Urban sustainability can be considered as both a desirable goal and an ongoing process [11]. Although over 80 definitions concerning different circumstances can be found in related literature from diverse disciplines, the widely accepted interpretation of urban sustainability always emphasizes balancing development in three primary domains: urban economy, society, and environment [5,12]. Overall, the definition of urban sustainability can be divided into two categories, holistic and narrow. The former highlights the general status of urban development with an offset view of different domains while the latter focuses more on one or two relevant domains concerning different circumstances. Accordingly, USIs as “parameters or values derived from parameters” are developed to form different USI models for quantitatively measuring and evaluating urban sustainability from both the holistic and narrow perspectives [13,14,15,16,17,18,19,20,21,22,23,24,25].

To systematically monitor and assess urban development progress towards sustainability, previous studies have described step by step procedures for developing USI models [13,26]. First of all, an unambiguous interpretation of urban sustainability is required to provide the overall direction of the USI system. Establishing a suitable framework and defining a set of selection criteria are deemed two essential steps in this whole process [27]. Based on the previously accepted indicator selection criteria of specific domains concerning sustainability (e.g., social indicators, urban indicators, and environmental indicators) listed by [13] and the fundamental requirements suggested by [15,21,27,28], we summarized a more general set of six standards (clearly defined and scientifically representable, responsive to target goals and audience, data available, numerically measurable, spatially and temporally comparable, and cost-effective) with the most consensuses to fit the selection of USIs. These USIs are supposed to characterize the relevant phenomena or aspects of urban development very well and be informative to the public in a spatial-temporal dynamic pattern. They should also be easily quantified with available measurement data in an acceptable expense.

According to the identified selection criteria, a preliminary list of possible urban sustainability indicators are supposed to be proposed based on specialized knowledge as well as the suggestions from other stakeholders (general public or policy makers). Then both equal and unequal weighting of the importance of identified selection criteria can be used to evaluate each proposed initial indicator. This indicator evaluation process can be based on the one-step procedure, a sequential procedure, or a hybrid procedure summarized by [26]. The specific goal for local development and adopted methodologies may also contribute to adding to or reducing from the original indicator set. Finally, after the selection of final indicators for the USI system, the effectiveness of this system should be tested in real case studies for evaluating urban sustainability [29].

Currently established USI models include “Sustainable Seattle indicators” [23], “Hamilton-Wentworth Indicators” [30], “British Columbia State of Sustainability Indicators” [16], “Taipei’s sustainability index” [31], and “London’s urban sustainability indicators” [32]. However, several limitations can be found in these USI models.

Firstly, efficient USI models are supposed to reflect inter-generational equity (future generation and current generation), intra-generational equity (social equity and geographical equity), ecological equity (species conservation, minimizing environmental impact, and efficient resources use), and human well-being [13,14]. Assessing inter-generational equity requires temporal tracking over periods based on the same evaluation standard, while monitoring intra-generational equity needs spatial comparison among geographical areas at different levels (international, national, regional, urban, neighbourhood). However, the current USIs normally measured by traditional census or tabular data cannot easily provide such temporal and spatial pattern of urban sustainability change. To this point, lacking spatial and temporal analysis can be identified as a dominant limitation of current USIs.

For the measurement and quantification of USIs, it is difficult to collect the expected census data by conducting social surveys which are time-consuming, expensive, and labor-intensive. In most USI models, the census data collected for quantifying USIs are commonly objective data merely indicative of the physical status or socioeconomic conditions of urban development. Subjective data (e.g., people’s perceptions, feeling, and sense) is almost ignored. This is likely due to the increased difficulty in acquiring such data which involves face-to-face interviews, detailed questionnaire design, collaborations with multiple agencies, and ethics approvals. However, as principle participants in urban activities, urban residents’ living experience and perspectives impact and in turn are impacted by urban development in an interactive manner. Most current USI models highly dependent on census data lack such encompassing measurement and quantification from both objective and subjective aspects, which can degrade the accuracy of urban sustainability monitoring and also hamper the investigation of linkages between indicators in different domains.

Another shortage in existing USIs lies in the weak integration of possible indicators from different domains to provide a comprehensive understanding of urban development. Maclaren argued that multiple indicators are required to measure the multiple dimensions of urban sustainability because indicators devoted to any single domain fail to draw the whole picture [27]. However, more indicators included for a wider coverage of urban development results in more complexity caused by measurement and quantification. Therefore, minimizing the amount of indicators without sacrificing the information content is another requirement for maximizing the efficiency of urban sustainability evaluation. To solve this problem, constructing composite USI can be a possible way by conjoining multiple specific indicators into a more integrated one [33]. For example, QOL is such a composite indicator that incorporating multiple dimensions of urban life (economic, social, educational, and environmental) into one single well-being index. This can be found in a growing body of research by [34,35,36,37,38]. Similarly, urban sustainability can also be evaluated by such composite indicator indices at different integrated levels for planning and policy-making purposes.

It is therefore of great significance to establish an integrated USI model based on a hierarchical indices system. This model should incorporate both objective and subjective information to provide a more accurate evaluation of urban sustainability. Additionally, geospatial data (satellite imagery and maps) and traditional census data are both needed to guarantee an effective spatial-temporal pattern analysis of urban sustainability distribution which is indispensable for decision makers in formulating and implementing adaptive strategies.

Urbanization can be simultaneously considered both an issue and a process that links different physical and social systems to a high concern of sustainability with regards to environmental problems, public infrastructures, service development, and policy making [39]. Global economic growth has driven this issue to experience an unprecedented period. An increasing number of unsustainable problems arising from this urbanization process include urban population growth, urban sprawl, and land cover/land use change, which has drawn great concern from a variety of stakeholders such as policymakers, urban planners, social scientists, geographers, economists, environmentalists, entrepreneurs, and urban residents [40]. Particularly, such urbanization issues have negatively affected sustainable spatial planning of environmental and socioeconomic development [17]. Key elements of urban spatial sustainability include balancing population growth, guaranteeing sufficient green space, and minimizing urban sprawl, natural area change, and use of environmentally-damaging materials (e.g., impervious surface) [17,41,42].

An urban environmental system is a necessary component of the whole urban ecosystem, which is of primary importance to urban sustainability. With respect to urban environmental policies which emphasize either “ecology within the city” or “city in ecology”, environmental quality always to a high extent determines urban residents overall well-being (e.g., health, recreation, etc.). Urban environmental issues generally involve energy use, biodiversity conservation, landscape amenity, and natural resource protection. Specific urban environmental concerns comprise aspects of green vegetation space (GVS), impervious surface area (ISA), water quality, and urban air condition (temperature and components). All of these elements are associated closely with the overall quality of the physical environment as well as residential QOL [15]. It is of great necessity to assess urban environmental conditions by investigating such features for further examining the urban QOL from a sustainable perspective.

Urban sprawl is commonly quantified by deriving the built-up area difference between different periods from physical field survey, historical maps, or remotely sensed imagery [75]. LULC change can be conducted by combining remote sensing change detection techniques and GIS spatial analysis functions [51]. Weng [73] detected the trend and spatial pattern of land use change in the Zhujiang Delta of South China based on geospatial analysis and stochastic modeling. He also found a large loss of cropland in the past decade resulting from an abnormal urban growth which demonstrated the impacts of urban sprawl on land use dynamic conversion. Seto and Kaufmann [74] studied the same region by applying an econometric model with inputs of Landsat Thematic Mapper (TM) imagery of 30 m spatial resolution and economic-demographic data, which indicated that the external investment in industry is the paramount driver of local land use change. Similarly, the spatial-temporal dynamics of urban sprawl in the city of Shijiazhuang (China) was explored by [43] who integrated socioeconomic census data, historical maps, and satellite imagery to the GIS spatial analysis platform. LULC change detection was also carried out in this study and was found to be greatly driven by this urban expansion phenomenon as well as other influencing factors. In addition, urban sprawl and LULC studies are also conducted in Africa. Tewolde and Cabral [45] employed remote sensing object-oriented classification and a multi-layer perception neural network to develop an urban growth model which can predict the land use change and urban expansion in the next decade. In this study high spatial resolution imagery (IKONOS-2 of 1 m and Quickbird of 0.6 m), DEM data, and land cover maps were used simultaneously as ancillary data to assist classification and accuracy assessment.

Previous literature shows great potential and availability for geomatic techniques and statistical approaches in assessing QOL from a spatial perspective. Lo and Faber [80] estimated QOL in Athens-Clarke County (Georgia, US) by incorporating TM imagery-derived environmental information (e.g., land use land cover percentage, NDVI, and surface temperature) and socioeconomic census data (e.g., population density, per capita income, median home value, and percentage of college graduates). The intercorrelation of different variables was also investigated by this study which presented a comprehensive explanation of the multiple contributors to the overall QOL. The methodologies proposed by this study can perfectly match the essential characteristics of QOL (subjectivity and objectivity) interpreted in early studies [102,103] by conducting both the subjective and objective QOL assessments respectively based on PCA and GIS overlay approaches. Likewise, Li and Weng [59] adopted the same idea in the QOL study of Indianapolis city (US) and they expanded the original variable set to encompass more input information such as the impervious surface area extracted from satellite imagery, poverty, employment rate and house characteristics for the quantitative QOL model. In addition, they used ETM+ imagery instead of TM for surface temperature retrieval which improved the spatial resolution of the thermal band from 120 m to 60 m. Moreover, they conducted Pearson’s correlation and factor analysis to reduce the redundant information of the dataset to derive combined information representing different aspects of QOL. Similar QOL studies can also be found in [82,104,105].

These aforementioned approaches are nothing new since they have been widely applied in a number of studies including environmental studies, urban planning, and some social sciences. However, most of these techniques were just used separately to investigate one or two aspects of urban development. Some focus on urban sprawl or land use while others concentrate on impervious surface extraction, urban heat island study, or urban vegetation mapping as well as for quality of life modeling. Such issues are highly associated with urban sustainability; they have rarely been studied together based on integrated geomatic approaches for a comprehensive urban sustainability study. In a sense, the promising remote sensing and GIS techniques have still not been fully explored to depict a whole picture of urban sustainability due to their detached applications. Some statistical data can further illustrate this point. For example, in the world-famous citation database Web of Science users can access 3043 articles on urban sustainability if “urban sustainability” has been selected for the search topic. However, if the search topic is set to “urban sustainability” and “remote sensing” at the same time, the search result is only 47 records. Similarly, only 120 articles can be found if the search topic is set to ‘urban sustainability’ and ‘GIS’. If the topic is set to “urban sustainability”, “remote sensing”, and ‘GIS’ simultaneously, only 20 articles are returned. This indicates that approximately 1.5% ~ 3.9% of urban sustainability studies have used advanced geospatial techniques. Accordingly the spatial-temporal comparability of these USIs is somehow challenged. It is necessary to explore the potential of advanced geomatic approaches in such aspects for improving the evaluation of urban sustainability.

We adopt a holistic interpretation of urban sustainability which could be explained as: A sustainable city is one undergoing development in harmony with its environment, economy, society and people’s well-being. Although this definition has been rejected by the United Kingdom’s Local Government Management Board in their description of urban sustainability owing to a lack of innovation, we still suggest this traditional way in explaining urban sustainability. Because this interpretation is comprehensive enough to cover different aspects of urban development all of which are highly relevant to sustainability. The USI model established based on this interpretation of urban sustainability can be more suitable for application in different urban contexts due to its generality and comprehensiveness. In addition, people’s well-being can provide more subjective information of urban sustainability status while the other three basic domains (environment, society, and economy) can reflect more objective evidence. People as an essential component of urban activities are supposed to be taken into consideration for explaining urban sustainability. Since there is a large portion of overlap between social and economic domains, we would prefer to simplify the interpretation of urban sustainability into a balance of three dimensions: environment, socioeconomic, and well-being (Figure 1). This identified coverage will constitute the fundamental domains of a hierarchical multiple-indicator system for our proposed USI model.

First, this integrated USI model will be a domain-based framework. As Keirstead and Leach [21] suggested, USIs should be presented within clearly allocated categories, themes, and sub-themes focusing on policy goals. Our proposed USI model will encompass three basic categories (environment, socioeconomic, and well-being) according to the definition of the appropriate urban sustainability aforementioned. The specific themes and sub-themes within the three broad domains are determined based on the key influential elements extracted from the literature. The socioeconomic domain contains the theme of urbanization consisting of sub-themes such as population growth, urban sprawl, and urban land use change. The environmental domain comprises four key aspects which are green space, impervious surfaces, water quality, and surface temperature. The well-being domain mainly focuses on the theme of QOL which encompasses multiple sub-themes explained by material living, surrounding status, and individual’s development [59]. In addition, this USI model can also be considered as a process-outcome framework in that the socioeconomic domain acts as a stressor agent while the environment and well-being domains represent the condition agents. The environment domain can also play the stressor role in the outcome of well-being. This model can be shown by Figure 2 in a graphic illustration. The accessible and measureable indicators can be found in following Section 5.3. and the discussion of this proposed model is in Section 6 with regards to its advantages and limitations.

Different interpretations of urban sustainability can lead to different selection of indicator set, so both academic and practical literature were studied to identify suitable USIs capable to evaluate the fundamental domains within the USI model. This step is critical to determine whether an accurate measurement and assessment of urban sustainability can be achieved using a synthesized framework. Existing USI models including State-of-Environment Reporting, Health City Reporting, QOL Reporting, The United Kingdom’s Local Government Management Board, Hodge’s Framework for Systematic Sustainability Reporting, and Developing Indicators of Urban Sustainability were examined to draw appropriate indicators for our proposed integrated USI model. The advantages and disadvantages of each possible indicator should be assessed according to the identified indicator selection criteria. The validity, accessibility, and measurement of the potential indicators should be considered and evaluated as well.

To efficiently measure and quantify this proposed USI model, a hierarchical indices system was developed at three integrated levels (Figure 3). In this system, all urban sustainability indicators will be quantified in the format of index; this allows standardized calculation of different indicators as well as easy comparability among indicators to overcome such weakness pointed out by researchers in most existing USIs models. The hierarchical relationships exist based on the degree of integrity or synthesis for different indicators. Level 1 indices represent the most specific USI set some of which can be directly measured using the census data or imagery information, such as Population index, Land use index, Green space index, and Safety index. Level 2 indices are more integrated and they can be calculated based on a function with Level 1 indices as input parameters, such as Environmental index, Quality of Life index, and Socioeconomic index. The single Level 3 index called Urban sustainability index is the most integrated compared to indices in the other two levels. It will be derived based on all Level 2 indices. The formula for indices calculation will be weighted linear equations. All objective variables in equations can be obtained from available census data or imagery information while the weights for objective variables are from social surveys or set equal.

In this study, a preliminary USI model was established with a selected USI set composed of fourteen indicators ranged from Level 1 to Level 3, which was aimed to achieve evaluation of urban sustainability from detail to more synthetic perspectives. Ten Level 1 USI indices (e.g., Population index, Land use change index, Urban sprawl index, Economic index, etc.) are defined as specific subsets of three Level 2 USI indices (Urbanization index, Quality of Life index, and Environmental index). All USI indices in Level 1 and Level 2 are considered as subsets of the most integrated and comprehensive Level 3 index (Urban sustainability index). Each Level 1 USI index can describe one specific dimension of its superset of Level 2 indices. In similar, each Level 2 USI index depicts a specific picture of the superset of Level 3 index- urban sustainability.

A brief description is provided below to show how to calculate the multiple indices for quantification of the proposed USI model. First, because Level 1 USI indices are single ratio based indices or derived based on a group of single ratio based indices, they can be very easily calculated using the available census data or imagery-derived information (Figure 4). For example, Land use change index can be derived from the ratio of changed urban area to the total urban area in the current year. Or Green space density index can be obtained based on the ratio of green vegetation area to the total urban area. Second, for Level 2 indices, they can be quantified based on a weighted addition of all Level 2 indices in its subset. Weights for objective indices including Urbanization index and Environmental index are set equal because we assumed there is no priority for objective change of urban development. Weights for the composite index Quality of life are set according to the social survey results since the human well-being status is determined by both objective conditions and subjective perceptions. For the final Urban sustainability index, three input variables which are Level 2 indices are given equal weights because three fundamental domains are considered to be the same important in determining urban sustainability. However, concerning the different urban sustainability goals set in different urban polies, the weights can be flexibly set for the identified USIs. As Gosselin [25] suggested, the balance of different domains within urban sustainability and the reference from other studies should be taken into consideration.

This hierarchical system can provide good quantification of the USI model. All the Level 1 and Level 2 indices are developed according to different domains, themes, and sub-themes of the integrated USI model shown in Figure 2. In particular, the Level 3 Urban sustainability index enables a quantitatively overall evaluation of the progress towards urban sustainability from a spatial-temporal perspective.

To build trust and approval of the proposed USI model, the initial indicators would be evaluated against the identified selection criteria to fit the context of specific urban sustainability study. To maintain the integrity and hierarchy of the proposed USI model, the actual evaluation should be directly conducted on the specific measurements (e.g., Population growth rate, Population density, Green Space area) in replace of urban sustainability indicators [25,29]. Equal weights are supposed to give each USI selection criteria based on a simple dichotomy by marking yes or no. Measurements with more than three criteria marked yes should be selected as the final USI measurement for calculating the indicators in the integrated USI system. The evaluation process was suggested to take the form of simple questionnaire surveys among policy makers and the general public. Alternatives in more consistency with the selection criteria are possibly included in the original list of USI measurements.

To test the feasibility of the hierarchical indices system in quantifying the proposed USI model, a case study on Saskatoon Neighbourhood Quality of life index (or score) in 2007 was carried out and is shown here. This case study displays only partial results of an urban sustainability evaluation of Saskatoon based on the integrated USI model. To calculate all the USI indices for the entire hierarchical system is time-consuming due to the data collection (satellite imagery, census data, and historical maps), data processing, statistical and spatial analysis, so we could not provide a full set of quantified USI indices presently. The main purpose for this paper is to establish the conceptual integrated USI model as well as to prove the feasibility of quantifying this USI model by a hierarchical indices system. However, this expected full set of quantified indices will finally be derived to achieve the goal of monitoring and evaluating urban sustainability in our case study cities (e.g., Saskatoon).

Saskatoon (52.12°N, 106.67°W), located in the province of Saskatchewan in west-central Canada, is geographically situated in townships 36 and 37, ranging 4, 5 and 6 over 300 km north of the U.S. border. It is a medium sized prairie city with an extent of approximately 218 km² formed during the Great Transitions of Canadian cities (1850–1945) [106,107]. Saskatoon is the commercial and educational center of Saskatchewan. It has become the fastest growing city in Canada due to its highest population growth rate driven by a dramatic increase in annual immigration [108]. Urban sustainability has been set as a goal for future development by the city council (City of Saskatoon, 2012). The total population within Saskatoon increased from less than 50,000 in the 1950s to more than 236,600 in the 2012. Concomitant with the rapid population growth is the dramatic urbanization processes (e.g., urban sprawl, land use change, etc.), which has posed great challenges on the city’s sustainable development (e.g., bare soil or green land is converted to urban areas every year). As of 2007 the City of Saskatoon consisted of 86 intra-city units 64 of which were residential neighbourhoods with the remaining 22 being used as industrial, management area, development area, and recreational park (Figure 5). Of the 64 residential neighbourhoods only 58 were included in the study due to the availability of the data for the remaining 6 neighbourhoods (Willowgrove, Blairmore Suburban Centre, Stonebridge, Hampton Village, Rosewood, and University Heights DA) in 2007 (Figure 6).

Research data including census data, subjective weights, and spatial data were acquired from secondary sources including the City of Saskatoon, Elections Saskatchewan, the Saskatoon Police Service, the Saskatoon Health Region, and the Community University Institute for Social Research (CUISR)’s report entitled “Tracking Quality of Life in Saskatoon: Summary of Research, 2007 Iteration”. This data was stored, manipulated, and analyzed with the use of a GIS in order to carry out index calculation and spatial analysis.

The calculated results of the subjectively weighted QOL index for the 58 study neighbourhoods in Saskatoon are displayed in Table 1. Statistical results show that QOL score ranges from 73.24 in Forest Grove to 137.11 in University Heights SC with a mean score of 118.82. A negative skew value of -1.47 suggests that the majority of Saskatoon neighbourhoods have a QOL score above the mean. The kurtosis value of 2.37 indicates that a significant number of neighbourhoods are concentrated in a certain QOL index range (Figure 7).

To determine the accuracy of the subjectively weighted QOL index, Saskatoon neighbourhood QOL scores were classified using two classification schemes: natural break method (or Jenks method) and standard deviation method. The former scheme represents the most distinct groupings of QOL while the latter shows how much a neighbourhood’s QOL index deviates from the mean. Both classification results were then compared to the socio-economic status (SES) of the neighbourhoods as determined by CUISR. This was done to discern if higher QOL scores corresponded with High and Middle SES neighbourhoods and if lower QOL scores corresponded with the Low SES neighbourhoods. Results (Figure 8 and Figure 9) demonstrated that the majority of Low SES neighbourhoods corresponded with the neighbourhoods having the lowest QOL scores. In both schemes, 11 out of 17 Low SES neighbourhoods represented the lowest QOL scores. It was also found that all 7 High SES neighbourhoods corresponded with the group of highest QOL scores. An overwhelming majority of the Middle SES neighbourhoods were either classified as above the mean or close to the mean (30 out of 34) or as good, very good, and advantaged (29 out of 34). The map comparison led to the conclusion that the QOL gap described by CUISR is reflected in our subjectively weighted QOL index’s results.

The spatial pattern of Saskatoon neighbourhood QOL index was detected based on two cluster statistical analysis named Luc Anselin’s Local Indicators of Spatial Autocorrelation (LISA) and Getis-Ord Gi. The mapped results (Figure 10, Figure 11) show the clusters and outliers found to be significant at the 95% confidence interval. Both spatial conceptualizations in Figure 10 show a cluster of low values on the west side in the core neighbourhoods and a cluster of high values in the southeast corner of the city. The k-nearest conceptualization differs slightly and recognizes Montgomery Place as a feature with a high value surrounded by lower values. Nutana SC is also recognized as an outlier but is a low value surrounded by higher values. The spatial autocorrelation statistics calculated by Getis-Ord Gi* (Figure 11) strongly resemble the findings of LISA. Z score returned determines whether high or low values cluster together. The larger statistically significant positive Z score represents a more intense clustering of high values or hot spot; the smaller statistically significant negative Z score indicates a more intense clustering of low values or cold spot. Both spatial conceptualizations depict a cold spot in the core neighbourhoods of the west side. The same clustering of high values found by LISA is also reported in the southeast part of the study area. However this hot spot is more expansive than reported by LISA. Both local measures of spatial autocorrelation and both spatial conceptualizations reaffirmed one another’s findings.

Like a significant amount of inner city neighbourhoods in cities across Canada and North America, neighbourhoods with lowest QOL scores are among the oldest areas of Saskatoon are the core areas where poverty flourishes. There is a disparity between these neighbourhoods and other parties of the city when it comes to things such as health and crime [109,110]. Neighbourhoods with high QOL scores clustered on the east side of the city further highlight this disparity. There is a perceived east-west divide that is common among Saskatoon residents and the authors, as residents of the city, are well aware of it. Clearly, this divide does exist in some capacity. On the other hand, not all neighbourhoods fit into this perception perfectly. Outliers included the neighbourhoods Montgomery Place and Nutana SC. Montgomery Place, although on the outskirts of the core neighbourhoods, was recognized as having significantly higher QOL than its surrounding neighbourhoods. It was developed in the mid1940s as a neighbourhood for returning WWII veterans and their families, more considered as a resort village than a neighbourhood due to an open park-like feeling. Nutana SC, which had a significantly lower QOL scores than its surrounding neighbourhoods, consists mainly of senior citizens and this may be in part why QOL is significantly lower here. Age was found to be a significant determinant when QOL was rated as lower [111].

As part of our research on urban sustainability evaluation based on the integrated USI model, the case study in this paper provides an example of QOL index calculation and spatial pattern detection. For other USI indices calculation, information extracted from satellite imagery and historical maps (e.g., urban land use change, urban sprawl, green space, etc.) are needed to be incorporated. In addition, to detect the urban sustainable progress between inter-generation, USI indices in different time periods are required to be compared; this could be conducted in GIS platform based on its spatial-temporal analysis functions.