Automated registration of potential locations for solar energy production with Light Detection And Ranging (LiDAR) and small format photogrammetry

Else ie Edi o ial Sys em( m) o Jou nal o Cleane P oduc ion Manusc ip D a Manusc ip Numbe : JCLEPRO-D-15-00934R1 Ti le: Au oma ed egis a ion o po en ial loca ions o sola ene gy p oduc ion wi h LiDAR and small o ma pho og amme y A icle Type: O iginal Resea ch Pape Co esponding Au ho : D . Szilá d Szabó, Ph.D. Co esponding Au ho 's Ins i u ion: Uni e si y o Deb ecen Fi s Au ho : Szilá d Szabó, Ph.D. O de o Au ho s: Szilá d Szabó, Ph.D.; Pé e Enyedi; Miklós Ho á h; Zol án Ko ács; Pé e Bu ai, Phd; Tamás Csoknyai, PhD; Ge gely Szabó, Phd Abs ac : Ene gy p oduc ion and consump ion is a key elemen in u u e de elopmen which is in luenced bo h by he echnical possibili ies a ailable and by decision make s. Sus ainabili y issues a e closely linked in wi h ene gy policy, gi en he desi e o inc ease he p opo ion o enewable ene gy. Acco ding o he Ho izon 2020 clima e and ene gy package, EU membe coun ies ha e o educe he amoun o g eenhouse gases hey emi by 20%, o inc ease he p opo ion o enewable ene gy o 20% and o imp o e ene gy e iciency by 20% by 2020. In his s udy we aim o assess he oppo uni ies a ailable o exploi sola adia ion on oo s wi h LiDAR and pho og amme y echniques. The su eyed a ea was in Deb ecen, he second la ges ci y in Hunga y. An ae ial LIDAR su ey was conduc ed wi h a densi y o 12 poin s/m2, o e a 7×1.8 km wide band. We ex ac ed he building and oo models o he buildings om he poin cloud. Fu he mo e, we applied a low-cos d one (DJI Phan om wi h a GoP o came a) in a smalle a ea o he LIDAR su ey and also c ea ed a 3D model: buildings and oo planes we e iden i ied wi h mul i esolu ion segmen a ion o he digi al su ace models (DSM) and o hopho o co e ages. Building heigh s and building geome y we e also ex ac ed and alida ed in ield su eys. 50 buildings we e chosen o he geode ic su ey and he esul s o he accu acy assessmen we e ex apola ed o o he buildings; in addi ion o his, 100 building heigh s we e measu ed. We ocused p ima ily on he oo s, as hese su aces o e possible loca ions o he mal and pho o ol aic equipmen . We de e mined he slope and aspec o oo planes and calcula ed he incoming sola ene gy acco ding o oo planes be o e compa ing he esul s o he poin cloud p ocessing o LiDAR da a and he segmen a ion o DSMs. Ex ac ed oo geome ies showed a ying deg ees o accu acy: he esea ch p o ed ha LiDAR-based oo -modelling is he bes choice in esiden ial a eas, bu he esul s o he d one su ey did no di e signi ican ly. Gene ally, bo h app oaches can be applied, because he sola adia ion alues calcula ed we e simila . The ae ial echniques combined wi h he mul i esolu ion p ocessing demons a ed can p o ide a aluable ool o es ima e po en ial sola ene gy. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 1 Au oma ed egis a ion o po en ial loca ions o sola ene gy p oduc ion wi h LiDAR and small o ma pho og amme y Szilá d SZABÓ1, Pé e ENYEDI2, Miklós HORVÁTH3, Zol án KOVÁCS1, Pé e BURAI2, Tamás CSOKNYAI3, Ge gely SZABÓ1 1 Depa men o Physical Geog aphy and Geoin o ma ics, Uni e si y o Deb ecen, Egye em é 1. 4032, Deb ecen, Hunga y 2 Resea ch Ins i u e o Remo e Sensing and Ru al De elopmen , 3Uni e si y o Deb ecen, Ká oly Róbe College, Má ai ú 36. 3200, Gyöngyös, Hunga y 3 Depa men o Building Se ice and P ocess Enginee ing, Budapes Uni e si y o Technology and Economics Add ess o co espondence: Szilá d Szabó Depa men o Physical Geog aphy and Geoin o ma ics, Uni e si y o Deb ecen, Egye em é . 1. 4032, Deb ecen, Hunga y, el.:+36 52 512900/22326 (swi chboa d), ax: +36 52 512945, e-mail: szabo.szila [email protected] *Manusc ip Click he e o iew linked Re e ences 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 2 Abs ac Ene gy p oduc ion and consump ion is a key elemen in u u e de elopmen which is in luenced bo h by he echnical possibili ies a ailable and by decision make s. Sus ainabili y issues a e closely linked in wi h ene gy policy, gi en he desi e o inc ease he p opo ion o enewable ene gy. Acco ding o he Ho izon 2020 clima e and ene gy package, EU membe coun ies ha e o educe he amoun o g eenhouse gases hey emi by 20%, o inc ease he p opo ion o enewable ene gy o 20% and o imp o e ene gy e iciency by 20% by 2020. In his s udy we aim o assess he oppo uni ies a ailable o exploi sola adia ion on oo s wi h LiDAR and pho og amme y echniques. The su eyed a ea was in Deb ecen, he second la ges ci y in Hunga y. An ae ial LiDAR su ey was conduc ed wi h a densi y o 12 poin s/m2, o e a 7×1.8 km wide band. We ex ac ed he building and oo models o he buildings om he poin cloud. Fu he mo e, we applied a low-cos d one (DJI Phan om wi h a GoP o came a) in a smalle a ea o he LiDAR su ey and also c ea ed a 3D model: buildings and oo planes we e iden i ied wi h mul i esolu ion segmen a ion o he digi al su ace models (DSM) and o hopho o co e ages. Building heigh s and building geome y we e also ex ac ed and alida ed in ield su eys. 50 buildings we e chosen o he geode ic su ey and he esul s o he accu acy assessmen we e ex apola ed o o he buildings; in addi ion o his, 100 building heigh s we e measu ed. We ocused p ima ily on he oo s, as hese su aces o e possible loca ions o he mal and pho o ol aic equipmen . We de e mined he slope and aspec o oo planes and calcula ed he incoming sola ene gy acco ding o oo planes be o e compa ing he esul s o he poin cloud p ocessing o LiDAR da a and he segmen a ion o DSMs. Ex ac ed oo geome ies showed a ying deg ees o accu acy: he esea ch p o ed ha LiDAR-based oo -modelling is he bes choice in esiden ial a eas, bu he esul s o he d one su ey did no di e signi ican ly. Gene ally, bo h app oaches can be applied, because he sola adia ion alues calcula ed we e simila . The ae ial echniques combined wi h he mul i esolu ion p ocessing demons a ed can p o ide a aluable ool o es ima e po en ial sola ene gy. Keywo ds: oo plane, sola i adia ion, poin cloud, mul i esolu ion segmen a ion, d one 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 3 1. In oduc ion Renewable ene gy esou ces a e becoming inc easingly impo an in he s uc u e o ene gy p oduc ion. As non- enewable sou ces (such as pe oleum o coal) a e o en conside ed pollu e s o he en i onmen , g eenhouse gas p oduce s, o as posing a high isk (nuclea ene gy), i is c ucial o ind solu ions o eplace hem wi h en i onmen - iendly al e na i es. A he same ime, he EU in oduced he Ho izon 2020 F amewo k P og am o Resea ch and Inno a ion: he e iciency o ene gy should be inc eased by 20%, he p opo ion o enewable ene gy should be inc eased by 20%, and g eenhouse gas emissions should be educed by 20% (Eu opean Commission, 2014). Conside ing he p i a e con ibu ion by esiden s, an inc ease in he numbe o passi e houses can ep esen a genuine miles one in e iciency (Kozma e al., 2013), while local ene gy p oduc ion can dec ease he GHGs and imp o e he p opo ion o enewable ene gy sou ces (Fa kas, 2010; Lázá , 2011; Lewis, 2007). In his s udy, we ocus on sola ene gy as a possible solu ion o p i a e ene gy p oduc ion. I is a solu ion which has bo h ad an ages and disad an ages. In he cu en economic en i onmen , p i a e p ope ies a e no suppo ed o ins all pho o ol aic (PV) sola sys ems in Hunga y. Consequen ly, he high cos o ins alla ion is a se ious disad an age, bu i is a solu ion which can o e comple e o pa ial con inuous ene gy o bo h ins i u ions and households. Acco dingly, ema kable e o s ha e been conduc ed o de e mine he sola po en ial o winemaking acili ies (Smy h, 2012). Besides, he e is no loss in ol ed in he anspo a ion o he ene gy. A limi ing ac o is ha no all oo s a e app op ia e o ins alling sola panels, as his depends on he size, aspec and slope o he oo planes. Shadows gene a ed by he oo elemen s, chimneys, an ennas, o by he ees and pylons in he s ee can se iously educe e iciency (S e ano i s, 2013). Roo s can be de ec ed wi h emo e sensing echniques (e.g. Nagy á adi e al., 2013); howe e , a simple iden i ica ion is no su icien o assess which oo s a e sui able o he ins alla ion o PV panels, as me hods mus be employed ha can e eal he oo s’ geome y. Pho og amme y and Ligh De ec ion And Ranging (LiDAR) a e he wo possible me hods sui able o his ask. While pho og amme y equi es ae ial pho og aphs, and he ou come depends on he geome ical esolu ion and he quali y o he images, LiDAR wo ks wi h lase beams and he e lec ing signs a e eco ded. Pho og amme y yields a digi al su ace model (DSM), while LiDAR, based on he emi ed and backsca e ed signs wi h di e en e u ning imes, p o ides a model bo h o he g ound (digi al e ain model, DTM) and he su ace (digi al su ace model, DSM). In e ms o oo de ec ion, bo h echniques a e sui able; we only need in o ma ion abou he su ace o he objec s (i.e. he oo s). The LiDAR echnique was de eloped in he 1960s, bu became popula only in he i s decade o he 2000s. Recen ly, se e al s udies ha e deal wi h e ain and su ace models de i ed om LiDAR poin clouds. Highly de ailed digi al ele a ion models a e he mos popula applica ion ields (e.g. Chasse eau e al., 2011; Liu, 2008) in na u al o u ban en i onmen s (Ghu a e al., 2013; Zlinszky e al., 2014) o o ex ac di e en elemen s o he su ace, such as geomo phic o ms (Do ninge e al., 2011), ees (Mücke e al., 2013), ci y buildings, o s ee u ni u e (P ies nall e al., 2000). 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 4 Nume ous publica ions ha e discussed he de ec ion o buildings based on LiDAR. In esea ch conduc ed by Yu e al., (2010) he accu a e de ec ion o ci y buildings was he aim, as in he case o Zhou and Neumann (2013). Fil e ing buildings was also he objec i e o he wo ks o Mongus e al (2014) and Li e al., (2013). While Alexande e al., (2009) deal wi h oo s uc u e, Lukac e al., (2014) ocused pa icula ly on he po en ial sola adia ion o buil -up a eas wi h LiDAR da a. The e a e se e al esea ch s udies which ha e adop ed a pho og amme ic app oach, oo, anging om he digi al ep esen a ion o he sola panels (Sho is e al., 2008) h ough sola po en ial es ima ion on a ci y-scale (Nex e al., 2013) o a comple e su ey o oo geome y (Lin and Zhang, 2014; P o ic e al., 2012). LiDAR has a ele an ad an age agains pho og amme y as i p o ides da a o he g ound e en i is co e ed by ee ege a ion (Demi e al., 2008; Ko pela e al., 2012). Bo h echniques ha e hei ad an ages and limi s. LiDAR can be conside ed mo e eliable han pho og amme y in e ms o he way da a is collec ed: a lase beam has a oo p in (i.e. a 20-40 cm diame e ci cle) on he su ace and i s size is he unc ion o he di e gence angle and he abo e- a ge ligh heigh (Bin e al., 2008). Thus, lase beams ha e mul iple echoes and o en can pene a e ege a ion and oo s co e ed by ee canopy, so hese can also be su eyed (Shan and To h, 2008). Howe e , due o he oo p in , he ho izon al accu acy is wo se han he e ical (Csanyi and To h, 2007). A majo issue wi h 3D poin clouds is how o handle he da ase , especially in he case o su eys p o iding a e y high poin densi y. Pho og amme y is biased by he ege a ion as i can only p oduce su ace models. Fu he mo e, he echnique is sensi i e o homogenous a ea sec ions, pe iodic objec s and shadows, while LiDAR is independen o hem (Papa odi is and Polido i, 2004). Acco ding o Bal sa ias (1999) he wo echnologies can be used in a complemen a y way o exploi he ad an ages o bo h. Incoming sola i adia ion can be compu ed wi h he in ol emen o slope, aspec , and shadows cas by opog aphic ea u es (e.g. mounds) o o he su ace objec s (e.g. buildings, ees, chimneys, pylons e c., Boehne and An onic, 2009; Quazi e al., 2015). I all o hese pa ame e s a e in ol ed in a model, esul s can be ega ded as eliable (Iqbal, 1983). Calcula ions can be conduc ed based on he app op ia e equa ions, o so wa e, such as A cGIS, SAGA GIS and GRASS GIS, which p o ide sola adia ion models (Wh.m-2.day-1, Ho ie ka and Šu i, 2002; Ho ie ka and Kañuk, 2009; Hengl e al., 2009). All models ha e e o s due o he unde lying concep o o a lack o app op ia e da a, bu in mos cases we do no equi e exac alues, because a good app oxima ion o he possible maximum summed by a gi en ime in e al is su icien . S udies ha e usually been designed o de e mine he a ea o he oo planes and he incoming sola ene gy, bu ha e no compa ed he di e en su eying me hods. Ou aim was o in es iga e and compa e he su ace models o a LiDAR su ey and an ae ial imaging ca ied ou wi h a low cos d one sys em om he pe spec i e o oo de ec ion. We compa ed he esul ing oo shapes and e alua ed hei sui abili y o sola panel ins alla ion o bo h models; u he mo e, we also compa ed 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 5 he incoming sola i adia ion o he models. We also compa ed he cos -bene i issues o he d one and LiDAR based echniques. 2. Ma e ials and me hods 2.1. Da a collec ion A combined LiDAR and high esolu ion ae ial imaging was ca ied ou o e a 7 km2 a ea in Deb ecen (Eas e n-Hunga y). A Leica ALS70-HP and a Leica RDC 30 RGBN 60 MP we e used in he su ey (1000 m ligh heigh , 780 m swa h, sinusoid scan pa e n, 20% o e lap). Poin densi y was 12 poin /km2, which was in acco dance wi h he sugges ion made by Cekada e al., (2010). An accu acy assessmen was ca ied ou on he whole s udy a ea; howe e , we used only a smalle pa in he analysis o in es iga e incoming sola i adia ion, whe e he d one su ey was possible (Fig. 1). The oo ing ma e ial o he buildings in he s udy a ea was ed ile, ensu ing he c ea ion o a uni o m da abase independen o LiDAR in ensi y alues. # Fig. 1. app oxima ely he e The d one su ey was conduc ed a a e age al i ude o 93 m wi h a DJI Phan om quad ocop e and a GoP o He o 3 Black edi ion came a ( ocus leng h: 2.77 mm, lens size: 14 mm) combined wi h an NDVI s ess came a (XNi eCanonELPH110NDVI, ocus leng h: 4.30 mm, lens size: 14 mm; LDP LLC L d.). The pilo a ea was 12 ha, alling wi hin he a ea o he LiDAR su ey, in he uni e si y campus (Uni e si y o Deb ecen). 2.2. Poin cloud p ocessing The LiDAR poin cloud was il e ed by Te aSolid’s Te aScan module in he Mic oS a ion en i onmen (h ps://www. e asolid.com/download/ scan.pd ) o e he whole a ea. TIN in e pola ion wi h na u al densi ica ion (Lin and Zhang, 2014) was ca ied ou o he sepa a ion o g ound poin s, hen e ical ou lying poin s we e emo ed using il e s. Following his, we il e ed ou he buildings wi h pa ame e ized algo i hms o Te aScan. A e wa ds, we ex ac ed he ec o ea u es o he buildings as an elemen o a semi-au oma ed oo iden i ica ion. We aimed o ind he op imal pa ame e s o ex ac he minimal oo -pa size o ob ain he mos accu a e and de ailed oo models. Besides, a digi al su ace model (DSM) was gene a ed om he poin cloud wi h 20 cm cell size (20 cm is he la ges easonable esolu ion which can be ob ained om he 12 poin s/m2 poin cloud) in o de o make a compa ison (Fig. 2). 2.3. Pho og amme ic analysis We applied Agiso Pho oscan P o 1.1.0. (Agiso LLC) o he pho og amme ic e alua ion o 188 images aken by he GoP o 3 came a. We used 16 GCP poin s (measu ed wi h a S onex S9 RTK 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 6 sys em) as ie poin s and he highes p ecision op ion was applied o p oduce he model. Ou lie poin s we e il e ed ou wi h he agg essi e dep h il e ing mode. The p ocedu e esul ed in a classic DSM wi h a densi y o 107 poin /m2, and also an o hopho og aph compiled om he ae ial pho os. The p ocedu e yielded a ue o hopho o (p oposed by Amha e al., 1998); acco dingly, bo h spa ial co e ages we e used in he analysis. Bo h he DSM and he o opho og aph had a esolu ion o 20 cm. 2.4. Analysis o digi al su ace models Image segmen a ion was ca ied ou on he DSMs, aspec and slope co e ages (bo h he la e we e de i ed om LiDAR and ae ial images) using eCogni ion De elope . Image segmen a ion is an objec -o ien ed analysis echnique ha akes in o accoun no only he pixel alues bu also he ex u e (Blaschke, 2010); hus, con a y o pixel-based classi ica ion, ―sal and peppe ‖ ype e o s can be a oided (Weih and Riggan, 2010). DSM, aspec and slope co e ages we e segmen ed using mul i esolu ion segmen a ion wi h ou di e en scale pa ame e s (L10, L50, L100, L200) and ound ha he p ocedu e wi h wo s eps om he supe -objec o he sub-objec using L200 and L100 alues ul illed he aims, i.e. sepa a ing he inpu as e co e ages in o he la ges homogenous segmen s (Kuma e al., 2014; Shao e al., 2014). This p ocedu e was epea ed wi h he use o he o hopho o. An XNi eCanon came a was used o p oduce a pseudo-colo o hopho o wi h blue-g een-in a ed bands, which was used o calcula e no malized di e ence ege a ion (NDVI, Rouse, 1973) alues. NDVI anges om -1 o 1 and alues below ze o indica e high e lec ance which is cha ac e is ic o ba e soil/ ock o an h opogenic objec s (e.g. buildings, oads e c.; Rouse e al., 1973). We applied a oo -mask compiled om NDVI alues (<0) and building heigh s (>3 m). 2.5. Digi al building models Finally, ou digi al ep esen a ions we e p oduced o he buildings. The ep esen a ion ha p o ided he oo plane geome y in he mos ealis ic way was he one om he poin cloud p ocessed in a CAD en i onmen (LPC), and a segmen ed digi al su ace model (SDSM) was p oduced wi h he segmen a ion me hod om he su ace model o he poin cloud. PDSM (segmen ed DSM) and OPDSM (common segmen a ion o he o hopho o and he DSM) we e p oduced om he su ace model o he pho og amme y app oach. 2.6. Valida ion We ob ained ield measu emen s wi h a S onex S9 RTK GPS pai o 50 buildings o check he con ou s o he buildings. Besides, 100 measu emen s we e ca ied ou o 100 buildings wi h a Leica Dis o D5 o con ol he building heigh s. Roo mean squa e e o (RMSE) and qua iles we e epo ed o he calcula ed di e ences be ween he LiDAR da a and he measu ed da a. Sola adia ion was alida ed by he compa ison o a building’s (s uden hos el) oo planes based on he bluep in and he LiDAR su ey. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 7 2.7. Calcula ion o he incoming sola i adia ion We il e ed ou hose segmen s o he oo planes ha we e sui ed o he ollowing condi ions (modi ying he app oach o Kassne e al., 2008): slope: 20-60o; aspec : 90-270o; a ea: >2 m2; compac ness: >0.3 (Fig. 2). #Fig. 2 app oxima ely he e In o de o e alua e he esul s o he modeling p ocedu e, a sample building was analyzed. The building analyzed was he s uden hos el building o he campus si e (N: 47° 33' 18.9000''; E: 21° 37' 24.1932''). The esul s o he model we e compa ed o a alida ed me hod based on a manual app oach. The da a ega ding he building’s oo we e acqui ed om wo sou ces. On he one hand, hey we e gene a ed au oma ically om he desc ibed p ocedu e and, on he o he hand, manually om a digi al map. We ob ained he ollowing da a: a ea, slope and azimu h. Roo a eas ha e e o s due o he su eying me hod used, i.e. he op iew o he oo s esul s in a smalle a ea o oo planes. We co ec ed he oo a eas wi h he cosine o he slope angles (F üh and Zakho , 2003); he co ec ion was made as in (Eq. 1). )cos( M h A A   (Eq. 1) whe e Ah is he oo a ea measu ed om abo e (ho izon al oo ) [m2]; A is he calcula ed a ea o he il ed oo [m2]. The numbe o PV panels by oo planes was de e mined manually, and au oma ically wi h a Py hon plugin de eloped o A cGIS. In bo h solu ions a 0.5 m bu e was omi ed om he calcula ion, he es o he oo plane was co e ed wi h PV panels. In he i s s ep all a ailable oo a eas we e co e ed wi h sola panels, in he second case he no h acing pa s o he oo we e le emp y. The incoming sola i adia ion was calcula ed o he geome ical da a acqui ed. Sola yield calcula ions we e pe o med wi h an aniso opic sola i adia ion model (Reindl e al., 1990). The di ec , di use and e lec ed adia ion componen s we e calcula ed acco ding o (Eq. 2-4) and he global adia ion was calcula ed as a sum o he componen s (Eq. 5).   b RDGI  (Eq. 2)                           bi M Mi RA ADD 2 sin1 2 1 cos11 3   (Eq. 3) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 8   2 1 cos1  M AGR  (Eq. 4) RDIG  (Eq. 5) whe e A is he albedo alue [-]; Ai is he aniso opy index; D is he di use adia ion on a ho izon al plane [kW/m2]; D is he di use adia ion on a il ed plane [kW/m2]; is he modula ing ac o o cloudiness; G is he global adia ion on a ho izon al plane [kW/m2]; G is he global adia ion on a il ed plane [kW/m2]; I is he beam adia ion on a il ed plane [kW/m2]; Rb is he a io o beam adia ion on a il ed plane o he beam adia ion on a ho izon al plane; R is he e lec ed adia ion on a il ed plane om he su oundings [kW/m2]; αM is he il angle o he il ed plane [°]. The me eo ological condi ions o he building si e we e desc ibed by he insola ion ime and we applied he Angs öm-P esco me hod (Paulescu e al., 2013) o de e mine he global i adia ion. Insola ion da a used in he calcula ions we e measu ed be ween 1981 and 2000. Calcula ions we e pe o med o h ee cases. Fi s ly, he incoming i adia ion was calcula ed o he en i e oo a ea o he building. In he second case we calcula ed he incoming sola i adia ion o he sola panels which we e alloca ed o he oo planes. In he hi d case he panels acing in a di ec ion anging om no heas o no hwes we e emo ed since hese loca ions lead o an economically non- iable solu ion. In he alida ion p ocess, he adi ional (manual) app oach based on he bluep in s was ega ded as p o iding he mos ealis ic da a. Following his we included all he egis e ed oo planes in he analysis and calcula ed he incoming sola ene gy o each o hem. We summa ized he oo planes o he 13 buildings which can be ound in he su eyed pa o he campus a ea. 2.8. S a is ical analysis We applied non-pa ame ic es s due o he non-no mal da a dis ibu ion o he oo a ea and sola i adia ion. Ou null hypo hesis (H0) was ha sola i adia ion de i ed om he wo su ace models had he same mean ank, and he al e na i e hypo hesis (H1) was ha he mean anks o he i adia ion alues we e di e en a he p<0.05 le el. Acco dingly, F iedman’s ANOVA and he Wilcoxon pai ed es we e applied in he hypo hesis es ing phase, combined wi h a Bon e oni co ec ion (Za , 1999). In he alida ion p ocess o he building geome y, we calcula ed Cohen’s Kappa (Kappa Index o Ag eemen , KIA; Cohen, 1960). KIA is a measu e o associa ion, equi es nominal alues, and indica es whe he he ag eemen occu s by chance ( anges a e be ween 0 and 1; 0 indica es ag eemen by chance, while 1 is o al ag eemen ; Rosen eld and Fi zpa ick-Lins, 1986). We compa ed he building con ou s based on GPS measu emen s and hose ha we e de i ed om he 3D poin cloud. Heigh s and a ea su aces o he sample building we e compa ed wi h he Wilcoxon pai ed es , and RMSE alues we e calcula ed. We e alua ed s a is ical es s wi h p alues (p<0.05), and also wi h e ec size as a s anda dized measu e o he di e ence be ween g oups (Cohen, 1992). 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 15 Ga ziolis, D., Ande sen, H-E., 2008. A Guide o LIDAR Da a Acquisi ion and P ocessing o he Fo es s o he Paci ic No hwes , USDA Fo es Se ice, Paci ic No hwes Resea ch S a ion, Gene al Technical Repo PNW-GTR-768, h p://www. s. ed.us/pnw/pubs/pnw_g 768.pd Ghu a , S., Székely, B., Ronca , A., P ei e , N., 2013. Landslide displacemen moni o ing using 3D Range Flow on ai bo ne and e es ial LiDAR da a. Remo e Sens. 5, 2720–2745. Heiskanen, E., Nissila, H., Lo io, R., 2015. Demons a ion buildings as p o ec ed spaces o clean ene gy solu ions – he case o sola building in eg a ion in Finland. J. Clean. P od. doi:10.1016/j.jclep o.2015.04.090 Hengl, T., G ohmann, C.H., Bi and, R.S., Con ad, O., Lobo, A., 2009. SAGA s GRASS: A Compa a i e Analysis o he Two Open Sou ce Desk op GIS o he Au oma ed Analysis o Ele a ion Da a, P oc. Geomo phome y 2009, Zu ich, Swi ze land, pp. 22–27. Ho ie ka, J., Kaňuk, J., 2009. Assessmen o pho o ol aic po en ial in u ban a eas using open–sou ce sola adia ion ools. Renew. Ene g. 34, 2206–2214. Ho ie ka, J., Šu i, M., 2002. The sola adia ion model o open sou ce GIS: implemen a ion and applica ion, P oc. Open-sou ce GIS-GRASS use s con e ence, T en o, I aly, pp. 1–19. Iqbal, M., 1983. An In oduc ion o Sola adia ion, Academic P ess Canada, On a io Izquie do, S., Rod igues, M., Fueyo, N., 2008. A me hod o es ima ing he geog aphical dis ibu ion o he a ailable oo su ace a ea o la ge-scale pho o ol aic ene gy-po en ial e alua ions. Sol. Ene g. 82, 929–939. Jakubiecz, J.A., Reinha , C.F., 2013. A me hod o p edic ing ci y-wide elec ici y gains om pho o ol aic panels based on LiDAR and GIS da a combined wi h hou ly Daysim simula ions. Sol. Ene g. 93, 127–143. Jochem, A., Hö le, B., Ru zinge , M., P e ie , N., 2009. Au oma ic Roo Plane De ec ion and Analysis in Ai bo ne LiDAR Poin Clouds o Sola Po en ial Assessmen . Senso s 9, 5241–5262. Kampou aki, M., Wood, G.A., B ewe , T.R., 2008. Oppo uni ies and limi a ions o objec –based image analysis o de ec ing u ban impe ious and ege a ed su aces using ue–colou ae ial pho og aphy, in Blaschke, T., Lang, S., Hay, G.J. (Eds), Objec -Based Image Analysis, Sp inge , Be lin–Heidelbe g, pp. 555–569. Kassne , R., Koppe, W., Schü enbe g, T., Ba e h, G., 2008. Analysis o he sola po en ial o oo s by using o icial LiDAR da a. In . A ch. Pho og amm. Remo e Sens. 37 Pa B4, 399–404. Ko pela, I., Ho i, A., Mo sdo , F., 2012. Unde s o y ees in ai bo ne LiDAR da a — Selec i e mapping due o ansmission losses and echo- igge ing mechanisms. Remo e Sens. En i on. 119, 92–104. Kozma, G., Molná , E., Czim e, K., Pénzes, J., 2013. Geog aphical aspec di usion o passi e houses. In . Re . Appl. Sci. Eng. 4, 151–156. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 16 Kuma M., Singh, R.K., Raju, P.L.N., K ishnamu hy, Y.V.N., 2014. Road Ne wo k Ex ac ion om High Resolu ion Mul ispec al Sa elli e Image y Based on Objec O ien ed Techniques. ISPRS Ann. Pho og amm. Remo e Sens. Spa ial In o . Sci. II–8, 107-110. Lázá , I., 2011. E ec s o clima e change on enewable ene gy sou ces (in Hunga ian), in Szabó V, Fazekas I. (Eds.), Kö nyeze uda os ene gia e melés- és elhasználás, MTA DAB Megújuló Ene ge ikai Munkabizo ság, Deb ecen, 2011, pp. 92–98. Lewis, N.S., 2007. Towa ds cos –e ec i e sola ene gy use. Science 315, 798–801. Lin, X., Zhang, J., 2014. Segmen a ion–Based Fil e ing o Ai bo ne LiDAR Poin Clouds by P og essi e Densi ica ion o Te ain Segmen s. Remo e Sens. 6, 1294–1326. Liu, X., 2008. Ai bo ne LiDAR o DEM gene a ion: some c i ical issues. P og. Phys. Geog. 32, 31– 49. Lukac, N., Seme, S., Zlaus, D., S umbe ge , G., Zalik, B., 2014. Buildings oo s pho o ol aic po en ial assessmen based on LiDAR (Ligh De ec ion And Ranging) da a. Ene gy 66 (2014) 598–609. Lukac, N., Zlaus, D., Seme, S., Zalik, B., S umbe ge , G., 2013. Ra ing o oo s’ su aces ega ding hei sola po en ial and sui abili y o PV sys ems, based on LiDAR da a. Applied. Ene g. 102, 803–812. May, C.N., To h, C.K., 2007. Poin posi ioning accu acy o ai bo ne LiDAR sys ems: a igo ous analysis. In . A ch. Pho og amm. Remo e Sens. 36(3/W49B), 107–111. Mongus, D., Lukac, N., Zalik, B., 2014. G ound and building ex ac ion om LiDAR da a based on di e en ial mo phological p o iles and locally i ed su aces, In . A ch. Pho og amm. Remo e Sens. 93, 145–156. Mücke, W., Deák, B., Sch oi , A., Hollaus, M., P ei e , N., 2013. De ec ion o allen ees in o es ed a eas using small oo p in ai bo ne lase scanning da a. Can. J. Remo e Sens. 39, 32-40. Nagy á adi, L., Gyenizse, P., Szebényi, A., 2011. Moni o ing he changes o a subu ban se lemen by emo e sensing. Ac a Geog aphica Deb ecina Landscape En . 5, 76–83. Nex, F., Remondino, F., Agugia o, G., de Filippi, R., Pole i, M., Fu lanello, C., Menegon, S., Dallago, G., Fon ana i, S., 2013. 3D sola web: A sola cadas e in he I alian alpine landscape. In . A ch. Pho og amm. Remo e Sens. XL-7/W2, 173–178. Nguyen, H.T., Pea ce, J.M., Ha ap, R., Ba be , G., 2012. The Applica ion o LiDAR o Assessmen o Roo op Sola Pho o ol aic Deploymen Po en ial in a Municipal Dis ic Uni . Senso s, 12, 4534–4558. Papa odi is, N., Polido i, L., 2004. O e iew o digi al su ace models, In Egels, Y and Kasse , M. (Eds.), Digi al Pho og amme y, Taylo and F ancis, London, pp. 159–163. Paulescu, M., Paulescu, E., G a ila, P., Badescu, V., 2013. Wea he modeling and o ecas ing o PV sys ems ope a ion, Sp inge , London P ies nall, G., Jaa a , J., Duncan, A., 2000. Ex ac ing u ban ea u es om LiDAR digi al su ace models. Compu . En . U ban Sys . 24, 31, 65–78. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 17 P o ic, D., Kiliba da, M., Vuce ic, I., Nes o o , I., 2002. 3D oo modelling o accu a e assessmen o sola po en ial, P oc. Eu oSun 2002 In . Con e ence, Bologna, I aly, 5 p. Quazi, A., Fayaz, H., Wadi, A., Raj, R.G., Rahim, N.A., Khan, W.A., 2015. The a i icial neu al ne wo k o sola adia ion p edic ion and designing sola sys ems: a sys ema ic li e a u e e iew. J. Clean. P od. 104, 1–12. Reindl, D.T., Beckman, W.A., Du ie, J.A., 1990. E alua ion o hou ly pi ched su ace adia ion models. Sola Ene gy 45, 9–17. Rosen eld, G.H., Fi zpa ick-Lins, K., 1986. A Coe icien o Ag eemen as a Measu e o Thema ic Classi ica ion Accu acy. Pho og amm. Eng. Remo e Sens. 52, 223–227. Rouse, J.W., Haas, R.H., Schell, J.A., Dee ing, D.W., 1973. Moni o ing ege a ion sys ems in he G ea Plains wi h ERTS, Thi d ERTS Symposium, NASA SP-351, Vol. 1. NASA, Washing on, DC, pp. 309–317. Shan, J., To h, C.K., 2008. Topog aphic Lase Scanning and Ranging: P inciples and P ocessing, CRC P ess, Boca Ra on Shao, P., Yang, G., Niu, X., Zhang, X., Zhan, F., Tang, T., 2014. In o ma ion Ex ac ion o High- Resolu ion Remo ely Sensed Image Based on Mul i esolu ion Segmen a ion. Sus ainabili y 6, 5300–5310. Sho is, M.R., Johns on, G.H.G., Po le , K., Lüp e , E., 2008. Pho og amme ic analysis o sola collec o s. In . A ch. Pho og amm. Remo e Sens. 37 Pa B5, 81–88. Smy h, M., 2012. Sola pho o ol aic ins alla ions in Ame ican and Eu opean winemaking acili ies. J. Clean. P od. 31, 22–29. S e ano ic, S., 2013. Op imiza ion o passi e sola design s a egies: A e iew. Renew. Sus . Ene g. Re . 25, 177–196. Weih, R.C., Riggan, N.D., 2010. Objec -Based Classi ica ion s. Pixel–Based Classi ica ion: Compa a i e Impo ance o Mul i-Resolu ion Image y. In . A ch. Pho og amm. Remo e Sens. (GEOBIA 2010), XXXVIII–4/C7 Yu, B., Liu, H., Wu, J., Hu, Y., Zhang, L., 2010. Au oma ed de i a ion o u ban building densi y in o ma ion using ai bo ne LIDAR da a and objec -based me hod, Landscape U ban Plan. 98, 210–219. Za , J.H., 1999. Bios a is ical Analysis. P en ice Hall, New Je sey, Zhou, Q., Neumann, U., 2013. Comple e esiden ial u ban a ea econs uc ion om dense ae ial LiDAR poin clouds. G aph. Models 75, 118–125. Zlinszky, A., Sch oi , A., Kania, A., Deák, B., Mücke, W., Vá i, Á., Székely, B., P ei e , N., 2014. Ca ego izing G assland Vege a ion wi h Full-Wa e o m Ai bo ne Lase Scanning: A Feasibili y S udy o De ec ing Na u a 2000 Habi a Types. Remo e Sens. 6, 8056–8087. Web e e ences 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 18 h ps://www.maxmax.com/Remo eSensingcame asi.h m, LDP-LLC L d. h ps://www. e asolid.com/download/ scan.pd , Te asolid, L d., Te a Scan Use ’s Guide, 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 19 Table 1. Di e ence o oo heigh s ( ela i e heigh s o LiDAR su ace sub ac ed om measu emen s) Roo ype Lowe qua ile Median Uppe qua ile RMSE Fla -0.01 0.17 0.45 0.30 Shed -0.18 0.03 0.20 0.31 Gable -0.02 0.11 0.29 0.34 Combina ion -0.25 0.01 0.23 0.43 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 20 Table 2. Numbe and a ea o oo planes in he campus a ea Me hod Numbe o de ec ed planes Sum o he a ea [m2] LPC 68 5432 SDSM 79 4893 PDSM 78 5197 OPDSM 52 5239 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 21 Table 3 S a is ical cha ac e is ics o oo planes o he hos el building, conside ing he ou calcula ion me hods (N=4 acco ding o he ou me hods; uni : m2) Roo plane IDs 16 20 23 24 31 Minimum 150.76 78.17 244.30 76.44 30.60 Maximum 182.54 100.47 284.89 96.97 70.32 Mean 160.19 90.53 257.15 88.12 52.08 S anda d e o 7.49 5.35 9.45 5.21 9.48 S anda d de ia ion 14.97 10.70 18.90 10.43 18.97 Median 153.72 91.74 249.69 89.53 53.70 25 pe cen ile 151.45 79.90 244.75 77.88 33.42 75 pe cen ile 175.39 99.95 277.00 96.94 69.13 Coe icien o a ia ion 9.35 11.82 7.35 11.83 36.42 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 22 Table 4. Compa ison o he incoming sola ene gy on he oo planes o he hos el building (p alues, Wilcoxon es wi h Bon e oni co ec ion; (LPC: LiDAR poin cloud p ocessing; SDSM: DSM om LiDAR + segmen a ion; PDSM: DSM om ae ial pho og aphs + segmen a ion; OPDSM: DSM om ae ial pho og aphs combined wi h o hopho o + segmen a ion) LPC SDSM PDSM SDSM 0.5455 PDSM 0.9091 1 OPDSM 1 1 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 23 Table 5. Incoming sola ene gy on he hos el building’s oo planes (Incoming sola i adia ion [kWh]; Roo plane IDs co espond o Fig. 4) Roo plane ID Roo plane su ace Calcula ed oo plane su ace Su ace o PV panels Calcula ed su ace o PV panels Su ace o PV panels excep no he n di ec ions Calcula ed su ace o PV panels excep no he n di ec ions 0 462203 438925 245800 256056 245800 256056 1 6622 0 0 2 7839 0 0 3 163404 109964 79207 33421 0 0 4 27739 25715 11044 7305 0 0 5 197275 54191 32885 19126 0 0 6 21756 26707 7650 7586 7650 7586 7 22204 24842 0 5565 0 0 8 74579 76779 26940 32458 26940 32458 9 117252 82262 37154 29033 37154 29033 10 30465 36849 2055 8254 2055 8254 11 428893 448378 228987 290448 228987 290448 12 62024 76814 21623 32472 21623 32472 13 158239 146923 72752 63890 72752 63890 14 127699 122964 56923 61851 0 0 15 275257 284305 142410 176715 142410 176715 16 37727 76814 7437 32472 7437 32472 17 276441 275325 139278 173186 139278 173186 18 158103 152502 73845 73886 73845 73886 19 68755 76779 25157 32458 25157 32458 To al 2710015 2551500 1211148 1336183 1031089 1208917 Ra io1 [%] 100 94.2 100 110.3 100 117.2 Ra io2 [%] 100 94.2 44.7 49.3 38.0 44.6 1sola ene gy calcula ed o a ea p o ided by manual me hod/sola ene gy calcula ed o a ea p o ided by au oma ed me hod 2sola ene gy calcula ed o sepa a e cases /sola ene gy calcula ed o he o al oo a ea p o ided by he au oma ed me hod 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 24 Table 6. Compa ison o di e en su eying echniques and possible e u ns in e ms o sola panels Calcula ions based on bluep in s D one su ey LiDAR su ey a ea/objec usually 1 house ~1-5 km2/day ~800-1000 km2/day absolu e cos low low High ela i e cos (p ice/building) high low Low cos /bene i can be inanced by a single household can be inanced by a local au ho i y o i m should be inanced by a p ojec und e u n soon soon he e is no di ec e u n IT in as uc u e equi emen low medium high HR expe ise equi emen medium high high Figu e 4 Click he e o download high esolu ion image Figu e 5 online Click he e o download high esolu ion image Figu e 5 p in ed Click he e o download high esolu ion image