Do climate teleconnections modulate wildfire-prone conditions over the Iberian Peninsula?

LETTER • OPEN ACCESS Do clima e eleconnec ions modula e wild i e-p one condi ions o e he Ibe ian Peninsula? To ci e his a icle: Ma cos Rod igues e al 2021 En i on. Res. Le . 16 044050 View he a icle online o upda es and enhancemen s. This con en was downloaded om IP add ess 155.210.59.216 on 12/04/2021 a 08:58 En i on. Res. Le . 16 (2021) 044050 h ps://doi.o g/10.1088/1748-9326/abe25d OPEN ACCESS RECEIVED 25 June 2020 REVISED 21 Janua y 2021 ACCEPTED FOR PUBLICATION 2 Feb ua y 2021 PUBLISHED 6 Ap il 2021 O iginal con en om his wo k may be used unde he e ms o he C ea i e Commons A ibu ion 4.0 licence. Any u he dis ibu ion o his wo k mus main ain a ibu ion o he au ho (s) and he i le o he wo k, jou nal ci a ion and DOI. LETTER Do clima e eleconnec ions modula e wild i e-p one condi ions o e he Ibe ian Peninsula? Ma cos Rod igues1,2,8,∗, Dhais Peña-Angulo3, Ana Russo4, Ma ía Z´ uñiga-An ón5 and Ad i´ an Ca dil2,6,7 1Depa men o Ag icul u al and Fo es enginee ing, Uni e si y o Lleida, Lleida, Spain 2Join Resea ch Uni CTFC—AGROTECNIO—CERCA, Solsona, Spain 3Ins i u o Pi enaico de Ecología, Consejo Supe io de In es igaciones Cien í icas (IPE–CSIC), Za agoza, Spain 4Ins i u o Dom Luiz (IDL), Faculdade de Ciˆ encias, Uni e sidade de Lisboa, Campo G ande, 1749–016 Lisboa, Po ugal 5GEOT G oup, IUCA, Depa men o Geog aphy and Te i o ial Planning, Uni e si y o Za agoza, Za agoza, Spain 6Depa men o C op and Fo es Sciences, Uni e si y o Lleida, Lleida, Spain 7Technosyl a Inc, La Jolla, CA, Uni ed S a es o Ame ica 8GEOFOREST G oup, IUCA, Depa men o Geog aphy and Te i o ial Planning, Uni e si y o Za agoza, Za agoza, Spain ∗ Au ho o whom any co espondence should be add essed. E-mail: ma cos. od [email p o ec ed] Keywo ds: wild i es, clima e eleconnec ions, i e dange , bu ned a ea, Ibe ian Peninsula Supplemen a y ma e ial o his a icle is a ailable online Abs ac Clima e eleconnec ions (CT) synch onize and in luence wea he ea u es such as empe a u e, p ecipi a ion and, subsequen ly, d ough and uel mois u e in many egions ac oss he globe. CTs hus may be ela ed o cycles in wild i e ac i i y, and he eby help i e manage s o an icipa e i e-p one wea he condi ions as well as en isaging hei u u e e olu ion. A wide numbe o CTs modula e wea he in he Ibe ian Peninsula (IP), exe ing di e en le els o in luence a di e en spa ial and seasonal scales on a wide ange o wea he ac o s. In his wo k, we in es iga ed he link be ween he mos ele an CT pa e ns in he IP and i e ac i i y and dange , explo ing di e en spa ial and empo al scales o agg ega ion. We analyzed a pe iod o 36 yea s (1980–2015) using his o ical eco ds o i e e en s (>100 ha bu ned) and he Canadian Fi e Wea he Index (FWI). C oss-co ela ion analysis was pe o med on mon hly ime se ies o CTs and i e da a. Resul s poin ed owa ds he No h A lan ic Oscilla ion (in he wes e n hal o he IP) and Medi e anean Oscilla ion Index (along he Medi e anean coas ) as he key CTs boos ing bu ned a ea (BA) and i e wea he dange in he IP. Bo h CTs ela e o he ela i e posi ion o he Azo ean an iclone, os e ing haza dous i e wea he condi ions du ing hei posi i e phases, i.e. low ain all and wa m empe a u e leading o low uel mois u e con en . The Scandina ian pa e n index also played an impo an ole in he wes e n hal o he Peninsula, linked o a dec ease in ain all du ing i s nega i e phases. None heless, he associa ion be ween he CTs and BA (up o 0.5 Pea son’s R p< 0.05) was weake han he obse ed be ween CTs and FWI (up o 0.75 Pea son’s R p < 0.05). 1. In oduc ion Modes o sea su ace empe a u e (SST) and associ- a ed clima e eleconnec ions (CT) a e known o in lu- ence and synch onize wea he om sub-con inen al o local scales [1], including empe a u e, p ecipi a- ion and, subsequen ly, d ough [2–4]. These wea he ea u es ha e a s ong impac on he occu ence and beha io o wild i es, by al e ing uel mois u e and igni ion p obabili y o igge ing ligh ning i e e en s [5–9]. The ole o clima e- ela ed ac o s is expec ed o s eng hen in he u u e, se ing a con ex in which CTs a e expec ed o ampli y hei e ec s [10–12]. Se e al au ho s ha e conduc ed esea ch in o CT pa e ns and hei ela ion wi h clima ic a i- ables a di e en empo al and spa ial scales wo ld- wide [13–16] and, speci ically, in he Ibe ian Pen- insula (IP [17–21]); one o he mos i e-a ec ed egions in Medi e anean Eu ope [22]. A wide ange o CTs in luence wea he ac o s in he Wes e n © 2021 The Au ho (s). Published by IOP Publishing L d En i on. Res. Le . 16 (2021) 044050 M Rod igues e al Medi e anean basin [23], namely he No h A lan ic Oscilla ion (NAO), he A lan ic Mul idecadal Oscil- la ion (AMO), he Eas A lan ic (EA) eleconnec- ion pa e n index, he ‘Niño’ 3.4 SST index (ENSO), he Medi e anean Oscilla ion (MOI), he Paci ic Decadal Oscilla ion (PDO), he Scandina ian pa - e n (SCAND) and he Wes e n Medi e anean oscil- la ion (WeMOi). Fo ins ance, he annual amoun o ain all was signi ican ly linked o he WeMOi, EA, NAO, ENSO, whe eas blocking episodes in he IP we e linked o SCAND pa e ns [24]. The sea- sonal and spa ial a iabili y in mean empe a u e was also associa ed o he di e en phases o CTs. The NAO and EA pa e ns we e posi i ely co ela ed wi h annual mean empe a u e in he IP. In u n, he co - ela ion be ween WeMOi and empe a u e depic s a g adien , being nega i e in he no h bu posi i e o e he Medi e anean coas [18]. Mo eo e , he AMO has been epo ed o con ol he a iabili y in sum- me leng h o e Eu ope, pa icula ly a e 1979 [25]. Ex ensi e esea ch has been de o ed o un a el- ling he link be ween CTs and i e ac i i y, showing how changes in wea he media ed by CTs can di ec ly impac bu ned a ea (BA) in many egions ac oss he wo ld [23,26–34]. The analysis o he e ec s o CTs on wea he , i e dange and ac i i y is no s aigh o - wa d [35] as hey a e non-s a iona y and associa ions may a y o e space a sub-con inen al scales. Like- wise, me hodological app oaches a y depending on he pu sued goals. Co ela ion and eg ession analysis using ime se ies [23,33–35] and - es [36] we e e- quen ly applied when dealing wi h ‘local’ spa ial pa - e ns, while mo e sophis ica ed echniques (wa ele cohe ence o supe posed epoch analysis) ha e been used o p o ide in-dep h insigh s a sub-con inen al scales [32,37–39]. Some s udies op ed by analyzing he in luence o clima e on i es based on ci cula ion- wea he pa e ns [5,6,40] o local-scale wea he o cing condi ions [41]. Recen ad ances in Global Clima e Models (GCMs) o e an al e na i e o s a - is ical app oaches [42] hough uncommon in he li - e a u e [38,39]. Howe e , aw GCM’s ou pu s canno be di ec ly used o quan i a i e wild i e e alua ions. Sys ema ic biases o he models om o he obse ed clima e leads o signi ican de iancies o i s s a is ical p ope ies [43]. Fu he mo e, hei coa se spa ial es- olu ion is usually no sui able o add essing he local i e-wea he condi ions [42]. Howe e , s udies add essing CTs and wild i es in he Medi e anean Eu ope a e sca ce, e en hough he unde s anding o he clima e- i e dependen ela- ionships would allow an icipa ing i e dange , hus imp o ing p e en ion and p epa edness (e.g. s a- egically a ange supp ession esou ces o educe uel loads). Mul iple ac o s may exe an in luence in i e ac i i y and i e wea he . As a consequence, he IP is qui e di e se in e ms o i e egimes and i s unde lying d i e s, which a y ac oss ime and space [44–46]. A ecen wo k by [23] epo ed no iceable co ela ion be ween SCAND, NAO and MOI/WeMOI in Spain and Po ugal, hough hei CT- ela ed ana- lyses we e conduc ed a he na ional (coun ywide) and yea ly scales in di e en ime ames (s a - ing in he 70s un il 2017). None heless, he e is e idence o CT in e ac ions modula ing and exace - ba ing wea he a iabili y [32,36,37], subsequen ly in luencing i e ac i i y [47]. Howe e , he e is no a clea pic u e o he e ec s o CT on i e ac i i y in he IP, and speci ically on he coupled e ec s o CTs, on i e ac i i y a mul iple empo al scales o agg ega ion. The e o e, wi h his pape we in end o ill he e e ed lacunae, aiming a iden i ying he in luence o he mos signi ican SST modes and CTs, indi idu- ally o coupled, a ec ing i e dange and ac i i y in he IP. We de eloped a no el amewo k o un a el he spa ial pa e ns o he mos in luen ial CTs while iden i ying syne gies and empo al scales modula - ing hei signal. We u he ex end o e o me ana- lyses by digging in o he spa ial dis ibu ion o he links be ween wild i es and CTs, p o iding a ans- bounda y assessmen using homogeneous i e egime zones, deepe insigh s in o mon hly and seasonal associa ions, and inco po a ing wild i e dange a ing indices in o he assessmen . Speci ically, we add ess he ollowing esea ch ques ions: (a) how do CTs ela e o me eo ological i e-wea he a ing indices and i e ac i i y? (b) do CT-wild i e ela ionships a y o e space? (c) a e he e coupled e ec s among CTs leaning owa ds a empo al window wi h highe i e dange and/o ac i i y? (d) is he e ec o CTs indices on i e dange and i e ac i i y lagged in anyway? 2. Ma e ials and me hods 2.1. Da a 2.1.1. Fi e da a His o ical i e da a was e ie ed om he Spanish EGIF (‘Es adís ica Gene al de Incendios Fo es ales’) [48] and Po uguese DECIF (‘Disposi i o Especial de Comba e a Incˆ endios Flo es ais’) [49] o icial i e da abases. We compiled a ha monized da ase ga h- e ing i e eco ds om bo h da abases. Acco ding o Pe ei a e al [49], i es below he 100 ha h eshold we e no compiled consis en ly du ing he 80s. To ensu e he homogenei y (missing eco ds, lack o in o ma ion, e c) in he inal i e da ase , we selec- ed only hose i e e en s la ge han 100 ha in he pe iod 1980–2015. Addi ional in o ma ion including he s a ing loca ion (a NUTS3 le el) and he da e o igni ion was also e ie ed om i e da abases o enable u he analyses. Fi es we e o ganized acco ding o hei py o e- gion o o igin (see igu e 1) as ou lined by T igo e al [44], who de ined ou mac o egions wi h homo- geneous i e egimes (N: No h, NW: No hwes , SW: Sou hwes and E: Eas ). Then, i e e en s we e agg eg- a ed on a mon hly basis in e ms o BA. 2 En i on. Res. Le . 16 (2021) 044050 M Rod igues e al Figu e 1. Desc ip ion o he s udy egion. (A) Spa ial dis ibu ion o numbe o i e e en s and no malized bu ned a ea (bu ned a ea di ided by egion size) a NUTS3 le el, 1980–2015; (B) physical map o he Ibe ian Peninsula. 2.1.2. Fi e wea he dange We used he Canadian Fi e Wea he Index (FWI) as an indica o o i e-conduci e wea he condi- ions. The FWI [50] is a nume ical a ing sys em ha summa izes he chances o a i e o igni e and p opaga e, widely used o moni o and o esee haza dous condi ions (e.g. he Eu opean Fo es Fi e In o ma ion Sys em). I is based on a combina ion o componen s buil -up om aw wea he in o ma ion such as ela i e humidi y, accumula ed ain all in he las 24 h, empe a u e and wind speed. The FWI da a we e gene a ed by he Eu opean Cen e o Medium- Range Wea he Fo ecas s om he ERA5 eanalysis da ase [51]. Daily g ids o FWI a 0.25◦ esolu ion we e e ie ed om he Cope nicus Clima e Sys em (C3S), la e agg ega ed in o mon hly ime se ies as he mean FWI pe mon h and g id cell. No e ha he g id cell se ed as spa ial uni o analysis in FWI analyses whe eas BA was add essed a egion le el. 2.1.3. Clima e eleconnec ions We chose he mos impo an CT pa e ns (a mon hly le el) in he Wes e n Medi e anean Basin based on p e ious s udies ([2,24,44,49,50]; able S1 (a ailable online a s acks.iop.o g/ERL/16/ 044050/mmedia)): NAO, AMO, EA, ENSO (El Niño), MOI, PDO, SCAND, and WeMOi. The NAO pa e n has been quan i ied by means o he dipole-like s anda dized su ace p essu e di e ence be ween he No h A lan ic Sub opical High (whose da a can be aken om Pon a Delgada- Azo es, Lisbon o Gib al a ) and a second s a ion close o he Icelandic Low (Reykja ik) [52]. The pos- i i e phase in ol es a subs an ial de elopmen o a low p essu e cen e in Iceland and a high p essu e cen e o e he Azo es, which is associa ed o below- no mal ain all o e sou he n Eu ope. The oppos- i e pa e n is ound du ing nega i e phases, when he Icelandic Low and he Azo es High a e weake han usual. The EA is s uc u ally simila o NAO. I consis s o a no h–sou h dipole o anomaly cen- e s spanning he No h A lan ic om eas o wes [53]. The posi i e phase o EA is associa ed wi h a wa m condi ions in Wes e n Eu ope and below- a e age p ecipi a ion ac oss sou he n Eu ope. The AMO pa e n is de ined as a de ended a e age o 3 En i on. Res. Le . 16 (2021) 044050 M Rod igues e al SST anomalies in he No h A lan ic [54]. The pos- i i e phase o he AMO is associa ed wi h inc eased mean su ace ai empe a u e, especially p onounced o e No h Ame ica and Eu ope du ing summe [55]. The SCAND consis s o a p ima y ci cula ion cen e o e Scandina ia, wi h weake cen e s o opposi e sign o e wes e n Eu ope and eas e n Russia/wes e n Mongolia [54]. The posi i e phase o SCAND is associa ed wi h abo e-a e age p ecipi a ion ac oss cen al and sou he n Eu ope. The MOI was de ined in o de o explain opposing a mosphe ic dynam- ics be ween he wes e n and eas e n sec o s o he Medi e anean basin. The MOI is ma ked as he no - malized p essu e di e ence be ween Algie s and El Cai o [53]. A second e sion o his index can be calcula ed as he di e ence o s anda dized p essu e anomalies a Gib al a and he Is aeli me eo ological s a ion o Lod. The WeMOi pa e n is calcula ed as a s anda dized p essu e di e ence be ween no he n I aly (Padua) and sou he n Spain (San Fe nando). The posi i e phase is cha ac e ized by he an icyc- lone o e he Azo es and a low-p essu e cen e o e he Gul o Genoa; he nega i e phase comp ises he Cen al Eu opean an icyclone loca ed in he no h o I aly, and low-p essu es in he Gul o Cadiz [24]. The ENSO/‘Niño’ 3.4 SST index was calcula ed om he HadISST1 as he a e age o SST anomalies o e he egion om 5◦S–5◦N o 170–120◦W [56]. The PDO is mos equen ly e e ed o as a long-li ed El-Niño like pa e n o he Paci ic clima e a iabili y. The PDO pa e n is de i ed as he leading p incipal componen (PC) o mon hly SST anomalies in he No h Paci ic Ocean, polewa d o 20◦N [57]. 2.2. S a is ical analyses 2.2.1. PC analysis PC analysis (PCA) o mon hly ime se ies o CTs was applied o add ess po en ial synch onic e ec s o in e ac ions be ween hem, and o build a smal- le subse o CT indices. PC we e selec ed acco ding o he Kaise c i e ion, i.e. e aining hose PCs wi h s anda d de ia ion highe han 1 [56]. The con ibu- ion o each CT o each PC was add essed ia a - imax o a ion [58]. PCA loadings and sco es we e subsequen ly compa ed o mon hly ime se ies o BA, analyzing each egion sepa a ely. 2.2.2. C oss-co ela ion ime se ies analysis The co e o ou s a is ical analysis is based on he calcula ion o Pea son’s c oss-co ela ion R coe i- cien s a mul iple empo al and spa ial scales. Pai ed mon hly ime se ies o BA, CTs and FWI we e explo ed o in es iga e he lagged associa ion o BA and CTs, and FWI and CTs, espec i ely. Fou di - e en iming windows o FWI and BA we e in es ig- a ed (Ap il–June, AMJ; May–July, MJJ; June–Augus , JJA; and July–Sep embe , JAS) o accoun o po en- ial di e ences be ween ea ly o la e summe condi- ions ( igu e 2). Se e al lag in e als (i.e. synch ony, Figu e 2. Illus a ion o empo al scales o agg ega ion. O ange indica es he empo al window o FWI o BA (JJA in he example). (A) 3 mon h agg ega ion (scale) o CTs, wi h lag 0 (synch onic o FWI/BA); (B) 4 mon h scale, lag 0; (C) 5 mon hs scale, lag −1; (D) 6 mon h scale, lag −2. om 0 o 6 mon hs be o e) and scales o agg ega ion ( om 3 o 6 a e aged mon hs) o CTs wi h BA and CTs we e analyzed (see igu e 2). S a iona i y in he ime se ies was es ed be o e he co ela ion analysis was pe o med using he Aug- men ed Dickey–Fulle es [57]. Those se ies ound non-s a iona y we e de ended by sub ac ing he end componen a e applying a Seasonal-T end decomposi ion [59]. Co ela ions be ween BA and coupled CTs we e add essed a egional le el, calcu- la ing he co ela ion be ween 3 mon h a e aged BA ( o accoun o a season-like ime span) and PCA’s sco es. BA assessmen s we e conduc ed in he ou i e egions ( igu e 1) sepa a ely. Co ela ion ou pu s o BA we e o ganized based on he combina ion o egion and ime window, epo ing he mos co - ela ed lag and scale ( able 2). Co ela ions be ween FWI and indi idual CTs we e calcula ed a FWI pixel le el (0.25◦). Co ela ions we e summa ized as a se o maps epo ing he mos and second mos co ela ed CT pe pixel ( igu e 4) and i s co esponding ime window, lag and scale o agg ega ion ( igu e 5). 3. Resul s 3.1. Associa ion be ween CT pa e ns and BA We e ained ou PCs (labeled om PC1 o PC4) acco ding o he Kaise c i e ion, explaining 72% o he o iginal a iance. Each PC depic ed a speci ic pa e n o associa ion wi h CT eleconnec ion pa - e ns ( igu e 3; see h ps://c s-ba-ip.ne li y.app [60] o addi ional ou pu s) ha was cha ac e ized acco d- ing o he loading con ibu ion o each indi idual CT p ojec ed o each componen ( able 1). Below we desc ibe he ou main componen s a ending o he mos ele an CTs: •PC1: synch onic MOI, NAO (bo h show a close loading alue and sign, −0.54) and WeMOi (−0.38) opposed o he AMO phase (0.30). The highes he PC’s sco e he lowe he associa ion wi h MOI/NAO/WeMOi. Likewise, high sco es poin owa ds a low AMO index. 4 En i on. Res. Le . 16 (2021) 044050 M Rod igues e al Figu e 3. In e ac ion be ween CTs (PC1 and PC2) and seasonal dis ibu ion o bu ned a ea in he Ibe ian Peninsula (1980–2015). Colo indica es season whe eas size ela e o bu ned a ea. Each do ep esen s he BA in an indi idual mon h. Table 1. Summa y o p incipal componen analysis. Sd, s anda d de ia ion; Exp. Va , explained a iance; Cum. Va , accumula ed a iance. Bold ace indica es signi ican loadings in Va imax o a ion. Loadings highe han 0.4 we e highligh ed in g ay. Summa y Loadings Sd Exp. a Cum. a AMO EA MOI NAO ENSO PDO SCAND WeMOi PC1 1.36 0.23 0.23 0.30 −0.23 −0.54 −0.54 −0.20 −0.23 0.20 −0.38 PC2 1.30 0.21 0.44 −0.47 −0.31 −0.22 −0.25 0.42 0.61 0.15 −0.02 PC3 1.06 0.14 0.58 0.36 0.56 −0.07 0.04 0.58 0.15 −0.06 −0.44 PC4 1.05 0.14 0.72 −0.16 −0.30 −0.04 0.15 −0.08 0.01 −0.79 −0.47 •PC2: his PC summa izes he in luence o PDO and ENSO (loadings 0.61 and 0.42, espec i ely) opposed o AMO (−0.47). The highes he sco e he highes he PDO/ENSO. High sco es ela e o −AMO phase. •PC3 ep esen s synch onic (simila loading alue) ENSO/EA/AMO opposed o WeMOi. High sco es o he PC indica e high ENSO/EA indices; low sco es depic ed low WEMOI. High alues ela e o +AMO phase •PC4: conjunc ion o SCAND (−0.79) and WeMOi (−0.47) phases. The highes he PC’s sco e he low- es SCAND/WeMOi index alue. PCA e ealed in e es ing pa e ns o associa ion be ween CTs and he seasonal dis ibu ion o BA size ( igu e 3; see igu e S1 in supplemen a y ma e i- als o addi ional biplo combina ions). MOI, NAO, WeMOi and, o a lesse ex en EA, we e posi i ely linked o summe BA (mainly in PC1), i.e. he main i e season in he IP. Likewise, AMO showed an in e se associa ion, hough sligh ly linked o BA du ing all. None heless, spa ial and empo al di e ences we e obse ed a egion le el since p essu e cen e s a ec di e en ly he di e en a eas o he IP in e ms o p ecipi a ion and he e o e ha is e lec ed in e ms o d ough and ege a ion de elopmen ( able 2). PC1 (MOI/NAO/WeMOi) showed signi ican associa ions wi h BA in all egions, excep in he SW a ea. Posi - i e phases o ei he NAO o MOI os e ed inc eased BA. This associa ion was obse ed consis en ly a mul iple empo al scales hough i was mo e in ense a 4–6 mon hs (lag) be o e June. The conjunc ion o posi i e ENSO and PDO (PC2) was ela ed o an inc eased BA in ea ly and la e summe in he N (MJJ and JAS) and, wi h he main i e season in he SW (JJA). An immedia e associa ion (lag 0) was epo - ed in he N whe eas he signal was sligh ly delayed in he SE (3 mon hs lag). Synch onic (lag 0) low WeMOi index unde posi i e phases o EA and ENSO (PC3) was associa ed wi h a la ge BA in he E egion. The wes e n hal o he peninsula (NW and SW) showed con as ing associa ions be ween ea ly and la e sum- me . BA in he MJJ pe iod beha ed like he E egion while he s onges associa ion in JJA and JAS depic- ed in e se co ela ions in he NW, hus BA co el- a ion was nega i e ENSO and EA phases and posi - i e wi h WEMOI. Finally, PC4 (SCAND +WEMOI) showed nega i e sho - e m (lag 0 and scale 3) co el- a ions wi h BA in la e summe in all egions excep in he SW a ea. Con e sely, high WEMOI and SCAND du ing he win e p io summe was associa ed wi h an inc eased BA in AMJ. 3.2. The in luence o CTs on i e wea he dange In o de o comple e he analysis o his o ical i e incidence we in es iga ed he po en ial associa ion be ween CTs and FWI ( igu e 4; see h ps://c s- wi- ip.ne li y.app [61] o see he indi idual maps o CTs’ co ela ion pa e ns). The mos in luen ial CTs we e MOI, SCAND, NAO, PDO and AMO. MOI, NAO 5 En i on. Res. Le . 16 (2021) 044050 M Rod igues e al Table 2. Co ela ion coe icien s pe egion, ime window, lag and scale be ween bu ned a ea and CTs. Columns heade s deno e he empo al window o bu ned a ea agg ega ion: AMJ, Ap il–June; MJJ, May–July; JJA, June–Augus ; JAS, June–Sep embe . E, Eas egion; N, no h egion; NW, no hwes egion; SW, sou hwes egion. Bold alues indica e signi ican co ela ions (p< 0.05). N indica es he numbe o i es analyzed. AMJ MJJ JJA JAS Region RLag Scale N R Lag Scale N R Lag Scale N R Lag Scale N PC1 E0.49 3 5 213 −0.41 4 3 640 −0.45 5 3 1092 −0.48 6 3 1118 N−0.56 6 3 312 −0.42 6 4 258 −0.37 1 5 752 −0.34 3 3 1384 NW −0.50 3 5 247 −0.28 4 3 896 −0.33 3 3 2956 0.42 6 5 3917 MOI/WeMOi and NAO SW 0.20 6 4 422 0.28 5 5 1745 −0.26 4 5 3665 −0.21 5 3 4231 PC2 E−0.34 6 5 213 0.24 6 5 640 −0.28 5 3 1092 0.32 0 5 1118 N0.38 0 3 312 0.43 1 3 258 −0.31 5 4 752 0.43 0 3 1384 NW −0.26 6 3 247 −0.19 4 4 896 0.33 2 4 2956 0.23 6 5 3917 PDO and ENSO SW −0.31 1 3 422 −0.22 4 3 1745 0.37 3 3 3665 0.32 4 3 4231 PC3 E0.39 6 5 213 0.19 0 4 640 0.36 0 3 1092 0.40 1 3 1118 N 0.32 5 4 312 −0.28 5 3 258 0.31 2 3 752 −0.20 4 3 1384 NW 0.30 4 5 247 0.47 0 5 896 −0.41 0 3 2956 −0.42 1 3 3917 ENSO. EA and AMO SW −0.23 1 4 422 0.38 1 3 1745 −0.27 6 3 3665 0.23 4 4 4231 PC4 E 0.26 6 5 213 −0.34 6 4 640 −0.24 0 3 1092 −0.39 0 3 1118 N0.41 5 3 312 0.40 6 5 258 −0.24 1 3 752 −0.45 0 4 1384 NW 0.32 2 4 247 0.28 4 3 896 −0.29 2 5 2956 −0.38 6 3 3917 SCAND and WeMOi SW 0.59 5 5 422 −0.27 0 4 1745 −0.41 6 5 3665 −0.31 5 5 4231 6 En i on. Res. Le . 16 (2021) 044050 M Rod igues e al Figu e 4. Spa ial dis ibu ion o mos in luencing CTs on FWI. Top mos (A) and second mos (B) co ela ed eleconnec ion. Bo om, Pea son’s R co ela ion coe icien o mos (C) and second mos (D) co ela ed CT. All co ela ion alues we e signi ican a p< 0.05. and AMO displayed a posi i e ela ionship, i.e. he highe he index he highe he i e wea he dange . These CTs ela e o sus ained an icyclone condi ions du ing hei posi i e phases. Con e sely, SCAND and PDO showed an in e se associa ion pa e n, possibly linked o he lack o p ecipi a ion du ing hei neg- a i e phases. The spa ial dis ibu ion o he mos co - ela ed CTs e ealed a dissimila pa e n ac oss he IP. The MOI domina es he Medi e anean coas (E) and he eas e n hal o he SW egion (R> 0.6, p< 0.05), in e mingling wi h smalle encla es o AMO in he sou h and SCAND in he no heas . The seasonal e olu ion o hese CTs, app aised in he supplemen - a y ma e ials, e eals a s ong and pe sis en in lu- ence o AMO om June o Sep embe , wi h SCAND being s onge du ing he cen al summe mon hs (JJA). In he wes e n hal o he Peninsula (SW and NW), NAO (R> 0.5, p< 0.05), PDO and SCAND (R<−0.6, p< 0.05) modes we e he mos co ela ed. The i s was main ope a ing in ea ly summe (MJJ a scale 3) whe eas he la e exe ed inc easing in lu- ence om ea ly o la e summe , especially he PDO, which a ec s JJA and JAS he mos (see h ps://c s- wi-ip.ne li y.app). 7 En i on. Res. Le . 16 (2021) 044050 M Rod igues e al Figu e 5. Spa ial dis ibu ion o empo al pa ame e s o mos in luencing CTs on FWI. Top pe iod wi h highes co ela ion o he mos (A) and second mos (B) co ela ed eleconnec ion. Mid, scale (numbe o a e aged mon hs) wi h highes co ela ion o he mos (C) and second mos (D) co ela ed CT. Bo om, lag (mon hs delay) wi h highes co ela ion o hemos (E) and second mos (F) co ela ed CT. All co ela ion alues we e signi ican a p< 0.05. 8