, a da abase composed by K d MODIS and MERIS images o No embe 2011, as shown in Figu e 2, displaying he e olu ion o he eddy is used o desc ibe he segmen a ion me hodology de eloped. Figu e 2. Mul i empo al K d p oduc gene a ed om MODIS (Te a and Aqua) and MERIS (ENVISAT) da a du ing Oc obe and No embe 2011. On he o he hand, he subma ine olcano was also moni o ed using medium and high esolu ion da a. In pa icula , he e y high esolu ion Wo ld iew-2 sa elli e was selec ed due o i s excellen cha ac e is ics ha ing eigh mul ispec al bands wi h 2 m esolu ion. Thus, Wo ldView-2 spec al bands we e a mosphe ically co ec ed and deglin ed p io o ob ain a e y high- esolu ion K d model. This no el model was de eloped and alida ed using ma chup da a se s wi h MERIS, MODIS da a, in-si u ansmi ances measu emen s and a seawa e adia i e ans e model [20]. This ex ao dina y esolu ion allows he analysis o he ma ine s uc u es a he highes le el o de ail. Senso s 2015, 15 8737 Figu e 3 displays a colou composi e Wo ldView-2 image acqui ed on 26 Oc obe . The image was p e-p ocessed o emo e he a mosphe ic and sun-glin e ec s in o de o eliably ob ain he e lec ance emi ed om he sea su ace. We can no ice he b own pa ch in he olcano cone loca ion and g eene wa e s indica ing he p esence o lowe concen a ions o olcanic ma e ial. Thus, we can see ex emely u bid wa e s wi h alues abo e 1.5 m−1 in he cen e o he e up ion and g adually dec easing in he su oundings. The di use a enua ion coe icien was es ima ed using ou imp o ed algo i hm [20]. Un o una ely, due o he la ge ho izon al ex ension o eddies, hei con inuous moni o ing and acking using his sou ce o in o ma ion is no sui able and ex emely expensi e. Wo ldView-2 was p og ammed o collec da a du ing hese mon hs bu , un o una ely, only he displayed image was app op ia e (Oc obe 2011), while o he es , he olcano ac i i y had empo a ily s opped o was e y low in he sensed a ea. (a) (b) Figu e 3. Wo ld iew-2 image o Oc obe 26, 2011: (a) colou composi e (bands 5-3-2) and (b) e y high esolu ion di use a enua ion coe icien map (clouds masked in black). 2.2. Me hodology o he P ecise Segmen a ion o Eddies Me hods o de ec ing eddies a e e y impo an and can help o be e unde s and he p ocesses in ol ed in eddy gene a ion, e olu ion and decaying, as well as hei impac s on ma ine en i onmen and clima e change. In addi ion, such eddy segmen a ion allows o he e alua ion o he nume ical modelling pe o mance in e ms o eddy cen e, posi ions, sizes and mo ion. The p ecise de ec ion app oach we ha e de eloped is p esen ed in Figu e 4. The me hodology is composed by wo s eps: (i) Ini ial eddy de ec ion: Fi s , a p e-p ocessing is applied o he MODIS/MERIS Kd images. Nex , he ini ial segmen a ion s age is applied wi h he possibili y o selec di e en clus e ing and h esholding algo i hms. Finally, a pos -p ocessing is equi ed o emo e isola ed noisy pixels o Senso s 2015, 15 8738 o ill holes in he eddy s uc u e due o clouds. A e he ini ial de ec ion, he s uc u e is well iden i ied; howe e he ine de ail o he eme ging ilamen s canno be p ese ed wi hou inco po a ing unwan ed a eas. (ii) Eddy g ow h: The p ecise delimi a ion is o undamen al impo ance when analysing he s uc u e and i s e olu ion, so con olled g owing echniques a e applied only whe e ilamen s appea . Poin s o g ow and hei di ec ions a e ound using he skele on o he ini ial s uc u e de ec ed. A each poin , a window wi h he co esponding o ien a ion and size masks he image and new pixels a e added o he s uc u e a e he applica ion o h esholding o wa e shed egion g owing algo i hms. This p ocess is i e a i ely applied o achie e he desi ed ine segmen a ion. Figu e 4. Me hodology o he au oma ic segmen a ion o eddies using op ical image y. The di e en modules ha in eg a e he me hodology a e nex explained in de ail. 2.2.1. Ini ial Eddy De ec ion P e-P ocessing The p ope segmen a ion o eddies is in e e ed by he p esence o clouds, land, e c. So, a cloud mask (CM), as well as he land mask (LM) a e applied o he geome ically co ec ed (GC) Kd image. Despi e he good pe o mance o he cloud de ec ion algo i hm [29], in e ms o de ec ion and alse ala m p obabili ies, no all he exis ing cloud pixels a e masked. The e o e hey may appea as noise in he segmen ed image. In o de o minimize i s e ec , an op ional 3 × 3 median il e ing (median) can be applied o he co ec ed and masked image. This il e ing me hod has also been selec ed because o i s smoo hing e ec wi hou edge de e io a ion. Thus, he p e-p ocessed image ( ) is ob ained as ollows:   (,)=󰇟󰇝(,)󰇞∩(,)∩(,)󰇠 (1) Ini ial Segmen a ion Se e al segmen a ion algo i hms ha e been de eloped o e he yea s, howe e oceanog aphic s uc u es a e di icul o be p ope ly de ec ed wi h con en ional algo i hms. A e es ing a a ie y o echniques and analysing hei esul s wi h expe oceanog aphe s, he ollowing me hods o he Senso s 2015, 15 8739 p elimina y eddy segmen a ion ha e been selec ed: K-means Clus e ing, Fuzzy K-means Clus e ing, Mean Shi Clus e ing and Th esholding. In [30,31] a de ailed e alua ion o 36 au oma ic global and local h esholding echniques was pe o med and he bes me hods belonging o di e en ca ego ies (his og am shape-based, clus e ing-based, en opy-based, objec a ibu e-based, spa ial me hods, e c.) we e inco po a ed o he eddy segmen a ion ool. In ou s udy, sa is ac o y esul s o segmen ma ine s uc u es a e achie ed by echniques like Riddle , O su, Gaussian Mix u e Modelling, Pun o Li. In pa icula , O su algo i hm p o ides an in e media e, ye accu a e, solu ion and i can be selec ed as a comp omise echnique, p o iding obus ness and excellen esul s o he majo i y o images. A e he applica ion o he segmen a ion algo i hm, we ge a bina y image co esponding o he Ini ial Segmen ed Eddy (ISE): (,) =  (,) (2) Some examples o possible au oma ic segmen a ions a e included in Figu e 5. As i can be app ecia ed, oceanog aphic s uc u es a e no simple o au oma ically de ec due o he ac ha g adien s a e weak, gene a ing se e al unconnec ed bo de lines and making o g ea di icul y he s uc u e iden i ica ion based on edge echniques; he p esence o noise, mainly due o a mosphe ic phenomena; he low signal ecei ed a senso le el; and he s ong mo phological a ia ion ha impedes an accu a e geome ic ep esen a ion and he absence o a alid analy ical model o he s uc u es. (a) (b) (c) (d) Figu e 5. S uc u es de ec ed by di e en segmen a ion algo i hms: (a) K d p e-p ocessed image y; (b) K-means; (c) Fuzzy K-Means; and (d) O su. T ying a segmen a ion using manual h esholding echniques p o ides he esul s included in Figu e 6a–c. I can be app ecia ed ha e en using a manual p ocedu e and adjus ing he K d h eshold o di e en alues (0.04 o 0.06) i is no possible o achie e a he same ime he ine de ail o he comple e ilamen s wi hou adding undesi ed lowe le els o K d . Pos -P ocessing The segmen ed image includes he eddy plus addi ional a e ac s due o he p esence o noise, isola ed u bid wa e zones and he emaining cloud pixels no p ope ly masked p e iously. Thus, a pos -p ocessing s ep is equi ed consis ing mainly on he applica ion o mo phological ope a o s [32]. Senso s 2015, 15 8740 A closing ope a o (∅), using a s uc u ing elemen ( ypically o size 3 pixels) o smoo h and connec pixels, mainly belonging o b oken ilamen s; ollowed by a emo al o holes (RH) ope a ion usually o ill cloudy pixels belonging o he eddy. The esul ing ini ial segmen ed eddy ( ) is ob ained as ollows:   (,)=󰇟∅󰇝(,)󰇞󰇠 (3) An addi ional capabili y included in he me hodology is he elimina ion o isola ed objec s whose size is smalle han a h eshold de ined by he use . (a) (b) (c) (d) Figu e 6. S uc u es de ec ed by manual h esholding algo i hms wi h global K d h eshold alues o : (a–c) 0.06, 0.05 and 0.04, espec i ely; (d) Segmen a ion desi ed by he oceanog aphe s. 2.2.2. Eddy G owing A e he p eceding asks, he s uc u e is iden i ied and, selec ing he co esponding pa ame e s, i is possible o achie e di e en le els o segmen a ions o ge he desi ed eddy de ec ion. Un o una ely, none o he es ed me hods is able o p oduce he esul s equi ed by ou oceanog aphe s, as p esen ed in Figu e 6d, who wan o p ese e he co e o he eddy bu also o pe ec ly delimi a e he eme ging ilamen s bu wi hou including noise associa ed o wa e wi h less u bidi y su ounding i s co e. The p ecise delimi a ion o he ilamen s p o ides undamen al in o ma ion abou he submesoscale a iabili y such as he ilamen a ion o he eddy no obse ed be o e wi h such. In addi ion, such p ecise delimi a ion is essen ial o pe o m a compa ison o a model ou come wi h obse a ional da a om emo e sensing image y. To achie e his ine segmen a ion, a wo-s ep p ocedu e has been de eloped, as shown in Figu e 4. I allows he applica ion o con olled g owing echniques bu only in egions whe e ilamen s appea . The p ocedu e is as ollows: Iden i ica ion o Poin s o G ow and Di ec ion Fi s , he skele on (SK) o he segmen ed eddy is used o iden i y he poin s o g ow and i p o ides, as well, he g owing di ec ion (see Figu e 7a in ed colou ). In his wo k, he skele on is ob ained by Senso s 2015, 15 8747 6. Sang à, P.; Auladell, M.; Ma e o-Díaz, A.; Peleg í, J. L.; F aile-Nuez, E.; Rod íguez-San ana, A.; Ma ín, J.M.; Mason, E.; He nández-Gue a, A. On he na u e o oceanic eddies shed by he Island o G an Cana ia. Deep-Sea Res. I 2007, 54, 687–709. 7. Jiménez, B.; Sang à, P.; Mason, E. A nume ical s udy o he ela i e impo ance o wind and opog aphic o cing on oceanic eddies shedding by deep wa e islands. Ocean Model. 2008, 22, 146–157. 8. Las e nas, S.; Piedeleu, M.; Sang à, P.; Dua e, C.M.; Agus í, S. Fo cing o dissol ed o ganic ca bon elease by phy oplank on by an icyclonic mesoscale eddies in he sub opical NE A lan ic Ocean. Biogeosciences 2013, 10, 2129–2143. 9. Sang à, P.; Pascual, A.; Rod íguez-San ana, A.; Machín, F.; Mason, E.; McWilliams, J.C.; Peleg í, J.L.; Dong, C.; Rubio, A.; A ís egui, J.; e al. The Cana y Eddy Co ido : A majo pa hway o long-li ed eddies in he sub opical No h A lan ic. Deep-Sea Res. I 2009, 56, 2100–2114. 10. Qua ly, G.; Chen, G. In oduc ion o he Special Issue on “Sa elli e Al ime y: New Senso s and New Applica ions”. Senso s 2006, 6, 616–619. 11. Liu, Y.; Dong, C.M.; Guan, Y.; Chen, D.; McWilliams, J.; Nencioli, F. Eddy analysis in he sub opical zonal band o he No h Paci ic Ocean. Deep-Sea Res. I 2012, 68, 54–67. 12. Xiu, P.; Chai, F.; Xue, H.; Shi, L.; Chao, Y. Modeling he mesoscale eddy ield in he Gul o Alaska. Deep-Sea Res. I 2012, 63, 102–117. 13. Ruiz, S.; Peleg i, J.L.; Emeliano , M.; Pascual A.; Mason, E. Geos ophic and ageos ophic ci cula ion o a shallow an icyclonic eddy o Cape Bojado . J. Geophys. Res.-Oceans 2014, 119, 1257–1270. 14. Mason, E.; Pascual, A.; McWilliams, J.C. A New Sea Su ace Heigh -Based Code o Oceanic Mesoscale Eddy T acking. J. A mos. Ocean. Technol. 2014, 31, 1181–1188. 15. Loh enz, S.E.; Cai, W.-J.; Chen, X.; Tuel, M. Sa elli e Assessmen o Bio-Op ical P ope ies o No he n Gul o Mexico Coas al Wa e s Following Hu icanes Ka ina and Ri a. Senso s 2008, 8, 4135–4150. 16. Shapi o, G.I.; S anichny, S.V.; S anychna, R.R. Ana omy o shel -deep sea exchanges by a mesoscale eddy in he No h Wes Black Sea as de i ed om emo ely sensed da a. Remo e Sens. En i on. 2010, 114, 867–875. 17. Wa anabe, E. Beau o shel b eak eddies and shel -basin exchange o Paci ic summe wa e in he wes e n A c ic Ocean de ec ed by sa elli e and modeling analyses. J. Geophys. Res.-Oceans 2011, 116, C08034. 18. Alpe s, W.; B and , P.; Laza , A.; Dago ne, D.; Sow, B.; Faye, S.; Hansen, M.W.; Rubino, A.; Poulain, P.M.; B ehme , P. A small-scale oceanic eddy o he coas o Wes A ica s udied by mul i-senso sa elli e and su ace d i e da a. Remo e Sens. En i on. 2013, 129, 132–143. 19. Dong, C.; Nencioli, F.; Liu, Y.; McWilliams, J.C. An Au oma ed App oach o De ec Oceanic Eddies om Sa elli e Remo e Sensed Sea Su ace Tempe a u e Da a. IEEE Geosci. Remo e Sens. Le . 2011, 8, 1055–1059. 20. Eugenio, F.; Ma in, J.; Ma cello, J.; F aile-Nuez, E. En i onmen al moni o ing o El Hie o Island subma ine olcano, by combining low and high esolu ion sa elli e image y. In . J. Appl. Ea h Obs. Geoin . 2014, 29, 53–66. Senso s 2015, 15 8748 21. F aile-Nuez, E.; González-Dá ila, M.; San ana-Casiano, J.M.; A ís egui, J.; Alonso-González, I.J.; He nández-León, S.; Blanco, M.J.; Rod íguez-San ana, A.; He nández-Gue a, A.; Gelado-Caballe o, M.D.; e al. The Subma ine Volcano E up ion a he Island o El Hie o: Physical-chemical Pe u ba ion and Biological Response. Sci. Rep. 2012, doi:10.1038/s ep00486. 22. Yi, J.; Du, Y.; He, Z.; Zhou, C. Enhancing he accu acy o au oma ic eddy de ec ion and he capabili y o ecognizing he mul i-co e s uc u es om maps o sea le el anomaly. Ocean Sci. 2014, 10, 39–48. 23. Van a am, S.R.; Sabe , E. Su ey o con empo a y ends in colo image segmen a ion. J. Elec on. Imaging 2012, 21, 9011–90128. 24. Thaku , P.; Madaan, N. A su ey o Image Segmen a ion Techniques. In . J. Res. Compu . Appl. Robo . 2014, 2, 158–165. 25. Aga wal, M.; Singh, V. A Me hodological Su ey and P oposed Algo i hm on Image Segmen a ion using Gene ic Algo i hm. In . J. Compu . Appl. 2013, 67, 7–17. 26. Dey, V.; Zhang, Y.; Zhong, M. A e iew o image segmen a ion echniques wi h emo e sensing pe spec i e. In P oceedings o he ISPRS TC VII Symposium–100 Yea s ISPRS, Vienna, Aus ia, 5–7 July 2010. 27. Na khede, H.P. Re iew o Image Segmen a ion Techniques. In . J. Sci. Mod. Eng. 2013, 1, 54–61. 28. Zhang, Y.; Liu, X.; Yin, Y.; Wang, M.; Qin, B. A simple op ical model o es ima e di use a enua ion coe icien o pho osyn he ically ac i e adia ion in an ex emely u bid lake om su ace e lec ance. Op . Exp ess 2012, 20, 18. 29. Acke mann, S.; F ey, R.; S abala, K.; Liu, Y.; ; Gumley, L.; Baum, B.; Menzel, P.H. Disc imina ing clea -sky o m cloud wi h MODIS. In NASA_MODIS Algo i hm Theo e ical Basis Documen (MOD35); Uni e si y o Wisconsin: Madison, WI, USA, 2010; Ve sion 6.1. 30. Ma cello, J.; Ma qués, F.; Eugenio, F. Au oma ic ool o he p ecise de ec ion o upwelling and ilamen s in emo e sensing image y. IEEE T ans. Geosci. Remo e Sens. 2005, 43, 1605–1616. 31. Ma cello, J.; Eugenio, F.; Ma qués, F. E alua ion o h esholding echniques applied o oceanog aphic emo e sensing image y. In P oceedings o he SPIE-Remo e Sensing Eu ope, Maspalomas, Spain, 10 No embe 2004. 32. Soille, P. Mo phological Image Analysis: P inciples and Applica ions; Sp inge -Ve lag: Be lin, Ge many, 2003. 33. AVISO. A ailable online: h p://www.a iso.al ime y. /en/home.h ml (accessed on 3 Ma ch 2014). 34. Dong, C.; McWilliams, J.C.; Shchepe kin, A. Island wakes in deep wa e . J. Phys. Oceanog . 2007, 37, 962–998. 35. Koumou sakos, P. In iscid Axisymme iza ion o an Ellip ical Vo ex. J. Compu . Phys. 1997, 138, 821–857. © 2015 by he au ho s; licensee MDPI, Basel, Swi ze land. This a icle is an open access a icle dis ibu ed unde he e ms and condi ions o he C ea i e Commons A ibu ion license (h p://c ea i ecommons.o g/licenses/by/4.0/).