In order to implement the concept of ecosystem services, various frameworks distinguish between the different components of ecosystem delivery (Bastian et al. 2013, Potschin and Haines-Young 2011, Villamagna et al. 2013). We distinguish between the ecosystem service capacity and flow based on the definitions in Villamagna et al. (2013). The ecosystem service capacity is defined as the maximum potential of ecosystems to deliver an ecosystem service based on the biophysical and social characteristics. The capacity of cultural ecosystem services may be more difficult to quantify, in comparison to provisioning and regulatory ecosystem services, as it varies across the landscape as a function of the range of biophysical (including land use and cover, climate, habitat and topography) and social (e.g. site management and accessibility) features (Villamagna et al. 2014, Villamagna et al. 2013). The ecosystem service flow is defined as the actual use of the service, which can be measured directly as the amount of a services delivered or indirectly as the number of beneficiaries served. The ecosystem service flow is different from the ecosystem service demand as the latter is a measure of the amount of ecosystem service required or desired by society and, for cultural ecosystem services, may include measures of the number of people wanting to experience that service (Villamagna et al. 2013).

Within this study, we use geocaching data as a proxy for recreational ecosystem services in a multifunctional landscape (Table 1). The distinction between the capacity and flow is particularly useful given that availability of different forms of geocaching data (Cord et al. 2015) and, namely, information about the availability of geocaches within landscapes as a measure of the ecosystem service capacity and the actual ecosystem service flow measures as the number of quests and number of favourite points. Through a spatial analysis, this study assesses the influence of the different site characteristics, such as the predominant land use, accessibility and landscape aesthetic value, on the capacity of ecosystems to provide recreational ecosystem services measured through geocache point data. Similarly, an analysis of the factors impacting on ecosystem service flow (i.e. how much is delivered) is also carried out. In addition, similar to Cord et al. (2015) and following on the approach in this study, we have used an adapted questionnaire with local geocachers in order to obtain further information on the choices of geocachers and to assess the degree of attraction for different geocache environmental characteristics.

Within this study, we have adapted the methodology used by Cord et al. (2015) to the multifunctional landscapes of the study area (Balzan et al. 2018). In a multifunctional landscape perspective, ecosystems serve more than one purpose, are considered in terms of their interactivity and their role in leading to improved well-being in different ways. Multifunctional landscapes can be thought of as a heterogeneous matrix that results from tangible interactions between natural and cultural systems (O’Farrell and Anderson 2010, Selman 2009). Within this study, we have therefore considered all geocaches, matching the inclusion criteria identified in this study and considered different types of caches across the cultural landscapes of the study area independent of the LULC category. Previous research has shown that the attraction of the cache tends to be very strongly associated with the attractiveness of the landscape (Telaar et al. 2014), suggesting the importance of an analysis that considers the entire landscape matrix (sensu Selman 2009). Hence, we consider all geocaches within Maltese Islands in order to obtain a measure of the (natural and social) features of the landscape that permit the delivery of this service.

The island state of Malta is situated in the Central Mediterranean Sea at 96 km south of Sicily and almost 300 km east of Tunisia. With a surface area of around 316 km², the archipelago consists of three inhabited islands (Malta, Gozo and Comino) and several uninhabited islets. The first evidence of human settlement in Malta dates back to around 7000 BP (Patton 1996) and the landscapes of the Islands have been moulded over millennia by human exploitation, resource use and the geoclimatic conditionswhich harbour considerable biodiversity (Schembri 1997). Today agricultural land cover occupies around 51% of the territory, whilst built-up, industrial and urban areas occupy more than 30% of the Maltese Islands (M.E.P.A. 2010). Malta has a population density of 1,346 persons per km² (N.S.O. 2014), the highest in the European Union and a growing tourism industry which, in 2016, surpassed 1.6 million tourists (N.S.O. 2017).

Recent research within the area of study has indicated the important role of ecosystems for the delivery of the key ecosystem services and that semi-natural and agricultural habitats are associated with a high ecosystem service capacity within this island environment. However, the actual use of ecosystem services (flow) was higher in small green urban and peri-urban areas. Coastal habitats were associated with cultural ecosystem services, in the form of habitats of conservation value (ecosystem service capacity) and aesthetic value (ecosystem service flow), but green urban areas and urban environments were also identified by survey respondents as having a high aesthetic value and for the provision of recreational ecosystem services (Balzan et al. 2018).

The influence of land cover and geocache type on the number of quests and favourite points was assessed through the use of a generalised linear model (GLM) with a negative binomial distribution due to the overdispersion of the count geocache data. The time since the placement of the geocache was included as an offset vector in order to account for different levels of visitation arising from the variation in time since the geocache was set up (Venables and Ripley 2002) (Appendix 3, Suppl. material 1). Subsequently, generalised linear mixed models (GLMM), with a negative binomial distributon, were used in order to assess the influence of categorical and continuous fixed variables on the number of quests and favourite points, whilst keeping geocache type as a random variable (Fournier et al. 2012, Skaug et al. 2013). Given the influence of urban land cover on the quest and favourite points data (Results), the GLMM for protected areas, including both the geocache type and the intersecting land use land cover of the geocache,were included as a random variable. The structure of the random variable for the model with the lowest AICc value was selected. The significance of the fixed variables was tested through elimination of the least significant effects or interaction, using likelihood ratio tests in which deviance with and without the term in the models were compared using a χ² test.

All spatial operations and statistical analysis were carried out using R language and environment for statistical computing (R Development Team 2016), whilst QGIS 2.18 Las Palmas geographic information system was used in order to visualise spatial data (QGIS Development Team 2016).

A total of 318 geocaches from the study area were investigated in this study (Fig. 1). The geocaches were distributed in five cache type categories, with traditional caches (n=231) and mystery caches (50) being the most common geocache types (Fig. 2a). Most of the caches were located in the urban LULC category (n=111, Fig. 2b). Geocache type and LULC category were significantly associated with the number of quests and favourite points but no significant interactions between these two categorical factors were recorded. Urban caches were associated with a significantly higher number of quests in comparison to marine and other terrestrial LULC categories (Table 2), while marine caches had a significantly higher number of favourite points when compared to urban areas. When using the reclassified urban atlas land use spatial data, the geocaches were distributed amongst agricultural and semi-natural areas (168), roads (41), industrial (34), marine (22), urban high density (15), urban low density (13), urban green, sport and leisure areas (11) and other (12). The highest cache density was recorded in woodland areas, but the number of caches in this category was low, followed by orchards and shrub communities, sparsely vegetated land and grassland land cover (Fig. 3a). The EEA Urban Atlas land use categories were used to provide a clearer indication of the type of urban land uses of geocaches. Urban green and sports and leisure areas had the highest geocache density, after forests which accounted for only one geocache and a small fraction of the study area (Fig. 3b). The number of quests and favourite points also differed between different land use categories (Fig. 4). The type of urban and industrial land use had a significant impact on the number of quests and favourite points (Table 3). Geocaches in urban low density and agricultural land uses and the marine environment had a lower number of quests. Weaker associations were generally recorded from the GLMM for the number of favourite points but geocaches in green urban, sports and leisure areas and the marine environment had the highest rating while mines and construction sites were associated with the lowest number of favourite points.

Figure 1.  

A land use land cover (LULC) map of the Maltese Islands study area and the location of the investigated geocaches according to the cache type.

a

b

Figure 2.

Number of geocaches according to the (a) cache type and (b) land use land cover (LULC) category.

a: Number of geocaches according to the cache type.  
b: Number of geocaches according to the land use land cover (LULC) category.  

a

b

Figure 3.

Relationship between cache density and the land use and land cover when using (a) a land use land cover (LULC) map generated using Sentinel 2 satellite images and (b) the EEA urban atlas land uses for the study area.

a: Relationship between cache density and the land use and land cover when using the land use land cover (LULC) map generated using Sentinel 2 satellite images  
b: Relationship between cache density and the land use and land cover when using the EEA urban atlas land uses for the study area.  

a

b

Figure 4.

Mean number of (a) quests and (b) favourite points (± standard error of the mean) for the reclassified Urban Atlas land use categories

a: Mean number of quests (± standard error of the mean) for the reclassified Urban Atlas land use categories  
b: Mean number of favourite points (± standard error of the mean) for the reclassified Urban Atlas land use categories  

In order to assess the contribution of gene pool protection ecosystem services in geocaching activities, the influence of sites designation as protected areas on recreational ecosystem services was analysed. Nationally designated areas had a higher geocache density (2.3 caches/km²) in comparison to sites that did not have any designation (0.4 caches/km²). Most of the caches found in designated areas were located in areas of high landscape value and in Special Areas of Conservation and Special Protection Areas (Table 3-1, Appendix 3, Suppl. material 1). The influence of protected area cover on geocache quests and favourite points was assessed using GLMM. No significant difference was recorded for quests and favourite point data between areas designated as protected areas and those which do not have any designation. Similarly, we analysed the association between the number of geocache quests and favourite points and areas protected for their landscape value. Results indicate a higher number of quests and favourite points in area protected for their landscape value, but only favourite points data was significantly positively associated with these landscapes (ꭕ²=7.84, p=0.005 for favourite points; ꭕ²=3.4 and p=0.065 for geocache quests data; the difference in deviance of the two compared models following backward elimination and its significance using a ꭕ² test are shown).

Most caches were located in proximity to residential areas and the road network but distance from public transport network and the coast appears to follow a normal distribution (Appendix 3, Suppl. material 1). The number of quests and favourite points were significantly negatively associated with increasing distance from the national road and public transport networks, indicating that accessibility of the geocache location may be an important factor in determining the preference for the geocache (Table 4). Similarly, the number of quests and favourite points were negatively associated with increasing distance from the coastal environment, which, in previous studies, has been shown to be positively associated with experiential use of ecosystems and the aesthetic value of the natural environment (Balzan et al. 2018). The number of geocache favourite points was positively associated with increasing distance from residential areas, indicating a higher number of favourite points in areas located less closely to these zones.

The analysis of geocaching data demonstrates that most geocaches were located in, or in close proximity to, urban land cover. However, the highest geocache density was not recorded in the urban land cover but in semi-natural categories and in urban green and sports and leisure areas, indicating that these environments have a higher capacity to deliver recreational ecosystem services to urban communities, as measured through the use of geocaching data. Accessibility to semi-natural and urban green areas appears to be a particularly important characteristic influencing the capacity of recreational ecosystem services and most of the geocaches were located in close proximity to residential areas and the road network. Similarly, recreational ecosystem services flow data also appear to be strongly linked to the accessibility of the environment. This was also confirmed from questionnaire data for geocache placers, where proximity to work or residence areas was considered as being important for the respondents and these also explained that their geocaches were placed in sites considered as being of personal interest or in urban green areas.

Results presented here are supported by those in Cord et al. (2015), whose study found the highest total number of geocaches and highest geocache density in urban green spaces. Similarly, in a study carried out in the United Kingdom, sub-urban and urban sites were associated with higher recreation potential compared to rural sites (Radford and James 2013) while, in another study in Barcelona, urban areas had the lowest recreational ecosystem services capacity but the highest recreation flow was recorded in forest areas surrounding urban settlements (Baró et al. 2016). Accessibility was also a key consideration when mapping the outdoor recreational ecosystem service potential (Paracchini et al. 2014) and was also identified as one of the most important characteristics of recreational sites in another study carried out in Finland (Yli‐Pelkonen 2013).

Similar to other studies investigating recreational ecosystem services (Yli‐Pelkonen 2013), geocaching was seen by geocache hunters as an opportunity to experience nature, spend time with family and friends and getting physical exercise. Caches located in urban areas, as well as in urban green and sports and leisure areas, had the highest number of quests and urban green and sports and leisure areas were also characterised with the highest favourite rating, demonstrating a preference for these caches. Proximity to the coastal environment and landscapes of higher conservation value were also associated with a higher number of quests and favourite points, whilst increasing distance from residential areas was associated with a higher number of favourite points. These results were also supported by those obtained from the questionnaires with geocache hunters, as geocaches that enable a ‘nature experience’ were also favoured by geocache hunters. These identified woodland, urban green areas, the coast and beaches as the environments in which they log most caches, whilst most of the geocache placers participating in the questionnaire in this study placed their geocaches in woodland, urban or coastal areas and in urban green spaces. These ranked coastal areas and urban built and green spaces higher in terms of their potential to provide a positive recreational experience to the geocache hunter. Similar results have been obtained by previous studies investigating the motivations of geocachers (Cord et al. 2015, Telaar et al. 2014), while other studies have similarly shown that the coastal environment is associated with increased recreational ecosystem services flow (Balzan et al. 2018, Paracchini et al. 2014).

This study contrasts with previous observations from the study area highlighting the importance of semi-natural habitats for aesthetic (cultural) ecosystem services (Balzan et al. 2018). Similar results were also obtained by Radford and James (2013), who, in their study, found a lower aesthetic value of highly urbanised environment. In the study by Balzan et al. (2018), aesthetic ecosystem services were associated with areas having a higher number of habitats of conservation interest from the study area. In contrast, within this study, designated protected areas were not significantly associated with increased flow of recreational ecosystem services, demonstrating that there are different factors affecting cultural ecosystem service capacity and flow (Radford and James 2013) and indicating the importance of identifying suitable indicators that directly reflect the different types of cultural ecosystem services and associated benefits, in the implementation of ecosystem service assessment and mapping activities.

The strong positive association of recreational ecosystem services with landscape value demonstrates that processes which essentially degrade the landscape are also likely to have a negative impact on the capacity and flow of this service. This is particularly important in the context of strong urbanisation and tourism development trends within the study area and, in general, within the Mediterranean region, leading to intense pressure on ecosystems as a consequence of land use change. This is exacerbated within the study area as a consequence of the limited land resources available within an insular Mediterranean environment, in particular as traditional management practices decline and are replaced by more intensified management associated with strong tourism and industrial development (Aretano et al. 2013, Petrosillo et al. 2013, Tzanopoulos and Vogiatzakis 2011).

Results obtained here demonstrate that the development of an understanding of the spatial variation in availability of green infrastructure and of ecosystem service capacity and flows is critical in order to implement the ecosystem service concept for the development of policies that promote the sustainable use of ecosystem services (Schröter et al. 2014). Urban areas do not necessarily provide fewer ecosystem services compared to other regions, as urban green infrastructure, such as tree cover or peri-urban agriculture, can significantly contribute to support biodiversity and ecosystem service delivery (Dennis and James 2016, Larondelle and Haase 2013). However, in this case, the high recreational ecosystem service flow in green urban areas appears to be unbalanced with the capacity of these spaces to deliver these services given the relatively low number of caches in these environments. This imbalance between capacity and flow would be expected to be higher in areas with higher population density and lower availability of urban green infrastructure (Balzan 2017, Tratalos et al. 2007). Similarly, in a recent study measuring accessibility of green urban areas, several Southern European cities had below-average availability and accessibility of green urban areas to the city inhabitants (Kabisch et al. 2016). These observations demonstrate the need to develop our understanding of composition and spatial configuration of landscapes, especially in terms of the availability of green spaces and the delivery of ecosystem services contributing to human well-being and the importance of policies and urban planning practices that maintains and improves the availability of green infrastructure.

To provide guidance in the choice of appropriate ecosystem service assessment and mapping methods, tiered approaches have been proposed, with tier 1 approaches being dependent on readily available data, tier 2 includes more specific information from the study area whilst tier 3 approaches are more data intensive and are based on an improved understanding of the underpinning processes (Grêt-Regamey et al. 2015, Gret-Regamey et al. 2017). This study has used different types of data to assess recreational ecosystem services and demonstrates that practitioners can use such an approach to develop a more detailed understanding of the links between recreational ecosystem services and the socio-ecological conditions and to predict areas with higher (or lower) urban recreational ecosystem service capacity and flow, which is necessary for the preparation of ecosystem services maps. Such approaches move away from traditional proxy-based (tier 1) approaches that are based on more easily available land cover data, which are common in the assessment of cultural ecosystem services (Fagerholm and Palomo 2017) and are often constrained by the availability of data. An improved understanding of the relationship between the components of the ecosystem service delivery chain and their relationship with socio-ecological factors, permits the development of more precise process-based (tier 3) ecosystem service assessments and mapping approaches (Grêt-Regamey et al. 2015, Lavorel et al. 2017). These are considered as being more useful in providing answers to urban planners and for the evaluation of management measures (Grêt-Regamey et al. 2015, Gret-Regamey et al. 2017).