Repositorio Institucional de Documentos
Abstract
Los espacios forestales son una fuente de servicios, tanto ambientales como económicos, de gran importancia para la sociedad. La caracterización de estos ambientes ha requerido tradicionalmente de un laborioso trabajo de campo. La aplicación de técnicas de teledetección ha proporcionado una visión más amplia a escala espacial y temporal, a la par que ha generado una reducción de los costes. La utilización de sensores óptico-pasivo multiespectrales y de sensores radar posibilita la estimación de parámetros forestales, si bien el desarrollo de sensores LiDAR, como el caso de los escáneres láser aeroportados (ALS), ha mejorado la caracterización tridimensional de la estructura de los bosques. La disponibilidad pública de dos coberturas LiDAR, generadas en el marco del Plan Nacional de Ortofotografía Aérea (PNOA), ha abierto nuevas líneas de investigación que permiten proporcionar información útil para la gestión forestal. <br />La presente tesis utiliza datos LiDAR aeroportados de baja densidad para estimar diversas variables forestales, con ayuda de trabajo de campo, en masas forestales de Pino carrasco (Pinus halepensis Miller) en Aragón. La investigación aborda dos cuestiones relevantes como son la exploración de las metodologías más adecuadas para estimar variables forestales considerando escalas locales y regionales, teniendo en cuenta las posibles fuentes de error en el modelado; y, además, analiza la potencialidad de los datos LiDAR del PNOA para el desarrollo de aplicaciones forestales que valoricen las áreas forestales como recursos socio-económicos. <br />La tesis se ha desarrollado según la modalidad de compendio de publicaciones, incluyendo cuatro trabajos que dan respuesta a los objetivos planteados. En primer lugar, se realiza un análisis comparativo de distintos modelos de regresión, paramétricos y no paramétricos, para estimar la pérdida de biomasa y las emisiones de CO2 en un incendio, mediante la utilización de datos LiDAR-PNOA y datos ópticos del satélite Landsat 8. En segundo lugar, se explora la idoneidad de distintos métodos de selección de variables para estimar biomasa total en masas de Pino carrasco utilizando datos LiDAR de baja densidad. En tercer lugar, se cuantificó y cartografió la biomasa residual forestal en el conjunto de masas de Pino carrasco de Aragón y se evaluó el efecto de diversas características de la tecnología LiDAR y de las variables ambientales en la precisión de los modelos. Finalmente, se analiza la transferibilidad temporal de modelos para estimar a escala regional siete variables forestales, utilizando datos LiDAR-PNOA multi-temporales. A este respecto, se compararon dos enfoques que permiten analizar la transferibilidad temporal: en primer lugar, el método directo ajusta un modelo para un determinado punto en el tiempo y estima las variables forestales para otra fecha; por otra parte, el método indirecto ajusta dos modelos diferentes para cada momento en el tiempo, estimando las variables forestales en dos fechas distintas. <br />Los resultados obtenidos y las conclusiones derivadas de la investigación indican que la técnica basada en coeficientes de correlación de Spearman y el método de selección por todos los subconjuntos constituyen los métodos de selección de métricas LiDAR más apropiados para la modelización. El análisis de métodos de regresión para la estimación de variables forestales indicó <br />que su idoneidad variaba de acuerdo con el tamaño y complejidad de la muestra. El método de regresión linear multivariante arrojó mejores resultados que los métodos no-paramétricos en el caso de muestras pequeñas. Por el contrario, el método Support Vector Machine produjo los mejores resultados con muestras grandes. El incremento de la densidad de puntos y de los valores de penetración de los pulsos LiDAR en el dosel, así como la presencia de ángulos de escaneo pequeños, incrementó la exactitud de los modelos. De forma similar, el incremento de la pendiente y la presencia de arbustos en el sotobosque implican una reducción en la exactitud de los modelos. En la estimación de variables forestales utilizando datos LiDAR multi-temporales, aunque la utilización del enfoque indirecto arrojó generalmente una mayor precisión en los modelos, se obtuvieron resultados similares con el enfoque directo, el cual constituye una alternativa óptima para reducir el tiempo de modelado y los costes de realización de trabajo de campo. La fusión de datos LiDAR y datos óptico-pasivos ha evidenciado la conveniencia de los métodos aplicados para cuantificar las emisiones de CO2 a la atmósfera generadas por un incendio. Esta metodología constituye una alternativa adecuada cuando no existen datos multi-temporales LiDAR. La estimación de variables de inventario forestal, así como de diversas fracciones de biomasa, como la biomasa total y la biomasa residual forestal, proporciona información valiosa para caracterizar las masas forestales mediterráneas de Pino carrasco y mejorar la gestión forestal<br /> Forest ecosystems provide environmental and economic services of great importance to the society. The characterization of these environments has been traditionally accomplished with intense field work. In comparison, the application of remote sensing tools provides a greater overview over large spatial and temporal scales while minimizing costs. Although optical data and Synthetic Aperture Radar (SAR) allow estimating forest stand variables, the development of LiDAR sensors such as Airborne Laser Scanner (ALS) have improved three-dimensional characterization of forest structure. The availability of two ALS public data coverages for the Spanish territory, provided by the National Plan for Aerial Ortophotography (PNOA), opens new research opportunities to generate useful information for forest management. This PhD Thesis used low-density ALS-PNOA data to estimate different forest variables, with support in fieldwork, in the Aleppo pine (Pinus halepensis Miller) forests of Aragón region. The addressed research is relevant mainly for two reasons: first, the examination of suitable methodologies and error sources in forest stand variables prediction at local (small area) and regional scales (large area), and second, the application of ALS data to the characterization of forest areas as a socio-economic reservoir. This PhD Thesis is a compendium of four scientific papers, which sequentially answer the objectives established. Firstly, a comparative analysis of different parametric and non-parametric models was performed to estimate biomass losses and CO2 emissions using low-density ALS and Landsat 8 data in a burnt Aleppo pine forest. Secondly, we assess the suitability of variable selection methods when estimating total biomass in Aleppo pine forest stands using low-density ALS data. In the third manuscript, the quantification and mapping of forest residual biomass in Aleppo pine forest of Aragón region and the assessment of the effect of ALS and environmental variables in model accuracy were accomplished. Finally, the temporal transferability of seven forest stands attributes modelling using multi-temporal ALS-PNOA data in Aleppo pine forest at regional scale was explored. In this case, the temporal transferability was assessed comparing two methodologies; the direct and indirect approach. The first one fits a model for one point in time and estimates the forest variable for another point in time. The indirect approach adjusts two models in different points in time to estimate the forest variables in two different dates. The results derived from this research indicated that Spearman’s rank and All Subset Selection are the most appropriate methods in the ALS metrics selection step commonly applied in modelling. The suitability of the regression methods depends on the sample size and complexity. Thus, multivariate linear regression outperformed non-parametric methods with small samples while support vector machine was the most accurate method with larger samples. Model accuracy increased with higher point density and canopy pulse penetration, while decreasing with wider scan angles. Furthermore, the presence of steep slopes and shrub reduced model performance. In the case of forest stand variables prediction using multi-temporal ALS data, although the indirect approach produced generally a higher precision, the direct approach provided similar results, constituting a suitable alternative to reduce modelling time and fieldwork costs. The fusion of ALS and passive optical data have evidenced the suitability of this information for quantifying wildfire CO2 emissions to atmosphere, constituting a good alternative when multi-temporal ALS data is not available. The estimation of forest inventory variables as well as different biomass fractions, such as total biomass and forest residual biomass, provided valuable information to characterize Mediterranean Aleppo pine forests and improve forest management.<br /> Domingo Ruiz, Darío; De la Riva Fernández, Juan; Lamela Gracia, María Teresa
Full text
2021
48
Da ío Domingo Ruiz
Cha ac e ia ion o
Medi e anean Aleppo
pine o es using low-
densi y ALS da a
Depa amen o
Di ec o /es
Geog a ía y O denación del Te i o io
De la Ri a Fe nández, Juan
Lamela G acia, Ma ía Te esa
© Uni e sidad de Za agoza
Se icio de Publicaciones
ISSN 2254-7606
Da ío Domingo Ruiz
CHARACTERIATION OF MEDITERRANEAN
ALEPPO PINE FOREST USING LOW-DENSITY ALS
DATA
Di ec o /es
Geog a ía y O denación del Te i o io
De la Ri a Fe nández, Juan
Lamela G acia, Ma ía Te esa
Tesis Doc o al
Au o
2019
UNIVERSIDAD DE ZARAGOZA
Reposi o io de la Uni e sidad de Za agoza – Zaguan h p://zaguan.uniza .es
Da ío Domingo Ruiz
Di ec o es: Mª Te esa Lamelas G acia y Juan de la Ri a Fe nández
PhD Thesis
Za agoza 2019
Cha ac e iza ion o Medi e anean Aleppo pine o es
using low-densi y ALS da a
The au ho o his PhD Thesis was suppo ed by Go e nmen o Spain, Depa men o Educa ion
Cul u e and Spo s unde G an (FPU G an BOE, 14/06250). Fu he mo e, he wo k was
suppo ed by FoResBiomALS esea ch p ojec om he Cen o Uni e si a io de la De ensa de
Za agoza (2017-09), SERGISAT esea ch p ojec (CGL2014-57013-C2-2-R) om he Spanish
Na ional Plan o Scien i ic and Technical Resea ch and Na ional Inno a ion Plan (Minis y o
Economy and Compe i i eness), and Geo o es -IUCA esea ch g oup (G oup S51_17R, inanced by
FEDER 2014-2020 and Go e nmen o A agon, “Cons uyendo Eu opa desde A agón”).
Co e page image: RGB and NIR colou ed poin clouds ( ile 664-4644) om PNOA ©INSTITUTO
GEOGRÁFICO NACIONAL DE ESPAÑA ‐ Au onomous Communi y o A agón.
This PhD Thesis is de eloped as compendium o pape s acco ding o he Doc o al p og am in O denación del
Te i o io y Medio Ambien e a Uni e si y o Za agoza. The PhD s uden , Da ío Domingo, is he i s au ho
and esponsible o each and e e y a icle lis ed below. The e e ences o he wo ks ha cons i u e he PhD
Thesis body a e he ollowing:
1. Domingo, D., Lamelas, M.T., Mon ealeg e, A.L., de la Ri a, J. 2017. Compa ison o eg ession models
o es ima e biomass losses and CO2 emissions using low-densi y ai bo ne lase scanning da a in a
bu n Aleppo pine o es . Eu opean Jou nal o Remo e Sensing, 50 (1), 384-396. doi:
10.1080/22797254.2017.1336067.
2. Domingo, D., Lamelas, M.T., Mon ealeg e, A.L., Ga cía-Ma ín, A., de la Ri a, J. 2018. Es ima ion o
o al biomass in Aleppo pine o es s ands applying pa ame ic and nonpa ame ic me hods o low-
densi y ai bo ne lase scanning da a. Fo es s, 9, 158-175. doi: 10.3390/ 9030158.
3. Domingo, D., Mon ealeg e, A.L., Lamelas, M.T., Ga cía-Ma ín, A., de la Ri a, J., Rod íguez, F,
Alonso, R. 2019. Quan i ying o es esidual biomass in Pinus halepensis Mille s ands using Ai bo ne
Lase Scanning da a. GIScience and Remo e Sensing, 56 (8), 1210-1232. doi:
10.1080/15481603.2019.1641653.
4. Domingo, D., Alonso, R, Lamelas, M.T., Mon ealeg e, A.L., Rod íguez, F, de la Ri a, J. 2019.
Tempo al T ans e abili y o Pine Fo es A ibu es Modeling Using Low-Densi y Ai bo ne Lase
Scanning Da a. Remo e Sensing, 11 (3), 261. doi:10.3390/ s11030261.
La p esen e esis doc o al se ha elabo ado como compendio de publicaciones siguiendo la modalidad
o ecida po el p og ama de Doc o ado en O denación del Te i o io y Medio Ambien e de la Uni e sidad
de Za agoza. El doc o ando, Da ío Domingo, igu a como p ime au o y esponsable de odos y cada uno
de los a ículos publicados. A con inuación, se de allan las e e encias de los abajos que cons i uyen el
cue po de la esis:
1. Domingo, D., Lamelas, M.T., Mon ealeg e, A.L., de la Ri a, J. 2017. Compa ison o eg ession models
o es ima e biomass losses and CO2 emissions using low-densi y ai bo ne lase scanning da a in a
bu n Aleppo pine o es . Eu opean Jou nal o Remo e Sensing, 50 (1), 384-396. doi:
10.1080/22797254.2017.1336067.
2. Domingo, D., Lamelas, M.T., Mon ealeg e, A.L., Ga cía-Ma ín, A., de la Ri a, J. 2018. Es ima ion o
o al biomass in Aleppo pine o es s ands applying pa ame ic and nonpa ame ic me hods o low-
densi y ai bo ne lase scanning da a. Fo es s, 9, 158-175. doi: 10.3390/ 9030158.
3. Domingo, D., Mon ealeg e, A.L., Lamelas, M.T., Ga cía-Ma ín, A., de la Ri a, J., Rod íguez, F,
Alonso, R. 2019. Quan i ying o es esidual biomass in Pinus halepensis Mille s ands using Ai bo ne
Lase Scanning da a. GIScience and Remo e Sensing, 56 (8), 1210-1232. doi:
10.1080/15481603.2019.1641653.
4. Domingo, D., Alonso, R, Lamelas, M.T., Mon ealeg e, A.L., Rod íguez, F, de la Ri a, J. 2019.
Tempo al T ans e abili y o Pine Fo es A ibu es Modeling Using Low-Densi y Ai bo ne Lase
Scanning Da a. Remo e Sensing, 11 (3), 261. doi: 10.3390/ s11030261.
Ag adecimien os
“ap ende a camina es más ácil si alguien e iende su b azo”
En p ime luga , me gus a ía exp esa mi más since o ag adecimien o a mis di ec o es de esis, la D a. Ma ía
Te esa Lamelas G acia y el D . Juan de la Ri a Fe nández po su apoyo incondicional y po su buen sabe
hace que han pe mi ido que la in es igación desa ollada haya llegado a buen pue o. G acias po habe
con iado en mí desde el inicio, po habe me dado ideas y ambién po habe me dejado abo da las mías
p opias. Me habéis enseñado a c ece en lo p o esional y en lo pe sonal y es oy segu o de que sin ues o
esón y alien o es o no hubie a sido posible.
También quie o ag adece a los coau o es de las dis in as publicaciones, el D . An onio Luis Mon ealeg e, el
D . Albe o Ga cía Ma ín, el D . Ra ael Alonso y el D . F ancisco Rod íguez, en especial po sus aliosas
apo aciones, po sus ideas y po odo el es ue zo que han dedicado.
Asimismo quie o ag adece a odos los miemb os del G upo de In es igación GEOFOREST-IUCA, del
Depa amen o de Geog a ía y O denación del Te i o io y al Cen o Uni e si a io de la De ensa po
acoge me y da me la opo unidad de o ma me como doc o ando, así como po p opo ciona me los
ecu sos económicos y ma e iales pa a pode desa olla las in es igaciones. En conc e o, el Depa amen o
ci ado es la casa donde inicié mi o mación y ha sido un g a o place pode segui o mando pa e de es a
amilia geog á ica de inmejo able calidad humana.
Ag adezco a los in es igado es que me acogie on du an e mis es ancias en el ex anje o po odo el apoyo
mos ado, po los conocimien os que me enseña on y po es a siemp e pendien es de mí. Muchas g acias a
Wa en B. Cohen y a Yang Zhiqiang po acoge me en el Labo a o y o Applica ions o Remo e Sensing in
Ecology (LARSE) del USDA Fo es Se ice en Co allis. Así mismo, muchas g acias a E ik Naesse , a Te je
Gobakken y a Hans Ole Ø ka po el a o magní ico, la dedicación y el buen sabe hace cuando me
acogie on en No wegian Uni e si y o Li e Sciences de Ås. No me puedo ol ida de An oine y Nemo que
hicie on la es ancia en Co allis mucho más amena y g a i ican e pese a es a lejos de casa. Tampoco puedo
ol ida me de Ma ie-Claude, Ana, Ida, Lenna y Vic o po odos los a os de pingpong y las salidas al
campo. También quie o ag adece a los e iso es ex e nos de la esis, a Ole Ma in Bollandsås y a Hooman
La i i po su disponibilidad y apo aciones al p esen e documen o.
Mi más sen ido ag adecimien o a odos los miemb os del p oyec o SERGISAT con los que he pasado
in ensas jo nadas de campo que e hacen c ece en lo pe sonal y labo al, po que el campo une y de qué
mane a. G acias a Mai e, a Te esa, a Juan, a Paloma, a Albe o, a Pe e, a Raúl a Ma cos y a Deme io.
Ag adezco a odos los compañe os y doc o andos que me han alen ado y ayudado du an e es e pe iodo.
Daniel Bo ini, Ad ián Jiménez, An onio Mon ealeg e, Xa ie Ga a e, Es ela Pé ez, Olga Rose o, Daniel
Balla ín, Daniel Mo a, Aldo A ánz, Rica do Badía, Samuel Es eban y And ea U gilez. Muchas g acias po
odos y cada uno de esos ca és jun os.
Del mismo modo ag adezco a Ka alin Va ga y a An onio Mon ealeg e po los buenos momen os que hemos
compa ido en di e sos cu sos y cong esos. También quie o aco da me de Yago, hace 9 años me dijis e
“apún a e a Geog a ía que e gus a el campo y segu o que e a bien”. Muchas g acias, no allas e.
1
1. In oduc ion
This chap e desc ibes he main concep s and he concep ual
amewo k in which his PhD Thesis was de eloped. Fi s ly,
ALS echnology and i s use o o es y pu poses a e
desc ibed. Secondly, he use o ALS da a o es ima ing o es
s and a iables, a local and egional scales, using di e en
eg ession algo i hms is p esen ed. Then, he esea ch
jus i ica ion, hypo hesis and aims a e de ined. Finally, he
chap e depic s he PhD Thesis s uc u e, including he
de eloped esea ch i ems and hei link o he di e en
objec i es ha o m a hema ic uni y.
In oduc ion
3
1.1. Backg ound
1.1.1. Ai bo ne lase scanning
The cha ac e iza ion and quan i ica ion o o es esou ces s a ed in Eu ope du ing he la e 18 h
cen u y when socie y was conce ned abou wood a ailabili y, being he main sou ce o uel. The
es ima ion o o es y me ics, especially olume, was pe o med by isual in e p e a ion. The
de elopmen o o es su eys, in en o y ools, sampling and s a is ical me hods and he ad ance
in compu e s du ing 19 h and 20 h cen u y yield g ea p og ess in o es y science. The g ow h o
emo e sensing ools such as ae ial images, op ical passi e sa elli e images, and Syn he ic Ape u e
Rada (SAR) da a p o ided a g ea e o e iew o o es esou ces o e la ge a eas (Boyd &
Danson, 2005). Howe e , he expansion o Ligh De ec ion and Ranging (LiDAR) echnology has
imp o ed he h ee-dimensional (3D) cha ac e iza ion o o es ecosys ems being a sui able ool o
o es y a iables es ima ion (Zhao e al., 2018).
LiDAR echnology measu es he dis ance be ween a lase ansmi e and an objec o su ace
using a monoch oma ic beam o ligh , cohe en and di ec ional (Ande sen e al., 2005). The o igin
o his echnology s a ed in he ea ly 1960s when Theodo e Ha old Maimam de eloped he i s
uby lase ha emi s powe ul pulses o collima ed ed ligh . In he 1980s he p o ile LiDAR was
used o o es y applica ion (Ald ed & Bonno , 1985; Maclean & K abill, 1986). Fu he g ow h
occu ed in he 90s wi h he gene a ion o Digi al Te ain Models (DTM) and o es in en o y
a iables es ima ion (Le sky e al., 1999; Means e al., 1999; Næsse , 1997). Du ing he las wo
decades he use o LiDAR echnology ha e exponen ially g own, being de eloped di e se
ha dwa e, so wa e and applica ions wi hin he o es y opics.
Acco ding o he pla o m used, h ee main ypes o LiDAR echnologies exis : e es ial lase
scanne s (TLS), ai bo ne lase scanne s (ALS) and sa elli e lase scanne s (SLS). ALS is one o he
mos widesp ead o o es y pu poses (Mal amo e al., 2014). Topog aphic ALS senso s emi hei
own elec omagne ic lux wi hin he in a ed wa eleng hs (900 o 1,064 nm). These wa eleng h
deno es high e lec ance alues o ege a ion and ansmissi i y o a mosphe e (Le sky e al.,
2002). The basic in o ma ion cap u ed by ALS is denomina ed poin cloud, e e ing o a dense se
o x, y and z coo dina es ha egis e he objec e lexions eached by he lase ligh .
ALS echnology is usually classi ied in wo main ypes, acco ding o he way i measu es dis ances
be ween he senso and he objec eached by he lase beam: ull-wa e o m sys ems and disc e e
e u n sys ems. Full-wa e o m sys ems comple ely egis e he e lec ed ene gy. The dis ance
be ween ansmi e and objec is measu ed using phase di e ence be ween emi ed signal and
sca e ed adia ion. Full-wa e o m senso s p o ide iche da a han disc e e e u n sys ems, while
he p ocessing is mo e demanding. Al hough se e al p ocessing me hodologies has been p oposed
such as oxeliza ion, some imes he wa eleng hs a e con e ed in o poin clouds simila o he
ones p o ided by disc e e senso s bu wi h highe numbe o e u ns pe pulse. On he o he hand,
disc e e e u n sys ems cap u e one o i e e u ns pe emi ed lase pulse. The dis ance be ween
he ansmi e and he objec is measu ed as a unc ion o ime. The gene alized use o disc e e
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
4
e u n senso s o o es y and opog aphic pu poses may be explained by he highe
implemen a ion in he comme cial sec o (Shan & To h, 2008).
ALS disc e e- e u n sys ems ha e been widely used o es ima ing o es a iables. T ees a e po ous
objec s om he lase pulse poin o iew. The lase beam can a el h ough he canopy and he
sys em is able o cap u e up o i e e u ns. The i s e u n in a o es ed a ea may each he op o
he ee o canopy su ace, he in e media e o low e u ns migh be sca e ed by he ee b anches
and lea es o e en he unde s o y, while he las e u n migh each he e ain. This capabili y o
ALS senso s p o ides a eliable 3D ep esen a ion o o es s uc u e. Canopy pene a ion pulse
a ies acco ding o canopy closu e, de e mining he poin s ha each he e ain. Acco ding o
Chasme e al. (2006b) only 50% o las e u ns in o es ed a eas a e backsca e ed by he e ain.
ALS senso s also cap u e, h ough a pho odiode, he ene gy e lec ed by he objec s. This
in o ma ion is denomina ed in ensi y, egis e ed in 8 o 12 bi s. The in ensi y e e s o each lase
echo wi h a oo p in a ying om 0.2 o 1.0 m in small oo p in disc e e e u ns ALS senso s
(Ande sen e al., 2005; E ans e al. 2009). In ensi y alues a ies acco ding o di e en pa ame e s
such as su ace oughness and we ness, ligh heigh , angle o incidence, ins umen al
cha ac e is ics, a mosphe ic condi ions, be ween o he s. Consequen ly, he use o in ensi y alues
equi es he no maliza ion o calib a ion o he da a in each acquisi ion. In his sense, al hough
in ensi y da a has been used o some applica ions, such as species classi ica ion (Ko pela e al.,
2010; Wa e al., 2007), i is s ill no b oadly implemen ed o o es y pu poses.
The main componen s o an ALS disc e e senso a e he pla o m, he lase senso , he Global
Na iga ion Sa elli e Sys em (GNSS), he Ine ial Measu emen Uni (IMU) and he da a manage o
compu e wi h speci ic so wa e. The lase scanne includes he lase pulse ansmi e , he
scanning mechanism and he ecep o o eco d he dis ance o he objec s. The lase scanne emi s
lase pulses, wi h a scanning equency o up o 300 kHz, di ec ed o he e ain su ace. The
scanning mechanism (oscilla ing mi o , o a ing polygon, palme scan, ib e scanne , e c.) d aws
speci ic scanning pa e ns o cap u e he e ain su ace in each ligh line (Vosselman & Maas,
2010). The scanning angle o senso ield o iew (FOV) and he ligh heigh de e mine he s ip wid h.
Acco dingly, he o e lap be ween s ips modi ies he ligh ime and accu acy (E ans e al., 2009).
The GNSS o e uni , loca ed inside o he plane, collec he posi ion om he GNSS sa elli es. The
enhancemen o GNSS posi ion accu acy, pe o med a eal ime using e e ence s a ions o base
di e en ial GNSS a he g ound le el, p o ide a cen ime e nominal accu acy. The IMU includes
he Ine ial Na iga ion Sys em (INS), managing he pi h, oll and yaw o he plane (Bal sa ias,
1999). The use o compu e s o manage he da a collec ed by he GNSS and IMU wi h speci ic
so wa e allows p o iding he x, y and z coo dina es o each e u n pulse along he ligh
acquisi ion, cons i u ing he ALS poin cloud (Bal sa ias, 1999).
1.1.2. Fo es a iable es ima ion using emo e sensing da a
ALS da a ha e been p o en as a sui able echnique o mapping o es e ical and ho izon al
s uc u e as well as o de i e o es y a iables (Zhao e al., 2018). The use o s uc u al and
In oduc ion
5
ex u al in o ma ion de i ed om passi e op ical da a ha e been explo ed o de i e o es y
pa ame e s (Dube & Mu anga, 2015; P ei e e al., 2012). In his sense, he a ailabili y o wide
empo al se ies ha e p o ided be e esul s on o es y a iables es ima ion when cha ac e izing
dis u bance his o y (Cohen e al., 1996; Coops & Wa ing, 2001). Howe e , he da a cap u ed by
passi e op ical senso s end o sa u a e unde closed canopy condi ions and in dense o es s (Lu,
2006). The use o o hopho og aphy o de i e o es y pa ame e s such as s and densi y, heigh o
co e , be ween o he s, has also been explo ed, p o iding lowe accu acies han ac i e senso s
(Campbell, 2006) such as SAR o LiDAR.
SAR sys ems allow cha ac e izing o es s uc u e a global scale. This echnology uses di e en
wa eleng hs om he mic owa e o p o ide in o ma ion abou lea es, b anches and s ems
(Pe iasamy, 2018; Tanase e al., 2014). Al hough SAR da a a ailabili y ha e inc eased and he e
exis ecen ad ances in p ocessing so wa e, as o example he ools p o ided by he Cope nicus
p og am, di icul ies s ill a ise in es ima ing o es y a iables in he e ogeneous and dense o es s
(Hyde e al., 2006). The imp o emen o s uc u e om mo ion (S M) algo i hms and
pho og amme ic echniques p o ides new insigh s in he 3D cha ac e iza ion o o es s uc u e.
In his sense, unmanned ai c a ehicles (UAV) ha e been p oposed o es ima e o es y a iables
(Gianne i e al., 2018; Kachamba e al., 2017; Puli i e al., 2017), cha ac e ize o es uels
(Fe nández-Ál a ez e al., 2019), de ec ion o canopy gaps (Baga am e al., 2018) o ee-s ump
(Puli i e al., 2018) mainly o small geog aphical a eas (Shin e al., 2018). Fu he mo e, he
de i a ion o 3D poin clouds om o opho og aphy may inc ease 3D da a a ailabili y and
imp o e he subsequen es ima ion o o es y a iables, especially hose ha desc ibe canopy
heigh (Noo de mee e al., 2019).
ALS ha e a b oad ange o applicabili y wi hin o es managemen , as o example es ima ion o
o es in en o y a iables (Gue a-He nández e al., 2016a; Mon ealeg e e al., 2016), i e-induced
change quan i ica ion (McCa ley e al., 2017) o cha ac e iza ion o o es s uc u al di e si y (Kane
e al., 2011). ALS poin densi y e e s o he numbe o poin s pe squa e me e , deno ing he
spa ial esolu ion. The e a e wo app oaches wi hin LiDAR li e a u e o es ima e o es a iables:
he indi idual ee-based app oach (ITB) and he a ea-based app oach (ABA) (La i i e al., 2015).
The ITB app oach gene ally in ol es a sequence o ee de ec ion, ea u e ex ac ion, and
es ima ion o ee a iables (Mal amo e al., 2014). Fu he mo e, i also equi es ield
measu emen s a ee le el. Se e al algo i hms ha e been p oposed o ee segmen a ion as well
as ea u e ex ac ion ( as e -based, poin based o mul isou ce-based), p o iding di e en
accu acies unde di e en o es complexi ies (Sačko e al., 2019). The ITB app oach gene ally
equi es poin densi ies highe han 4-5 poin s m-2 (Ande sen e al., 2006).
The ABA app oach was c ea ed by Næsse (1997) and i is also known as he wo phase app oach
in en o y o double sampling in en o y (Næsse , 2002). The i s phase consis on de e mining he
ela ionships be ween ALS me ics and he o es s and a iables es ima ed using ield da a
measu emen s. The second phase i s models ha a e subsequen ly ex apola ed o he whole
s udy a ea. The ABA app oach has been widely implemen ed o he es ima ion o o es y
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
6
a iables (González-Fe ei o e al., 2013; La i i e al., 2010; Noo de mee e al., 2018), being a
sui able app oach o applica ions using low poin densi y da ase s, as he case o he ALS da a
om he Na ional Plan o Ae ial O opho og aphy (ALS-PNOA da a) in Spain.
The e o made by coun ies and o ganiza ions (i.e.: Finland, Spain, Czech Republic) o p o ide
open low densi y ALS da a a egional and na ional scales c ea es new oppo uni ies o o es
managemen . The analysis o he p ocessing me hodology o low densi y ALS-PNOA da a in
Medi e anean o es s has been add essed by Mon ealeg e e al., (2015a and b) who compa ed
se e al il e ing and in e pola ion ou ines o assess he mos sui able me hods o wo k wi hin
Aleppo pine o es ed a eas.
The gene a ion o models equi es he selec ion o he mos sui able ALS me ics and eg ession
me hods. Va iable selec ion, also known as ea u e selec ion, cons i u es a ele an s ep in
modelling gene a ion. The ecen ly inc ease in size o da ase s, wi h ens, hund eds o housands
o a ailable a iables, inc eased he esea ch in e es on selec ion echniques (Guyon & Elissee ,
2003). This g ow h in he a ailabili y o a iables gene a es a dimensionali y p oblem, ypical in
many ields o science (Mehmood e al., 2012), which e e s o he exis ence o a highe numbe o
a iables han samples. These p oblem is also known as la ge p small n p oblem (Ma ens e al.,
1992). The la ge numbe o ALS me ics ha a e de i ed om he poin cloud has subs an ially
inc eased he numbe o a iables in o es y modelling, being some imes e en highe han he
numbe ield plo s sampled.
Dealing wi h la ge ea u e se s p esen s se e al disad an ages such as echnical and model
dec ease o accu acy. Al hough ha dwa e and so wa e ha e imp o ed in he las decades, he use
o a la ge numbe o a iables akes oo many compu a ional esou ces and slows down eg ession
algo i hms. Fu he mo e, in acco dance wi h Koha i & John (1997), he e may be a dec ease o
model accu acy when he numbe o a iables is signi ican ly highe han he op imal.
Conce ning a iable selec ion, a a ie y o me hods exis s in o de o educe he dimensionali y
p oblem as well as o deal wi h la ge ea u e se s. Acco ding o Guyon & Elissee (2003) a iable
selec ion p ocess ha e h ee objec i es: imp o e he p edic ion pe o mance o he p edic o s,
educe ime and cos when de e mining he p edic o s and p o ide a be e unde s anding o he
gene a ed models. A simila de ini ion was p oposed by Peña (2002), who es ablished ha his
selec ion p ocess should ollow he p inciple o pa simony; il e ing o educing he p edic o
a iables in o de o gene a e models as simple as possible, while maximizing hei powe .
Va iable selec ion cons i u es he i s s ep in model i ing. The eg ession analysis cons i u es he
p ocess o es ima e o model a ela ionship be ween a iables. This analysis can be ca ego ized in
wo main ypes: pa ame ic and non-pa ame ic eg ession. The classical eg ession app oach is
he pa ame ic one, which assumes he exis ence o a ini e se o pa ame e s. On he con a y, non-
pa ame ic me hods conside ha da a dis ibu ion canno be de ined by a ini e se o pa ame e s.
Pa ame ic eg ession elies on s ong assump ions such as no mali y, homoscedas ici y, linea i y,
independence, no-collinea i y and absence o a ypical alues. In non-pa ame ic models, mee ing
In oduc ion
7
hese equi emen s is no necessa y, u ning hem in o a mo e lexible ool. The e exis s a wide
a ie y o pa ame ic and non-pa ame ic me hods; om simple linea eg ession models o
complex neu al ne wo ks models. Acco ding o Hazel on (2015), non-pa ame ic me hods a e
di ided in ke nel and local polynomial eg ession, spline-based eg ession o eg ession ees.
The use o machine lea ning algo i hms, showed good pe o mance in se e al esea ch ields as
da a mining. Machine lea ning algo i hms, de ined as algo i hms and s a is ical models ha do no
equi e speci ic ins uc ions and whose o m o he unc ion is unknown, a e no di ec ly
associa ed o he wo es ablished ypes o eg essions. Pa ame ic eg ession ha e been
adi ionally used in o es y o s and a iable p edic ion wi h ALS da a (Penne e al., 2013).
Recen ly, he applica ion o non-pa ame ic and machine lea ning me hods o his opic has
inc eased popula i y (Bollandsås e al., 2013b; Chi ici e al., 2008; Liaw & Wiene , 2002).
Model accu acy a ies acco ding o se e al ac o s such as ield and ALS da a cha ac e is ics, o es
complexi y, a iable selec ion and eg ession me hod used, be ween o he s. In his sense, ALS
ligh con igu a ion de e mines he inal da a cha ac e is ics, as he quali y o he poin cloud,
a ec ing model pe o mance. ALS senso s a e con igu ed wi h di e en scanning pa e ns, pulse
equencies, scanning equencies, scanning angles and beam di e gence. This con igu a ion,
summed up o he ligh al i ude and speed p oduce di e en oo p in sizes and poin densi ies.
All hese se ings a e no mally di e en om one ligh o ano he .
Se e al au ho s ha e explo ed he e ec o some o hese se ings on he p edic ion e o in o es
a ibu es modelling. Yu e al. (2004) concluded ha an inc ease in ligh al i ude dec eases
accu acy in ees de ec ion and ee heigh p edic ion, a ec ing mo e o deciduous species han
coni e ous ones. The e ec o oo p in size has ecei ed li le a en ion in he li e a u e, howe e
se e al au ho s ag eed ha , in small oo p in applica ions, he la ges oo p in s gene a e a highe
bias in he p edic ion o ee heigh s (Ande sen e al., 2006; Hi a a, 2004; Roussel e al., 2017). An
inc ease o pulse equency, de ined as he numbe o pulses pe second and exp essed in kilohe z
(kHz), gene a es a lowe pene a ion o he pulses h ough he canopy (Chasme e al., 2006b;
Næsse , 2009), implying a dec ease in ee heigh p edic ion accu acy (Chasme e al., 2006a). The
e ec o scan angle is ela i ely low up o ~20°, p oducing conside ably highe p edic ion e o s in
o es y me ics wi h highe alues (Liu e al., 2018; Mon aghi, 2013). Acco ding o Ande sen e al.
(2006), he beam di e gence also a ec ee heigh p edic ions; a dec ease in he angle inc eases
accu acy. Fu he mo e, he analysis o he e ec o poin densi y de e mines ha a dec ease o his
ac o gene ally does no p oduce a dec ease in accu acy (Ga cia e al., 2017; Roussel e al., 2017).
Finally, some en i onmen al a iables such as slope ha e been conside ed a sou ce o e o in ALS
p ocessing in o es ed a eas. The p esence o s eep slopes gene ally inc ease he e o s in poin
cloud il e ing and in e pola ion p ocesses (Mon ealeg e e al., 2015b).
Fo es s uc u e e e s o he size, shape, and ho izon al and e ical dis ibu ion o lea es,
b anches and s ems. These cha ac e is ics a y along ime, being a ec ed by na u al o an h opic
dis u bances. Fi e is caused by na u al ac o s such as olcanic e up ions o ligh ning and has
his o ically ans o med he landscape. T adi ional human ac i i ies used i e o manage di e en
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
8
land uses, p o iding an an h opogenic dimension o wild i es. These dis u bances cons i u e some
o he mos impo an socio-en i onmen al haza ds in Medi e anean o es ecosys ems, ha migh
be enhanced by clima e change, since ex eme me eo ological condi ions o long d ough s inc ease
he i e isk (González-de Vega e al., 2016; Sebas ián-López e al., 2008). Al hough s a is ic
egis e s showed a educ ion in he numbe o i e e en s du ing he las decade (2001-2010), he
occu ence o la ge i es (>500 ha) in Spain has inc eased (Rod igues e al., 2014).
Aleppo pine (Pinus halepensis Mille ) is a lammable species, equen ly a ec ed by wild i es,
cha ac e ized by a high s and densi y and a con inuous p esence o b anches along he s em
(Pausas e al., 2008). Pine o es s ha e a high esilience o i e, bu hei egene a ion p ocess migh
ail when i e ecu ence is high. Fo es i es ha e impo an e ec s in ege a ion dynamic and
a mosphe e, as may emi la ge quan i ies o g eenhouse gases (GHGs) and ep esen an impo an
ca bon sink (Akagi e al., 2013; an de We e al., 2010; Wiedinmye e al., 2011). The
quan i ica ion o wild i e ca bon dioxide (CO2) emissions is i al o clima e egula ion policies
(Mie ille e al., 2010) as well as o highligh ing he se ice ha o es s p o ides o socie ies (Lal,
2008). The es ima ion o i e GHGs emissions equi es p e- i e biomass es ima ion , he assessmen
o he ac ion o biomass consumed by i e, usually ela ed o i e se e i y, and, subsequen ly, he
use o con e sion ac o o es ima e GHG emissions (De San is e al., 2010). The use o passi e
emo e sensing o es ima e i e se e i y ha e been b oadly analysed in he li e a u e (Ga cía-
Llamas e al., 2019). Thus, di e en indexes ha e been p oposed o accoun o i e damage in
ege a ion as No malized Bu n Ra io (Key & Benson, 2005), Rela i e del a No malized Bu n Ra io
(Mille & Thode, 2007), be ween o he s. Fu he mo e, he use o ALS da a o quan i y biomass
ha e been es ed o di e en ecosys ems (Ga cía e al., 2010; Næsse , 2011).
The es ima ion o o es y in en o y a iables is one o he mos ele an applica ion o o es y
pu poses (La i i e al., 2010; Mon ealeg e e al., 2016). As men ioned abo e, o es ecosys ems
cons i u e impo an ca bon sinks and play a majo ole in managing GHGs emissions. In his
sense, he es ima ion o biomass and ca bon con en has g owing in e es . Se e al s udies ha e
explo ed he es ima ion o abo eg ound ee biomass using ALS da a in di e en ecosys ems
(Gue a-He nández e al., 2016b; Mauya e al., 2015). Howe e , he es ima ion o some biomass
ac ions, such as sh ub biomass o o es esidual biomass, ha e been less s udied (Es o nell e al.,
2012; Hauglin e al., 2014). The p esence o unde s o y in o es ed a eas and he exis ence o
sh ubland a eas a e e y common in he Medi e anean basin land co e . Thus, he quan i ica ion
o hese biomass ac ions may imp o e he accoun o ca bon ese oi s. Fu he mo e, he use o
some biomass ac ions o bioene gy pu poses may educe he CO2 emissions o he a mosphe e
p oduced by o he uels and migh help o each he clima e and ene gy a ge s o he Eu opean
Ene gy Roadmap (Hamelin e al., 2019). Finally, he managemen o hese ac ions ha e se e al
bene i s o u al de elopmen , as he educ ion o wild i e isk o he eme gence o new business
oppo uni ies o o es land owne s (Hauglin e al., 2012).
ALS da a p o ides a wide ange o applica ions when mul i- empo al da a is a ailable. Fo es
manage s could use mul i- empo al ALS da a o applica ions such as: cha ac e izing na u al o
In oduc ion
9
an h opic changes, de e mining unde sampled a eas and selec i ely add plo s o u u e
in en o ies, applying exis ing ALS-based models o subsequen acquisi ions in o es s wi h simila
cha ac e is ics, educing ield wo k (Feke y e al., 2015) and imp o ing he accu acy o long pe iod
o es y end analysis using usion o ac i e and passi e op ical da a. Despi e he g ea po en ial
o mul i- empo al analysis, i s applica ion is s ill limi ed by he acquisi ion cos s as well as he need
o empo al-concomi an ield da a (Cao e al., 2016; Dubayah e al., 2010; Fe az e al., 2018).
The ecen e o made by coun ies and o ganiza ions o p o ide mul i- empo al da ase s c ea es
new oppo uni ies o upg ading o es in en o ies a local and egional scales. Local scale e e s o
small a eas whose s and cha ac e is ics a e simila while egional scale de e mine la ge a eas wi h
highe s and and s uc u al a iabili y. In addi ion, se e al au ho s ha e es ima ed heigh g ow h
(Ga ziolis e al., 2010; Socha e al., 2017) as well as biomass and ca bon dynamics (Hudak e al.,
2012; Poudel e al., 2018). The es ima ion o olume (Næsse & Gobakken, 2005; Poudel e al., 2018;
Yu e al., 2008), basal a ea (Næsse & Gobakken, 2005) and si e index (Noo de mee e al., 2018) as
well as he quan i ica ion o wild i e changes (McCa ley e al., 2017) and gaps p esence
(Vepakomma e al., 2008) o he analysis o de olia ion e ec (Solbe g e al., 2006) ha e also been
pe o med. Two app oaches, di ec and indi ec , ha e been p oposed o model o es a ibu es
using mul i- empo al ALS da a o e ime (Bollandsås e al., 2013). The di ec app oach i s one
model o one poin in ime and empo ally ans e s he model o o he poin in ime. The indi ec
one i s wo di e en models o each poin in ime. The explo a ion o hese app oaches p o ides
use ul in o ma ion o o es manage s o he educ ion o ield da a acquisi ions (Noo de mee e
al., 2018).
1.2. Impo ance and jus i ica ion
LiDAR echnology, and speci ically ai bo ne lase scanne s, has become a aluable sou ce o 3D
in o ma ion o cha ac e ize o es ecosys ems. Di e en p oduc s can be de i ed using ALS da a,
om he gene a ion o p ecise DTM and digi al su ace models (DSM) o he cha ac e iza ion o
o es s ands a iables. Thus, he combina ion o ALS da a wi h ield wo k as well as wi h da a
om o he emo e sensing sou ces p o ides accu a e in o ma ion o quan i y and e alua e o es
esou ces.
Wild i es cons i u e a socio-en i onmen al haza d in Medi e anean ecosys ems. These e en s a e
caused by na u al o an h opogenic ac o s and cons i u e a ele an sou ce o g eenhouse gases
emission o he a mosphe e. Aleppo pine, being a py ophy e species, is one o he mos a ec ed by
i es in Spain. The es ima ion o i e emissions equi es he quan i ica ion o p e- i e biomass and
he ac ion o biomass consumed by he i e. The adi ional es ima ion o p e- i e biomass was
pe o med wi h ield da a campaigns, while he es ima ion o pos - i e biomass was based in
isual examina ion o ield-based weigh ing. In his sense, he use o emo e sensing ools o
de e mine i e se e i y, and subsequen ly, bu n e iciency has been widely analysed. The
capabili ies o ALS da a o desc ibe o es s uc u e and quan i y o es biomass and he
ad an ages p o ided by i s usion wi h op ical passi e da a, migh enhance quan i ica ion o
g eenhouse gases emissions sou ced om wild i es.
17
2. S udy a ea, ma e ials and me hods
This chap e desc ibes he s udy a ea, which includes ou
di e en zones wi hin A agón egion, as well as he ma e ial
and me hods u ilized in he esea ch. Fi s ly, we desc ibe he
ield in en o y da a, he allome ic equa ions used o
compu ing he di e en o es a iables and he p e-
p ocessing pe o med o he da a. Secondly, he emo e
sensing in o ma ion used; he ALS-PNOA da a and passi e
op ical images, a e p esen ed, including he p e-p ocesses
applied. Thi dly, we include he selec ion and eg ession
me hods u ilized o modelling di e en a iables a s and
le el. Then, we de ine he me hods o analyse he empo al
ans e abili y o mul i- empo al ALS da a. Fu he mo e, he
me hods u ilized o assess he e ec o ALS pa ame e s and
en i onmen al condi ions in model accu acy a e p esen ed.
Finally, he mapping p ocess and i s impo ance in he
gene a ion o in o ma ion a local o egional scale is
add essed.
S udy a ea, ma e ials and me hods
19
2.1. S udy a ea
Aleppo pine is he mos b oadly dis ibu ed species om genus Pinus in he Medi e anean basin.
The egions wi h highe p esence a e he no h o A ica and Spain, which ep esen s ~850,000 ha,
acco ding o Cáma a (2001). The wide al i udinal and la i udinal g adien o his dis ibu ion
allows his species o li e om sea le el up o 1,600 m in he Saha an A las (Cabanillas 2010).
Al hough Aleppo pine is a limes one species, i g ows in di e en ypes o soils such as siliceous,
qua zi e o g ani e. Fu he mo e, he species is heliophilous, he mophile, xe ophile and
py ophy e, being adap ed o d ough s and wild i es.
This PhD Thesis s udied he Aleppo pine o es o A agón egion. A agón is an Au onomous
Communi y loca ed in he no heas o Spain. This egion occupies 47,720.3 km2 and ep esen s
9.4% o he Spanish e i o y. Th ee p o inces; Za agoza, Huesca, and Te uel, cons i u e his
Au onomous Communi y. A agón limi s o he no h wi h F ance, o he eas wi h Ca aluña and
Valencia and o he wes pa wi h Cas illa-La Mancha, Cas illa y León, La Rioja and Na a a.
Th ee main elie uni s compose A agón, he Py enees, he Eb o Basin and he Ibe ian Ranges. The
Py enees a e ep esen ed by he Axial Py enees, he in e io moun ains, he In apy enean
opog aphic dep ession and he ex e io moun ains (p e-Py enees). Axial Py enees, cons i u ed by
g ani es, qua zi e, sla es and limes one, p esen s he highes al i udes such as Ane o (3,404 m) o
Malade a (3,308 m). The in e io moun ains include calca eous c es s adhe ed o he axial
Py enees. The In apy enean opog aphic dep ession con ains se e al pe pendicula i e alleys,
ending in “San Juan de la Peña” and “Peña O oel” conglome a ic esca ps. The p e-Py enees,
cons i u ed by calca eous ocks, p esen heigh s be ween 1,500 and 2,000 m (Peña & Lozano, 2004).
The “Somon anos” connec he p e-Py enees wi h he le bank o he Eb o Basin. The Eb o Valley
was a sea du ing Mesozoic and Eocene s ages. Nowadays i is a opog aphic dep ession, co e ed
by e ia y ma e ials and allu ial sedimen s. The e osion p ocesses ha e gene a ed di e en
abula elie s om 500 up o 800 m in bo h banks o Eb o Ri e .
The Ibe ian Ranges is a moun ain chain o lowe al i ude han Py enees. The “Sie a del Moncayo”
cons i u es he no hwes pa , “Pue os de Becei e” and “Gúda -Maes azgo” a e loca ed in he
eas e n pa , while “Ja alamb e” and “Alba acín” in he sou heas e n pa . Qua zi e and
Palaeozoic sla e a e he main ma e ials p esen in he highe moun ains, while Ju assic and
C e aceous limes one and dolomi es cons i u e he lowe elie s uc u es.
The clima e o A agón is Medi e anean wi h con inen al ea u es, cha ac e ized by cold win e s,
d y summe s and i egula and sca ce ain all. The a iabili y in o og aphy modi ies he
empe a u es and p ecipi a ions, gene a ing a wide di e si y o clima ic ambien om semi-
dese ic a eas such as Moneg os o high moun ains in he Py enees. Acco ding o Cuad a (2004),
A agón clima e is de ined by ou main cha ac e is ics. The Eb o alley p esen s low annual
p ecipi a ion alues, ~300 mm m2, gene a ed by he shadow e ec o Py enees and Ibe ian Ranges;
i s loca ion in a con inen al a ea gene a es a b oad ange o empe a u es; p ecipi a ions a e
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
20
i egula and he winds come om he no hwes in win e and sou heas in summe , being
equen and hea y.
A agón includes wo Hola c ic biogeog aphical egions: Eu osibe ian and Medi e anean. The
Eu osibe ian egion is occupied by o es s and g asslands dis ibu ed in h ee al i udinal s a a:
alpine, subalpine, and mon ane. The Medi e anean egion includes he Eb o Valley, he
“Somon anos”, he i e alleys loca ed in he igh ma gin o Eb o Ri e and he opog aphic
dep ession in which Te uel ci y is si ua ed. Que cus ilex, Pinus halepensis, Pinus nig a and Junipe us
sabina cons i u e he species ha domina e Medi e anean o es s (Longa es, 2004). Acco ding o
he Spanish Fo es Map, he o es ed a ea in A agón ep esen s 1.58 million o ha, o which
259,057.45 ha a e occupied by Aleppo pine o es s. Conc e ely, 124,473.12 ha a e loca ed in
Za agoza, 37,817.19 ha in Huesca and 96,767.14 ha in Te uel egions. In he whole, semi-na u al
o es s ep esen 211,013.49 ha and a o es ed o es 48,043.96 ha.
As men ioned be o e, ou s udy zones we e delimi ed o p o ide answe s o he speci ic esea ch
objec i es (Figu e 1). The s udy zones a e desc ibed below:
Zone A is loca ed in “Las Cinco Villas” egion, no hwes o A agón. The elie is
cha ac e ized by a opog aphic dep ession close o he p e-Py enees. Ele a ions ange om
430 o 1150 m abo e sea le el and slopes om 0° o 39°. The clima e is Medi e anean wi h
con inen al ea u es wi h an annual a e age p ecipi a ion o 525 mm. The Zone A includes
wo di e en es a eas: “Luna” wild i e and ield in en o y 3 (Figu e 3). Luna wild i e was
caused by ag icul u al machine y on 4 h July 2015. The i e sco ched 14,263 ha, 3,390.4 ha
co e ed by woodland. Field in en o y 3 is loca ed in an unbu ned a ea close o he
wild i e, wi h simila en i onmen al, clima ic and o es cha ac e is ics. The unbu ned
Aleppo pine o es is he e ogeneous om he s uc u al poin o iew, being accompanied
by an e e g een unde s o ey wi h species such as Que cus ilex subsp. o undi olia, Que cus
cocci e a, Junipe us oxyced us, Buxus sempe i ens and Junipe us phoenicea.
Zone B is loca ed in he Eb o Basin, No heas Spain. This zone includes wo di e en es
a eas: in en o y 2 (Figu e 3) and in en o y 3. Bo h a eas a e ep esen a i e o Aleppo pine
Medi e anean o es , occupying 11,400 ha. The in en o y 3 a ea was desc ibed abo e. The
majo i y o In en o y 2 plo s a e loca ed in he mili a y aining cen e “San G ego io”
(CENAD), loca ed in he no h o Za agoza ci y. The a ea p esen s a hilly opog aphy wi h
ele a ions anging om 300 o 750 m and slopes om 0° o 39°. Clima e is Medi e anean
wi h con inen al ea u es, while he a e age annual p ecipi a ion is lowe han 350 mm.
Mos o he pine s ands a e semi-na u al, while he s ands loca ed in he sou h eas e n pa
we e plan ed app oxima ely o y yea s ago. The e e g een unde s o ey is cha ac e ized by
xe ophiles species such as Que cus cocci e a, Junipe us oxyced us, Rosma inus o icinalis and
Thymus ulga is.
S udy a ea, ma e ials and me hods
21
Figu e 3. Loca ion o o es in en o y campaigns. High spa ial esolu ion o hopho og aphy om Spanish
Na ional Plan o Ae ial O hopho og aphy spa ial da a in as uc u e (SDI) included as backd op.
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
22
Zone C is loca ed in he Eb o Basin and Ibe ian Ranges, including a b oad pa o Aleppo
pine o es in A agón egion, excep om he o es ed a eas close o he p e-Py enees and
he s ands loca ed in he sou h o Te uel. This zone includes h ee di e en es a eas:
in en o y 1 (Figu e 3), in en o y 2 and mos o he plo s om in en o y 4. In en o y 2 was
desc ibed p e iously. In en o y 1 is loca ed in Da oca Municipali y in he Ibe ian Range.
The a ea includes wo o es s denomina ed “Dehesa de los eneb ales” and “Valdá y
Ca ilanga”, co e ing 1,102 ha. Bo h o es s we e a o es ed om 1908 o 1979, being
occupied by a monospeci ic Aleppo pine o es wi h sca ce unde s o y. The a ea p esen s a
hilly opog aphy, wi h ele a ions anging om 860 up o 980 m. In en o y 4 includes
se e al s ands om he Middle Eb o Basin up o he Ibe ian Ranges. The s ands we e
a o es ed app oxima ely o y o six y yea s ago, keeping a low p esence o ha dwood
species. The di e en en i onmen al sample condi ions include a g ea a iabili y o
Aleppo pine o es s.
Zone D ep esen s 197,951.24 ha o he Aleppo pine o es ed a ea o A agón, including all
he species dis ibu ion ange. The Zone D includes ou di e en es a eas: in en o ies 1
o 4 (Figu e 3), which ha e been p e iously desc ibed. The b oad a ea shows a a iabili y
o geomo phological o ms om opog aphic dep essions up o moun ains ha each mo e
han 2,000 m abo e sea le el. Consequen ly, empe a u e changes ac oss he al i udinal
g adien and he annual p ecipi a ion ange om less han 350 mm up o 1,000 mm. This
a iabili y and he p esence o semi-na u al and a o es ed s ands a e cha ac e is ic o
Aleppo pine o es a A agón egion.
2.2. Ma e ials and me hods
2.2.1. Field in en o y da a
Field in en o ies p o ide an accu a e quan i ica ion o o es a iables wi hin a small ac ion o
he s udy a ea. Field plo in o ma ion is he g ound- u h e idence, cons i u ing he dependen
a iables ha a e es ima ed o a b oad a ea. In his sense, ield plo da a mus be ep esen a i e
o he s udy a ea, cap u ing he maximum a iabili y o minimize ex apola ion e o s.
Sampling design is a ele an ac o o cap u e o es s and a iabili y. Se e al sampling ypes
exis , such as sys ema ic sampling, andom sampling o s a i ied andom sampling. T adi ional
in en o ies, pe o med using ield campaigns, equi e he di ision o o es s ands in o
homogeneous s a a, conside ing ac o s such as ee species, si e p oduc i i y, o es de elopmen
s age, la i ude, ele a ion and s and s uc u e. Thus, pe o ming in en o ies using ALS da a is
easie as his echnology p o ides in o ma ion abou some o hose ac o s a a s and le el. This
PhD Thesis applied a s a i ied andom sampling, conside ing e ain slope, canopy heigh and
canopy co e a iabili y, acco ding o Næsse & Økland (2002), in monospeci ic Aleppo pine
o es . Field da a was acqui ed in 192 plo s in ou campaigns pe o med du ing 2013, 2014, 2015
and 2016 (he eina e men ioned as i s , second, hi d and ou h campaign, espec i ely). Field
da a we e ela ed o ALS da a using an a ea-based app oach (Næsse & Økland, 2002). The
S udy a ea, ma e ials and me hods
23
numbe o plo s allowed he es ima ion o o es s and a iables a local and egional scales. The
speci ica ions o each in en o y a e desc ibed below:
The acquisi ion o ield da a om he i s campaign was pe o med om June o July 2013,
in 53 ci cula plo s wi hin he Mas e Thesis o Jesús Cab e a (Cab e a, 2013). The cen e
poin o each ci cula plo wi h 15 m adius was posi ioned using a Leica VIVA® GS15
CS10 eal- ime kinema ic GNSS wi h a planime ic accu acy o 0.30 m. We used a diame e
ape, wi h millime e p ecision, o measu ing ee diame e a b eas heigh (dbh) in hose
ees wi h a dbh la ge han 7.5 cm, which is he s anda d dbh o in en o ied ees in
Spain. A Suun o® hypsome e was used o measu ing g een c own heigh and ee heigh
o up o 4 andomly selec ed ees wi hin each plo . The selec ion o he sample ees, om
7.5 cm up o 42.5 cm, conside s he diame ic classes de ined as ep esen a i e o he s udy
a ea in he hi d na ional o es in en o y. The heigh o hose ees no measu ed in he
ield plo s was p edic ed by using a heigh -diame e model de eloped om he sampled
ees (equa ion 1). The model pe o mance o he heigh model ga e a RMSE o 1.36 m and
R2 o 0.63. No mali y, homoscedas ici y and independence o no au o-co ela ion in he
esiduals we e e i ied o he i ed model.
ℎ𝑡=0.776·𝐺0.179·𝑑𝑏ℎ𝑖0.660·1.009 (1)
whe e h is ee heigh (m), dbhi is he diame e a b eas heigh (cm) and G is ield plo basal
a ea (m2 ha-1).
The second and hi d ield campaigns use he same in en o y me hodology. The second
campaign includes 43 ci cula plo s acqui ed om July o Sep embe 2014 (Mon ealeg e e
al., 2016). The hi d ield plo campaign sampled 45 plo s om June o July 2015, being used
o objec i es 1, 2 and 4. The same GNSS ins umen used o he i s campaign posi ioned
he 30 m diame e plo s, ob aining a planime ic accu acy o 0.15 and 0.18 m in 2014 and
2015, espec i ely. A Haglö Sweden® Man ax P ecision Blue diame e callipe allowed
measu ing he dbh o hose ees wi h a dbh la ge han 7.5 cm. We used a Haglö Sweden®
Ve ex ins umen o measu e he g een c own heigh and he heigh o all ees in he
plo . Fu he mo e, he pe cen age o sh ub canopy co e and he a e age heigh o he
di e en sh ub species ha ep esen he unde s o y we e measu ed (used o objec i e 2).
The ou h campaign in en o ied 51 ield plo s in Ap il 2016 being ca ied ou by ö a
o es echnologies wi hin he p ojec RF-64079 (Ru al De elopmen P og am o A agón
2014–2020). A T imble subme ic GNSS was used o posi ion he cen e o each plo wi h a
subme ic accu acy in planime y. A a iable plo adius was selec ed (5.6 m, 8.5 m, 11.3 m,
and 14.10 m), in o de o ob ain da a om a simila numbe o ees in each plo , due o he
di e ence in s and densi y. A Haglö Sweden® Man ax P ecision Blue diame e callipe
allowed measu ing hose ees wi h a dbh la ge han 7.5 cm. A Haglö Sweden® Ve ex
was used o measu ing he g een c own heigh and he heigh o up o 6 ees, he nea es
o he plo cen e . The sample was comple ed o achie e 100 dominan s ems ha-1,
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
24
conside ing hose wi h la ge dbh. The heigh o hose ees no measu ed in he ield was
es ima ed by using a heigh -diame e model de eloped om he sampled ees (equa ion
2). The model pe o mance o he heigh model ga e a RMSE o 0.80 m and R2 o 0.93.
ℎ𝑡=(1.32.5511+(𝐻02.5511−1.32.5511)·1−𝑒𝑥𝑝(−0.025687·𝑑𝑏ℎ)
1−exp(−0.025687·𝐷0))12.5511
⁄ (2)
whe e h is ee heigh (m), dbh is he diame e a b eas heigh (cm), Ho is he Assmann
dominan heigh (m) and Do is he Assman dominan diame e (cm).
2.2.2. Es ima ion o o es s and a iables
The es ima ion o o es a iables a ee le el equi es he use o allome ic equa ions. These
equa ions gene ally need a des uc i e sampling, d ying and subsequen weigh ing o a
ep esen a i e sample. The majo i y o ee o es species ha e an allome ic equa ion. Howe e ,
he accessibili y o sh ub allome ic equa ions in Spain is mo e limi ed, being a ailable o he
es ima ion o some speci ic a iables such as biomass (Mon e o e al., 2013). This PhD Thesis
es ima ed nine s and a iables using allome ic equa ions as a g ound- u h. The es ima ed
a iables a e: s and densi y (N), basal a ea (G), squa ed mean diame e (Dg), dominan diame e
(Do), dominan heigh (Ho), imbe olume o e ba k o s em (V), abo e g ound biomass (Wag),
o al ee biomass (W) and o es esidual biomass (FRB). In addi ion, he equa ions de e mined by
Mon e o e al. (2013) o di e en sh ub o ma ions de ined in he Spanish Fo es Map (MFE) we e
used o calcula e sh ub biomass. Finally, o al biomass including abo eg ound ee biomass and
sh ub biomass (TW) was compu ed by summing up o al ee biomass and sh ub biomass.
T ee equa ions
The equa ions 3 o 16 we e used o es ima e he abo e men ioned ee a iables.
𝑁 (𝑠𝑡𝑒𝑚𝑠 ℎ𝑎−1)=𝑠𝑡𝑒𝑚𝑠
𝑎 (3)
𝐺 (𝑚2 ℎ𝑎−1)=∑𝜋
4∙𝑑𝑖2𝑛
𝑖𝑎 (4)
𝐷𝑔 (𝑐𝑚)=100·√4∙𝐺
𝜋∙𝑁 (5)
𝐷𝑜=∑𝑑𝑖
𝑘
𝑖
𝑎/100 (6)
𝐻𝑜=∑ℎ𝑖
𝑘
𝑖
𝑎/100 (7)
𝑣𝑖=𝜋
40000∫[(1+1.121163·𝑒(−10.23293·ℎ𝑖
ℎ𝑡))·0.696362·𝑑𝑖
ℎ𝑡
0·((1−ℎ𝑖
ℎ𝑡)1.266261−(0.003553∗𝐸)−1.865418·(1−ℎ
ℎ𝑡))]2𝑑𝑖ℎ𝑖
𝑉 (𝑚3ℎ𝑎−1=10000·∑𝑣𝑖
𝑛
𝑖𝑎 (8)
S udy a ea, ma e ials and me hods
25
Wag (kg ℎ𝑎−1)=𝐶𝐹∙𝑒𝑎∙𝑑𝑖𝑏
𝑎∙10,000 (9)
Ws (kg)=0.0139 ∙ 𝑑𝑖2∙ℎ𝑖 (10)
Wb7 (kg)= [3.926∙(𝑑𝑖−27.5)]∙Z; (11)
I 𝑑𝑖≤27.5 cm hen Z=0;I 𝑑𝑖>27.5 𝑐𝑚 𝑡ℎ𝑒𝑛 𝑍=1
Wb2−7 (kg)=4.257+0.00506∙𝑑𝑖2∙ℎ𝑖−0.0722∙𝑑𝑖∙ℎ𝑖 (12)
Wb2+n (kg)=6.197+0.00932∙𝑑𝑖2∙ℎ𝑖−0.0686∙𝑑𝑖∙ℎ𝑖 (13)
W (kg)=0.0785∙𝑑𝑖2 (14)
W (kg)=𝑊𝑠+𝑊𝑏7+𝑊𝑏2−7+𝑊𝑏2+𝑛 (15)
FRB (kg)=𝑊𝑏7+𝑊𝑏2−7+𝑊𝑏2+𝑛 (16)
whe e d is he no mal diame e in cm; a is he a ea o he plo exp essed in m2; ∑n e e s o he
numbe o ees inside o a plo ; ∑k e e s o he numbe o k ees being k he hickes ees; hi is he
heigh o he ees; h is he o al ee heigh ; E is he ee slende ness (h /di); i is he olume o each
ee in m3; CF is a co ec ion ac o (𝐶𝐹=𝑒𝑆𝐸𝐸2/2) being e he Eule numbe and SEE he s anda d
e o (0.151637); a is (−2.0939) and b (2.20988) a e he speci ic pa ame e s o Aleppo pine; Ws is he
biomass weigh o he s em ac ion, Wb7 is he biomass weigh o he hick b anch ac ion
(diame e la ge han 7 cm), Wb2−7 is he biomass weigh o medium b anch ac ion (diame e
be ween 2 and 7 cm), Wb2+n is he biomass weigh o he hin b anch ac ion (diame e smalle han
2 cm) wi h needles, and W is he biomass weigh o he oo s.
Sh ub equa ions
The sh ub equa ions used o each o ma ion a e p esen ed below (equa ions 17 o 21):
Sh ub hedges, bo de s, galle ies, e c.:
ln(W𝑠)=0.494×ln (CC) (17)
Que cus cocci e a and Pis acia len iscus:
ln(W𝑠)=−2.892+1.505×ln(hm)+0.462× ln (CC) (18)
Leguminosae aulagoideas and ela ed sh ubs:
ln(W𝑠)=−2.464+0.808×ln(hm)+0.761× ln (CC) (19)
Labia ae and Thymus o ma ions:
ln(W𝑠)=−1.877+0.643×ln(hm)+0.661× ln (CC) (20)
Gene al sh ub biomass:
ln(W𝑠)=−2.560+1.006×ln(hm)+0.672×ln (C) (21)
whe e Ws is he biomass weigh o each species in ons/ha, hm is he a e age sh ub heigh a plo
le el and CC is he pe cen age o sh ub canopy co e a plo le el.
Equa ion 17 was applied o C a aegus monogyna, Rhamnus lycioides and Rosa canina; equa ion 18
was used o Que cus cocci e a; equa ion 19 was applied o Genis a sco pius; equa ion 20 was used
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
32
GPS ime indica es he da e and ime o GNSS lase poin egis a ion when he plane is
lying. The ime is exp essed in seconds.
Table 4. Technical speci ica ions o LiDAR-PNOA plan o A agón egion om he i s and second
co e age.
Cha ac e is ic
Desc ip ion
Fi s co e age
Second co e age
Senso
Leica ALS-50
Leica ALS-80
Geode ic e e ence sys em
ETRS89
Ca og aphic p ojec ion
Uni e sal T ans e sal Me ca o (UTM) 30 and 31
Geoid model
EGM2008-REDNAP
Field o iew (FOV)
The maximum allowed FOV is 50°
Scanning equency
Minimum o 70 Hz and up o 40 Hz wi h a FOV o 50°
Pulse equency
Minimum o 45 kHz wi h a FOV o 50° and up o 3,000 m
Poin densi y
0.5 poin s m-2 implying a spacing be ween poin s ≤1.41 m
Radiome ic esolu ion o
mul iple in ensi ies
Dynamic ange wi h a leas 8 bi s
Abili y o cap u e mul iple e u ns
o he same pulse
Up o 4 e u ns pe pulse when he e ical dis ance is
highe han 4 m
GNSS na iga ion sys em
Double equency GNSS wi h a leas 2 Hz
Ine ial sys em (IMU/INS)
Da a egis e equency ≥200Hz and d i <0.1° h-1
Plane eloci y when LiDAR da a
cap u ing
Va iable
Fligh heigh
Va iable
Maximum ligh line leng h
90 km
Ho izon al and e ical p ecision
a e p ocessing
The ho izon al global p ecision a nadi ha e a RMSEx,y <30
cm (1 sigma), while he e ical global posi ion a nadi ha e
a RMSEz <20cm (1 sigma). E o s up o 3xRMSE in dense
ege a ion o s eep slopes migh occu . The edge o ligh
line migh p esen an e o up o 2xRMSE
Al ime y p ecision and maximum
e o
≤0.40 m o 95% o he cases. Poin s canno ha e an e o
highe han 0.60 m
Al ime y disc epancies be ween
ligh lines
≤0.40 m
Poin cloud o ma
*.las 2x2 km iles
LAS ile classi ica ion
Au oma ically classi ied
(g ound, ege a ion,
buildings, o e lap)
Non classi ied
S udy a ea, ma e ials and me hods
33
Da a om bo h ALS-PNOA co e ages a e u ilized in his PhD Thesis. In his sense, mo e han
4,000 *.las iles om he i s co e age we e downloaded om he CNIG webpage. Fu he mo e,
he Geog aphic Ins i u e o A agón (IGEAR) and CNIG p o ided 147 iles om he second
co e age be o e being included in CNIG pla o m.
The ligh echnical speci ica ions, he poin cloud cha ac e is ics and he ile nomencla u e is
included in Table 4, Table 5 and Figu e 5, subsequen ly.
Table 5. Cha ac e is ics o LiDAR-PNOA da a acco ding o he s udy a ea.
Cha ac e is ics
Zone A
Zone B
Zone C
Zone D
Acquisi ion
da e
Oc obe 2010
( i s co e age)
Oc obe 2010,
Janua y 2011 and
Feb ua y 2011
( i s co e age)
July 2010 o
Feb ua y 2011
( i s co e age)
July 2010 o Feb ua y
2011 ( i s co e age)
Sep embe o No embe
2016 (second co e age)
Senso
Leica ALS50
Leica ALS50
Leica ALS50
Leica ALS50 ( i s
co e age)
Leica ALS80 (second
co e age)
A e age ligh
heigh (m)
3146
3022
3240
3138 ( i s co e age),
2943 (second co e age)
A e age plane
eloci y (km/h)
241
241
184
184 ( i s co e age), ~240
(second co e age)
Numbe o iles
42
690
3800
147 ( i s co e age) and
147 (second co e age)
Figu e 5. Nomencla u e o LiDAR-PNOA iles.
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
34
2.2.5. ALS p e-p ocessing and me ic compu a ion
Quali y con ol o ALS poin clouds is equi ed o accu a e p e-p ocess he da a. The ALS da a
p o ided by he PNOA is no aw da a, bu use s should check he quali y o downloaded poin
clouds be o e pe o ming any u he analysis. The i s ALS p ocessing s ep conside ed was he
emo al o noise e u n hi s. ALS-PNOA da a is classi ied and noise e u ns a e deno ed as class 7.
These poin s we e emo ed using LAS ools so wa e implemen ed in A cGIS 10.5.
O e lapping s ips migh gene a e ho izon al and e ical disc epancies in lase scanning su eys
playing and impo an ole in quali y con ol (Vosselman, 2012). Two app oaches could be
conside ed when dealing wi h o e lapping s ips: emo ing hose disc epan s ips o using
me hods o compensa e o disc epancies be ween wo da ase s in he o e lapping a eas (La ypo ,
2002). In his PhD he alidi y o o e lapping e u ns was e i ied by isualizing he 3D poin
clouds, gene a ing epo s abou ALS cha ac e is ics and analysing he gene a ed DEMs. The
analysis pe o med de e mined ha some iles om he sou h o A agón, named as “ARA_SUR”
p esen e ical and/o ho izon al displacemen s. These displacemen s a e classi ied wi h he
numbe 12 and, consequen ly, we e emo ed using he same p ocedu e as noise e u n emo al.
The ALS poin clouds in o es ed en i onmen s con ain e u ns om se e al su ace objec s, such
as sh ubs, ees, elec ical wi es and buildings ha should be sepa a ed om g ound e u ns
(Mon ealeg e e al., 2015a). The p ocess o sepa a ing he g ound and non-g ound poin s,
pe o med p io o DEM gene a ion, is called il e ing o classi ica ion. This is a key p ocess in
o es y applica ions, allowing no malizing poin clouds o de e mine abo eg ound e u n heigh s.
The e a e se e al il e ing algo i hms which can be classi ied in ou ypes (Meng e al., 2010;
Si hole & Vosselman, 2004): in e pola ion-based, slope-based, segmen a ion-based and
mo phological me hods. Mos il e ing me hods p o ide accu a e esul s in la and non-complex
a eas while s eep slopes and complex o es s ands inc eases classi ica ion e o s (Si hole &
Vosselman, 2004).
P e ious esea ch de eloped by Mon ealeg e e al. (2015a) compa ed se en il e ing me hods
implemen ed in open so wa e in Aleppo pine o es ed a eas wi h mode a e o s eep slopes. The
s udy de e mined ha Mul iscale Cu a u e Classi ica ion (MCC) me hod, de eloped by E ans &
Hudak (2007), was he mos accu a e o il e ALS-PNOA poin clouds in hese Medi e anean
en i onmen s. Acco dingly, in his PhD MCC algo i hm, implemen ed in MCC-LiDAR so wa e,
was selec ed o classi ying he poin clouds.
MCC is an i e a i e-in e pola ion-based il e . The algo i hm calcula es an in e pola ed su ace by
using a hin-pla ed spline and disca ds he ALS e u ns ha exceed a h eshold cu a u e. MCC
c ea es h ee scale domains de ining h ee p ocessing window sizes. The algo i hm i e a es in he
h ee scale domains un il he numbe o emaining e u ns changes by less han 1%, less han 0.1%,
and less han 0.01%, espec i ely (E ans & Hudak, 2007). Two pa ame e s mus be de ined o
unning MCC: he scale pa ame e (s) and he cu a u e h eshold ( ). The scale depends on objec
sizes and ALS poin spacing, while he cu a u e h eshold is ela ed o cu a u e ole ance o a
S udy a ea, ma e ials and me hods
35
scale domain. In his esea ch, bo h pa ame e s we e de e mined acco ding o Mon ealeg e e al.
(2015b). The scale was se o 1 m and he cu a u e h eshold o 0.3.
ALS da a c ea e a andom sampling o he e ain su ace, being necessa y o apply in e pola ion
p ocesses o gene a e a con inuous su ace (Vosselman & Maas, 2010). The selec ion o he
app op ia e in e pola ion algo i hm and DEM spa ial esolu ion is ele an as may cons i u e a
sou ce o inaccu acy in ege a ion me ics p edic ion (Aguila e al., 2006). Se e al in e pola ion
me hods, such as na u al neighbou , T iangula ed I egula Ne wo k (TIN) o as e , k iging, poin
o as e , a e commonly implemen ed in Geog aphical In o ma ion Sys em (GIS) so wa e.
Mon ealeg e e al. (2015b) compa ed he sui abili y o six in e pola ion ou ines o e a ange o
e ain oughness in Aleppo pine o es ed s ands. The analysis de e mined ha he TIN o as e
in e pola ion me hod p oduces he bes esul when gene a ing a DEM wi h 1 m esolu ion.
Consequen ly, he TIN o as e ou ine, implemen ed in A cGIS 10.5 so wa e, was selec ed o
in e pola ing and gene a ing he DEMs in his esea ch.
TIN o as e me hod includes wo p ocedu es. Fi s ly, he in e pola ion me hod build a e ain
su ace using Delaunay i egula iangles by connec ing ALS e u ns. The ele a ion is eco ded
o each iangle node, while ele a ions be ween nodes can be in e pola ed o gene a e a
con inuous su ace. Secondly, he TIN s uc u e is ans o med o a as e s uc u e using na u al
neighbou in e pola ion om p e ious iangles nodes o gene a e a alue a he cen e o each
as e cell.
The s a is ical me ics de i ed om ALS da a we e ela ed o ield a iables o i eg ession
models. The compu a ion o ALS me ics equi es he clipping o he poin cloud o he spa ial
ex en o each ield plo , being pe o med using he “ClipDa a” command o FUSION LDV 3.60
open sou ce so wa e (McGaughey, 2009). Fu he mo e, he heigh o he poin cloud e u ns we e
no malized using he DEM o compu e a wide ange o s a is ics using he “Cloudme ics”
command (Table 6). These s a is ical me ics a e commonly used as independen a iables in
o es y (E ans e al., 2009).
The compu a ion o ALS me ics no mally equi es he applica ion o a h eshold alue o emo e
g ound and/o unde s o ey lase hi s. In his PhD Thesis, se e al es s we e pe o med o
de e mine he mo e sui able h esholds acco ding o he ype o es ima ed s and a iable and
applica ion. A h eshold alue o 2 m heigh was selec ed acco ding o sh ub heigh ag eeing wi h
Nilsson (1996) and Næsse & Økland (2002) o p edic ing he analysed o es y me ics excep
om o al biomass, ha equi ed he selec ion o a di e en h eshold o include unde s o ey lase
hi s. The explo ed h esholds anging om 0 up o 1 m de e mined ha 0.2 is he mos sui able one
being ela ed o he ALS senso p ecision in coo dina e Z, which has a oo mean squa e e o
below 0.2 m.
Table 6 desc ibes he compu ed ALS me ics s uc u ed in h ee main g oups: me ics ela ed wi h
canopy heigh , me ics associa ed o canopy heigh a iabili y and canopy densi y me ics. The
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
36
compu ed me ics migh be selec ed o be used in modelling and all o hem p esen a cohe en
ela ionship wi h ege a ion s uc u e (E ans e al., 2009; McGaughey, 2009).
Table 6. De i ed me ics om ALS poin clouds, whe e xi is he heigh alue o he e u n, N is he o al
numbe o obse a ions, i is he e u n, and p is he pulse.
Me ic
Desc ip ion
Canopy heigh
me ics (CHM)
Pe cen iles o he e u n
heigh s 1, 5, 10, 20, 25, 30, 40,
50, 60, 70, 75, 80, 90, 95 and
99 (P01, P05, P10, e c.)
The pe cen iles we e compu ed acco ding o he ollowing
me hodology:
(𝑁−1)𝑃=𝐼+𝑓{ 𝐼 𝑖𝑠 𝑡ℎ𝑒 𝑖𝑛𝑡𝑒𝑔𝑒𝑟 𝑝𝑎𝑟𝑡 𝑜𝑓 (𝑁−1)𝑃
𝑓 𝑖𝑠 𝑡ℎ𝑒 𝑓𝑟𝑎𝑐𝑡𝑖𝑜𝑛𝑎𝑙 𝑝𝑎𝑟𝑡 𝑜𝑓 (𝑁−1)𝑃
whe e N is he numbe o obse a ion and P is he
pe cen ile alue di ided by 100.
𝑖𝑓 𝑓=0 𝑡ℎ𝑒𝑛 𝑃𝑒𝑟𝑐𝑒𝑛𝑡𝑖𝑙𝑒 𝑣𝑎𝑙𝑢𝑒=𝑥𝑖+1
𝑖𝑓 𝑓>0 𝑡ℎ𝑒𝑛 𝑃𝑒𝑟𝑐𝑒𝑛𝑡𝑖𝑙𝑒 𝑣𝑎𝑙𝑢𝑒=𝑥𝑖+1+𝑓(𝑥𝑖+2−𝑥𝑖+1)
whe e xi is he obse a ion alue conside ing ha
obse a ions a e anked in ascending o de .
Minimum ele a ion
𝑥𝑖 𝑚𝑖𝑛𝑖𝑚𝑢𝑚
Mean ele a ion
∑𝑥𝑖
𝑁
𝑖=1
𝑁
Mode ele a ion
xi alue mo e equen in he plo
Ele a ion quad a ic mean
(1
𝑁∑𝑥𝑖2
𝑁
𝑖=1 )1
2
Ele a ion cubic mean
(1
𝑁∑𝑥𝑖3
𝑁
𝑖=1 )1
3
L momen s (λ1 o λ4)
λ1=1𝐶1
𝑛∑𝑥(𝑖)
𝑁
𝑖=1
λ2=121𝐶2
𝑛∑( 𝐶1− 𝐶1
𝑛−1𝑖−1 )𝑥(𝑖)
𝑁
𝑖=1
λ3=131𝐶3
𝑛∑( 𝐶2−2 𝐶1
𝑖−1𝑖−1 𝐶1+ 𝐶2
𝑛−1𝑛−1 )𝑥(𝑖)
𝑁
𝑖=1
λ4=141𝐶4
𝑛∑(𝐶3−3 𝐶2
𝑖−1𝑖−1 𝐶1+3 𝐶1
𝑖−1𝑛−1 𝐶2− 𝐶3
𝑛−1𝑛−1 )𝑥(𝑖)
𝑁
𝑖=1
whe e x(i), i=1, 2, …, n, a e sample alues anked in ascending o de and
𝐶𝑘=(𝑚
𝑘)
𝑚=𝑚!
𝑘!(𝑚−𝑘)!
is he numbe o combina ions o any k i ems o m m i ems
and is equal o ze o when k > m.
Maximum ele a ion
𝑥𝑖 𝑚𝑎𝑥𝑖𝑚𝑢𝑚
S udy a ea, ma e ials and me hods
37
Canopy heigh
a iabili y
me ics
(CHVM)
S anda d de ia ion o poin
heigh dis ibu ion
√∑(𝑥𝑖−𝜇)2
𝑁
𝑖=1 𝑁
Va iance o poin heigh
dis ibu ion (σ2)
∑(𝑥𝑖−𝜇)2
𝑁
𝑖=1 𝑁
Coe icien o a ia ion o
poin heigh dis ibu ion
𝜎
𝜇100
Skewness o poin heigh
dis ibu ion
∑(𝑥𝑖−𝜇)3
𝑁
𝑖=1
(𝑁−1)𝜎3
ku osis o poin heigh
dis ibu ion
∑(𝑥𝑖−𝜇)4
𝑁
𝑖=1
(𝑁−1)𝜎4
In e qua ile dis ance o
poin heigh dis ibu ion
[𝑃75(𝑥)−𝑃25(𝑥)]
A e age Absolu e De ia ion
o poin heigh dis ibu ion
∑(𝑥𝑖−𝜇)
𝑁
𝑖=1 𝑁
L momen coe icien o
a ia ion o poin heigh
dis ibu ion (𝜏2)
λ2
λ1
0< 𝜏2<1
L momen skewness o poin
heigh dis ibu ion (𝜏3)
λ3
λ2
-1< 𝜏3<1
L momen ku osis o poin
heigh dis ibu ion (𝜏4)
λ4
λ2
14(5τ32−1)≤τ4<1
Canopy
densi y me ics
(CDM)
Pe cen age o i s e u ns
abo e a heigh -b eak, abo e
he mean o he mode
∑𝑟𝑖 𝑓𝑖𝑟𝑠𝑡 𝑟𝑒𝑡𝑢𝑟𝑛𝑠>ℎ𝑒𝑖𝑔ℎ𝑡𝑏𝑟𝑒𝑎𝑘
𝑁
𝑖=1 ∑𝑟𝑖 𝑓𝑖𝑟𝑠𝑡 𝑟𝑒𝑡𝑢𝑟𝑛𝑠
𝑁
𝑖=1 100
Pe cen age o all e u ns
abo e a heigh -b eak, abo e
he mean o he mode
∑𝑟𝑖 >ℎ𝑒𝑖𝑔ℎ𝑡𝑏𝑟𝑒𝑎𝑘
𝑁
𝑖=1 𝑁100
Canopy elie a io
𝜇−𝑥𝑖 𝑚𝑖𝑛𝑖𝑚𝑢𝑚
𝑥𝑖 𝑚𝑎𝑥𝑖𝑚𝑢𝑚−𝑥𝑖 𝑚𝑖𝑛𝑖𝑚𝑢𝑚
All e u ns abo e a heigh -
b eak, abo e he mean o he
mode x 100
∑𝑟𝑖 𝑎𝑙𝑙 𝑟𝑒𝑡𝑢𝑟𝑛𝑠>ℎ𝑒𝑖𝑔ℎ𝑡𝑏𝑟𝑒𝑎𝑘
𝑁
𝑖=1 ∑𝑟𝑖 𝑎𝑙𝑙 𝑟𝑒𝑡𝑢𝑟𝑛𝑠
𝑁
𝑖=1 100
2.2.6. Op ical da a and spec al indices
Al hough he PhD Thesis mainly ocus on he use o ALS-PNOA da a o o es y applica ions, he
combina ion wi h op ical da a allowed o es ima e i e se e i y and, subsequen ly, de e mine he
CO2 emissions.
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
38
Wild i es cons i u e a socio-en i onmen al haza d in py ophy e Aleppo pine Medi e anean
ecosys ems. Fi e se e i y usually eaches high le els, gene a ing impo an changes in ege a ion
igou , colou , wa e con en as well as o es composi ion, s uc u e and densi y. Op ical emo e
sensing da a ha e been widely used o cha ac e ize i e se e i y (Ga cía-Llamas e al., 2019)
conside ing in a ed spec al egions.
Landsa p og am is he longes middle esolu ion emo e sensing p og am and p o ides cohe en
and con inuous global da a since 1972. Speci ically, Landsa 8 was launched in 2013 ca ying on
boa d he Ope a ional Land Image (OLI) senso . The mission p esen s a pola low ea h o bi and
16 days o empo al esolu ion. The spa ial esolu ion o he 9 op ical spec um bands is 30 m,
excep om he panch oma ic band wi h 15 m esolu ion. The mos aluable bands o cha ac e ize
i e se e i y a e numbe 5, which co esponds o he nea in a ed (NIR), and 7, which e e s o he
sho wa e in a ed (SWIR). NIR is linked o olia s uc u e and mo phology while SWIR is
ela ed o ege a ion and soil wa e con en (So e el e al., 2010).
In his esea ch, wo images we e selec ed o cha ac e ize p e and pos - i e condi ions and es ima e
i e se e i y. The p e- i e image was acqui ed on June 30 (pa h 200, ow 31) 2015 and he pos - i e
image on July 9 (pa h 199, ow 31) 2015. The selec ion o hese images was condi ioned by he need
o empo a y p oximi y be ween images in o de o cap u e he a iabili y o ege a ion spec al
esponse. The images we e p o ided by he Uni ed S a es Geological Su ey (USGS) and
downloaded using Ea h Explo e pla o m. Bo h images p esen a high p ocessing le el. Images
we e geome ically and adiome ic co ec ed. Speci ically, he Landsa Su ace Re lec ance Code
(LaSRC) was applied acco ding o Ve mo e e al. (2016). The algo i hm uses a adia i e ans e
model, clima e da a om MODIS and he coas al ae osol band o pe o m hose co ec ions.
The use o spec al indices, de i ed om he combina ion o di e en spec al bands, p o ides
ele an in o ma ion abou su ace p ope ies (Chu ieco, 2010). The No malized Bu n Ra io (NBR)
index, de eloped by Key & Benson (2006), ela es Landsa band 5, om NIR, wi h band 7 om
SWIR. NBR alues ange om -1 o 1. Su aces wi h no p esence o ege a ion o low ege a ion
igou p esen nega i e alues while pho osyn he ically ac i e a eas show posi i e alues. In his
PhD Thesis, NBR was calcula ed o p e- i e and pos - i e images acco ding o equa ion 27. Then,
he Di e enced o Del a NBR index (∆NBR) was compu ed o p o ide a quan i a i e measu e o
he bu ned a ea (equa ion 28).
NBR= (ρNIR − ρSWIR )/ (ρNIR + ρSWIR) (27)
∆NBR= NBRp e i e – NBRpos i e (28)
whe e
NIR (nea in a ed) and
SWIR (sho wa e in a ed) e e o bands 5 and 7 Landsa 8 OLI
e lec ance, espec i ely.
The ∆NBR alues we e mul iplied by 1,000 o p o ide a alid con inuous ange o alues.
Nega i e alues a e associa ed wi h as eg ow h om he baceous, while posi i e alues a e
ela ed o di e en deg ee o i e se e i y.
S udy a ea, ma e ials and me hods
39
2.2.7. Wild i e biomass loss and CO2 emission es ima ion
Wild i es cons i u es a ele an sou ce o ca bon monoxide emissions (Pé on e al., 2004) in he
Medi e anean basin, which yea ly eco ds an a e age o 45,000 i es (Oli ei a e al., 2012). The
accoun o ca bon dioxide emissions is ele an o unde s anding ca bon cycling and p o ides
in o ma ion o de elop clima e egula ion policies (Mie ille e al., 2010). Thus, di e en
app oaches ha e been es ed o es ima e he emissions o wild i e as he one p oposed by he
In e go e nmen al Panel on Clima e Change (IPCC) (equa ion 29):
𝐿𝑓𝑖𝑟𝑒=𝐴×𝐵×𝐶×𝐷×10−6 (29)
whe e L i e e e s o he quan i y o GHG eleased due o i e ( onnes o GHG), A is he bu ned a ea
(ha), B is he mass o “a ailable” uel including biomass, g ound li e and dead wood ( ons ha-1), C
is he combus ion e iciency o ac ion o combus ed biomass (dimensionless), D is he emission
ac o (g kg-1 o d y ma e bu n ).
A simila app oach, p oposed by De San is e al. (2010), adap ed he IPCC one o es ima e i e
GHG emissions, using emo e sensing echniques. In his sense, ou s eps we e equi ed: he
delimi a ion o he bu ned a ea; he es ima ion o p e- i e biomass; he assessmen o he ac ion o
biomass consumed by he i e o bu ning e iciency, associa ed wi h i e se e i y and; he use o
con e sion ac o s o es ima e GHG emissions.
T adi ionally he es ima ion o p e- i e biomass was pe o med by using ield da a and allome ic
equa ions, while pos - i e biomass was assessed ei he by isual examina ion (Roy e al., 2005) o
ield-based weigh ing (Sá e al., 2005). Howe e , he implemen a ion o emo e sensing echniques
p o ided new me hodological app oaches. In his esea ch, ALS da a (used o de e mine p e- i e
biomass) and passi e op ical da a (applied o es ima e i e se e i y) we e combined wi h he inal
objec i e o es ima ing biomass losses and, subsequen ly CO2 emissions. The ollowing i e s eps
we e conduc ed in his app oach:
i. Es ima ion o Aleppo pine p e- i e biomass using ALS da a cap u ed p e iously o he
occu ence o he wild i e.
ii. Es ima ion o i e se e i y. ∆NBR index is compu ed using wo Landsa 8 OLI images om
June 30 h and July 9 h 2015.
iii. Mapping o p e- i e Aleppo pine o es . The Spanish Na ional Fo es Map as well as a
canopy heigh model de i ed om he ALS da a we e used o delimi he bu ned s ands.
i . Selec ion o bu ning e iciency ac o s and biomass losses quan i ica ion. Mos app oaches
conside ha o es biomass is comple ely consumed by i e (F ench e al., 2004). Howe e ,
mo e accu a e app oaches om ou poin o iew conside di e en bu n se e i y le els,
ela ed o di e en biomass losses. Thus, in his esea ch h ee bu ning e iciency ac o s
we e applied o h ee di e en se e i y le els (De San is e al., 2010). Key & Benson (2006)
gene ic se e i y anges we e eclassi ied o ma ch he h ee bu ning e iciency ac o s
(Table 7). Low bu ning e iciency e e s o low consump ion o lea es and e y low
consump ion o b anches. Mode a e bu ning e iciency deno es in e media e consump ion
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
40
o lea es and mode a e consump ion o small b anches. High bu ning e iciency
co esponds o a comple e consump ion o lea es and high loss o small b anches and
wigs.
Table 7. Se e i y le els, ΔNBR gene ic anges de ined by Key & Benson (2006) and bu ning e iciency
ac o s used o es ima e biomass losses ollowing De San is e al. (2010).
Se e i y le el
ΔNBR ange
Bu ning e iciency ac o s
Unbu ned
−100 o +99
0.00
Low se e i y
+100 o +269
0.25
Mode a e–low se e i y
+270 o +439
0.42
Mode a e–high se e i y
+440 o +659
0.57
High se e i y
+660 o +1300
. Con e sion o biomass losses o ca bon con en and, subsequen ly, o CO2 emissions.
Biomass can be con e ed o ca bon con en by applying con e sion ac o s. The mos
common ac o is 0.5, bu , in his PhD Thesis, a alue o 0.499 was se o Aleppo pine, in
acco dance wi h Mon e o e al. (2005). These au ho s de e mined se e al con e sion ac o s
o Medi e anean species based in ield wo k. Secondly, he con e sion om ca bon o
CO2 was pe o med acco ding o T ozzi e al. (2002) equa ion. This equa ion includes he
a iables p oposed by Seile & C u zen (1980), while modi ying he speci ic ac o s o be e
ep esen Medi e anean en i onmen s (equa ion 30)
CO
2
= ε*δ*
C
(30)
whe e
ε
is he ac ion o o al ca bon emi ed as CO
2
(0.888);
δ
is he ac o o
con e sion om he emissions in on o ca bon o he emissions in on o CO
2
(44/12);
and
C
is he ca bon con en .
2.2.8. Modelling: a iable selec ion, eg ession me hods and alida ion
Va iable selec ion
Va iable selec ion includes a a ie y o me hods o educe he dimensionali y p oblem as well as
deal wi h la ge ea u e se s, imp o ing model pe o mance, educing ime and cos s and
gene a ing unde s andable models (Guyon & Elissee , 2003). This PhD Thesis explo ed he
pe o mance o i e selec ion p ocesses o es ima e o es s and a iables: Spea man’s ank
co ela ion, S epwise selec ion, P incipal componen analysis (PCA) and Va imax o a ion, Las
absolu e sh inkage and selec ion ope a o (LASSO) and All subse selec ion. These me hods a e
desc ibed below.
Spea man’s ank co ela ion coe icien is gene ally deno ed by he G eek le e p ( ho) (Spea man,
1904). This coe icien de e mines he s eng h and di ec ion o he ela ionship be ween wo
S udy a ea, ma e ials and me hods
41
a iables being ca ied ou using “co .spea man” unc ion in R en i onmen . Al hough o he
co ela ion coe icien s as Pea son o Kendall exis , Spea man’s ank coe icien has been widely
used o o es y applica ions, showing a uni o m powe o linea and non-linea ela ionships.
The ho coe icien a ies be ween +1 and -1. The alue +1 indica es a pe ec posi i e deg ee o
associa ion be ween wo a iables while he alue -1 deno es a pe ec nega i e deg ee o
associa ion. The weake he co ela ion, he close o 0 is he alue. Thus, he co ela ion alue o 0
indica es a null ela ionship be ween he analysed a iables. In his PhD, he selec ion o he ALS
a iables was made conside ing a minimum posi i e and nega i e ho alue, anging om 0.2 up
o 0.5. Depending on he highe o lowe s eng h o he ela ionship be ween he a iables he
ange was changed o be e educe a iable edundancy and mul icollinea i y.
S epwise selec ion, o also called s epwise eg ession, allows selec ing p edic i e a iables in an
au oma ic p ocedu e (E oymson, 1960). The me hod i e a i ely adds o d ops a iables a se e al
s eps o de e mine he bes subse o a iables, which deno es he bes model pe o mance and
lowe p edic ion e o . The p edic o , ha is added o d opped in each s ep, is based in he Akaike
in o ma ion c i e ion (AIC). The e exis h ee ways o pe o ming s epwise eg ession: o wa d
selec ion, backwa d selec ion and bidi ec ional s epwise selec ion. Fo wa d selec ion begins wi h
no a iables in he model, i e a i ely selec s he a iables ha con ibu e he mos and s ops
adding p edic o s when he imp o emen is no longe s a is ically signi ican . Backwa d selec ion
begin wi h all a iables in he model, i e a i ely d ops he one ha con ibu e he leas and s ops
d opping p edic o s when all he selec ed a iables a e s a is ically signi ican . Bidi ec ional
s epwise selec ion combines o wa d and backwa d selec ions. Fi s ly, his me hod begins wi h
o wa d selec ion and hen d ops hose a iables no s a is ically signi ican using backwa d
selec ion. The s epwise selec ion was ca ied ou using “s ep” unc ion in R en i onmen .
PCA is a dimension- educ ion me hod ha educes la ge se s o p edic o s o a smalle size,
keeping he majo i y o he in o ma ion. Al hough is a classic s a is ical me hod ha has been
b oadly used in many esea ch ields, i is no a common app oach in ALS me ic selec ion (Sil a e
al., 2016). PCA was compu ed using R package “la ice” and speci ically wi h he unc ion
“p comp”, which uses he singula alue decomposi ion o examine he co a iance and co ela ions
be ween indi iduals. Acco ding o he Kaise C i e ion, in his esea ch he gene a ed componen s
wi h g ea e alues han 0.1 we e e ained. Then, a Va imax o a ion, which maximizes he sum o
he a iance, was applied o be e in e p e he PCA esul s (Da ling on & Ho s , 1966; Kaise ,
1958).
LASSO is a me hod ha pe o ms wo p ocesses: egula iza ion and ea u e selec ion. The me hod
was popula ized and imp o ed by Tibshi ani (1996). This echnique applies a sh inking o
egula iza ion p ocess o penalize some o he coe icien s o ze o by using a penal y e m, which
e e s o he sum o he absolu e coe icien s. Those a iables ha ha e a non-ze o coe icien a e
egula iza ion will be selec ed o be pa o he model, minimizing p edic ion e o s and
gene a ing in e p e able models. LASSO was compu ed in R using he “glmne ” package.
All subse selec ion me hods allow a sui able g oup o me ics being selec ed, while dis ega ding
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
48
The indi ec app oach equi es highe cos o modelling and da a cap u ing. Thus, we ound
di e en esul s o he e alua ion o he wo app oaches in he li e a u e (Næsse & Gobakken,
2005; Noo de mee e al., 2018; Zhao e al., 2018). Some au ho s ge sligh ly be e pe o mance o
he di ec app oach in he p edic ion o biomass and ca bon luxes (Bollandsås e al., 2013; Cao e
al., 2016; Skow onski e al., 2014) while Meye e al. (2013) and Zhao e al. (2018) achie ed be e
esul s wi h he indi ec app oach.
This PhD Thesis explo es he bene i s o he wo men ioned app oaches in he p edic ion o se en
o es y a ibu es a egional scale: s and densi y, basal a ea, squa ed mean diame e , dominan
diame e , ee dominan heigh , imbe olume and o al ee biomass. We pe o med a
compa ison o he di ec and indi ec app oaches o assess empo al ans e abili y. Fi s ly,
ollowing he indi ec app oach, wo di e en models we e i ed o he a ailable ALS-PNOA
da a (2011 and 2016), es ima ing he s and a ibu es o each poin in ime, using di e en ALS-
me ics and model pa ame e s. In his app oach in en o y da a was upda ed using single- ee-
g ow h models o gene a e concomi an in o ma ion o he ALS-PNOA yea s (see Sec ion 2.2.3).
Secondly, he di ec app oach was es ed. The models, i ed o one poin in ime, we e
ex apola ed o he o he poin in ime using he same a iables and model pa ame e s. The
p ocess was pe o med wo imes: he 2011 models we e ex apola ed o 2016, and in e sely he
2016 models we e ex apola ed o 2011. The modi ica ion o he o iginal me hodology es ed he
alidi y o mul i- empo al ALS da a o pe o m u u e o e ospec i e analyses.
2.2.10. Va iable in luence assessmen
ALS senso s ha e di e en con igu a ions ha de e mine he inal da a cha ac e is ics and quali y
o he poin clouds. These con igu a ions a e no mally di e en om one ligh o ano he , a ying
acco ding o he speci ic senso capabili ies, and migh ha e an e ec in model accu acy. In his
sense, he e ec o h ee ALS cha ac e is ics (poin densi y, scan angle, canopy pene a ion pulse)
and wo en i onmen al condi ions (slope and sh ub p esence) on he p edic ion o o es esidual
biomass ha e been add essed.
The e ec o poin densi y has been explo ed in he p edic ion o di e en o es in en o y
a ibu es such as heigh , basal a ea o olume (Roussel e al., 2017; Gobakken & Næsse , 2008).
Howe e , less s udies ha e conside ed poin densi y e ec on biomass p edic ion (Ga cía e al.,
2017; Singh e al., 2015; Ruiz e al., 2014). The e ec o scan angle on ee heigh p edic ion has been
analysed by Disney e al. (2010), Holmg en (2004), Liu e al. (2018) and Mon aghi (2013). Disney e
al. (2010) p oposed o minimize he use o da a collec ed a scan angles g ea e han ∼15°.
Acco ding o Holmg en (2004) and Liu e al. (2018), he p edic ion o canopy closu e is a ec ed by
he scan angle, being necessa y o a oid o -nadi angles om 23 o 38°. Simila esul s we e
concluded by Mon aghi (2013) in he p edic ion o o es y me ics using an a ea based app oach,
de e mining ha scan angles highe han 20° had a g ea e ec in o es pa ame e s p edic ions.
The s uc u al cha ac e is ics and ALS ligh se ings modi y canopy pulse pene a ion (CPP),
dec easing DTM accu acy (Cowen e al., 2000; Hollaus e al., 2006; Hyyppä e al., 2000). The highe
S udy a ea, ma e ials and me hods
49
he densi y o he o es , he lowe CPP a es a e ound (Hollaus e al., 2006). Seconda y e ec s o
densely co e ed a eas would be a dec ease in CPP in lowe s a a (Chasme e al., 2006b; Wasse e
al., 2013) and he dec ease o DTM accu acy (Cowen e al., 2000; Hollaus e al., 2006; Hyyppä e al.,
2000). In con as , lea o condi ions in deciduous o es imp o es CPP a es (Hill e al., 2009;
Wasse e al., 2013).
The p esence o s eep slopes educes he accu acy in ee heigh p edic ion (B eidenbach e al.,
2008; Cla k e al., 2004; Ø ka e al., 2018), as well as in ee diame e , basal a ea, numbe o s ems
and olume (Ø ka e al., 2018). Fu he mo e, dec eases he abili y o de ec ee ops (Khos a ipou
e al., 2015). Al hough ALS p edic ions a e a ec ed by he inc ease o slope, he de ec ed e ec was
no se e e. Slope e ec migh be pa ially explained by he lowe accu acy o DTMs in hese a eas,
conside ing ha il e s ha e mo e di icul ies on de e mining g ound poin s on s eep slopes
(Mon ealeg e e al., 2015a).
The e ec o sh ub p esence has no been p e iously analysed. The hypo hesis ha he p esence o
sh ub migh a ec he p edic ion o o es esidual biomass is associa ed o he lowe CPP in hose
a eas and, consequen ly, a lowe numbe o g ound e u ns, which may a ec DTM gene a ion.
Two me hodological app oaches we e es ed o analyse he e ec o he i e abo emen ioned
a iables in o es esidual biomass p edic ion. Fi s ly, a g aphical assessmen using boxplo s,
including he a e age mean p edic ion e o pe de ined class and a iable was applied. In
addi ion, se e al s a is ical es s, desc ibed below, we e applied o de e mine whe he he
di e ences be ween model pe o mances we e s a is ically signi ican .
The de ined classes o each analysed a iable a e he ollowing:
Poin densi y. Plo s we e classi ied in o wo classes: up o 1 poin m-2 and highe han 1
poin m-2 acco ding o Mon ealeg e e al. (2015b). Jakubowski e al. (2013) and Ga cía e al.
(2017) also concluded ha his h eshold allows o ha e ela i ely high accu acies in o es
pa ame e s modelling.
Scan angle. Th ee scan angle classes we e es ed. The i s class was se close o nadi wi h
a e age scan angle o up o 5°. The second class was de ined as plo s wi h an a e age scan
angle be ween 5 and 15°. The hi d class included hose plo s wi h an a e age scan angle
highe han 15°. The b eakpoin be ween he second and hi d classes was de ined by he
maximum a e age scan angle es ablished by he PNOA mission in o de o each he
minimum nominal poin densi y speci ied by he p ojec .
Canopy pulse pene a ion. The in luence o ou ca ego ies we e analysed: 0%-25%, 25%-
50%, 50%-75% and 75%-100% ollowing Mon ealeg e e al. (2015b). The p opo ion o he
pulses ha pene a e canopy and each he g ound was calcula ed using a g ound ole ance
o 2 m.
Te ain slope. The e ec o he e ain slope was assessed using wo ca ego ies: smoo h
slopes o up o 15% and s eep slopes highe han 15%.
Sh ub p esence was bina y ca ego ized in plo s wi h and wi hou unde s o y.
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
50
The mean p edic ed e o (MPE) was selec ed o analyse he di e ences be ween he es ablished
classes. MPE was ob ained o each ield plo and, a e wa ds, he a e age mean pe class was
compu ed (equa ion 36). Fu he mo e he pe cen age o mean p edic ed e o (%MPE) espec o
he obse ed mean alue was compu ed acco ding o equa ion 37.
𝑀𝑃𝐸=∑(𝑦𝑖−𝑦𝑖)
𝑛
𝑖=1𝑛 (36)
%𝑀𝑃𝐸=𝑀𝑃𝐸
𝑦×100 (37)
whe e 𝑦𝑖 is he obse ed alue o plo i; 𝑦𝑖 is he p edic ed alue o sample plo i; 𝑛 is he numbe
o plo s and 𝑦 is mean obse ed alue o all plo s.
P e ious o signi icance analysis, no mali y and homogenei y es we e compu ed. A e
conside ing loga i hmic and squa e oo ans o ma ion we concluded ha a iables we e no
no mally dis ibu ed. In his sense, non-pa ame ic Mann-Whi ney and median es s we e applied
o analysing di e ences be ween wo ca ego ies. K uskal Wallis es was applied o hose
analyses wi h mo e han wo classes.
The Mann-Whi ney es is conside ed he main non-pa ame ic al e na i e o he independen
sample - es . This es is designa ed o compa e wo popula ions. The null hypo hesis o he es
es ablishes ha bo h samples come om di e en popula ions, while he al e na i e o he null
hypo hesis indica es ha bo h samples come om he same popula ion.
The median es is a non-pa ame ic one used o de e mine whe he wo o mo e samples di e in
hei cen al endency o median alue, consequen ly i can be in e ed whe he he samples a e
om he same popula ion o no . The null hypo hesis es ablishes ha bo h samples come om he
same popula ion.
K uskal Wallis es is conside ed he main non-pa ame ic al e na i e o he One Way ANOVA
es . This me hod is a ank-based es ha de e mines whe he he medians o wo o mo e g oups
a e di e en . The null hypo hesis conside s ha he samples a e om he same popula ion, while
he al e na i e hypo hesis es ablishes ha a leas one o he samples come om a di e en
popula ion.
2.2.11. Mapping o o es a iables
The mapping o o es s and a iables ha e been adi ionally pe o med by linking ee me ics o
he es ima ed s and a iable using allome ic equa ions and, subsequen ly, ex apola ing hese
es ima es a s and-le el o egional scale (Boud eau e al., 2008). The use o emo e sensing ools
p o ides a g ea e o e iew o la ge a eas and wi h highe empo al esolu ion (Cas o e al.,
2003). The use o op ical da a and, specially, he cha ac e iza ion o dis u bance his o y wi h
empo al se ies, ha e been p oposed o es ima ing o es y a iables (Cohen e al., 1996; Coops &
Wa ing, 2001). Howe e , he in o ma ion cap u ed by passi e op ical senso s ends o sa u a e
unde closed canopy condi ions and in dense o es s (Lu, 2006). Al hough ac i e SAR emo e
S udy a ea, ma e ials and me hods
51
sensing imp o ed he accu acy on o es y p edic ions (Tanase e al., 2014), he e a e s ill
di icul ies in he e ogeneous and dense o es s (Hyde e al., 2006).
ALS da a ha e been p o en as he mos sui able echnique o mapping 3D s uc u e (Zhao e al.,
2018). Howe e , he use o ALS da a a egional scales is s ill limi ed by he acquisi ion cos s, he
absence o global co e age and he huge da a olumes ha need o be p ocessed. The use o ALS
da a o sampling some a eas o by c ea ing s ips and he subsequen usion wi h passi e op ical
da a ha e been explo ed o es ima ing o es s and a iables a egional scales (e.g.: Ma asci e al.,
2018; P lugmache e al. 2014). Recen ly, he inc ease o ALS da a a egional scale o coun y le el,
as he case o Spain, has opened new oppo uni ies. In his sense, al hough compu a ional e o is
equi ed, he imp o emen in ha dwa e and so wa e has allowed p ocessing ime educ ion. In
his sense, his PhD Thesis explo ed he use o ALS da a a egional scales, p o iding ca og aphic
in o ma ion o o es manage s.
In his esea ch, he gene a ion o ca og aphy implied he p edic ion o a dependen a iable o
he whole s udy a ea by applying a eg ession model p e iously gene a ed wi h he abo e
men ioned me hodology. Pixel size is one o he mos ele an ac o s o be conside ed in
mapping. The de e mina ion o pixel size should conside ield plo size and ALS poin densi y,
bu migh be modi ied o accoun o speci ic equi emen o o es manage s. Commonly, pixel
size is simila o ield plo size. An inc ease in pixel size migh dec ease mapping accu acy when
he numbe o ALS e u ns pe pixel is low. Fu he mo e, independen a iables ha a e included
in eg ession models should be con e ed o as e o ma . In his sense, FUSION so wa e,
speci ically designa ed o wo k wi h ALS da a o o es y pu poses, was used o gene a e he
as e iles using “G idme ics” and, subsequen ly “CSVG id” commands. The gene a ion o he
as e o he dependen a iable was pe o med in an R en i onmen ollowing h ee s eps:
unning he bes selec ed model; eading he independen a iables in as e o ma ; spa ializing
he esul s.
53
3. Resea ch con ibu ions
The pape s ha cons i u e he PhD Thesis body a e en i ely
included in his chap e as equi ed by compendium PhD
Thesis ype. The pape s p o ide di e en o es applica ions
using ALS-PNOA da a and ield su eys: he es ima ion o
CO2 emissions gene a ed by wild i es, se e al biomass
ac ions and in en o y a iables. Fu he mo e, he pape s
compa e se e al selec ion and eg ession me hods wi hin
di e en condi ions, analysing he e ec o ALS
cha ac e is ics and en i onmen al a iables in modelling
e o and assessing empo al model ans e abili y.
Resea ch con ibu ions
55
3.1. Compa ison o eg ession models o es ima e biomass losses
and CO2 emissions using low densi y ai bo ne lase scanning
da a in a bu n Aleppo pine o es
Compa ación de modelos de eg esión pa a es ima las pé didas de biomasa y emisiones de CO2
u ilizando da os de escáne láse ae opo ado de baja densidad en bosques de Pino ca asco
a ec ados po el uego
RESUMEN
El conocimien o de la pé dida de biomasa o es al p oducida po un incendio puede se de
u ilidad pa a la es imación de las emisiones de los gases de e ec o in e nade o a la a mós e a. Es e
es udio se cen a en la es imación de la pé dida de biomasa y las emisiones de CO2 po la
combus ión de masas o es ales de Pino ca asco en un incendio ocu ido en el municipio de Luna
(España). La disponibilidad de da os de escáne láse ae opo ado (ALS) de baja densidad pe mi ió
es ima la biomasa a bó ea p e- uego. Se ealizó una compa ación de nue e modelos de eg esión
con obje o de elaciona la biomasa, es imada en 46 pa celas de campo, con dis in as mé icas
ex aídas de los da os ALS. El mé odo de eg esión linea mul i a ian e seleccionado como
óp imo, incluyó en e las a iables independien es el po cen aje de p ime os e o nos sob e 2 m y
el pe cen il 40 de la al u a de los e o nos. El modelo se alidó u ilizando una écnica de alidación
c uzada “lea e-one-ou -c oss- alia ion” (RMSE: 6.1 on ha-1). Las pé didas de biomasa se es ima on
u ilizando una ap oximación en es ases: (i) la se e idad del incendio ue ob enida u ilizando el
índice de di e encia no malizado (ΔNBR), (ii) los pina es de Pino ca asco se delimi a on
u ilizando el Mapa Fo es al Nacional y da os ALS y, (iii) es ac o es de e iciencia de combus ión
se aplica on conside ando los ni eles de se e idad. La biomasa pos - uego se ans o mó en
emisiones de CO2 (426.754,8 on). Es e es udio e idencia la u ilidad de los da os ALS de baja
densidad pa a es ima de o ma p ecisa la biomasa p e- uego y e alua las emisiones de CO2 en
masas o es ales medi e áneas de Pino ca asco.
Resea ch con ibu ions
57
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
64
Resea ch con ibu ions
65
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
66
Resea ch con ibu ions
67
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
68
Resea ch con ibu ions
69
Resea ch con ibu ions
71
3.2. Es ima ion o o al biomass in Aleppo pine o es s ands
applying pa ame ic and nonpa ame ic me hods o low-
densi y ai bo ne lase scanning da a
Es imación de biomasa o al en bosques de pino ca asco aplicando mé odos pa amé icos y no
pa áme os median e da os de escáne láse ae opo ado de baja densidad
RESUMEN
La cuan i icación de la biomasa o al es de u ilidad pa a la e aluación de las polí icas de egulación
climá icas desde escalas locales a escalas globales. Es a in es igación es ima la biomasa o al,
incluyendo la biomasa a bó ea y a bus i a, en bosques de Pino ca asco localizados en la egión de
A agón (España), u ilizando da os de escáne láse ae opo ado (ALS) y abajo de campo. La
compa ación de cinco mé odos de selección y cinco modelos de eg esión se ealizó con obje o de
elaciona la biomasa o al, es imada en 83 pa celas de campo median e ecuaciones alomé icas,
con di e sas a iables ex aídas de la nube de pun os ALS. Pa a el cálculo de las a iables ALS se
u ilizó un umb al de 0.2 m. La mues a se di idió en en enamien o y alidación componiéndose
de 62 y 21 pa celas de campo, espec i amen e. El modelo con meno e o cuad á ico medio
después de la alidación (15,14 ons ha-1) ue el modelo de eg esión linea mul i a ian e. Dicho
modelo incluyó es a iables ALS: el pe cen il 25 de la al u a de e o nos, la a ianza y el
po cen aje de p ime os e o nos sob e la media. El es udio con i ma la u ilidad de los da os ALS
de baja densidad pa a es ima con exac i ud la biomasa o al, y po consiguien e mejo a la
cuan i icación de biomasa disponible y con enido de ca bono en Pina es medi e áneos de Pino
ca asco.
Resea ch con ibu ions
73
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
80
Resea ch con ibu ions
81
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
82
Resea ch con ibu ions
83
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
84
Resea ch con ibu ions
85
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
86
Resea ch con ibu ions
87
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
88
Resea ch con ibu ions
89
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
96
1
Resea ch con ibu ions
97
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
98
Resea ch con ibu ions
99
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
100
Resea ch con ibu ions
101
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
102
Resea ch con ibu ions
103
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
104
Resea ch con ibu ions
105
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
112
Resea ch con ibu ions
113
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
114
Resea ch con ibu ions
115
Resea ch con ibu ions
117
3.4. Tempo al T ans e abili y o Pine Fo es A ibu es Modelling
Using Low-Densi y Ai bo ne Lase Scanning Da a
T ans e ibilidad empo al de modelos de a iables o es ales u ilizando da os de escáne láse
ae opo ado de baja densidad
RESUMEN
Es e es udio e alúa la ans e ibilidad empo al de modelos gene ados u ilizando da os de escáne
láse ae opo ado (ALS) y adqui idos en dos echas di e en es. La es imación de sie e a iables
o es ales (densidad de pies, á ea basal, diáme o cuad á ico medio, diáme o dominan e, al u a
dominan e, olumen made able y biomasa a bó ea) se ealizó u ilizando un en oque basado en
á eas en masas o es ales medi e áneas de Pino ca asco. Los da os ALS de baja densidad se
adqui ie on en 2011 y en 2016, mien as que las 147 pa celas de campo se mues ea on en 2013,
2014 y 2016. La gene ación de da os de campo pa a las echas de los uelos ALS se ealizó
median e la aplicación de modelos de c ecimien o de á bol indi idual. Cinco mé odos de selección
y cinco modelos de eg esión ue on compa ados pa a elaciona las obse aciones omadas en
campo espec o a las mé icas ALS. La selección de los mejo es modelos de eg esión ajus ados
pa a cada a iable o es al, y sepa adamen e pa a los años 2011 y 2016, se ealizó u ilizando un
en oque indi ec o. El ajus e de los modelos y la ans e ibilidad empo al de los mismos se analizó
ex apolando los mejo es modelos ajus ados pa a 2011 a 2016, e in e samen e de 2016 a 2011. El
modelo de eg esión no pa amé ico suppo ec o machine con ke nel adial mos ó los mejo es
esul ados. La di e encia en el e o cuad á ico medio en po cen aje de los modelos ajus ados y
ex apolados ue de 2,13% pa a los modelos de 2011 y 1,58% pa a los de 2016.
Resea ch con ibu ions
119
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
120
Resea ch con ibu ions
121
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
128
Resea ch con ibu ions
129
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
130
Resea ch con ibu ions
131
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
132
Resea ch con ibu ions
133
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
134
Resea ch con ibu ions
135
Cha ac e iza ion o Medi e anean Aleppo pine o es using low-densi y ALS da a
136
Resea ch con ibu ions
137