scieee Science in your language
[en] (orig)

Repositorio Institucional de Documentos

Abstract

El fuego ha coexistido de forma intrínseca en diversos ecosistemas a nivel global. En el caso de los ambientes más humanizados la acción del hombre ha alterado esos regímenes de incendio naturales por uno fundamentalmente de carácter antrópico. En el contexto de la Europa Mediterránea, el número de incendios forestales y su área quemada observados han experimentado un descenso general durante el final del siglo XX. Esto ha supuesto un declive de la incidencia del fuego en la mayoría de los ecosistemas mediterráneos históricamente afectados por incendios recurrentes. Por tanto, es evidente la alteración de los regímenes de incendio pasados, debido principalmente a la intervención humana con una política de exclusión total del fuego muy exigente.<br />No obstante, la evolución reciente de los regímenes de incendio presenta una alta variabilidad espacial y temporal. Por otro lado, las perspectivas de futuro vaticinan un impacto creciente del factor humano (abandono del campo, gestión de los bosques y mantenimiento de la supresión excluyente), lo que consecuentemente derivará una mayor actividad de incendios debido a una mayor cantidad de combustible disponible. Asimismo, se prevén unas condiciones climáticas cada vez más propensas a generar incendios de gran superficie (mayores valores de temperatura, mayor frecuencia de olas de calor y sequías), lo que sin duda afectará negativamente tanto a los ecosistemas como las sociedades futuras.<br />Todos estos factores hacen necesaria una adecuada zonificación de los regímenes de incendio desde una perspectiva espacio-temporal, la cual permita conocer la relación existente entre el régimen de incendios alterado y los factores socio-económicos y ambientales asociados. Así como detectar tendencias en el tiempo en regiones que experimenten un descenso de la actividad, o, por el contrario, incremento de la incidencia de incendios. Por tanto, conociendo estas zonas se podrá mejorar la gestión y prevención contra incendios forestales.<br />Esta tesis doctoral se enfoca en enriquecer el conocimiento sobre la identificación e interpretación de regiones homogéneas de regímenes de incendio. Para ello se recurre a un amplio abanico de métodos de análisis estadísticos y de modelado espacial. La tesis se estructura de acuerdo a los siguientes objetivos: el objetivo 1 se centra en analizar la distribución espacio-temporal de las principales métricas que definen el régimen de incendio durante el periodo reciente. El objetivo 2 pretende profundizar en la influencia del riesgo meteorológico en la evolución de la actividad de los incendios. El objetivo 3 evalúa el cambio de la contribución relativa de los factores antropogénicos en los incendios forestales. El objetivo 4 se enfoca en explicar la evolución y causas de los cambios o transiciones de los regímenes de incendios durante el periodo reciente (1974-2015) y futuro (2016-2036). Finalmente, el objetivo 5 pone la atención en la traslación de la zonificación de tipologías de regímenes de incendios hacia una cartografía integral de piroregiones.<br />Los resultados indican que los regímenes de incendio en la España peninsular han experimentado diversos cambios, principalmente una disminución considerable de la actividad de incendios en la mayor parte del territorio, aunque todavía persiste una alta actividad en el extremo norte (especialmente en invierno). Los diversos métodos de aprendizaje automático empleados, especialmente Random Forest, han demostrado su potencial en términos de revelar los factores que impulsan la evolución del régimen de incendios. Además, la proyección ARIMA ha confirmado la tendencia actual hacia una menor incidencia de incendios. Todo apunta a que las medidas preventivas deben tomar más protagonismo en áreas con un abrupto descenso de la ocurrencia, ya que son significativamente más propensas a grandes incendios a corto y medio plazo.<br /> Fire has always been an intrinsic feature in various ecosystems around the world. In environments heavily populated by humans, their actions have altered these natural fire regimes for others that are fundamentally anthropogenic in nature. In the context of Mediterranean Europe, the number of forest fires and their observed burnt area fell into a general decline during the late twentieth century, which led to a reduced incidence of fire in most Mediterranean ecosystems historically affected by recurrent fires. Therefore, the change in past fire regimes is evident, mainly due to human intervention instigating a very demanding policy of total exclusion of fire. However, the recent evolution of fire regimes presents a high spatial and temporal variability. On the other hand, future scenarios predict a growing impact of the human factor (more land abandonment, poor management of forests and adhering exclusively to suppression methods), which will result in increased fire activity due to a greater amount of available fuel. In addition, climatic conditions are expected to cause increasingly larger burned areas (higher temperatures, more frequent heat waves and droughts), which will undoubtedly have a negative effect on both ecosystems and future societies. All these factors make an adequate zoning of fire regimes necessary from a spatial-temporal perspective, which allows the relationship between the altered fire regime and associated socio-economic and environmental factors to be determined, as well as detecting temporal trends in regions with decreasing activity, or on the contrary, an increase in the incidence of fires. Therefore, finding these areas will lead to improved management and prevention of forest fires. This doctoral thesis focuses on enriching knowledge for identifying and interpreting homogeneous regions of fire regimes. A wide range of methods of statistical analysis and spatial modeling are employed. The thesis is structured according to the following objectives: Objective 1 focuses on analyzing the spatial-temporal distribution of the main features defining the fire regime during the recent period. Objective 2 aims to further describe the influence of meteorological danger on the evolution of fire activity. Objective 3 evaluates the change in the relative contribution of anthropogenic factors on forest fires. Objective 4 focuses on explaining the evolution and causes of changes or transitions in fire regimes during the recent (1974-2015) and future (2016-2036) periods. Finally, Objective 5 centers on the transfer of the zoning of fire regime typologies into an integral mapping of pyroregions The results indicate that fire regimes in mainland Spain have undergone several changes, mainly a considerable decrease in fire activity in most of the territory, although it still remains high in the north (especially in winter). The diverse machine-learning methods employed, especially Random Forest, have demonstrated their potential in terms of revealing the fire drivers behind fire regime evolution. Moreover, forecasting by the ARIMA model has confirmed the ongoing tendency towards a lower incidence of fire. All indications are that preventive measures should take greater prominence in areas with an abrupt decrease in wildfires, as they are significantly more prone to large ones in the short and medium term.<br /> Jiménez Ruano, Adrián; Rodrigues Mimbrero, Marcos; De La Riva Fernández, Juan

Read accessible full text

Repositorio Institucional de Documentos

Publisher: Universidad de Zaragoza, Prensas de la Universidad
Year: 2019
Source: https://zaguan.unizar.es/record/99374/files/TESIS-2021-054.pdf
2021
54
Ad ián Jiménez Ruano
Cha ac e izing py o egions in
mainland Spain om spa ial-
empo al pa e ns o i e egime
and hei unde lying d i e s
Depa amen o
Di ec o /es
Geog a ía y O denación del Te i o io
Rod igues Mimb e o, Ma cos
De la Ri a Fe nández, Juan
© Uni e sidad de Za agoza
Se icio de Publicaciones
ISSN 2254-7606
Ad ián Jiménez Ruano
CHARACTERIZING PYROREGIONS IN MAINLAND
SPAIN FROM SPATIAL-TEMPORAL PATTERNS OF
FIRE REGIME AND THEIR UNDERLYING DRIVERS
Di ec o /es
Geog a ía y O denación del Te i o io
Rod igues Mimb e o, Ma cos
De la Ri a Fe nández, Juan
Tesis Doc o al
Au o
2019
Reposi o io de la Uni e sidad de Za agoza – Zaguan h p://zaguan.uniza .es
UNIVERSIDAD DE ZARAGOZA

Tesis Doc o al
Cha ac e izing py o egions in mainland Spain om
spa ial- empo al pa e ns o i e egime and hei
unde lying d i e s
Au o
Ad ián Jiménez Ruano
Di ec o /es
de la Ri a Fe nández, Juan
Rod igues Mimb e o, Ma cos
Geog a ía y O denación del Te i o io
2019
El au o de es a esis doc o al dis u ó pa a su desa ollo, así como pa a la es ancia en un cen o de
in es igación ex anje o, de la inanciación del P og ama de ayudas FPU del Minis e io de Educación, Cul u a y
Depo e. Re e encia de la ayuda FPU 13/06618. Desea hace cons a , po an o, su ag adecimien o.
Jiménez-Ruano A, Rod igues M, Jolly W M, de la Ri a Fe nández J. (2018). The ole o d ough and
magni ude in he empo al e olu ion o i e occu ence and bu ned a ea size in mainland Spain.
Geophysical Resea ch Abs ac s
(Pos e con ibu ion). Vol. 20, EGU2018-13520, Vienna, Aus ia.
In his wo k, he PdD s uden , Ad ián Jiménez Ruano, is esponsible o mos o he wo k, ha ing
ca ied ou he s a is ical and spa ial analysis, and being he main esponsible o elabo a ing he
pos e . D . Ma cos Rod igues and D . Juan de la Ri a ha e collabo a ed e y closely in he concep ion
o he me hodology and in he e iew o he pos e . D . Ma Jolly has also con ibu ed o he
me hodology design as well as o he e iew o he esul s.

La p esen e esis doc o al se ha elabo ado siguiendo la modalidad de compendio de publicaciones. El
doc o ando, Ad ián Jiménez Ruano, igu a como p ime au o y esponsable de casi odos los a ículos.
Seguidamen e se de allan los abajos que cons i uyen el cue po de la esis, su ac o de impac o y el de alle
de las a eas ealizadas po cada uno de los au o es en cada uno de ellos:
Jiménez-Ruano A, Rod igues M, de la Ri a J (2017) Unde s anding wild i es in mainland Spain. A
comp ehensi e analysis o i e egime ea u es in a clima e-human con ex .
Applied Geog aphy
89:100-111. h ps://doi.o g/10.1016/j.apgeog.2017.10.007
Fac o de impac o JCR: 3,117 (1e cua il, “Geog aphy”).
En es e abajo el doc o ando, Ad ián Jiménez Ruano, es esponsable de la mayo pa e del abajo,
desa ollando g an pa e del análisis es adís ico y espacial, siendo el esponsable úl imo de la edacción
de los con enidos. El D . Ma cos Rod igues Mimb e o ha colabo ado en el desa ollo de algunas
a eas y ambién ayudó en el p oceso de esc i u a. Tan o el D . Juan de la Ri a and Ma cos Rod igues
Mimb e o igu an como coau o es en calidad de di ec o es de esis, siendo esponsables del ema
gene al de la in es igación y habiendo colabo ado en la e isión de los esul ados.
Jiménez-Ruano A, Rod igues M, de la Ri a J (2017) Explo ing spa ial– empo al dynamics o i e egime
ea u es in mainland Spain.
Na u al Haza ds and Ea h Sys em Sciences
17:1697-1711.
h ps://doi.o g/10.5194/nhess-17-1697-2017
Fac o de impac o JCR: 2,281 (2º cua il, “Geosciences, Mul idisciplina y”).
En es e abajo el doc o ando, Ad ián Jiménez Ruano, es el esponsable de la mayo pa e del abajo,
desa ollando g an pa e del análisis es adís ico y espacial, siendo el esponsable úl imo de la edacción
de los con enidos. El D . Ma cos Rod igues Mimb e o y el D . Juan de la Ri a ambién han
colabo ado muy es echamen e en la e isión de la me odología y los esul ados.
Jiménez-Ruano A, Rod igues M, Jolly W.M, de la Ri a J (2019) The ole o sho - e m wea he condi ions
in empo al dynamics o i e egime ea u es in mainland Spain.
Jou nal o En i onmen al
Managemen
241:575-586. h ps://doi.o g/10.1016/j.jen man.2018.09.107
Fac o de Impac o JCR: 4,865 (1e cua il, “En i onmen al Sciences”).
En es e abajo el doc o ando, Ad ián Jiménez Ruano, es el esponsable de la mayo pa e del abajo,
desa ollando g an pa e del análisis es adís ico y espacial, siendo el esponsable úl imo de la edacción
de los con enidos. El D . Ma Jolly igu a como coau o po su colabo ación en la cons ucción de
pa e de la me odología y e isión de los esul ados p elimina es siendo además el esponsable de la
es ancia en la que se desa olló la in es igación. El D . Ma cos Rod igues Mimb e o y el D . Juan de
la Ri a ambién han colabo ado muy es echamen e en la e isión de los esul ados.
Rod igues M, Jiménez-Ruano A, de la Ri a J. (2016) Analysis o ecen spa ial– empo al e olu ion o
human d i ing ac o s o wild i es in Spain.
Na u al Haza ds
84(3):2049-2070.
h ps://doi.o g/10.1007/s11069-016-2533-4
Fac o de impac o JCR: 1,833 (2º cua il, “Geosciences, Mul idisciplina y”).
En es e abajo, el D . Ma cos Rod igues es esponsable de la mayo pa e del abajo, habiendo
ealizado el análisis es adís ico y espacial, y siendo el p incipal esponsable de la edacción de odos
los con enidos. El doc o ando Ad ián Jiménez Ruano ha colabo ado en la ejecución de pa e de la
me odología. El D . Juan de la Ri a es el esponsable del ema gene al de la in es igación y ha
colabo ado en la e isión de los esul ados.
Rod igues M, Jiménez-Ruano A, Peña-Angulo D, de la Ri a J. (2018) A comp ehensi e spa ial- empo al
analysis o d i ing ac o s o human-caused wild i es in Spain using Geog aphically Weigh ed Logis ic
Reg ession.
Jou nal o En i onmen al Managemen
225: 177-192.
h ps://doi.o g/10.1016/J.JENVMAN.2018.07.098
Fac o de impac o JCR: 4,865 (1º cua il, “En i onmen al Sciences”).
En es e abajo, el D . Ma cos Rod igues es esponsable de la mayo pa e del abajo, habiendo
ealizado el análisis es adís ico y espacial, y siendo el p incipal esponsable de la edacción de odos
los con enidos. El doc o ando, Ad ián Jiménez Ruano, ha colabo ado en la ealización de pa e de la
me odología y pa e de la ca og a ía del abajo inal. La D . Peña-Angulo ha p opo cionado los
da os climá icos empleados en la in es igación. El D . Juan de la Ri a es el esponsable del ema
gene al de la in es igación y ha colabo ado en la e isión de los esul ados.
Además, se han incluido o os abajos en el cue po de la in es igación, pe o que aún no han sido
publicados:
Rod igues M, Jiménez-Ruano A, de la Ri a J (En e isión). Fi e egime dynamics in mainland Spain. Pa
1: d i e s o change.
Science o he To al En i onmen .
Fac o de impac o JCR: 5,589 (1º cua il, “En i onmen al Sciences”).
En es e abajo el D . Ma cos Rod igues Mimb e o, es el esponsable de la mayo pa e del abajo,
desa ollando g an pa e del análisis es adís ico y especial, siendo el esponsable úl imo de la edacción
de los con enidos. El doc o ando, Ad ián Jiménez Ruano, ha colabo ado muy es echamen e en la
con ección del p oceso me odológico y en ob ene pa e de los esul ados, y el D . Juan de la Ri a
ambién ha pa icipado ac i amen e en la e isión de los esul ados.
Jiménez-Ruano A, de la Ri a J, Rod igues M (En e isión). Fi e egime dynamics in mainland Spain. Pa
2: a nea - u u e p ospec i e o i e ac i i y.
Science o he To al En i onmen .
Fac o de impac o JCR: 5,589 (1º cua il, “En i onmen al Sciences”)
En es e abajo el doc o ando, Ad ián Jiménez Ruano, es el esponsable de la mayo pa e del abajo,
desa ollando g an pa e del análisis es adís ico y espacial, siendo el esponsable úl imo la edacción
de los con enidos. El D . Ma cos Rod igues Mimb e o ha colabo ado muy es echamen e en la
con ección del p oceso me odológico inal y e isión del ex o, y el D . Juan de la Ri a ambién ha
pa icipado es echamen e en la e isión de los esul ados.
Po úl imo, se han incluido a ias con ibuciones de cong esos en dos apéndices especí icos:
Apéndice D
Jiménez-Ruano A, Rod igues M, de la Ri a Fe nández J. 2018. Iden i ying py o egions by means o Sel
O ganizing Maps and hie a chical clus e ing algo i hms in mainland Spain, in: Viegas, D.X. (Ed.),
Ad ances in Fo es Fi e Resea ch
(VIII In e na ional Con e ence on Fo es Fi e Resea ch).
Imp ensa da Uni e sidade de Coimb a, Coimb a, pp. 495–505.
h ps://doi.o g/h ps://doi.o g/10.14195/978-989-26-16-506_54
En es e abajo, el doc o ando Ad ián Jiménez Ruano es el esponsable de la mayo pa e del abajo,
habiendo ealizado el análisis es adís ico y espacial, y siendo el p incipal esponsable de la edacción
de odos los con enidos. El D . Ma cos Rod igues ha colabo ado muy es echamen e en la
p epa ación del p oceso me odológico y en la e isión de la edacción. El D . Juan de la Ri a ambién
ha pa icipado es echamen e en la e isión de los esul ados.
Apéndice E
Jiménez-Ruano A, Rod igues M, de la Ri a Fe nández J. (2017). An analysis o wild i e equency and
bu ned a ea ela ionships wi h human p essu e and clima e g adien s in he con ex o i e egime.
Geophysical Resea ch Abs ac s
(Pos e con ibu ion). Vol. 19 EGU2017-15084, Vienna, Aus ia.
En es e abajo, el doc o ando, Ad ián Jiménez Ruano, es el esponsable de la mayo pa e del abajo,
habiendo ealizado el análisis es adís ico y espacial, y siendo el p incipal esponsable de la elabo ación
del pós e . El D . Ma cos Rod igues ha colabo ado muy es echamen e en la concepción de la
me odología y en la e isión del diseño del pós e . El D . Juan de la Ri a ambién ha pa icipado en
la e isión de los esul ados.
Jiménez-Ruano A, Rod igues M, de la Ri a Fe nández J. (2017). Assessing he in luence o small i es on
ends in i e egime ea u es a mainland Spain.
Geophysical Resea ch Abs ac s
(Pos e
con ibu ion). Vol. 19, EGU2017-15755, Vienna, Aus ia.
En es e abajo, el doc o ando, Ad ián Jiménez Ruano, es el esponsable de la mayo pa e del abajo,
habiendo ealizado el análisis es adís ico y espacial, y siendo el p incipal esponsable de la elabo ación
del pós e . El D . Ma cos Rod igues ha colabo ado muy es echamen e en la concepción de la
me odología y en la e isión del diseño del pós e . El D . Juan de la Ri a ambién ha pa icipado en
la e isión de los esul ados.
Jiménez-Ruano A, Rod igues M, Jolly W M, de la Ri a Fe nández J. (2018). Assessing he in luence o
i e wea he dange indexes on i e equency and bu ned a ea in mainland Spain.
Geophysical
Resea ch Abs ac s
(O al p esen a ion). Vol. 20, EGU2018-13196, Vienna, Aus ia.
En es e abajo, el doc o ando, Ad ián Jiménez Ruano, es el esponsable de la mayo pa e del abajo,
habiendo ealizado el análisis es adís ico y espacial, y siendo el p incipal esponsable de la p epa ación
de la p esen ación. El D . Ma cos Rod igues y el D . Juan de la Ri a han colabo ado muy
es echamen e en la concepción de la me odología y en la e isión de la comunicación o al. El D .
Ma Jolly ambién ha con ibuido al cálculo de los índices me eo ológicos de incendios, así como a
la e isión de los esul ados.
Jiménez-Ruano A, Rod igues M, Jolly W M, de la Ri a Fe nández J. (2018). The ole o d ough and
magni ude in he empo al e olu ion o i e occu ence and bu ned a ea size in mainland Spain.
Geophysical Resea ch Abs ac s
(Pos e con ibu ion). Vol. 20, EGU2018-13520, Vienna, Aus ia.
En es e abajo, el doc o ando, Ad ián Jiménez Ruano, es el esponsable de la mayo pa e del abajo,
habiendo ealizado el análisis es adís ico y espacial, y siendo el p incipal esponsable de la elabo ación
del pós e . El D . Ma cos Rod igues y el D . Juan de la Ri a han colabo ado muy es echamen e en
la concepción de la me odología y en la e isión del pós e . El D . Ma Jolly ambién ha con ibuido
al diseño de la me odología, así como a la e isión de los esul ados.
ABSTRACT
Fi e has always been an in insic ea u e in a ious ecosys ems a ound he wo ld. In en i onmen s hea ily
popula ed by humans, hei ac ions ha e al e ed hese na u al i e egimes o o he s ha a e undamen ally
an h opogenic in na u e. In he con ex o Medi e anean Eu ope, he numbe o o es i es and hei
obse ed bu n a ea ell in o a gene al decline du ing he la e wen ie h cen u y, which led o a educed
incidence o i e in mos Medi e anean ecosys ems his o ically a ec ed by ecu en i es. The e o e, he
change in pas i e egimes is e iden , mainly due o human in e en ion ins iga ing a e y demanding policy
o o al exclusion o i e.
Howe e , he ecen e olu ion o i e egimes p esen s a high spa ial and empo al a iabili y. On he o he
hand, u u e scena ios p edic a g owing impac o he human ac o (mo e land abandonmen , poo
managemen o o es s and adhe ing exclusi ely o supp ession me hods), which will esul in inc eased i e
ac i i y due o a g ea e amoun o a ailable uel. In addi ion, clima ic condi ions a e expec ed o cause
inc easingly la ge bu ned a eas (highe empe a u es, mo e equen hea wa es and d ough s), which will
undoub edly ha e a nega i e e ec on bo h ecosys ems and u u e socie ies.
All hese ac o s make an adequa e zoning o i e egimes necessa y om a spa ial- empo al pe spec i e,
which allows he ela ionship be ween he al e ed i e egime and associa ed socio-economic and
en i onmen al ac o s o be de e mined, as well as de ec ing empo al ends in egions wi h dec easing
ac i i y, o on he con a y, an inc ease in he incidence o i es. The e o e, inding hese a eas will lead o
imp o ed managemen and p e en ion o o es i es.
This doc o al disse a ion ocuses on en iching knowledge o iden i ying and in e p e ing homogeneous
egions o i e egimes. A wide ange o me hods o s a is ical analysis and spa ial modeling a e employed.
The disse a ion is s uc u ed acco ding o he ollowing objec i es: Objec i e 1 ocuses on analyzing he
spa ial- empo al dis ibu ion o he main ea u es de ining he i e egime du ing he ecen pe iod.
Objec i e 2 aims o u he desc ibe he in luence o me eo ological dange on he e olu ion o i e ac i i y.
Objec i e 3 e alua es he change in he ela i e con ibu ion o an h opogenic ac o s on o es i es.
Objec i e 4 ocuses on explaining he e olu ion and causes o changes o ansi ions in i e egimes du ing
he ecen (1974-2015) and u u e (2016-2036) pe iods. Finally, Objec i e 5 cen e s on he ans e o he
zoning o i e egime ypologies in o an in eg al mapping o py o egions
The esul s indica e ha i e egimes in mainland Spain ha e unde gone se e al changes, mainly a
conside able dec ease in i e ac i i y in mos o he e i o y, al hough i s ill emains high in he no h
(especially in win e ). The di e se machine-lea ning me hods employed, especially Random Fo es , ha e
demons a ed hei po en ial in e ms o e ealing he i e d i e s behind i e egime e olu ion. Mo eo e ,
o ecas ing by he ARIMA model has con i med he ongoing endency owa ds a lowe incidence o i e.
All indica ions a e ha p e en i e measu es should ake g ea e p ominence in a eas wi h an ab up dec ease
in wild i es, as hey a e signi ican ly mo e p one o la ge ones in he sho and medium e m.

RESUMEN
El uego ha coexis ido de o ma in ínseca en di e sos ecosis emas a ni el global. En el caso de los
ambien es más humanizados la acción del homb e ha al e ado esos egímenes de incendio na u ales po
uno undamen almen e de ca ác e an ópico. En el con ex o de la Eu opa Medi e ánea, el núme o de
incendios o es ales y su á ea quemada obse ados han expe imen ado un descenso gene al du an e el inal
del siglo XX. Es o ha supues o un decli e de la incidencia del uego en la mayo ía de los ecosis emas
medi e áneos his ó icamen e a ec ados po incendios ecu en es. Po an o, es e iden e la al e ación de
los egímenes de incendio pasados, debido p incipalmen e a la in e ención humana con una polí ica de
exclusión o al del uego muy exigen e.
No obs an e, la e olución ecien e de los egímenes de incendio p esen a una al a a iabilidad espacial y
empo al. Po o o lado, las pe spec i as de u u o a icinan un impac o c ecien e del ac o humano
(abandono del campo, ges ión de los bosques y man enimien o de la sup esión excluyen e), lo que
consecuen emen e de i a á una mayo ac i idad de incendios debido a una mayo can idad de combus ible
disponible. Asimismo, se p e én unas condiciones climá icas cada ez más p opensas a gene a incendios
de g an supe icie (mayo es alo es de empe a u a, mayo ecuencia de olas de calo y sequías), lo que sin
duda a ec a á nega i amen e an o a los ecosis emas como las sociedades u u as.
Todos es os ac o es hacen necesa ia una adecuada zoni icación de los egímenes de incendio desde una
pe spec i a espacio- empo al, la cual pe mi a conoce la elación exis en e en e el égimen de incendios
al e ado y los ac o es socio-económicos y ambien ales asociados. Así como de ec a endencias en el
iempo en egiones que expe imen en un descenso de la ac i idad, o, po el con a io, inc emen o de la
incidencia de incendios. Po an o, conociendo es as zonas se pod á mejo a la ges ión y p e ención con a
incendios o es ales.
Es a esis doc o al se en oca en en iquece el conocimien o sob e la iden i icación e in e p e ación de
egiones homogéneas de egímenes de incendio. Pa a ello se ecu e a un amplio abanico de mé odos de
análisis es adís icos y de modelado espacial. La esis se es uc u a de acue do a los siguien es obje i os: el
obje i o 1 se cen a en analiza la dis ibución espacio- empo al de las p incipales mé icas que de inen el
égimen de incendio du an e el pe iodo ecien e. El obje i o 2 p e ende p o undiza en la in luencia del
iesgo me eo ológico en la e olución de la ac i idad de los incendios. El obje i o 3 e alúa el cambio de la
con ibución ela i a de los ac o es an opogénicos en los incendios o es ales. El obje i o 4 se en oca en
explica la e olución y causas de los cambios o ansiciones de los egímenes de incendios du an e el pe iodo
ecien e (1974-2015) y u u o (2016-2036). Finalmen e, el obje i o 5 pone la a ención en la aslación de la
zoni icación de ipologías de egímenes de incendios hacia una ca og a ía in eg al de pi o egiones.
Los esul ados indican que los egímenes de incendio en la España peninsula han expe imen ado di e sos
cambios, p incipalmen e una disminución conside able de la ac i idad de incendios en la mayo pa e del
e i o io, aunque oda ía pe sis e una al a ac i idad en el ex emo no e (especialmen e en in ie no). Los
di e sos mé odos de ap endizaje au omá ico empleados, especialmen e Random Fo es , han demos ado su
po encial en é minos de e ela los ac o es que impulsan la e olución del égimen de incendios. Además,
la p oyección ARIMA ha con i mado la endencia ac ual hacia una meno incidencia de incendios. Todo
apun a a que las medidas p e en i as deben oma más p o agonismo en á eas con un ab up o descenso de
la ocu encia, ya que son signi ica i amen e más p opensas a g andes incendios a co o y medio plazo.
TABLE OF CONTENTS
CHAPTER 1: INTRODUCTION ........................................................................................................................ 1
1.1. The wild i e phenomenon ............................................................................................................................... 3
1.2. The concep o i e egime: de ini ions and componen s ......................................................................... 4
1.3. Me hodological app oaches in i e egime modelling ................................................................................ 5
1.4. Fi e egime s py o egion ................................................................................................................................ 7
CHAPTER 2: OBJECTIVES AND RESEARCH DESIGN ...................................................................... 9
2.1. Resea ch ques ions .......................................................................................................................................... 11
2.2. Resea ch s uc u e ........................................................................................................................................... 13
CHAPTER 3: STUDY AREA ............................................................................................................................... 15
CHAPTER 4: MATERIALS AND METHODS ........................................................................................... 21
4.1. Da ase s and sou ces ....................................................................................................................................... 23
4.1.1. The Spanish i e da abase...................................................................................................................... 23
4.1.2. Fi e da a and i e ea u es ...................................................................................................................... 24
4.1.3. Clima e and wea he ............................................................................................................................... 27
4.1.4. An h opogenic d i e s ........................................................................................................................... 29
4.2. Modelling echniques ...................................................................................................................................... 33
4.2.1. Desc ip i e and explo a i e .................................................................................................................. 33
4.2.2. Time se ies analysis ................................................................................................................................ 35
4.2.3. Classi ica ion and eg ession................................................................................................................. 39
CHAPTER 5: SPATIAL-TEMPORAL DISTRIBUTION OF FIRE REGIME FEATURES ..... 43
CHAPTER 6: THE INFLUENCE OF FIRE-WEATHER ON THE EVOLUTION OF
FIRE ACTIVITY ........................................................................................................................ 73
CHAPTER 7: CHANGE IN ANTHROPOGENIC DRIVERS .............................................................. 87
CHAPTER 8: EVOLUTION AND CAUSES OF FIRE REGIME CHANGE ............................... 127
CHAPTER 9: TRANSLATING FIRE REGIME ZONING SCHEMES INTO
PYROREGIONS ...................................................................................................................... 181
CHAPTER 10: CONCLUSIONS AND FUTURE RESEARCH .......................................................... 195
REFERENCES ....................................................................................................................................................... 207
APPENDIX A: SUPLEMENTARY MATERIAL OF FIRE REGIME FEATURES ................... 219
APPENDIX B: SUPLEMENTARY MATERIAL OF FIRE-WEATHER ........................................ 229
APPENDIX C: SUPLEMANTARY MATERIAL OF DRIVERS OF CHANGE ........................... 237
APPENDIX D: PRELIMINARY PYROREGIONS DELIMITATION ........................................... 251
APPENDIX E: CONFERENCES CONTRIBUTIONS ......................................................................... 265
Chap e 1: In oduc ion
6
speci ic-de o ed simula ion models, as change poin s (Mouillo e al., 2002) o powe law (Malamud, 1998;
Malamud e al., 2005; Pe e a and Cui, 2010). In Spain, se e al a icles sugges ha al e a ions in i e egimes
ha e been d i en by clima e, land use changes and supp ession policies (Mo eno e al., 2014) as well as
di e en p opaga ion pa e ns in Ca alonia (Duane e al., 2015).
The majo i y o s udies ha e used eg ession models in combina ion wi h simula ed da a om gene al
clima e models (GCM) (Boulange e al. 2013; DaCama a e al. 2014; Kilpeläinen e al. 2010; K awchuk e
al. 2009; Pechony and Shindell 2010; Te ie e al. 2014; Wes e ling e al. 2011) based on IPPCC p ojec ions
o u u e emission scena ios o Regional Clima e Models (RCM). Mos o hese s udies en isage an
inc easing bu ned a ea in egions such as Po ugal (DaCama a e al., 2014), Cali o nia (Wes e ling e al.,
2011) and he Ibe ian Peninsula (Sousa e al., 2015). Howe e , se e al au ho s poin ou di e en ends
depending on he egions o he wo ld (K awchuk e al., 2009; Pechony and Shindell, 2010), including
showing opposi e endencies wi h inc easing equency and a sligh decline o bu ned a ea in he No heas
o Spain (Tu co e al., 2014).
In he i e-clima e amewo k, many au ho s ha e analyzed he ela ionship be ween clima e change and
shi s in ce ain cha ac e is ics o i e egimes ( i e equency, su ace a ea, seasonali y, a e age i e ange,
maximum i e size, e c.) in many egions. Fo example, in he bo eal o es s o No h Ame ica (Kasischke
and Tu e sky, 2006) hey eso o his o ical eco ds, he analysis o indi idual yea s by ca ego ies o eco-
zones and he s a ime o indi idual e en s. In Canada, he Fi e G ow h Model has been used o model
he isk o ligh ning and human-induced igni ions (Ni schke and Innes, 2013). On he o he hand, mos o
he s udies ha e assumed u u e p ojec ions wi h simila en i onmen al and an h opic condi ions o he
cu en ones (Boulange e al., 2014, 2012), hus showing ce ain limi a ions in end de ec ion since hey
assume a “s a ic” o non-clima e condi ions o he u u e. The e o e, he g owing impo ance o es ima ing
he p esen and u u e impac o clima e change on i e egime has become a key issue in isk assessmen
and adap a ion s a egies, eme ging as he co ne s one in na ional and in e na ional clima e p og ams
(Tu co e al., 2014), such as he Eu opean p ojec FUME (2010-2013).
Howe e , i is well-known ha he democ a ic and massi e use o u u e clima e change scena ios implies
a high deg ee o unce ain y. In o he wo ds, he mos complica ed issue is he alida ion o p ojec ed da a,
especially hose by GGM o RCM models, as he e is s ill no ime se ies wi h which o co ela e. This is
why some au ho s leaned owa ds he “sa es ” al e na i es, such as he au o- eg ession and mo ing a e age
models (ARIMA). ARIMA models a e known o hei good pe o mance in ields such as ma ke s and he
economy (Loi and Ng, 2018; Ma yjaszek e al., 2019), as well as in he en i onmen al amewo k: ege a ion
(REF) o clima e change. In he con ex o o es i e, P eisle and Wes e ling (2007) employed ARIMA
using empe a u e o ecas ing o assess i e dange in wes e n USA, whe eas, Boube a e al. (2016) applied
a simpli ied e sion o ARIMA (ARMA) wi hou he in eg a ed componen in o de o p edic bu n a ea
in Galicia. The main i ue o ARIMA models lies in he ac ha hey p edic u u e ends and seasonali y,
wi h he his o ical ime se ies o da a as hei only e e ence. As a esul , he p inciple o pa simony is
gua an eed in he model, since he minimum numbe o a iables is used, being mo e easily ep oducible
and wi hou c ea ing an o e adjus men .
When discussing he use o spa ial modeling me hods and he p edic ion o o es i e cha ac e is ics in
speci ic a eas, a wide epe oi e o me hodological app oaches and explana o y a iables can be b ough o
bea . The scale o analysis (global, egional o local), he p oposed objec i es and he na u e o he da a
used will condi ion he analysis amewo k. Among he mos widely applied models o da e, GLM and
GAM (Gene alized Linea Models and Gene al Addi i e Models, espec i ely) s and ou as lexible

Chap e 1: In oduc ion
7
gene aliza ions o linea eg ession, able o deal wi h non-no mal dis ibu ions o he a iable unde s udy
( i e ac i i y) o he explana o y a iables (clima e, wea he , opog aphy, popula ion, wildland u ban-
ag icul u al in e aces, oad ne wo k, e c.). This modeling amewo k is adequa e, since o es i e da a
usually depic non-linea esponse unc ions. In addi ion, Geog aphically Weigh ed Reg ession (GWR) is a
mo e ad anced al e na i e ha has also been applied in he con ex o wild i es (Kou sias e al., 2010; Sá e
al., 2011), whose main ad an age is ha i allows he calcula ion o local eg ession pa ame e s, use ul o
analyzing he spa ial beha io o each explana o y a iable and de e mining hei le el o signi icance. The
ew wo ks conduc ed in mainland Spain poin o a ce ain deg ee o spa ial a iabili y (Ma ínez-Fe nández
e al., 2013), con i ming ha human d i ing ac o s a y o e bo h space and ime(Rod igues e al., 2016)
and a e losing explana o y powe in a o o clima ic condi ions (Rod igues e al., 2018).
Ano he impo an aspec when es ima ing he p obabili y o he occu ence o wild i es is o analyze he
cha ac e is ic o uels and how hey in e ac wi h clima ic a iables (p ecipi a ion, empe a u e, wind, ela i e
humidi y, e c.). The ole o o es uels no only la gely de e mines he likelihood o igni ion, bu also he
speed o p opaga ion, and ul ima ely he se e i y. In his espec , nume ous i e wea he dange indices
ha e been used o ela e me eo ological da a o i e (Fi e Wea he Index: FWI, S anda dized P ecipi a ion-
E apo anspi a ion Index: SPEI, Palme D ough Index: PDSI, among o he s). Some s udies ca ied ou
in Po ugal (Fe nandes e al., 2014) s essed he posi i e ela ionship be ween i e and wea he oge he
wi h uel haza d and he inal bu ned a ea. In eas e n-Spain, (Ca dil e al., 2019) poin ed ou he impo ance
o a mul i- empo al pe spec i e when s udying he link be ween d ough and bu ned a ea o di e en
ege a ion communi ies. In any case, ha he e is a s ong in luence om he lammabili y o he uel and
i s spa ial con inui y on i e equency and bu ned a ea has been demons a ed by uel model classi ica ions
(P ome heus, NFFL, NFDRS, McA hu , FBP - see A oyo, Pascual, and Manzane a (2008) o mo e
de ails.
1.4. Fi e egime s py o egion
Because o he po en ial use ulness and in e es in p edic ing how he beha io o i e egimes e ol es, his
PhD Thesis aims o op imize he iden i ica ion and cha ac e iza ion o i e egimes in mainland Spain,
beginning wi h he iden i ica ion o he mos ele an ea u es o i e, and con inuing by e alua ing he
di ec ion and ex en o egional ends bo h in space and ime. Un il now, mos wo ks ocused on b oad-
scale i e egime modeling based on la ge ecological and adminis a i e uni s. In Canada, he nex s ep was
o ou line homogeneous i e egime (HFR) zones wi hou his adi ional app oach, since i does no cap u e
he spa ial he e ogenei y o i e egimes and could lead o spa ially inaccu a e es ima ions o u u e i e
ac i i y (Boulange e al., 2014). In Spain, only a ew pape s ha e de ined i e egime uni s bu by a ibu ing
a s a ic image in hei delimi a ion (Mo eno and Chu ieco, 2013), i.e., no inco po a ing he non-s a iona y
beha io o i e ea u es. To o e come his limi a ion, ou goal is o p o ide a p ojec ion o he possible
u u e e olu ion o hese homogeneous i e egime zones, since un il now he ew p ojec s ca ied ou in
he Ibe ian Peninsula ha e ocused on only o ecas ing selec ed componen s, such as he a ec ed su ace
(Sousa e al., 2015).
Gene ally, he spa ial delimi a ion o i e egimes is based exclusi ely on he conside a ion o he main,
de ining ea u es o i e. Howe e , his zoning has o be inco po a ed in o a mo e comp ehensi e spa ial
con ex ha also in eg a es he d i ing ac o s (bo h clima ic and human) mos di ec ly ela ed o o es
i es. In his espec , he i s s udy ha pu o wa d his new concep was F éja ille & Cu (2015), which
added he concep o “py oclima es” o he wild i e li e a u e. They de eloped a new amewo k o
analyzing egional changes in i e egimes om speci ic spa ial- empo al pa e ns o o es i es and clima e,
Chap e 1: In oduc ion
8
de ining i as a geog aphical en i y displaying homogenous a ibu es wi h espec o i e egime, clima e
condi ions (bioclima ic a iables and i e dange indices) and he empo al ends o bo h. The e o e, we
adop ed pa o he inno a ion o his concep in ou e m “py o egions”, bu in ou case, we added human
ac o s in o he de ini ion and no only he clima e condi ions. As esul , we de ine py o egion as “a
geog aphical a ea sha ing homogeneous i e egime ea u es, clima e-human condi ions and he e olu ion
o bo h”.
To sum up, he main di e ence be ween “ i e egime” and “py o egion” lies undamen ally in he na u e
o hei unde lying ac o s. In he case o i e egime, i b oadly e e s o he a e age condi ions in e ms o
i e ea u es o e ime and space. The py o egion anscends i e egime being a geog aphical en i y ha
cha ac e izes by uni o m o homogeneous i e ac i i y, bu is also in luenced by sel -de ining clima ic and
human condi ions.
2
CHAPTER 2: OBJECTIVES
AND RESEARCH DESIGN
This chap e summa izes he objec i es and s uc u e o he
hesis, connec ing he o me wi h hei co esponding
publica ions and appendices ha compose he whole esea ch.
Chap e 2: Objec i es and esea ch design
11
The wo king hypo hesis o his PhD Thesis is ha mainland Spain p esen s di e en i e egimes
de ined by speci ic i e equency, bu ned a ea, seasonali y and cause, which a e non-s a iona y o e space
and ime, hus allowing modeling and en isaging hei e olu ion. To unde s and he complexi y o
he phenomenon, we had o in es iga e he d i ing o ces o i e egimes, which ul ima ely would lead o
he de ini ion o dynamic py o egions, hus imp o ing i e managemen , p e en ion and p epa edness
wi hin a con ex o clima e and socio-economic change.
The e o e, he main objec i e o his esea ch was o ansla e he a ie y o homogenous zones o i e
egimes in o py o egions, p o iding insigh s in o hei possible e olu ion h ough he iden i ica ion
and cha ac e iza ion o hei main componen s ( equency, size, seasonali y, cause, e c.) and d i ing
ac o s (clima e, wea he , human p essu e, e c.).
2.1. Resea ch ques ions
In o de o add ess he main objec i e s a ed be o e, i e speci ic esea ch ques ions (RQs) o objec i es
we e o mula ed and add essed by s udying se e al esea ch pape s. Table 1 shows he co ela ion be ween
each speci ic objec i e and i s co esponding publica ions.
RQ 1: Wha is he spa ial- empo al dis ibu ion o he main i e egime ea u es and wha is
i s ela ionships wi h clima e-human ac o s?
1s Objec i e: Explo e he spa ial- empo al dis ibu ion o i e egime ea u es and hei ela ion wi h
clima e-human ac o s.
RQ 2: Wha ole does i e-wea he dange play in he empo al e olu ion o i e egime
ea u es?
2nd Objec i e: Es ima e he con ibu ion o i e-wea he dange on he obse ed e olu ion o i e
ac i i y.
RQ 3: Wha ha e been he spa ial- empo al changes in he in luence o human ac o s on
wild i es?
3 d Objec i e: Analysis spa ial- empo al changes in he ole o an h opogenic d i e s on wild i es.
RQ 4: Wha changes ha e been expe ienced by i e egimes and which ac o s a e behind
hese dynamics?
4 h Objec i e: Cha ac e ize he dynamics o ecen - u u e i e egimes and know he d i e s o hei
changes.
RQ 5: How a e py o egions dis ibu ed in space on he basis o he obse ed e olu ion o i e
egime ypologies and d i e s?
5 h Objec i e: T ansla e he i e egime ypologies scheme in o py o egions.

Chap e 2: Objec i es and esea ch design
12
Table 1. Summa y o he speci ic objec i es and hei co esponding publica ion o con ibu ion.
Objec i e
Publica ion
1s Objec i e: Explo e
he spa ial- empo al
dis ibu ion o i e egime
ea u es and hei ela ion
wi h clima e-human
ac o s.
CHAPTER 5
-Jiménez-Ruano A,
Rod igues M, de la Ri a J (2017) Unde s anding wild i es in mainland
Spain. A comp ehensi e analysis o i e egime ea u es in a clima e-human con ex .
Applied
Geog aphy 89:100-111. h ps://doi.o g/10.1016/j.apgeog.2017.10.007
-Jiménez-Ruano A
, Rod igues M, de la Ri a J (2017) Explo ing spa ial– empo al dynamics
o i e egime ea u es in mainland Spain. Na u al Haza ds and Ea h Sys em Sciences 17:1697-
1711. h ps://doi.o g/10.5194/nhess-17-1697-2017
APENDIX A
-Supplemen a y ma e ial om “Unde s anding wild i es in mainland Spain. A comp ehensi e
analysis o i e egime ea u es in a clima e-human con ex ”.
APPENDIX E
-
Jiménez-Ruano A
, Rod igues M, de la Ri a Fe nández J. (2017). An analysis o wild i e
equency and bu ned a ea ela ionships wi h human p essu e and clima e g adien s in he con ex
o i e egime. Geophysical Resea ch Abs ac s (Pos e con ibu ion). Vol. 19 EGU2017-15084,
Vienna, Aus ia.
-
Jiménez-Ruano A
, Rod igues M, de la Ri a Fe nández J. (2017). Assessing he in luence o
small i es on ends in i e egime ea u es a mainland Spain. Geophysical Resea ch Abs ac s
(Pos e con ibu ion). Vol. 19, EGU2017-15755, Vienna, Aus ia.
2nd Objec i e: Es ima e
he con ibu ion o i e-
wea he dange on he
empo al e olu ion o i e
ac i i y.
CHAPTER 6
-Jiménez-Ruano A
, Rod igues M, Jolly W.M, de la Ri a J (In P ess) The ole o sho - e m
wea he condi ions in empo al dynamics o i e egime ea u es in mainland Spain. Jou nal o
En i onmen al Managemen 17:1697-1711. h ps://doi.o g/10.1016/j.jen man.2018.09.107
APPENDIX B
-
Supplemen a y ma e ial om: “The ole o sho - e m wea he condi ions in empo al
dynamics o i e egime ea u es in mainland Spain”
APPENDIX E
-
Jiménez-Ruano A
, Rod igues M, Jolly W M, de la Ri a Fe nández J. (2018). Assessing he
in luence o i e wea he dange indexes on i e equency and bu ned a ea in mainland Spain.
Geophysical Resea ch Abs ac s (O al p esen a ion). Vol. 20, EGU2018-13196, Vienna,
Aus ia.
-
Jiménez-Ruano A
, Rod igues M, Jolly W M, de la Ri a Fe nández J. (2018). The ole o
d ough and magni ude in he empo al e olu ion o i e occu ence and bu ned a ea size in mainland
Spain. Geophysical Resea ch Abs ac s (Pos e con ibu ion). Vol. 20, EGU2018-13520,
Vienna, Aus ia.
3 d Objec i e: Analysis
o spa ial- empo al
changes in he ole o
an h opogenic d i e s on
wild i es.
CHAPTER 7
-Rod igues M,
Jiménez-Ruano A
, de la Ri a J. (2016) Analysis o ecen spa ial– empo al
e olu ion o human d i ing ac o s o wild i es in Spain. Na u al Haza ds 84(3):2049-2070.
h ps://doi.o g/10.1007/s11069-016-2533-4
-Rod igues M,
Jiménez-Ruano A
, Peña-Angulo D, de la Ri a J. (2018) A comp ehensi e
spa ial- empo al analysis o d i ing ac o s o human-caused wild i es in Spain using Geog aphically
Chap e 2: Objec i es and esea ch design
13
2.2. Resea ch s uc u e
The con en s o he Thesis a e o ganized as ollows: Chap e 3 p esen s a desc ip ion o he s udy a ea.
Chap e 4 summa izes he da a sou ces and me hods employed in he esea ch, complemen ing he
in o ma ion al eady published. Chap e s 5 o 8 b ing oge he he o iginal e sion o accep ed and published
a icles. Las ly, he las wo chap e s (Chap e 9 and 10) po ay he inal ou line o py o egions and
summa ize he main conclusions, espec i ely. Figu e 1 summa izes he main da abases and me hodologies
employed in he in es iga ion acco ding o he i s ou speci ic objec i es.
In addi ion, a complemen a y sec ion p o ides u he in o ma ion, o ganized in o i e appendixes (A, B,
C, D and E). The i s h ee co espond o he supplemen a y ma e ial in h ee publica ions o he main
body o his hesis, appendix A belongs o he pape en i led “Unde s anding wild i es in mainland Spain. A
comp ehensi e analysis o i e egime ea u es in a clima e-human con ex ”, appendix B is pa o he a icle “The ole o
sho - e m wea he condi ions in empo al dynamics o i e egime ea u es in mainland Spain” and appendix C
co esponds o he manusc ip unde e iew “Fi e egime dynamics in mainland Spain. Pa 1: d i e s o change”.
Appendix D e e s o Chap e 3 o he book “Ad ances in Fo es Fi e Resea ch” edi ed by Domingos Xa ie
Viegas, as a esul o he con ibu ion in he “VIII In e na ional Con e ence on Fo es Fi e Resea ch”, held in he
ci y o Coimb a (Po ugal) om 9 o 16 No embe 2018. The la e appendix includes se e al abs ac s
om di e en con e ence con ibu ions held in EGU 2017 and EGU 2018.
Weigh ed Logis ic Reg ession. Jou nal o En i onmen al Managemen 225: 177-192.
h ps://doi.o g/10.1016/J.JENVMAN.2018.07.098
4 h Objec i e:
Cha ac e ize he dynamics
o ecen - u u e i e
egimes and know he
d i e s o hei changes.
CHAPTER 8
-Rod igues M,
Jiménez-Ruano A
, de la Ri a J. (In p ess). Fi e egime dynamics in mainland
in Spain. Pa 1: d i e s o change. Science o he To al En i onmen .
-Jiménez-Ruano A
, de la Ri a J, Rod igues M. (In p ess). Fi e egime dynamics in mainland
Spain. Pa 2: a nea - u u e p ospec i e o i e ac i i y. Science o he To al En i onmen .
APPENDIX C
-
Supplemen a y ma e ial om: “Fi e egime dynamics in mainland in Spain. Pa 1: d i e s
o change. Science o he To al En i onmen ”.
5 h Objec i e: T ansla e
he i e egime ypologies
scheme in o py o egions.
CHAPTER 9
-Jiménez-Ruano A
, Rod igues M, de la Ri a J. ( o be submi ed) Mapping ecen py o egions
on he basis o spa ial- empo al pa e ns o i e egimes and en i onmen al-human da ase s in
mainland Spain
APPENDIX D
-
Jiménez-Ruano A,
Rod igues M, de la Ri a J. (2018) Iden i ying py o egions by means o
Sel O ganizing Maps and hie a chical clus e ing algo i hms in mainland Spain. in: Viegas, D.X.
(Ed.), Ad ances in Fo es Fi e Resea ch (VIII In e na ional Con e ence on Fo es Fi e Resea ch).
Imp ensa da Uni e sidade de Coimb a, Coimb a, pp. 495–505.
h ps://doi.o g/h ps://doi.o g/10.14195/978-989-26-16-506_54
Chap e 2: Objec i es and esea ch design
14
Figu e 1 Concep ual wo k low o he hesis acco ding o he i s ou speci ic objec i es.
3
CHAPTER 3: STUDY AREA
This chap e p esen s a desc ip ion o he s udy a ea whe e he
hesis has had i s spa ial amewo k.

Chap e 4: Ma e ials and me hods
23
4.1. Da ase s and sou ces
4.1.1. The Spanish i e da abase
The Gene al S a is ics o Wild i es (Es adís ica Gene al de Incendios Fo es ales: EGIF) da abase s ands
ou o i s p ecision and comple eness, being one o he oldes wild i e da abases in Eu ope, beginning in
1968 (Mo eno e al., 2011; Vélez, 2001). I s incep ion coincided wi h he adop ion in he same yea o Law
81/1968 on Fo es Fi es, he i s legal manda e exp essly designed o add ess a se ious p oblem. by means
o p e en ion and con ol ac ions (López San alla e al., 2017). The Bu eau o De ense Agains Fo es Fi es
(Á ea de De ensa Con a Incendios Fo es ales: ADCIF) is he ins i u ion esponsible o s anda dizing,
main aining, d a ing and publishing hese s a is ics, based on he in o ma ion submi ed by au onomous
communi ies o e e y i e occu ing in he coun y. All he baseline in o ma ion collec ed is o ganized in
di e en sec ions in he Spanish Fo es Fi e Repo s (Pa e de Incendio Fo es al: PIF), which cu en ly
collec s mo e han 150 da a ields o each i e. I should be no ed ha his s uc u e, sec ions and ype o
in o ma ion ga he ed has a ied o e he yea s, unde going a o al o eigh modi ica ions om i s i s
publica ion.
Sys ema ic collec ion o s a is ical da a on o es i es began in 1956. Un il hen hey we e collec ed manually
and on an i egula basis by he p o incial se ices. In 1967, he Calcula ion O ice o he Ins i u e o
Fo es y Resea ch and Expe iences acqui ed a compu e , which enabled a new model o PIF o be c ea ed
ha came in o ope a ion in he second semes e o ha yea . The e o e, he i s Annual Fo es Fi e Repo
was published in 1968, bu included da a on i es ha had occu ed since 1961. Wi h ega d o he quali y
o he da a, i should be no ed ha , in he ea ly yea s, i only included i es ha a ec ed o es masses o
la ge non- o es a eas, al hough subsequen ly, he es o he i es we e aken in o accoun , e en hose o
less han 1 ha.
The spa ializa ion o in o ma ion has changed o e he yea s om i s beginning, when he minimum spa ial
uni o e e ence was he p o ince (NUTS3), wi h a 10 x 10 km e e ence g id adop ed a e 1974. Un il
1979, only hose i es occu ing in public and e o es a ion o es s we e eco ded. La e , in he pe iod 1980-
1988, all i e e en s we e collec ed, ega dless o owne ship. Since 1982, he municipali y was added as a
ield in each i e loca ion. La e , in 1989-1992, he PIF was e o med o inco po a e impo an ields such
as ime, use o ai , means o mo i a ions ela ed o in en ionali y. Since 1990, he Gene al S a is ic has been
submi ed o he Eu opean Commission o in eg a ion in o he Communi y da abase EFFIS (Eu opean
Fo es Fi e In o ma ion Sys em).
On he o he hand, he adi ional dema ca ion by egion employed in he annual i e epo s was he same
o he pe iod 1968-1977 wi h a o al 7 egions (excluding he Balea ic and Cana y Islands): Galicia, No h,
No heas , Eb o, Le an e, he Hin e land and Andalusia. Since 1978, he numbe o egions inc eased o
10 (excluding he Cana y Islands): Galicia, No h o Can ab ic (As u ias and Can ab ia), Eb o (A agón),
No heas (Ca alonia-Balea es), Due o (Cas illa-León), Cen e (Cas illa–La Mancha), Le an e (Valencia-
Mu cia), Ex emadu a, Wes Andalusia and Eas Andalusia. Mo eo e , om 1982 a mo e ex ensi e sec ion
e e ing o wea he condi ions h oughou he pa icula yea was added, p o ided by he Na ional Ins i u e
o Me eo ology. F om 1983, he p e ious egions we e eplaced by he Au onomous Communi ies. The
sec ions o he cu en PIF con ain he ollowing common in o ma ion:
a) Loca ion da a: Includes he ID o he i e (IDPIF) as 10 digi s. The codes o he au onomous
communi y, p o ince, municipali y con aining he i e igni ion poin (c ea ed in 1983), ile and g id
(c ea ed in 1974) and UTM coo dina es.
Chap e 4: Ma e ials and me hods
24
b) Time da a: Day, mon h, yea , hou and minu es when he i e was de ec ed, bu also i s a i al
o engines by land (c ea ed in 1988), i s a i al o i e- igh ing ai c a (c ea ed in 1989), i s
ai bo ne b igade a i al (c ea ed in 2005), and ime when he i e was con olled and ex inguished.
c) De ec ion: Who i s de ec ed he i e (pe manen gua d, o es y o ice , ai c a , e c.) and place
o o igin ( oad, pa h, house, ain ail, c ops, e c.).
d) Igni ion causes: Di e ences be ween known and supposed cause (since 1998), ligh ning,
negligence and acciden al causes, a son (c ea ed in 1989), unknown cause, ekindled i e (since
1998), iden i ica ion o o he wise o he pe son causing i , and ype o day ( es i al, Sa u day, es i al
e e and wo king day).
e) Dange condi ions when he i e s a s: Me eo ological da a (days om las ain, maximum
empe a u e, ela i e humidi y, wind) uel model (since 1989) and p obabili y o igni ion.
) Type o i e: Su ace, c own o subsoil (since 1989).
g) Fi e supp ession media: Type o land anspo ( ehicles, helicop e s), numbe o di e en
pe sonnel ( echnical s a , o es y agen s, p o essional i e igh e s, ci il s a , a my, e c.) and
ex inguishing me hods (ai c a s, helicop e s, e a dan s, e c.).
h) Fi e supp ession echniques: Di ec o indi ec a acks, i ewall opening, e c.
i) Losses: People killed and inju ed, ci il p o ec ion inciden s, ype o su ace a ec ed, en i onmen al
impac s.
4.1.2. Fi e da a and i e ea u es
Fi e ea u es we e e ie ed om he Gene al Wild i es S a is ics (EGIF) da abase. Gene ally, i e eco ds
o 1974-2015 we e selec ed and spa ialized acco ding o he 10 x 10 km UTM e e ence g id which is used
by i e igh ing c ews o app oxima e loca ions o i e igni ion poin s. Fi e coun da a, o al bu ned a ea
size, igni ion igge ing da e and i e cause we e e ie ed o each e en . In all cases, only in o ma ion on
i es la ge han 1 ha was e ained because small i es (i.e. i es wi h less han 1 ha a ec ed) we e no ully
compiled un il 1988. This is a well-known issue a ec ing o he egions in he Medi e anean, such as
Po ugal (Pe ei a e al., 2011). Addi ionally, i is impo an o emembe ha in he au onomous communi y
o Na a e, i e da a we e only a ailable om 1988. Hence, all he analyses conduc ed in Na a e we e
based on a sligh ly di e en s udy pe iod ( om 1988 o 2010, 2013 o 2015).
The s a yea was se as 1974, since i was he i s yea o use he 10 x 10 km g id. P io o ha ime, i e
da a we e only eco ded a p o ince le el, so g id in o ma ion was no a ailable. The end yea luc ua es
depending on he empo al ame o o he da abases equi ed o di e en analyses. Fo he i s objec i e,
he end yea (2010) was chosen because o he a ailabili y o clima e da a om he MOTEDAS and
MOPREDAS da ase s (desc ibed below). Fo he second objec i e, he inal yea was se a 2013, because
he sole inpu was he EGIF da abase, and a he ime o he esea ch, i e da a was only a ailable un il hen.
As s a ed in sec ion 1.2., egions we e ou lined ollowing MAGRAMA speci ica ions. In u n, wo i e
seasons we e de ined acco ding o Mo eno e al. (2014). Thus, annual da a we e di ided in o a sp ing-
summe season (S), om Ap il o Sep embe ; and an au umn-win e season (W) om Oc obe o Ma ch.
F om all a ailable i e da a in o ma ion, se e al i e egime ea u es we e cons uc ed sepa a ely o he
season, egion, NUTS3 and g id le el. The inal numbe o i e ea u es changes acco ding o each speci ic
objec i e (see Table 2).
Chap e 4: Ma e ials and me hods
25
Table 2. Summa y o i e egime ea u es cons uc ed, hei desc ip ion and co esponding ime pe iod o each
speci ic objec i e.
Objec i e
Fi e Fea u e
Desc ip ion
Time
Pe iod
1s . Explo e he spa ial- empo al
dis ibu ion o i e egime ea u es and
hei ela ion wi h clima e-human
ac o s
Fi e equency
(F)
To al numbe o i es, ega dless
o size o igni ion sou ce
1974-
2010
1974-
2013
Bu ned a ea (B)
To al i e a ec ed a ea, ega dless
o size o igni ion sou ce
Numbe o la ge
i es (N500)
Numbe o i es abo e 500 ha
bu ned, ega dless o igni ion
sou ce
Bu ned a ea om
la ge i es (B500)
O e all a ec ed a ea om i es
abo e 500 ha, ega dless o
igni ion sou ce
Numbe o
na u al i es (NL)
Numbe o i es igge ed by
ligh ning
Bu ned a ea om
na u al i es (BL)
O e all bu ned a ea om i es
igge ed by ligh ning
Numbe o
human i es
(NH)
Numbe o i es igge ed by an
an h opogenic sou ce
Bu ned a ea om
human i es (BH)
O e all bu ned a ea om i es
igge ed by an an h opogenic
sou ce
2nd. Es ima e he con ibu ion o i e-
wea he dange on he empo al
e olu ion o i e ac i i y.
Fi e equency
(F)
To al numbe o i es, ega dless
o size o igni ion sou ce
1979-
2013
Bu ned a ea (B)
To al i e a ec ed a ea, ega dless
o size o igni ion sou ce
3 d. Analysis o spa ial- empo al
changes in he ole o an h opogenic
d i e s on wild i es.
Fi e coun s
Numbe o i es by g id
1988-
2010
1988-
2013
Fi e p esence o
absence
Recoded in o a bina y p esence o
absence o i e eco ded
25 subse s o
occu ence
Combina ion o wo pe iods, wo
seasons, wo causes and h ee i e
sizes.
4 h. Cha ac e ize he dynamics o
ecen - u u e i e egimes and know he
d i e s o hei changes.
Fi e equency
(F)
To al numbe o i es, ega dless
o size o igni ion sou ce
1974-
2015
Bu ned a ea (BA)
To al i e a ec ed a ea, ega dless
o size o igni ion sou ce
Bu ned a ea om
na u al i es
(BAL)
O e all bu ned a ea om i es
igge ed by ligh ning
Chap e 4: Ma e ials and me hods
26
Bu ned a ea om
la ge i es
(BA100)
O e all a ec ed a ea om i es
abo e 100 ha, ega dless o
igni ion sou ce
Win e i e
equency (FW)
Numbe o i es occu ed in
au umn-win e season, (W)
ega dless o size o igni ion
sou ce
In o al, he calcula ion o i e occu ence (a o al o 229,068 i es in he pe iod 1974-2015, excluding small
i es – i.e. less han 1 ha) was cons uc ed by he me hod de eloped by De la Ri a e al. (2004). This me hod
consis s in spa ializing i e da a as an inpu o i e modeling by using a ke nel app oach o in e pola e
his o ic i e obse a ions. In e ms o mon hly mean and o al alues o he main i e ea u es, a double
annual peak can be ound ( he highes is usually ound in Augus , wi h a second one in Ma ch, see Table
3). Howe e , he second peak disappea s o he a ea bu ned by la ge i es (>100 ha) and hose caused by
ligh ning, he la e showing a displacemen o he summe peak o July.
Table 3. Summa y o mon hly mean, s anda d de ia ion (sd) and o al numbe o i es and bu ned a ea o each
i e ea u e in he pe iod 1974-2015 (small i es less han 1 ha a e excluded).
Fi e ea u e
Mon h
Mean
Sd
To al
Fi e equency
Janua y
0.05
188.33
7,199
Feb ua y
0.13
479.82
18,691
Ma ch
0.24
754.77
33,329
Ap il
0.12
428.94
17,311
May
0.06
135.09
7,921
June
0.07
163.06
9,892
July
0.17
387.86
24,226
Augus
0.34
675.74
47,657
Sep embe
0.29
808.65
40,296
Oc obe
0.09
398.99
13,160
No embe
0.03
103.22
3,979
Decembe
0.04
167.94
5,407
Bu ned a ea
Janua y
2,477.85
3,278.01
104,069.51
Feb ua y
6,133.04
8,977.37
257,587.55
Ma ch
11,321.46
11,995.31
475,501.25
Ap il
6,277.16
7,694.76
263,640.90
May
2,985.79
4,329.85
125,403.29
June
5,778.87
9,068.18
242,712.68
July
28,265.16
37,616.06
1,187,136.62
Augus
44,804.75
34,226.26
1,881,799.50
Sep embe
25,413.33
35,333.18
1,067,360.03
Oc obe
7,011.48
12,163.58
294,482.01
No embe
1,865.65
3,731.95
78,357.49
Decembe
3,024.84
5,501.75
127,043.30
La ge bu ned a ea (> 100 has)
Janua y
1,004.03
1,775.17
42,169.42
Feb ua y
2,349.64
4,567.64
98,684.74
Ma ch
3,859.71
4,371.90
162,107.73
Chap e 4: Ma e ials and me hods
27
Ap il
2,623.81
4,282.10
110,199.93
May
1,520.87
3,105.52
63,876.42
June
4,033.21
8,588.28
169,394.96
July
23,179.12
35,556.11
973,523.08
Augus
34,094.29
27,675.81
1,431,960.15
Sep embe
16,189
25,104.63
679,938.17
Oc obe
4,087.58
7,829.04
171,678.17
No embe
1,071.2
3,344.38
44,990.17
Decembe
1,795.32
3,886.1
75,403.49
Na u al bu ned a ea
Janua y
0.39
1.51
16.50
Feb ua y
5.01
22.38
210.30
Ma ch
38.91
226.98
1,634.38
Ap il
20.30
53.86
852.43
May
49.33
93.17
2,071.79
June
593.73
1,194.33
24,936.83
July
5,772.42
16,780.78
242,441.65
Augus
2,469.47
4,578.96
103,717.77
Sep embe
523.36
1,081.92
21,980.98
Oc obe
16.06
46.77
674.57
No embe
5.89
30.26
247.40
Decembe
0.22
0.76
9.30
On he o he hand, se e al explana o y a iables can be aken in o accoun when add essing a i e egime
cha ac e iza ion. These a e usually di ided in o wo g oups: na u al and human. In he i s case, ac o s
ela ed o en i onmen al condi ions we e selec ed o ep esen he gene al clima e g adien s and i e-
wea he . The second g oup e e s o an h opogenic condi ions ela ed wi h he i e igni ion and we e
chosen on he basis o o he p e ious esea ch (Rod igues e al 2014).
4.1.3. Clima e and wea he
Clima e da a we e ex ac ed om MOTEDAS (Mon hly Tempe a u e Da ase o Spain) and MOPREDAS
(Mon hly P ecipi a ion Da ase o Spain) da ase s. These da abases p o ide mon hly clima e in o ma ion a
a spa ial esolu ion o 10 x 10 km. They we e cons uc ed om eal measu emen s om he Spanish
Me eo ological Ne wo k o wea he s a ions in he pe iod 1951-2010 (González-Hidalgo e al., 2015, 2011).
MOTEDAS and MOPREDAS s and ou as one o he mos accu a e da abases in he con ex o clima e
da a o mainland Spain.
Thei de elopmen was based on he econs uc ion o a me eo ological da a ime se ies om each wea he
s a ion in he egion. This p ocess includes a quali y con ol, consis ing o wo s eps: suspec da a
iden i ica ion and inhomogenei y de ec ion. Fi s ly, a se o e e ence se ies was calcula ed o each o iginal
s a ion by means o a mon hly co ela ion ma ix be ween he candida e se ies and all he o he s, and
selec ing he neighbo ing se ies wi h he highes posi i e mon hly co ela ion coe icien (mean g ea e han
0.60 and 0.50, o MOPREDAS and MOTEDAS, espec i ely) wi hin a c i ical h eshold dis ance o 25
km ( o MOTEDAS) and 50 km ( o MOPREDAS). The minimum o e lapping pe iod equi ed o he
co ela ion compu a ion was se a 7 yea s om MOTEDAS and 10 yea s o MOPREDAS.

Chap e 4: Ma e ials and me hods
28
To assess he suspec da a, au ho s use bo h a io and in e -qua ile me hods, as well as di ec and in e se
a ios o a oid he ze o e ec . On he o he hand, o homogenei y analyses hey applied a combina ion o
es s: Single No mal Homogenei y Tes -SNHT, Bi a ia e, -S uden and Pe i es . Finally, in o de o ill
he gaps in he da a, he me hod consis s in p oducing a combina ion o neighbo ing se ies wi h no
o e lapping pe iods.
The inal da abase o MOTEDAS consis s o 3,066, and MOPREDAS o 2,670 selec ed homogenous
se ies wi hou suspec da a om he di e en s a ions o AEMET (Agencia Es a al de Me eo ología). A e
in e pola ing he s a ions’ da a on o he 10 x 10 km g id cells, using an imp o ed e sion comp ising a
combina ion o wo weigh s: one adial weigh wi h a Gaussian shape, and an angula weigh . The adial
weigh p e en s undesi ed exchange o in o ma ion be ween di e en clima ic egions, and be ween ei he
side o he la ges moun ain chains. The angula weigh a oids undesi ed o e weigh ing o he a eas wi h
he highes s a ion densi y. The mean numbe o s a ions in ol ed in he es ima ion o each g id is
app oxima ely 4 o MOTEDAS and 6 o MOPREDAS.
Fo he i s objec i e o his disse a ion, mon hly da a on annual a e age maximum empe a u e (T -
Figu e 4) and o al p ecipi a ion in mm (P - Figu e 5) in he pe iod 1974-2010 we e ex ac ed and adap ed
o he i e g id using a nea es neighbo p ocedu e. Bo h maximum empe a u e and p ecipi a ion we e la e
eclassi ied in o 10 homogeneous (equal in e al) ca ego ies used o cons uc clima e codes o he la e
i e ea u es ela ionship plo s.
Figu e 4. Spa ial dis ibu ion o a e age maximum empe a u e (in ºC) om
MOTEDAS.
Chap e 4: Ma e ials and me hods
29
Figu e 5. Spa ial dis ibu ion o o al p ecipi a ion (in mm) om MOPREDAS.
The ERA-In e im Reanalysis da ase s p oduced by he Eu opean Cen e o Medium-Range Wea he
(ECMWF) da ase (Dee e al., 2011) was used o cons uc h ee di e en i e dange indexes (Fi e Wea he
Index: FWI, US Bu ning Index: BI and Aus alian McA hu Fo es Fi e Dange Index: FFDI) necessa y
o achie e he 2 d objec i e. The main eason o his choice was due o he ac ha his sou ce has a highe
spa ial and empo al esolu ion (a ound 78 km). Mo e speci ically, 3-hou ly 2 m ai empe a u e, dew poin
empe a u e, su ace o al p ecipi a ion, and 10 m wind componen s we e ex ac ed o de i e he ollowing
clima e a iables: maximum and minimum empe a u e, maximum and minimum ela i e humidi y,
maximum wind, o al daily p ecipi a ion amoun and o al daily p ecipi a ion du a ion (see Jolly 2015 o
mo e de ails).
The Wo ldClim da abase is an in e pola e clima e su ace o global land a eas a a spa ial esolu ion o 1
km (Hijmans e al., 2005). Mon hly p ecipi a ion and mean, minimum, and maximum empe a u e we e
included as clima e elemen s, and all inpu da a came om di e en sou ces, es ic ed o all eco ds o he
1950-2000 pe iod. Wo ldClim was pa icula ly chosen o c ea e he Aus alian McA hu FFDI, p o iding
he annual mean p ecipi a ion da a, which when combined wi h he ECMWF maximum daily p ecipi a ion
and empe a u e, p oduced he D ough Index (see Figu e 2 in Chap e 6 and Jolly e al. (2015) o u he
de ails on he calcula ion p ocess). I is he e o e pa o he achie emen o he hi d objec i e o his
disse a ion.
4.1.4. An h opogenic d i e s
The i s o he human d i e s e e s o land use da a, and was e ie ed om Co ine Land Co e 1990
(CLC), since i is cen e ed on he s udy pe iod. CLC in o ma ion was used o ou line he Wildland-
Ag icul u al In e ace (WAI) and he Wildland-U ban In e ace (WUI), wo a iables s ongly ela ed o
an h opogenic igni ions (V. Leone e al., 2009; Ma ínez e al., 2004; Rod igues e al., 2014). WAI ep esen s
he leng h o he bounda y be ween ag icul u al and wildland a eas, and WUI, he leng h be ween popula ed
and wildland a eas. Bo h we e calcula ed a i e g id le el (Rod igues e al., 2016).
Chap e 4: Ma e ials and me hods
30
On he o he hand, in o de o ep esen he human p essu e o e he wildlands, we ha e chosen he
Demog aphic Po en ial (DP), which is an agg ega e index o he ul ima e u u e po en ial o he popula ion,
was e ie ed om (J. L. Cal o and Pueyo, 2008) and based on he ollowing o mula:
𝑃𝑂𝑇𝑖= ∑( 𝑃𝑗
𝑑𝑖𝑗
2)+𝑃𝑖
𝑛
𝑗=1
( 1 )
whe e POTi is he popula ion po en ial accumula ed in cell i, Pj a e he inhabi an s coun ed in each o he
emaining accoun ing cells o he sys em and Pi a e hose o cell i i sel , while d2 is he kilome e dis ance
be ween each pai o cells i and j.
In he ca og aphic alues o POTi, hose co esponding o i s own esiden popula ion (Pi) plus hose
in e ed by he es o he sys em as a consequence o i s posi ioning in he whole a e accumula ed, ob ained
by he sum o he popula ion alues o Pj di ided by he dis ances (d) o which each accoun ing cell (j) is
di ided wi h espec o (i), and he la e ele a ed o an exponen , which in his case is 2, coinciding wi h
he g a i a ional o mula p oposed by New on.
The demog aphic po en ial in 1991 was used a a spa ial esolu ion o 5 x 5 km, la e escaled o he i e
g id as he a e age alue inside each cell (Figu e 6). WAI, WUI (Figu e 7 and Figu e 8, espec i ely) and
DP we e no malized o a 0-1 in e al and hen agg ega ed o de elop a Human P essu e Index (HPI, Figu e
9), ep esen ing he o e all p essu e o human ac i i ies likely o esul in i e igni ion.
Figu e 6. Spa ial dis ibu ion o he Demog aphic Po en ial (DP) in 1991.
Chap e 4: Ma e ials and me hods
31
Figu e 7. Spa ial dis ibu ion o he wildland ag icul u al in e ace (WAI) leng h in
me e s.
Figu e 8. Spa ial dis ibu ion o he wildland u ban in e ace (WUI) leng h in me e s.
Chap e 4: Ma e ials and me hods
38
Remainde : he componen ha is le o e om he wo p e ious ones, and which he e o e can
be unde s ood as anomalies o ex eme e en s (bo h excep ionally high and low alues) ha a e
ou side he a e age alues o he end and seasonal ime se ies.
Figu e 12. Example o i e equency ime se ies decomposi ion in he Medi e anean egion o mainland Spain.
Au oco ela ion Func ion (ACF)
This is one o he simples me hods o check ha a ime se ies ul ills he cha ac e is ic o being s a iona y.
Speci ically, he idea is o obse e whe he e e y signal di e s by a high deg ee o 0 o each ime lag. Wi h
his pu pose in mind, he ACF signal g aph is isualized. In pa icula , a s a iona y signal p oduces ew
signi ican delays exceeding he ACF con idence in e al. In compa ison, ano he ime se ies wi h a end
would show ha , in mos o i s ime lags, he con idence in e al o he ACF is exceeded.
Au o eg essi e In eg a ed and Mo ing A e age (ARIMA)
To o ecas he e olu ion o i e ea u es, a se o au o- eg essi e, in eg a ed and mo ing a e age (ARIMA)
models we e employed. They can be iewed as a “ il e ” ha ies o sepa a e he signal om he noise, and
he signal is hen ex apola ed in o he u u e o ob ain o ecas s. Thei main ad an age is ha hey adjus
exclusi ely o he his o ical se ies o he inpu a iable, which g ea ly educes he complexi y o he analysis,
since i is no necessa y o inco po a e o he explana o y a iables. Howe e , he main condi ion o ARIMA
models is ha ime se ies a e s a iona y, i.e. cons an in mean and a iance. As his condi ion is e y di icul
o ind, all i e ea u e ime se ies we e p e iously ans o med h ough he squa e oo and hen a de-
ans o ma ion was applied o e u n o hei o iginal uni s. The u u e a ge pe iod was se a 2016-2036,
so ha i would be he same leng h as he es o pe iods. This explo a ion p esupposed a con inuous
scena io in which i is assumed ha he e olu ion o he ac o s associa ed wi h i e ac i i y de elop as
obse ed in he whole his o ic pe iod (1974-2015).
Mon hly ime se ies o i e ea u es o he cu en pe iod (1995-2015) we e en e ed in o he ARIMA. The
eason was o include he seasonal componen (in a-annual peaks and d ops) because hey would o e
mo e in o ma ion o he model so ha he u u e p ojec ion would be as consis en and ealis ic as possible.

Chap e 4: Ma e ials and me hods
39
ARIMA o e s se e al ou pu da a, he mos impo an o which was he mean o he o ecas , as well as
he uppe and lowe limi s o wo con idence in e als (80% and 95%).
An au oma ic ARIMA was applied o ob ain u u e i e egime ea u es, e u ning he bes model acco ding
o he minimum Akaike in o ma ion c i e ion (AIC) alue, so i s algo i hm au oma ically calcula es he p, i
and q pa ame e s. As epo ed by Hyndman and Khandaka (2008), he seasonal ARIMA o mula is
es ablished as ollows:
Φ(𝐵𝑚)∅(𝐵)(1−𝐵𝑚)𝐷(1−𝐵)𝑑𝑦𝑡=𝑐+Θ(𝐵𝑚)𝜃(𝐵)𝜀𝑡
( 6 )
whe e Φ(z) and Θ(z) a e polynomials o o de s P and Q espec i ely, each con aining no oo s inside he
uni ci cle. I c≠0, he e is an implied polynomial o o de d + D in he o ecas unc ion. The main ask in
au oma ic ARIMA o ecas ing is selec ing an app op ia e model o de , ha is he alues p, q, P, Q, D, d.
When d and D a e known, he es o o de s a e chosen ollowing an in o ma ion c i e ion such as he AIC:
AIC= −2log(𝐿)+2(𝑝+𝑞+𝑃+𝑄+𝑘)
( 7 )
whe e k=1 i 𝑐≠0 and 0 o he wise, and L is he maximized likelihood o he model i ed o he
di e enced da a (1−𝐵𝑚)𝐷(1−𝐵)𝑑𝑦𝑡. The likelihood o he ull model o 𝑦𝑡 is no ac ually de ined
and so he alue o he AIC o di e en le els o di e encing a e no compa able. In o de o o e come
his di icul y, o ou case o seasonal da a, we selec ed he seasonally di e enced da a D =1.
4.2.3. Classi ica ion and eg ession
Clus e ing
The i e egime delimi a ion was done by Wa d’s clus e ing me hod om he NbClus R package. This
package p o ides a o al o 30 indices o choosing he mos adequa e numbe o clus e s and p oposing
he bes clus e ing scheme om he di e en esul s ob ained by a ying all combina ions o clus e s
(minimum and maximum desi ed), dis ance measu emen s and clus e ing me hods. A e se e al ial-and-
e o changing unc ion pa ame e s and acco ding o he esul s ob ained, he minimum and maximum
numbe o clus e s was es ablished a 5 and 7, espec i ely. The dis ance selec ed was Canbe a (Cd), and
he me hod o he clus e ing ou lined was Wa d.D2. Cd was p oposed by Lance and Williams (1967) and
examines he sum o se ies o a ac ion o di e ences be ween he coo dina es o a pai o obse a ions
(Teknomo, 2015). In gene al, he Cd e ms wi h ze o nume a o and denomina o a e omi ed om he
sum and ea ed as i he alues we e missing (Cha ad e al., 2014). The o mula o Cd is as ollows:
𝐶𝑑 (𝑥,𝑦)=∑ |𝑥𝑗−𝑦𝑗|
|𝑥𝑗|+|𝑦𝑗|
𝑑
𝑗=1
( 8 )
whe e xj is he i s obse a ion wi h coo dina es o he ea u es and yj is he second obse a ion wi h i s
co esponding coo dina es o he same ea u es. Each e m o ac ion di e ence has alue be ween 0 and
1, al hough in i sel i is no eally be ween ze o and one. I one o coo dina es is ze o, he e m becomes 1
Chap e 4: Ma e ials and me hods
40
ega dless o he o he alue, hus he dis ance will no be a ec ed. Consequen ly, Cd is e y sensi i e o a
small change when bo h coo dina es a ea nea o ze o. The e o e, Cd has he ad an age o no being
a ec ed by he p esence o ze os, which a e abundan in some cells o he s udy a ea, mo e especially in
i e ea u es such as na u al bu ned a ea and la ge bu ned a ea (abo e 100 has).
A each s ep he pai o clus e s is chosen which leads o a minimum inc ease in he o al wi hin-clus e
a iance a e me ging. The Wa d.D2 op ion implemen s Wa d’s clus e ing c i e ion in which he
dissimila i ies a e squa ed be o e clus e ing upda ing.
K-Nea es Neighbo
In o de o ans e cu en i e egime clus e s o he emaining ime pe iods (pas and u u e), a KNN
classi ica ion was pe o med. KNN is a nonpa ame ic echnique used in s a is ical es ima ion and pa e n
ecogni ion (Ripley, 1996) widely used since he 1970’s. The cu en pe iod was aken as a benchma k,
because i has mo e obus and eliable da a. KNN ains o each g id in he es da ase (pas and u u e
i e ea u es), inds he nea es K by a dis ance measu e (Canbe a dis ance), and he clus e class is decided
by a majo i y o e o i s neighbo s. The K pa ame e means he maximum numbe o nea es neighbo s
conside ed in he algo i hm (Venables and Ripley, 2002), being se in 5.
Gene alized Addi i e Models (GAM)
In o de o ul il he i s esea ch objec i e o un a elling po en ial cause-and-e ec ela ionships be ween
i e ea u es and clima ic/human a iables, se e al GAM eg essions we e calib a ed o each
Mul idimensional Sca e plo (MDS) subse . Gene alized Addi i e Models (GAM) a e Gene alized Linea
Models (GLM) in which he usual linea ela ionships be ween he esponse and p edic o a iables a e
eplaced by non-linea ‘smoo hs’ (Has ie and Tibshi ani, 1986; Jones and Almond, 1992). The same as
GLM, GAM can use p obabili y dis ibu ions o he han Gaussian, so we applied Nega i e Binomial o
model he numbe o i es (N) and log linea dis ibu ion in bu ned a ea a iables (B500, BL). NB is
pa icula ly sui able o deal wi h ze o-in la ed esponse a iables, as is he case o N (Boadi e al., 2015). On
he o he hand, we applied a log linea amily in bu ned a ea i e ea u es (He nandez e al., 2015). Model
selec ion, is based on he educ ion o Gene alized c oss alida ion (GCV, C a en and Wahba, 1978; Golub
e al., 1979). GVC de e mines he op imal amoun o smoo hing and es ima es he mean squa ed p edic ion
e o o e all da ase s whe e a single obse a ion is omi ed om he model i ing, and hen p edic ed
De iance is explained (analogous o a iance in a linea eg ession) and pa ial e ec s in he p edic o s we e
also calcula ed. All analyses o GAM modeling we e conduc ed using he R package mgc , e sion 1.8–9.
Random Fo es
In o de o assess he ole o he d i e s in i e egime change, we selec ed Random Fo es (RF; B eiman,
2001) as he modeling algo i hm, gi en i s p o en p edic i e accu acy (Ba Massada e al., 2011; Leuenbe ge
e al., 2018a; Rod igues and de la Ri a, 2014a). RF is a ee-based ensemble algo i hm ha ains mul iple
decision ees by andomly boo s apping he aining sample, keeping 67% o he obse a ions o ain he
decision ee and he emaining 33% (Ou -o -bag, OOB) o e alua e he ela i e in luence o he p edic o s
and he model i sel . The inal s age assembles all ees in o a inal p edic ion as he a e age o all indi idual
ee p edic ions (Bagging; B eiman, 2001).
Chap e 4: Ma e ials and me hods
41
Fo each i e egime ansi ion ype, we ained and alida ed 100 RF models, using a andom sample o
70% o aining and he emaining 30% o es ing he pe o mance o he model. A he aining s age,
we conduc ed a 10- old calib a ion p ocedu e o iden i y he op imal pa ame e s (m y and n ees) o he
model. A he same ime, we also e alua ed he in luence o each d i e by calcula ing he pe cen age
inc ease in he Mean Squa e E o (no malized be ween 1 and 0), and i s explana o y sense by means o
pa ial dependence plo s (J.H. F iedman, 2001). To es ima e he p edic i e pe o mance o each model
ca ied ou , we calcula ed he A ea Unde he Recei e Ope a ing Cha ac e is ic Cu e (AUC; B adley,
1997). Addi ionally, he explana o y sense o he co a ia es (ei he posi i ely o nega i ely ela ed) was
explo ed by isual inspec ion o pa ial dependence plo s.
Geog aphically Weigh ed Reg ession Models (GWLR)
GWR is a s a is ical echnique o explo a o y spa ial da a analysis de eloped wi hin he amewo k o Local
Spa ial Models o S a is ics. Local models could be desc ibed as he spa ial disagg ega ion o global s a is ics
whose main cha ac e is ic is ha i is calib a ed om a se o spa ially limi ed samples and, hence, yielding
local eg ession pa ame e es ima es (Fo he ingham e al., 2002). The e o e, GWR echniques ex end he
adi ional use o global eg ession models, enabling local eg ession pa ame e s o be calcula ed.
Ma hema ically, a con en ional GWR is desc ibed by he ollowing equa ion:
𝑦𝑖= ∑𝛽𝑘 (𝑢𝑖 ,𝑣𝑖) 𝑋𝑘,𝑖
𝑘+ 𝜀𝑖
( 9 )
whe e yi, xk,i, and 𝜀i a e dependen a iables, k h is he independen a iable, and he Gaussian e o a
loca ion i;(ui, i) is he x–y coo dina e o he i h loca ion; and coe icien s 𝛽 (ui, i) a e a ying condi ionals on
he loca ion.
Such modeling is likely o a ain highe pe o mance han adi ional eg ession models, and eading he
coe icien s can lead o a new in e p e a ion o he phenomena unde s udy. Howe e , GWR models a e
no jus a simple local eg ession model like, i.e., mo ing window eg essions. In a mo ing window example,
a egion is d awn a ound a eg ession poin and all he da a poin s wi hin his egion (neighbo hood) o
window a e hen used o calib a e a model. This p ocess is epea ed o e all he eg ession poin s, esul ing
in a se o local eg ession s a is ics. Howe e , in his example, each poin wi hin he neighbo hood is ea ed
equally o eg ession pu poses, no ma e i s dis ance o he a ge eg ession poin . GWR o e comes his
limi a ion by applying a dis ance weigh pa e n; hence, da a poin s close o he eg ession poin a e
weigh ed mo e hea ily in he local eg ession han da a poin s a he away. In addi ion o he eg ession
coe icien s, a GWR model calcula es se e al use ul s a is ical pa ame e s o analyze he spa ial beha io o
each explana o y a iable, such as he alue o he S uden ’s es , which is used o de e mine he le el o
signi icance. On he o he hand, GLM app oaches such as Geog aphically Weigh ed Logis ic Reg ession
(GWLR) and Geog aphically Weigh ed Poisson Reg ession (GWPR) ha e been inco po a ed o GWR o
ex end i s unc ionali y (Fo he ingham e al., 2002; Nakaya and Fo he ingham, 2009). The GWR app oach
has been al eady been explo ed in se e al pape s such as Kou sias, Ma ínez-Fe nández, & Allgöwe (2010),
Ma ínez-Fe nández e al. (2013) and Rod igues e al. (2014). These wo me hodologies—GWLR and
GWPR—a e used in his s udy o complemen he esul s om GLM. Se e al pa ame e s ha e been
included when calib a ing GWR models. Ke nel shape and ype, bandwid h selec ion and op imiza ion
pa ame e s, o he local o global na u e o he p edic o s (see Nakaya and Fo he ingham, (2009) o u he
de ails o bo h me hod and so wa e). In his p ojec , GWR model i ing was ca ied ou using Fixed
Chap e 4: Ma e ials and me hods
42
Gaussian Ke nel bandwid h, op imized acco ding o he alue o AICc, conside ing all he p edic o s as
local co a ia es.
5
CHAPTER 5: SPATIAL-
TEMPORAL DISTRIBUTION OF
FIRE REGIME FEATURES
This chap e desc ibes he esul s, discussion and main
conclusions ob ained om he analyses ela ed o he
iden i ica ion o he majo i e egime ea u es and hei empo al
dynamics. We e alua e he ela ionships o i e ea u es wi h
clima e g adien s and human p essu e, as well as he assessmen
o he con ibu ion o small i es in o he i e egime
cha ac e iza ion. Mul i-G oup P incipal Componen analysis,
GAM models, change poin de ec ion me hods, Mann-Kendall
and Sen’ slope ha e been applied o i e ea u es a egional and
p o incial le el. The main goals a e: desc ibe and cha ac e ize
he i e egime, iden i y i s shi s and ends, de e mina e he
ex en o which i e egime is linked o clima e-human ac o s
and disco e po en ial ela ions in he e olu ions o i e ea u es.
The e o e, we seek o imp o e he unde s anding o he spa ial-
seasonal pa e ns o he key i e egime ea u es.

Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
45
Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
46
Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
47
Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
54

Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
55
Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
56
Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
57
Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
58
Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
59

Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
60
Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
61
Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
62
Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
63
Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
70

Chap e 5: Spa ial- empo al dis ibu ion o i e egime ea u es
71
6
CHAPTER 6: THE INFLUENCE
OF FIRE-WEATHER ON THE
EVOLUTION OF FIRE ACTIVITY
This chap e desc ibes he esul s, discussion and main
conclusions ob ained om he analysis o spa ial and empo al
associa ions be ween mon hly ime se ies o i e wea he dange
indices (Fi e Wea he Index, Bu ning Index and Fo es Fi es
Dange Index) a egional and local le el. Decomposi ion o ime
se ies was he i s s ep, hen apply c oss-co ela ion o explo e
seasonal associa ions a egional scale, as well as, a Pea son’s
co ela ion was calcula ed be ween each index and 18 i e-ac i i y
subse s by i e size and cause a local scale.
Chap e 6: The in luence o i e-wea he on he e olu ion o i e ac i i y
75

Chap e 6: The in luence o i e-wea he on he e olu ion o i e ac i i y
76
Chap e 6: The in luence o i e-wea he on he e olu ion o i e ac i i y
77
Chap e 6: The in luence o i e-wea he on he e olu ion o i e ac i i y
78
Chap e 6: The in luence o i e-wea he on he e olu ion o i e ac i i y
79
Chap e 6: The in luence o i e-wea he on he e olu ion o i e ac i i y
86

7
CHAPTER 7: CHANGE IN
ANTHROPOGENIC DRIVERS
This chap e desc ibes he esul s, discussion and main
conclusions ob ained om he analyses o spa ial and empo al
e olu ion o human d i e s ac o s in o he i e egime ea u es.
We employed a ious eg ession models (Logi and Poisson
Gene alized Linea Models), as well as, end analysis by means
o Mann-Kendall. In addi ion, Geog aphically Weigh ed
Reg ession Models a e applied o assess spa ial- empo al
pa e ns.
Chap e 7: Change in an h opogenic d i e s
89
Chap e 7: Change in an h opogenic d i e s
90
Chap e 7: Change in an h opogenic d i e s
91

Chap e 7: Change in an h opogenic d i e s
92
Chap e 7: Change in an h opogenic d i e s
93
Chap e 7: Change in an h opogenic d i e s
94
Chap e 7: Change in an h opogenic d i e s
95
Chap e 7: Change in an h opogenic d i e s
102

Chap e 7: Change in an h opogenic d i e s
103
Chap e 7: Change in an h opogenic d i e s
104
Chap e 7: Change in an h opogenic d i e s
105
Chap e 7: Change in an h opogenic d i e s
106
Chap e 7: Change in an h opogenic d i e s
107

Chap e 7: Change in an h opogenic d i e s
108
Chap e 7: Change in an h opogenic d i e s
109
Chap e 7: Change in an h opogenic d i e s
110
Chap e 7: Change in an h opogenic d i e s
111
Chap e 7: Change in an h opogenic d i e s
118

Chap e 7: Change in an h opogenic d i e s
119
Chap e 7: Change in an h opogenic d i e s
120
Chap e 7: Change in an h opogenic d i e s
121
Chap e 7: Change in an h opogenic d i e s
122
Chap e 7: Change in an h opogenic d i e s
123

Chap e 7: Change in an h opogenic d i e s
124
Chap e 7: Change in an h opogenic d i e s
125
Chap e 7: Change in an h opogenic d i e s
126
8
CHAPTER 8: EVOLUTION AND
CAUSES OF FIRE REGIME
CHANGE
This chap e desc ibes he main esul s, discussion and
conclusions o he ou lining o i e egime zones, hei empo al
e olu ion owa ds he nea u u e and he analysis o he
in luence o d i e s o i e ac i i y in he obse ed i e egime
ajec o ies. Random Fo es is employed o e alua e he
indi idual con ibu ion o each i e d i e , as well as, ARIMA
models a e used o o ecas he immedia e u u e end o he
main i e egime ea u es.

Appendix B
231
Appendix B
232
Appendix B
233
Appendix B
234
Appendix B
235

C
APPENDIX C: SUPLEMANTARY
MATERIAL OF DRIVERS OF
CHANGE
This appendix p esen s he supplemen a y ma e ial o he
accep ed pape en i led “Fi e egime dynamics in mainland
Spain. Pa 1: d i e s o change” which shows complemen a y
esul s ob ained in his publica ion.
Appendix C
239
Appendix C
246

Appendix C
247
Appendix C
248
Appendix C
249
Appendix C
250
D
APPENDIX D: PRELIMINARY
PYROREGIONS DELIMITATION
This appendix p esen s he wo k “Iden i ying py o egions by
means o Sel O ganizing Maps and hie a chical clus e ing
algo i hms in mainland Spain” o ming pa o he con e ence
p oceedings in
Ad ances in Fo es Fi e Resea ch
2018. I
summa izes a p elimina y a emp o de ine spa ial- empo al
de ini ion py o egions employing Sel O ganizing Maps (SOM),
including he s uc u al and end componen o i e egime
ea u es.

252
Appendix D
253
Appendix D
254
Appendix D
255
Appendix D
262

Appendix D
263
265
E
APPENDIX E: CONFERENCES
CONTRIBUTIONS
This appendix b ings oge he se e al abs ac s om di e en
con e ence con ibu ions, mainly om he Eu opean
Geosciences Union Gene al Assembly (EGU) held in 2017 and
2018. Mos o he abs ac s e e s o pos e p esen a ions and one
o an o al disse a ion.
Appendix E
267

Appendix E
268
Appendix E
269
Appendix E
270