ORIGINAL PAPER
Recons uc ing o es canopy om he 3D iangula ions
o ai bo ne lase scanning poin da a o he isualiza ion
and planning o o es ed landscapes
Ja i Vauhkonen
1,2,3
&Roope Ruo salainen
1,2
Recei ed: 15 Ma ch 2016 /Accep ed: 4 No embe 2016
#The Au ho (s) 2017. This a icle is published wi h open access a Sp inge link.com
Abs ac
&Key message We p esen a da a-d i en echnique o i-
sualize o es landscapes and simula e hei u u e de el-
opmen acco ding o al e na i e managemen scena ios.
Gen le ha es ing in ensi ies we e p e e ed o main ain-
ing scenic alues in a es o elici ing public’s p e e ences
based on he simula ed landscapes.
&Con ex Visualiza ions o u u e o es landscapes acco ding
o al e na i e managemen scena ios a e use ul o elici ing
s akeholde s’p e e ences on he al e na i es. Howe e , con-
en ional compu e isualiza ions equi e labo ious ee-wise
measu emen s o simula o s o gene a e hese obse a ions.
&Aims We desc ibe and e alua e an al e na i e app oach, in
which he isualiza ion is based on econs uc ing o es can-
opy om spa se densi y, lea -o ai bo ne lase scanning da a.
&Me hods Compu a ional geome y was employed o gene -
a e il a ions, i.e., o de ed se s o simplices belonging o he
h ee-dimensional iangula ions o he poin da a. An app o-
p ia e deg ee o il e ing was de e mined by analyzing he
opological pe sis ence o he il a ions. The opology was
u he u ilized o simula e changes o canopy biomass, esem-
bling ha es s wi h a ying e en ion le els. Rela i e p io i ies
o ec ea ional and scenic alues o he ha es s we e es ima -
ed based on pai wise compa isons and analy ic hie a chy p o-
cess (AHP).
&Resul s The canopy elemen s we e co-loca ed wi h he ee
s ems measu ed in he ield, and he isualiza ions de i ed
om he en i e landscape showed easonably ealis ic, despi e
a low nume ical co espondence wi h plo -le el o es a i-
bu es. The po en ial and limi a ions o imp o e he p oposed
pa ame e iza ion a e discussed.
&Conclusion Al hough he c i e ia o e alua e he landscape
isualiza ion and simula ion models we e no conclusi e, he
esul s sugges ha o es scenes may be easibly econs uc -
ed based on da a al eady co e ing b oad a eas and eadily
a ailable o p ac ical applica ions.
Keywo ds Spa ial mul ic i e ia decision analysis .Public
pa icipa iongeog aphicin o ma ionsys em(PPGIS) .Remo e
sensing .Ligh de ec ion and anging (LiDAR) .Pe sis en
homology .Alpha shape
1 In oduc ion
En i onmen al and o es y decision making equi es he
iden i ica ion and compa isons o di e en managemen
Handling Edi o : Jean-Michel Leban
Con ibu ion o he co-au ho s Ja i Vauhkonen designed he s udy,
implemen ed all s ages ela ed o econs uc ing he o es scenes, and
d a ed he manusc ip .
Roope Ruo salainen ca ied ou he su ey and analyzed i s esul s wi h
pai wise compa isons and he analy ic hie a chy p ocess. He also ead
and app o ed he inal manusc ip .
Elec onic supplemen a y ma e ial The online e sion o his a icle
(doi:10.1007/s13595-016-0598-6) con ains supplemen a y ma e ial,
which is a ailable o au ho ized use s.
*Ja i Vauhkonen
ja i. auhkonen@ue . i
Roope Ruo salainen
oope. u@ou look.com
1
School o Fo es Sciences, Uni e si y o Eas e n Finland,
Yliopis oka u 7, P.O. Box 111, FI-80101 Joensuu, Finland
2
Depa men o Fo es Sciences, Uni e si y o Helsinki,
La oka anonkaa i 7, P.O. Box 27, FI-00014 Helsinki, Finland
3
P esen add ess: Na u al Resou ces Ins i u e Finland (Luke),
Economics and Socie y, Yliopis oka u 6, P.O. Box 68,
FI-80101 Joensuu, Finland
Annals o Fo es Science (2017) 74:9
DOI 10.1007/s13595-016-0598-6
al e na i es based on mul iple objec i es and s akeholde s
(Kangas e al. 2008). In eg a ing in o ma ion on he impac s
o managemen decisions wi h p e e ences o he s akeholde s
esul s in a amewo k called mul iple c i e ia decision analy-
sis (MCDA). When inco po a ed wi h a geog aphic in o ma-
ion sys em (GIS), he esul ing applica ions a e nowadays
called “spa ial MCDA”o “public pa icipa ion geog aphic
in o ma ion sys em”(PPGIS; Siebe 2006), depending on
he in ol emen o he s akeholde . Ea ly o es y applica ions
o spa ial MCDA and PPGIS a e p o ided by S o e and
Kangas (2001) and Kangas and S o e (2003), espec i ely.
P ojec ing u u e o es landscapes acco ding o al e na i e
managemen scena ios is a pa icula ly use ul MCDA compo-
nen o pa icipa o y o es planning. Fo example, impac s o
g ow h, managemen p ac ices, o es (logging) ope a ions,
and loca ions o logging a eas may a ec he s akeholde s’
p e e ences on he managemen al e na i es. I has long been
ecognized ha e en echnically simple compu e isualiza-
ions imp o e unde s anding o o es s and dynamics and
e ec s o managemen decisions (Pukkala and Kellomäki
1988; Bu kha 1992) and acili a e elici ing he ela ed p e -
e ences (Tah anainen e al. 2001; Ka jalainen and Ty äinen
2002). Visualiza ion echniques and applica ions a e e iewed
in he o es y con ex by Mendoza e al. (2006)andFalcão
(2008) and echniques o ex ac ing he decision make s’
p e e ences om he isualiza ions by Kangas e al. (2008).
Recen p ac ical examples o inco po a ing he in o ma ion
ob ained in decision suppo and GIS amewo ks a e p esen -
ed by Wa en-K e zschma and on Haa en (2014) and Lämås
e al. (2015).
The isualiza ions o o es ed landscapes equi e balancing
be ween he image ealism, he desi ed esolu ion, and he
measu emen s a ailable (c . Be gen e al. 1998;Uusi aloand
O land 2001). E en i ex emely pho o ealis ic o es scenes
may be ob ained by combining bo anical models o ee a chi-
ec u e wi h line g aphics o eal ee ex u es p esen ed in
i ual eali y en i onmen s (e.g., Aono and Kunii 1984;
Honjo and Lim 2001;Fujisakie al.2008), he eeand
b anch-le el measu emen s equi ed by hese echniques a e
ime-consuming and expensi e. Unless ee-wise in en o y
da a a e a ailable, he landscapes need o be popula ed wi h
ees based on less de ailed in en o y da a. A ypical app oach
is o simula e he loca ions and appea ances o he indi idual
ees based on mean diame e , heigh , and s uc u e in o ma-
ion (e.g., Ka jalainen and Ty äinen 2002). Pa icula ly, he
use o mean-based a ibu es and hus he equi emen o sim-
ula e he ees will esul in gene aliza ions and es ic ions o
eali y (Uusi alo and O land 2001;Wange al.2006).
Techniques based on emo e sensing p o ide signi ican
ad ances o e con en ional o es measu emen s.
Pa icula ly due o i s capabili y o p esen o es s uc u e as
a poin cloud mapped in 3D, ai bo ne lase scanning (ALS;
also e e ed o in some ins ances as “ai bo ne scanning
LiDAR”) has become an inc easingly popula echnique o
a ious o es y applica ions (Mal amo e al. 2014). Howe e ,
he use o ALS o isualizing o es y decision making has
been a e, o da e, e en hough p oposed al eady by Hill and
Vei ch (2002), McGaughey and Ca son (2003), and Ahlbe g
e al. (2004). Despi e he a ailabili y o ea u e- ich so wa e
packages o isualizing ALS poin clouds, ende ing ealis ic
ee geome y om he poin da a poses a p oblem (Simons
e al. 2014). Typical app oaches o pa ame e ize a o es scene
in ALS-based analyses include popula ing a lis o ee loca-
ions and dimensions using simple a i icial u bid media such
as cones, ellipses, oxels (Schneide e al. 2014), o o he
ypes o 3D p imi i es (Koch e al. 2014). A common p oblem
o hese app oaches is, howe e , he equi emen o a e y
high da a densi y (e.g., ens o pulses pe m
2
), whe eas p ac-
ically a ailable da a se s a e ypically collec ed o o he pu -
poses such as g ound ele a ion modeling wi h densi ies as low
as <1 pulses pe m
2
(e.g., No d-La sen and Schumache 2012;
Villikka e al. 2012).
T iangula ing poin da a and subsequen il e ing o he
iangula ions (e.g., Del inado and Edelsb unne 1995) has
been p oposed as an al e na i e means o ep esen he 3D
canopy su ace (Vauhkonen e al. 2014). As opposed o ap-
p oaches ha use ixed image elemen s and he e o e equi e
speci ying an a i icial pixel o oxel esolu ion, he iangu-
la ions a e en i ely based on he p ope ies o he poin da a.
Applying il a ions, one can adjus he le el o de ail in he
iangula ed poin cloud and hus accoun o canopy gaps and
de ailed p ope ies exis ing in he da a (see Sec . 2.2. o
de ails). The op imal deg ee o il e ing is ound o be o es
s uc u e speci ic (Vauhkonen e al. 2016), whe eas less de-
ailed in o ma ion could su ice o gene a ing ealis ic isu-
aliza ions o essella ed landscapes (Vauhkonen 2015). We
addi ionally p opose ha he opology based on he il a ions
could be u ilized as a mechanism o simula e changes in he
o es ed landscapes isualized (Sec . 2.3.), which is
complica ed based on simula ed o es s and dynamics (c .
D ey us 2012).
E en hough he low esolu ion ob ainable by he p ac ical-
ly a ailable spa se da a mos likely es ic s he de ail o he
isualiza ions, he decision make may be expec ed o bene i
om ep esen ing eal ee geome y (c . Uusi alo and O land
2001) and isual desc ip ions (c . Tah anainen e al. 2001)
ins ead o simula ed and e bally o nume ically desc ibed
o es scenes. Ano he p ac ical bene i o ou app oach is ha
equi ed da a can be easibly ex ac ed om any loca ion co -
e ed by an ALS campaign such as hose designed o
collec ing spa se densi y da a o g ound ele a ion modeling.
Al hough he sui abili y o such da a o o es in en o ies was
e i ied (No d-La sen and Schumache 2012; Villikka e al.
2012), he bene i s o using he spa ially explici 3D scenes
ob ainable ha e no been ully employed despi e he po en ial
iden i ied ea lie (Hill and Vei ch 2002; McGaughey and
9 Page 2 o 13 Annals o Fo es Science (2017) 74:9
Ca son 2003; Ahlbe g e al. 2004). The e o e, a he han de-
eloping ano he i ual eali y pla o m aiming a imp o ed
pho o ealism wi h high cos s o collec ing he equi ed da a,
we p esen a wi e ame model on how o ex ac in o ma ion
om he isualized o es scenes o a spa ial MCDA o open
up discussion on he po en ial and limi a ions o he b oadly
a ailable spa se ALS poin da a o suppo o es y decision
making.
The pu pose o his s udy is o desc ibe a echnique o
isualizing o es scenes om spa se densi y ALS da a and
e alua e hei po en ial o o es y decision suppo , making
a di e ence be ween (i) close- ange and (ii) landscape le els.
The close- ange isualiza ions co espond o a wi hin- o es
iew and up o a plo scale, whe eas he landscape le el shows
o es a eas in scales o ens o hund eds o hec a es co e-
sponding o ypical p ope ies managed by a single o es
owne . The o es scenes we e ob ained applying a iangula -
ing and il e ing echnique o ALS da a acqui ed o iginally o
g ound ele a ion modeling. Rega ding (i), we assessed he
co espondence o he o es canopy elemen s modeled as e -
ahed a and ob ained by il e ing he iangula ions wi h o es
a ibu es obse ed in he ield. Rega ding (ii), we in oduce an
in ui i e concep based on using he il a ions o simula e
changes o canopy biomass. The ini ial and simula ed land-
scapes a e isualized and demons a ed in elici ing public’s
p e e ences on scenic alues esul ing om biomass ha es s
applied wi h a ying in ensi ies.
2 Ma e ial and me hods
2.1 S udy a ea and da a
The s udy a ea is a ypical, managed bo eal o es in eas e n
Finland. An a ea o 1000 × 1200 m loca ed app oxima ely on
62° 31.5′N, 30° 11′E was selec ed o he isualiza ions due
o he p esence o a o es and lake mosaic, which likely p o-
duced high ec ea ional and scenic alues o e he a ea. The
da a o he isualiza ions included he loca ions o he lakes
and ALS da a ex ac ed om a opog aphic da abase. In addi-
ion, he ALS da a we e ex ac ed o a numbe o ield sample
plo s exis ing in he a ea, and he ield measu emen s co-
loca ed wi h he ALS da a we e used o assess he co espon-
dence o he modeled canopy wi h espec o he ield da a.
The plo s we e loca ed up o 2.5 km apa he isualized a ea
and we e ea lie used o s udy he spa ial pa e n and diame e
dis ibu ion o he ees om bo h he ecological and econom-
ic pe spec i es. Sco s pine (Pinus syl es is L.) and No way
sp uce (Picea abies [L.] Ka s .), wi h a mino p opo ion o
deciduous species, we e p esen in he plo s, whe eas he a ea
used o he isualiza ions was assumed o be nea ly pu ely
pine-domina ed.
The ALS da a used in he s udy we e acqui ed by he
Na ional Land Su ey o Finland as a pa o hei da a acqui-
si ion campaign o c ea ing a na ionwide g ound ele a ion
model o Finland. The da a we e downloaded om a ile se -
ice (h ps:// iedos opal elu.maanmi auslai os.
i/ p/ka a?lang=en), om which hey a e a ailable o ee
and wi h ex ensi e pe missions o use. The da a we e
acqui ed on Ap il 30, 2012, wi h Leica ALS60 scanne
ope a ed in a mul ipulse mode. The lying al i ude was 2350
m, yielding a nominal pulse densi y o 0.8 m
−2
. The da a
p o ide had de ec ed and classi ied he g ound le el, on
which he no maliza ion o he ege a ion heigh alues was
based. As he da a a e mean speci ically o g ound ele a ion
modeling, we assumed he accu acy o his classi ica ion o be
app op ia e o ou pu poses. The analyses we e ocused only
on he i s echoes (i.e., “only”and “ i s o many”o up o ou
echoes eco ded pe pulse), aiming o ob ain he main
in o ma ion om he da a (Vauhkonen e al. 2014), while
e aining mos gene aliza ion abili ies o e senso s ha eco d
a di e en numbe o echo ca ego ies (e.g., Næsse 2014).
The ield plo s we e measu ed in May–June, 2010. All
ees wi h ei he diame e a b eas heigh (DBH) ≥4cmo
heigh ≥4 m we e mapped o loca ions and measu ed o
species, DBH, and heigh . The measu emen s a e desc ibed
in mo e de ail by Packalén e al. (2013). Fo simplici y, he
plo size used in his s udy was s anda dized o 400 m
2
, a he
han a ying om 400 o 900 m
2
as in he ea lie publica ions.
The s anda diza ion was achie ed by c ea ing ec angula
windows o 20 × 20 m wi h he cen e and o ien a ion co e-
sponding o he o iginal plo . The ees loca ed wi hin his
window based on he measu ed ee coo dina es we e ex ac -
ed o he analyses.
Plo -le el basal a ea was compu ed based on summing
om he diame e measu emen s. Dominan heigh was de e -
mined as he mean heigh o 100 hickes ees pe hec a e
( ou ees pe plo ). Indi idual s em olumes we e es ima ed
by models o Laasasenaho (1982), employing he DBH,
heigh , and ee species as p edic o s. The models o bi ch
we e used o all deciduous ees. The spa ial pa e n o he
ees was assessed by means o he Cla k-E ans index (CEI;
Cla k and E ans 1954) o he agg ega ion o a poin pa e n.
The CEI alues we e compu ed based on he ee coo dina es
using he spa s a package o he R s a is ical compu ing en-
i onmen (Baddeley and Tu ne 2005), applying an edge co -
ec ion p oposed by Donnelly (1978). Gene al desc ip i e s a-
is ics o he ield da a a e shown in Table 1.
2.2 Modeling he ini ial o es canopy using iangula ions
In ou app oach, “ isualiza ion”essen ially e e s o isualizing
o es canopy geome y based on he 3D measu emen s collec -
ed by an ALS senso (e.g., Sun and Ranson 2000). The geom-
e y is composed o n-dimensional basic elemen s (poin s,
Annals o Fo es Science (2017) 74:9 Page 3 o 13 9
edges, ace s, e ahed a), which a e e e ed o using gene ic
e m “simplex”unless an elemen o a speci ic dimension is
pa icula ly deno ed. The main canopy elemen s a e e ahed a
ob ained by subdi iding he unde lying space o he 3D mea-
su emen s, which is called iangula ing. The en i e iangula-
ion, e e ed o below also as “complex,” hus ep esen s bo h
he canopy and emp y space, which need o be di ided o sep-
a a e subcomplexes cons uc ing a il e . The iangula ions
we e il e ed using a well-known compu a ional geome ic o
opological concep called 3D alpha shapes (Edelsb unne and
Mücke 1994), while an app op ia e deg ee o il e ing was de-
e mined by analyzing pe sis ence o he obse ed opology,
i.e., opological pe sis ence o pe sis en homology
(Edelsb unne e al. 2002). All compu a ions we e based on
Delaunay iangula ions and implemen ed wi h C++ and
open-sou ce lib a ies (Mo ozo 2012;Dae al.2013;Pion
and Teillaud 2013). The applied wo k low is desc ibed below,
bu he ma hema ical o malism is in en ionally le o he o ig-
inal publica ions. Illus a ions o all he concep s applied o
ALS da a a e p esen ed by Vauhkonen (2015), o example.
Following ea lie analyses wi h ALS poin da a
(Vauhkonen e al. 2014,2016), he landscape was essella ed
o a egula g id wi h a cell size co esponding o he plo size
(20 × 20 m). The ALS da a o each cell o plo was iangu-
la ed and il e ed sepa a ely. The p oblem ela ed o using he
alpha shapes is he de e mina ion o a p ope alue o alpha
(α), which is a h eshold o he squa ed adius o he
ci cumsc ibing sphe e o each simplex and he e o e a c i e i-
on de e mining which simplices o he ull iangula ion be-
long o i s cu en subcomplex, i.e., α-complex o α-shape.
He e, he alue o αwas selec ed based on il a ion ob ained
by o de ing he α-complexes acco ding o he alues o α(see
also Del inado and Edelsb unne 1995).
Du ing he il a ion, i.e., by inc easing he alue o α,
simplices o a dimension djoin oge he wi h o he d-dimen-
sional simplices o o m s uc u es wi h a highe dimensional-
i y. Pa icula ly, when he alue o αallows wo poin s (d=0)
o connec , an edge o d= 1 be ween hese wo poin s is
o med. The edges u he join o o m ace s (d= 2) and
e ahed a (d= 3), he la e being he highes dimension con-
side ed he e. Each simplex can hus be desc ibed by i s pe -
sis ence in he s uc u e o med. This index o pe sis ence was
ob ained as he absolu e di e ence be ween he index alues
o αcausing he bi h and dea h o simplices o he gi en
dimension.
Following he logic o he p e ious pa ag aph, a bi h and
dea h diag am was compu ed o indices wi h d= 1 and d=2.
Only hese dimensions need o be conside ed, since hose wi h
d=0o d= 3 do no bo n o die, espec i ely, in his p ocess.
In he diag am, he mos in e es ing obse a ions a e hose o -
diagonal: These simplices a e in e p e o cause he mos un-
damen al changes o he iangula ed s uc u e, ypically a he
lowes alues o α, while he diagonali y o he obse a ions
indica es s abili y (pe sis ence) o hose ea u es. The maxi-
mum o -diagonal αa he dea h o bo h d= 1 and d= 2 was
hypo hesized o e lec he p ima y pe sis ing s uc u es o he
ob ained iangula ion and was hus used as he α alue de e -
mining he deg ee o il e ing and hus he ini ial s a e o he
o es canopy.
The a ionali y o he ob ained canopy was e alua ed using
he ield da a. The obse ed ee (s em) posi ions and dimen-
sions we e isualized wi h he canopy o assess he co-loca ion
o hese elemen s. Desc ip i e cha ac e is ics we e ex ac ed
om he iangula ion-based canopy models and ela ed wi h
he o es a ibu es measu ed in he ield (Table 1). The alue o
he pa ame e α esul ing om he pe sis en homology analysis
and he o al e ahed al olume, pe cen o he plo a ea co -
e ed by he e ahed al, and op heigh o he unde lying space
o he speci ied α-complex we e ex ac ed, assuming hese o
be ela ed o he o al ee a ibu es o a ield plo . Addi ionally,
he numbe o connec ed e ahed a wi h a leas one join edge
was compu ed, assuming his o be ela ed o he numbe o ee
pa ches wi hin he plo . The co espondence was assessed by
means o he coe icien o de e mina ion (R
2
).
2.3 Simula ing changes o he canopy and assessing hei
impac s o he scene y
2.3.1 Visualizing he landscape and ha es s o canopy
biomass
The ull landscapes we e isualized using iangula ion ep e-
sen a ion (T iRep) and iangula su ace plo ( isu ) unc-
ions implemen ed in Ma lab, e sion R2012a (Ma hWo ks,
Table 1 Cen al cha ac e is ics
o he 71 ield plo s A ibu e Mean S d Min Max
S em olume, m
3
/ha 199.6 73.2 69.6 482.2
Basal a ea, m
2
/ha 24.9 6.6 12.2 43.7
Numbe o s ems, 1/ha 1275.3 590.8 400.0 2875.0
Basal-a ea weigh ed mean diame e , cm 21.6 5.2 12.8 32.3
Dominan heigh , m 20.4 4.1 13.5 32.5
CEI, index alue 1.09 0.16 0.69 1.41
S d s anda d de ia ion, Min minimum, Max maximum
9 Page 4 o 13 Annals o Fo es Science (2017) 74:9
Inc., Na ick, MA, USA). The landscapes we e composed o
he adjacen 20 × 20 m cells p ocessed acco ding o Sec . 2.2.
Due o depic ing solely canopy geome y, ce ain simpli ica-
ions o he isualizing o he elemen s we e adop ed. The
numbe o he spa ial pa e n o he ee s ems could no be
p ecisely modeled, which is a ypical esul o ALS-based
analyses (e.g., Packalén e al. 2013). Howe e , he
landscape-le el isualiza ions we e ound o be mo e ealis ic
wi h a leas some shadowing caused by he s ems. A numbe
o he highes e ahed a pe cell we e hus accompanied wi h
simula ed ee s ems. The numbe o s ems was de e mined as
ceil(n
p
× ×V
p
/H
p
), whe e n
p
was he numbe o e ahed a,
V
p
was he o al olume o hese e ahed a, H
p
was he max-
imum ALS heigh alue o he cell p,and was a uni a ea
ac o , in his case 1/400. The diame e s o he s ems (in me-
e s) we e simula ed as H
/50, whe e H
is he maximum
heigh o he conside ed s em in me e s. The applied ules
esul ed in oo ew and oo la ge ees, in gene al, bu p o ided
he landscapes wi h such a ealism ha only he la ges ee
s ems we e isible when looking om a dis ance. The p o-
duced landscapes we e colo ed as pu e pine o es s, i.e., spe-
cies ecogni ion was no a emp ed due o he low p opo ion
o o he species. The g ound le el was depic ed as pe ec ly
la , since he luc ua ions in he e ain ele a ion we e known
o be mino in his a ea.
In addi ion o isualizing he ini ial canopy, educ ions o
he canopy olume we e simula ed o po ay ha es s wi h
a ying in ensi ies. In he app oach, simplices belonging o
he canopy we e e-assigned as emp y space by adjus ing he
alue o αpa ame e (see Sec . 2.2.) downwa ds. As a esul ,
he ini ial α-complexes we e sub ac ed s a ing om he la g-
es e ahed a, assuming his o co espond wi h ha es s wi h
a ying e en ion le els based on emo ing he la ges ees
om he o es . The ha es s we e ocused on one compa -
men wi h a dense ini ial canopy. Al oge he , i e ha es ing
in ensi ies we e de ined by mul iplying he ini ial αby a ac o
o 0.95, 0.75, 0.68, 0.6, and 0.5, which co esponded o an
a e age educ ion o 17, 71, 83, 93, and 99%, espec i ely, o
he ini ial canopy olume (11,673 m
3
/ha on he a ea o he ull
compa men , on a e age). The isualiza ions o he ull land-
scape ob ained as a esul o he ha es ing we e labeled as
images A–E acco ding o he inc easing ha es ing in ensi y
and a e p esen ed as Elec onic Supplemen a y Ma e ial
(Online Resou ce 1).
2.3.2 Pai wise compa isons o he simula ed landscapes
To ob ain p e e ence in o ma ion on how he ha es ing a -
ec ed he landscape, a pai wise compa ison o he isualized
landscapes was implemen ed using he Su eyMonkey® on-
line pla o m (h p://www.su eymonkey.com/). In he su ey,
wo simula ed landscapes we e shown a he ime, asking a
ques ion o which o he wo images was be e wi h espec o
he ec ea ional use and he scenic alues o he landscape and
how much be e i was. The esponden was asked o exp ess
his/he opinion on whe he ei he image was (i) equally good,
(ii) sligh ly be e , (iii) clea ly be e , (i ) conside ably/s ongly
be e , o ( ) absolu ely be e compa ed o he o he . In he
ollowing ex , his decision is called he p io i y a io o he
esponden be ween he wo landscapes.
A link o he su ey was submi ed o he mailing lis o he
o es y s uden s o he Uni e si y o Eas e n Finland and
accompanied wi h b ie ins uc ions ( ansla ed om Finnish):
“The cen al compa men o [ his] landscape is ma ked
o be ha es ed. A walking ail and a ‘laa u’(a empo-
a y shel e o ield lunch o camping) a e si ua ed in
he a ea o he compa men . The ha es ing in ensi y is
no ye de e mined, bu he aim is o main ain he ec e-
a ional alue o he o es as high as possible.
In he su ey, you a e asked o compa e wo al e na i e
landscapes and selec which ha es ing in ensi y is be -
e sui ed wi h espec o he ec ea ional use and scenic
alues. The compa men o be ha es ed is delinea ed
wi h a ed bounda y in he image below. The o he pa s
o he landscape will no be a ec ed. S a he su ey by
clicking ‘Nex image pai ’bu on.”
The su ey was held open be ween 21 Janua y and 12
Feb ua y, 2015. The su ey yielded al oge he 57 esponses,
o which 7 we e incomple e (no esponse o one o mo e
compa isons). Al oge he , 50 esponses we e hus analyzed
in he u he s udy.
2.3.3 P e e ence elici a ion
On he basis o he pai wise compa isons (Sec . 2.3.2.), ela-
i e p io i ies o he scenic alues wi h espec o he ha es
in ensi ies we e es ima ed using he eigen alue echnique o
analy ic hie a chy p ocess (AHP; Saa y 1977,1980). This
echnique was used o con e e bal e alua ions o a io-
scale nume ical alues in a o es y decision suppo applica-
ion bea ing a close esemblance o ou s udy (Kangas e al.
1993).
In he app oach, he p io i y a ios i– (Sec . 2.3.2.) a e
ansla ed in o nume ical alues o 1:1, 3:1, 5:1, 7:1, and
9:1, espec i ely. The alues a e a anged in o a ecip ocal
ma ix A, and using he ma ix as inpu , he nume ic weigh s
o p io i ies o he pai wise compa isons a e compu ed as he
igh eigen ec o o he la ges eigen alue (λ
max
)o A(Saa y
1977).
To assess he cohe ence o he pai wise compa isons, a
consis ency index (CI; Saa y 1980), which es ima es he le el
o consis ency wi h espec o he en i e compa ison p ocess,
is de i ed by ela ing he alue o λ
max
wi h he numbe o
compa isons made. The obse ed CI is u he compa ed wi h
Annals o Fo es Science (2017) 74:9 Page 5 o 13 9

an a e age CI ob ained by andomly gene a ed compa isons
in a ma ix wi h a size co esponding o A(Saa y 1980). The
esul is a consis ency a io (CR), which measu es he cohe -
ence on a scale o 0.1. Acco ding o a gene al ule o humb,
compa isons wi h CR > 0.1 a e deemed inconsis en .
3Resul s
Excep o op heigh , he iangula ed and il e ed canopy
models we e in low co espondence wi h any o es a ibu e
when e alua ed wi h espec o he ield measu emen s.
Ne e heless, he R
2
alues o he o al olume and op heigh
o he canopy model in pa icula we e s a is ically signi ican
(Table 2), and he o es s uc u e a ec ed he models ob ained
in he way i was expec ed. The spa ial pa e n o he ees,
quan i ied in e ms o CEI, signi ican ly a ec ed he alues o
he αpa ame e (Table 2), indica ing ha a di e en α alue
was equi ed o di e en spa ial pa e ns. The o al e ahed al
olume o he econs uc ed canopy was signi ican ly ela ed
o all o he size and s uc u e a ibu es expec o he s em
numbe (Table 2), and he olume inc eased acco ding o in-
c easing alues o hese a ibu es. The numbe o connec ed
e ahed a and he co e age o he e ahed a ela i e o he
plo a ea howe e had a low deg ee o de e mina ion wi h any
o he ield a ibu es. The op heigh o he canopy model had
he bes o e all co espondence wi h he ield a ibu es and a
s ong linea ela ionship especially wi h he dominan ee
heigh . No ing ha no allome ic knowledge was used in he
cons uc ion o he canopy, he models based on he geome y
and opology o he poin da a s ill e lec he a ia ion in he
o es p ope ies and agg ega ed s em a ibu es.
Based on a isual assessmen , he canopy elemen s o he
coni e ous dominan canopy in pa icula we e ai ly accu a e-
ly co-loca ed wi h he ee s ems measu ed in he ield (Fig. 1).
The posi ions o he canopy elemen s we e also somewha in
line wi h he spa ial pa e ns o he ield measu ed ees, pa -
icula ly hose showing clus e ed pa e ns. Ne e heless,
Fig. 1also shows he limi a ions o he p oposed echnique
owa d close- ange isualiza ions. The le el o de ail and he
ep esen a ion o he deciduous and unde s o ey ees a e in
pa icula limi ed, he easons o which a e u he discussed in
Sec . 4.
Despi e he low de ail in close- ange isualiza ions (Fig. 1),
he isualiza ions de i ed o he en i e landscape showed a
easonable le el o de ail and ealism (Fig. 2). Fo ins ance,
oads, elec ic lines, o o he wise open a eas could be poin ed
ou om he image. Fu he , despi e he po en ial o edge e -
ec s, i.e., inco ec ly cu ing he ALS da a due o an edge o a
cell (Fig. 1), he bo de s o he 20 × 20 m cells did no show in
he isualiza ion o he ull landscape (Fig. 2).
The canopy biomass educ ions simula ed o he landscape
we e ound o easonably esemble e ec s o eal-wo ld ha -
es ing ope a ions (Fig. 2). Howe e , he simula ed ha es s
clea ly had a mo e conside able e ec on he ho izon al co -
e age han on he heigh o he modeled canopy (Fig. 3).
Especially, he op heigh changed e y mode a ely based on
ha es ing in ensi ies A, B, and C acco ding o Fig. 3, which
shows he e ec s o he simula ed educ ion o pa ame e α o
he e ical and ho izon al s uc u e o he landscape.
Acco ding o he su ey, he landscape ea ed wi h hea y
ha es ing in ensi ies (images E and D) was p e e ed by none
o he esponde s, while lowe in ensi ies (images C, B, and A)
we e selec ed as mos p e e able by 10 (20%), 16 (32%), and
24 (48%) esponde s, espec i ely. Due o hese p e e ences,
he (accumula ion o ) ela i e p io i y be ween he landscapes
was signi ican ly indi e en o di e en ypes o esponde s
(Fig. 4) and he gene al end was di icul o model wi hou
addi ional in o ma ion.
Acco ding o he AHP, he consis encies o he esponses
a ied conside ably. Only 10 (20%) o he pai wise compa i-
sons had a CR < 0.1, while 30 (60%) and 10 (20%) esul ed in
CRs be ween 0.1–0.3 and >0.3, espec i ely. The maximum
and mean CR alues we e 0.99 and 0.22. When examined
agains he ela i e p io i ies (Fig. 4), he inconsis encies we e
nei he clus e ed in any p e e ence g oup no e lec ed in any
o he pa icula way, howe e .
4Discussion
Acco ding o he esul s p esen ed abo e, he canopy models
and isualiza ions based on he spa se densi y ALS da a a e
Table 2 The deg ee o
de e mina ion (R
2
)be ween
selec ed ield a ibu es and alpha
alue (α), o al olume (V),
numbe o connec ed e ahed a
(N), co e age (C), and op heigh
(H) o he canopy models
Field obse a ion αVNCH
S em olume, m
3
/ha 0.01 0.26* 0.07* 0.09* 0.54*
Basal a ea, m
2
/ha 0.00 0.11* 0.06* 0.07* 0.16*
Numbe o s ems, 1/ha 0.02 0.02 0.00 0.03 0.19
Basal-a ea weigh ed mean diame e , cm 0.05 0.27* 0.04 0.05 0.78*
Dominan heigh , m 0.02 0.32* 0.06* 0.07* 0.94*
CEI 0.06* 0.01 0.05 0.06* 0.00
As e isk ma ks signi icance o he co ela ion a he 95% con idence le el, while R
2
> 0.5 a e p in ed in i alics
9 Page 6 o 13 Annals o Fo es Science (2017) 74:9
mainly sui ed o landscape-le el analyses. The low deg ee o
de e mina ion be ween he close- ange models and o es a -
ibu es indica es ha ei he a ield sample o op imize he
canopy model (Vauhkonen e al. 2014,2016) o a comple ely
di e en app oach based on highe densi y da a ob ained by
ALS o e es ial lase scanning is equi ed o de ailed
modeling. On he o he hand, he p oposed app oach appea s
in e es ing due o he possibili y o gene a e he isualiza ions
om e y spa se ALS da a ha o en eadily exis and he
inhe en p ope y o he modeling echnique o simula e
changes o he landscape by adjus ing he il a ion pa ame e
α. The ob ained isualiza ions a e a ec ed by bo h inpu da a
and p ope selec ion o he alue o α, as discussed below.
As opposed o con en ional o es isualiza ion echniques
ha equi e labo ious ee-by- ee measu emen s o he use o
simula o s o gene a e hese obse a ions, he p esen s udy
equi es a wall- o-wall co e age o ALS da a o e he land-
scape o be isualized. While he acquisi ions o ALS da a o
g ound ele a ion modeling nowadays allow applica ions o
hese da a o as a eas, p ope ies such as he sca ci y o
he inpu poin da a may es ic some o es y analyses (e.g.,
Vauhkonen e al. 2016). As seen om Fig. 1o his s udy, o
ins ance, he ue canopy heigh was unde es ima ed due o he
pene a ion o he pulses h ough he op canopy (Ga eau and
Hill 2003). Ne e heless, he op heigh ex ac ed om he
iangula ion-based canopy models had a s ong ela ionship
wi h he dominan heigh (Table 2), which was compa able o
he heigh me ics ex ac ed om he ini ial poin da a (de-
ailed esul s no shown). No in o ma ion ela ed o he spa ial
pa e n o he ees o ee numbe could be de i ed, which is
in line wi h ea lie s udies based on ALS da a (e.g., Packalén
e al. 2013). The pe cen co e age o he e ahed a could be
compa ed o he canopy co e me ic compu ed om he poin
da a as he p opo ion o he echoes abo e a ce ain heigh
h eshold o all echoes (Ko honen e al. 2011). This me ic
had a co ela ion o 0.8 wi h he e ahed al co e age, bu
simila ly low deg ees o de e mina ion wi h he ield-
measu ed o es a ibu es.
Since he da a we e cap u ed unde a lea -o pe iod o he
ege a ion phenology, less i s - e u n da a we e ob ained
om he deciduous han coni e ous ees, e en i hose we e
loca ed in he dominan canopy (Fig. 1; see also Villikka e al.
2012). Also, only limi ed in o ma ion was ob ained om he
unde s o ey. Al hough some esea che s ha e de ec ed mo e
ees unde lea -o han lea -on condi ions, when ocusing on
indi idual ee de ec ion in dense ALS da a (Duncanson e al.
Fig. 1 Examples o isualized plo s wi h clus e ed (CEI = 0.7, abo e),
andom (CEI = 1.0, middle), and egula (CEI = 1.4, below) pa e ns o
ee s ems. The loca ions and dimensions o he cylinde s co espond wi h
he ee s ems measu ed in he ield. The b own and g ay cylinde s
ep esen sco s pine and deciduous ees, espec i ely. The in e al
be ween ick ma ks co esponds o 5 m in he e ical axis and 2 m in
he plane
Annals o Fo es Science (2017) 74:9 Page 7 o 13 9
2014), also he esul s o Ko pela e al. (2012) indica e ha
di ec measu es o he unde s o ey a e di icul o ob ain due o
ansmission losses occu ing in he uppe canopy (see also
Mal amo e al. 2005). Fo he easons s a ed abo e, he spa ial
pa e ns o he deciduous and unde s o ey ees may be inco -
ec in some pa s o he landscape, bu his was no assumed
o be a majo p oblem in he pine-domina ed es a ea consid-
e ed in ou s udy. Whe he a emp ed in mo e complex cano-
pies in u he s udies, howe e , conside a ions on, e.g., he
species and pulse pene a ion p ope ies a e equi ed.
Whe he he limi a ions desc ibed in he p e ious pa a-
g aph a e accep ed, he p esen ed app oach is p oposed easi-
ble o isualizing landscapes such as ha p esen ed in Fig. 2.
The me hodology p esen ed he e allows a desc ip ion ha is
based on eal canopy elemen s, as ex ac ed om he iangu-
la ions based on ALS da a, bu no explici p esc ip ion o lea
o shoo loca ions and o ien a ions, o example. The e o e,
ega ding he deg ee o de ail in desc ibing he dis ibu ion o
canopy elemen s, he p oposed app oach is a comp omise be-
ween a i icial ee/c own-le el u bid media pa ame e ized
by a lis o ee loca ions and dimensions (e.g., Schneide
e al. 2014) and mo e ealis ic 3D models o ege a ion
elemen s po en ially ob ainable by e es ial lase scanning.
An applica ion o o es managemen planning is desc ibed in
he p esen s udy, bu he modeling app oach could also be
sui ed o o he applica ions equi ing geome ically explici
pa ame e iza ions o o es landscapes. An example is ligh
in e ac ion and adia i e ans e modeling (e.g., Schneide
e al. 2014), in which inco ec assump ions on c own shapes
and inapp op ia e oxel g id esolu ions a e epo ed o esul
in signi ican e o s in e ie ed c own pa ame e s (Calde s
e al. 2013; Widlowski e al. 2014).
Visualiza ion is a s aigh o wa d applica ion o he iangu-
la ions and il a ions o ALS da a collec ed o e a o es ed
a ea, which a e ea lie es ed in a ious o es in en o y appli-
ca ions (Vauhkonen e al. 2014,2016; Vauhkonen 2015). As
opposed o he ypical “su ace modeling”app oaches using 2D
o 2.5D as e images, he use o iangula ions and il a ions
p o ides means o cons uc ing genuine 3D pa ame e iza ions
o o es ed landscapes ins ead o jus ep esen ing coun s o
heigh alues wi hin pixel o oxel cells. Besides iangula ions,
geome ic modeling echniques ha could be conside ed o
sol ing he p esen ed p oblem a e an i e a i e su ace w apping
echnique based on delinea ed ee c owns (Ka o e al. 2009),
Fig. 2 The ull landscape
isualized wi h he app oach
p oposed. The sub- igu es on he
igh gi e examples on ha es s
simula ed o he compa men in
he cen e o he landscape. All
i e simula ed landscapes a e
p esen ed as Elec onic
Supplemen a y Ma e ial (Online
Resou ce 1)
Fig. 3 The e ec s o he
simula ed educ ion o pa ame e
α o he op heigh and p opo ion
o he plo a ea co e ed by he
e ahed a o he modeled canopy.
The do s and lines indica e he
mean and s anda d de ia ion,
espec i ely, wi hin he a ea.
Ha es ing in ensi y 0
co esponds o he ini ial canopy
9 Page 8 o 13 Annals o Fo es Science (2017) 74:9
3D clus e ing ollowed by con ex poly ope econs uc ion
(Gup a e al. 2010), s acking ho izon al slices o he heigh
alues (Tang e al. 2013), o oxel-based echniques
(Schneide e al. 2014; see also Koch e al. 2014). All hese
me hods in ol e ce ain pa ame e s o be se by he ope a o ,
and he ob ained esul s depend on hese. Fu he , he me hods
a e so a es ed wi h da a densi ies a ying om 4 o 5 m
−2
(Gup a e al. 2010) o10–20 m
−2
(Ka o e al. 2009;Tange al.
2013; Schneide e al. 2014) as opposed o 0.8 m
−2
o his s udy.
The da a s udied he e would ha e esul ed in an ex emely
coa se oxel g id esolu ion. On he o he hand, equi ing da a
wi h a highe densi y would mean dis inc da a acquisi ions and
he e o e conside ably highe in en o y cos s.
Applying il a ions, one can adjus he le el o de ail in he
iangula ed poin cloud and hus accoun o canopy gaps and
de ailed p ope ies exis ing in he da a, which is con olled by
il a ion pa ame e α. The app oaches o de e mine he il a-
ion pa ame e ha e included using he mos easible
popula ion-speci ic ixed alue like in Vauhkonen e al.
(2012), who e alua ed a ange o α’s, bu a he han selec ing
one pa icula alue, hey used me ics quan i ying he di e -
ence o he shape ob ained wi h he ange o α’s o he con ex
hull o he poin da a, which co esponds o α-shapes wi h
α→∞. Vauhkonen e al. (2014,2016) op imized he deg ee
o il a ion wi h espec o ield-measu ed o es a ibu es and
p edic ed his deg ee by means o linea eg ession. In he
p esen s udy, he selec ion o αwas based on analyzing he
pe sis en homology o he il a ions (Edelsb unne e al.
2002). This app oach esul ed in less accu a e pa ame e es i-
ma es (e.g., R
2
= 0.26 o he olumes in he p esen s udy s.
0.49–0.83 epo ed in he p e ious s udies) bu elimina ed he
need o a ield aining da a equi ed o he op imiza ion
(Vauhkonen e al. 2014,2016).
The pe sis en homology app oach was es ed o eplace he
calib a ion ield da a by analyzing he inhe en opological
p ope ies o he poin da a. A simple app oach based on he
bi h and dea h diag ams o he a ious simplices
(Edelsb unne e al. 2002) u ned ou o p oduce easonably
good in o ma ion. The pe sis en homology o ege a ion
poin clouds should be mo e comp ehensi ely analyzed a e
his ini ial es . Fo example, one could de i e he Eule cha -
ac e is ic and he Be i numbe s om he il a ions (Robins
2002) and use hem in a simila manne han he bi h and
dea h diag am was used. One could u he a emp o quan i y
he ob ained il a ion by o he geome ic ea u es han o al
olumes such as edge leng hs o ace a eas. The app oach has
a p o ound ma hema ical o malism on opological connec i -
i y and implemen a ions based on e icien da a s uc u es a e
a ailable (Pion and Teillaud 2013;Dae al.2013).
A de ini e di e ence and a s eng h o he iangula ion-
based app oach o e al e na i e 3D modeling o desc ip ion
echniques is he abili y o simula e changes o he geome ic
models ob ained. This is an inhe en p ope y o he echnique
adop ed and may be simply implemen ed by adjus ing he
deg ee o il e ing, bu , o he bes o ou knowledge, his
op ion has no been u ilized elsewhe e beyond he p esen
s udy. He e, i was used o depic impac s o biomass ha es -
ing o he scenic alues o he a ea by simula ing ha es s wi h
a ying in ensi ies. Acco ding o he high CR alues, some
incohe ence exis ed in he pai wise compa isons, which can
possibly be explained by a compa ison o Figs. 3and 4.
Acco ding o Fig. 3, he educ ion o he α alue changed
he op heigh o he canopy models based on ha es ing in-
ensi ies A, B, and C mode a ely, whe eas he educ ion o he
op heigh was clea ly mo e p onounced o ha es s D and E.
This co esponds well wi h he dispe sion be ween A, B, and
C as he mos p e e ed ha es ing in ensi y and he clea ly
lowe p e e ence owa d D and E (Fig. 4). Al hough he can-
opy co e age changed mo e han heigh (Fig. 3), he mode a e
change in he op heigh may ha e domina ed especially in he
Fig. 4 The dis ibu ion (le ) and accumula ion o he ela i e p e e ences be ween ha es ing in ensi ies. The hick line depic s he mean o all
p e e ences. The dashing o he indi idual p e e ences a ies acco ding o which ha es ing in ensi y was p e e ed mos
Annals o Fo es Science (2017) 74:9 Page 9 o 13 9