Methods

Data Collection

This study was carried out using a mixed-mode approach. A mixed mode study combines a qualitative component, such as interviews or focus groups, and quantitative data retrieved through a questionnaire (Leedy & Ormrod, 2005). Mixed mode studies are frequently suggested as an ideal approach because they can have lower costs, reduce coverage error, improve response rate and reduce nonresponse error, and reduce measurement error (Dillman, Smyth, & Christian, 2009, pp. 302-306). In this study, focus groups constituted the qualitative portion of the research. A mailback questionnaire randomly distributed in the project's study areas constituted the quantitative component of the research. The study areas in this project were the cities of Bend, Eugene, Portland, and Springfield, Oregon. Because Eugene and Springfield, Oregon, are adjacent cities, we considered them as a single metropolitan area for the purposes of our study. All research activities were reviewed and approved by the Oregon State University Institutional Review Board (IRB Study #5039).

Qualitative component.

Focus Groups. The research team conducted focus groups to help develop background information, stimulate new ideas, and learn about how participants thought and talked about urban forest ecosystem services (Berg, 2007).

Recruitment. We recruited two categories of focus group participants: natural resource professionals and non-governmental stakeholders. We recruited prospective participants via purposive sampling given that we needed to speak with people who could be expected to be knowledgeable and interested in issues related to urban forests and ecosystem services. We initially contacted, by email or phone, natural resource professionals with whom members of the research team had established relationships. For natural resource professionals we initially contacted by phone, we briefly described the project, discussed their possible participation in a focus group, and asked them to suggest names of their colleagues who we could invite to participate. We used an invitation email template for all prospective participants, both those we contacted by phone initially, and those for whom the email invitation was the initial contact. We advised those we contacted by phone during our conversation with them that they would also receive the email invitation at the same time as other invitees. The email invitation template described the project, its goals, and how the individual we were contacting could contribute in their capacity as a natural resource professional. The email soliciting participation was resent to those who had not responded approximately three weeks after the initial email invitation. Those who did not respond to the second email contact were not contacted again. For professionals who responded and agreed to participate, we continued communication, via email or phone, to further discuss the project, and to coordinate dates when the greatest number of participants would be available. We contacted about 15 to 20 resource professionals in each of the study areas (Bend, Eugene-Springfield, and Portland). We contacted no more than 20 resource professionals per study area reasoning that we could realistically expect that about half would likely participate. Under that assumption, we expected to obtain about ten focus group participants from among the natural resource professionals we contacted. Average group size for professionals was eight individuals.

Non-governmental stakeholders were contacted using a similar strategy. We asked our professional contacts to recommend citizens from local environmental or community organizations. Professional contacts from each area (Bend, Eugene-Springfield, and Portland) provided us with three to five names and accompanying email addresses or phone numbers of non-governmental stakeholders. We contacted these individuals primarily via email using a similar template as that used to recruit professionals (with slightly altered verbiage to reflect the different role of non-governmental stakeholders in resource management). In cases where only a phone number was provided for a non-governmental stakeholder, we phoned that potential participant and used verbiage from the email template as a guide in the conversation; no script for initial phone contact was developed. We reasoned that a more unrehearsed conversation would be more natural and better received by members of the public than a pre-written script. Having established contact with a potential non-governmental participant, we asked that person if s/he could recommend other non-governmental stakeholders we could contact. As we had done with the resource professionals, we made two attempts at direct contact with a non-governmental stakeholder, after which no further attempts were made. Contacting a large number of non- governmental stakeholders proved difficult as individuals we spoke with frequently had little or no contact information for others and many of those we attempted to contact did not respond. In order to remain on schedule, we proceeded with a relatively smaller number of non- governmental stakeholders. We coordinated with both groups to establish a time and day that was most convenient for the largest number of potential participants, and then formally invited prospective participants via another email that contained time and location information. Average group size for non-governmental stakeholders was about three.

Focus Group Process. In Portland, Oregon, the focus groups were conducted in the offices of a regional governmental agency building, and in Eugene-Springfield and Bend, Oregon, in the office space of parks and recreation departments. For logistical reasons, focus groups in each city were organized to occur on a single weekday, with professionals participating in the morning, and non-governmental stakeholders meeting in the afternoon.

We conducted all focus groups using customary methods (Krueger & Casey, 2009). Both groups were asked identical questions pertaining to urban forest ecosystem services, using the same wording, with some slight variation in introductory statements given the differing roles professionals and non-governmental stakeholders have concerning natural resource management. The first and second author conducted each focus group, adhering to a script (Appendix 5) that guided the flow of the meeting and contained interview questions. The first author acted as a moderator for the groups, providing instructions for focus group participants. The first author asked pre-written questions from the script, pre-written probing questions to ensure coverage of topics of interest to the study, and unscripted probing questions to participants' responses when clarification was required. The second author assisted by taking notes and asking follow-up questions for further clarification as well. We asked both groups of participants questions that covered areas including how they defined urban forest ecosystem services, how important they thought each service was, what they felt management goals for urban forests ought to be, and what they thought were indicators of successful management. Focus groups were audio recorded and lasted about an hour.

Quantitative Component.

Sampling. We obtained our list of Bend, Eugene, Portland, and Springfield residents using zip codes and state issued driver's licenses or identification cards. First, we retrieved zip codes for each city from the U.S. Postal Service website. Then, using those zip codes as the geographical criteria for selection, we obtained names and mailing addresses from Oregon Department of Motor Vehicles (OR DMV) records for people 18 years of age or older in possession of a state driver's license or identification card. We did not include zip codes for PO Boxes given that a PO Box owner may or may not reside within the city in which the PO Box is located. Only zip codes corresponding to street addresses were used. The OR DMV supplied us with 899,515 names and mailing addresses total, from which we randomly selected a sample of 3900, 1300 each for Portland, Eugene-Springfield, and Bend. The figure of 1300 contacts per city was calculated based on available budget for printing and mailing services.

Having combined Eugene and Springfield into a single metropolitan center, we had to select proportionate numbers from each city to sum to a sample total of 1300 for both cities combined. The total number of names combined for Eugene and Springfield retrieved from the OR DMV was 223,160. The number of names we received from the OR DMV for Springfield totaled 60,009. Therefore, the proportion of Springfield respondents from the total for the combined area was 60,009/223,160 or 27%. This 27% value was used to calculate the number of names from Springfield we would select, 351 (.27*1300 = 351), leaving the balance of names for the combined Eugene-Springfield area to come from Eugene at 949 (1300-351 = 949).

Mail questionnaire. We developed the mail questionnaire for our study based on analysis of focus group outcomes and thorough review of literature concerning ecosystem services and urban forests and green spaces. We mailed the questionnaire in the fall of 2011 to recipients in Bend, Eugene, Portland, and Springfield, Oregon, using a modified Dillman approach (Dillman, Smyth, & Christian, 2009). First, we mailed a postcard announcement to recipients describing the project and advising respondents that they would be receiving the questionnaire shortly. About a week after receiving the announcement postcard, recipients received a full questionnaire package that contained the questionnaire, a cover letter, and a self-addressed, postage paid return envelope. A second postcard was mailed to all recipients two weeks after we mailed the questionnaire. This postcard thanked those who had already mailed back the completed questionnaire, and also offered a reminder that if the respondent had not responded, we would greatly appreciate their doing so. A second full survey package was then mailed out about two weeks after the reminder postcard to those who had not responded by the time of the second mailing of the full survey package.

Addresses for 144 of the 3900 people we attempted to contact by mail were not valid, and we received 12 refusals. There were 734 completed surveys returned for a final response rate of 20% (734/3744). The response per study area was about 33% each. We received 244 completed surveys from those we contacted in Portland, 246 from Bend, and 244 from Eugene-Springfield. We conducted a nonresponse bias check by asking a sample of non-respondents a selection of questions from the full questionnaire in phone interviews. We compared respondent and non- respondent answers to the selected questions using t-tests, chi square, and Wilcoxon rank-sum tests. Non-respondents did not significantly differ from respondents, with two exceptions. Non- respondents rated tree roots causing damage to sidewalks as a more severe problem than respondents and also reported feeling more attached to their city because of the trees and landscape than respondents. Though statistically significant differences were found, the practical significance of the differences was minimal (point bi-serial correlations were below .100 in each case).

Questionnaire development. The questionnaire contained multiple sections, each covering a specific topic area. It also contained a section of questions assessing general attitudes about urban forests and ecosystem services, and a final section on basic demographic information. What follows are descriptions of each section.

Importance-Performance analysis. Importance-Performance Analysis (IPA) was originally proposed as a simple way to convey information on consumer satisfaction (Hammit, Bixler, & Noe, 1996; Martilla & James, 1977; Siegenthaler, 1994). IPA relies on using a simple grid, through which business owners or service providers can more easily determine what people's expectations are for a good or service, and how well the good or service is being provided. Though concerns about IPA have been raised, including whether importance is unidirectional (zero importance to very important) or bidirectional (least important to most important), absence of a clear definition of importance, and potential misclassification of attributes on the IPA grid (Oh, 2001,/a>), it has still demonstrated usefulness in many natural resource studies (Backlund, Stewart, McDonald, & Miller, 2004; Hendricks, Schneider, & Budruk, 2004; Needham, Tynon, Ceurvorst, Collins, Connor, & Culnane, 2008).

IPA uses a Cartesian coordinate grid to compare the importance of a good or service to how well that good or service is being provided. The midpoints of each axis (one axis for performance, one axis for importance) customarily intersect at the midpoint of each axis, though this can be modified to reflect a higher standard (Bruyere, Rodriguez, & Vaske, 2002). For this study, our scales ranged from zero to six, and we intersected the axes at slightly above the midpoint of each scale, at 3.5, to reflect a slightly higher standard for the provision of ecosystem services. We refer to provision of ecosystem services, not performance as in the original scale. We used "provision" instead because it conveys the same nature of the relationship as the original scale that uses "performance," and is more meaningful in the present context.

Knowledge about urban forests and ecosystem services. Non-professionals commonly have differing levels of knowledge about forest management, and such varying knowledge can have an impact on their responses and attitudes about management (Loomis, Bair, & Gonzalez- Caban., 2001). For instance, knowledge about fire ecology and fire management practices have been found to be significantly associated with public responses to wildland fire management (Toman, Shindler, & Brunson, 2006). In this study, we asked respondents how familiar they were with three concepts - urban forests, ecosystems, and ecosystem services. We decided to ask about level of familiarity with these terms, instead of testing respondent knowledge, reasoning that testing questions might be objectionable to many respondents. Available scale responses ranged from 0, "Never heard this term before," to 6, "Very familiar with this term."

Ecosystem disservices/problems. Though ecosystem functions and structures provide outputs that are of service to people (a benefit), they also produce certain outputs that are troublesome for some (costs or disservices) (Lyytimäki, Petersen, Normander, & Bezák, 2008; Lyytimäki & Sipilä, 2009). For example, natural processes and systems can produce virulent diseases that are lethal to humans (Dunn, 2010). Less dramatic examples of ecosystem disservices include allergens, leaf litter, and sidewalks damaged by tree roots (Lyytimäki & Sipilä, 2009).

In this study, we asked respondents to rate a variety of disservices commonly associated with urban forests, including those mentioned above, as well as places for trash and litter to accumulate, safety concerns, and financial burden on citizens for street tree care. We asked people to rate the severity of the disservice or cost on a scale from 0, "Not a problem at all," to 6, "A severe problem."

Threats/Challenges to urban forests. Public attitudes about the impact of human activities on natural areas need to be considered in urban natural resource planning (Buijs, 2009; Haggett, 2011). It is often the case that non-governmental stakeholders possess different attitudes concerning natural resource management than those who have received professional training (Eisenhauer & Nicholson, 2007; Raik, Lauber, Decker, & Brown, 2005,/a>). In order to understand how members of the public may respond to various management activities, which are often based on specialized knowledge that non-professionals may not possess, urban forest managers need to consider public perceptions about challenges or threats to urban forest ecosystem health (Barro & Dwyer, 2000; Di Giulio, Holderegger, & Tobias, 2009). If managers can obtain a better understanding of public perceptions and attitudes about which issues are of greater or lesser importance to the public, and can improve their understanding of where the public's perceptions of threats to urban forests differ from managers', then natural resource professionals will be prepared to reduce the likelihood of misunderstandings and better respond to public concerns (Bright, Cordell, Hoover, & Tarrant, 2003).

We included a series of questions assessing public perceptions of the severity of a selection of challenges or threats to urban forest health were. We asked respondents about such challenges as habitat loss, invasive species, fertilizer use, and changes in climate. Respondents assessed the threat level on a scale from 0, "No threat at all," to 6, "A severe threat."

Preferred sources of information. Effective communication with members of the public is a key responsibility for natural resource agencies and professionals (Barro & Dwyer, 2000; Fazio & Gilbert, 1986). The content and nature of messaging matters greatly, but of equal importance to consider is through which media do members of the public most prefer to receive or acquire information (Kaplowitz, Yeboah, Thorp, & Wilson, 2009; Tucker & Napier, 2002).

To address the need for effective means of communication, we asked respondents about their preferred media sources using a list we adapted from a prior study (Charnley & Engelbert, 2005). We asked respondents for their level of preference for different means of communication including mailings, web-based platforms, and public meetings. We asked respondents to indicate their level of preference for each source on a scale from 0, "Least preferred," to 6, "Most preferred." We also asked respondents about whether they found different sources of information trustworthy. We asked for respondents' level of agreement with statements such as "you trust commercial television and radio to provide you with reliable information about natural resource issues," and "government agencies are reliable and trustworthy sources of information about natural resource issues." Respondents could select answers on a scale from -3, "Strongly disagree," to 3, "Strongly agree."

Management goal importance. In order for managers and decision-makers to practice responsive and efficient natural resource management, they need to understand how the public perceives management actions (Kneeshaw, Vaske, Bright, & Absher, 2004; Shindler, Brunson, & Stankey, 2002). More specifically, managers are well served by improving their understanding of what the public believes management priorities ought to be, since management agencies are moving towards more collaborative management that includes public input (Clendenning, Field, & Kapp, 2005) and public response to management actions will depend, in part, on the public's priorities (Brunson & Shindler, 2004; Bruskotter, Schmidt, & Teel, 2007).

We developed a list of management goals based on our review of ecosystem services literature (e.g., Bastian, Haase, & Grunewald, 2012; Dobbs, Escobedo, & Zipperer, 2011; Gobster & Westphal, 2004; Pickett, Buckley, Kaushal, & Willliams, 2011; Tyrväinen, Silvennoinen, & Kolehmainen, 2003; Young, 2010) and discussions with members of the public and natural resource professionals during focus group meetings. We asked questionnaire respondents about their attitudes concerning such goals as managing for sustainability, managing for watershed health, managing to increase property values, and increasing public involvement in urban forest management. Respondents indicated how important they felt each goal was by answering on a scale from 0 "Least important," to 6, "Most important."

Indicators of successful management. Public perceptions of natural area management are partially dependent upon the values people hold (Tyrväinen, Mäkinen, & Schipperijn, 2007). For example, people may value a green space or natural area for tangible benefits like recreation or consumable products, or intangible benefits like aesthetic and spiritual value (Winter & Lockwood, 2005). How the public values a landscape, the valued uses or benefits the landscape provides, will influence public perceptions of a well-managed landscape which, in turn, will impact public responses to changes in landscape condition and to management actions (Barro & Dwyer, 2000; Kirkpatrick, Davison, & Daniels, 2012). Members of the public may perceive a landscape to be well-managed and healthy because it is aesthetically pleasing, but trained professionals may hold a different evaluation of the same landscape for ecological reasons (Ryan, 2011). For example, public values and perceptions about wildland fires may be at odds with forest managers' and scientists' values and perceptions that identify (some) fires as a natural and necessary part of the ecosystem (Dale, 2006). Similarly, invasive species may be flourishing in an area, upsetting the native ecosystem integrity. If the invasives are particularly attractive, members of the public may perceive the landscape as beautiful and therefore, healthy; though, in terms of ecological integrity, it is not. Understanding what the public believes to be indicators of a well-managed and healthy urban forest can help managers better inform the public (where misinformation exists) and to better understand how the public may respond to management actions (e.g., removal of problem trees).

We asked respondents about their level of agreement with a series of statements to assess what they believed were indictors of successful management of urban forests. We asked, for example, if respondents believed seeing more trees was a sign of successful urban forest management. We also asked respondents if less trash and litter, and more green spaces in general, were signs that urban forests were being well-managed in their city. Respondents indicated their level of agreement with such statements by answering on a scale from -3, "Strongly disagree," to 3, "Strongly agree."

Responsibility for management. Incorporating multiple perspectives and values into planning and management has grown increasingly central to sustainable natural resource management (Harshaw, 2010; Kangas, Saarinen, Saarikoski, Leskinen, Hujala, & Tikkanen, 2010). For instance, natural resource professionals increasingly recognize that members of the public want a more active role in resource planning and management, and that including greater public input is important for more responsive and effective resource management (Martineau- Delisle & Nadeau, 2010; Tuler & Webler, 1999). Public involvement in natural resource decision making and planning involves issues of trust in government and public empowerment (Parkins & Mitchell, 2005; Rowe & Frewer, 2004) that need to be considered. How members of the public see their role and involvement in urban natural resource management and who they trust to responsibly manage public resources will impact how members of the public respond to management actions and public participation efforts (Charnley & Engelbert, 2005; Westphal, 2003).

In a general attitudes section of the questionnaire, we asked respondents their level of agreement with statements describing public influence in decision making, other groups' influences on management, and who respondents felt were best qualified to manage urban forests. For example, we asked respondents' level of agreement with statements such as "citizens have as much responsibility as trained experts for urban forest management," "special interests and industries have a lot of influence in natural resources management," and "you trust your city government to responsibly manage local natural resources." Respondents indicated their level of agreement with statements by answering on a scale from -3 "Strongly disagree," to 3, "Strongly agree."

The final two sections, general attitudes and demographics, are not discussed here. They may be viewed in Appendix 4 which contains a copy of the questionnaire.

Data Analysis

Statistical Analysis.

Statistical Analysis. We used t-tests and ANOVA to compare different categories of respondents. Our discussions with natural resource professionals revealed that groups that are easily identifiable and distinct would be the most helpful for resource professionals. Resource professionals we spoke with suggested that identifying differences between females and males would be helpful and learning more about differences between homeowners and renters would also be desirable. For these two groups, we conducted t-tests to evaluate statistically significant differences between women and men, and between homeowners and renters. T-tests are efficient and reliable statistical analyses that reveal differences in group means. T-tests are also robust tests even if target variables are non-normally distributed when sample size is large enough, as in our study (Vaske, 2008). To compare results across cities, we used ANOVA, a standard statistical test used to compare means across groups.

Weighting.

Though randomly generated, characteristics of our sample were somewhat dissimilar to U.S. Census demographics data for the Oregon cities included in the study. Additionally, our sample size was moderate, and response rate was somewhat low, though consistent with a general trend in mailback survey response rates (Van Horn, Green, & Martinussen, 2009).

Therefore, to strengthen the representativeness of our sample, we used 2010 Census data to weight our data by age, race/ethnicity, and sex. The weighting factor we used was -

Wp = %pop/%respondents.

To find the weighting factor, first we calculated the total population for the four cities in our study (Portland, Eugene, Springfield, and Bend). Then, we calculated the number of each ethnic/racial group (Whites, Blacks, Latinos, Asians, Hawaiians, Native Americans, and Mixed races) by age and gender, for each city. We split age into two categories, 18-44, and over 45, based on U.S. Census data.

As an example, to calculate the weighting factor for Black males, 18-44, we found the percentage of Black males, 18-44, from the total population for the four cities that we calculated. This number was divided by the total population to give us the percentage of Black males, 18-44, from the total population (%pop, in the above equation). Then we calculated the same percentage for our sample. That is, we calculated the total number of Black males, 18-44, who responded to our survey, and divided this number by the total number of respondents (% respondents, in above equation). Then, using these two values (expressed in decimal form), we were able to calculate the weighting factor for all Black male respondents, aged between 18 and 44, and apply that weighting factor to those respondents using a weighting function in SPSS.