Geo-questionnaire is a scalable method for soliciting and collecting public input by way of answering a set of prepared questions on topics that have explicit and/or implicit spatial connotations (Jankowski et al. 2016). These connotations are usually evoked in geo-questionnaire through a map and can be communicated by map sketching/marking and by answering questions that are triggered by map interactions. The theoretical-conceptual roots of geo-questionnaire method are broad and include planning theories, especially communicative planning (Khakee 1998), participatory geographical information systems (PGIS) and public participation geographical information systems (PPGIS) (Dunn 2007), and critical geographical information systems (GIS) (Elwood et al. 2011). Methodically, the geo-questionnaire method is rooted in participatory mapping (Corbett, Keller 2006), social surveys and pools, and most directly in
Geo-questionnaire can be classified as a PPGIS method, albeit due to its one-way communication mode (from citizens to planners) it ranks in the middle of the communication spectrum linking citizens with planners (Carver 2001). Its primary purpose is to inform the planners and decision-makers about people’s preferences and evaluations, which if taken seriously can contribute to making a planning process more inclusive, anticipate conflict among stakeholders, and provide a basis for socially acceptable, legitimate, and sustainable solutions. Several studies carried out in the last two decades sought to develop methods for eliciting public preferences by using map annotations and free-hand sketches. Early work involved sketching and annotating land development scenarios on a common base map by a small group of stakeholders in face-to-face meetings (Faber et al. 1995), incorporating local knowledge and preferences of residents into desktop GIS databases available for planners (Talen 2000), and using web GIS to identify areas of environmental value and assessing the degree of agreement between conservation and development preferences in small stakeholder groups (Dragićević, Balram 2004). Marking and annotating analogue and digital maps has been also used as a way of collecting indigenous knowledge about land and natural resources (Kayem 2004), and mapping landscape values (Brown, Weber 2012).
Geo-questionnaires and other PGIS/PPGIS methods have been used in a variety of contexts, and processes initiated by various types of stakeholders with diverse goals and expected outcomes. In this paper, the focus is on using geo-questionnaires in processes initiated by researchers from the theoretical point of view and spatial planners from the practical point of view. That is why the geo-questionnaire will first be described as a method, then the types of data collected with geo-questionnaires, and techniques used to analyse these data will be presented. Next, the methods of recruiting survey participants and the quality of collected data will be analysed. Finally, examples of geo-questionnaire applications will be discussed. The paper is closed by recommendations for geo-questionnaires’ implementation in spatial planning.
Geo-questionnaires belong to a broader category of Computer-Assisted Web Interviewing (CAWI) methods or Geoweb methods. They allow to simultaneously collect qualitative, quantitative and spatial data from relatively larger population samples than during face-to-face meetings. They differ from other CAWI methods in that they provide geographical context for surveys and offer functionalities enabling the input of geographic objects, i.e. points, lines and/or polygons by respondents (Jankowski et al. 2016). Geo-questionnaires typically contain multiple pages of which some are complemented with an interactive map, i.e. a map that, at a minimum, allows for panning and zooming. The data are usually contributed online in an individual unsupervised setting, but it is also possible to use geo-questionnaires in a group and/or supervised setting (e.g. Chaix et al. 2012, Bugs, Kyttä 2016). Other names used rather inconsistently in the literature to describe geo-questionnaire include “surveys that include a spatial mapping component” (Brown 2006),
As mentioned above, responses to a geo-questionnaire may contain point, line, and polygon objects sketched by participants on a map. The act of object sketching will often be used in geo-questionnaire as a trigger to present the respondent with additional questions appearing in the context windows (Fig. 1). Answers to such questions are stored in a geo-database as feature attributes. Points are commonly used in geo-questionnaires to represent the locations of spatial variables, mostly because they are simple to interpret and use by respondents (Brown, Pullar 2012). However, in cases involving development preferences that pertain to features with defined boundaries, the use of polygons may be warranted (Jankowski et al. 2016). Some geographical features and phenomena, such as routes taken by respondents, are best represented in geo-questionnaires by lines. In some applications, instead of drawing new geographical features, the respondents may select map features provided by the geo-questionnaire designers and answer the pertinent questions (e.g. Schmidt-Thome et. al. 2013).
Data collection in geo-questionnaire is a one-way communication, in which the data flows from respondents to researchers and/or planners. The respondents usually do not see the contributions of others and there is little possibility for interaction between people. This characteristics differentiates geo-questionnaires from other online PPGIS tools such as argumentation maps (Rinner 2001), geo-discussions (Leahy, Hall 2010) or Canela PPGIS (Bugs 2010), which allow for interaction between participants.
Geo-questionnaires are in general best suited to measure spatial attributes that are subjective in nature, and based on respondents’ local knowledge, experience, perception, and opinion. Such spatial local knowledge is often ambiguous, fuzzy, and does not have distinct boundaries (McCall, Dunn 2012). Data on spatial variables collected with geo-questionnaires and other PPGIS methods typically fall into four broad and mutually related categories:
Data collected with a geo-questionnaire can be analysed and visualised in GIS using a variety of methods. Participant attributes may be georeferenced using their residential locations, and analysed similarly to other geodemographic data. For instance, they may be aggregated to administrative units such as districts or postal code areas and visualised in order to identify spatial patterns. The data may also be analysed (in both aggregated and disaggregated manner) using methods of exploratory and confirmatory spatial data analysis (Anselin 1999).
Geo-questionnaire data may be also used to summarize the behavioural patterns of individual respondents by calculating their activity spaces (e.g. Perchoux et al. 2014), and to build models predicting travel mode choices of individuals and households (e.g. Salonen et al. 2014). The data may be also used to calculate travel origin and destination matrices, and to study travel demand and route choices. Thus, in the context of transportation and mobility research and planning, geo-questionnaire data provide an alternative to or complement of traditional travel diaries and newer data collection methods such as GPS and mobile tracking techniques (Czepkiewicz et al. 2016a).
Spatial variables representing respondents’ experiences, evaluations, values and preferences may be analysed in a disaggregated (individual) or aggregated manner (Talen 2000). When analysed individually, respondent contributions are visualised and explored one by one in a desktop or web GIS environment. When treated in aggregate, each contribution is treated in the same way and data contributed by multiple respondents are brought to a single output (e.g. a density map). With the aggregate approach, some subjectivity and individuality in data may be lost, especially when the attribute data contain the details of participants’ local knowledge. Yet, the aggregated data may be preferable, for instance if the decision-making informed by the data is focused on locational choice (Talen 2000). when an in-depth insight into the respondent’s contributions is necessary, a combination of aggregated and disaggregated approaches is recommended (Fig. 2).
Various GIS techniques can be used for data aggregation. Points and lines may be aggregated using simple density and kernel density estimation techniques (e.g. Alessa et al. 2008) (Fig. 2), while polygons may be aggregated by counting the polygons overlapping regularly shaped map tessellation units such as square or hexagonal cells (Jankowski et al. 2016) (Fig. 3).
Visualising the density maps can provide valuable insights into a given decision-making or research problem, but the maps can be also analysed further with GIS methods. Spatial statistics and visual analytics may be used to identify high and low density clusters (Alessa et al. 2008, Brown, Pullar 2012). Distributions of spatial variables (e.g. landscape characteristics) may be also analysed using landscape ecology metrics (Brown, Reed 2012). Preferences and subjective evaluations of geo-questionnaire respondents are often combined with categorical and quantitative data such as land use/land cover maps and other derivatives of remote sensing data. For instance, the geo-questionnaire derived social values and use patterns may be spatially joined with land use data or remote sensing images to evaluate urban green spaces (Brown et al. 2014a, Pietrzyk-Kaszyńska et al. 2017), create sociotope maps (Ståhle 2006) or identify socio-ecological hotspots (Alessa et al. 2008).
Spatial variables that represent development preferences may be aggregated into agreement-disagreement maps depicting areas, in which favourable or unfavourable opinions dominate (Jankowski et al. 2016, Fig. 4). These maps may be further combined with spatial variables representing social values to show the locations of potential conflict (Brown, Raymond 2014, Kahila-Tani et al. 2015). Such maps allow to identify areas and locations, in which the high level of subjective value overlaps with the high proportion of views favourable for development. The aggregated preference maps may be also combined with locations of known development proposals or initiatives, and enable planners to anticipate the areas of potential conflict and disagreement during the planning process.
Geo-questionnaires and related PPGIS methods have been applied in various contexts and served diverse stakeholder needs. Most broadly, the geo-questionnaire applications may be divided into those related to academic research and the ones related to the non-statutory collecting of public preferences and opinions informing spatial planning. In academic research, geo-questionnaire data are often expected to allow the generalization of preferences and behavioural patterns from a sample to broad population. In such context, probability sampling is preferred over voluntary or other non-probabilistic sampling methods used to identify and recruit potential respondents. However, the use of geo-questionnaires as a public participation method requires a more nuanced consideration of the question of who participates in data collection (Schlossberg, Shuford 2005).
Random sampling used to recruit participants allows for the data collected through PPGIS applications to be interpreted as representative of the views of general public, and provides a
Brown and Kyttä (2014) see the choice of sampling methods as one of key issues in PPGIS and call for further research on their effects on participation rates and data quality. Other researchers have called for the concurrent use of probability and voluntary sampling to study the effect of sampling methods (Brown, Reed 2009) on the quality of public participation (Brown et al. 2014a). Another important issue for geo-questionnaire data collection is the quality of data required to effectively inform planning (Brown, Kyttä, 2014, Czepkiewicz et al. 2016b).
The criteria for evaluating quality of data obtained from geo-questionnaire depend on the goal and the context of geo-questionnaire use. Some of the criteria are the same as in other survey methods, others are specific to geo-questionnaires and PPGIS methods. The quality criteria can be divided into internal and external (Lechner et al. 2014, Devillers et al. 2010).
Internal data quality pertains to the degree, to which the attributes have been correctly measured, interpreted and represented by data. In survey research, such criteria include construct validity, logical consistency, and completeness (Lechner et al. 2014). Additionally, in reference to spatial data the criteria include positional and attribute accuracy (Goodchild, Li 2012).
Description of selected geo-questionnaire cases in Poland.
Case | City | Year | Topic | Participants | Recipients |
---|---|---|---|---|---|
Kasprowicz Park | Poznań (population 545,000) | 2015 | Local urban planning | local area residents and users (N = 1,009) | Poznań Urban Planning Office |
Local needs map | Poznań (population 545,000) | 2016 | Downtown urban renewal | Downtown residents and users (N = 709) | Poznań City Hall, Poznań neighborhood councils |
Sustainable public transportation model | Łódź (population 701,000) | 2016 | Public transportation planning | Public transportation users (N = 2,387) | Łódź City Hall |
New Rokietnica centre | Rokietnica (population 5,500) | 2016 | Suburban town planning | Rokietnica residents (N = 435) | Rokietnica Municipal Office, land use plan designer |
Construct validity and logical consistency refer to how well spatial variables collected with geo-questionnaires reflect phenomena they are intended to measure, and are largely dependent on the survey design and to a lesser extent on participant characteristics (Jankowski et al. 2016). Locational accuracy in PPGIS is influenced by the nature of spatial variable itself (e.g. its ambiguity), instructions provided to participants, quality of mapping environment (e.g. background map, navigation tools), map zoom level, as well as the respondents’ mapping skills and familiarity with the study area (Brown 2012b). Some PPGIS data do not require high spatial accuracy because the participants mark regions and places with undefined boundaries to begin with (Brown, Pullar 2012). Attribute accuracy in PPGIS is closely related to locational accuracy, since it may result from miss-attributing certain characteristics of geographical features as well as from misplacing them. The attribute accuracy may be validated in some cases by the comparison with expert-derived data. Such comparison is not always possible, and the researchers have suggested proxy measures of PPGS data quality, such as the amount of participant effort put into mapping activities (Brown 2012b).
External data quality relates to how well knowledge and values of the target population group(s) are represented in geo-questionnaire data. It is as important in academic research as it is in the practical uses of geo-questionnaire data. In the former, the main question refers to the ability to generalize findings to a target population, whereas in the latter the question pertains to whose perspective has stronger influence on decisions.
Social representativeness refers to how well various social groups, and by extension their values and perspectives, are represented in the data. It is usually measured by comparing basic socio-demographic characteristics (e.g. age, gender, education, ethnicity or employment status), between the sample and the population. As a predominantly online data collection method, geo-questionnaire is subject to digital divide. Therefore, any use of the method may reproduce inequalities in access to and the ability of using information and communication technologies by a given social group. One way to alleviate socio-demographic biases may be to use the supervised data collection mode in workshop settings with the recruitment based on quota sampling to ensure equal participation of social groups, and to facilitate those participants who have a low level of Web browsing and mapping skills. Biases related to browsing and mapping skills may also be alleviated by user-centric design (Haklay, Tobón 2003, Bugs 2012), and tailoring interfaces to the needs of specific groups, such as children (Kahila, Kyttä 2009) or older adults (Gottwald et al. 2016).
An aspect of data representativeness, which is rarely evaluated in other social survey methods is related to the spatial distribution of participants. If the spatial distribution of respondent sample does not reflect well the spatial pattern of target group, it may be the source of bias in the analysis of data collected with geo-questionnaire. Participants’ distribution is typically represented by locations of their residences and it may be compared with the spatial pattern of area residents. In PPGIS, the level of knowledge about places affects how spatial variables are mapped, and the familiarity with locations closer to one’s place of residence is typically higher than with locations that are farther away (Brown, Reed 2009). In cases of attributing values and desirable land uses to specific locations, the distance to the domicile might influence the assigned values/preferred land use categories due to spatial discounting: place value and attachment develop with proximity and familiarity, and people prefer to have positive aspects close to their homes while keeping the negative aspects away (Brown et al. 2002, Pocewicz, Nielsen-Pincus 2013, Brown, Kyttä 2014). Spatial representativeness may be assessed by comparing the distribution of respondents with the pattern of plan area residents aggregated to spatial units (Czepkiewicz et al. 2016b).
Geo-questionnaires and similar PPGIS methods have been applied to elicit public preferences on various subjects and at various spatial scales in a number of countries including Finland (Kahila, Kyttä 2009, Schmidt-Thomé et al. 2013, Kahila-Tani et al. 2015), Sweden (Babelon et al. 2016), Iceland (Pánek, Benediktsson 2017), Czech Republic (Pánek et al. 2017), USA (Alessa et al. 2008, Brown, Reed 2009, Schmidt-Thomé et al. 2014), New Zealand (Brown, Weber 2012), Portugal and Brazil (Bugs, Kyttä 2016), and Australia (Brown 2012a). A more comprehensive review of PPGIS data collection cases in academic research can be found in Brown and Kyttä (2014). Here, we briefly present four recent cases of geo-questionnaire applications conducted in Poland in 2015 and 2016, initiated by the authors as part of applied research projects, and related to ongoing urban planning processes.
Each of the cases is described below in terms of its context, topical content, and situation in urban planning process:
The majority of PPGIS applications reported in different journals so far have involved prototypes developed and tested by researchers for academic purposes and with little or no linkages to actual planning processes (e.g. Rinner, Bird 2009, Bugs et al. 2010). More recently, several cases of PPGIS use in municipal planning have been reported (Kahila-Tani et al. 2015, Jankowski et al. 2016), and some commercial and open-source tools have been commissioned and applied by municipalities in Sweden, Finland, and Poland Examples include Geopanelen and Bästa Platsen in Sweden, Maptionnaire and eHarava in Finland, and Geoankieta in Poland.
An in-depth description of the geo-questionnaire, as well as geo-discussion, can be found in two separate elaborations, extensively discussing all the details of both Geoweb applications (Jankowski et al. 2017b, 2018).
The article’s aim has been to provide an overview of geo-questionnaire as a PPGIS method for eliciting spatial local knowledge and public preferences. We have discussed geo-questionnaire data content, data visualisation, aggregation and analysis methods. We have reviewed several issues underpinning the applications of geo-questionnaires in academic research and in municipal planning, including respondent selection and recruitment, data quality, and data representativeness. To substantiate the discussion of method with application examples we have briefly described four recent geo-questionnaire use cases from Poland. The described case studies show the versatility of geo-questionnaire as a method for collecting data on public preferences, behavioural patterns, and local knowledge that can be relevant for planning practice in a variety of domains including land use, public transportation, green spaces, and urban functions, to name a few. Whether or not this particular method of public input becomes more broadly adopted by planners will depend on the perceived and real costs of deploying geo-questionnaires vs. the value of collected data, and on public perception of the effectiveness of geo-questionnaire to have their preferences incorporated in the decision-making outcomes.