Unravelling the mechanisms linking cultural ecosystem services and human wellbeing

Ecosystems contribute significantly to human wellbeing through the provision of ecosystem services. 39 Despite the growing literature on cultural ecosystem services (CES), there is a lack of systematic 40 understanding of how they are linked with human wellbeing. Here we conduct a systematic review of the 41 peer-reviewed literature to identify the mechanisms and pathways underpinning the linkages between 42 different CES and constituents of human wellbeing. Furthermore we identify their complex associations 43 through Latent Class Analysis, Multiple Correspondence Analysis, and different visualisation tools. Overall 44 we identify 16 major mechanisms linking CES and human wellbeing, via 70 distinct pathways. Beyond that 45 we find five major assemblages of pathways featuring consistent associations among mechanisms mediating 46 CES and human wellbeing. We critically discuss the main research trends and gaps, and propose future 47 directions for research and practice in order to leverage the potential of CES for human wellbeing, and 48 sustainability more broadly. 49

The academic community has repeatedly emphasised on the necessity to understand the complex 54 human-nature relationships and unravel the pathways through which ecosystems contribute to human 55 wellbeing via the provision of ecosystem services 1,2 . Understanding the underlying processes linking 56 ecosystem services and human wellbeing, designing appropriate interventions to leverage the 57 contribution of ecosystem services on human wellbeing, and mitigating the negative impacts of 58 human activity on ecosystem services are essential for sustainable natural resource management 2,3 59 Cultural ecosystem services (CES) are the diverse non-material contributions of nature to humans, 60 such as, among others, recreation, spiritual enrichment, cognitive development, social relations, and 61 aesthetic experiences 1, 4,5 . Despite the growing attention of the academic community on CES, it has 62 been exceptionally challenging to systematise in concrete terms their linkages with human 63 wellbeing 6,7 . Unlike provisioning, regulating, and supporting services whose assessment, although 64 often complicated, is usually possible through quantitative and consistent metrics, CES are often 65 intangible, subjective, socially constructed, and dependent on human perception, requiring an entirely 66 different set of tools, metrics, and approaches 4,8,9 . Hence, despite the importance of CES for 67 sustainable resource management 10,11 , their assessment and contributions to human wellbeing is often 68 a context-specific and qualitative endeavour, which makes the effective incorporation of its outcomes 69 into policy-making processes very challenging 6 . In this sense the systematic understanding of the 70 linkages between CES and human wellbeing can offer valuable insights for policy and practice for 71 ecosystem management and broader sustainability. 72 However, despite the ever-expanding body of literature discussing the interface of CES and human 73 wellbeing, the current evidence is highly fragmented. First, the relevant literature tends to adopt 74 different theoretical frameworks with various terminologies [4][5][6][7] , resulting in fragmented information 75 and inconsistent CES assessments and valuations. This is largely due to the fact that the underlying 76 research comes from a very diverse range of academic fields, with often limited effort to synthesise 77 knowledge in a cohesive manner 12 . Second, as both the provision of CES and the linkages to human 78 wellbeing are highly context-dependent, their generalisation and systematic understanding tends to be 79 riddled with challenges 12 . 80 Here we aim to bridge these gaps through the comprehensive and cohesive systematisation of the 81 linkages between CES and human wellbeing. In particular we conduct a systematic review and 82 analysis of the relevant peer-reviewed literature to (a) delineate the mechanisms and pathways linking 83 different CES and constituents of human wellbeing (b) compare the effects of the different 84 mechanisms underpinning these linkages on human wellbeing, and (c) identify possible associations 85 among the mechanisms. The outcome of this Review is a theoretical framework consisting of 16 86 individual mechanisms linking different combinations of CES and constituents of human wellbeing, 87 as well as their complex interaction in terms of synergies and trade-offs. Beyond this we identify and 88 critically discuss research trends and gaps at the interface of CES and human wellbeing, and offer 89 recommendations for future research and opportunities for ecosystem management to leverage the 90 potential of CES for human wellbeing, and sustainability more broadly.

General literature patterns 98
The systematic review identified literature on the linkages between CES and human wellbeing across 99 all continents. The reviewed studies span a total of 62 countries at various spatial scales, with 81.8% 100 of papers focusing on the local scale (n=247 studies), 8.3% at the national scale (n=25 studies), 6.3% 101 at the regional scale (n=19 studies), and 3.6% at the global scale (n=10 studies). In terms of the 102 stakeholders represented across the different studies, almost all studies consider local communities, 103 followed by tourists, indigenous communities, and farmers, fishers and business owners ( Figure S5-104 S6, Supplementary Material). 105 Figure 1 shows the geographical distribution of the study sites considered in the respective literature, 106 and the number of publications by region and ecosystem type. Most studies focus on Europe (42.1% 107 of articles), Asia (21.7% of articles), and North America (18.5% of articles). Only a minority of 108 studies focuses on Central and South America (6.5% of articles), Africa (5.8% of articles), and 109 Oceania (5.4% of articles), despite being biodiversity-rich and large fractions of their populations 110 depending on ecosystem services for their livelihoods. The reviewed studies mostly focus on CES 111 from urban and semi-urban ecosystems (26.2% of articles), forests and woodlands (20.2% of articles), 112 inland water (12.5% of articles), and coastal areas (8.9% of articles). Some of the reviewed studies 113 also document the linkages between CES and human wellbeing in relatively less studied ecosystems 114 such as the arctic and mountain tundra, desserts and scrublands, and savannas.  Platform on Biodiversity and Ecosystem Services -IPBES), relevant studies came from a larger 123 diversity of the academic fields. By 2020, studies were coming from diverse fields of the social 124 sciences and humanities, cultural studies, psychology, pharmacology, medicine, and international 125 relations, among others. It is worth noting that the reviewed studies have gradually adopted new and 126 innovative tools from different disciplines for data collection and analysis. However, a closer 127 examination of the theoretical frameworks and research tools (see Table S7, Supplementary Material) 128 shows that knowledge integration cross the disciplines is still rather shallow and the diversity of the 129 methodological portfolio is generally low. 130

Pathways and mechanisms linking CES and human wellbeing 131
Among the 301 reviewed studies and the 1138 observations of the pathways linking CES and human 132 wellbeing, the 979 observations (86%) represent positive contributions, the 137 observations (12%) 133 negative contributions, and the 18 observations (1.6%) two-way interactions, while 4 observations 134 (0.4%) could not be categorised in terms of the direction of impact. 135 Our results suggest that the pathways linking CES and human wellbeing are multi-faceted and 136 intricate. We identify 70 unique such pathways, each of which depicts a linkage through which the 137 provision of (or change in) a single CES affects a single constituent of human wellbeing (see Methods  138 for deeper explanation). Of these 70 pathways, 45 denote positive contributions to human wellbeing 139 and 25 negative contributions to human wellbeing. Table S12-S14 in the Supplementary Material  140  provide an explanation of these 70 pathways, including examples from the literature. Subsequently  141 through a critical analysis we systematise similar pathways across various "channels of interaction" 142 and "mechanisms". 143 The four channels of interactions essentially denote the different ways in which people consciously 144 and unconsciously engage with ecosystems and experience their benefits. According to our results, the 145 four channels are form, cultural practices, intellectual practices, and spiritual practices. Form 146 essentially denotes the interactions with nature through the physical and tangible aspects of 147 ecosystems. People perceive the physical structure of nature via multiple qualities that reflect visual 148 and other sensory experiences, with examples including the interactions from looking at the shape of 149 the cliffs, feeling the sea breeze, smelling the flowers scent among others [13][14][15] Cultural practices 150 denote the interactions with nature that provide an opportunity for playing and exercising, creating 151 and expressing, producing and caring, and gathering and consuming 9,16 . Intellectual practices denote 152 the interactions with nature that provide an environment for learning and gaining new knowledge, 153 including, for example, the interactions that emerge from researching, learning, thinking about or 154 knowing an ecosystem or its components 12,17 . Spiritual practices denote the interactions with nature 155 that provide an opportunity for spiritual and religious activities, as for example rituals and religious 156 activities carried out in sacred natural places or using plants and animals 18,19 . 157 Through these channels, CES contribute to human wellbeing via very diverse mechanisms. We 158 identify 16 types of mechanisms, namely (a) cognitive, satisfactive, (l) transactive, (m) transcendentive, (n) apprehensive, (p) destructive, and (q) irritative 161 (Table 1). Of these, six mechanisms were adapted from a previous study 15 and ten mechanisms were 162 newly defined by the authors following the qualitative data analysis (see Methods). 163 164 <<Insert Table 1>>  165 166 The empirical research on these mechanisms linking CES and human wellbeing constituents is uneven. 167 Figure 2 depicts an alluvial diagram illustrating the frequency of the documented mechanisms across 168 the reviewed studies. Recreation and tourism and aesthetic value are the most popular CES among the 169 identified studies, accounting respectively for 31.8% and 17.6% of the total observations. Similarly 170 there is large representation for some mechanisms and constituents of human wellbeing, as for 171 example CES contributions to "mental health" (15.8% of observations), "physical health" (10.6% of 172 observations), and "subjective wellbeing" (13% of observations) via regenerative mechanisms.

173
Equally well documented are CES benefits to "social connectedness and belonging" (12.7% of 174 observations) via communicative and cohesive mechanisms. However, the role of CES in influencing 175 many constituents of human wellbeing is still overlooked, such as for example the contributions of 176 CES to "learning and capacity", "personal identity and autonomy", and "sense of security and 177 certainty".

Relative contribution of individual mechanisms 184
Upon unravelling the pathways, channels of interactions, and the mechanisms between CES and 185 human wellbeing, we normalise the effects for every observation through expert judgement, assigning 186 scores between -2 to +2 (see Methods). We develop three matrix maps ( Figure 3) that represent the 187 level of impact of each mechanism to specific constituents of human wellbeing (denoted by the 188 colours) and the overall quantity of the empirical literature (denoted by the size of the squares). 189 Overall, there is a higher prevalence of high magnitude positive CES impacts on human wellbeing, 190 while the prevalence of lower magnitude or negative impacts is relatively lower in the literature.  tourism and aesthetic value exhibit the highest contributions to human health via the regenerative 198 mechanism. "Connectedness and belonging" is the wellbeing constituent that benefits the second most 199 from CES with an average score of 1.92 (n=131, SE=0.03). CES can also have significant positive 200 effects for personal "learning and capability", with an average impact level of 1.91 (n=114, SE=0.04).

201
The average scores are more moderate for other constituents of human wellbeing such as "economic 202 wellbeing" (score=1.88, SE=0.09, n=61), "certainty, sense of control and security" (score=1. 86,203 SE=0.072, n=19, "identity and autonomy" (score=1.84, SE=0.042, n=71), "spirituality" (score=1. 79,204 SE=0.05, n=81), "inspiration and fulfilment of imagination" (score=1.72, SE = 0.03, n=84), and 205 "subjective wellbeing" (score=1.71, SE=0.34, n=125). 206 The negative contribution of CES to human wellbeing manifest through (a) cultural ecosystem 207 disservices (Table S12, Supplementary Material), and (b) the degradation of CES provision (Table  208 S13, Supplementary Material) (see Methods for definitions). Among all wellbeing constituents, the 209 results suggest that the highest negative effects are liked to "mental health" (score=-1.98, SE=0.02, 210 n=26), via the destructive mechanisms. Some disservices also profoundly hamper "certainty, sense of 211 control and security" (score=-1.88, SE=0.44, n=15 via apprehensive mechanisms. These are mainly 212 associated with aesthetic value and recreational and tourism, with a common underlying concern over 213 safety, which is directly associated with how some natural elements are perceived. CES degradation 214 sometimes also has negative effects on "spirituality" via the destructive mechanism (score=-1.94, 215 SE=0.06, n=10. Interestingly, in terms of "economic well-being", via the remunerative mechanism 216 human wellbeing can be significantly affected by the financial loss caused by the degradation of CES 217 (score=-1.96, SE=0.08, n=17). 218 219 Pathway assemblages, synergies and trade-offs. 220 Beyond the individual effects outlined above, CES and the pathways through which they contribute to 221 human wellbeing, also tend to interact with each other in complex ways. Following the LCA we 222 identify 'CES pathway assemblages', which refers to a collection of pathways linking CES and 223 wellbeing that appear significantly related (see Methods for definitions). 224 Overall, we identify five CES pathway assemblages that refer to (a) sensory affection, (b) learning 225 and development, (c) health and leisure fulfilment, (d) social vibrancy, and (e) spiritual and heritage 226 resources. Table 2 summarises the main features and synergies among these assemblages, the relevant 227 constituents of human wellbeing, as well as the underlying mechanisms and affected groups. 228

230
Subsequently we identify trade-offs between these assemblages through Multiple Correspondence 231 Analysis (MCA). Regarding the associations between CES, wellbeing constituents, and affected 232 groups, the significant associations of CES account for 7.3% of the first dimension (axis x) and 6.9% 233 of the second dimension (axis y) ( Figure 4). Trade-offs occur among three assemblages, namely 234 "health and leisure fulfilment", "spiritual and heritage resources", and "learning and development". 235 Below we examine some specific cases to illustrate these trade-offs better. On the second dimension, 236 "spiritual and heritage resources" is assessed as a trade-off to "health and leisure fulfilment" ( Figure  237 4). These trade-offs are mainly associated with traditional and indigenous communities. Furthermore, 238 we observe trade-offs between "health and leisure fulfilment" and "learning and development" on the 239 second dimension (Figure 4).

Mechanisms linking CES and human wellbeing 247
Following the critical analysis and synthesis of the findings of the 285 empirical case studies and their 248 1138 observations, we identify 16 distinct mechanisms that mediate the linkages between CES and 249 human wellbeing (Table 1). Six of these mechanisms are adapted from the literature 15 , while the 250 remaining ten were systematised by the authors. This has resulted in a comprehensive mapping of the 251 interface between CES and human wellbeing, and constitutes a conceptual advancement that can have 252 both theoretical and practical application. 253 In terms of theoretical application, these mechanisms can influence the development of conceptual 254 frameworks that explore the interface of ecosystem services and human wellbeing in more nuanced 255 ways. For example, this can help refine conceptual frameworks proposed by large-scale 256 assessments 1,2,20 , international initiatives 21 , and individual studies 4,5,7,8 . 257 In terms of practical applications, the findings of this study can guide studies that seek to identify 258 better how such mechanisms unfold in different real-life contexts and inform the design of appropriate 259 interventions to enhance human wellbeing through the provision of CES. This latter point echoes a 260 large number of studies arguing for the need to capitalise on the intangible benefits provided by green 261 spaces for enhanced human wellbeing, especially in urban contexts [22][23][24][25] . wellbeing. To the authors' best knowledge this is the first comprehensive attempt to systematically 266 quantify the impacts of CES on human wellbeing through a global systematic review. As the value 267 and contribution of many CES is subjective and intangible, the literature often adopts descriptive and 268 qualitative approaches, which make it challenging to undertake a systematic analysis across studies 4,7,9 . 269 The results of our analysis suggest that CES tend to have the highest positive contribution to 270 constituents of human wellbeing such as "physical health" and "mental health" via regenerative 271 mechanisms ( Figure 3). These denote interactions with nature that create positive restorative 272 outcomes such as stress reduction, relaxation, tranquillity, escapism, physical exercises, increased 273 longevity, and recovery from sickness 12,[26][27][28][29] . 274 Strong positive effects are also observed for "connectedness and belonging" via "cohesive" 275 mechanisms ( Figure 3). Here people communicate and develop meaningful personal relationships 276 through interactions with ecosystems 15,30-32 . For example, nature-based activities such as recreation, 277 hiking, and camping foster social cohesion via socially healthy behaviours and stewardship 30 . Studies 278 have consistently pointed that the social bonding mediated by interactions with nature can create 279 networks that emerge beyond the physical boundary of the sites and reinforce the existing relationship 280 at both personal and collective levels 33 . 281 Strong positive effects are also observed for "learning and capability" through cognitive and evolutive 282 mechanisms ( Figure 3). For example via the evolutive mechanism, nature-based recreation has a 283 positive effects on childhood growth and gradually equips children with knowledge and skillsets that 284 are beneficial in the future 34,3617,35 . These contributions can also be manifested via the cognitive 285 mechanism through which nature acts as the source of learning about history, culture, the natural 286 world and social relationships 37 , or an opportunity for scientific development, outdoor education and 287 learning from previous generations 38,39 . It is worth noticing that "learning and capability" is one of the 288 constituents of human wellbeing that has been relatively overlooked in the CES research landscape, as 289 for example it was not included in the conceptual framework of the Millennium Ecosystem 290 Assessment 1 . However, here we find the centrality of this constituent as almost all CES provide 291 benefits to it via various mechanisms. For example, via the cohesive and communicative mechanisms, 292 educational values, knowledge systems, social relations and recreation and tourism are the CES that 293 contribute the most to this wellbeing constituent, but again through very different mechanisms. 294 When it comes to negative contributions, the constituents of human wellbeing that are most negatively 295 affected by cultural ecosystem disservices and CES loss tend to be "mental health" and "certainty, 296 sense of control and security" via apprehensive 14,40 and destructive 41-43 mechanisms ( Figure 3). 297 It is well documented that cultural ecosystem disservices such as noise from wildlife, wild and messy 298 landscapes, and the presence and movement of pests give rise to perceptions of disorder, while animal 299 waste and plant litter many cause disgust 40,44 . For example, obsessive fear can emerge through 300 encounters with natural features via visual (or sometimes auditory) interactions, such as scary animals, 301 dangerous predators, animal blood, and areas that are dark and covered by high trees 12,14,40 . Many 302 people have a limited frame of reference for recognising and construing such unfamiliar sensory 303 experiences, and may develop a sense of overwhelming "cognitive chaos" and alienation towards 304 nature 12 . Landscape planning, eco-tourism development, socioeconomic background, and childhood 305 interactions with nature are some of the external factors that tend to mediate these mechanisms 44,45 . 306 When it comes to destructive mechanisms, the point of departure are the benefits that nature provides 307 to many people through spiritually transcendental experiences which transform something from 308 within 42,43 . These are closely linked with religion, places for customary rituals and worship 18 . Often 309 environmental degradation, urbanisation and/or overexploitation cause the gradual loss of spiritually 310 important landscapes 46,47 or plants/animals that with religious/spiritual importance 48 , causing the 311 decline/loss of ritual activities and related spiritual wellbeing for some people 49 . 312 It is worth noting that many people might not always be well aware of the benefits offered by 313 ecosystems. Yet they can be significantly affected by the financial loss caused by the degradation of 314 CES via the remunerative mechanism ( Figure 3). For example, climate change or environmental 315 degradation can cause significant loss in tourism revenue, thereby reducing the incomes of people 316 working in these sectors and hindering their capability meeting basic needs 50,51 . Some of the reviewed 317 studies suggest that when money enters the picture, it can sometimes shift the way people frame their 318 wellbeing, appreciation of nature, and motivations behind their interactions with nature 52,53 . The 319 extent of how these changes manifest varies from case to case and between different social groups. 320 Thus it is not possible to elicit a universal conclusion here. Nevertheless, in the discussion related to 321 wellbeing and CES, the usefulness of money to meet several wellbeing needs remains elusive 54 . 322 When looking these findings critically there seems to be a dominance of studies about the positive 323 contribution of CES on human wellbeing associated with transient, immediate, and significant 324 impacts. Much fewer studies have focused on possibly negative or low magnitude CES impacts. 325 Furthermore, Figure 3 visually highlights many blank areas, which indicate many possible missing 326 pathways between CES and constituents of human wellbeing via the identified mechanisms. These 327 imbalances and missing pathways could be attributable to three possible reasons. The first could be 328 publication bias, which refers to the selective publication of studies based on the magnitude and 329 direction of the results and/or the areas of interest of the authors 55 . However, due to the type of data 330 and analytical procedure it is not possible to formally test for publication bias, as is common in meta-331 analyses 56 (see limitations in Methods). Second, the missing pathways in Figure 3 linking a specific 332 CES to a particular wellbeing constituent via a specific mechanism might not exist in reality. Third, 333 these pathways might exist but have not been empirically identified in academic studies. 334 Considering the above, this systematic review seeks to provide a level of evidence and possibilities to 335 inform future research and practice at the interface of CES and human wellbeing to reduce the biases 336 in the areas that "we know", fill in the knowledge gaps in the areas that "we do not know", and hint to 337 explore the areas that "we do not know we do not know" (see Implications, recommendations and 338 future research directions). 339 340

Synergies and trade-offs between mechanisms 341
We find some consistent associations among the pathways and mechanisms through which CES 342 contribute to human wellbeing between studies. Overall, the results seem to confirm that different 343 mechanisms are more relevant to certain CES types and affect specific sets of wellbeing constituents. 344 This points to that it might be practical and beneficial to identify such synergistic relationships to 345 inform landscape and urban planning, natural resource management, and biodiversity conservation. In 346 particular we identify five assemblages representing the collections of associated pathways namely 347 "sensory affection", "learning and development", "health and leisure fulfilment", "social vibrancy", 348 and "spiritual and heritage resources". Some implications can emanate from the characteristics of 349 these assemblages. 350 First, when it comes to "sensory affection", while natural aesthetic values can enhance subjective 351 wellbeing 12,57 , authentic wilderness with disordered and frightening landscapes can also cause fear and 352 negative feelings to some people 14,40 . This could raise some interesting debates about the promotion of 353 "authentic" natural landscapes or planning "false wilderness" to reconnect people to nature 58 . 354 Second, "health and leisure fulfilment" and "social vibrancy" assemblages tend to be more prevalent 355 in human-dominated landscapes such as urban areas (Table 2). This points to the synergistic effects 356 and cost-effectiveness that urban green and blue infrastructure can have for meeting multiple needs 357 for urban residents 14,59,60 . 358 Third, related to "learning and development", the results show many similarities in how ecosystems 359 shape the way people think 61 , their choices in life 13 , and the development of their worldview and 360 cultural significance between Indigenous people and modern communities 12 . Regardless of their 361 culture and level of dependence on ecosystems for their livelihoods, this points to longstanding 362 associations between ecosystems and the personal lives of people through intimate knowledge of (and 363 adaptive integrity with) the local environment, which significantly contributes to personal growth 61 . 364 Fourth, in terms of "spiritual and heritage resources", identity appears to be the core determinant of 365 the synergies and trade-offs effects 62,63 . Thus, the inclusion of local communities' identities and 366 cultural practices can create substantial benefits for ecosystems management in areas that such 367 associations are visible 64 . 368 Finally, we identified trade-offs among three assemblages, namely "learning and development", 369 "health and leisure fulfilment", "spiritual and heritage resources" (Figure 4). Trade-offs between 370 "spiritual and heritage resources" and "learning and development" seem to be linked to religious or 371 sacred landscapes. In particular some traditional and Indigenous communities are sceptical about the 372 research potential and educational value of such areas 65 . This is often due to diverse challenges and 373 barriers (e.g. technical, perception, communication), as well as differences in values, which have 374 alienated on many occasions the active engagement of indigenous communities in the formulation of 375 ecosystem management plans in such areas [65][66][67] . 376 Another trade-off is between "spiritual and heritage resources" and "health and leisure fulfilment" 377 ( Figure 4). Such trade-offs are again mainly associated with traditional and Indigenous communities 378 for which ecosystems (and nature more broadly) invoke spiritual experiences, e.g. the Earth and its 379 elements are perceived as living entities valued for their own sake 17,42 . Sometimes tourism and 380 recreational activities that can provide leisure opportunities for improving health and subjective 381 wellbeing to tourists are sometimes perceived to violate sacred places 68 . Indeed, some studies have 382 suggested that tourism activities sacrifice spiritual and intrinsic values (e.g. sacredness and the 383 spiritual connections between the sites and people) for instrumental benefits (e.g. tourism revenue) 69 . 384 A final trade-off is between "health and leisure fulfilment" and "learning and development" (Figure 4). 385 This is usually associated with the fact that some tourism and recreational activities can alter 386 livelihoods at community level and encourage young people to leave their traditional livelihoods 37,70 . 387 The environmental degradation associated with intensive tourism combined with the risks of changing 388 livelihood structure may result in significant loss of local knowledge systems and skills 71 . In other 389 cases, the inappropriate planning of tourism activities can hinder the educational value of historically, 390 culturally, and ecologically important areas 69,72 . 391 The above findings could effectively inform practice and decision-making processes to anticipate 392 what types of human wellbeing trade-offs are to be expected in areas where such CES are provided, 393 for example due to tourism [73][74][75] or economic development 76,77 . This can guide the identification of 394 possible context-specific solutions to prevent or mitigate CES-driven trade-offs in human wellbeing. constituents to the CES that would be needed to achieve this (e.g. provide green spaces for exercise, 410 promote landscape elements with aesthetic values associated with stress release and escapism, 411 enhance spiritual fulfilment in designing urban green infrastructure). By knowing the mechanisms 412 permeating these pathways (e.g. regenerative, satisfactive, transcendentive), and comprehending the 413 most likely beneficiaries and landscapes (and landscape elements) that can deliver these CES then it 414 could be possible to inform the development of specific interventions and policies that meet these 415 objectives, and ideally promote synergistic effects while reducing trade-offs. In this sense, the 416 comprehensive systematisation of the 70 pathways of possible non-material interactions between 417 human and nature (Table S12-S13-S14, Supplementary Material)  belonging' and psychological connections to enhance 'mental health' are more likely to have a 430 stronger effect on human wellbeing outcomes rather than the 'outer' connections such as remunerative 431 connections ( Figure 3, Table S8 and S9). As CES have often received less attention in policy-making 432 progress than provisioning or regulating services 4,8 , the authors emphasise the necessity of 433 reconnection strategies that aim to influence the behaviour of individuals and alter the paradigms that 434 underpin the actions and decision-making for ecosystem management. 435 Despite the wealth of studies exploring the interface of CES and human wellbeing we identify several 436 knowledge gaps that future research should target. First, the research at this interface tends to focus on 437 individuals. While this is undoubtedly important, the fact remains that there has been less focus in the 438 reviewed studies on understanding the effect of CES on collective wellbeing. However, in several 439 studies we observed that due to trade-offs the provision of CES has improved the wellbeing of 440 individuals, but reduced collective wellbeing 80 , and vice versa 81 . Although this has been recognised in 441 the individual studies, there is a lack of multi-level wellbeing assessments, which would be necessary 442 for better assessing ecosystem services trade-offs and synergies. 443 Second, when observing the evidence imbalances and missing pathways (see Figure 3 and 'Relative 444 effects of mechanisms'), there is a need to fill in the knowledge gap in the areas that (a) 'we know', 445 (b) 'we do not know', and (c) 'we do not know we do not know'. Regarding (a), there is a need to 446 advance the currently available knowledge and address publication biases. For the former, research 447 should explore in-depth how these mechanisms manifest in the less studied ecosystems and 448 understand their differentiated effects to various stakeholders. The underlying factors mediating these 449 impacts and the drivers of changes in CES provision would also need more dedicated attention. For 450 the latter, scholars should be able to publish high quality research regardless of "uninteresting" results 451 or novelty. Low magnitude, negative or incremental impacts of CES provision on human wellbeing 452 should also be captured. Regarding (b), our work could be utilised as a summary of the current 453 research landscape, which highlights the many missing pieces that need to be found. The blank areas 454 in Figure 3 could offer some starting points to explore whether the missing pathways exist or not in 455 reality. Regarding (c), we should point that there is a possibility that more mechanisms link CES and 456 human wellbeing considering the large biological and cultural diversity across the globe, and the often 457 very tight human-nature interactions in many geographical contexts. In this sense there is a need to 458 move beyond the conventional way of thinking and upgrade research approaches and framings to 459 unravel the unknown unknowns in human-nature relationships. We hypothesise that missing 460 mechanisms could be present in ecosystem-dependent communities, and especially traditional and 461 Indigenous communities, considering their very unique relations with nature. In this sense there would 462 be a need to enhance even more the current efforts to promote the collaboration between scientists and 463 Indigenous and Local Knowledge (ILK) holders 82 . 464 Overall, following this systematic review, we argue that the fuller understanding of the complex 465 linkages between CES and human wellbeing can help navigate towards outcomes that promote 466 effectively both wellbeing and ecosystem management and contribute to meeting global sustainability 467 challenges. The conceptual framework develop can possibly move the current debate forward. diversity, (f) inspiration, (g) sense of place, (h) knowledge system, and (i) social relations. We then 492 complement this initial list with other CES found in the reviewed documents that are not explicitly 493 delineated in the MA, but identified as such in the source literature. Table S2 in the Supplementary  494 Material provides the full list of CES (and their definitions) considered in this study. 495

Similar to CES, human wellbeing is a broad and contested term that has been interpreted in various 496
ways without a commonly agreed definition 12,84 . At a generalised level, human well-being can be 497 perceived as a synergistic and multi-dimensional concept that consists of multiple constituents, which 498 when combined, they characterise the positive state of individuals 12 . Although the concept of human 499 wellbeing has drawn the attention of policymakers, researchers, and practitioners globally, there is 500 insufficient knowledge within the literature on how it is linked with the natural environment and the 501 ecosystem services it provides 84 . 502 In our systematic review we use eleven constituents of human wellbeing, most of which are adopted We need to point that although we have used these conceptualisations and typologies of CES and 509 constituents of human wellbeing to form the conceptual framework of the systematic review, we have 510 not limited the review to the studies that only used these explicitly. For example, for CES we do not 511 only review studies using the MA terminology/typology, but included studies that adopted different 512 terminologies/typologies (e.g. IPBES, CICES, TEEB). Acknowledging the slight differences among 513 terminologies [i.e. non-material nature's contributions to people (IPBES) vs. cultural ecosystem 514 services (MA)], we use these terms interchangeably in the context of this review. Similarly we 515 acknowledge that the current frameworks of the constituents of human wellbeing are imperfect and 516 that there are blurry distinctions among them 12 . 517 Considering the above this review does not seek to present an argument on the accuracy of the 518 adopted typologies, but instead focuses on covering all the studies using different terminologies to 519 ensure the widest possible capture of studies to elicit the linkages between CES and human wellbeing. 520 Our use of certain conceptual frameworks does not seek to imply the superiority of the one over the 521 other, but their functionality within this review. 522

Literature identification and selection 523
For this systematic review we identified peer-reviewed literature that report observations about the 524 contribution of CES to human wellbeing both quantitatively and qualitatively. We identified the 525 literature though Elsevier Scopus and ISI Web of Science Core Collection. We employed three 526 categories of search words that were guided by the conceptual framework presented above. The three 527 levels of search words reflect (a) ecosystems or ecosystem services, (b) specific CES, (c) human 528 wellbeing or quality of life. 529 The specific keywords are: ("Ecosystem*" OR "Ecosystem service*" OR "social-ecological system*" 530 OR "Nature's contribution*") AND ("cultural ecosystem service*" OR "aesthetic*" OR "recreation*" 531 OR "spiritual*" OR "inspiration*" OR "place attachment" OR "social relation*" OR "knowledge 532 system" OR "sense of place" OR "educational value* OR "Non-material nature's contribution*") 533 AND ("Quality of life" OR "wellbeing" OR "human needs" OR "well-being"). 534 The literature search was conducted for the literature title, abstract and keywords, and was limited to 535 peer-reviewed articles in English. The search was performed in July 2020 with no restriction on the 536 publication time frame. We followed the PRISMA principles for systematic review 87 . 537 In total, 463 articles were found in Elsevier Scopus, and 251 documents were found in ISI Web of 538 Science Core Collection. We then removed duplicates leaving 502 articles for further screening. 539 Subsequently two filters were applied. For the first round, the first author read the studies' titles and 540 abstracts to remove non-relevant literature. For the second round, the remaining articles were 541 downloaded and read by the first author in full to determine whether they met the inclusion criteria. 542 The inclusion criteria were: 543 a) The study should report cultural services provided by nature or ecosystems (i.e. non-544 ecosystem related cultural services were excluded); 545 b) The study should report CES or non-material contributions of nature (i.e. other ecosystem 546 services or material contributions were excluded); 547 c) The study should be empirical or a review of empirical studies (i.e. conceptual, theoretical 548 and simulation studies were excluded); 549 d) The study should report observed changes in human wellbeing (i.e. studies not mentioning 550 change in wellbeing were excluded) 551 e) The study documents should be Articles or Reviews (i.e. Editorials, Books, and the 552 Proceedings of conferences and meetings were excluded). 553 Among the 502 documents identified after the search, a total of 356 documents appeared to match the 554 inclusion criteria mentioned above after the first screening round. The first author then read the full 555 text of these 356 documents and ended up with 302 documents (285 empirical studies and 17 review 556 papers) that were deemed eligible for further analysis. Figure S1 in the Supplementary Electronic 557 Material contains the different stages of study selection. 558

Critical appraisal of studies 559
As systematic reviews draw conclusions based on multiple individual studies, it is necessary to 560 evaluate the reliability of evidence at the individual study level 88 . In this study we adopted a series of 561 appraisal guidelines for ecosystem services and conservation studies 88 and created a checklist for 562 assessing the reliability of the evidence contained in each reviewed study. The checklist includes 563 questions related to internal validity in terms of the research aim, data collection, data analysis, results 564 and conclusions, and design-specific aspects (see Table S3, Supplementary Material). Each study is 565 then categorised as having "very strong evidence" (score: >75%), "strong evidence" (score: 50-74%), 566 "moderate evidence" (25-49%), or "weak evidence" (<25%). 567 Overall, the quality appraisal indicated that 92.4% of all studies included in this systematic review 568 (279 out of 302 studies) are categorised as having "very strong evidence", 7.3% (22 studies), as 569 having "strong evidence", and only one study as having "weak evidence". The mean value of the 570 quality score across all studies is 83.5%. 571 To ensure the high quality of the database, while at the same time highlighting the diversity of the 572 research landscape, we include in this systematic review the broadest possible range of the studies. 573 Thus we only remove the single study with "weak evidence". Thus the final database used for the data 574 extraction included 301 studies, of which 285 were empirical studies and 16 review papers of 575 empirical studies. 576

Coding and meta-data extraction 577
Three broad categories of meta-data was extracted from each paper, and subsequently used for the 578 analysis and visualisation. Table S1 in Supplementary Material shows the summary of coding for 579 meta-data extraction. 580 The first type of meta-data reflected the general study characteristics, including the (a) site location, 581 (b) publication year, (c) spatial and temporal scale, (d) research types and objectives, and (e) the types 582 of stakeholder engagement. We used Google map to collect the longitude and latitude coordinates of 583 the studied sites for those studies that did not provide actual coordinates. We created a heat map using 584 ArcGIS version 10.5, illustrating the geographical distribution of the study sample. 585 The second type of meta-data extraction focused on the study methodologies. This includes the 586 information related to (a) data collection tools, (b) data analysis methods, (c) research framework, and 587 (d) the broad academic field. This meta-data was used to explore the interdisciplinarity and the 588 evolution of research methodologies through time using visualisation tools that illustrate the diversity 589 of the disciplines and fields represented in the reviewed studies. 590 The third type of meta-data forms essentially the central part of the analysis, and relates to the 591 mechanisms through which CES contribute to human wellbeing. Data extraction was guided by a 592 series of questions designed to unravel the mechanisms, including information related to: (a) type of 593 ecosystem, (b) type of CES, (c) observed changes in CES provision, (d) reason for changes in CES 594 provision, (e) affected group, (f) constituents of human wellbeing that CES contribute to, (g) direction 595 of the impact, (h) magnitude of the impact, (i) outcome of the impact, and (j) description of the 596 mechanism in open text. 597 These variables are both closed-ended using coded ranges and open-ended using narrative answers. 598 The former facilitates quantitative categorical analysis, while the latter facilitates the narrative for 599 qualitative content analysis. Observations of mechanisms in which CES contribute to human 600 wellbeing were extracted only from the empirical studies, and not from the review papers in the 601 authors' database (see above). 602 From the 285 empirical studies, the authors identified 1138 observations of mechanisms linking 603 different CES to different constituents of human wellbeing, which were divided as explained below. 604 The elicitation of meta-data described above was performed by the first author, in close consultation 605 with the second author on a case-by-case basis in case of inconsistencies or emerging new categories. 606 This was to allow for the consistent elicitation of the meta-data, while at the same time ensuring an 607 added lens for challenging cases. 608

Elicitation of mechanisms linking CES and human wellbeing 609
A relational content analysis was conducted for the 1138 observations of the mechanisms linking 610 different CES with different constituents of human wellbeing. The relational analysis allowed for the 611 exploration of the relationships between the concepts and the identification of themes and patterns 89 . 612 Inductive coding was applied to allow for the new concepts and narratives to emerge from the data 613 itself. Figure S2 (Supplementary Material) shows the flowchart of data analysis. 614 In order to identify the pathways linking CES and human wellbeing we conducted two coding 615 iterations. During the first coding, the 1138 observations were systematised across 231 pathways. For 616 the purpose of this study, we define a pathway as the linkage through which the provision or change 617 in a single CES affects a single constituent of human wellbeing. 618 During the second coding iteration we condensed the initial 231 pathways into 70 pathways based on 619 similarity. These are explained in greater detail in Table S12-S14 in the Supplementary Material, with 620 45 having a positive effect on human wellbeing, and 25 negatively (of which 17 are associated with 621 CES degradation/loss and 8 are disservices). For the purpose of this paper we define ecosystem 622 disservices as "the ecosystem generated functions, processes, and attributes that result in perceived or 623 actual negative impacts on human wellbeing" 90 , e.g. unwanted pests, pollen allergy, vector-spread 624 diseases, noise from wildlife, and frightening natural landscapes 91,92 . We view CES degradation as the 625 process through which natural and anthropogenic drivers of ecosystem change disrupt the provision of 626 individual or multiple CES, having detrimental impact on human wellbeing 13,46,[93][94][95][96] . 627 Further content analysis allowed the grouping of these 70 pathways into 4 channels of human 628 interactions with ecosystems, and 16 types of mechanisms. This is because although some of the 629 pathways link different CES with different wellbeing constituents, these linkages manifest in 630 relatively similar ways in their functions to how they affect human wellbeing. Of the 16 identified 631 mechanisms, 6 were adapted from another study 15 and 10 were defined by the authors. 632 Following an iterative content analysis, we systematised the evidence from the literature in this new 633 typology and conceptual framework that links CES and human wellbeing, which is the main 634 conceptual contribution of this systematic review. Finally, the authors re-coded the entire dataset using this new typology of mechanisms. The final 638 coded variables were all categorical variables, which were then used for further quantitative data 639 analysis and visualisation as outlined below. We used alluvial diagrams to visualise the frequency of 640 the mechanisms documented in the reviewed studies. 641

Quantification of the effect of different mechanisms 642
Considering that the different studies used quite different quantitative and qualitative measures for the 643 contribution of CES (or their change) to human wellbeing it was not possible to conduct a proper 644 meta-analysis. Instead we used a semi-quantitative normalisation approach that normalised the 645 contribution of CES to human wellbeing between studies. This normalisation approach relied on 646 expert judgement and followed the process proposed in Berrang-Ford et al 97 . 647 In summary, the criteria used for the normalisation were the magnitude and direction of the impacts. 648 The direction of impact was coded as: (a) positive, (b) two-way, (c) negative, and (d) not concluded.

649
The magnitude was designated as (a) high negative impact (score=-2), (b) low negative impact 650 (score=-1), (c) no significant effect (score=0), (d) low positive impact (score=+1), and (e) high 651 positive impact (score=+2). 652 The data for the magnitude of impact (score -2 to + 2) was extracted in two steps. First, this came 653 from the text of each study where the studies' authors explicitly indicated whether the impacts are 654 high or low as perceived by their respondents. Second, for the studies that did not clearly clarify the 655 impact magnitude, we used expert judgement to provide the scores using a series of criteria, namely 656 depth, scale, and speed (Table S11, Supplementary Material) 97 . Any observation, that met one of the 657 conditions for high depth, scope or speed, was classified as a high magnitude impact. 658 Upon calculating the impact scores for each observation, we grouped together the observations 659 belonging to the same pathway linking a particular CES to a constituent of wellbeing. Three matrix 660 heat maps were produced to show (a) the average impact scores for these mechanism, and (b) the 661 frequency of their presence in the reviewed studies. 662

Latent Class Analysis and Multiple Correspondence Analysis 663
We performed Latent Class Analysis (LCA) to identify possible CES synergies. LCA is a statistical 664 tool that allows for the analysis of multivariate categorical data to identify the latent classes based on 665 similar patterns 98 . In this study we use LCA to identify the CES assemblages through unobserved or 666 "latent" classes 99 . We extracted eight unweighted variables that characterised the observed 667 mechanisms including (a) type of ecosystem, (b) type of CES, (c) channel of interaction, (d) affected 668 group, (e) constituent of human wellbeing, (f) type of mechanism, (g) direction of impact, and (h) 669 magnitude of impact. Table S1 in the Supplementary Material provides the actual codes of these 670 variables. 671 The analysis was conducted using the open access PoLCA R package. We conducted the analysis for 672 up to 6 classes, re-estimating the model until identifying the maximum likelihood solution. The 673 Bayesian information criteria (BIC) were used to determine the appropriate number of classes to 674 select. Detailed information of the BIC, sensitivity, specificity and accuracy tests can be found in 675 Table S8 in the Supplementary Material. The general patterns and characteristics of each class were 676 drawn to characterise the underlying ecosystem services assemblages and synergies. 677 We have to point here that we introduce this notion of assemblages, as CES (and the pathways 678 through which they contribute to human wellbeing) also tend to interact with each other in complex 679 ways. By the terms "CES pathway assemblages" we refer to subsets of the 231 identified pathways 680 linking CES and wellbeing that appear significantly related and interactive. Each assemblage tends to 681 contain a set of pathways that likely link a specific set of CES and contribute to a specific set of 682 human wellbeing constituents via some explicit mechanisms with some potential synergies and trade-683 offs 1 . We loosely adopt this concept of assemblages from philosophy 100 to emphasise the complexity 684 and the relationship between the part (single CES, single pathways) and the whole (assemblage of 685 CES and pathways). In this sense while a single CES via a single pathway can influence 686 autonomously human wellbeing, when these interact in the assembled whole they create synergies, 687 trade-offs and a dynamic whole. 688 We then performed a Multiple Correspondence Analysis (MCA) to supplement the results of the LCA 689 and to explore further the trade-offs and synergies among specific sets of variables. The MCA method 690 could be seen as a generalisation of the Principle Component Analysis (PCA) when the analysed 691 variables are categorical instead of quantitative 101 , which is applicable to our dataset. Through the 692 MCA we produced plots that summarise and display the relationships between categorical variables 693 by calculating the chi-square distance between the categories of the variables and individuals 102 . We 694 conducted the MCA with subsets of variables to investigate a more explicit correspondence among 695 ecosystems, users, and their wellbeing. 696 For all analyses, we presented and interpreted the first two dimensions as the eigenvalues decrease 697 regularly with small difference after the third dimension 101 . We filtered results by selecting variable 698 categories with higher contributions to a dimension, which exceeded the expected average value. The 699 MCA and related visualisations were performed through the FactoMineR packages in R software 103 . 700 701

Challenges and limitations 703
Despite its extensive focus, the systematic review presented in this paper has a series of limitations. 704 These include the (a) non-inclusion of grey literature, (b) keyword selection, (c) quantification method 705 and synergies/trade-offs analysis, and (d) publication bias. 706 Regarding (a), the systematic review included only peer-reviewed literature and excluded grey 707 literature. The authors consciously made this decision to ensure the reliability and reproducibility of 708 the results. We are aware that a large fraction of the documents reporting the benefits people obtain 709 from ecosystems are not peer-reviewed papers. This is because practitioners and government agencies 710 that implement relevant projects are less likely to write academic papers about their actions. 711 Furthermore, most of the relevant knowledge linking CES and human wellbeing from Indigenous and 712 local communities is not found in peer-reviewed papers 7 despite its importance for understanding 713 human-nature relations 104 In this sense while this systematic review can indicate the current scientific 714 evidence about the pathways linking CES and human wellbeing, it should not be taken as the totality 715 of the evidence about these linkages. 716 Regarding (b), even though this review uses a wide range of keywords, these terms were confined to 717 the conception of CES and human wellbeing in the broad fields of ecosystem services and 718 biodiversity conservation. In this sense it was not possible to include all keywords related to possible 719 constituents of human wellbeing and the interaction between humans and nature brought up in other 720 fields and disciplines such as sociology 105 and psychology 106 . With that in mind the authors carefully 721 considered and refined all search terms based on the prevailing terminologies in the field. Although 722 we believe that the search terms allow for the very good identification of the research landscape and 723 related trends at the interface of CES and human wellbeing, we also acknowledge that the keyword with green spaces and biodiversity more intensively, which in turn, increases their attachment and 734 inclination towards nature 34 . Due to the complexity of the quantification such feedback loops were not 735 captured in our analysis. Finally, the normalisation approach used for the quantifying the effects of 736 CES on human wellbeing was based on expert judgement. Although we followed an established 737 approach 97 and very clear criteria (  12,28 Benefits to physical health through interaction with nature include (a) lower body mass index, (b) reduced prevalence of disease, (c) reduced obesity, (d) lower somatisation level, (e) decreased cognitive decline, (f) reduced blood pressure, heart rate and muscle tensions, (g) improved immune system, (h) increased restoration and healing, and (i) lower mortality risk 12 . 9 Remunerative The economic benefits people obtain from ecosystems through non-material benefits in cash or other forms of money For many indigenous communities, particular species carry a special cultural heritage value that can be utilised for exchange and trades among kins to sustain the reciprocal relationships essential to their functioning 114 13 Transcendentive The benefits that lie beyond the ordinary experiences and the regular physical realm, more often associated with religious or spiritual values through interaction with nature Spiritual value Spiritual practice Spirituality Many people and communities experience ecosystem-inspired feelings related to "entities larger than themselves" 115 . For some people being in natural settings makes them appreciate people's connection to all things in the Universe 17 .
14 Apprehensive The anxious and fearful feeling generated from interaction with nature.

Aesthetic
Form Certainty, sense of control and security Some people are afraid of their safety when encountering certain natural features via visual or auditory interactions, such as scary animals, dangerous predators, animal blood, and areas that are dark with high tree cover, among others 12,14,40 15 Destructive The direct damages caused to health, relationships, finance, and capability (among others) due to interaction with nature of the loss of CES Note: The six mechanism indicated with (*) are adapted from previous study 15 , while the other 10 mechanisms are defined by the authors. For each of the 753 mechanisms we provide as an example a unique pathway of CES-mechanism-constituent of human wellbeing. It should be noted that some mechanisms 754 mediate more CES-human wellbeing connections. A comprehensive explanation of the different pathways for each mechanism can be found in Table S12-14  755 in the Supplementary Material. Due to certain overlaps the positive (No. 1-13) and negative (No. 14-16) mechanisms are presented in alphabetical order 756 rather than some other taxonomy.

Channel of interaction
Type of mechanism Constituent of human wellbeing Figure 3: Relative contribution of individual pathways to human wellbeing. Each box or empty space represents a unique pathway of a single CES to a single constituent of human wellbeing via a single mechanism. For simplicity we list all of the 16 mechanisms explained in Table 1 for each CES (y-axis) and cross map them to each constituents of human wellbeing (x-axis). Boxes indicate a unique combination of CES-mechanism-human wellbeing found in the literature (i.e. in the 1138 observations divided into 231 unique pathways during the first iteration of the coding; see Methods -Elicitation of mechanisms linking CES and human wellbeing), while blank spaces indicate unique combinations not found in the literature. The size of each box represents the number of studies that captured the specific pathway. The colour of each box represents the average effect of the specific CES to the specific constituent of human wellbeing via the specific mechanism after normalising each relevant observation with a score of -2 to +2.