Sources and types of uncertainties in the information on forest-related ecosystem services

Con en s lis s a ailable a ScienceDi ec Fo es Ecology and Managemen jou nal homepage: www.else ie .com/loca e/ o eco Sou ces and ypes o unce ain ies in he in o ma ion on o es - ela ed ecosys em se ices A. Kangas, K.T. Ko honen, T. Packalen, J. Vauhkonen ⁎ Na u al Resou ces Ins i u e Finland (Luke), Bioeconomy and En i onmen Uni , Yliopis oka u 6, 80100 Joensuu, Finland ARTICLE INFO Keywo ds: Fo es in en o y Su ey Indica o Mapping Unce ain y Value o in o ma ion ABSTRACT The concep o ecosys em se ices has gained impo ance in he o es managemen and o es policy p ocesses in ecen yea s. Ensu ing he sus ainable p o ision o ecosys em se ices equi es accu a e in o ma ion o he cu en p o ision and me hods o p edic ing he impac o impo an d i e s, such as changes in land co e and land use. In his e iew, we define he sou ces o unce ain ies in o es - ela ed ecosys em se ice assessmen s and discuss hei impo ance o he usabili y o he in o ma ion o diffe en pu poses. The unce ain ies a e due o e.g. a ia ion in he selec ed indica o s o he ecosys em se ices, lack o p ima y in o ma ion on hem, poo co ela ion wi h he da a used o mapping he ecosys em se ices o la ge scale and o p edic ing he impac s o human in e en ions. The unce ain ies can be andom o non- andom and hei assessmen is o en igno ed, especially in he case o he non- andom e o s. As a esul , diffe en assessmen s and subsequen decision ecommenda ions can be highly conflic ing. We do no expec ha he accu acies would significan ly imp o e in he sho e m. The bes way o p oceed is he e o e o assess he unce ain ies and ake hem in o accoun in he decision making o o es managemen . 1. Ecosys em se ices concep as a means o p omo e sus ainable o es managemen Ecosys em se ices (ES) ep esen he goods and se ices de i ed om he unc ions o ecosys ems u ilized by he humani y (Cos anza e al., 1997, 2017; C ossman e al., 2013). The concep o ecosys em se ices was o iginally designed as an educa ional and communica ion ool (Daily, 1997) o acknowledge ha human wellbeing is igh ly connec ed o he p o ision o hese se ices. Nowadays he concep o ecosys ems se ices is he main amewo k o en i onmen al policies and moni o ing (No gaa d, 2010). The Eu opean Commission empha- sizes he impo ance o accu a e in o ma ion on ecosys em se ices as he basis o he EU Biodi e si y S a egy o 2020 (Eu opean Commission, 2011). Land use change is he mos impo an d i e a - ec ing he ecosys ems (Dong e al., 2015). As a esul , ecosys em se - ices ha e been emphasized in na ional and egional land use policies and planning (e.g. F ank e al., 2015, Haakana e al., 2017, Tammi e al., 2017). Policy make s a e inc easingly ecognizing he po en ial o ecosys em se ice mapping in s a egic planning (Vo s ius and Sp ay, 2015). In he cascade model (Fig. 1), he ecosys em se ices a e add essed h ough he s uc u e and p ocess o he ecosys ems and hei unc- ioning, benefi s and alue ob ained om he used ecosys em se ices. The biophysical s uc u es and p ocesses c ea e he basis o he unc- ioning o he ecosys em and he unc ions c ea e he capaci y o p o- ide se ices. The capaci y o deli e a se ice exis s independen ly o whe he anyone wan s o needs ha se ice, bu ha capaci y becomes a se ice only i a beneficia y can be clea ly iden ified. The alue o he benefi can be defined as economic, social, heal h o in insic alue (Haines-Young and Po chin, 2010). The ecosys em se ices app oach has been c i icized, howe e , o aking a ully an h opocen ic iew and hiding he in insic alues o na u e (Fü s , 2015). Ecosys em se ices can be g ouped in many diffe en ways (e.g. de G oo e al., 2002, MA, 2005; TEEB, 2010). In Common In e na ional Classifica ion o Ecosys em Se ices (CICES), which is used in his e- iew, he ecosys em se ices a e di ided o p o isioning, egula ion and main enance, and cul u al (Haines-Young and Po chin, 2010). Se ices ha a e mos ele an om o es managemen poin o iew include p o isioning se ices such as imbe , be ies and mush ooms, game, eindee , and bioene gy; egula ing and main enance se ices such as clima e egula ion; and cul u al se ices such as ec ea ion and na u e ou ism. In he ollowing ex , we use he e m ‘ecosys em se - ices’ o e e o all possible o es - ela ed ecosys em se ices in gene al and diffe en ia e be ween hem only when i is ele an om he poin o iew o da a acquisi ion. In hose cases, we always spell ou he specific ecosys em se ices o s eps o he cascade model (Fig. 1) we a e h ps://doi.o g/10.1016/j. o eco.2018.05.056 Recei ed 2 Ma ch 2018; Recei ed in e ised o m 24 Ap il 2018; Accep ed 24 May 2018 ⁎ Co esponding au ho . E-mail add ess: ja i. auhkonen@luke.fi(J. Vauhkonen). Fo es Ecology and Managemen 427 (2018) 7–16 0378-1127/ © 2018 The Au ho s. Published by Else ie B.V. This is an open access a icle unde he CC BY license (h p://c ea i ecommons.o g/licenses/BY/4.0/). T e e ing o. The ecosys em se ices a e ope a ionalized h ough a selec ed se o indica o s (e.g. Mülle and Bu kha d, 2012). The pu pose o he in- dica o s is o suppo he managemen o ecosys ems and o commu- nica e on hei condi ion. Thus, hey simpli y he complexi y o he ecosys ems o manageable concep s. The se o ele an ecosys em se ices and hei indica o s a ies om egion o ano he . Fo ins ance, Mononen e al. (2016) and Hansen and Malmaeus (2016) ha e p e- sen ed a diffe en se o indica o s o Finland and Sweden, espec i ely, e en hough he wo coun ies esemble each o he e y closely in e ms o o es s uc u e. This a ia ion is one o he challenges when compa ing in e na ional ecosys em se ice assessmen s (Maes e al., 2012b, Mononen e al., 2016). One possible eason o he a ia ion is ha he alues o he expe s who ca y ou he selec ion o he c i e ia implici ly eflec o he selec ion o he indica o s (Menzel and Teng, 2010). Sus ainable managemen o na u al esou ces can be seen as max- imizing he social wel a e ob ainable om hem (Kan and Lee, 2004). Sus ainabili y means ha he u u e gene a ions can consume he eco- sys em se ices o he same ex en as he cu en one (e.g. No gaa d, 2010). Sus ainable p o ision o ecosys em se ices hus equi es a non- declining p o ision o all se ices o e an infini e pe iod in ime. Only changes ha a e ina guably sus ainable a e Pa e o imp o emen s, whe e he supply o ecosys em se ices imp o es wi h espec o one o mo e indica o s bu does no de e io a e wi h espec o any o he o he se ices. T ade-offs a e ine i ably ela ed o all o he changes in he cu en and u u e p o ision o he ecosys em se ices. The decision on whe he such changes a e sus ainable o no depends on he alues o he humans making he e alua ion (e.g. Fü s e al., 2010, Vo s ius and Sp ay, 2015, Ha ikainen e al., 2016). Including ecosys em se ices in o decision making is one way o s i e o sus ainable o es man- agemen (e.g. Ma inez-Ha ms e al., 2015). Acco ding o Meye and Schulz (2017), howe e , o es s a e cu en ly unde ep esen ed in he s udies ela ed o ecosys em se ices. Ensu ing ha ecosys em se ices a e p o ided sus ainably equi es in o ma ion o he cu en s a e. Sample-based in o ma ion is adequa e o make decisions on sus ainabili y a na ional and egional scales. Fo decision conce ning loca ions, such as whe e i is impo an o p o ec , es o e o imp o e ecosys ems o hei se ices, a map –i.e. spa ially explici in o ma ion –is equi ed. The maps can be used, o ins ance, o de ec ho spo s o coldspo s, i.e. a eas wi h high o low supply o eco- sys em se ices (e.g. Pagella and Sinclai , 2014). Co-occu ence o diffe en ecosys em se ices in an a ea implies syne gies and adeoffs (Maes e al., 2012a). The maps can also be used o de ec a eas whe e he supply o ecosys em se ices dec eases o inc eases due o changes in land use (Pagella and Sinclai , 2014); o iden i y p o iding and benefi ing a eas (Sy be and Wal z, 2012); and o communica e he e - ec s o policies o he land use and ecosys em se ice p o ision (Vo s ius and Sp ay, 2015). Real policy decisions equi e wo o mo e decision op ions o choose om and p edic ions o he u u e consequences o hese op ions (Co ona, 2016). To ensu e sus ainable p o ision o ecosys em se ices, in o ma ion needs o be a ailable and o sufficien quali y. We also need o ha e decision suppo ools o p edic ing he u u e de elop- men o he se ices affec ed by he decisions execu ed. We e iew he acquisi ion o p ima y da a (Sec ion 2) and mapping o ecosys em se ices (Sec ion 3), concen a ing on hose se ices e- le an om o es managemen poin o iew. We e iew he me ho- dology a ailable o assess he unce ain y in he ecosys em se ices da a (Sec ion 4) and no e ha in mos cases unce ain y assessmen is lacking o inadequa e. The sea ch o e e ences was ca ied ou using Web o science on 15 No embe 2016. We used one keywo d desc ibing he unce ain y assessmen (e.g. “e o ”,“unce ain y”,“ alida ion”, “e alua ion”), one keywo d desc ibing he da a collec ion and usage (e.g. “mapping”,“in en o y”,“da a acquisi ion”) and as he las key- wo d “ecosys em se ices”. While he inclusion o all esul ing a icles o his e iew is by no means exhaus i e (Web o science ga e 22 652 hi s o he keywo d “ecosys em se ices”), we specifically a emp ed o ocus on a icles ha acknowledged unce ain ies. Finally, we discuss ou findings on he gap be ween he in o ma ion demand and supply in e ms o con en s, scale, accu acy and unce ain y assessmen wi h e- spec o decision making. 2. Acqui ing ecosys em se ices da a 2.1. Indica o s o ecosys em se ices The da a acquisi ion o ecosys em se ices is ope a ionalized h ough a se o indica o s ha can be assessed. P ima y da a a e needed o he indica o s o he s uc u e, unc ion, benefi s, and alue in he cascade model (Fig. 1). Fo ins ance, he habi a a ea (ha), p o- duc ion (kg/ha/A), yield (kg), and mone a y alue (€) could se e as he indica o s o s uc u e, unc ion, benefi s, and alue, espec i ely, i o es be ies and mush ooms we e conside ed as an example se ice (Mononen e al., 2016). Selec ing a good se o indica o s is impo an , as hose a y in quali y o decision making. Au inen e al. (2007) e alua ed indica o s o biodi e si y using se e al c i e ia: ele ance, impac , effec i eness, cos -effec i eness, accep abili y, incen i e alue, anspa ency and op- po uni ies o pa icipa ion, equi y, flexibili y, p edic abili y, and Biophysical s uc u e o p ocess Func ion Se ice Bene i Value Fig. 1. The cascade model (modified om Haines-Young and Po chin, 2010). A. Kangas e al. Fo es Ecology and Managemen 427 (2018) 7–16 8 pe manence. Fo es in en o y expe s defined 41 a iables measu able a field ha could be used as indica o s o biodi e si y (Chi ici e al., 2012). These indica o s we e anked om low o high acco ding o hei easibili y and impo ance and hose ha had bo h high easibili y and impo ance we e ound o be he mos use ul. P ima y da a o many o es ela ed indica o s o s uc u e and unc ion can be ob ained om field plo s o o es in en o y (Table 1). I p ima y da a a e a ailable, hose can be used o o m (s a is ical o p ocess-based) models. Models a e o impo ance, as hey can be used o mapping he se ices and especially o assessing he impac o u- u e changes. P ima y da a o benefi s and alue equi e su eys in- ol ing he human beneficia ies. These da a can also be p esen ed as a (benefi ans e ) model. As i is o en difficul o ob ain he equi ed da a, mos o he se ices lack p ima y in o ma ion (Eigenb od e al., 2010). The only op ion in ha case is o assess he ecosys em se ices by means o expe judgmen , e en hough such da a a e highly sub- jec i e and difficul o alida e (Table 1). 2.2. Field da a In many coun ies a sample-based na ional o es in en o y (NFI) has been es ablished o moni o ing o es s (Chi ici e al., 2011). The main benefi o he sample-based da a acquisi ion is he possibili y o assess he unce ain y included (Table 1). The NFIs ha e adi ionally measu ed indica o s sui able o imbe p oduc ion. Fo ins ance, he NFIs p o ide es ima es o he a ea o managed o es s (ha) and he inc emen o g owing s ock (m 3 /ha), which we e iden ified as in- dica o s o wood and bioene gy p o ision (Mononen e al., 2016). Ca bon- ela ed indica o s a e also eadily a ailable, because hose can be calcula ed om he g owing s ock in o ma ion. Cu en ly, he NFIs also measu e o he han imbe - ela ed in- dica o s, such as hose ela ed o deadwood, e ical s uc u e, im- po an habi a s, and ee species mix u es. Fo ins ance, he global Fo es Resou ces Assessmen (FRA) s i es o collec ing in o ma ion on he p o ision o ecosys em se ices globally (Miu a e al., 2015). Howe e , ob aining p ima y in o ma ion on ecosys em se ices o he han imbe is s ill a challenge. Sample-based in en o y da a a e ac- cu a e and, in p inciple, could include many o he indica o s o s uc- u e and non-wood p oduc ion. In p ac ice, he collec ion o such da a may be esou ce in ensi e (Maes e al., 2012b) and he e o e p ac ically in easible. A smalle sample ailo ed specifically o he gi en eco- sys em se ice(s) is o en needed o p o ide adequa e s a is ics con- ce ning, o ins ance, he p oduc ion o o es be ies o game habi a s (e.g. Ku ki e al., 2000, Tu iainen e al., 2011, Melin e al., 2016). To enhance he a ailabili y o p ima y da a, many diffe en en- i onmen al moni o ing schemes ha e been and a e being de eloped (see Geijzendo ffe and Roche, 2013 o a e iew). Mos o hem a e ela ed o a subse o possible ecosys em se ices. Se e al moni o ing schemes ha e been de eloped o habi a o species di e si y (e.g. S åhl e al., 2011, Co ona e al., 2011, Chi ici e al., 2012), while he poli ical ocus o en i onmen al p o ec ion has ecen ly been based mo e on main aining o es o ing ecosys em se ices (Geijzendo ffe and Roche, 2013). The ecosys ems p o ide se ices in diffe en scales, which need o be accoun ed o also in he moni o ing schemes (Geijzendo ffe and Roche, 2013). The scales can be local, egional, global, o ela ed o di ec ional flow o dis ance (e.g. Cos anza, 2008). Each indica o should ob iously be moni o ed on he ele an spa ial scale. Some in- dica o s like landscape pa e ns equi e obse a ions om wide geo- g aphical ex en s han sample plo s. The pa e ns can be desc ibed using landscape me ics such as he numbe o pa ches, mean pa ch size, o shape and di e si y indices (F ank e al., 2012; Ramezani and Ramezani, 2015). Ano he simila example equi ing landscape mea- su es a e he ec ea ional and aes he ic alues (Pukkala e al., 1995; F ank e al., 2012). 2.3. Su ey da a The benefi s and alue, i.e. he up ake o se ices, a e ypically difficul o assess in he field. Some indica o s, such as he deg ee o g azing by game, may be moni o ed wi h sample-based in en o ies, bu su eys o in e iews o he beneficia ies a e o en needed. I i was possible o in e iew he beneficia ies li ing close o he field plo s, he su eys could be di ec ly linked wi h he field measu emen s, bu usually he su eys need o be ca ied ou as a sepa a e effo . T ade s a is ics can also p o ide use ul in o ma ion. Fo he p o i- sion o mos o he ecosys em se ices, ele an s a is ical in o ma ion exis s a he na ional le el. Fo ins ance, he amoun and alue o be ies, mush ooms o o he non-wood o es p oduc s ha ing mone- a y alue a e assessed in se e al coun ies. The Food and Ag icul u e O ganiza ion o he Uni ed Na ions (FAO), Cen e o In e na ional Fo es y Resea ch (CIFOR), In e na ional Fo es y Resou ces and Ins i u ions Resea ch Ne wo k (IFRI) and Wo ld Bank ha e e en pub- lished good p ac ice guidance o collec ing socio-economic da a wi h household su eys (FAO, CIFOR, IFRI and Wo ld Bank, 2016). Specific alua ion s udies a e needed o ecosys em se ices which do no ha e ma ke alue. In gene al, alua ing non-ma ke goods can be ca ied ou using ei he p e e ences e ealed in eal decision Table 1 Da a acquisi ion o he diffe en le els o he cascade model o ecosys em se ices. Me hod S eps o he cascade model in ol ed Ad an ages Disad an ages Field da a - S uc u e - Func ion (- Use) - Objec i e (unbiased) in o ma ion - Unce ain y assessmen s a ailable - Cos ly - Labo ious and imp ac ical o some indica o s - Unobse able o some indica o s Models - S uc u e - Func ion (- Use) - Easy o u ilize in mapping he ecosys em se ices - Cheape han p ima y da a - Unce ain y assessmen s possible - Possible o assess he impac s o changes - Requi es p ima y da a o es ima ion Su ey da a - Use - Benefi - Value - Unce ain y assessmen s possible - P one o subjec i i y - Regional a e ages difficul o u ilize in mapping Benefi ans e - Benefi - Value - Cheape han o iginal alua ion s udies - P one o bias due o diffe ences be ween egions and con ex - Valida ion and unce ain y assessmen difficul Expe judgmen - S uc u e - Func ion - Use - Benefi - Value - Easy o u ilize in mapping he ecosys em se ices - Cheape han p ima y da a - Possible o assess he impac s o changes - Highly subjec i e - P one o bias - Valida ion and unce ain y assessmen difficul A. Kangas e al. Fo es Ecology and Managemen 427 (2018) 7–16 9 si ua ions, o hose s a ed in hypo he ical decision con ex s (Lou ie e e al., 2000). In many cases, howe e , he e a e no o he possibili ies excep elying on he s a ed p e e ences. The s a ed p e e ences me hods include e.g. con ingen alua ion o choice modeling (Lou ie e e al., 2000). 2.4. Benefi ans e da a Mos o he e alua ions a e case s udies, which in ol e one o a couple o se ices in one egion (e.g. Ho ne e al., 2005; Japelj e al., 2016). When he alue o an ecosys em se ice is e alua ed in one o mo e egions, he benefi s can be ans e ed also o o he a eas o ob ain na ional o global le el e alua ions (benefi ans e ). This is o en he only op ion because o budge cons ain s. The main difficul y in he benefi ans e is he ques ion o how well he findings om one egion can be gene alized o he o he s (Table 1). Fo ins ance, he es ima e o Cos anza e al. (1997) on he o al alue o he global ecosys em se ices o 33 illion US dolla s annually has been widely c i icized o he lack o a en ion o he con ex and consequen ly in alid benefi ans e (Bul e and an Koo en, 2000). Howe e , such es ima es po en ially p o ide easonable app oxima ions and he main goal in he benefi ans e app oach may no be accu acy, bu use ulness (Richa dson e al., 2015). The ans e abili y depends on he con ex and he spa ial scale o he o iginal s udies (Richa dson e al., 2015). Fo ins ance, i is possible o compu e a uni alue o one o iginal e alua ion s udy ha bes ma ches he cha ac e is ics o he si e o in e es and o use ha alue o he gene aliza ion. Howe e , a me a-analysis modelling he eco- sys em se ice alues obse ed in he exis ing s udies as a unc ion o he cha ac e is ics o he s udy si es could be mo e use ul. Such me a- analysis models a e a ailable o many ecosys em se ices (Nelson and Kennedy, 2009). 2.5. Expe judgmen E en i we seldom ha e sui able p ima y field o su ey da a o ecosys em se ices, i does no mean ha we ha e no in o ma ion a all. Ins ead, expe opinion may p o ide he bes op ion a ailable o as- sessing he ecosys em se ices in ques ion o some cases. Expe opinions a e ypically collec ed in he o m o a model, meaning ha he expe s a e asked o assess e.g. be y p oduc ion po- en ial, biodi e si y, ec ea ion alue o habi a sui abili y in a gi en o es si e ype o o es age class, and he esul is p esen ed as a unc ion o he o es cha ac e is ics. The mos popula inpu da a a e p obably he land use / land co e (LU/LC) ma ices (Balz e e al., 2015). The expe opinions a e ypically assessed using quali a i e echniques, such as on a e bal scale om e y low o e y high supply (Koschke e al., 2012, Fü s e al., 2013, Jacobs e al., 2015). I is also possible o use quan i a i e echniques such as pai wise compa isons, which enable p oducing s a is ical models om he judgmen s (Kangas e al., 1993, Ihalainen e al., 2002, Leskinen e al., 2003). In o ma ion based on expe opinions is a less cos ly al e na i e compa ed o acqui ing p ima y field o su ey da a on any ecosys em se ice. The d awback is ha he quali y o such da a canno be gua - an eed. La ge disag eemen s be ween he expe s imply poo in o ma- ion quali y (Kangas e al., 1998). The esul s depend on who, how many, and wha kinds o expe s a e used. Fo ins ance, bo h scien is s and ou is s could be used as expe s in ques ions conce ning ec ea- ion. Techniques such as Delphi ounds ha e been used in o de o in- c ease he consensus be ween he expe s (Scolozzi e al., 2012), bu new ounds o assessmen s may also inc ease he a ia ion be ween he expe s (Kleemann e al., 2017). The judgmen s o expe s could be weigh ed based on hei expe ise o he mos unce ain expe s could be le ou (Leskinen e al., 2003). Howe e , e en when all expe s ag ee on he p o ision o a gi en ecosys em se ice, hey may be col- lec i ely w ong o he expe disag eeing wi h all he o he s could be he one wi h he bes knowledge. This means ha expe opinion is e y suscep ible o bias (Table 1). 2.6. Assessing he impac o changes I is impo an o model he se ices based on in o ma ion ha al- lows o p edic ing hei de elopmen in ime and unde human in e - en ions –i.e., no jus o mapping he cu en s a e. Land use change is he mos impo an d i e in he p o ision o ecosys em se ices and o p edic consequences o land use changes, he p o ision o ecosys em se ices needs o be modelled as a unc ion o he land use. In he simples case, he expe s assess he p o ision o ecosys em se ices di ec ly as a unc ion o LU/LC classes in wo ime poin s and he im- pac s o land use changes can be calcula ed as he sub ac ion o he be o e and a e maps (e.g. Fü s e al., 2013, Kleeman e al., 2017). The landscape pa e n also plays a majo ole in many cases, which means ha he models need o accoun o he p ope ies o he neighbo hood o pixels o s ands ins ead o jus indi iduals. A modelling app oach can be used o cap u e he con inuum o ecosys em se ices (K ishnaswamy e al., 2009). I is possible o model he p o ision o ecosys em se ices using ei he inpu a iables fixed in ime, such as opog aphy, o hose changing o e ime, such as canopy co e , ee species o biomass (e.g. And ew e al., 2014, Ma inez- Ha ms e al., 2016). Many ypes o modelling echniques, such as niche- based, ai -based o ull p ocess models based on ac ual causal e- la ionships (La o el e al., 2017), ha e been used. The effec o land use change on he p o ision o ecosys em se ices can be p edic ed by fi s assessing he effec o he land use change on he inpu alues. In a case o o es esou ces, ma ix-based o es scena io models can be u ilized (e.g. Vauhkonen and Packalen, 2017). Thus, i is possible o p edic he consequences o diffe en policies on ecosys em se ices by fi s p e- dic ing he u u e de elopmen o o es esou ces unde hese policies. 3. Mapping o o es ecosys ems Mapping ecosys em se ices means ha he in o ma ion a ailable is gene alized o e an a ea based on emo e sensing (RS) o o he in- o ma ion p esen ed in a Geog aphical In o ma ion Sys em (GIS). The simples –bu o en also he leas accu a e –way is o u ilize p e iously collec ed and in e p e ed RS ma e ial, such as LU/LC maps p o ided by CORINE. I is possible o map all aspec s o he cascade model (Fig. 1), in p inciple. In p ac ice, howe e , pa ame e s ela ed o in si u p o- duc ion like ca bon seques a ion can be mapped ai ly easily, whe eas hose ela ed o benefi s like clima e egula ion a e used globally and hus mapping hem is mo e challenging (Pagella and Sinclai , 2014). As he e is a lack o p ima y in o ma ion conce ning mos o he ecosys em se ices, he mos easily a ailable map da a sou ce o eco- sys em se ices is one based on expe assessmen and LU/LC classes (Fig. 2,Fü s e al., 2010, 2013; Jacobs e al., 2015). The esul ing map is ine i ably hea ily simplified. Fo ins ance, Me zge e al. (2006) conside ed all non-u ban lands o ha e an equal po en ial o ec ea ion and c opland and u ban a eas o ha e no ec ea ional po en ial a all. Such maps also concen a e on he land use composi ion, bu igno e land use configu a ion and in ensi y (La o el e al., 2017). Mo e de ailed analyses equi e mo e de ailed in o ma ion, such as soil maps, ege a ion o bio ome maps o addi ional RS da a (e.g. Vihe aa a e al., 2015; Kaise e al., 2013). Fo ins ance, Vauhkonen and Ruo salainen (2017) used expe judgemen based models o gen- e alize he ecosys em se ice p edic ions de i ed om o es esou ce maps (e.g. Tomppo e al., 2008). Fo es esou ce maps may o en in- clude mo e in o ma ion han he LU/LC maps: beside he main ee species, also es ima es o mean age, mean size and o es s uc u e a e included. I p ima y da a a e a ailable, he selec ed a ibu es can be gen- e alized o la ge a eas by calcula ing he a e age es ima es o he LU/ LC classes ins ead o using expe assessmen s (e.g. Eigenb od e al., A. Kangas e al. Fo es Ecology and Managemen 427 (2018) 7–16 10 2010). The use o a e age wi hin-class es ima es may o e -simpli y he si ua ion, as o es s belonging o he same LU/LC class may ha e a e y diffe en po en ial o p o ide ecosys em se ices acco ding o s and mean age, o example. P ima y da a can also be used in physical o s a is ical models, whe e he p o ision o ecosys em se ices is p e- dic ed o each pixel o s and using a ailable GIS and RS in o ma ion (e.g. And ew e al., 2014; Ma inez-Ha ms e al., 2016). Such models can be di ec ly u ilized in mapping. The accu acy o he p oduced map imp o es by a s onge e- la ionship be ween he inpu a iables and he ecosys em se ices in ques ion. Thus, h ee-dimensional (3D) emo e sensing p oduc s such as ai bo ne lase scanning (ALS) da a allow p oducing mos de ailed maps o o es ecosys em se ices (Da ies and Asne , 2014; Vihe aa a e al., 2015). Un o una ely, he a ailabili y o 3D emo e sensing da a seldom suppo s la ge-scale analyses in he same way as he a ailabili y o sa elli e images. 4. Unce ain ies in he ecosys em se ices mapping 4.1. Assessing he unce ain y The e may be unce ain ies due o indica o s chosen o desc ibe he ecosys em se ices; GIS da a used o mapping he se ices; ela ion- ships be ween he GIS da a and he indica o s; and due o he e- la ionships be ween human in e en ions and he ecosys em se ices (Table 2). Un o una ely, he accu acy assessmen p o ocols used wi h map p oduc s do no clea ly in o m on i a map is use ul o no (Ayanu e al., 2012; McRobe s, 2011; Pagella and Sinclai , 2014). The fi s and las ca ego ies o Table 2 apply o decision making in ol ing ecosys em se ices o e all, and he o he wo ca ego ies o decisions conce ning loca ion-specific decision making. The unce ain ies can be andom o non- andom (o s uc u al, see Boi hias e al., 2016). Non- andom e o s can be due o e.g. incohe en defini ions o he ecosys em se ices, selec ion o indica o s, spa ial scale, missing explana o y a iables, o analysis echniques (Ma ínez- Ha ms and Bal ane a, 2012; Schulp e al., 2014; Ba edo e al., 2015). Non- andom e o s a e also ine i able whene e expe judgmen is used o assess he ecosys em se ice, e.g. due o he selec ion o he expe s. In he alua ion o ecosys em se ices, he numbe o se ices conside ed and he numbe o benefi s e alua ed o each se ice may in oduce non- andom e o s (Boi hias e al., 2016). Random e o s, on he o he hand, come om measu emen e o s, sampling e o s and esidual e o s o models. The fi s sou ce o unce ain y in he ecosys em se ices es ima es is he use o p ima y da a (Table 3). In he case o andom e o s, basic s a is ical analyses o measu emen and sampling e o s a e alid as- sessmen me hods applied o indica o s such as imbe and ca bon p oduc ion. Non- andom e o s a e much mo e difficul o assess, bu empi ical alida ion s udies could be used o assess he effec o selec ed indica o s o he a ia ions o indica o s be ween egions. Such s udies could also be used o analyze he effec s o missing indica o s. The agueness o defini ions could be desc ibed using e.g. uzzy numbe s o se s, bu om decision making poin o iew he bes app oach would be o efine he defini ions. On he o he hand, he a ia ion o he measu es used o desc ibe he ES in ques ion could in i sel desc ibe he agueness o he defini ions (Boe ema e al., 2017). Unce ain y in he ela ionship o he chosen indica o and he ecosys em se ice in ques ion is mo e complica ed. To be use ul, he indica o should ha e a high posi i e co ela ion wi h he se ice in ques ion. Wi h empi ical alida ion da a, i is possible o analyze he deg ee by which he chosen indica o s succeed. The second sou ce o unce ain y is he ac ual GIS da a used o mapping, which can con ain andom e o s due o misclassifica ions o he LU/LC classes, o ins ance. Such unce ain y can be assessed using Mon e Ca lo simula ion (e.g. Dong e al., 2015; Foody, 2015). The andom e o s can also be p edic ion e o s, esul ing om model-based gene aliza ions o e.g. soil samples ac oss egions. The non- andom e o s o GIS da a may be ela ed o a ying amoun s o LU/LC classes o a ying esolu ion in he RS ma e ial be ween diffe en s udies. As- sessing such unce ain ies in he con ex o ecosys em mapping has la gely been ca ied ou by compa ing es ima es o ecosys em se ices ob ained om diffe en land use maps (e.g. Bení ez e al., 2007; Schulp and Alkemade, 2011; Schulp e al., 2014; Van de Bies e al., 2015). Fo ins ance, Schulp e al. (2014) compa ed ou maps a he Eu opean le el. As a esul o he unce ain ies, he p oduced maps s ongly dis- ag eed on he po en ial o clima e egula ion in Sweden and Finland despi e he high p opo ion o o es s in hese coun ies. In such a compa ison, he p oblem is ha none o he maps is necessa ily he ue one and he accu acy o each map emains unknown. Ano he p oblem is ha he a ying sou ces o e o a e con ounded and i is ypically no possible o sepa a e he effec s o jus one sou ce. The hi d sou ce o e o is he ela ionship be ween he indica o s Fig. 2. Ecosys em se ice mapping based on LU/LC map and expe es ima es o ecosys em se ices p o ision in he classes (Jacobs e al., 2015). A. Kangas e al. Fo es Ecology and Managemen 427 (2018) 7–16 11 o ecosys em se ices and he GIS da a used o he mapping. The andom e o s a e due o he p edic ion models be ween inpu da a and he p ima y da a on indica o s o he ecosys em se ices. Non- andom e o s a e due o, o ins ance, using expe judgmen o modelling. Assessing he unce ain ies in expe assessmen s o ecosys em se ices is no sel -e iden , bu one way o assess he unce ain ies is o analyze he disag eemen s be ween he expe s using s a is ical me hods (Alho and Kangas, 1997,Table 3). E en i p ima y da a a e a ailable and he ela ionship is modelled wi h s a is ical models, non- andom e o s due o missing explana o y a iables o o he model misspecifica ions is possible, e en likely. Tha sou ce o e o can be assessed by compa ing diffe en models in he p edic ions. The e o s in he ela ionships be ween a iables can be assessed using empi ical e alua ion. Land co e -based p oxies a e known o be poo p edic o s, because hey ha e a poo co ela ion wi h he ac ual ecosys em se ices (Eigenb od e al., 2010; Geijzendo ffe and Roche, 2013). Fo ins ance, Eigenb od e al. (2010) gene alized sample-based da a o h ee ecosys em se ices o 10 × 10 km squa es and compa ed he esul ing maps o hose p oduced as a e ages o he land co e classes. Bo h he app oaches we e hus based on he same da a, bu classified ei he wi h o wi hou using he land co e classes. When he a eas wi h highes po en ial o p o ision o ecosys em se ices we e sea ched o , he cong uence be ween hese wo maps was no e y high. Fo ins ance, when looking a ho spo s ( he bes 10% o he a ea), he wo maps we e o e lapping in 23% o cases o biodi e si y, 17% o ec ea ion and 62% o ca bon s o age. A model ha uses local biophysical in o ma ion as explana o y a iables in addi ion o land use can p oduce much mo e accu a e in- o ma ion. Fo ins ance, in he compa ison ca ied ou by Ma inez- Ha ms e al. (2016), a model could p edic 60% o he a ia ion o he fi ewood, compa ed o he 15% o he look-up- able. Wi h his kind o model, i is also possible o analyze he impo ance o he diffe en componen s o unce ain y (e.g. Li ne and S o ay, 2011). The las sou ce o e o is he ela ionship be ween he ecosys em se ice p oduc ion and human in e en ions. The assessmen me hods o his sou ce o e o a e simila as in he p e ious case. The unce ain y assessmen s should be spa ially explici ins ead o agg ega ed in applica ions aiming o use he ecosys em maps o loca e he a eas o ho spo s. One p oblem is ha he e o s may be e y he - e ogeneously dis ibu ed o e he landscape o be ween he diffe en ecosys em se ices conside ed (Dong e al., 2015). Fo ins ance, G ê - Regamey e al. (2013) used a Bayesian ne wo k o a spa ially explici unce ain y assessmen . I is also possible o map he obus ness and sensi i i y o he expe opinions (Ligmann-Zielinska and Jankowski, 2014; Ga cía Má quez e al., 2017). The unce ain ies om he diffe en sou ces cumula e in he ana- lysis (e.g. Ba on e al., 2018). Fo ins ance, i ecosys em s uc u e is used as a basis o assessing ecosys em unc ion(s), which is u he used as a basis o p edic ing he benefi (s) and alue, he e o s ela ed o he diffe en s eps o he cascade model cumula e. This cumula ing effec hus mainly affec s he p ima y da a conce ning he benefi (s) and alue. Howe e , he cumula ing effec also applies o cases whe e decisions a e ca ied ou based on locally p edic ed impac s o human in e en ions, which in u n a e based on he p edic ed p o ision o he ecosys em se ices. In such cases, all sou ces o e o s in oduced in Table 2apply, bu he use s o he in o ma ion should especially be cau ioned on he cumula ing effec o he non- andom e o s. 4.2. The effec o unce ain y in decision making The unce ain ies should be quan ified and communica ed o deci- sion make s in o de o make in o med decisions. E o s can lead o biased alua ion o he ecosys em se ices and, consequen ly, e oneous decisions (Foody, 2015). The dis inc ion be ween andom and non- Table 2 Sou ces o e o s in he mapping. Sou ce o e o Random e o Non- andom e o s Indica o s o ecosys em se ices •Measu emen e o s o indica o a iables •Sampling e o s •Poo co ela ion be ween he indica o and he ES •Ambiguous defini ions o ES •Selec ion o indica o s •Va ia ion in indica o s used o a gi en ES ac oss egions and s udies •Misspecified ela ionship be ween he indica o and he ES GIS da a (used as a basis o mapping) •Misclassifica ion e o s •Numbe o LU/LC classes •Spa ial esolu ion o he da a •Missing da a Rela ionship be ween GIS da a and indica o s o ES •Residual e o s o models (poo co ela ion) •E o s in model coefficien s •Selec ion o expe s •Bias in expe judgmen •Misspecified (s a is ical o expe ) models •Missing explana o y a iables •Regional bias (in benefi ans e and o he models) Impac o human in e en ions on ES •Residual e o s o models (poo co ela ion) •E o s in model coefficien s •Selec ion o expe s •Bias in expe judgmen •Misspecified (s a is ical o expe ) models •Missing explana o y a iables •Regional bias Table 3 Me hods sui able o assessing diffe en ypes o e o s. Sou ce o e o Assessmen o andom e o s Assessmen o non- andom e o s Indica o s o ecosys em se ices •S a is ical analysis o sampling and measu emen e o s •Empi ical alida ion o co ela ion be ween he se ice and he indica o •Fuzzy logic •Empi ical alida ion o selec ion o indica o s used GIS da a (used as a basis o mapping) •S a is ical analysis o model e o s •Mon e Ca lo simula ion •Compa ison o maps p oduced Rela ionship be ween GIS da a and indica o s o ESImpac o human in e en ions •S a is ical analysis o model e o s •Mon e Ca lo simula ion •Empi ical alida ion o co ela ion be ween he se ice and he indica o •Va ia ion among expe s •Va ia ion among p edic ion models •Empi ical alida ion o model misspecifica ion A. Kangas e al. Fo es Ecology and Managemen 427 (2018) 7–16 12 andom e o s (Table 4) is especially impo an when accoun ing o he unce ain ies. Fo all andom e o ypes, i is possible o simula e se e al ou comes o a gi en decision and use hese simula ions ei he o sensi i i y o scena io analyses. When scena ios can be simula ed, i is also possible o use s ochas ic p og amming o Bayesian decision analysis o he decision making. S ochas ic p og amming makes sense when he e a e cons ain s in he analysis (e.g. Ey indson and Kangas, 2014; Ha ikainen e al., 2016). Bayesian analysis is especially use ul i he p oblem can be desc ibed using a decision ee wi h only a ew possible ou comes o each decision (Smi h, 2012). Scena io analyses wi h scena ios based on he assessmen s o di - e en expe s o diffe en models a e also one possibili y o assess he non- andom e o s, such as unce ain ies in he expe judgmen s. In he assessmen s o he effec s o human in e en ions, i is possible o u ilize adap i e managemen , i.e. lea n om expe iences and adap he managemen acco dingly (Bi ge e al., 2016). Howe e , adap i e managemen based on wo-s age s ochas ic p og amming is also a possibili y, when addi ional in o ma ion is acqui ed (Kangas e al., 2015). Vague defini ions a e he mos challenging non- andom e o , bu decision suppo me hods such as uzzy op imiza ion and uzzy decision analysis (Kangas e al., 2015) can s ill be u ilized when all unce ain ies a e desc ibed using uzzy app oach a he han s a is ical analysis. Diffe en ypes o unce ain ies may be impo an o he diffe en ypes o decisions. The e o e, i may no be necessa y o accoun o all ypes o e o s in e e y decision p oblem. As he effec s o hese e o s a e e y poo ly known, i is difficul o assess he impo ance o each e o sou ce on he decisions. Howe e , based on he map ypologies o Pagella and Sinclai (2014), we ha e assessed he impo ance o ac- coun ing o e o s o some applica ions. Table 5 p esen s he esul s o his analysis using h ee classes (low, medium, o high), which indica e how impo an i is o include unce ain y analyses in he example applica ions o hose o p oduce u h ul ou comes. Random e o s in p ima y da a o he indica o s and GIS da a likely ha e a small effec on ade-offanalyses be ween diffe en ecosys em se ices. In con a y, he mos impo an sou ce o e o o he ade- offanalyses is he ela ionship be ween GIS da a and he indica o s o he ecosys em se ice. The non- andom e o s a e especially impo an . I he ela ionship is biased, i is e iden ha he ela ionship be ween wo ecosys em se ices analyzed o ade-offs will be biased as well (Table 5). Likewise, ague defini ions o ecosys em se ices in he fi s place a e also likely e y impo an . Howe e , he bes way o a oid his sou ce o unce ain y is using unambiguous defini ions. On he o he hand, he accu acy o he loca ions is impo an , i we wish o find ho o cold spo s o he conside ed ecosys em se ices. In such case, also he andom e o s in he GIS da a may be impo an . Howe e , he non- andom e o s in he ela ionships be ween he GIS da a and he indica o s o ecosys em se ices a e likely o be he mos impo an sou ces o e o when de ec ing he ho spo s. Ob iously, he ela ionships be ween he impac s and he ecosys em se ices a e mo e impo an han in he p e ious decision ypes, i we wish o map he impac s. The unce ain ies ela ed o he impac s a e also impo an , i we wish o de ec loca ions wi h g ea es possibili ies o inc easing he p o ision o selec ed ecosys em se ices. The unce ain ies in he GIS da a a e also ai ly impo an in de ec ing he co ec spo s. He e, as well as in he p e ious cases, we assess he non- andom unce ain ies o ha e he highes impac s in he analyses. 5. Discussion Conside ing he high ele ancy o ecosys em se ices o policy making, he in o ma ion basis is no adequa e o mos o he o es - ela ed ecosys em se ices. Maps o ecosys em se ices a e only as good as he da a a ailable on he se ices o be included (Blacks ock e al., 2015). The e o e, i is conce ning ha in a ecen e iew o 405 pape s on ecosys em se ices (Boe ema e al., 2017), only a fi h o he s udies included ac ual field measu emen s, i.e. p ima y da a, and 31% o s udies did no epo any da a. I is possible o make in o med decisions conce ning some o he o es y- ela ed se ices, especially imbe p oduc ion and ca bon se- ques a ion o which good-quali y da a a e usually a ailable om NFIs. Fo some o he s, like be y and mush oom p oduc ion, in o ma- ion is a ailable, bu i s quali y may no be high enough o in o med decisions (see Kilpeläinen e al., 2016). Maps conce ning he benefi s and alues also in ol e conside able e o s (Schägne e al., 2013). Mo eo e , he quali y o he indica o s hemsel es should be alida ed. Many o he indica o s cu en ly used a e e y weakly co ela ed wi h he ecosys em se ices (Boe ema e al., 2017). Measu ing indica o a iables in na ional and global o es in en o ies would p o ide a b eak h ough o sus ainable o es managemen (e.g. Miu a e al., 2015). I would mean ha p ima y da a on he se ices a e a ailable bo h o he analyses and o he alida ion. Assessing he unce ain ies in he es ima ed p o ision o he eco- sys em se ices is conside ed o be highly impo an (e.g. Boe ema e al., 2017, Ba on e al., 2018). In spi e o his, Boe ema e al. (2017) conclude ha unce ain y and alida ion analyses a e mos ly igno ed. E en i unce ain ies a e assessed, he assessmen ypically co e s only one o wo possible sou ces o unce ain y. Many o he po en ial sou ces lis ed in he p e ious sec ions ha e only a ely been comp e- hensi ely add essed. In he pape s we e iewed, he non- andom un- ce ain ies ela ed o he numbe o LU/LC classes and RS da a e- solu ion we e add essed mos o en. We also disco e ed cases whe e he andom misclassifica ion e o s o GIS da a, he non- andom un- ce ain y due o possible model misspecifica ion, and he andom un- ce ain y due o poo co ela ion we e add essed. Such cases we e s ill a e, conside ing ha ou sea ch was specifically ocused on pape s add essing he unce ain y. I is p obable ha we did no find all e- le an pape s, bu conside ing he ob ious lack o exis ence o such pape s, unlikely many we e missed. Fo some ecosys em se ices, he only a ailable analyses a e based on expe opinion and LU/LC classes, which may be highly biased and poo ly co ela ed wi h he ac ual se ices. Conside ing ha p ima y da a a e a ailable only o a mino i y o s udies, he expe opinion is a e y impo an sou ce o da a. Rela ed o he impo ance o his da a sou ce, i is no able ha we we e no able o find any ES s udies ha Table 4 Possibili ies o accoun o he e o s in decision making. Sou ce o e o Random e o s in decision analysis Non- andom e o s in decision analysis Indica o s o ecosys em se ices •Sensi i i y analysis •Scena io analysis •S ochas ic op imiza ion •Bayesian decision analysis •Fuzzy decision analysis •Fuzzy op imiza ion GIS da a (used as a basis o mapping) •Sensi i i y analysis •Scena io analysis •S ochas ic op imiza ion •Bayesian decision analysis •Scena io analysis Rela ionship be ween GIS da a and indica o s o ES •Sensi i i y analysis •Scena io analysis •S ochas ic op imiza ion •Bayesian decision analysis •Scena io analysis Impac o human in e en ions •Sensi i i y analysis •Scena io analysis •S ochas ic op imiza ion •Bayesian decision analysis •Adap i e managemen •Scena io analysis A. Kangas e al. Fo es Ecology and Managemen 427 (2018) 7–16 13 would ha e add essed he unce ain y due o (non- andom) a ia ion among he expe s o possible biases o misspecifica ions o he expe opinion models. In he u u e, i is e y impo an o alida e hese expe judgmen s udies. As he majo i y o expe s can be w ong, i is impo an o exe cise cau ion wi h hese analyses. The e is an ine i able adeoffbe ween he co e age, le el o de ail, and accu acy o he da a, which eflec s o he decision making. Fo in e na ional policy making, global assessmen s a e equi ed and a low le el o de ail and low accu acy migh be accep able. The equi emen s o de ail and accu acy inc ease om he na ional o egional and local le els o policy making. Unce ain ecosys em se ice maps wi h a low le el o de ail a e ill-sui ed o decision suppo o o iden i ying a eas ha p oduce mul iple se ices a a local le el. E en so, such maps migh s ill be used o de ec la ge-scale ends (Eigenb od e al., 2010). When local p ima y da a a e a ailable o he mapping, maps ha a e be e sui ed also o local decision making can be ob ained. The unce ain ies in he ecosys em se ices analyses a e high, bu p oducing significan ly mo e accu a e in o ma ion on ecosys em se - ices may ake yea s. Ne e heless, Cos anza e al. (2017) p oposed imp o ing he measu emen s and hei u iliza ion as a key o iden i y diffe ences in ou comes among policy choices. Moo e e al. (2017) e- commend ha in some cases quali a i e a he han quan i a i e in- o ma ion on ecosys em se ices is used, due o he high unce ain y. We ag ee wi h he p oblem b ough up by bo h he e e ences abo e, bu disag ee on means o sol e i . Al eady in he sho e m and in he absence o be e measu emen s, app op ia e and anspa en in- o ma ion can be p o ided o decision make s using quan i a i e in- o ma ion and decision analysis echniques ha accoun o he in- he en unce ain ies (Table 4). The quali y equi emen s conce ning he in o ma ion on ecosys em se ices o planning and policy making should be assessed. This as- sessmen depends on he scale o he decision p oblem (local, egional o na ional). I is also impo an o ela e he quali y needs o he empo al scale o he decisions and policies. I is clea ha in a spa ial scale o pixels, s ands o es a es, he decisions o be made a e diffe en han hose o na ional o global le els. As a consequence, also he in- dica o s ha a e ele an and/o use ul a e diffe en , as well as he accu acy equi emen s se o hese indica o s. We also need o assess he quali y equi emen s sepa a ely o diffe en ypes o p oblems (Table 5), as he da a equi emen s a e ob iously diffe en o diffe en asks. The mos p essing need is o assess non- andom unce ain ies. The sufficien accu acy o in o ma ion can be add essed based on he concep o alue o in o ma ion (VOI) o each specific decision p oblem. I can (ex an e) be calcula ed as he diffe ence o expec ed alue o a gi en decision wi h and wi hou a sou ce o new in o ma ion (Law ence, 1999, Kangas, 2010). The VOI o moni o ing ecosys em se ices has no been es ima ed in any s udy o ou knowledge. How- e e , Eigenb od e al. (2010) conclude ha benefi s om imp o ing he in o ma ion on ecosys em se ices by sample-based mapping a ou - weigh he cos s. The e is also an u gen need o communica e he scope and lim- i a ions o a ious ecosys em se ice maps o he use s (Vo s ius and Sp ay, 2015). Mo eo e , he e is a need o in oduce he unce ain ies in o he decision making p ocess, in o de o imp o e he ac ual deci- sions. Un o una ely, Ba on e al. (2018) conclude ha only 2% o he 313 pape s hey e iewed ac ually a ge ed decision making. Add es- sing he unce ain y in he decision p ocess could imp o e he decisions and educe he isks in a mo e cos -efficien way han imp o ing he accu acy o p edic ions by be e p ima y da a (e.g. Ey indson and Kangas, 2014). This is possible when he decisions a e good o many scena ios o he u u e p o ision o ecosys em se ices, ins ead o being he bes o op imal o jus one ou come. The e is a need o change he ocus om assessing and mapping he ecosys em se ices o ac ually using he in o ma ion in decision making, and doing i in elligen ly, i.e. aking also he unce ain ies in o accoun . 6. Conclusions The quali y o in o ma ion o diffe en ecosys em se ices a ies be ween he diffe en se ices. Fo es in en o ies p o ide accu a e in- o ma ion o imbe p oduc ion, whe eas o many o he se ices, ex- pe opinions se e as he main sou ce o in o ma ion. The la ge he spa ial scale o he analyses, he less in o ma ion is ypically a ailable and his si ua ion canno be expec ed o ma kedly imp o e in he u- u e. Inco po a ing he unce ain ies in o he decision making and policy analyses enables us o p oduce obus decisions and policies, meaning ha he ecommenda ions apply o a whole se o possible scena ios o u u e and e y poo ou comes can be a oided. Thus, as- sessing he unce ain ies and acknowledging hem in decision making is he as es and cheapes way o imp o e he p o ision policies o o es - ela ed ecosys em se ices. Re e ences Alho, J.M., Kangas, J., 1997. Analyzing unce ain ies in expe s’opinions o o es plan pe o mance. Fo es Sci. 43, 521–528. And ew, M.E., Wulde , M.A., Nelson, T.A., 2014. Po en ial con ibu ions o emo e sen- sing o ecosys em se ice assessmen s. P og . Phys. Geog . 38, 328–353. Au inen, A.-P., Hildén, M., Toi onen, H., P imme , E., Niemelä, J., Aapala, K., Bäck, S., Hä mä, P., Ikä alko, J., Jä enpää, E., Kaipiainen, H., Ko honen, K.T., Kumela, H., Kä kkäinen, L., Lankoski, J., Laukkanen, M., Manne koski, I., Nuu inen, T., Nöjd, A., Pun ila, P., Salminen, O., Söde man, G., Tö mä, M., Vi kkala, R., 2007. E alua ion o he Finnish Na ional Biodi e si y Ac ion Plan 1997-2005. Monog aphs o he Bo eal En i onmen Resea ch No. 29. 54 p. Ayanu, Y.Z., Con ad, C., Nauss, T., Wegmann, M., Koellne , T., 2012. Quan i ying and mapping ecosys em se ices supplies and demands: a e iew o emo e sensing ap- plica ions. En i on. Sci. 46, 8529–8541. Balz e , H., Cole, B., Thiel, C., Schmullius, C., 2015. Mapping CORINE land co e om sen inel-1A SAR and SRTM digi al ele a ion model da a using andom o es s. Remo e Sens. 7, 14876–14898. Ba edo, J.I., e al., 2015. Mapping and assessmen o o es ecosys ems and hei se ices –Applica ions and guidance o decision making in he amewo k o MAES. Repo EUR 27751 EN, Join Resea ch Cen e, Eu opean Union, 78 p. doi: 10.2788/720519. Ba on, D.N., Kelemen, E., Dick, J., Ma in-Lopez, E., Gómez-Bagge hun, S., Jacobs, S., Hend icks, C.M.A., Te mansen, M., Ga cía-Llo en e, M., P imme , E., Dun o d, R., Ha ison, P.A., Tu kelboom, F., Saa ikoski, H., an Dijk, J., Rusch, G.M., Palomo, I., Yli-Pelkonen, V.J., ... Lapola, D.M., 2018. (Dis) in eg a ed alua ion –Assessing he in o ma ion gaps in ecosys em se ice app aisals o go e nance suppo . Ecosys . Se . 29 (C), 529–541. Table 5 Assessmen o he impo ance o unce ain y assessmen o a ious example applica ions (R – andom and N –non- andom e o s). Applica ion Sou ce o e o Indica o s o ecosys em se ices GIS da a (used as a basis o mapping) Rela ionship be ween GIS da a and ES Impac o human in e en ions RN R N R N R N Analyses o ade-offs and syne gies Low High Low Low Medium High Low Low De ec ing cold and ho spo s Low High Medium Medium Medium High Low Low Mapping o impac s Low High Low Low Medium High Medium High Mapping o oppo uni ies o imp o emen s Low High Medium Medium Medium High Medium High A. Kangas e al. Fo es Ecology and Managemen 427 (2018) 7–16 14 Bení ez, B.C., McCallum, I., Obe s eine , M., Yamaga a, Y., 2007. Global po en ial o ca bon seques a ion: Geog aphical dis ibu ion, coun y isk and policy implica ions. Ecol. Econ. 60, 572–583. Bi ge, H.E., Allen, C.R., Ga mes ani, A.S., Pope, K.L., 2016. Adap i e managemen o ecosys em se ices. J. En i on. Manage. 183, 343–352. Blacks ock, K., Ma in-O ega, J., Sp ay, C.J., 2015. Implemen a ion o he Eu opean wa e amewo k di ec i e: wha does aking an ecosys em se ices-based app oach add? In: Ma in-O ega, J., Fe ie , R.C., Go don, I.J., Khan, S. (Eds.), Wa e Ecosys em Se ices: A Global Pe spec i e. Camb idge Uni e si y P ess. (In e na ional Hyd ology Se ies), Camb idge, pp. 57–64. Boe ema, A., Rebelo, A.J., Bodi, M.B., Esle , K.J., Mei e, P., 2017. A e ecosys em se ices adequa ely quan ified? J. Appl. Ecol. 54, 358–370. Boi hias, L., Te ado, M., Co ominas, L., Zi , G., Kuma , V., Ma gués, M., Schumache , M., Acuña, V., 2016. Analysis o he unce ain y in he mone a y alua ion o ecosys em se ices –A case s udy a he i e basin scale. Sci. To al En i on. 543, 683–690. Bul e, E., Van Koo en, G.C., 2000. Economic science, endange ed species, and biodi e si y loss. Conse . Biol. 14, 113–119. Chi ici, G., McRobe s, R.E., Win e , S., Be ini, R., B aendli, U.-B., Asensio, I.A., Bas up- Bi k, A., Rondeux, J., Ba soum, N., Ma che i, M., 2012. Na ional o es in en o y con ibu ions o o es biodi e si y moni o ing. Fo es Sci. 58, 257–268. h p://dx. doi.o g/10.5849/ o sci.12-003. Chi ici, G., Win e , S., McRobe s, R.E. (Eds.), 2011. Na ional Fo es In en o ies: Con ibu ions o Fo es Biodi e si y Assessmen s. Sp inge . Managing Fo es Ecosys ems. Cos anza, R., d’A ge, R., de G oo , R., Fa be k, S., G asso, M., Hannon, B., Limbu g, K., Naeem, S., O’Neill, R.V., Pa uelo, J., Raskin, R.G., Su on, P., an den Bel , M., 1997. The alue o he wo ld's ecosys em se ices and na u al capi al. Na u e 387 (6630), 253–260. Cos anza, R., de G oo , R., B aa , L., Kubiszewski, I., Fio amon i, L., Su on, P., Fa be , S., G asso, M., 2017. Twen y yea s o ecosys em se ices: how a ha e we come and how a do we s ill need o go? Ecosys . Se . 28, 1–16. Cos anza, R., 2008. Ecosys em se ices: Mul iple classifica ion sys ems a e needed. Biol. Conse . 141, 350–352. Co ona, P., 2016. Consolida ing new pa adigms in la ge-scale moni o ing and assessmen o o es ecosys ems. En . Res. 144, 8–14. h p://dx.doi.o g/10.1016/j.en es.2015. 10.017. Co ona, P., Chi ici, G., McRobe s, R.E., Win e , S., Ba ba i, A., 2011. Con ibu ion o la ge-scale o es in en o ies o biodi e si y assessmen and moni o ing. Fo . Ecol. Manage. 262, 2061–2069. h p://dx.doi.o g/10.1016/j. o eco.2011.08.044. C ossman, N.D., Bu kha d, B., Nedko , S., Willemen, L., Pe z, K., Palomo, I., D akou, E.D., Ma ín-Lopez, B., McPhea son, T., Boyano a, K., Alkemade, R., Egoh, R., Dunba , M.B., Maes, J., 2013. A bluep in o mapping and modelling ecosys em se ices. Ecosys . Se . 4, 4–14. Daily, G. (Ed.), 1997. Na u e’s Se ices: Socie al Dependence on Na u al Ecosys ems. Island P ess. Da ies, A.B., Asne , G.P., 2014. Ad ances in animal ecology om 3D-LiDAR ecosys em mapping. T ends Ecol. E ol. 29, 681–691. Dong, M., B yan, B.A., Conno , J.D., Nolan, M., Gao, L., 2015. Land use mapping e o in oduces s ongly-localized, scale-dependen unce ain y in o land use and eco- sys em se ices modelling. Ecosys . Se . 15, 63–74. Eigenb od, F., A mswo h, P.R., Ande son, B.J., Heinemeye , A., Gillings, S., Roy, D.B., Thomas, C.D., Gas on, K.J., 2010. The impac o p oxy-based me hods on mapping he dis ibu ion o ecosys em se ices. J. Appl. Ecol. 47, 377–385. Eu opean Commission, 2011. Communica ion om he Commission o he Eu opean Pa liamen , he Council. The Economic and Social Commi ee and he Commi ee o he Regions, B ussels, 3.5.2011 COM(2011) 244 final. Ey indson, K., Kangas, A., 2014. S ochas ic goal p og amming in o es planning. Can. J. Fo . Res. 44, 1274–1280. FAO, CIFOR, IFRI and Wo ld Bank. 2016. Na ional socioeconomic su eys in o es y: guidance and su ey modules o measu ing he mul iple oles o o es s in household wel a e and li elihoods, by R.K. Bakkegaa d, A. Ag awal, I. Animon, N. Hoga h, D. Mille , L. Pe sha, E. Rame s eine , S. Wunde , A. Zezza. FAO Fo es y Pape No. 179. Food and Ag icul u e O ganiza ion o he Uni ed Na ions, Cen e o In e na ional Fo es y Resea ch, In e na ional Fo es y Resou ces and Ins i u ions Resea ch Ne wo k and Wo ld Bank. Foody, G.M., 2015. Valuing map alida ion: The need o igo ous land co e map ac- cu acy assessmen in economic alua ions o ecosys em se ices. Ecol. Econ. 111, 23–28. F ank, S., Fü s , C., Koschke, L., Makeschin, F., 2012. A con ibu ion owa ds a ans e o he ecosys em se ice concep o landscape planning using landscape me ics. Ecol. Ind. 21, 30–38. F ank, S., Fü s , C., Pie zsch, F., 2015. C oss-sec o al esou ce managemen : how o es managemen al e na i es affec he p o ision o biomass and o he ecosys em se - ices. Fo es s 6, 533–560. h p://dx.doi.o g/10.3390/ 6030533. Fü s , C., Volk, M., Pie zsch, K., Makeschin, F., 2010. Pimp you landscape: a ool o quali a i e e alua ion o he effec s o egional planning measu es on ecosys em se ices. En i on. Manage. 46, 953–968. Fü s , C., F ank, S., Wi , A., Koschke, L., Makeschin, F., 2013. Assessmen o he effec s o o es land use s a egies on he p o ision o ecosys em se ices a egional scale. J. En i on. Manage. 127. Fü s , C., 2015. Does using he ecosys em se ices concep p o oke he isk o assigning i ual p ocess ins ead o eal alues o na u e? Some eflec ions on he benefi o ecosys em se ices o planning and policy consul ing. Eu . J. Ecol. 1, 39–44. Ga cía Má quez, J.R., K uege , T., Páez, C.A., Ruiz-Agudelo, C.A., Beja ano, P., Mu o, T., A jona, F., 2017. Effec i eness o conse a ion a eas o p o ec ing biodi e si y and ecosys em se ices: a mul i-c i e ia app oach. In . J. Biodi e si y Sci., Ecosys . Se ices Manage. 13 (1), 1–13. Geijzendo ffe , I.R., Roche, P.K., 2013. Can biodi e si y moni o ing schemes p o ide indica o s o ecosys em se ices? Ecol. Indica o s 33, 148–157. G ê -Regamey, A., B unne , S.H., Al wegg, J.H., Bebi, P., 2013. Facing unce ain y in ecosys em se ices-based esou ce managemen . Ecosys . Se . 15, 63–74. de G oo , R.S., Wilson, M.A., Boumans, R.M.J., 2002. A ypology o he classifica ion, desc ip ion and alua ion o ecosys em unc ions, goods and se ices. Ecol. Econ. 41, 393–408. Haakana, H., Hi elä, H., Hanski, I.K., Packalen, T., 2017. Compa ing egional o es policy scena ios in e ms o p edic ed sui able habi a s o he Sibe ian flying squi el (P e omys olans). Scand. J. Fo . Res. 32, 185–195. Haines-Young, R., Po schin, M., 2010. P oposal o a Common In e na ional Classi-fica- ion o Ecosys em Goods and Se ices (CICES) o In eg a ed En i onmen aland Economic Accoun ing. Backg ound Documen , Repo o he EEA (21 Ma ch2010). Hansen, K., Malmaeus, M., 2016. Ecosys em se ices in Swedish o es s. Scand. J. Fo . Res. 31, 626–640. Japelj, A., Ma sa , R., Hodges, D.G., Ko ač, M., Ju ančič, L., 2016. La en p e e ences o esiden s ega ding an u ban o es ec ea ion se ing in Ljubljana, Slo enia. Fo es Policy Econ. 71, 70–78. Ha ikainen, M., Ey indson, K., Mie inen, K., Kangas, A., 2016. Da a-based o es man- agemen wi h unce ain ies and mul iple objec i es. In: Giuff ida, G., Nicosia, G., Pa dalos, P. (Eds.), The Second In e na ional Wo kshop on Machine Lea ning, Op imiza ion and big Da a - MOD 2016. Sp inge LNCS 10122. Ho ne, P., Boxall, P.C., Adamowicz, W.L., 2005. Mul iple-use managemen o o es e- c ea ion si es: a spa ially explici choice expe imen . Fo es Ecol. Manage. 207, 189–199. Ihalainen, M., Alho, J., Kolehmainen, O., Pukkala, T., 2002. Expe models o bilbe y and cowbe y yields in Finnish o es s. Fo . Ecol. Manage. 157, 15–22. Jacobs, S., Bu kha d, B., Van Daele, T., S aes, J., Schneide s, A., 2015. ‘The Ma ix Reloaded’: A e iew o expe knowledge use o mapping ecosys em se ices. Ecol. Modellling 295, 21–30. Kaise , G., Bu kha d, B., Röme , H., Sangkaew, S., G a e ol, R., Hai ook, T., S e , H., Sakuna-Schwa z, D., 2013. Mapping sunami impac s on land co e and ela ed ecosys em se ice supply in Phang Nga, Thailand. Na . Haza ds Ea h Sys . Sci. 13, 3095–3111. Kangas, A., Ku ila, M., Hujala, T., Ey indson, K., Kangas, J., 2015. Decision suppo o o es managemen . In: Managing Fo es Ecosys ems, second ed. Sp inge , pp. 307. Kangas, A., 2010. Value o o es in o ma ion. Eu . J. Fo es Res. 129, 863–874. Kangas, J., Alho, J., Kolehmainen, O., Mononen, A., 1998. Analyzing consis ency o ex- pe s' judgmen s - Case o assessing o es biodi e si y. Fo es Sci. 44 (4), 610–617. Kangas, J., Laasonen, L., Pukkala, T., 1993. A me hod o es ima ing o es landowne 's landscape p e e ences. Scand. J. Fo . Res. 8, 408–417. Kan , S., Lee, S., 2004. A social choice app oach o sus ainable o es managemen : an analysis o mul iple o es alues in No hwes e n On a io. Fo es Policy Econ. 6, 215–227. Kilpeläinen, H., Miina, J., S o e, R., Salo, K., Ku ila, M., 2016. E alua ion o bilbe y and cowbe y yield models by compa ing model p edic ions wi h field measu emen s om No h Ka elia, Finland. Fo . Ecol. Manage. 363, 120–129. Kleeman, J., Baysal., G., Bulley, H.N.N. and Fü s , C., 2017. Assessing d i ing o ces o land use and land co e changes by a mixed-me hod app oach in no he n-eas e n Ghana, Wes A ica. J. En i on. Manage. 196, 411–442. Koschke, L., Fü s , C., F ank, S., Makechin, F., 2012. A mul i-c i e ia app oach o an in eg a ed land-co e -based assessmen o ecosys em se ices p o ision o suppo landscape planning. Ecol. Ind. 21, 54–66. Ku ki, S., Nikula, A., Helle, P., Lindén, H., 2000. Effec s o landscape agmen a ion and o es composi ion on b eeding success o g ouse in bo eal o es s. Ecology 81, 1985–1997. K ishnaswamy, J., Bawa, K.S., Ganeshaiah, K.N., Ki an, M.C., 2009. Quan i ying and mapping biodi e si y and ecosys em se ices: U ili y o a mul i-season NDVI based Mahalanobis dis ance su oga e. Remo e Sens. En i on. 113, 857–867. La o el, S., Baye , A., Bondeau, A., Lau enbach, S., Ruiz-F au, A., Schulp, N., Seppel , R., Ve bu g, P., an Teeffelen, A., Vannie , C., A ne h, A., C ame , W., Ma ba, N., 2017. Pa hways o b idge he biophysical ealism gap in ecosys em se ices mapping ap- p oaches. Ecol. Ind. 74, 241–260. Law ence, D.B., 1999. The Economic Value o In o ma ion. Sp inge , pp. 393. Leskinen, P., Kangas, J., Pasanen, A.-M., 2003. Assessing ecological alues wi h depen- den explana o y a iables in mul i-c i e ia o es ecosys em managemen . Ecol. Model. 170, 1–12. Ligmann-Zielinska, A., Jankowski, P., 2014. Spa ially-explici in eg a ed unce ain y and sensi i i y analysis o c i e ia weigh s in mul ic i e ia land sui abili y e alua ion. En i on. Modell. So wa e 57, 235–247. Li ne, E., S o ay, T., 2011. Componen s o unce ain y in p ima y p oduc ion model: he s udy o DEM, classifica ion and loca ion e o . In . J. Geog aphical In . Sci. 25, 473–488. Lou ie e, J.D., Henshe , D., Swai , J., 2000. S a ed Choice Me hods. Analysis and Applica ion. Camb idge Uni e si y P ess, pp. 402. M.A., 2005. Millennium ecosys em assessmen . In: Ecosys ems and Human Well-Being: Syn hesis. Island P ess, Washing on, D.C. Ma inez-Ha ms, M.J., Quijas, S., Me enlende , A.M., Bal ane a, P., 2016. Enhancing ecosys em se ices maps combining field and en i onmen al da a. Ecosys . Se . 22, 12–40. Ma inez-Ha ms, M.J., B yan, B.A., Bal ane a, P., Law, E.A., Rhodes, J.R., Possingham, H.P., Wilson, K.A., 2015. Making decisions o managing ecosys em se ices. Biol. Conse . 184, 229–238. Ma ínez-Ha ms, M.J., Bal ane a, P., 2012. Me hods o mapping ecosys em se ice supply: a e iew. In . J. Biodi e si y Sci., Ecosys . Se . Manage. 8 (1–2), 17–25. A. Kangas e al. Fo es Ecology and Managemen 427 (2018) 7–16 15