Convergence in Perceptions of Ecosystem Services Supports Green Infrastructure Decision-making in a Semi-arid City

Effective management of cities using ecosystem services from green infrastructure (GI) requires explicit consideration of the linkages between provision of services and ecosystem service demands (i.e., governance priorities). Identification of stakeholder knowledge and objectives in GI decision-making contexts with respect to ecosystem services may improve urban planning; yet this information is rarely explicit in local contexts and cases. We address this gap by surveying environmental stakeholders and practitioners to investigate how perceptions of ecosystem services influence GI practice in Tucson, AZ. Results indicate that the semi-arid environment and urban design led to prioritizations that focus on water sustainability and urban heat mitigation. We found strong agreement in environmental perceptions between different management sectors. We observed matches (as well as mismatches) between the ecosystem service priorities and important environmental issues. Ecosystem services prioritization revealed a unique classification of ecosystem services that reflects stakeholder priorities. Our findings suggest the study of ecosystem services supply and demand can inform local urban management. These findings from a semi-arid city further suggest that understanding stakeholder knowledge, perceptions, and priorities should be important for cities in other regions where GI is being implemented as an environmental solution to provide ecosystem services.


Introduction
Green infrastructure (GI) is an approach to managing stormwater at its source that also delivers many environmental benefits or ecosystem services (USEPA 2019). Governments, organizations, and researchers are promoting an expansion of GI implementation to enhance ecosystem service provision to address issues of runoff reduction, air quality, microclimate and urban heat islands, and public health issues (Berkooz 2011;Livesley et al. 2016;Meerow and Joshua 2017;Mell 2016). In urban areas, GI is implemented in different forms (such as urban trees and forests, green roofs, rain gardens, etc.), each with differing abilities to provide ecosystem services (Ellis 2013;Gill et al. 2017;Pugh et al. 2012;Raje et al. 2013). A notable benefit of GI to city planners is its multifunctionality, or the ability to combine multiple functions and using limited space more efficiently to provide ecosystem services to diverse stakeholders, compared to gray infrastructure (Ahern 2011;Connop et al. 2016). The multifunctionality of GI does require consideration of trade-offs and the balance of service supply and demand in planning and design (Hansen and Pauleit 2014).
Despite the growing interest in GI, there are limits and challenges associated with GI in practice. For instance, stakeholders are skeptical of the ability of practices to provide the level of ecosystem services expected and if investing on GI will provide those benefits (Copeland 2014). There is a gap in "locallyspecific" application of ecosystem services in planning and governance that accounts for differences in climates and suitable plants between regions (Kabisch 2015;Koo et al. 2019). In developing best management practices in planning, planners and managers need to identify ecosystem services to address a wide range of stakeholder perspectives, however these perspectives may be poorly understood (de Groot et al. 2010). To meet the growing complexity and scale of ecological challenges, there are demands for higher levels of stakeholder engagement in developing solutions for these challenges (Young and McPherson 2013). Understanding the perceptions of local actors such as planners at city and local level, decision-makers, and local stakeholders to identify relationship between local demands and provision of locally relevant ecosystem services is essential in locally complex situation (Kabisch 2015;Koo et al. 2019).
To effectively manage cities using GI ecosystem services, the linkages between ecosystem services provision (ecosystem service supply) and governance priorities (ecosystem service demand) need to be explicit across sectors and spatio-temporal scales (Burkhard 2014;Klug and Jenewein 2018). Urban ecosystem service supply and demand have been well documented across spatial and temporal scales, primarily through mapping and geospatial approaches (Dobbs et al. 2014;Larondelle et al. 2014;McPhearson et al. 2013). While mapping approaches to ecosystem services have shown the spatial variation and ability of urban spaces to provide ecosystem services, for services to be effective in guiding planning and decision making, there is a need to connect to local governance practice and evaluation across sectors (Opdam 2013). Cortinovis and Geneletti (2018) showed a clear gap in application of ecosystem services in planning despite existing methods for mapping and evaluation, and proposed an approach that explicitly considers goal setting, multifunctionality, and service demand as a way to enhance the role of ecosystem service science in urban planning (Cortinovis and Geneletti 2018). Such an approach would address criticisms of models such as the ecosystem services cascade (Haines- Young and Potschin 2018), that give primacy to ecological structures and functions over perceptions and socio-cultural values that govern ecosystem service provision (Spangenberg et al. 2014;Zoderer et al. 2019).
Despite the need to consider perceptions and values in the balance of ecosystem services supply and demand, few studies investigate both supply and demand (Baró et al. 2017;Schirpke et al. 2019). Consideration of both supply and demand informs management by identifying potential mismatches between scales of supply and demand (Castro et al. 2014;Wei et al. 2017). Furthermore, interests, priorities, and needs differ among groups (Díaz et al. 2011;Geijzendorffer et al. 2015;Martín-López B et al. 2012;Martín-López J et al. 2012;Wei et al. 2017), and identification of how stakeholders perceive ecosystem services supply and demand and any potential conflicts can allow for more effective policy and management decisions (Zoderer et al. 2019). One of the potential challenges in implementing the ecosystem service concept are disconnects in how universal frameworks (MEA 2005;TEEB 2010;Wallace 2007) align with local and practical perceptions and framings of ecosystem services (Crossman et al. 2013;Seppelt et al. 2012).
The degree of ecosystem service knowledge transfer from research to policy and decision-making is low, despite the growing number of studies such as developments of tools and models, landscape planning, and increasing awareness and communication (Haase et al. 2014;Luederitz et al. 2015). Although general recommendations can be derived from such studies for land management and planning, they are less likely influence practice without involvement of local stakeholders (Haase et al. 2014). Yet, stakeholders' viewpoints on ecosystem services supply and demand are poorly described or limited to a few services (Baró et al. 2015;Cumming et al. 2006). Understanding how the landscape can be managed with respect to the multifunctionality of GI helps with understanding of the potential outcomes for policy and management and it can be achieved by involving the diversity of stakeholders to evaluate supply and demand (Kremer et al. 2016). In addition, stakeholder perceptions of ecosystem disservices (Lyytimäki and Sipilä 2009) can impact decisions about GI in practice as they affect perceptions of trade-offs. Thus, identifying stakeholder knowledge and objectives in decision-making contexts with respect to the transfer of ecosystem services may improve landscape planning and decision-making in urban areas (Haase et al. 2014;Luederitz et al. 2015).
This paper aims to answer, how do perceptions of ecosystem services influence environmental stakeholders' practice of GI in Tucson, AZ? To answer this, we used an online survey of managers and practitioners in Tucson, AZ to specifically ask: (1) what is stakeholder's knowledge of environmental and water-related issues in Tucson, (2) what ecosystem services stakeholders want from (demand) and think that different types of GI provide (supply), and (3) what connections exist between stakeholders' perceptions of environmental and water issues and their priorities (practice). Our results imply that the unique water resource challenges of the semi-arid environment connect to stakeholders' prioritization for GI. We found several matches between priorities and practices with respect to ecosystem supply, demand, and disservices. However, we also observed some mismatches between stakeholder ecosystem service priorities and their view of environmental issues that may be due to a scale mismatch. In this case, GI is implemented at local scales, but some concerns such as groundwater recharge is a challenge that is managed at larger spatial scales.

Study Area
We focus on Tucson, AZ (including the metropolitan area extending into Pima County, AZ) and how those places are using GI to address water sustainability. Tucson is a semiarid city, with very hot summers (temperatures reaching over 38°C) and average annual precipitation around 300 mm with summer monsoon rains accounting for half of the annual precipitation (Pessarakli 2019). Growing population in the area (Pima County was just over 1 million in 2018 (Bureau of Reclamation 2019) has increased demand for potable water. The city of Tucson is dependent on allocations from the Colorado River through the Central Arizona Project (CAP) and its own aquifer (Central Arizona Project 2017; Kuhn et al. 2017). Outdoor water consumption accounts for 45% of potable water consumption in Tucson and the primary outdoor water usage is for landscaping irrigation. There is a desire to increase green spacing in the area that requires water for irrigation. Considering the population growth and impact of climate change, the current system might not be able to provide that water need for irrigation in the future (Kuhn et al. 2017). On the other hand, Tucson receives annually more rainfall than the total demand that if harvested and used at its origin can cover notable portion of potable water that is used for outdoor activities (Korgaonkar et al. 2018;Kuhn et al. 2017).
Stormwater is managed to deal with flooding from monsoon rains primarily using streets to convey water out of neighborhoods. However, there is growing interest in harvesting stormwater flows to meet outdoor irrigation and revegetation goals and as a solution to water resource challenges in the Tucson region (Radonic 2019). Hydrologic simulations suggest water availability at the lot scale during monsoon events to meet the outdoor demands in drier months (Korgaonkar et al. 2018). Currently, there are a number of water-harvesting facilities in public, commercial, and residential places that capture runoff through GI. Water harvesting is considered a strategy that is beneficial both on the demand and supply side (Brooks 2006). Tucson considers rebates to the resident for such strategies called "Water Conservation Rebates" (City of Tucson 2019). Ongoing research seeks to identify connections between GI design and ecosystem services (Luketich et al. 2019;Pavao-Zuckerman and Sookhdeo 2017), and to evaluate water harvesting programs and policy motivations and effectiveness (Elder and Gerlak 2019; Radonic 2019).

Participants
We conducted a questionnaire between April 2019 and June 2019 to gather information about stakeholder perceptions of GI ecosystem services. We selected the stakeholders to represent critical managers and practitioners of environmental governance, and practice, and water harvesting and GI management in Tucson and Pima County. This includes representatives from city-level and county-level government and agencies, environmental utilities, non-profit organizations, and engineering and design firms. The questionnaire was distributed a total of 117 stakeholders, and we had a response rate of approximately 54% or 63 respondents.

Survey design
We conducted the survey through Qualtrics (Qualtrics Labs Inc 2009). In this survey, three types of GI were included: rain barrels and cisterns, rain gardens, and urban trees ( Fig. 1). Although rain barrels and cisterns are householdscale GI, the perceptions of practitioners and managers are important for developing design standards, policies, and incentives for residential implementation. The survey included both closed-ended and opened-ended questions in five sections. In the first section, the stakeholders' general and specific knowledge of Tucson's current and future environmental situation and knowledge of GI were investigated. The statements asked whether there is a need to substitute other sources for freshwater, control stormwater, recharge groundwater, and improve the quality of surface water. We also asked about the efficacy of current strategies for water management in Tucson, such as the delivery of water from the CAP to deal with water supply. Also, there were statements on the effect of future conditions such as climate change and population growth on water resources. General environmental knowledge was determined by asking the respondent to indicate on a 5-point Likert scale (1 = strongly disagree-3 = Neutral-5 = strongly agree) the level of agreement regarding Tucson's needs on waterrelated statements and usefulness of GI in providing water. This was followed by more specific questions on their familiarity with types of GI and their effectiveness in runoff capturing methods.
In the second section, stakeholders were asked on how important each type of GI is in providing various types of ecosystem services. We selected ecosystem services from Millennium Ecosystem Assessment (MEA 2005) relevant to water harvesting and water use through GI. Respondents were provided 15 types of ecosystem services and they were asked to rate, on the scale of 1-5, how important each type of GI is to provide the listed ecosystem services (ecosystem service supply). Ecosystem services supply was determined by asking stakeholders to answer a 5-point Likert scale (1 = not important, 2 = slightly important, 3 = important, 4 = fairly important, 5 = very important) on the level of ecosystem services each type of GI (urban tree, rain garden, and cistern) can provide. Again, respondents were provided the same 15 ecosystem services. Stakeholders were asked to prioritize GI ecosystem services that they think are important for GI to provide, representing service demand. Respondents ranked the 15 ecosystem services from most to least important. In the third section, the stakeholders were asked to assess ecosystem disservices that the three types of GI may cause from the stakeholders' point of view and any strategies that their offices might use to mitigate the concerns. Here, respondents were asked to rate on the scale of 1-5, how concerned they are with 10 ecosystem disservices for each type of GI. In the fourth section, the respondents were asked on a 5-point Likert scale, what they think about the environmental challenges and concerns currently in this area (i.e., flooding, environmental justice, poverty, property value, etc.) related to how GI could be useful to address the challenges. In the fifth section, we asked questions about the stakeholders' professional background and training and demographics.

Analysis
Descriptive statistics were used to determine central tendencies and frequencies for responses to knowledge questions. For general environmental knowledge, negatively worded questions were used as well as positively worded questions to reduce acquiescent bias. Reverse coding was used to remove mismatch for the level of agreement and the increasing scale. Thus, for those negatively worded statements, responses expressed disagreement interpreted as agreement with high level of environmental knowledge. As a result, higher scores on the 5-point Likert scale means a higher level of general knowledge on Tucson situation and efficacy of GI. Chi-squared test of associations (Rea and Parker 2014) was used to determine the relationships between level of knowledge and demographic variables of sector, role in the office, duration of employment, expertise, degree, major, duration of living in Tucson, duration of living in southwest, participation in low impact development (LID) conference, LID member, and if the office required to follow stormwater discharge for small municipal separate storm sewer system (Appendix 1).
A cluster analysis was performed in Minitab 18, (Minitab Inc. 2020) using complete linkage method. Cluster analysis uses a distance matrix to group factors that are different from other groups and identifies homogenous clusters when the grouping is not known. In this study, clustering was used to see the differences among stakeholder rankings of the 15 ecosystem services. The results are displayed as a dendrogram to highlight perceived groups of services.
A cumulative weighted score was used to compare ecosystem services ranked by stakeholders by demographics and professional background. Cumulative weighted score is the average of a set of scores where each set carries a different amount of importance regarding the score each stakeholder allocated to ecosystem services (from 1 to 15). Each score is calculated as shown in Eq. 1: ω is the rank value (1-15), x is the score stakeholders assigned to each ecosystem services (1-5), N is the total number of stakeholders that ranked the service, and S is the cumulative weighted score.
Kruskal-Wallis and Mann-Whitney U-tests were used (due to non-parametric data; Corder and Foreman 2011) to assess relationships between demographic variables and other response variables: environmental knowledge, ecosystem services supply, ecosystem services demand, ecosystem disservices, and severity of environmental concerns in Tucson (Appendix 1). The comparison of between means was carried out using non-parametric Tukey test to determine difference between type of GI (Zar 1984) (Fig. 2).

Results
We used a survey of environmental managers and practitioners in Tucson and Pima County to determine how perceptions of stakeholders about ecosystem services influence environmental priorities and practice of GI. Below we describe responses that characterize stakeholder knowledge and perception of environmental and water issue, perceptions on ecosystem services supply and demand, prioritization of services with respect to environmental issues, and perceived ecosystem disservices of GI.

Stakeholder Knowledge and Perception of Environmental and Water Issues Knowledge
Due to current water sustainability issues in the region and critical needs to substitute water resources in near future (Guo 2017), we evaluated stakeholder knowledge of the current water system in Tucson. We asked stakeholders if they agree or disagree with statements about water resources in the Tucson region. We did not find differences in knowledge and perception of environmental issues and challenges among professional sectors. We found stakeholders strongly agreed with statements designed to explore knowledge on water resource and environmental issues that are unique to the semi-arid setting of Tucson. For example, most of the respondents indicated that there is a need to substitute other sources for freshwater (87%, n = 55, M = 4.53, SD = 0.91) (percentage indicates the percent of stakeholders that strongly agree and agree with environmental and water issues). Furthermore, the majority of respondents stated that the current water supply alternatives (i.e., CAP delivery) (71%, n = 45, M = 2.25, SD = 1.30) and groundwater extraction (90%, n = 57, M = 4.52, SD = 0.86) are not sustainable ways to deal with water supply issues. Stakeholders expressed the need for alternative water resources and insufficiency of current water management strategies (82%, n = 52, M = 1.81, SD = 1.04). Moreover, the stormwater control (92%, n = 60, M = 4.67, SD = 0.74) and groundwater recharged (95%, n = 60, M = 4.65, SD = 0.72) were perceived as highly important priorities.
Stakeholders rated the severity of the general environmental challenges concerns that GI might be a potential solution for in Tucson (Fig. 3). Groundwater depletion, urban heat islands, and flooding are the environmental concerns ranked highest by stakeholders (Fig. 3). We assume that these ratings of concerns related to stakeholder practice. Groundwater depletion (80%, n = 57, M = 4.56, SD = 0.86) was the strongest environmental concern (percentage indicates the percent of stakeholders that ranked the environmental issue as important and very important). Flooding (68%, n = 57, M = 4.2, SD = 0.94), urban heat island and severe heat waves (68%, n = 57, M = 4.06, SD = 1.01), environmental justice (51%, n = 57, M = 4.1, SD = 1.067), and water quality (56%, n = 7, M = 3.91, SD = 1.03) are also among the highest concerns of stakeholders. Concerns about wildlife habitat, biodiversity, and property values were ranked lowest by the stakeholders.
We asked about stakeholders' perception of the efficacy of different types of GI to serve as an alternative water resource. The majority of respondents stated the importance of GI as a supplemental or even primary water resource (89%, n = 56, M = 4.28, SD = 0.93) and 95% of stakeholders expressed implementing GI as a helpful approach to deal with future challenges (n = 60, M = 4.72, SD = 0.68). A large number of respondents (75%, n = 47) indicated that infiltration and retention are among the most effective ways of capturing stormwater. Stakeholders perceived rain gardens as the most cost-effective means of water harvesting (54%, n = 34). Cisterns and rain barrels were perceived as the most difficult to install and maintain mostly due to the initial cost, maintenance cost, change in the quality of water, and lack of technical knowledge in installation and maintenance (56%, n = 35). We did not find any significant associations between stakeholder demographics (including professional sector, major, degree, duration of employment, duration of living in Tucson and southwest, role in the office, being a low impact development member, attendance of low impact development conferences, and gender) and environmental and water issues knowledge (Appendix 1- Table 2).

Stakeholder Perceptions on Ecosystem Services Supply and Demand
As a measure of ecosystem service supply, we asked stakeholders to rate the potential for different GI types to provide ecosystem services. Stakeholders indicated that water harvesting for future use, stormwater reduction, and education opportunities are the strongest ecosystem services provided by cisterns (Fig. 4). Rain gardens are viewed to provide more pollinator support compared to cisterns and urban trees. Urban trees are perceived to be more effective in air quality regulation, carbon sequestration, moderation of extreme heat events and urban heat island.  We asked stakeholders to rank their demand ecosystem services and they expressed a higher demand for urban heat island mitigation, stormwater reduction, water harvesting, and biodiversity enhancement (Fig. 4). To investigate the connection between ecosystem services supply and demand, we selected the highest ranked demands and compared the ratings for different types of GI (Table 1). Stakeholders rated urban trees as the most effective approach for moderation of extreme heat events and urban heat island reduction. They perceived rain gardens to be the most efficient GI for reducing stormwater, but cisterns were the most efficient for water harvesting and storage. Additionally, both urban trees and rain gardens were perceived as important factors for biodiversity enhancement.
We found that stakeholders in different sectors have different ecosystem service prioritizations (Fig. 5). While this survey focuses on GI practitioners in Tucson and thus is not a statistically randomized sample, interesting trends emerge between sectors when we compare the weighted average of each ecosystem service ranked by each sector/type of stakeholder (Fig. 5). For example, engineering firms valued stormwater reduction and water harvesting more than aesthetics compared to environmental utilities. Environmental utilities on the other hand, valued regulating services such as air and water quality compared to other sectors (the results of environmental utility sector must be considered with caution due to the small sample size). Stakeholders in city and county level governments valued urban heat island and energy usage reduction and moderation of extreme heat events more than other sectors. Stakeholders in non-profit organizations valued various ecosystem services without a clear top priority (Fig. 5).

Classification for Stakeholders' Perception
We used agglomerative hierarchical clustering to distinctly classify similar ecosystem services prioritized by stakeholders (Fig. 6). This revealed a unique classification of ecosystem services in Tucson, AZ relative to established schemes, such as the MEA (MEA 2005) and is indicative of local stakeholder priorities (Fig. 6). The water-related factors (water harvesting and storage, water quality, stormwater reduction, groundwater recharge, and water quality) grouped into one cluster and heat-related factors (moderation of extreme heat event and urban heat island reduction) also grouped into one cluster (Fig. 6). These two clusters are the most important services expressed as critical demands of stakeholders reflecting the unique urban and environmental conditions in Tucson (Fig. 4). Environmental quality factors (air quality and soil quality) were also grouped. Pollinator support and enhancement of biodiversity grouped into one cluster. Finally, aesthetic values, enhanced property values, and education opportunities were grouped (Fig. 6) and seem to be less prioritized by stakeholders (Fig. 5). We found no significant correlations between GI practice (as indicated by the priority of stakeholders) and their perceived environmental challenges and concerns.

Ecosystem Disservices
Stakeholders ranked the ecosystem disservices of GI and reported maintenance costs, installation and maintenance time, obstruction of views (and links to safety), and potential health risks as the most important GI disservices (Fig. 7). Stakeholders were concerned with different ecosystem disservices of different GI types (Table 2) when we compared the mean importance of the top five disservices (Fig. 7) for the three types of GI. Maintenance cost was the strongest disservices among the stakeholders and rain gardens are perceived to have the highest cost among other types of GI. Rain gardens are perceived to have the most time-consuming installation and maintenance. Urban trees have unique concerns as obstruction of views and the production of leaf litter, and damage to physical property, with less perceived health (Fig. 7-Table 2). Cisterns are thought to have slightly higher health risks.

Discussion
We investigated stakeholders' environmental knowledge, supply and demand for GI ecosystem services, and connections between stakeholder perception and practice Significance level a *** *** * *** *** (prioritization) among various environmental managers and professionals in the Tucson region. We found that stakeholders have a high level of knowledge reflecting the semiarid setting of Tucson and that there were no significant differences in environmental perception between sectors. We observed both matches and mismatches between the ecosystem service priorities and important environmental issues. Clustering approaches to classify similar ecosystem services prioritized by stakeholders revealed a unique classification of ecosystem services indicating stakeholder priorities that differed from established schemes (MEA 2005). These findings on stakeholder knowledge, perceptions, and priorities in Tucson, AZ provide implications for other regions where GI is being implemented as an environmental solution to provide ecosystem services.

Stakeholder Knowledge about Environmental Challenges and Concerns
We found stakeholders have a high level of knowledge of water-related environmental issues, reflecting the semi-arid setting of Tucson. We did not find differences in knowledge and perception of environmental issues and challenges among professional sectors. The degree of knowledge regarding current water-related issues in the Tucson region (Fig. 3) indicates that the respondents support the need for water resource alternatives and believe that current water strategies cannot meet demand. Our findings also confirm that practitioners view GI implementation as a multifunctional approach to deal with current urban water management problems. This perception reflects recent trends towards rethinking stormwater runoff as a novel water source, rather than as a waste product (Walsh et al. 2012). In fact, in Tucson harvested runoff at the lot scale may be sufficient to meet potential demands for outdoor irrigation (Korgaonkar et al. 2018). Residents in Tucson are already beginning to use harvested water for landscaping (Radonic 2019); however, barriers still exist to widespread adoption of GI for water harvesting purposes (Staddon et al. 2018). Urban water management needs transformative change in practice that addresses technical, economic, social and institutional challenges (Brown et al. 2009;Ferguson et al. 2013; Wilfong and Pavao-Zuckerman 2020) Such a fundamental change requires identification of the socioenvironmental benefits of stormwater as a resource (Walsh et al. 2012). Understanding stakeholder perceptions can promote systems thinking for water management and support such transformations in perception (Rebekah R Brown 2005; Lee and Yigitcanlar 2010). Despite the potential for disciplinary influences, we did not find differences in perceptions of environmental issues among professional sectors, nor did we find significant associations between environmental knowledge and factors such as, professional sector, education level, and duration of living in Tucson and southwest (Appendix 1- Table 2). This high level of knowledge on important current and future environmental issues shows the potential for agreement among stakeholders and movement toward an integrated approach to GI implementation (Cousins 2017), and the potential for fewer conflicts among sectors managing water-related issues. This high level of consensus on issues and also on the multifunctionality of GI could lead to novel strategies such as cost-benefit sharing among different sectors despite different backgrounds in order to reach a reliable and sustainable water resources management and associated socio-ecological benefits (Paavola 2016). In fact, solving current challenges of GI planning likely requires collaboration among different professionals to establish new cross-disciplinary collaboration approaches that are supported by system thinking (Hansen and Pauleit 2014).

Perceptions of Ecosystem Services Supply and Demand
Stakeholder knowledge about challenges and concerns for GI revealed the need for water resources alternatives and the inefficiency of current practices. The first step to address this issue using GI is to identify how experts view ecosystem services supply and demand. Stakeholders thought that the ecosystem services and disservices of different forms of GI were different (Figs. 4, 7 and Table 1). For example, stakeholders perceive urban trees and rain gardens as more effective than cisterns for moderation of extreme heat events and urban heat island mitigation ( Fig. 4 and Table 1). Stakeholders ranked high demand for the supply of moderation of extreme heat events, urban heat island mitigation, stormwater reduction, water harvesting and storage, and enhancement of biodiversity (Fig. 4). Significance level a *** * *** *** * All ecosystem disservices are ranked in Fig. 7 a Statistical differences between GI types within the column at *p ≥ 0.05; ***p ≥ 0.001 significance values Urban trees have previously been found to be the most effective and least costly method of urban heat island control, corroborating stakeholder opinion in this case (Norton et al. 2015;Solecki et al. 2005). However, to support decision making for GI planning, awareness of urban trees' potential benefits might not be enough to motivate planting and quantification of those benefits and the feasibility of implementation are crucial. We also found that stakeholders may have concerns about the cost of implementation and maintenance of urban tree cover (Fig.  7). For example, cisterns, the GI form that was thought to be the most effective at meeting the most important demands (Fig. 4) was also mentioned by the majority of respondents as one of the more difficult type of GI to install and maintain (Table 2) and is also perhaps not a costeffective means of water harvesting. The ultimate effectiveness of GI and its costs and benefits have uncertainties from both biophysical and socio-political factors that need to be characterized in order to better inform management decisions (Ding et al. 2015;Liu et al. 2016).
Knowledge of ecosystem services supply and demand can help to select the right GI to address the priorities of stakeholders vis-á-vis environmental issues in the Tucson region. The capacity of an ecosystem to provide ecosystem services differs from the actual services delivered to the society Nedkov and Burkhard 2012;Villamagna et al. 2013). GI may produce certain regulating, provisioning, cultural, and supporting services, but it is critical that the GI services provided match stakeholder demand. Thus, selecting the right type of GI that best matches the stakeholders need was aimed in this study. Our results imply that the ecosystem services that stakeholders want more are more likely to be provided by rain gardens and trees than by cisterns themselves (Fig. 4). In future planning and incentive programs (City of Tucson 2019), passive GI should be promoted or paired with active harvesting to meet ecosystem services demands.
We found strong familiarity and knowledge of environmental issues for stakeholders engaged in policy, design, maintenance, and implementation of GI. This might suggest a relatively low barrier to adoption of GI. However, knowledge and perceptions of managers might not be enough to overcome barriers go GI adoption, as policy and management options are also inhibited by the views of residents (Gartin et al. 2010). While residents' knowledge of GI types and its effectiveness for water management is typically lower than that of stakeholders (Baptiste 2014;Baptiste et al. 2015;Maeda et al. 2018;Turner et al. 2016), residents' knowledge and behavior can play a role in adoption that can be addressed with bottom-up policy and education approaches to enhance implementation (Maeda et al. 2018). Moreover, the interaction between residents' and practitioners' knowledge plays an important role in improving the quality of urbanized watersheds. This is especially important with GI such as rain gardens and cisterns, as they are mostly implemented in the residential areas but have high technical and maintenance considerations (Fig. 7). In addition to knowledge, there might be differences in goals (demand) and functionality (supply) of ecosystem services from point of view of stakeholders and residents. Although there may be differences in priorities (demand) and functionality (supply), interactions between stakeholders and residents can help bridge the effective supply and demand connection. Future work should address these differences and could also promote stakeholderresident interactions and partnerships.

Linking Stakeholder Knowledge and Perception to Practice
Stakeholder perception of ecosystem services supply and demand revealed the potential for GI to address water resources and environmental challenges. Bridging the knowledge, perceptions, and practice of stakeholders can help link past practices and future needs. We aimed at filling the current gap in application of the concept of ecosystem services of GI by connecting the ecosystem services to local governance. Our results suggest that in some instances, there is a strong connection between environmental perception and ecosystem services goals and priorities for GI (Figs. 3, 4). For example, urban heat island mitigation and flooding effects were rated as high priorities for ecosystem service provision and as significant environmental challenges (Figs. 3, 4). This connection may arise because stakeholders across sectors and with various backgrounds likely interact with similar issues in a similar way due to the basic and tangible needs of a semi-arid region facing water scarcity (Lau et al. 2018). Our respondents are generally stakeholders who work closely to address these common issues and are engaged with GI. Lau et al. (2018) found that in a resource management setting dealing with direct subsistence needs, that differences in perception and practice may be reduced. In our semi-arid setting, it is likely that water resources function similarly as a provisioning ecosystem service and set up convergences in perceptions and practice between different sectors and groups. Recognizing how different stakeholders perceive and prioritize ecosystem services is a vital step for effective ecosystem servicebased approaches (Daw et al. 2015;Sikor et al. 2014). Implementing ecosystem services in environmental planning requires connection to local governance perception and practices in order to be effective in guiding planning and decision making (Opdam 2013), which can be achieved through active research collaboration (Palo et al. 2016).
Surprisingly, we found some mismatches between stakeholders' perceptions of environmental challenges for the Tucson region and their demand for GI ecosystem services (Figs. 3, 4). For example, most stakeholders stated that ground water depletion is the most critical problem in Tucson (Fig. 3) yet did not prioritize recharge as a GI ecosystem service. Biodiversity enhancement was rated among the highest service priorities but was perceived as one of the least important environmental issues (Fig. 3). These mismatches in priority and practice may also reflect stakeholder views on multifunctionality of GI and the notion that GI is better at providing other ecosystem services in this case (Fig. 4a), despite GI being thought generally to provide groundwater recharge services (USEPA 2019). Differences in scales of management may also explain the mismatch between knowledge and practice of stakeholders. Often, scale mismatches appear between the scale that stakeholders have influence on and scales that ecological process occurs (Cumming et al. 2006;Lambin et al. 2006). In Tucson, GI is managed and installed at the lot and neighborhood scale, while groundwater depletion occurs and is managed at the watershed and city scale. One way to solve this mismatch is the collaboration of local actors (i.e., stakeholders involved in neighborhood planning and site design) to form a comprehensive governance and decisionmaking system at the larger scale (Bergsten et al. 2014;Termeer et al. 2010;Pelosi et al. 2010). A management system integrated across scales and that considers stakeholder demands, priorities, and shared knowledge systems in landscape management can both reduce mismatches and potential conflicts among stakeholders (Zoderer et al. 2019). This integration may have its own challenges such as cost and time to form the integrated system.
Clustering approaches to classify similar ecosystem services prioritized by stakeholders revealed a unique classification of ecosystem services that indicate stakeholder priorities (Fig. 6). This classification differs considerably from established schemes that include provisioning, regulating, cultural, and supporting services (MEA 2005). Here, water sustainability and heat mitigation are among the most critical priorities in the Tucson region (Figs. 4,5). We argue that service classification systems (de Groot et al. 2002;MEA 2005;Wallace 2007) should be applied with care when used in planning and management, as the local attributes of cases can influence how stakeholders perceive groups of services. Ecosystem services show complex patterns of utilization and perceptions by receivers and managers of services potentially limiting generalized classification schemes for local management and planning (Boyd and Banzhaf 2007;Fisher and Turner 2008;Fisher et al. 2009;Wallace 2007). Research collaborations to reveal stakeholder perceptions of ecosystem service clusters may be needed apply classification schemes for local environmental contexts.
Evaluating how stakeholders perceive ecosystem services supply and demand can assist the implementation of local knowledge and perceptions into management and long-term planning (Klug and Jenewein 2018;Luederitz et al. 2015). However, as we observed (Figs. 3, 4), mismatches between priorities of stakeholders and important environmental issues can arise due to varying spatial scales of management. As suggested by Burkhard et al. (2014), appropriate institutions should oversee the spatial and temporal scales that match with ecosystem services supply and demand (Burkhard et al. 2014). In the case of GI, focusing on appropriate types of GI to fulfill stakeholders' specific goals at both local and regional scales can help work against mismatches.
Urban ecosystem services that are effective in managing environmental challenges can be achieved by connecting supply to the demand and priorities of stakeholders. Overall, the study of ecosystem services supply and demand of various types of GI can help to identify these priorities. For example, here we identify ecosystem services that would be supported by passive vs. active GI approaches, which can then be selected to align with identified desired services. Moreover, identifying diverse stakeholder priorities for ecosystem services can foster partnerships and coalitions between agencies and organizations to meet diverse and multifunctional ecosystem services goals (Scarlett and Boyd 2015). In Tucson, such partnerships have emerged between stakeholders as GI policy has adapted and evolved over time in response to learning among policy agents and entrepreneurs (Gerlak et al. 2021).
The goal of achieving ecosystem service multifunctionality requires connecting research with planning and policy in practice (Olander et al. 2017;Luederitz et al. 2015).
Here, identifying perceptions of environmental concerns and ecosystem service demand among stakeholders shows the potential for collaboration and management of environmental assets. This is especially the case where multifunctionality of such assets can be demonstrated. Our work indicates that integration of stakeholder perceptions can serve as an important link between ecosystems and social values (Olander et al. 2017). It is through the engagement of stakeholders in ecosystem services research that knowledge and objectives related to those values becomes available to improve landscape planning and decision-making in urban areas (Luederitz et al. 2015). Participatory and collaborative approaches, such as science-policy dialogs (Pavao-Zuckerman and Gerlak 2019), can facilitate dialog among stakeholders to further bridge perceptions into practices to enhance ecosystem services in cities.

Conclusion
This study connects stakeholder perception and priorities of ecosystem service supply and demand with the role of GI to address local water and environmental challenges. Here, we demonstrate convergence in perception of environmental challenges and desired goals from GI. Despite this convergence, there are still challenges on the path to implementing GI regarding how it provides the level of expected ecosystem services and how services connect to the specific priorities of stakeholders. Often this requires a careful consideration of the appropriate and expected spatial and temporal scales that ecosystem services should be managed at. Here, ecosystem services prioritization revealed a unique classification of ecosystem services that reflected local stakeholder priorities. Our findings suggest the study of ecosystem services supply and demand can inform local urban management through an integrated understanding of stakeholder knowledge, perceptions, and priorities.