**4. Discussion**

The results show a rise in research interest in this area. In particular, the growing numbers of articles, especially since the year 2018, as well as the trend line confirm the increasing significance of this topic (see Figure 2). Already in the first four months of 2019, more than 50% of the previous year's total number of publications on the investigated topic had been published. The review still confirms the finding of Haase et al. [14] that most studies are carried out in European cities. Due to the higher numbers of publications in the last two years, it can be assumed that the topic is still evolving as a research field and that more studies will be published in the future.

Referring to key question (1), we found that Provisioning ES are the least assessed ES section for urban green infrastructure types (see Table 1). This also tallies with the finding of Haase et al. [14]. Unlike the ES sections Regulation & Maintenance and Cultural, Provisioning ES can generally be physically transported (e.g., deliveries of foodstuffs to stores). For this reason, cities mainly import provisioning ES from the surrounding countryside and even further afield. The demand for provisioning ES, however, is very high in densely populated areas. Our review indicates that the number of assessments

of provisioning ES has been increasing over the last six years. This trend can perhaps be attributed to new initiatives fostering urban agriculture such as urban gardening, Edible Cities, etc. In today's society, we can identify changes that are serving to highlight the issue of Edible Cities or the role of urban gardening [32]. In the foreseeable future, provisioning ES could become an important element of the urban environment. Moreover, with regard to promoting the sustainable development of cities and the reduction in urban ecological footprints, Gómez Baggethun and Barton [13] have pointed out that ES should not be imported from distant locations. Concurring with the review by Haase et al. in 2014 [14], Regulation & Maintenance were the most frequent forms of ES investigated in the publications (see Table 1). This ES section contains a number of ES classes that play a special role in securing the well-being of inhabitants, although they are generally only perceived when lacking. Such ES are mostly used indirectly and are in grea<sup>t</sup> demand [33]. One reason for the high frequency of these ES assessments may be the current lively debates on climate change and its impact on cities. Cities are facing increasing environmental pressure as well as concerns about human health and the well-being of local residents [34]. Regulation and Maintenance ES can have a considerable impact on human well-being by regulating the microclimate, air pollution or water flows [35]. There is an increasing demand on cities to plan and develop important green structures to counteract the negative impacts of heat islands, heavy rainfall, air pollution, etc. As we also found in our review, more and more research has been conducted over the last few years into these ES classes in urban areas [14]. Our results show that the Cultural ES section is the second most investigated in the reviewed publications (see Table 1). This fact is unsurprising if we remember that the Cultural section contains ES classes, which are commonly referred to as "recreation services". Other studies have shown that ES classes related to recreational aspects (such as 3.1.1.1 and 3.1.1.2) are the most frequently investigated cultural ES within urban areas and especially within urban green spaces or parks, where this ES is mostly directly used [14,36]. Other publications have considered some less obvious small green infrastructure types; Mathey et al. [37] and Pue ffel et al. [38], for instance, have conducted surveys on the recreational ES provided by urban brownfields (Table 2). Contrary to the findings of Haase et al. [14], our results show that many studies assess multiple ES. Those studies that evaluate only one class of ES, such as Takács et al. [39], Marando et al. [40] or Lehmann et al. [41], generally deal with Regulating & Maintenance ES. A closer look at these studies shows that they are largely based on primary data drawn from on-site inspections and include biotope mapping or climatic field measurements.

In regard to key question (2), our findings show that previous ES assessments have looked at a wide variety of urban green infrastructure types at larger and smaller spatial scales. The majority of examined urban structures considered the city-dimension (Figure 3). Taking a closer look at the studies dealing with site-dimension, we found that many applied time- and cost-intensive methods, in particular self-collected (primary) datasets (e.g., [42–44]). Consequently, the small number of studies at the site-dimension can be attributed to the requirement for more precise and site-dependent data, usually primary data collected in the field. The high number of publications that assessed ES in "urban green spaces" and "structural types" (Figure 3) can be explained by the aggregation of di fferent spatial structures, in particular treating diverse green spaces as one undi fferentiated study object (e.g., [42,45,46]). On the other hand, this frequency can also be attributed to the data used in the individual articles. Many of the articles took land-use data as a basis for their assessments, resulting in an investigation of di fferent land-use types that usually exist in aggregated classes (e.g., green spaces, streets, buildings, etc.). In addition to this, the frequently examined green infrastructure type "parks" (Figure 3) is one of those structures mostly associated with the urban context; we can thus expect a large number of investigations in such spaces. Many diverse analyses can be carried out in parks, which can also be investigated in regard to direct usage by local residents, especially within the section of Cultural ES. In contrast to parks, for example, which are present in almost all cities, the smaller number of ES assessments of green infrastructure types, such as "forest" or "green roof/wall" (Figure 3), can be attributed to their rarity in many cities. Consequently, ES assessments of such green infrastructure types are less frequent.

Based on this review, we identified four main motivations for assessing the ES of urban green infrastructure types on di fferent scales. Most studies based their investigations on the need to successfully implement ES in urban planning and for the development of appropriate measures to preserve and protect these ES (e.g., [46–48]). Others, such as Czembrowski et al. [49], justified their investigations by pointing out the services and value of urban green areas as well as the need to facilitate communication between environmental experts and decision makers. Another stated intention in assessing ES is to ensure the inclusion of user demands in the planning and design of cities and green spaces [33]. Studies such as Mathey et al. [37] aimed to increase the acceptance of less popular green structures such as brownfields by the public as well as by urban planners. This is achieved by highlighting the values of associated ES and striving to prevent the automatic redevelopment of urban brownfields, instead maintaining these as areas of green space.

In a large number of publications, green spaces were generally considered in an aggregated form and thus examined in a larger scale study, i.e., defined as "urban green structures" according to our classification (Figure 2). For such green spaces, the studies particularly investigated the Regulative ES. On the other hand, studies examining single and smaller sites (site-dimension) most often focused on spatial structures that have a stronger impact at small-scale levels, such as the usually high perception and significance of parks within a district.

While previous studies have particularly highlighted economic valuation methods to assess ecosystem services (e.g., [50,51]), we only identified a small number of articles applying "economic valuation methods" (Figure 4). Regarding key question (3), the most frequently used approach is the "spatial proxy method". This method is particularly applied to the assessment of ES in regard to "structural types", "urban green spaces" as well as "trees". As in the previous section, this observation can be explained by the adopted base data. In most cases, the land-use types of a study area provide the basis for the application of di fferent proxies, as performed in the studies of Speak et al. [52] and Kremer et al. [53], for example. Such studies used proxy variables such as land cover maps and other geoinformation system (GIS) datasets that depict special ecosystem processes as a basis for ES assessments. One reason for this high number of studies applying spatial proxy methods can be limited time and financial resources available to researchers as well as a lack of primary data. However, spatial proxy methods have been criticized for the overgeneralization of obtained results that represent only a fraction of reality [54]. On the other hand, they allow for the investigation of ES in study regions that lack primary data while o ffering the advantage of good comparability of results and generally low costs [54]. Only a few of the reviewed studies applied "social media-based methods" (Figure 4). The inclusion of "modern" (in particular social media-based) datasets such as photographs in Richards and Tunçer [55] is still an under-researched and rather new field in regard to ES assessments. In the near future, this method could potentially become more popular due to the continuous expansion of social media and uptake of mobile technology by the public as well as for research purposes (e.g., citizen science). However, there still exist several limitations in the use of such social-media-based datasets and methods. For example, it can be argued that the real world is not being accurately represented as data generation is currently mainly carried out by young people [55]. The finding that many studies use a mix of methods (Figure 4) agrees with previous work by Seppelt et al. [27]. "Model-based methods" such as "I-tree Eco", a free peer-reviewed software to assess the ES of trees provided by the USDA Forest Service, were applied by some studies, especially for the assessment of Regulating and Maintenance ES. In these cases, the particular issue of interest was indeed "trees", such as in the studies of Parsa et al. [56] and Baró et al. [57]. These models use primary data (in these cases, tree inventories) and apply benefit transfers under location-dependent variables such as climatic parameters. This can also limit the accuracy of results. Therefore, while such investigations can be compared in their underlying methodology, at the same time, they are based on generalizations, especially with regard to ecological parameters, rendering them di fficult to transfer to other geographic patterns. Although the popular "I-tree Eco" model also provides results expressed in monetary values and thus integrates economic valuation methods, it is surprising, that most of the

reviewed studies using this model are limited to the presentation of biophysical results [35,52,58–60]. This results in a lower number for "economic valuation methods" in Figure 4. Only a few studies also discuss the economic results provided by "I-tree Eco" [56,57,61].

Regarding our research design, it should be noted that the criteria used in this study placed restrictions on the literature review and thus led to a smaller sample size. In an initial search, in which we searched for the term "ecosystem services" instead of "urban ecosystem services", we received a high amount of results. After reviewing some of the results, we found that a large part did not match our search scope. Therefore, we decided to narrow the search terms to explicitly "urban ecosystem services" in order to reduce the large number of non-relevant articles identified after the initial scan of the databases. We are aware that our narrow selection meant that some potentially relevant studies, which assessed ES in cities but did not explicitly use the term "urban ecosystem services" (in article titles, keywords or abstracts), could not be included. Nevertheless, all studies that refer to "urban ecosystem services" are included in this review. A further restriction results from only searching for the keyword "green infrastructure" (apart from the di fferent green and blue elements) and with this not explicitly integrating other planning concept terms that are sometimes synonymously used especially in the context of regulative ES, e.g., Sustainable Drainage Systems (SuDS), Low Impact Development (LID), or the term Nature-Based Solutions (NBS). An investigation of the underrepresented blue elements and the inclusion of the mentioned concepts represent an interesting field for further research. In this context, it can also be expected that the results are increasingly reflecting studies from European cities, as the term "green infrastructure" is commonly used within the EU member states.
