Change Detection of Tree Biomass with Terrestrial Laser Scanning and Quantitative Structure Modelling

Remo e Sens. 2014, 6, 3906-3922; doi:10.3390/ s6053906 emo e sensing ISSN 2072-4292 www.mdpi.com/jou nal/ emo esensing A icle Change De ec ion o T ee Biomass wi h Te es ial Lase Scanning and Quan i a i e S uc u e Modelling Sanna Kaasalainen 1,*, Anssi K ooks 1, Ja i Liski 2, Pasi Raumonen 3, Ha i Kaa inen 1, Mikko Kaasalainen 3, Ee u Pu onen 1, Ka i An ila 1 and Raisa Mäkipää 4 1 Finnish Geode ic Ins i u e, Geodee in inne 2, FI-02431 Masala, Finland; E-Mails: Anssi.K ooks@ gi. i (A.K.); [email p o ec ed] (H.K.); [email p o ec ed] (E.P.); [email p o ec ed]i (K.A.) 2 Finnish En i onmen Ins i u e, Mechelininka u 34a, FI-00251 Helsinki, Finland; E-Mail: Ja i.Liski@ympa is o. i 3 Tampe e Uni e si y o Technology, Depa men o Ma hema ics, P.O. Box 553, FI-33101 Tampe e, Finland; E-Mails: pasi. aum[email p o ec ed] (P.R.); Mikko.Kaasalaine[email p o ec ed] (M.K.) 4 Finnish Fo es Resea ch Ins i u e, Jokiniemenkuja 1, PL 18, FI-01301 Van aa, Finland; E-Mail: Raisa.Makipaa@me la. i * Au ho o whom co espondence should be add essed; E-Mail: Sanna[email p o ec ed]; Tel.: +358-295-308-031; Fax: +358-929-555-211. Recei ed: 29 Janua y 2014; in e ised o m: 21 Ma ch 2014 / Accep ed: 9 Ap il 2014 / Published: 30 Ap il 2014 Abs ac : We p esen a new applica ion o e es ial lase scanning and ma hema ical modelling o he quan i a i e change de ec ion o ee biomass, olume, and s uc u e. We in es iga e he easibili y o he app oach wi h wo case s udies on ees, assess he accu acy wi h labo a o y e e ence measu emen s, and iden i y he main sou ces o e o , and he ways o mi iga e hei e ec on he esul s. We show ha he changes in he ee b anching s uc u e can be ep oduced wi h abou ±10% accu acy. As he cu en biomass de ec ion is based on des uc i e sampling, and he change de ec ion is based on empi ical models, ou app oach p o ides a non-des uc i e ool o moni o ing impo an o es cha ac e is ics wi hou labo ious biomass sampling. The e iciency o he app oach enables he epea ing o hese measu emen s o e ime o a la ge numbe o samples, p o iding a as and e ec i e means o moni o ing o es g ow h, mo ali y, and biomass in 3D. OPEN ACCESS Remo e Sens. 2014, 6 3907 Keywo ds: e es ial lase scanning; au oma ic ee modelling; o es moni o ing; b anch size dis ibu ion; change de ec ion 1. In oduc ion The moni o ing o o es esou ces has adi ionally concen a ed on he olume o s em wood. This is because human in e es in o es s has s ongly ocused on his economically mos aluable o es cha ac e is ic. Today, he impo ance o o es s is seen in a wide pe spec i e. I is acknowledged ha se e al ecosys em se ices ha o es s p o ide a e ela ed o he whole biomass o ees a he han he s em only. These se ices include, o example, ca bon seques a ion, o es bioene gy esou ces and o es biodi e si y alue. T ee biomass is moni o ed using es ablished bu a he coa se me hods. A common me hod o biomass moni o ing (such as ha in he IPCC Guidelines [1]) is based on allome ic equa ions (e.g., [1,2]). These equa ions gi e he biomass es ima es as a unc ion o s em cha ac e is ics, such as ee heigh and he diame e [3,4]. The same equa ions a e also used o es ima e changes in he biomass based on changes in he s em cha ac e is ics [1]. The mo ali y o ee biomass componen s is moni o ed using li e collec o s placed below ee canopies o de i ing es ima es om canopy measu emen s (e.g., [5]). The li e p oduc ion measu emen s a e combined wi h he biomass es ima es o ob ain u no e a es o ee biomass componen s; he u no e a e is equal o he li e p oduc ion di ided by he biomass (e.g., [6]). These u no e a es a e hen used o es ima e li e p oduc ion based on he biomass es ima es. Al hough hese me hods a e p ac ical, hei eliabili y and use ulness can s ill be imp o ed. Fi s , he allome ic equa ions and he biomass u no e a es can be made mo e eliable and applicable o new condi ions by aking mo e measu emen s. Since he cu en allome ic equa ions and biomass u no e a es a e based on labo ious biomass and li e measu emen s, hey a e s ill based on ela i ely small da a se s ha ep esen selec ed in ensi ely s udied si es. Second, hese me hods do no p o ide all he in o ma ion needed o cu en and u u e analyses. Fo example, he size o li e elemen s is an impo an a ibu e a ec ing he decomposi ion o li e [7,8]. This in o ma ion is hus ele an o he ca bon budge o o es s, bu is e y labo ious o measu e om he li e collec o s. In addi ion, he dis ibu ion o he biomass wi hin a canopy is a key cha ac e is ic in unde s anding he ligh -use e iciency and he compe i i e s a us o a ee [9]. The ligh -use e iciency (LUE); i.e., he amoun o ca bon ixed pe uni o abso bed pho osyn he ically ac i e adia ion, inc eases wi h he p opo ion o ligh ha is ecei ed a low i adiances and is he e o e highe o clumped canopies [10]. Thus, as measu emen s o he canopy b anching pa e n can help o p edic he e iciency wi h which canopies ha es ligh o ca bon assimila ion. Since he LUE is equen ly used wi h he emo e-sensed No malized Di e ence Vege a ion Index (NDVI) o calcula e p oduc i i y, me hods ha imp o e he unde s anding o he a ia ion o he LUE a e impo an o global ca bon balance es ima es [11]. Te es ial lase scanning (TLS) has become inc easingly impo an in o es s udies because o i s capabili y o p o iding accu a e 3D ee da a wi h e icien and ligh weigh ins umen s o ield use. Thus a , TLS-based me hods ha e been es ablished o de ec ing o es a ibu es such as ee loca ion, he diame e a b eas heigh (DBH), heigh , s em olume, and he o al biomass [12–14]. The o al Remo e Sens. 2014, 6 3908 biomass has been shown o co ela e wi h he TLS poin densi y o dis ibu ion [13,15], DBH, and ee heigh , om which i is possible o e ie e wi h allome ic biomass equa ions [3,4,16]. These me hods p o ide he o al biomass o a ee and a s and, wi hou in o ma ion on i s dis ibu ion in he canopy. T ee modelling has hus a mos ly ocused on e ie ing he s em olume o o es ( imbe ) in en o y pu poses [17,18]. Qui e ecen ly, me hods ha e also been p esen ed o ee 3D s uc u e including b anches [19–21]. These s udies ha e also ex ended in o 3D econs uc ions o he s ump- oo sys ems [22,23]. An e o ange o ±10% has been achie ed o he main s em olume, whe eas o b anches, he cumula i e b anch olumes ha e been es ima ed a ±30% accu acy o b anches down o 7 cm in diame e [20]. Inc easing he numbe o scans has been shown o educe he e o s somewha [21]. Ge ing quan i a i e s uc u e models o ees wi h small b anches has, howe e , s ill been a challenge, and he small b anches ha e mos o en been excluded om he analysis. A plan opology model was p esen ed in [24] o desc ibe he opology and geome y o plan s. They also eco ded he spa ial coo dina es o plan s (e.g., b anch ips). In ou p e ious s udy, we ha e shown ha he s uc u e o ees can be cha ac e ized in de ail based on TLS measu emen s combined wi h 3D quan i a i e s uc u e modelling (which we he ea e call QSM) [25]. I hese measu emen s and his modelling we e epea ed o e ime, i could po en ially p o ide a la gely au oma ic, non-des uc i e and as means o es ima e he g ow h and mo ali y o ee biomass componen s in 3D. In ou QSM o ees, we use he geome ic p imi i es app oach wi h ci cula cylinde s. Al e na i e me hods o modelling he ee olume based on oxels and oxel skele ons exis (e.g., [14,26,27]). While he oxel and skele oniza ion models ha e been used success ully o ex ac ing ee me ics, such as he diame e and heigh [28], he QSM is di e en om hese models because i has been designed o ollow he simple mo phological ules o ee s uc u e (such as b anches a ached o he s em and sub-b anches being a ached o he main b anches, c . [25]) as a s a ing poin o he calcula ion. A pa icula challenge in he oxel me hods is ha hey equi e a comple e sampling o he ee su ace in o de o ill he in e io oxels. This equi es a la ge numbe o measu emen s and scan posi ions (e.g., 20–60 million poin s om 4 o 5 scan posi ions pe ee [28]). Wi h cylinde s, he ine i able gaps in he su ace sampling and he ac ha mos b anches a e sampled only om he bo om side a e no so c i ical. Simila ly, he oxel skele on models do no equi e a la ge numbe o measu emen s because he oxels a e only needed o he econs uc ion o he skele on and hen he olume is modelled, e.g., wi h cylinde s [27]. Changes in he o al biomass ha e been moni o ed wi h TLS and oxel o con ex hull models [29], bu mo e s udies a e needed o quan i y he dis ibu ion o he changes. The objec i es o his s udy we e o (1) e alua e he sui abili y o his app oach (p esen ed in [25]) o es ima ing changes in ee biomass, e.g., g ow h and li e p oduc ion; and (2) iden i y he mos impo an a eas o imp o emen . The compa ison wi h e e ence measu emen s also enabled us o imp o e he da a p ocessing and modelling s eps o op imize he p ocedu e by inding and elimina ing he sou ces o majo sys ema ic e o s in he measu emen (e.g., emo ing ex a noise) o modelling he smalles b anch ips. In his way, we imp o ed he p ocedu e o mo e eliable esul s. The change de ec ion app oach p esen ed in his pape is applied o ee-s anding ees. Ou u u e objec i e is o ex end he app oach in o la ge a eas and o alida e such plo -based modelling. We ha e used and will use he ou pu s o his p ocedu e o compu e ca bon emissions ia, e.g., he Yasso soil ca bon model as in [23]. Remo e Sens. 2014, 6 3909 This a icle is o ganized as ollows: he ma e ials and me hods a e p esen ed in Sec ion 2. The esul s o bo h he labo a o y and ield cases a e in Sec ions 3.1–3.3, and he discussion and conclusions a e p esen ed in Sec ions 3.4 and 4, espec i ely. 2. Ma e ials and Me hods 2.1. Samples We demons a e he app oach wi h wo case s udies: he i s one was ca ied ou in a labo a o y o a la ge (abou 2 m in leng h) aspen (Populus emula) b anch. We c ea ed a ime se ies o olume and leng h measu emen s by cu ing o pieces o he sub-b anches, a e which he b anch was scanned wi h a e es ial lase scanne (see Sec ion 2.2). The scanning was epea ed ou imes ( om h ee di ec ions each ime) and he cu s we e ca ied ou be ween each scan. The aim o hese measu emen s was o p o ide e e ence o alida e he b anch size ( he olume and leng h o he s em and all sub-b anches) es ima ion wi h he QSM model. The second case s udy was a ield moni o ing o changes in ee biomass in Espoonlah i, Finland. We p oduced a ime se ies o TLS poin clouds o a maple (Ace pla anoides) shown in Figu e 1. Fi e scans we e ca ied ou : in Feb ua y 2011, No embe 2011, No embe 2012, Ap il 2013, and No embe 2013. The changes in ee b anch olume and b anch leng h, caused by g ow h and mo ali y, we e modelled om each poin cloud. Figu e 1. (Le ) he maple ee in Espoonlah i (pho og aphed Feb ua y 2011). Some o he whi e sphe ical e e ence a ge s used o egis a ion a e also isible, be ween he lamp pos and he ee; (Righ ) a sec ion o he poin clouds om Feb ua y 2011 ( ed) and No embe 2011 (g een), showing a missing b anch deno ed by an a ow. 2.2. Te es ial Lase Scanning Bo h he ee and he b anch sample we e measu ed wi h a phase-based e es ial lase scanne Leica HDS6100, see Table 1 o scanne pa ame e s. The same scanne (and he same scanning Remo e Sens. 2014, 6 3910 pa ame e s) was also used in ou p e ious s udies [25]. The ins umen al pa ame e s ela ed o each measu emen a e lis ed in Table 2. The scanning was ca ied ou wi h he “High” esolu ion se up. To p oduce a poin cloud, h ee s a iona y TLS scans we e ca ied ou o each sample om di e en di ec ions, and hese scans we e co- egis e ed using whi e sphe ical e e ence a ge s ( isible in Figu e 1 du ing he TLS o he Espoonlah i maple, c . [13]). The co- egis a ion accu acy is bes desc ibed in e ms o he e o in loca ing he cen e poin s o he sphe ical e e ence a ge s, which a ied be ween 1 and 3 mm in he labo a o y. Howe e , he e a e g ea e sou ces o unce ain y, especially in ou doo measu emen s, caused by, e.g., he b anches mo ing du ing he scans. In Espoonlah i, some pa s o he a ge (such as he canopy) we e u he om he scanne , and he co- egis a ion accu acy a ied om 4 o 7 mm. The dis ance be ween he scanne and he ee a ied om 1 o 2 m in he labo a o y and 9–20 m in Espoonlah i. The aspen b anch was scanned be o e any cu s and a e each h ee cu s, om h ee di e en di ec ions each ime. A e he hi d cu , we ca ied ou wo independen scans (deno ed wi h cu 3A and 3B in he ollowing sec ions), i.e., al oge he six scans we e made. This was done o compa e he epea abili y o he measu emen and modelling p ocedu es. Table 1. Te es ial lase scanne pa ame e s. Scanne Leica HDS6100 Wa eleng h 650–690 nm Field o iew 360° × 310° Poin sepa a ion 0.036° Beam diame e 3 mm Beam di e gence 0.22 m ad Maximum Range 79 m Table 2. Measu emen speci ica ions o he labo a o y and ield case. Measu emen Labo a o y (aspen) Espoonlah i (maple) Numbe o poin s 390,000–460,000 1–5 million Ho izon al dis ance be ween scanne and b anch/ ee s em 1.5–1.9 m 7–12 m A e age poin densi y 11–25 poin s /cm2 2–5 poin s/cm2 The p e-p ocessing o he TLS da a was ca ied ou wi h he Z+F Lase Con ol 8 so wa e (Zolle + F öhlich GmbH). The dis ance measu emen o he scanne , based on phase di e ence, causes inc eased mixed pixel noise in complex s uc u es, whe e he lase beam hi s mul iple a ge s a he same ime. The measu emen noise was minimized wi h in ensi y- and poin -densi y based il e s a ailable wi h he Z+F so wa e. In his case, noise il e ing esul ed in he ejec ion o 1%–3% o he da a poin s. 2.3. Re e ence Measu emen s To change he b anch leng h and olume in he labo a o y, he sample b anches o he aspen we e cu a e each measu emen o poin clouds. The leng hs o he cu b anch pieces we e measu ed manually (wi h a s anda d me ic measu e). The o al olume emo ed in each cu was measu ed by weighing Remo e Sens. 2014, 6 3911 he cu b anches, and app oxima ing hei olume on he basis o hei densi y. The densi y o he esh sample b anches was measu ed by subme ging he b anch pieces in wa e and eco ding he inc ease o weigh o he wa e eplaced by he b anches. The weigh ep esen s he olume, and he esh weigh di ided by his olume gi es he densi y, which was used as a densi y app oxima ion o he en i e b anch. In his way, he app oxima e esh densi y o ou sample b anches was 0.925 kg/dm3. We also measu ed he o al leng h and olume o he en i e b anch (including all he sub-b anches) a e he hi d cu o b anches (a e which we made wo se s o scans, which we e p ocessed as wo independen h ee-scan measu emen s, 3A and 3B). Because he o al olume e e ence was e ie ed using he app oxima e densi y measu ed om small b anch bi s, some unce ain y may occu in he densi y o he la ge pa s o he b anch. These b anch leng h and olume alues we e compa ed o hose p oduced by TLS and he QSM model [25]. 2.4. Quan i a i e S uc u e Modelling (QSM) The b anch size and olume o each sample was compu ed using he quan i a i e s uc u e modelling me hod [25]. In he QSM, he su ace o he isible ee pa s is econs uc ed by making a lexible su ace model o he ee o model he s em and b anch sizes and he opological b anching s uc u e. The me hod uses a local app oach in which he poin cloud is co e ed wi h small se s co esponding o connec ed su ace pa ches in he ee su ace. Wi h hese pa ches he en i e ee is segmen ed in o s em and b anches. The pa ches a e andomly bu e enly dis ibu ed along he isible ee su ace and hei size de e mines he smalles de ails ha can be sepa a ed o he ee model. The segmen s a e hen modelled wi h collec ion o cylinde s i ed o he de ails o he segmen s. Wi h hese cylinde s, he b anching s uc u e, olume, and b anch size dis ibu ions, e c. can be app oxima ed bo h o he whole ee and some o i s pa s indi idually. We use cylinde s because o all he geome ic p imi i es ha app oxima e he local s em and b anch shape well, hey a e he mos obus and eliable o i . Mo e de ails o he model and i s alida ion a e p o ided in [25]. We also compa ed he QSM wi h he T iangula ed I egula Ne wo k (TIN), commonly used o il e ing lase scanne da a o digi al ele a ion model (DEM) gene a ion and p oducing 3D models o di e en objec s [30,31]. The TIN model was a educed 3D Delaunay iangula ion. The algo i hm had he ollowing s eps: 1. The aw poin cloud was i s iangula ed wi h 3D Delaunay iangula ion. 2. Te ahed ons wi h side leng hs o e a p ede ined h eshold (3 cm) we e emo ed. 3. The su ace o he educed iangula ion was sea ched and hen di ided in o sepa a e laye s using poin s on he su ace as ini ial poin s. 4. The second e ahed on educ ion un wi h 2 cm h eshold was ca ied ou o he e ahed ons in he wo ou e mos laye s. 5. The ee olume was es ima ed by summing he olume o he e ahed ons emaining a e bo h educ ion uns. No addi ional smoo hing was done o he poin clouds, hus some noise poin s nea b anches and he s em we e le , esul ing in o e es ima ion o he ee olume. Remo e Sens. 2014, 6 3912 3. Resul s and Discussion 3.1. Aspen B anch Measu ed in Labo a o y: Valida ion o B anch Size Modelling The changes in he olume (Table 3) and leng h (Table 4) o b anches es ima ed wi h he QSM ag eed well wi h e e ence measu emen s. The modelling was ca ied ou as i he b anch we e a small ee, i.e., he main b anch was segmen ed as “s em”, whe eas he sub-b anches a e ea ed as b anches. All changes a e mean alues o 10 modelling uns, whe e he size o he a e age pa ch emains cons an bu hei numbe and loca ions a y andomly be ween di e en models which a ec s he segmen a ion and i ed cylinde s. Typical anges o he s anda d de ia ions o he 10 models o each cu a e 5%–15% o he b anch olume and 1%–2% o he b anch leng h. Table 3. Change o aspen b anch olume (in (li es)) a e each cu . Cu s 3A and 3B ep esen wo independen scans o he ee a e he hi d cu ( he e e ence being he same). Cu 1 (L) Cu 2 (L) Cu 3A (L) Cu 3B (L) Re e ence −0.06 −0.06 −0.08 −0.08 QSM −0.02 −0.06 −0.11 −0.11 Re e ence, cumula i e −0.06 −0.11 −0.19 −0.19 QSM, cumul. −0.02 −0.08 −0.18 −0.18 Table 4. Cumula i e change o aspen b anch leng h (in me es) a e each cu . Cu s 3A and 3B ep esen wo independen scans o he ee a e he hi d cu ( he e e ence being he same). In “Re e ence (>5 cm)”, he smalles sub-b anches (less han 5 cm leng h) we e no included in he o al b anch leng h. Cu 1 (m) Cu 2 (m) Cu 3A (m) Cu 3B (m) Re e ence −3.85 −7.25 −10.43 −10.43 Re e ence (>5 cm) −3.08 −5.37 −7.97 −7.97 QSM −2.29 −5.06 −6.53 −6.37 The poin clouds and models a e p esen ed in Figu es 2 and 3. To compa e wi h he e e ence measu emen s, which we e ca ied ou o he cu pieces only, he modelling esul s a e also p esen ed he e as changes om he o iginal condi ion o he b anch (be o e any cu s). The model has unde es ima ed he cumula i e changes, which is mos likely a esul o inaccu acies in he measu emen ; see Sec ion 3.2 o mo e de ails. A close examina ion o he poin clouds and models (see Figu e 4) e ealed ha he smalles sub-b anches ( hose wi h leng h less han 5 cm) we e ha dly isible in he poin cloud, and had mos ly been le ou by he model. The numbe o poin s o hose bi s is oo small o o m a cylinde . To ge some insigh in o he e ec o his ea u e on he esul s, we made ano he leng h e e ence, whe e he smalles (<5 cm) sub-b anches had also been omi ed (see Table 4). In addi ion, some segmen a ion e o s we e also p esen , esul ing in he cylinde s o e lapping each o he (Figu e 4). A u he analysis on he o e all accu acy o he app oach is p o ided in Sec ion 3.2. Remo e Sens. 2014, 6 3913 Figu e 2. TLS poin clouds o he aspen b anch: he o iginal b anch (g een) and ha a e he las ( hi d) cu ( ed). The poin cloud has been plo ed in o an xyz-coo dina e sys em o show he scaling (in (me es)). Figu e 3. Quan i a i e s uc u e models o he aspen b anch (wi h cylinde s plo ed o e he poin cloud): (a) he o iginal b anch; (b) he b anch a e he hi d cu . The o de o b anches has been deno ed wi h colou s: he s em ( he main b anch) is blue, he i s o de and second o de b anches a e g een and ed, espec i ely. Remo e Sens. 2014, 6 3914 Figu e 4. E o sou ces in he QSM, ma ked wi h g een ci cles: (a) B anches sho e han 5 cm a e poo ly isible in he poin cloud (blue poin s) and ha e been omi ed by he model (no cylinde ); (b) Some cylinde s o e lap in he model. The QSM and TIN models a e compa ed in Figu e 5, which also p esen s he s anda d de ia ion e o s o he QSM. In he ini ial es s, bo h models o e es ima ed he b anch olume change by se e al o de s o magni ude. This esul s om he inaccu acy in measu emen , because he diame e s o he b anch ips a e less han o equal o he lase spo size (abou 3 mm) owa ds he b anch ips. The e o e, we used a modi ied app oach o he QSM p esen ed in [25], whe e he sizes o b anches wi h diame e less han 1 cm we e app oxima ed wi h a b anch diame e ha dec eases linea ly owa ds he b anch end. This imp o ed he esul s, al hough some disc epancy be ween he measu emen s and he model s ill emained. Ano he imp o emen was he emo al o ex a noise. Figu e 5. Compa ison o he QSM (blue) and TIN ( ed) models o : (a) The o al olume. The dec easing end has no been ep oduced in cu 2 by ei he o he models; (b) The o al leng h. The QSM s anda d de ia ions a e p esen ed as e o ba s. Re e ence measu emen s o o al olume and leng h, plo ed wi h g een symbols, only exis o cu 3. The TIN model has o e es ima ed he b anch olumes mo e han he QSM (c . Figu e 5), whe eas he leng hs ha e been unde es ima ed somewha . The TIN model seems o be mo e sensi i e o measu emen noise in his case han he QSM, especially in he case o hin b anches, o which he TIN model ends o include he noise poin s in he calcula ion. The compu a ional e e ence o b anch Remo e Sens. 2014, 6 3921 11. Ruimy, A.; Ke goa , L.; Bondeau, A. Compa ing global models o e es ial ne p ima y p oduc i i y (NPP): Analysis o di e ences in ligh abso p ion and ligh -use e iciency. Glob. Chang. Biol. 1999, 5, 56–64. 12. Pueschel, P. The in luence o scanne pa ame e s on he ex ac ion o ee me ics om FARO Pho on 120 e es ial lase scans. ISPRS J. Pho og amm. Remo e Sens. 2013, 78, 58–68. 13. Kanka e, V.; Holopainen, M.; Vas a an a, M.; Pu onen, E.; Yu, X.; Hyyppä, J.; Vaaja, M.; Hyyppä, H.; Alho, P. Indi idual ee biomass es ima ion using e es ial lase scanning. ISPRS J. Pho og amm. Remo e Sens. 2013, 75, 64–75. 14. P ei e , N.; Go e, B.; Win e halde , D. Au oma ic econs uc ion o single ees om e es ial lase scanne da a. In . A ch. Pho og amm. Remo e Sens. Spa . In . Sci. 2004, 35, 114–119. 15. Kaasalainen, S.; Hyyppä, J.; Ka jalainen, M.; K ooks, A.; Lyy ikäinen-Saa enmaa, P.; Holopainen, M.; Jaakkola, A. Compa ison o e es ial lase scanne and syn he ic ape u e ada da a in he s udy o o es de olia ion. In . A ch. Pho og amm. Remo e Sens. Spa . In . Sci. 2010, 38, 82–87. 16. Hun e , M.O.; Kelle , M.; Vic o ia, D.; Mo on, D.C. T ee heigh and opical o es biomass es ima ion. Biogeosciences 2013, 10, 8385–8399. 17. Liang, X.; Hyyppä, J.; Kaa inen, H.; Holopainen, M.; Melkas, T. De ec ing changes in o es s uc u e o e ime wi h bi- empo al e es ial lase scanning da a. ISPRS In . J. Geo-In . 2012, 1, 242–255. 18. Holopainen, M.; Vas a an a, M.; Kanka e, M.; Rä y, M; Vaaja, M.; Liang, X.; Yu, X.; Hyyppä, J.; Hyyppä, H.; Vii ala, R.; e al. Biomass es ima ion o indi idual ees using s em and c own diame e TLS measu emen s. In . A ch. Pho og amm. Remo e Sens. Spa . In . Sci. 2011, 38, 91–95. 19. Cô é, J.-F.; Widlowski, J.-L.; Fou nie , R.A.; Ve s ae e, M.M. The s uc u al and adia i e consis ency o h ee-dimensional ee econs uc ions om e es ial LiDAR. Remo e Sens. En i on. 2009, 113, 1067–1081. 20. Dasso , M.; Colin, A.; San enoise, P.; Fou nie , M.; Cons an , T. Te es ial lase scanning o measu ing he solid wood olume, including b anches, o adul s anding ees in he o es en i onmen . Compu . Elec on. Ag ic. 2012, 89, 86–93. 21. Keigh ley, K.E.; Bawden, G.W. 3D olume ic modeling o g ape ine biomass using T ipod LiDAR. Compu . Elec on. Ag ic. 2010, 74, 305–312. 22. Danjon, F.; Reubens, B. Assessing and analyzing 3D a chi ec u e o woody oo sys ems, a e iew o me hods and applica ions in ee and soil s abili y, esou ce acquisi ion and alloca ion. Plan Soil 2008, 303, 1–34. 23. Liski, J.; Kaasalainen, S.; Raumonen, P.; Akujä i, A.; K ooks, A.; Repo, A.; Kaasalainen, M. Indi ec emissions o o es bioene gy: De ailed modeling o s ump- oo sys ems. GCB Bioene gy 2014, in p ess. A ailable online: h p://onlinelib a y.wiley.com/doi/10.1111/gcbb.12091/abs ac (accessed on 27 Janua y 2014). 24. Godin, C.; Cos es, E.; Sinoque , H. A me hod o desc ibing plan a chi ec u e which in eg a es opology and geome y. Ann. Bo . 1999, 84, 343–357. 25. Raumonen, P.; Kaasalainen, M.; Åke blom, M.; Kaasalainen, S.; Kaa inen, H.; Vas a an a, M.; Holopainen, M.; Disney, M.; Lewis, P. Fas au oma ic p ecision ee models om e es ial lase scanne da a. Remo e Sens. 2013, 5, 491–520. Remo e Sens. 2014, 6 3922 26. Hosoi, F.; Nakai, Y.; Omasa, K. 3-D oxel-based solid modeling o a b oad-lea ed ee o accu a e olume es ima ion using po able scanning LiDAR. ISPRS J. Pho og amm. Remo e Sens. 2013, 82, 41–48. 27. Bucksch, A.; Lindenbe gh, R. CAMPINO—A skele oniza ion me hod o poin cloud p ocessing. ISPRS J. Pho og amm. Remo e Sens. 2008, 63, 115–127. 28. Vonde ach, C.; Voeg le, T.; Adle , P. Voxel-based app oach o es ima ing u ban ee olume om e es ial lase scanning da a. In . A ch. Pho og amm. Remo e Sens. Spa . In . Sci. 2012, 39, 451–456. 29. Olsoy, P.J.; Glenn, N.F.; Cla k, P.E.; De ybe y, D.R. Abo eg ound o al and g een biomass o d yland sh ub de i ed om e es ial lase scanning. ISPRS J. Pho og amm. Remo e Sens. 2014, 88, 166–173. 30. Axelsson, P. DEM gene a ion om lase scanne da a using adap i e TIN models. In . A ch. Pho og amm. Remo e Sens. Spa . In . Sci. 2000, 33, 110–117. 31. Si hole, G.; Vosselman, G. Expe imen al compa ison o il e algo i hms o ba e-Ea h ex ac ion om ai bo ne lase scanning poin clouds. ISPRS J. Pho og amm. Remo e Sens. 2004, 59, 85–101. 32. Cao, J.; Tagliasacchi, A.; Olson, M.; Zhang, H.; Su, Z. Poin Cloud Skele ons ia Laplacian-Based Con ac ion. In P oceedings o 2010 IEEE Con e ence on Shape Modeling In e na ional, Aix-en-P o ence, F ance, 21–23 June 2010; pp. 187–197. 33. Yang, X.; S ahle , A.H.; Schaa , C.B.; Jupp, D.L.B.; Yao, T.; Zhao, F.; Wang, Z.; Cul eno , D.S.; Newnham, G.J.; Lo ell, J.L.; e al. Th ee-dimensional o es econs uc ion and s uc u al pa ame e e ie als using a e es ial ull-wa e o m lida ins umen (Echidna®). Remo e Sens. En i on. 2013, 135, 36–51. 34. Van Leeuwen, M.; Coops, N.C.; Hilke , T.; Wulde , M.A.; Newnham, G.J.; Cul eno , D.S. Au oma ed econs uc ion o ee and canopy s uc u e o modeling he in e nal canopy adia ion egime. Remo e Sens. En i on. 2013, 136, 286–300. © 2014 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 license (h p://c ea i ecommons.o g/licenses/by/3.0/).