Predicting Growing Stock Volume of Eucalyptus Plantations Using 3-D Point Clouds Derived from UAV Imagery and ALS Data

A icle
P edic ing G owing S ock Volume o Eucalyp us
Plan a ions Using 3-D Poin Clouds De i ed om
UAV Image y and ALS Da a
Juan Gue a-He nández 1,2,*, Diogo N. Cosenza 1, Ad ian Ca dil 3, Ca los Albe o Sil a 4,5 ,
B igi e Bo equim 1, Paula Soa es 1, Ma ga ida Sil a 6, Edua do González-Fe ei o 7and
Ramón A. Díaz-Va ela 8
1
Fo es Resea ch Cen e, School o Ag icul u e, Uni e si y o Lisbon, Ins i u o Supe io de Ag onomia (ISA),
Tapada da Ajuda, 1349-017 Lisboa, Po ugal; [email p o ec ed] (D.N.C.);
[email p o ec ed] (B.B.); [email p o ec ed] (P.S.)
23eda a. Cen o de inicia i as emp esa iais. Fundación CEL. O Paloma s/n, 27004 Lugo, Spain
3Tecnosyl a. Pa que Tecnológico de León. 24009 León, Spain; [email p o ec ed]
4Biosciences Labo a o y, NASA Godda d Space Fligh Cen e , G eenbel , MD 20707, USA;
[email p o ec ed]
5Depa men o Geog aphical Sciences, Uni e si y o Ma yland, College Pa k, Ma yland, MD 20740, USA
6RAIZ (Fo es and Pape Resea ch Ins i u e), The Na iga o Company, Rua JoséEs e ão, 3800-783 Eixo,
Po ugal; ma ga ida.sil a@ hena iga o company.com
7Depa amen o de Tecnología Mine a, Topog a ía y de Es uc u as, G upo de In es igación en Geomá ica e
Ingenie
í
a Ca og
á
ica GI-202-GEOINCA, Escuela Supe io y T
é
cnica de Ingenie os de Minas, Uni e sidad
de León, A . de As o ga s/n, Campus de Pon e ada, 24401 Pon e ada, Spain; [email p o ec ed]
8
Biodi e si y and Applied Bo any Uni (GI BIOAPLIC 1809), Depa men o Bo any, Uni e si y o San iago de
Compos ela, Escuela Poli écnica Supe io , R/Benigno Ledo, Campus Uni e si a io, 27002 Lugo, Spain;
[email p o ec ed]
*Co espondence: [email p o ec ed] o [email p o ec ed]; Tel.: +351-21-365-3356
Recei ed: 9 Sep embe 2019; Accep ed: 8 Oc obe 2019; Published: 15 Oc obe 2019


Abs ac :
Es ima ing o es in en o y a iables is impo an in moni o ing o es esou ces and
mi iga ing clima e change. In his espec , o es manage s equi e lexible, non-des uc i e me hods
o es ima ing olume and biomass. High- esolu ion and low-cos emo e sensing da a a e inc easingly
a ailable o measu e h ee-dimensional (3D) canopy s uc u e and o model o es s uc u al a ibu es.
The main objec i e o his s udy was o e alua e and compa e he indi idual ee olume es ima es
de i ed om high-densi y poin clouds ob ained om ai bo ne lase scanning (ALS) and digi al
ae ial pho og amme y (DAP) in Eucalyp us spp. plan a ions. Objec -based image analysis (OBIA)
echniques we e applied o indi idual ee c own (ITC) delinea ion. The ITC algo i hm applied
co ec ly de ec ed and delinea ed 199 ees om ALS-de i ed da a, while 192 ees we e co ec ly
iden i ied using DAP-based poin clouds acqui ed om Unmanned Ae ial Vehicles (UAV), ep esen ing
accu acy le els o espec i ely 62% and 60%. Add essing olume modelling, non-linea eg ession
i based on indi idual ee heigh and indi idual c own a ea de i ed om he ITC p o ided he
ollowing esul s: Model E iciency (Me ) =0.43 and 0.46, Roo Mean Squa e E o (RMSE) =
0.030 m3
and 0.026 m
3
, RMSE =20.31% and 19.97%, and an app oxima ely unbiased esul s (0.025 m
3
and
0.0004 m
3
) using DAP and ALS-based es ima ions, espec i ely. No signi ican di e ence was ound
be ween he obse ed alue ( ield da a) and olume es ima ion om ALS and DAP (p- alue om
- es s a is ic =0.99 and 0.98, espec i ely). The p oposed app oaches could also be used o es ima e
basal a ea o biomass s ocks in Eucalyp us spp. plan a ions.
Keywo ds:
unmanned ae ial ehicles (UAV); o es in en o y; olume; canopy heigh model (CHM);
objec based image analysis (OBIA); s uc u e om mo ion (S M)
Fo es s 2019,10, 905; doi:10.3390/ 10100905 www.mdpi.com/jou nal/ o es s
Fo es s 2019,10, 905 2 o 18
1. In oduc ion
Sus ainable o es managemen demands accu a e in o ma ion ha can be ob ained e icien ly and
apidly [
1
] in o de o desc ibe o es s uc u e and quan i y o es esou ces [
2
]. Howe e , al hough
accu a e, adi ional o es in en o y is esou ce- and ime-consuming, indica ing he need o ei he
al e na i e o complemen a y me hods ha may o e come he d awbacks o ield da a acquisi ion [
3
].
In addi ion, al hough ield-measu ed da a a e commonly assumed o be g ound u h alues o emo e
sensing es ima ions, he associa ed e o s end o be la ge [3–6].
A numbe o al e na i es o adi ional ield-based measu emen o mo phological pa ame e s o
cha ac e izing h ee-dimensional (3D) s uc u e o ees and canopies ha e eme ged [
4
,
7
]. Ai bo ne
Lase Scanning (ALS), Digi al Ae ial Pho og amme y (DAP) and Te es ial Lase Scanning (TLS) ha e
become widely es ablished as o es mapping and moni o ing me hods [
8
–
11
]. In he las wo decades,
DAP, ALS (e.g., [
12
–
14
]) o a combina ion o hese me hods (e.g., [
15
–
17
]) ha e been inc easingly used
o suppo o es in en o ies a di e en scales.
ALS has been he p ima y sou ce o 3D da a on o es e ical s uc u e since he 1990s [
18
,
19
].
The e has been an abundance o esea ch demons a ing he u ili y o ALS o p edic ing o es
biophysical a iables o suppo o es in en o ies a indi idual ee- and s and-le el [
20
,
21
]. Since he
la e 2000s DAP has p o ided a p omising al e na i e, as he accu acy o s and-based es ima es has
been ound o be simila o ha achie ed h ough ALS, a much lowe cos [17,22,23].
Bal sa ias [
12
] p o ided a comp ehensi e compa ison o DAP and ALS da a, highligh ing he
ad an ages and disad an ages o bo h echnologies wi h ega ds o acquisi ion, accu acy, ma u i y
and cos s. Al hough ALS da a ha e many ad an ages (e.g., di ec measu emen o heigh , highe
pene a ion h ough he ege a ion), DAP s ill ep esen s an essen ial sou ce o da a o he o es
in en o y analyses. In ac , pho og amme ic so wa e has de eloped apidly in he pas 15–20 yea s.
Since he i s s udies [
24
], ad ances in compu e ision, image ma ching algo i hms and compu ing
powe ha e p omo ed he use o ae ial images o gene a ing high- esolu ion 3D da a by image
ma ching [
25
]. Many pho og amme ic so wa e packages (p op ie a y and open-sou ce) ha e been
de eloped, o e ing unpa alleled oppo uni ies o p oduce 3D da a om 2D image collec ions wi h
high o e lap.
Recen ad ances in senso s and in image p ocessing –pa icula ly S uc u e om Mo ion (S M)
echnology– ha e also enabled he ex ac ion o dense poin clouds ob ained by DAP [
16
,
24
,
26
–
29
].
In his sense, DAP de i ed om S M is an eme ging sou ce o 3D da a, eaching quali y s anda ds
close o hose p o ided by ALS [
25
,
30
]. wo ks by [
26
,
31
–
33
] poin ed ou he po en ial o DAP o o es
applica ions. Du ing he pas i e yea s in pa icula , he e has been an inc easing in e es in he use o
DAP o gene a e 3D da a analogous o ALS da a, in o de o suppo o es in en o ies [
17
,
22
,
25
,
34
–
39
].
This in e es can be a ibu ed o he need o op imize cos s while imp o ing he empo al esolu ion.
Unmanned Ae ial Vehicles (UAVs), also known as d ones o Unmanned Ae ial Sys ems (UAS),
ha e eme ged as a cos -e ec i e al e na i e o con en ional me hods based on manned ixed-wing
ai c a o helicop e s o DAP image y and ALS da a collec ion [
40
–
43
]. Since he i s s udies in which
UAV-de i ed da a was used o o es in en o y pu poses [
44
,
45
], UAV-based o es y applica ions
o bo h ALS and DAP ha e inc eased subs an ially [
40
,
46
–
52
]. Indeed, ALS and Red-G een-Blue
(RGB) senso s moun ed on UAV pla o ms a e becoming cos -e ec i e ools o moni o ing o es
s uc u e because o hei high spa ial and empo al esolu ion, achie ed by he low ligh heigh ,
ope a ional lexibili y and ela i ely low cos o he ligh su eys, which mee mos o o es manage s
equi emen s [
53
]. In pa icula , ligh UAVs equipped wi h inexpensi e consume g ade came as ha e
ecen ly appea ed as a easible op ion o moni o ing 3D o es s uc u e [
54
]. In addi ion, mul i- empo al
UAV-acqui ed da a can also be used o apid, accu a e and cos -e ec i e ee g ow h assessmen ,
p o iding up- o-da e in o ma ion o suppo decision-making in o es managemen [48,54–58].
Two main s a egies ha e been adop ed o DAP and ALS-based analysis in o es y in en o ies:
(i) he A ea-Based App oach (ABA), a dis ibu ion-based echnique which ypically p o ides da a
a s and le el, and (ii) he indi idual ee c own (ITC) delinea ion, in which indi idual ee c owns,
Fo es s 2019,10, 905 3 o 18
heigh s and posi ions a e he basic uni s o assessmen . ABA has been used wi h ALS and DAP o
es ima e o es a ibu es o e a wide ange o o es ypes including Tempe a e (e.g., [
59
]), Bo eal
(e.g., [13,14,17,34,60]), A lan ic (e.g., [61,62]), T opical (e.g., [37]), Alpine (e.g., [63,64]), Medi e anean
o es s (e.g., [
65
–
69
]) and plan a ions [
70
]. A he s and le el, esul s om ecen esea ch on small- o
medium-sized bo eal and opical o es s ha e demons a ed he po en ial use o UAV-based DAP
da a o es ima ing o es biomass [
53
,
71
]. On he o he hand, ITC has also been applied o DAP poin
clouds [
72
] and o ALS clouds [
73
,
74
]. ITC p esen s se e al ad an ages o e ABA o es ima ion o
abo e-g ound biomass because i can be used o de i e biomass when an allome ic model is a ailable
a indi idual ee le el [
75
]. A he same ime, i is pa icula ly well sui ed o p ecision o es y, which
usually equi es in o ma ion abou indi idual ees.
Finally, ALS and S M app oaches o ee heigh es ima ion end o unde es ima e ee
heigh
[73,74,76,77]
. Recen s udies [
77
] ha e p esen ed a model ha explains he obse ed bias
using p obabili y heo y, de eloping me hods o co ec ing se e al ALS me ics used o ABA
p edic ion o s and s uc u e. Howe e , ew s udies ha e e alua ed he in luence o his bias a
indi idual- ee le el. In his espec , u he esea ch is needed in o de o analyze he in luence o his
bias in he indi idual- ee biomass and olume models o bo h echnologies.
The objec i es o his s udy we e as ollows: (i) o in es iga e he combined use o ALS- and
S M-de i ed indi idual- ee measu emen s (heigh and c own a ea) wi h non-linea eg ession models
o es ima e indi idual ee diame e and olume; and (ii) o compa e he es ima ion o ALS- and
S M-de i ed indi idual- ee olume models o es ima e g owing s ock olume in ela ion o ield da a.
2. Ma e ials and Me hods
2.1. S udy A ea
The s udy a ea is loca ed in he municipali y o Valongo (41.213
◦
N,
−
8.496
◦
W) in he dis ic o
Po o, Po ugal (Figu e 1). The si e consis s o a se en yea old plan a ion o Eucalyp us spp. clonal
ma e ial (G74), co e ing an a ea o 26 ha. T ee spacing was 3.70
×
2.5 m, yielding a densi y o one ee
pe 9.25 m
−2
. The ele a ion anges om 163 o 294 m abo e he WGS84 e e ence ellipsoid. The e ain
is opog aphically complex, wi h s eep slopes (mean slope =24.2%), and o ele a ion up o 131 m.
The mean annual ain all is 1568 mm, and 42.1% o he p ecipi a ion occu s be ween No embe and
Janua y. The mean annual empe a u e is 14.2
◦
C, anging om 8.6
◦
C in he coldes mon hs (Decembe
o Feb ua y) o 20.1
◦
C in he wa mes mon hs (July o Sep embe ). The s udy si e is cha ac e ized by
e enly plan ed ees o supe io gene ic ma e ial wi h a low mo ali y a e.
Fo es s 2019, 10, x FOR PEER REVIEW 3 o 18
ee g ow h assessmen , p o iding up- o-da e in o ma ion o suppo decision-making in o es
managemen [48,54–58].
Two main s a egies ha e been adop ed o DAP and ALS-based analysis in o es y in en o ies:
(i) he A ea-Based App oach (ABA), a dis ibu ion-based echnique which ypically p o ides da a a
s and le el, and (ii) he indi idual ee c own (ITC) delinea ion, in which indi idual ee c owns,
heigh s and posi ions a e he basic uni s o assessmen . ABA has been used wi h ALS and DAP o
es ima e o es a ibu es o e a wide ange o o es ypes including Tempe a e (e.g., [59]), Bo eal
(e.g., [13,14,17,34,60]), A lan ic (e.g., [61,62]), T opical (e.g., [37]), Alpine (e.g., [63,64]), Medi e anean
o es s (e.g., [65–69]) and plan a ions [70]. A he s and le el, esul s om ecen esea ch on small-
o medium-sized bo eal and opical o es s ha e demons a ed he po en ial use o UAV-based DAP
da a o es ima ing o es biomass [53,71]. On he o he hand, ITC has also been applied o DAP poin
clouds [72] and o ALS clouds ([73,74]). ITC p esen s se e al ad an ages o e ABA o es ima ion o
abo e-g ound biomass because i can be used o de i e biomass when an allome ic model is
a ailable a indi idual ee le el [75]. A he same ime, i is pa icula ly well sui ed o p ecision
o es y, which usually equi es in o ma ion abou indi idual ees.
Finally, ALS and S M app oaches o ee heigh es ima ion end o unde es ima e ee heigh
[73,74,76,77]. Recen s udies [77] ha e p esen ed a model ha explains he obse ed bias using
p obabili y heo y, de eloping me hods o co ec ing se e al ALS me ics used o ABA p edic ion
o s and s uc u e. Howe e , ew s udies ha e e alua ed he in luence o his bias a indi idual- ee
le el. In his espec , u he esea ch is needed in o de o analyze he in luence o his bias in he
indi idual- ee biomass and olume models o bo h echnologies.
The objec i es o his s udy we e as ollows: (i) o in es iga e he combined use o ALS- and S M-
de i ed indi idual- ee measu emen s (heigh and c own a ea) wi h non-linea eg ession models o
es ima e indi idual ee diame e and olume; and (ii) o compa e he es ima ion o ALS- and S M-
de i ed indi idual- ee olume models o es ima e g owing s ock olume in ela ion o ield da a.
2. Ma e ials and Me hods
2.1. S udy A ea
The s udy a ea is loca ed in he municipali y o Valongo (41.213° N, −8.496° W) in he dis ic o
Po o, Po ugal (Figu e 1). The si e consis s o a se en yea old plan a ion o Eucalyp us spp. clonal
ma e ial (G74), co e ing an a ea o 26 ha. T ee spacing was 3.70 × 2.5 m, yielding a densi y o one ee
pe 9.25 m−2. The ele a ion anges om 163 o 294 m abo e he WGS84 e e ence ellipsoid. The e ain
is opog aphically complex, wi h s eep slopes (mean slope = 24.2%), and o ele a ion up o 131 m.
The mean annual ain all is 1568 mm, and 42.1% o he p ecipi a ion occu s be ween No embe and
Janua y. The mean annual empe a u e is 14.2 °C, anging om 8.6 °C in he coldes mon hs
(Decembe o Feb ua y) o 20.1 °C in he wa mes mon hs (July o Sep embe ). The s udy si e is
cha ac e ized by e enly plan ed ees o supe io gene ic ma e ial wi h a low mo ali y a e.
Figu e 1. (a) UAV-de i ed RGB-o homosaic image o he s udy a ea and 10 g ound con ol poin s
(GCPs) ( ed do s) (b) UAV-de i ed Digi al Su ace Model (DSM) gene a ed using Pix4d® so wa e.
Figu e 1.
(
a
) UAV-de i ed RGB-o homosaic image o he s udy a ea and 10 g ound con ol poin s
(GCPs) ( ed do s) (b) UAV-de i ed Digi al Su ace Model (DSM) gene a ed using Pix4d®so wa e.
2.2. Field Measu emen s and Field Volume Es ima ion
The ield da a we e collec ed in Decembe 2016 ( o co espond o he da e o acquisi ion o ALS
da a) om 6 squa e plo s, each o app oxima ely 400 m
2
(Table 1). A o al o 323 e e ence ees we e
Fo es s 2019,10, 905 4 o 18
measu ed and loca ed in 6 squa e plo s (400 m
2
). The heigh o each ee (h, m) wi hin he plo s was
measu ed wi h a Haglo Ve ex IV hypsome e equipped wi h a T3 ansponde . The diame e a
b eas heigh (1.30 m abo e he g ound – d, cm) was measu ed wi h a s eel diame e measu ing ape.
Field plo s we e emeasu ed using he same me hods in Sep embe 2017 (ma ching wi h UAV-based
DAP acquisi ion).
Table 1. S a is ical desc ip ion o he ield da a wi hin six ield plo s o a o al o 323 e e ence ees
Plo d h ^
Mean Mean Mean
Dec 2016 Sep 2017 Dec 2016 Sep 2017 Dec 2016 Sep 2017
P1 13.2 13.4 19.4 20.9 0.13 0.14
P2 12.9 13.3 18.8 19.9 0.12 0.15
P3 13.4 13.8 18.6 20.0 0.13 0.15
P4 13.8 14.1 18.8 19.8 0.13 0.15
P5 13.7 14.0 18.3 19.7 0.13 0.15
P6 13.8 14.2 17.6 19.1 0.13 0.14
Min. 5.3 5.4 9.9 10.3 0.01 0.01
Mean 13.5 13.8 18.6 19.9 0.13 0.14
Max. 17.3 17.8 22.8 23.5 0.24 0.26
SD 1.7 1.7 1.5 1.6 0.03 0.04
A e age (Mean), minimum (Min), maximum (Max) and s anda d de ia ion (SD), alues o he indi idual ee
diame e (d, cm), heigh (h, m) and ield da a olume es ima ion (ˆ
, m3).
In o de o ob ain accu a e posi ions o he ees, opog aphic su eys we e conduc ed o de e mine
he posi ion o he cen e o each ee wi hin he plo s. A T imble
®
TSC3 GPS con olle wi h T imble
®
R8s In eg a e GNSS Sys em An enna (T imble, Sunny ale, CA, USA) (dual- equency eal- ime
kinema ic ecei e –RTK) was used o de e mine he coo dina es o a densi ied geode ic ne wo k o
he s udy a ea by applying eal ime kinema ic (RTK). Based on he ne wo k es ablished wi h GPS,
a opog aphic su ey o he plo s was conduc ed using a T imble
®
M3 Robo ic To al S a ion (T imble,
Sunny ale, CA, USA). Obse a ions on he posi ion o each ee wi hin he plo we e made du ing
he su ey.
Field-de i ed olumes we e es ima ed using he Equa ion (1), p o ided by [78].
ˆ
=0.2105 d
100!1.8191
h1.0703 (1)
whe e
ˆ
is he es ima ed olume (m
3
), dis he diame e (cm) a he b eas heigh (1.30 m) and his he
ee heigh (m).
2.3. ALS Acquisi ion
The ai bo ne su eys we e conduc ed on 17 Decembe 2016, co e ing an a ea o 100 ha. The da a
we e cap u ed wi h Leica ALS80-HP lase scanne ope a ing a pulse a e o 704 kHz, ield o iew
o 6.5
◦
and scan a e o 73.5 Hz, which was moun ed on a Cessna ai plane ha lew he a ea a an
app oxima ely ligh al i ude o 2750 m.a.s.l and an a e age speed o 250 km.h
−1
. The o e lap be ween
sweeps was 30%, achie ing an a e age lase pulse densi y o 43.33 pulses m−2.
2.4. UAV Da a Acquisi ion and Use
The ai bo ne su eys we e conduc ed on 6 Sep embe 2017. An RGB S.O.D.A. 10.2 (20 MP) came a
(senseFly Co, Cheseaux-Lausanne, Swi ze land) was moun ed, wi h nadi iew, on a ixed-wing
UAV (SenseFly eBee) (Figu e 2). The came a, which was equipped wi h a 12.75
×
8.5 mm senso
Fo es s 2019,10, 905 5 o 18
and
5472 ×3648 pixels
de ec o , was used in manual mode. Exposu e se ings (ISO 150 and shu e
speed o 1/1000 s) we e se be o e each ake-o acco ding o he ligh condi ions. This p o ided
~6 cm
pixel
−1
esolu ion o a a iable al i ude abo e g ound le el, which is especially use ul in a eas o
di e se ele a ion ange such as moun ainous egions. A mosphe ic condi ions du ing he ai bo ne
su eys we e cha ac e ized by calm winds, clea ligh ing a he ligh ime (be ween 11.30 am and
12.15 pm) o minimize he e ec o shadowing. Fligh pa ame e s we e de e mined using eMo ion
V. 3.2.4 ligh planning and moni o ing so wa e. The ligh plan co e ed he en i e s udy a ea wi h
longi udinal and la e al o e laps o 85% in bo h cases. The ligh line spacing was 25 m (Figu e 2).
In o al, 744 images we e used o gene a e o homosaics and Digi al Su ace Models (DSMs) by he S M
image econs uc ion p ocess. Two-block ligh s we e equi ed o cap u e he en i e o es s udy a ea
( he o homosaic co e ed an a ea 103.70 ha wi h a e age G ound Sample Dis ance (GSD) o 5.95 cm).
Fo es s 2019, 10, x FOR PEER REVIEW 5 o 18
and 5472 × 3648 pixels de ec o , was used in manual mode. Exposu e se ings (ISO 150 and shu e
speed o 1/1000 s) we e se be o e each ake-o acco ding o he ligh condi ions. This p o ided
~6 cm pixel−1 esolu ion o a a iable al i ude abo e g ound le el, which is especially use ul in a eas
o di e se ele a ion ange such as moun ainous egions. A mosphe ic condi ions du ing he ai bo ne
su eys we e cha ac e ized by calm winds, clea ligh ing a he ligh ime (be ween 11.30 am and
12.15 pm) o minimize he e ec o shadowing. Fligh pa ame e s we e de e mined using eMo ion
V. 3.2.4 ligh planning and moni o ing so wa e. The ligh plan co e ed he en i e s udy a ea wi h
longi udinal and la e al o e laps o 85% in bo h cases. The ligh line spacing was 25 m (Figu e 2). In
o al, 744 images we e used o gene a e o homosaics and Digi al Su ace Models (DSMs) by he S M
image econs uc ion p ocess. Two-block ligh s we e equi ed o cap u e he en i e o es s udy a ea
( he o homosaic co e ed an a ea 103.70 ha wi h a e age G ound Sample Dis ance (GSD) o 5.95 cm).
Figu e 2. UAV Came a (le ) and ligh design ( igh ).
2.5. 3D Model Gene a ion and P ep ocessing Poin Clouds
The absolu e o ien a ion o he ae ial pho os was de e mined using ae o iangula ion echniques
implemen ed in pix4D 3.3.29 (pix4D®, Ecublens, Swi ze land). A se o 10 g ound con ol poin s
(GCPs) measu ed in he ield wi h opog aphic me hods was used o geo e e ence he S M mosaics
o a p ojec ed coo dina e sys em o bo h da ase s. The g ound con ol pho og amme ic poin s
(GCPs) we e cap u ed wi h a T imble TSC3 con olle and a T imble R8s GNSS an enna (RTK
p ecision 8 mm + 1 ppm Ho izon al/15 mm + 0.5 ppm Ve ical) moun ed on a pole. The GCPs
ma ke s comp ised a se o 1 × 1 m c oss-shaped whi e pain ed imbe planks wi h some black and
whi e 50 × 50 cm pain ed checke boa ds. Fo eliable accu acy o GPS measu emen , all GCPs we e
loca ed in open a eas wi h no canopy co e . A each poin , GPS signals we e logged in RTK–global
na iga ion sa elli e sys em (GNSS) mode. The eco dings we e p ocessed wi h eal- ime co ec ion
da a e ie ed om he ixed base s a ion in Gaia (Po o) (la i ude: 41° 06' 21.67048" N, longi ude:
8° 35' 20.73434" W, and ellipsoidal ele a ion: 287.63 m abo e he WGS84 e e ence ellipsoid).
Pho og amme ic poin clouds we e compu ed using S M echniques, implemen ed in Pix4D
3.3.29. The ma ching pa ame e s o poin cloud densi ica ion we e se as ollows: mul iscale, image
scale = 1/2 (hal image size) and poin densi y = ‘op imal’. The minimum numbe o ma ched images
was also se o 3. DEMS M was gene a ed om he g ound poin s by using a na u al neighbo
in e pola ion echnique implemen ed in Pix4D (addi ional de ails o he algo i hms a e p op ie a y
and we e no disclosed by Pix4D).
The ALS and S M poin clouds p ep ocessed using FUSION/LDV 3.60 so wa e [79] and
LasTools [80]. Fo mo e de ails o poin cloud p ocessing see de ails in [74]. Finally, wo CHMs
(CHMS M and CHMALS) we e ob ained by sub ac ing he DEMs (DEMALS and DEMS M) om he DSMs
(DSMALS and DSMS M) in he FUSION LiDAR Toolki [79].
Figu e 2. UAV Came a (le ) and ligh design ( igh ).
2.5. 3D Model Gene a ion and P ep ocessing Poin Clouds
The absolu e o ien a ion o he ae ial pho os was de e mined using ae o iangula ion echniques
implemen ed in pix4D 3.3.29 (pix4D
®
, Ecublens, Swi ze land). A se o 10 g ound con ol poin s
(GCPs) measu ed in he ield wi h opog aphic me hods was used o geo e e ence he S M mosaics o a
p ojec ed coo dina e sys em o bo h da ase s. The g ound con ol pho og amme ic poin s (GCPs)
we e cap u ed wi h a T imble TSC3 con olle and a T imble R8s GNSS an enna (RTK p ecision 8 mm +
1 ppm Ho izon al/15 mm +0.5 ppm Ve ical) moun ed on a pole. The GCPs ma ke s comp ised a se
o 1
×
1 m c oss-shaped whi e pain ed imbe planks wi h some black and whi e 50
×
50 cm pain ed
checke boa ds. Fo eliable accu acy o GPS measu emen , all GCPs we e loca ed in open a eas wi h
no canopy co e . A each poin , GPS signals we e logged in RTK–global na iga ion sa elli e sys em
(GNSS) mode. The eco dings we e p ocessed wi h eal- ime co ec ion da a e ie ed om he ixed
base s a ion in Gaia (Po o) (la i ude: 41
◦
06
0
21.67048” N, longi ude: 8
◦
35
0
20.73434” W, and ellipsoidal
ele a ion: 287.63 m abo e he WGS84 e e ence ellipsoid).
Pho og amme ic poin clouds we e compu ed using S M echniques, implemen ed in Pix4D 3.3.29.
The ma ching pa ame e s o poin cloud densi ica ion we e se as ollows: mul iscale, image scale =
1/2 (hal image size) and poin densi y =‘op imal’. The minimum numbe o ma ched images was also
se o 3. DEM
S M
was gene a ed om he g ound poin s by using a na u al neighbo in e pola ion
echnique implemen ed in Pix4D (addi ional de ails o he algo i hms a e p op ie a y and we e no
disclosed by Pix4D).
The ALS and S M poin clouds p ep ocessed using FUSION/LDV 3.60 so wa e [
79
] and
LasTools [
80
]. Fo mo e de ails o poin cloud p ocessing see de ails in [
74
]. Finally, wo CHMs
(CHM
S M
and CHM
ALS
) we e ob ained by sub ac ing he DEMs (DEM
ALS
and DEM
S M
) om he
DSMs (DSMALS and DSMS M) in he FUSION LiDAR Toolki [79].

Fo es s 2019,10, 905 6 o 18
2.6. ITC P ocess o De i e ALS- and S M-Va iables
Indi idual ee posi ion (Xand Ycoo dina es), heigh (h
S M
, h
ALS
) and c own a ea (ca
S M
, ca
ALS
)
we e e ie ed om he espec i e CHM
S M
and CHM
ALS
(Figu e 2). Resampling o he CHMs o
20 cm esolu ion and subsequen smoo hing wi h mean il e (5
×
5 window) in he case o ALS
and median il e s (3
×
3 window) o S M we e conduc ed using he FUSION LiDAR Toolki [
79
].
C own delinea ion ollowed he p ocedu e de ailed in [
81
]. The p ocess is di ided in o h ee main
phases: segmen a ion, classi ica ion and i e a i e wa e shed segmen a ion The Chessboa d Segmen a ion
algo i hm was used o spli he image in o squa e image objec s. In he second phase, a Classi ica ion
algo i hm was used o classi y image objec s om he smoo hed CHM. Objec s wi h an ele a ion
alue o less han 5 m we e classi ied as gaps. The h eshold was es ablished empi ically om ield
obse a ions and by ial-and-e o es s. The emaining objec s we e assigned o he ‘ empo a y
canopy’ class. These objec s we e used o loca e ee ops and delinea e ee c owns in he ollowing
i e a i e wa e shed segmen a ion p ocesses. In he i e a ion, he Find Local Ex ema algo i hm was used
o classi y he image objec s o he ‘ empo a y canopy’ class, which ul ills a local ex eme condi ion
acco ding o image objec ea u es wi hin a sea ch domain in hei neighbo hoods. Howe e , because
o he o es s and and ee species cha ac e is ics, he ini ial maximum sea ch domain used in he
i e a i e p ocess (see Figu e 3 in [
81
]) o de ec op ees was changed om 5 o 3, and 4 in e ac ions
we e applied. A sea ch wi h a a iable squa e window enables de ec ion o apices o ees wi h a
la ge a ie y o c own sizes. Objec s less han 3 m away om any de ec ed ee op we e e ained
in he ‘ empo a y canopy’ class (candida es o wa e shed) and any o he objec s we e dis ega ded.
This dis ance was he maximum obse ed c own wid h in he plo s, which was conside ed he limi o
c own g owing in he nex s ep. Then, he c own delinea ion esul s (Figu e 3) and ee op posi ions
we e expo ed in ESRI
TM
shape iles as ec o polygons and poin s espec i ely, o subsequen analysis.
Fo es s 2019, 10, x FOR PEER REVIEW 6 o 18
2.6. ITC P ocess o De i e ALS- and S M-Va iables
Indi idual ee posi ion (X and Y coo dina es), heigh (hS M, hALS) and c own a ea (caS M, caALS) we e
e ie ed om he espec i e CHMS M and CHMALS (Figu e 2). Resampling o he CHMs o 20 cm
esolu ion and subsequen smoo hing wi h mean il e (5 × 5 window) in he case o ALS and median
il e s (3 × 3 window) o S M we e conduc ed using he FUSION LiDAR Toolki [79]. C own
delinea ion ollowed he p ocedu e de ailed in [81]. The p ocess is di ided in o h ee main phases:
segmen a ion, classi ica ion and i e a i e wa e shed segmen a ion The Chessboa d Segmen a ion
algo i hm was used o spli he image in o squa e image objec s. In he second phase, a Classi ica ion
algo i hm was used o classi y image objec s om he smoo hed CHM. Objec s wi h an ele a ion
alue o less han 5 m we e classi ied as gaps. The h eshold was es ablished empi ically om ield
obse a ions and by ial-and-e o es s. The emaining objec s we e assigned o he ‘ empo a y
canopy’ class. These objec s we e used o loca e ee ops and delinea e ee c owns in he ollowing
i e a i e wa e shed segmen a ion p ocesses. In he i e a ion, he Find Local Ex ema algo i hm was used
o classi y he image objec s o he ‘ empo a y canopy’ class, which ul ills a local ex eme condi ion
acco ding o image objec ea u es wi hin a sea ch domain in hei neighbo hoods. Howe e , because
o he o es s and and ee species cha ac e is ics, he ini ial maximum sea ch domain used in he
i e a i e p ocess (see Figu e 3 in [81]) o de ec op ees was changed om 5 o 3, and 4 in e ac ions
we e applied. A sea ch wi h a a iable squa e window enables de ec ion o apices o ees wi h a
la ge a ie y o c own sizes. Objec s less han 3 m away om any de ec ed ee op we e e ained in
he ‘ empo a y canopy’ class (candida es o wa e shed) and any o he objec s we e dis ega ded. This
dis ance was he maximum obse ed c own wid h in he plo s, which was conside ed he limi o
c own g owing in he nex s ep. Then, he c own delinea ion esul s (Figu e 3) and ee op posi ions
we e expo ed in ESRITM shape iles as ec o polygons and poin s espec i ely, o subsequen
analysis.
Figu e 3. Examples o he canopy heigh models (cool o wa m colo s ep esen ing low o high
heigh s), c own delinea ion (blue lines), and ee op posi ions ( ed do s) wi hin example plo s using
ai bo ne lase scanning (ALS) (a) and S uc u e om Mo ion (S M) (b).
Figu e 3.
Examples o he canopy heigh models (cool o wa m colo s ep esen ing low o high heigh s),
c own delinea ion (blue lines), and ee op posi ions ( ed do s) wi hin example plo s using ai bo ne
lase scanning (ALS) (a) and S uc u e om Mo ion (S M) (b).
Fo es s 2019,10, 905 7 o 18
2.7. Indi idual T ee Volume Es ima ion
Volume equa ion (Equa ion (1)) equi es he measu emen o ee diame e o ci cum e ence,
which is no a ailable om UAV image y. We he e o e es ed wo app oaches o es ima ing ALS- and
S M-de i ed indi idual- ee olumes (
S M
and
ALS
) (Figu e 4). In he i s app oach, he mul iplica i e
(powe unc ion) model in Equa ion (2) was i ed using d( om ield da a) as he dependen a iable
and he pai s o explana o y a iables h
S M
,ca
S M
o S M, o h
ALS
,ca
ALS
o ALS. The p edic ed
diame e ob ained by each me hod (d
S M
and d
ALS
) and hei espec i e heigh es ima es (h
S M
and h
ALS
)
we e hen included as independen a iables in Equa ion (1) o p edic he indi idual olumes o
he subse o 192 ees o ALS and 199 o S M ( S M and ALS, espec i ely). In he second app oach,
he mul iplica i e (powe unc ion) model in Equa ion (3) was also i ed o p edic
S M
and
ALS
o
he 192 and 199 ees espec i ely, bu was conside ed a dependen a iable (es ima ed using he
ield-measu ed dand hin Equa ion (1)), and he pai s h
S M
,ca
S M
( om S M) o h
ALS
,ca
ALS
( om ALS)
we e conside ed explana o y a iables, wi hou he need o es ima e he diame e s.
ˆ
d=hβ0caβ1+ε(2)
ˆ
=hβ0caβ1+ε(3)
whe e
ˆ
is he es ima ed olume (m
3
),
ˆ
d
is he es ima ed ee diame e (cm), his he ee heigh (m),
ca is he canopy a ea (m), gene a ed om ALS o S M,
β0
,
β1
, a e he exponen ial pa ame e s o be
es ima ed by non-linea eg ession analysis; and
ε
is he addi i e andom e o . The models we e i ed
using he Non-linea Leas Squa es nls unc ion implemen ed in he BASE package o R so wa e (R
Co e Team, 2018).
Finally, he Model E iciency (Me , Equa ion (4)), he o e all oo mean squa e e o (RMSE,
Equa ion (5)), he ela i e oo mean squa e e o ( RMSE, Equa ion (6)) and he Bias (Equa ion (7))
we e compu ed in o de o de e mine he accu acy o ALS and S M models o es ima ing diame e
and olume wi h he second app oach. Me compa es p edic ions di ec ly wi h obse ed da a using
a s a is ic analogous o R
2
[
82
]. This s a is ic p o ides a simple index o pe o mance on a ela i e
scale, whe e 1 indica es a ‘pe ec ’ i , 0 e eals ha he model is no be e han a simple a e age, and
nega i e alues indica e a poo model.
Me =1−
(n−1)Pn
i=1(yi−ˆ
yi)2
(n−p)Pn
i=1yi−y2
(4)
RMSE =sPn
i=1(yi−ˆ
yi)2
n(5)
RMSE =RMSE
y∗100 (6)
Bias =Pn
i=1(ˆ
yi−yi)
n(7)
whe e nis he numbe o ees; y
i
is he ield-measu ed ee diame e i;
y
is he he mean obse ed alue
o he ield-measu ed diame e s;
ˆ
yi
is he es ima ed alue o diame e de i ed om he non-linea
eg ession model and pis he numbe o pa ame e s in he models.
Finally, using he co ec ly de ec ed and delinea ed ees, dwas compa ed wi h d
S M
, d
ALS
and
wi h
S M
,
ALS
in he subsample o 192 ees o S M and 199 o ALS, espec i ely. Es ima ed and
obse ed alues we e plo ed and isually examined. A pai ed - es was conduc ed o compa e ALS-
and S M-p edic ed a iables (d
S M
, d
ALS
,
S M
, and
ALS
) o e i y he signi icance o he de ia ions
be ween he obse ed and es ima ed alues. Howe e , hese de ia ions we e p e iously checked using
Fo es s 2019,10, 905 8 o 18
he Shapi o-Wilk es [
83
], which indica ed ha he dis ibu ions mee he assump ion o no mali y.
The es s we e conduc ed a a 5%signi icance le el.
1
Figu e 4. Summa y s eps o indi idual ee c own (ITC) o map olume.
3. Resul s
Field, ALS and S M Volume Es ima ion
Table 2shows he pa ame e es ima es and goodness-o - i s a is ics o he models used o p edic
d(cm) in he i s app oach, and di ec ly es ima ed by S M- and ALS- a iables in he second app oach.
In he i s app oach, non-linea eg ession yielded an Me alue o 0.45 o he S M-es ima ed
diame e and 0.47 o ALS-es ima ed diame e (RMSE =1.17 and 1.12 cm, RMSE o 8.49 % and 8.31%,
espec i ely). Al hough he UAV-based DAP me hod ends o unde es ima e ee heigh ela i e o
ield measu emen s (hypsome e s), he e was no app eciable bias h oughou he obse ed diame e
(Figu e 5a,b). The bias alues (0.38 and 0.35 cm) indica ed a sligh endency o o e es ima e he ini ial
diame e alues om ield da a (Figu e 5a,b). On he o he hand, al hough dwas no di ec ly measu ed
Fo es s 2019,10, 905 9 o 18
in CHMs de i ed om UAV and ALS, h
S M
and h
ALS
we e, and hese a iables we e signi ican in
he S M and ALS equa ions. Fo d
S M
and d
ALS
modelling, he c own a ea (ca
S M
and ca
ALS
) was also
s a is ically signi ican (p<0.05 and p<0.001, espec i ely).
Table 2. Models selec ed o es ima ing S M and ALS de i ed indi idual ee diame e and olume.
App oach Dependen
a iable P edic o s Pa ame e
es ima e
S anda d
e o p- alue Me RMSE
(cm)
RMSE
(%)
bias
(cm)
1s
dS M
Cons an 0.863 1.170 <0.001
0.45 1.17 8.49 0.38
hS M 0.907 0.108 <0.001
caS M 0.037 0.037 0.013
dALS
Cons an 0.564 0.151 <0.001
0.47 1.12 8.31 0.35
hALS 1.042 0.090 <0.001
caALS 0.062 0.015 <0.001
App oach Dependen
a iable P edic o s Pa ame e
es ima e
S anda d
e o p- alue Me RMSE
(m3)
RMSE
(%)
bias
(m3)
2nd
S M
Cons an 0.004 0.002 0.082
0.43 0.030 20.31
0.0016
hS M 1.192 0.201 <0.001
caS M 0.151 0.035 <0.001
ALS
Cons an 0.001 0.000 0.106
0.46 0.026 19.97
0.0004
hALS 1.828 0.224 <0.001
caALS 0.024 0.037 <0.001
h
S M
and h
ALS
a e he S M and ALS-de i ed ee heigh (m), ca
S M
and ca
ALS
a e he S M and ALS-de i ed indi idual
c own a ea (m
2
), Me is he model e iciency s a is ic, RMSE is he oo mean squa ed e o and RMSE is he ela i e
oo mean squa e e o .
Fo es s 2019, 10, x FOR PEER REVIEW 9 o 18
hS M and hALS a e he S M and ALS-de i ed ee heigh (m), caS M and caALS a e he S M and ALS-de i ed indi idual
c own a ea (m2), Me is he model e iciency s a is ic, RMSE is he oo mean squa ed e o and RMSE is he
ela i e oo mean squa e e o .
In he i s app oach, non-linea eg ession yielded an Me alue o 0.45 o he S M-es ima ed
diame e and 0.47 o ALS-es ima ed diame e (RMSE = 1.17 and 1.12 cm, RMSE o 8.49 % and 8.31%,
espec i ely). Al hough he UAV-based DAP me hod ends o unde es ima e ee heigh ela i e o
ield measu emen s (hypsome e s), he e was no app eciable bias h oughou he obse ed diame e
(Figu e 5a,b). The bias alues (0.38 and 0.35 cm) indica ed a sligh endency o o e es ima e he ini ial
diame e alues om ield da a (Figu e 5a,b). On he o he hand, al hough d was no di ec ly
measu ed in CHMs de i ed om UAV and ALS, hS M and hALS we e, and hese a iables we e
signi ican in he S M and ALS equa ions. Fo dS M and dALS modelling, he c own a ea (caS M and caALS)
was also s a is ically signi ican (p < 0.05 and p < 0.001, espec i ely).
In he case o ALS modelling, he second app oach yielded an Me alue o 0.56. The mean RMSE
o es ima ion was 20.31% (0.030 m3) when calcula ed on he basis o he S M cloud, and 19.97%
(0.026 m3) when based on he ALS cloud. The e we e no app eciable biases om he models
h oughou he obse ed olume ange using bo h app oaches (Figu e 5 c,d,e, ). Howe e , he
endency o ALS and S M o unde es ima e h may be he main eason o he sligh unde es ima ion
o in he i s app oach (Figu e 5c,d). In he case o he second app oach, a sligh ly posi i e bias
(0.0004 and 0.0016 m3) indica ed sligh o e es ima ion when olume was modeled di ec ly o m ALS-
and S M- a iables (Figu e 5e, ).
Figu e 5. Sca e plo s o ALS and S M-de i ed a iables agains ield-de i ed a iables: (a) ield-
measu ed ee diame e (d) agains ALS-es ima ed ee diame e (dALS); (b) ield-measu ed ee
diame e (d) agains S M-es ima ed ee diame e (dS M); (c) ield-es ima ed olume ( ield) agains ALS-
es ima ed olume ( ALS) using he i s app oach; (d) ield-es ima ed olume ( ield) agains S M-
es ima ed olume ( S M) using he i s app oach; (e) ield-es ima ed olume ( ield) agains ALS-
es ima ed olume ( ALS) using he second app oach; ( ) ield-es ima ed olume ( ield) agains S M -
es ima ed olume ( S M) using he second app oach.
The - es (Table 3) showed ha he e we e no e idence o signi ican di e ences be ween obse ed
and es ima ed alues o diame e (p- alues o 0.98 o bo h app oaches in he subsample o 192 ees
Figu e 5.
Sca e plo s o ALS and S M-de i ed a iables agains ield-de i ed a iables:
(
a
) ield-measu ed ee diame e (d) agains ALS-es ima ed ee diame e (d
ALS
); (
b
) ield-measu ed
ee diame e (d) agains S M-es ima ed ee diame e (d
S M
); (
c
) ield-es ima ed olume (
ield
) agains
ALS-es ima ed olume (
ALS
) using he i s app oach; (
d
) ield-es ima ed olume (
ield
) agains
S M-es ima ed olume (
S M
) using he i s app oach; (
e
) ield-es ima ed olume (
ield
) agains
ALS-es ima ed olume (
ALS
) using he second app oach; (
) ield-es ima ed olume (
ield
) agains S M
-es ima ed olume ( S M) using he second app oach.
Fo es s 2019,10, 905 16 o 18
46.
Dandois, J.P.; Ellis, E.C. High Spa ial Resolu ion Th ee-Dimensional Mapping o Vege a ion Spec al Dynamics
Using Compu e Vision. Remo e Sens. En i on. 2013,136, 259–276. [C ossRe ]
47.
Zahawi, R.A.; Dandois, J.P.; Holl, K.D.; Nadwodny, D.; Reid, J.L.; Ellis, E.C. Using Ligh weigh Unmanned
Ae ial Vehicles o Moni o T opical Fo es Reco e y. Biol. Conse . 2015,186, 287–295. [C ossRe ]
48.
Gue a-H
é
nandez, J.; Gonz
á
lez-Fe ei o, E.; Sa men o, A.; Sil a, J.; Nunes, A.; Co eia, A.C.; Fon es, L.;
Tom
é
, M.; D
í
az-Va ela, R. Sho Communica ion. Using High Resolu ion UAV Image y o Es ima e T ee
Va iables in Pinus Pinea Plan a ion in Po ugal. Fo . Sys . 2016,25, 9. [C ossRe ]
49.
Mohan, M.; Sil a, C.A.; Klaube g, C.; Ja , P.; Ca s, G.; Ca dil, A.; Hudak, A.T.; Dia, M. Indi idual T ee
De ec ion om Unmanned Ae ial Vehicle (UAV) De i ed Canopy Heigh Model in an Open Canopy Mixed
Coni e Fo es . Fo es s 2017,8, 340. [C ossRe ]
50.
Thiel, C.; Schmullius, C. Compa ison o UAV Pho og aph-Based and Ai bo ne Lida -Based Poin Clouds
o e Fo es om a Fo es y Applica ion Pe spec i e. In . J. Remo e Sens. 2017,38, 2411–2426. [C ossRe ]
51.
Ca dil, A.; Vepakomma, U.; B o ons, L. Assessing Pine P ocessiona y Mo h De olia ion Using Unmanned
Ae ial Sys ems. Fo es s 2017,8, 402. [C ossRe ]
52.
Na a o, J.; Algee , N.; Fe n
á
ndez-Landa, A.; Es eban, J.; Rod
í
guez-No iega, P.; Guill
é
n-Climen , M.
In eg a ion o UAV, Sen inel-1, and Sen inel-2 Da a o Mang o e Plan a ion Abo eg ound Biomass
Moni o ing in Senegal. Remo e Sens. 2019,11, 77. [C ossRe ]
53.
Puli i, S.; Ø ka, H.O.; Gobakken, T.; Næsse , E. In en o y o Small Fo es A eas Using an Unmanned Ae ial
Sys em. Remo e Sens. 2015,7, 9632–9654. [C ossRe ]
54.
To esan, C.; Be on, A.; Ca o enu o, F.; Di Genna o, S.F.; Gioli, B.; Ma ese, A.; Miglie a, F.; Vagnoli, C.;
Zaldei, A.; Wallace, L. Fo es y Applica ions o UAVs in Eu ope: A Re iew. In . J. Remo e Sens.
2017
,38,
2427–2447. [C ossRe ]
55.
Whi ehead, K.; Hugenhol z, C.H. Remo e Sensing o he En i onmen wi h Small Unmanned Ai c a Sys ems
(UASs), Pa 1: A Re iew o P og ess and Challenges 1. J. Unmanned Veh. Sys . 2014,2, 69–85. [C ossRe ]
56.
Tang, L.; Shao, G. D one Remo e Sensing o Fo es y Resea ch and P ac ices. J. Fo . Res.
2015
, 1–7. [C ossRe ]
57.
Gue a-He n
á
ndez, J.; Gonz
á
lez-Fe ei o, E.; Monle
ó
n, V.; Faias, S.; Tom
é
, M.; D
í
az-Va ela, R. Use o
Mul i-Tempo al UAV-De i ed Image y o Es ima ing Indi idual T ee G ow h in Pinus Pinea S ands. Fo es s
2017,8, 300. [C ossRe ]
58.
P
á
dua, L.; H uška, J.; Bessa, J.; Ad
ã
o, T.; Ma ins, L.M.; Gonçal es, J.A.; Pe es, E.; Sousa, A.M.; Cas o, J.P.;
Sousa, J.J. Mul i-Tempo al Analysis o Fo es y and Coas al En i onmen s Using UASs. Remo e Sens.
2017
,
10, 24. [C ossRe ]
59.
Hall, S.A.; Bu ke, I.C.; Box, D.O.; Kau mann, M.R.; S oke , J.M. Es ima ing S and S uc u e Using Disc e e-Re u n
Lida : An Example om Low Densi y, Fi e P one Ponde osa Pine Fo es s. Fo . Ecol. Manag.
2005
,208, 189–209.
[C ossRe ]
60.
Jä ns ed , J.; Pekka inen, A.; Tuominen, S.; Ginzle , C.; Holopainen, M.; Vii ala, R. Fo es Va iable Es ima ion
Using a High-Resolu ion Digi al Su ace Model. ISPRS J. Pho og amm. Remo e Sens. 2012,74, 78–84. [C ossRe ]
61.
Gonz
á
lez-Fe ei o, E.; Di
é
guez-A anda, U.; Mi anda, D. Es ima ion o S and Va iables in Pinus Radia a D.
Don Plan a ions Using Di e en LiDAR Pulse Densi ies. Fo es y 2012,85, 281–292. [C ossRe ]
62.
Gonz
á
lez-Fe ei o, E.; Di
é
guez-A anda, U.; C ecen e-Campo, F.; Ba ei o-Fe n
á
ndez, L.; Mi anda, D.;
Cas edo-Do ado, F. Modelling Canopy Fuel Va iables o Pinus Radia a D. Don in NW Spain wi h Low-Densi y
LiDAR Da a. In . J. Wildland Fi e 2014,23, 350–362. [C ossRe ]
63.
Mon aghi, A.; Co ona, P.; Dalpon e, M.; Gianelle, D.; Chi ici, G.; Olsson, H. Ai bo ne Lase Scanning o
Fo es Resou ces: An O e iew o Resea ch in I aly as a Commen a y Case S udy. In . J. Appl. Ea h Obs.
Geoin o ma ion 2013,23, 288–300. [C ossRe ]
64.
Co ona, P.; Ca isano, R.; Sal a i, R.; Chi ici, G.; Flo is, A.; Di Ma ino, P.; Ma che i, M.; Sc inzi, G.;
Clemen el, F.; To esan, C. Ai bo ne Lase Scanning o Suppo Fo es Resou ce Managemen unde Alpine,
Tempe a e and Medi e anean En i onmen s in I aly. Eu . J. Remo e Sens. 2012,45, 27–37.
65.
Gonz
á
lez-Olaba ia, J.-R.; Rod
í
guez, F.; Fe n
á
ndez-Landa, A.; Mola-Yudego, B. Mapping Fi e Risk in he
Model Fo es o U bi
ó
n (Spain) Based on Ai bo ne LiDAR Measu emen s. Fo . Ecol. Manag.
2012
,282, 149–156.
[C ossRe ]
66.
Gue a-He n
á
ndez, J.; Gö gens, E.B.; Ga c
í
a-Gu i
é
ez, J.; Rod iguez, L.C.E.; Tom
é
, M.; Gonz
á
lez-Fe ei o, E.
Compa ison o ALS Based Models o Es ima ing Abo eg ound Biomass in Th ee Types o Medi e anean
Fo es . Eu . J. Remo e Sens. 2016,49, 185–204. [C ossRe ]

Fo es s 2019,10, 905 17 o 18
67.
Mon ealeg e, A.L.; Lamelas, M.T.; Tanase, M.A.; de la Ri a, J. Es imaci
ó
n de La Se e idad En Incendios
Fo es ales a Pa i de Da os LiDAR-PNOA y Valo es de Composi e Bu n Index. Re . Telede ec.
2017
,49, 1–16.
[C ossRe ]
68.
Mon ealeg e, A.L.; Lamelas, M.T.; de la Ri a, J.; Ga c
í
a-Ma
í
n, A.; Esc ibano, F. Use o Low Poin Densi y
ALS Da a o Es ima e S and-Le el S uc u al Va iables in Medi e anean Aleppo Pine Fo es . Fo . In . J. Fo .
Res. 2016,89, 373–382. [C ossRe ]
69.
Domingo, D.; Alonso, R.; Lamelas, M.T.; Mon ealeg e, A.L.; Rod
í
guez, F.; de la Ri a, J. 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 Sens. 2019,11, 261. [C ossRe ]
70.
Sil a, C.A.; Klaube g, C.; Hudak, A.T.; Vie ling, L.A.; Liesenbe g, V.; Ca alho, S.P.E.; Rod iguez, L.C. A
P incipal Componen App oach o P edic ing he S em Volume in Eucalyp us Plan a ions in B azil Using
Ai bo ne LiDAR Da a. Fo . In . J. Fo . Res. 2016,89, 422–433. [C ossRe ]
71.
Kachamba, D.J.; Ø ka, H.O.; Gobakken, T.; Eid, T.; Mwase, W. Biomass Es ima ion Using 3D Da a om
Unmanned Ae ial Vehicle Image y in a T opical Woodland. Remo e Sens. 2016,8, 968. [C ossRe ]
72.
S -Onge, B.; Aude , F.-A.; B
é
gin, J. Cha ac e izing he Heigh S uc u e and Composi ion o a Bo eal Fo es
Using an Indi idual T ee C own App oach Applied o Pho og amme ic Poin Clouds. Fo es s
2015
,6,
3899–3922. [C ossRe ]
73.
Mielca ek, M.; S e e´nczak, K.; Khos a ipou , A. Tes ing and E alua ing Di e en LiDAR-De i ed Canopy
Heigh Model Gene a ion Me hods o T ee Heigh Es ima ion. In . J. Appl. Ea h Obs. Geoin o ma ion
2018
,
71, 132–143. [C ossRe ]
74.
Gue a-He n
á
ndez, J.; Cosenza, D.N.; Rod iguez, L.C.E.; Sil a, M.; Tom
é
, M.; D
í
az-Va ela, R.A.;
Gonz
á
lez-Fe ei o, E. Compa ison o ALS-and UAV (S M)-De i ed High-Densi y Poin Clouds o Indi idual
T ee De ec ion in Eucalyp us Plan a ions. In . J. Remo e Sens. 2018,39, 5211–5235.
75.
Mal amo, M.; Ee ikäinen, K.; Packal
é
n, P.; Hyyppä, J. Es ima ion o S em Volume Using Lase Scanning-Based
Canopy Heigh Me ics. Fo es y 2006,79, 217–229. [C ossRe ]
76.
Hen z,
Â
.M.; Sil a, C.A.; Dalla Co e, A.P.; Ne o, S.P.; S age , M.P.; Klaube g, C. Es ima ing Fo es Uni o mi y
in Eucalyp us Spp. and Pinus Taeda L. S ands Using Field Measu emen s and S uc u e om Mo ion Poin
Clouds Gene a ed om Unmanned Ae ial Vehicle (UAV) Da a Collec ion. Fo . Sys . 2018,27, 5. [C ossRe ]
77.
Roussel, J.-R.; Caspe sen, J.; B
é
land, M.; Thomas, S.; Achim, A. Remo ing Bias om LiDAR-Based Es ima es
o Canopy Heigh : Accoun ing o he E ec s o Pulse Densi y and Foo p in Size. Remo e Sens. En i on.
2017,198, 1–16. [C ossRe ]
78.
Tom
é
, M.; Tom
é
, J.; Ribei o, F.; Faias, S. Equaç
õ
es de Volume To al, Volume Pe cen ual e de Pe il Do T onco
Pa a Eucalyp us Globulus Labill. Em Po ugal. Sil a Lusi . 2007,15, 25–39.
79.
McGaughey, R.J. FUSION/LDV: So wa e o LIDAR Da a Analysis and Visualiza ion; Ve sion 3.60+; Paci ic
No hwes Resea ch S a ion, Uni ed S a es Depa men o Ag icul u e Fo es Se ice: Sea le, WA, USA, 2017.
80.
Isenbu g, M. LAS ools—E icien Tools o LiDAR P ocessing; Ve sion 160921; Academic: Camb idge, MA, USA,
2016.
81.
Gonz
á
lez-Fe ei o, E.; Di
é
guez-A anda, U.; Ba ei o-Fe n
á
ndez, L.; Buj
á
n, S.; Ba bosa, M.; Su
á
ez, J.C.;
Bye, I.J.; Mi anda, D. A Mixed Pixel-and Region-Based App oach o Using Ai bo ne Lase Scanning Da a
o Indi idual T ee C own Delinea ion in Pinus Radia a D. Don Plan a ions. In . J. Remo e Sens.
2013
,34,
7671–7690. [C ossRe ]
82. Vanclay, J.K.; Sko sgaa d, J.P. E alua ing Fo es G ow h Models. Ecol. Model. 1997,98, 1–12. [C ossRe ]
83.
Shapi o, S.S.; Wilk, M.B.; Chen, H.J. A Compa a i e S udy o Va ious Tes s o No mali y. J. Am. S a . Assoc.
1968,63, 1343–1372. [C ossRe ]
84.
Iizuka, K.; Yoneha a, T.; I oh, M.; Kosugi, Y. Es ima ing T ee Heigh and Diame e a B eas Heigh (DBH)
om Digi al Su ace Models and O hopho os Ob ained wi h an Unmanned Ae ial Sys em o a Japanese
Cyp ess (Chamaecypa is Ob usa) Fo es . Remo e Sens. 2017,10, 13. [C ossRe ]
85.
Chisholm, R.A.; Cui, J.; Lum, S.K.; Chen, B.M. UAV LiDAR o Below-Canopy Fo es Su eys. J. Unmanned
Veh. Sys . 2013,1, 61–68. [C ossRe ]
86.
Cosenza, D.N.; Soa es, V.P.; Lei e, H.G.; Gle iani, J.M.; do Ama al, C.H.; G ipp J
ú
nio , J.; da Sil a, A.A.L.;
Soa es, P.; Tom
é
, M. Ai bo ne Lase Scanning Applied o Eucalyp us S and In en o y a Indi idual T ee
Le el. Pesqui. Ag opecuá ia B as. 2018,53, 1373–1382. [C ossRe ]
Fo es s 2019,10, 905 18 o 18
87.
Pe sson, A.; Holmg en, J.; Söde man, U. De ec ing and Measu ing Indi idual T ees Using an Ai bo ne Lase
Scanne . Pho og amm. Eng. Remo e Sens. 2002,68, 925–932.
88.
Popescu, S.C. Es ima ing Biomass o Indi idual Pine T ees Using Ai bo ne Lida . Biomass Bioene gy
2007
,31,
646–655. [C ossRe ]
89.
Zhao, K.; Popescu, S.; Nelson, R. Lida Remo e Sensing o Fo es Biomass: A Scale-In a ian Es ima ion
App oach Using Ai bo ne Lase s. Remo e Sens. En i on. 2009,113, 182–196. [C ossRe ]
90.
Ko pela, I. Indi idual T ee Measu emen s by Means o Digi al Ae ial Pho og amme y; Finnish Socie y o Fo es
Science: Helsinki, Finland, 2004; Volume 3.
91.
S -Onge, B.; Jumele , J.; Cobello, M.; V
é
ga, C. Measu ing Indi idual T ee Heigh Using a Combina ion o
S e eopho og amme y and Lida . Can. J. Fo . Res. 2004,34, 2122–2130. [C ossRe ]
92.
Jensen, J.L.; Ma hews, A.J. Assessmen o Image-Based Poin Cloud P oduc s o Gene a e a Ba e Ea h Su ace
and Es ima e Canopy Heigh s in a Woodland Ecosys em. Remo e Sens. 2016,8, 50. [C ossRe ]
93.
Hopkinson, C.; Chasme , L.; Hall, R.J. The Unce ain y in Coni e Plan a ion G ow h P edic ion om
Mul i-Tempo al Lida Da ase s. Remo e Sens. En i on. 2008,112, 1168–1180. [C ossRe ]
94.
Yu, X.; Hyyppä, J.; Kukko, A.; Mal amo, M.; Kaa inen, H. Change De ec ion Techniques o Canopy Heigh
G ow h Measu emen s Using Ai bo ne Lase Scanne Da a. Pho og amm. Eng. Remo e Sens.
2006
,72, 1339–1348.
[C ossRe ]
95.
Ga ziolis, D.; F ied, J.S.; Monleon, V.S. Challenges o Es ima ing T ee Heigh ia LiDAR in Closed-Canopy
Fo es s: A Pa able om Wes e n O egon. Fo . Sci. 2010,56, 139–155.
96. Milas, A.S.; A end, K.; Maye , C.; Simonson, M.A.; Mackey, S. Di e en Colou s o Shadows: Classi ica ion
o UAV Images. In . J. Remo e Sens. 2017,38, 3084–3100. [C ossRe ]
97.
Lalibe e, A.S.; He ick, J.E.; Rango, A.; Win e s, C. Acquisi ion, O ho ec i ica ion, and Objec -Based
Classi ica ion o Unmanned Ae ial Vehicle (UAV) Image y o Rangeland Moni o ing. Pho og amm. Eng.
Remo e Sens. 2010,76, 661–672. [C ossRe ]
98.
Ke, Y.; Quackenbush, L.J. A Re iew o Me hods o Au oma ic Indi idual T ee-C own De ec ion and
Delinea ion om Passi e Remo e Sensing. In . J. Remo e Sens. 2011,32, 4725–4747. [C ossRe ]
99.
Nuij en, R.J.; Coops, N.C.; Goodbody, T.R.; Pelle ie , G. Examining he Mul i-Seasonal Consis ency o
Indi idual T ee Segmen a ion on Deciduous S ands Using Digi al Ae ial Pho og amme y (DAP) and
Unmanned Ae ial Sys ems (UAS). Remo e Sens. 2019,11, 739. [C ossRe ]
100.
F ey, J.; Ko ach, K.; S emmle , S.; Koch, B. UAV Pho og amme y o Fo es s as a Vulne able P ocess. A
Sensi i i y Analysis o a S uc u e om Mo ion RGB-Image Pipeline. Remo e Sens.
2018
,10, 912. [C ossRe ]
101.
F ase , B.; Congal on, R. Issues in Unmanned Ae ial Sys ems (UAS) Da a Collec ion o Complex Fo es
En i onmen s. Remo e Sens. 2018,10, 908. [C ossRe ]
102.
Ni, W.; Sun, G.; Pang, Y.; Zhang, Z.; Liu, J.; Yang, A.; Wang, Y.; Zhang, D. Mapping Th ee-Dimensional
S uc u es o Fo es Canopy Using UAV S e eo Image y: E alua ing Impac s o Fo wa d O e laps and
Image Resolu ions Wi h LiDAR Da a as Re e ence. IEEE J. Sel. Top. Appl. Ea h Obs. Remo e Sens.
2018
,11,
3578–3589. [C ossRe ]
©
2019 by he au ho s. Licensee MDPI, Basel, Swi ze land. This a icle is an open access
a icle dis ibu ed unde he e ms and condi ions o he C ea i e Commons A ibu ion
(CC BY) license (h p://c ea i ecommons.o g/licenses/by/4.0/).