Developing Sustainable City Indicators for Cambodia through Delphi Processes of Panel Surveys

: Currently, Cambodia does not have sustainable city indicators, and green and clean city indicators are also limited compared to UN Sustainable Development Goal 11 (SDG 11) indicators. Therefore, this research aims to develop sustainable city indicators for Cambodia and to address the questions “Are the green and clean city indicators limited in terms of sustainability?” and “Are the UN SDG 11 indicators suitable for Cambodia?” Delphi processes of panel surveys were conducted to develop the indicators in Round 1, pre-validate the indicators in Round 2, and validate the indicators in Round 3. The results showed that 69 initial indicators were obtained from Round 1; 41 pre-validated indicators were obtained from Round 2; and ﬁnally, 32 validated indicators were obtained from Round 3. All of the 32 indicators reached consensus. Based on the consensus indicators, the green and clean city indicators were found to be limited in terms of sustainability, and the UN SDG 11 has nine indicators suitable for Cambodia. These ﬁndings could be useful for applying the UN SDG 11 indicators to Cambodia and transforming the green and clean city indicators to sustainable city indicators. The 32 consensus indicators could be used as alternative sustainable city indicators for Cambodia. demography: Population growth and density, urban-rural migrations, and household economic conditions; (ii) employment: Labor force management and improvement, unemployment reduction, and new job creation; (iii) housing: Low-income house development, quality of living, and slum upgrading; (iv) transport: Public transport system and accessibility, parking lots and sidewalk improvement, tra ﬃ c congestion reduction, and road infrastructure expansion; (v) safety: Crime prevention measure, disaster prevention facility, construction safety management, and insurance enhancement; (vi) water use: Potable water supply infrastructure improvement, sustainable water consumption and production, water reservoir creation, and natural water source conservation; (vii) waste management: Waste collection network, waste reduction and recycling, and wastewater treatment plant capacity improvement; (viii) air quality and energy: Air quality management and improvement, urban forest area conservation, e ﬃ cient energy consumption and production, and renewable energy system improvement; and (ix) urban space and tourism: Urban park creation and management, botanic garden preparation, quality playground development, sport and leisure area enhancement, cultural, historical, and heritage building conservation, and tourism infrastructure and facility development.


Introduction
Following global trends on low-carbon development, Cambodia produced the national green growth roadmap in 2009 (published in 2010) to suggest the win-win-win situations between the economy, environment, and society in order to achieve stable economic growth, environmental sustainability, and human well-being [1,2]. In 2012, Cambodia, after signing the agreement to establish the Global Green Growth Institute (GGGI) as an international organization (headquarters in Seoul) with fifteen other founding member countries at the United Nations Conference on Sustainable Development in Brazil [3,4], established the National Council on Green Growth (NCGG) to coordinate the low-carbon development of the country [5]. In 2013, the government approved the national policy and national strategic plan on green growth 2013-2030 to promote sustainable long-term economic, environmental, and social development in Cambodia [6,7].
With the new agenda on sustainable development goals of the United Nations (UN SDGs), Cambodia established the National Council for Sustainable Development (NCSD) in 2015 by combining the NCGG with other relevant institutions to promote the sustainable development in Cambodia [8]. In August 2016, Minister of Environment as the Chair of NCSD argued in the green city strategic

Relevant Indicator Selection, Review, and Classification
As mentioned in Section 2.2, this research selected, reviewed, and classified the five major sources of the indicators as follows: (1) UN Sustainable Development Goals (SDGs) are a major step forward and an improvement on the Millennium Development Goals (MDGs) [11,85] and agreed in the UN 2030 Agenda for Sustainable Development. The SDGs addressed 17 goals and 169 targets [86][87][88][89]. The goal 11 addressed 10 targets, and its indicators were reviewed in Table 1 in the column 'SDG 11 [30,31]. (2) ASEAN Environmentally Sustainable City (ESC) is the initiative of ASEAN, the Association of Southeast Asian Nations (Brunei, Cambodia, Indonesia, Laos, Malaysia, Myanmar, Philippines, Singapore, Thailand, and Vietnam), which was endorsed by the ASEAN Environment Ministers in 2005 in order to pursue environmental sustainability in the rapidly-growing cities of ASEAN [90]. The ESC indicators were reviewed in Table 1  The developed indicators were reviewed in Table 1 in the column 'HAN'. (4) GC refers to the green city development project in Cambodia. The government through its NCSD implemented the project named 'Green Urban Development Program' with the support from GGGI and produced the green city strategic planning methodology [9] and green city strategic plan for Phnom Penh 2017 to 2026 [16]. The indicators attached with their sectoral objectives were reviewed in Table 1 in the column 'GC' [16]. (5) CC refers to the clean city standard of Cambodia produced by the National Committee for Clean City Assessment in order to monitor and evaluate the cities through a clean city contest every Sustainability 2019, 11, 3166 4 of 32 three years. The winning cities will be awarded by the Prime Minister of Cambodia in the following three names 'Clean City Romduol I, II, and III' upon the winning score [32,33]. The CC indicators were reviewed in Table 1 in the column 'CC' [15].
categories for developing questionnaires for a Delphi panel survey in Round 1. Thirdly, the round-one Delphi process of the panel survey was conducted to initially develop sustainable city indicators. The initial indicators were developed as follows: (1) Similar or same input indicators were combined into one (the results are shown in the frequency of the panelists providing same indicators); and (2) measurement-lacked indicators were supplemented based on the measurements of the reviewed indicators. Fourthly, the round-two Delphi process of the panel survey was conducted to identify the level of importance of the indicators (to pre-validate the indicators). Fifthly, the round-three Delphi process of the panel survey was conducted to confirm the level of importance of the indicators (to validate the indicators). Finally, the consensus was analyzed after the level of importance confirmed. The research flow is illustrated in Figure 1.  Based on the explanation above, the UN SDG 11 indicators are globally related to Cambodia; the ASEAN ESC indicators are regionally related to Cambodia; and the Korean HAN indicators are potentially related to Cambodia. Furthermore, the green and clean city indicators are domestic available indicators that are related to sustainability. The relations and scopes of these indicators are illustrated in Figure 2.

Relevant Indicator Selection, Review, and Classification
As mentioned in Section 2.2, this research selected, reviewed, and classified the five major sources of the indicators as follows: (1) UN Sustainable Development Goals (SDGs) are a major step forward and an improvement on the Millennium Development Goals (MDGs) [11,85] and agreed in the UN 2030 Agenda for Sustainable Development. The SDGs addressed 17 goals and 169 targets [86][87][88][89]. The goal 11 addressed 10 targets, and its indicators were reviewed in Table 1 in the column 'SDG 11′ [30,31]. (2) ASEAN Environmentally Sustainable City (ESC) is the initiative of ASEAN, the Association of Southeast Asian Nations (Brunei, Cambodia, Indonesia, Laos, Malaysia, Myanmar, Philippines, Singapore, Thailand, and Vietnam), which was endorsed by the ASEAN Environment Ministers in 2005 in order to pursue environmental sustainability in the rapidly-growing cities of ASEAN [90]. The ESC indicators were reviewed in Table 1 in the column 'ESC' [91]. (3) Korean HAN refers to the Korean case study indicators, which were developed by Sang-mi Han and Myeong-hun Lee [92]. This research developed indicators based on the UN Sustainable Development Goal 11 indicators, HABITAT indicators and Korea's relevant indicators. The developed indicators were reviewed in Table 1 in the column 'HAN'. (4) GC refers to the green city development project in Cambodia. The government through its NCSD implemented the project named 'Green Urban Development Program' with the support from GGGI and produced the green city strategic planning methodology [9] and green city strategic plan for Phnom Penh 2017 to 2026 [16]. The indicators attached with their sectoral objectives were reviewed in Table 1 in the column 'GC' [16]. (5) CC refers to the clean city standard of Cambodia produced by the National Committee for Clean City Assessment in order to monitor and evaluate the cities through a clean city contest every three years. The winning cities will be awarded by the Prime Minister of Cambodia in the following three names 'Clean City Romduol I, II, and III' upon the winning score [32,33]. The CC indicators were reviewed in Table 1 in the column 'CC' [15].
Based on the explanation above, the UN SDG 11 indicators are globally related to Cambodia; the ASEAN ESC indicators are regionally related to Cambodia; and the Korean HAN indicators are potentially related to Cambodia. Furthermore, the green and clean city indicators are domestic available indicators that are related to sustainability. The relations and scopes of these indicators are illustrated in Figure 2. Moreover, the reviewed indicators which have parallel descriptions were merged and listed in the same row (see Table 1). These indicators, after classifying by their similarity and correlation,  Moreover, the reviewed indicators which have parallel descriptions were merged and listed in the same row (see Table 1). These indicators, after classifying by their similarity and correlation, have nine relative categories, such as (1) Demography, (2) Jobs and Tourism, (3) Housing, (4) Transport, (5) Safety, (6) Clean Air and Energy, (7) Waste Management, (8) Water Use, and (9) Public Space and Heritage. These categories are also shown in Table 1 in the column 'Category'.

Questionnaire Development
Delphi processes of panel surveys were conducted in three rounds; therefore, the development of questionnaires was not done in one time in this research. Three different questionnaires were developed at different times. The development of questionnaires was as follows: Round 1: As mentioned in previous sections, the questionnaires for Round 1 were developed based on the relative categories of the reviewed indicators. Generally, sustainable city indicators are various and broad, so it would be difficult and complicated for panelists if they were asked to provide indicators without categories. Therefore, this research developed questionnaires, based on the nine relative categories, into 9-part questions. A sample of a questionnaire in brief for the Demography is shown in Figure 3a. More detail on how it works is explained in Section 2.5.3.
Round 2: Questionnaires for Round 2 were developed based on the initial indicators obtained from Round 1. The purpose of this round was to identify the level of importance of the indicators (preliminarily validate the initially developed indicators). Therefore, this research developed the questionnaires for this round by using a 5-point Likert-type scale. A sample of a questionnaire, in brief, is shown in Figure 3b. More detail on how it works is explained in Section 2.5.3.

Questionnaire Development
Delphi processes of panel surveys were conducted in three rounds; therefore, the development of questionnaires was not done in one time in this research. Three different questionnaires were developed at different times. The development of questionnaires was as follows: Round 1: As mentioned in previous sections, the questionnaires for Round 1 were developed based on the relative categories of the reviewed indicators. Generally, sustainable city indicators are various and broad, so it would be difficult and complicated for panelists if they were asked to provide indicators without categories. Therefore, this research developed questionnaires, based on the nine relative categories, into 9-part questions. A sample of a questionnaire in brief for the Demography is shown in Figure 3a. More detail on how it works is explained in Section 2.5.3.
Round 2: Questionnaires for Round 2 were developed based on the initial indicators obtained from Round 1. The purpose of this round was to identify the level of importance of the indicators (preliminarily validate the initially developed indicators). Therefore, this research developed the questionnaires for this round by using a 5-point Likert-type scale. A sample of a questionnaire, in brief, is shown in Figure 3b. More detail on how it works is explained in Section 2.5.3.
Round 3: Questionnaires for Round 3 were developed based on the pre-validated indicators obtained from Round 2. The purpose of this round was to confirm the level of importance of the indicators (to finally validate the indicators). Therefore, questionnaires were developed by adding Round 2 mean values in front of the 5-point Likert-type scale. A sample of a questionnaire in brief for this round is shown in Figure 3c. More detail on how it works is explained in Section 2.5.3.  indicators (to finally validate the indicators). Therefore, questionnaires were developed by adding Round 2 mean values in front of the 5-point Likert-type scale. A sample of a questionnaire in brief for this round is shown in Figure 3c. More detail on how it works is explained in Section 2.5.3.

Place of Panel Surveys
With the administrative support from the Ministry of Environment, Department of Green Economy, and General Secretariat of the National Council for Sustainable Development, this research conducted the 3-round Delphi processes of panel surveys at the 'Tung Hall' meeting room of the Ministry of Environment in Phnom Penh capital, the Kingdom of Cambodia.

Panelist Selection, Invitation, and Participation
In general, Delphi studies use different sizes of panels [63]. Weidman and colleagues (2011) did not specify the number of panelists needed for a Delphi study; however, it is recognized that a minimum appropriate size should include 7 or 8 panelists [79]. Mitchell and McGoldrick (1994) argued that the size should be no less than 8 to 10 members [38]. Hallowell and Gambatese (2010) mentioned that most studies incorporated 8 to 16 panelists and suggested a minimum of 8 [78]. Following these studies, this research determined that the number of panelists for each round should be no less than 10. Consequently, this research selected 20 professional panelists whose fields were in and related to urban sustainability, such as sustainable urban and rural development (4), eco-labelling and green buildings (2), green energy and economic policies (3), climate change policy coordination (2), greenhouse gas inventory and reduction strategies (2), administration, planning, and finance (2), ecosystem services (1), biodiversity conservation (1), sustainable consumption and production (1), disaster risk assessment (1), and green technologies (1). All panelists were invited to participate in the panel surveys by the General Secretariat of the National Council for Sustainable Development. A total of 16 of them came to both Round 1 and 2, and 10 of them came to Round 3.

Panel Survey Processes
In this research, every panelist was briefed on the research purposes and Delphi processes of panel surveys in the concept note attached with the invitation. Most importantly, these matters were explained again after a set of questionnaires were distributed immediately before starting the survey in Round 1. This explanation was to further focus on how to complete the questionnaires. With the fill-in-blank questions, see Figure 3a, panelists were asked to provide at least 4 sustainable city indicators for Cambodia for each given category. Extra blanks for other categories were also provided in case panelists wanted to add the indicators that were not related to the given categories. Similarly, Sourani and Sohail (2015) asked respondents to provide five major criteria, and respondents were notified that they were welcome to provide more than five criteria [63].
In Round 2, the panelists were instructed how to complete the questionnaires after a set of questionnaires were distributed immediately before starting the survey. As shown in Figure 3b, the questionnaires were developed based on the initial indicators obtained from Round 1 by adding a 5-point Likert-type scale in front of the indicators as follows-1 represents "not important"; 2 represents "less important"; 3 represents "important"; 4 represents "very important"; and 5 represents "extremely important" [63,76]. The panelists were asked to assess the level of importance of the initial indicators.
In Round 3, the panelists were also told about how to complete the questionnaires after distribution immediately before starting the survey. As shown in Figure 3c, the questionnaires were developed based on the pre-validated indicators obtained from Round 2 by adding Round 2 mean value and 5-point Likert-type scale in front of the indicators. The purpose of this round was to re-check the level of importance of the pre-validated indicators because the individual panelists could change their opinions and judgements based on the Round 2 mean value. Likewise, Gene and George (1999) argued that with the iteration, the individuals are given the opportunity to change their opinions and judgments [57]. Therefore, the panelists in this round were asked to re-assess the level of importance of the pre-validated indicators based on the statistical mean value obtained from Round 2.

Analyzing Consensus and Investigation
In this research, the consensus was not analyzed in Round 2 because, as explained in the previous sections, the level of importance of the indicators in Round 2 was not yet definitely confirmed; however, the Round 2 mean value helped panelists to make final decision to confirm the level of importance of the indicators in Round 3. Therefore, this research analyzed the consensus in Round 3. According to Hughes (2003) [84], the consensus is to determine the extent to which panelists agree about a given issue. Quite commonly, it is defined by more than 83% of the responses voting for the issue as "important or very important or extremely important". Sourani and Sohail (2015) [63] proposed that the developed consensus is based on identifying and assigning, for each criterion, the mean value, standard deviation, and percentage of respondents agreeing on ranking the criterion. On a 5-point Likert-type scale, the consensus can be established based on reaching 75% or above of the respondents agreeing on a given rank that is equal to or above 3.00. Therefore, this research accordingly determined the consensus based on the confirmed level of importance of the indicators at equal to or above 3.00, and the percentage of panelists agreeing on a given rank from 3 to 5 at equal to or above 75%. The validated indicators that reached this consensus were chosen as sustainable city indicators (SCIs) for Cambodia.
Furthermore, the consensus sustainable city indicators are, based on the validations and agreements, suitable for Cambodia's urban context for the development of sustainable cities. Therefore, this research sought to address the questions "Are the green and clean city indicators limited in terms of sustainability?" and "Are the UN SDG 11 indicators suitable for Cambodia?" based on the consensus indicators. In addition, how the green and clean city and UN SDG 11 indicators are limited or suitable for Cambodia in terms of sustainability was also investigated through the correlation of these indicators with the consensus indicators.

Indicator Development in Delphi Round 1
After analyzing the results of Delphi Round 1, this research obtained the categories and initial indicators as follows. The number of categories remained the same as the number of the reviewed indicators (9 categories), but three categories underwent a name change. The 'jobs and tourism' and 'public space and heritage' were changed to 'employment' and 'urban space and tourism'. The 'clean air and energy' was changed to 'noise, air quality, and energy' (see Table 2). These changes were based on the obtained initial indicators and comments and suggestions from the panelists. Percentage of the population living in slums SCI18 Percentage of new residential buildings (aged less than 30 years) SCI19 Percentage of the population living in owned houses SCI20 Percentage of the population living in rent houses SCI21 Percentage of aging residential buildings (aged more than 30 years) SCI22 The radio of middle-income houses compared to the low-income houses Transport SCI23 Public transport sharing rate SCI24 Number of initiatives to reduce traffic congestion SCI25 Number of initiatives for sidewalk improvement SCI26 Number of public parking lots in the city SCI27 Percentage of the population living within 0.5 km of public transport access SCI28 Percentage of environmentally friendly vehicles SCI29 Number Percentage of households with tap water that meets WHO drinking water standard SCI40 Number of natural or artificial reservoirs in or nearby the city SCI41 Average of the water consumption rate per person or household SCI42 Percentage of water use in production SCI43 The ratio of water use from underground and surface water SCI44 Percentage of school at all levels with water conservation education programs Table 2. Cont.

Category Indicator
Waste Management

SCI45
Percentage of solid waste regularly collected SCI46 The capacity of wastewater treatment plants in the city SCI47 Percentage of households linked to the sewerage system SCI48 Number of initiatives to reduce wastes SCI49 Percentage of reduction in total waste generated per year SCI50 Percentage of recycled waste from total waste generation SCI51 Number of households with secured sanitation systems SCI52 Percentage of waste collected from door to door or collection point Noise, Air Quality and Energy

SCI53
Average of the energy consumption rate per person or household SCI54 Percentage of the urban forest SCI55 Fine dust levels SCI56 Number of measures or initiatives to control noise in the city SCI57 Absence or presence of greenhouse gas reduction strategies SCI58 Percentage of renewable energy contributed to the electricity supply SCI59 Percentage of hotels using energy saving devices SCI60 Percentage of green buildings in the city Urban Space and Tourism

SCI61
The ratio of public and green spaces compared to the total area of the city SCI62 Number of natural or artificial parks in the city SCI63 Percentage of residents living within 0.5 km of public and green space access SCI64 Tourism growth rate per year SCI65 Number of conserved cultural, historical, and heritage buildings SCI66 Number of tourism firms providing tourism services in the city SCI67 Number of botanic gardens in the city SCI68 Number of playgrounds in the city SCI69 Number of zoological gardens in the city Furthermore, the maximum number of the initial indicators is 9 and the minimum number of the initial indicators is 7. Number of indicators obtained in each category are 8 in 'demography', 7 in 'employment', 7 in 'housing', 8 in 'transport', 7 in 'safety', 7 in 'water use', 8 in 'waste management', 8 in 'noise, air quality and energy', and 9 in 'urban space and tourism' (see Figure 4).  The highest frequency is 100% (16/16) and the lowest frequency is 6% (1/16). The highest-frequency indicators are SCI1 in 'demography', SCI23 in 'transport', SCI31 in 'safety' and SCI45 in 'waste management'. The top indicator in each category are SCI1 in 'demography', SCI9 in 'employment', SCI16 in 'housing', SCI23 in 'transport', SCI31 in 'safety', SCI38 in 'water use', SCI45 in 'waste management', SCI53 and SCI54 in 'noise, air quality and energy', and SCI61 in 'urban space and tourism' (see Figure 5). The highest frequency is 100% (16/16) and the lowest frequency is 6% (1/16). The highest-frequency indicators are SCI1 in 'demography', SCI23 in 'transport', SCI31 in 'safety' and SCI45 in 'waste management'. The top indicator in each category are SCI1 in 'demography', SCI9 in 'employment', SCI16 in 'housing', SCI23 in 'transport', SCI31 in 'safety', SCI38 in 'water use', SCI45 in 'waste management', SCI53 and SCI54 in 'noise, air quality and energy', and SCI61 in 'urban space and tourism' (see Figure 5).

Indicator Pre-Validation in Delphi Round 2
After analyzing the results of Delphi Round 2, this research obtained the pre-validated indicators as follows. The maximum number of indicators is 6 and the minimum number of indicators is 3. Number of the indicators obtained in each category are 4 in 'demography', 5 in 'employment', 3 in 'housing', 5 in 'transport', 4 in 'safety', 5 in 'water use', 4 in 'waste management', 5 in 'noise, air quality and energy', and 6 in 'urban space and tourism' (see Figure 6).

Indicator Pre-Validation in Delphi Round 2
After analyzing the results of Delphi Round 2, this research obtained the pre-validated indicators as follows. The maximum number of indicators is 6 and the minimum number of indicators is 3. Number of the indicators obtained in each category are 4 in 'demography', 5 in 'employment', 3 in 'housing', 5 in 'transport', 4 in 'safety', 5 in 'water use', 4 in 'waste management', 5 in 'noise, air quality and energy', and 6 in 'urban space and tourism' (see Figure 6). The maximum mean value is 4.63 and the minimum mean value is 3.00. The maximum-mean-value indicator is SCI38 in 'water use'. The top indicator (maximum-mean-value indicator) in each category are SCI1 in 'demography', SCI12 in 'employment', SCI6 in 'housing', SCI25 in 'transport', SCI31 and SCI33 in 'safety', SCI38 in 'water use', SCI46 in 'waste management', SCI54 in 'noise, air quality and energy', and SCI62 in 'urban space and tourism' (see Figure 7). The maximum mean value is 4.63 and the minimum mean value is 3.00. The maximum-mean-value indicator is SCI38 in 'water use'. The top indicator (maximum-mean-value indicator) in each category are SCI1 in 'demography', SCI12 in 'employment', SCI6 in 'housing', SCI25 in 'transport', SCI31 and SCI33 in 'safety', SCI38 in 'water use', SCI46 in 'waste management', SCI54 in 'noise, air quality and energy', and SCI62 in 'urban space and tourism' (see Figure 7). The maximum mean value is 4.63 and the minimum mean value is 3.00. The maximum-mean-value indicator is SCI38 in 'water use'. The top indicator (maximum-mean-value indicator) in each category are SCI1 in 'demography', SCI12 in 'employment', SCI6 in 'housing', SCI25 in 'transport', SCI31 and SCI33 in 'safety', SCI38 in 'water use', SCI46 in 'waste management', SCI54 in 'noise, air quality and energy', and SCI62 in 'urban space and tourism' (see Figure 7).

Indicator Validation in Delphi Round 3
After analyzing the results of Delphi Round 3, this research obtained the categories and validated indicators as follows. One category named 'noise, air quality, and energy' was changed to 'air quality and energy' because noise indicator in this category obtained mean value (level of importance) lower than 3.00 after final validation. The maximum number of indicators is 5 and the minimum number of indicators is 3. Number of the indicators obtained in each category are 3 in 'demography', 3 in 'employment', 3 in 'housing', 4 in 'transport', 4 in 'safety', 3 in 'water use', 3 in 'waste management', 4 in 'air quality and energy', and 5 in 'urban space and tourism' (see Figure 8).
The maximum mean value is 4.80 and the minimum mean value is 3.50. The maximum-mean-value indicator is SCI25 in 'transport'. The top indicator (maximum-mean-value indicator) in each category are SCI3 in 'demography', SCI12 in 'employment', SCI16 in 'housing', SCI25 in 'transport', SCI31 in 'safety', SCI38 in 'water use', SCI45 in 'waste management', SCI55 in 'air quality and energy', and SCI67 and SCI62 in 'urban space and tourism' (see Figure 9). validated indicators as follows. One category named 'noise, air quality, and energy' was changed to 'air quality and energy' because noise indicator in this category obtained mean value (level of importance) lower than 3.00 after final validation. The maximum number of indicators is 5 and the minimum number of indicators is 3. Number of the indicators obtained in each category are 3 in 'demography', 3 in 'employment', 3 in 'housing', 4 in 'transport', 4 in 'safety', 3 in 'water use', 3 in 'waste management', 4 in 'air quality and energy', and 5 in 'urban space and tourism' (see Figure 8). The maximum mean value is 4.80 and the minimum mean value is 3.50. The maximum-mean-value indicator is SCI25 in 'transport'. The top indicator (maximum-mean-value indicator) in each category are SCI3 in 'demography', SCI12 in 'employment', SCI16 in 'housing', SCI25 in 'transport', SCI31 in 'safety', SCI38 in 'water use', SCI45 in 'waste management', SCI55 in 'air quality and energy', and SCI67 and SCI62 in 'urban space and tourism' (see Figure 9).  The maximum mean value is 4.80 and the minimum mean value is 3.50. The maximum-mean-value indicator is SCI25 in 'transport'. The top indicator (maximum-mean-value indicator) in each category are SCI3 in 'demography', SCI12 in 'employment', SCI16 in 'housing', SCI25 in 'transport', SCI31 in 'safety', SCI38 in 'water use', SCI45 in 'waste management', SCI55 in 'air quality and energy', and SCI67 and SCI62 in 'urban space and tourism' (see Figure 9).

Consensus Analysis
Based on the determined consensus, the developed indicators need to be confirmed with regard to the level of importance at equal to or above 3.00 and percentage voting as 5 or 4 or 3 at equal to or above 75%. According to the consensus analysis, percentage of panelists voting as 5 or 4 or 3 for the 32 validated indicators is 100% for 28 indicators and 90% for 4 indicators. Therefore, all of the 32 indicators reached consensus (see Table 3).   Table A2 (Appendix B).

Correlation of Green and Clean City, UN SDG 11, and Consensus Indicators
According to Table 4, the green and clean city indicators correlate with 11 consensus indicators. However, these indicators have no correlation in 'demography', 'employment', and 'water use' and only one for each in 'housing' and 'air quality and energy'. Furthermore, the UN sustainable development goal 11 indicators correlate with nine consensus indicators. This number is less than the number of green and clean city indicators correlated with the consensus indicators. However, these indicators correlate with all categories of the consensus indicators (one indicator for each category). The consensus indicators and their short explanations are shown in Table A3.  The number of indicators in each category decreased from one round to the next round until reaching the consensus. The number of indicators decreased from 8 to 3 in 'demography', 7 to 3 in 'employment' and 'housing', 8 to 4 in 'transport', 7 to 4 in 'safety', 7 to 3 in 'water use', 8 to 3 in 'waste management', 8 to 4 in 'air quality and energy', and 9 to 5 in 'urban space and tourism' (Figure 11). Furthermore, the top indicator in each category has also changed the rank from Round 1 to Round 3. There are only three top indicators have not changed their rank, since Round 1 until reaching the consensus. These top indicators are SCI16 in 'housing', SCI31 in 'Safety', and SCI38 in 'water use'. Moreover, how all indicators have changed their ranks is also shown in Table A2 (Appendix B).

Correlation of Green and Clean City, UN SDG 11, and Consensus Indicators
According to Table 4, the green and clean city indicators correlate with 11 consensus indicators. However, these indicators have no correlation in 'demography', 'employment', and 'water use' and only one for each in 'housing' and 'air quality and energy'. Furthermore, the UN sustainable development goal 11 indicators correlate with nine consensus indicators. This number is less than the number of green and clean city indicators correlated with the consensus indicators. However, these indicators correlate with all categories of the consensus indicators (one indicator for each category). The consensus indicators and their short explanations are shown in Table A3.

Correlation of Green and Clean City, UN SDG 11, and Consensus Indicators
According to Table 4, the green and clean city indicators correlate with 11 consensus indicators. However, these indicators have no correlation in 'demography', 'employment', and 'water use' and only one for each in 'housing' and 'air quality and energy'. Furthermore, the UN sustainable development goal 11 indicators correlate with nine consensus indicators. This number is less than the number of green and clean city indicators correlated with the consensus indicators. However, these indicators correlate with all categories of the consensus indicators (one indicator for each category). The consensus indicators and their short explanations are shown in Table A3.

Discussion
This research started from the review of five major source indicators, such as UN Sustainable Development Goal 11, ASEAN Environmentally Sustainable City, Korean HAN, and domestic green and clean city indicators. These reviewed indicators were classified based on their similarity and correlation in order to identify relative categories for developing questionnaires for Delphi Round 1. These indicators were also used to supplement the measurement-lacked indicators obtained from Delphi Round 1. Even though the indicator development was initially started by these processes, the obtained 69 initial indicators are various compared to the reviewed indicators. This shows that the reviewed indicators are not inclusively suitable for the development and management of sustainable cities in Cambodia based on the panelists' opinions. Furthermore, the 69 initial indicators were reduced to 41 indicators through the pre-validation process in Round 2. This shows that 28 of the initial indicators were assessed by the panelists as 'not or less important' in Round 2. Moreover, the 41 indicators were further reduced to 32 indicators through the final validation process in Round 3. This shows that nine of the pre-validated indicators were re-assessed by the panelists as 'not or less important' in Round 3 even though these indicators were assessed as 'important' in Round 2. In this case, perhaps individual panelist firstly thought that the indicators they provided in Round 1 were important, but after seeing all initial indicators, they changed their opinions and judgments accordingly. Likewise, after seeing the mean value (level of importance of the pre-validated indicators) obtained from Round 2, individual panelist made the final decision to further reduce the not-or-less-important indicators. However, all of the 32 validated indicators reached consensus. Especially, the confirmed levels of importance and consensus rates of the indicators are strong and high. The levels of importance are all equal to or above 3.50 and the consensus rates are all equal to or above 90.00%. The 32 consensus indicators could, therefore, be significant for the measurement of sustainable city development in Cambodia.
The green and clean city indicators have no correlation with the consensus indicators in 'demography', 'employment', and 'water use' and only one for each in 'housing' and 'air quality and energy'. This shows that current green and clean city indicators are limited in terms of sustainability. Especially, these indicators are very limited in terms of 'housing, air quality, and energy' indicators. Most importantly, these indicators are extremely limited in terms of 'demography, employment, and water use' indicators. These findings could be useful for making decisions on the improvement of the green and clean city indicators and transformation of these indicators to sustainable city indicators. Based on the correlation with consensus indicators, transforming these indicators to sustainable city indicators should be made as follows: (a) Indicators for demography need to be comprised of household income, population density, and urban-rural migration. (b) indicators for employment need to be covered on labor force participation, unemployment, and new job creation. (c) indicators for housing need to be enhanced to residential quality improvement and slum upgrading. (d) indicators for transport need to be heightened to the public parking lot and sidewalk improvement. (e) indicators for safety need to be enhanced to construction safety and insurance services. (f) indicators for water use needs to be comprised of potable water accessible, water consumption rate, and reservoir conservation and maintenance. (g) indicators for air quality and energy need to be enhanced to fine dust level, urban forest areas, and energy consumption rate. (h) indicators for urban space and tourism need to be covered on urban park creation and maintenance, botanic garden preparation, and playground quality and management.
The UN Sustainable Development Goal 11 (SDG 11) indicators have correlations with nine consensus indicators. These indicators correlate with SCI1 (population density) in 'demography', SCI10 (unemployment rate) in 'employment', SCI17 (slum population) in 'housing', SCI23 (public transport sharing rate) in 'transport', SCI32 (disaster prevention) in 'safety', SCI38 (potable water supply) in 'water use', SCI45 (regular solid waste collection) in 'waste management', SCI55 (fine dust level) in 'air quality and energy', and SCI65 (conserved cultural, historical, and heritage buildings) in 'urban space and tourism'. This shows that the UN SDG 11 has nine indicators suitable for applying to Cambodia. However, when applying these indicators to Cambodia, their measurement should be revised for Cambodia's urban context. In this case, the consensus indicators and their explanations could be useful for this modification. Moreover, these correlated indicators of UN SDG 11 distribute in all categories of the consensus indicators. This shows that the consensus indicators' categories are appropriate to use for developing and/or classifying the sustainable city indicators for Cambodia towards achieving urban sustainability. Consequently, policymaking for the development and management of sustainable cities should focus on these categories and their consensus indicators. The policymaking processes must be involved by all relevant agencies, especially the agencies that have roles and responsibilities related to these categories and indicators. The most relevant agencies should play roles as implementing agencies for the development projects, and other relevant agencies could play roles as supporting or participating agencies.
Based on the findings of this research and the roles and responsibilities of the government agencies [8,93], the sectoral development of sustainable cities should be responsible by the Ministry of Planning (sustainable urban demographic management and improvement), Ministry of Labor and Vocational Training (sustainable urban employment development and management), Ministry of Land Management, Urban Planning, and Construction (sustainable urban housing development and urban spaces management), Ministry of Public Works and Transport (sustainable urban transport development and management), Ministry of Interior (sustainable urban safety system development and management), Ministry of Industry and Handicraft (sustainable urban water consumption and production improvement), Ministry of Environment (sustainable urban waste management, noise control, and air quality improvement), Ministry of Mines and Energy (sustainable urban energy consumption and production improvement), and Ministry of Tourism (sustainable urban tourism development and management). The other agencies which have roles and responsibilities related to the consensus indicators should be supporting and involving in the policymaking processes as well. The National Council for Sustainable Development (NCSD) should be responsible for the policy alignment for the inclusive development of sustainable cities. Most importantly, the NCSD should be responsible for the inclusive assessment of sustainable city development by coordinating with the above-mentioned nine sectoral responsible agencies. In this assessment, inclusive sustainable city indicators are needed. The indicators could be developed by upgrading the green and clean city indicators. The 32 consensus indicators could be the alternative indicators.
Based on the consensus indicators, policymaking for the development of sustainable cities should focus on (i) demography: Population growth and density, urban-rural migrations, and household economic conditions; (ii) employment: Labor force management and improvement, unemployment reduction, and new job creation; (iii) housing: Low-income house development, quality of living, and slum upgrading; (iv) transport: Public transport system and accessibility, parking lots and sidewalk improvement, traffic congestion reduction, and road infrastructure expansion; (v) safety: Crime prevention measure, disaster prevention facility, construction safety management, and insurance enhancement; (vi) water use: Potable water supply infrastructure improvement, sustainable water consumption and production, water reservoir creation, and natural water source conservation; (vii) waste management: Waste collection network, waste reduction and recycling, and wastewater treatment plant capacity improvement; (viii) air quality and energy: Air quality management and improvement, urban forest area conservation, efficient energy consumption and production, and renewable energy system improvement; and (ix) urban space and tourism: Urban park creation and management, botanic garden preparation, quality playground development, sport and leisure area enhancement, cultural, historical, and heritage building conservation, and tourism infrastructure and facility development.
Practically, Schumann (2016) argued that many aspects need to be considered to apply the indicators at the national level, particularly at the sub-national level; the role that indicators can play in encouraging cooperation among different sub-national governments should be taken into the account [94]. In this case, the departments and divisions of the above-mentioned responsible agencies at both sub-national and local levels could be the important agencies contributing to the efficient policymaking and play important roles in encouraging the local relevant stakeholders and residents to take part in the policymaking processes.
The above-mentioned findings and discussions could be useful to (i) the relevant institutions for making efficient policies, especially the policies for sectoral development, (ii) the development partners for seeking out collaboration with government agencies for sectoral development and inclusive assessment of sustainable cities in Cambodia, (iii) the National Council for Sustainable Development for clarifying roles and responsibilities between the sectoral development agencies, and for collaborating on the inclusive assessment of sustainable city development in Cambodia, and (iv) the government for considerably applying the UN sustainable development goal 11 indicators to Cambodia, and transforming the green and clean city indicators to sustainable city indicators. The 32 consensus indicators could be used as alternative sustainable city indicators for Cambodia.
These findings and discussions could also be useful to future research on the development, management, and assessment of the sustainable cities in Cambodia, especially the specific fields of sustainable cities. The sectoral policies for sustainable city development could be further exploring and investigating, especially how to align the sustainable urban policies with the existing urban and relevant policies. The research on a framework of the sustainable city indicators seems to be required and the prioritization of the developed indicators seems to be essential as priority weight is necessary for the sectoral development and inclusive assessment of sustainable cities.

Conclusions
Through the Delphi processes of panel surveys, this research obtained 69 initial indicators in Round 1, 41 pre-validated indicators in Round 2, and 32 validated indicators in Round 3. All of the 32 indicators reached consensus. The confirmed levels of importance were all equal to or above 3.50. The consensus rates were all equal to or above 90.00%. Based on the consensus indicators, the green and clean city indicators were found to be limited in terms of sustainability, especially in terms of 'housing, air quality, energy, demography, employment, and water use' indicators. The UN SDG 11 has nine indicators suitable for Cambodia. These indicators, however, should be modified when applying to Cambodia. The consensus indicators and their explanation would be useful for this modification. Moreover, the nine indicators distribute in all categories of the consensus indicators. These categories, therefore, could be appropriate to use for developing and/or classifying the sustainable city indicators for Cambodia towards achieving urban sustainability. Consequently, the nine categories, consensus indicators, and their explanation could be useful for policymaking for sustainable city development in Cambodia, especially policymaking for sectoral development.
This research suggests implications based on its lessons learned for future research as follows: In order to develop the indicators through Delphi processes of panel surveys which Round 1 conducts to develop the initial indicators, there should be at least two rounds for validating the initial indicators because (i) the statistical mean value from the group is very important for individual panelists to make the final decision to validate the indicators; and (ii) this situation can provide an opportunity for individuals to change their judgments based on statistical evident [57]. The consensus should be analyzed after the level of importance of the indicators is confirmed. It should be in Round 3 because (a) panelists provide judgments based on the mean value-the level of error is small, and (b) panelists get familiar with the indicators, due to three-time experiences-the judgment is accurate. In order to use the method more efficiently, a short time is efficient and identifying categories for panel surveys is a good option particularly, since the type of professional involved in this process is often busy. Especially, the availability of professional panelists and their willingness to join a panel must be considered in order to make sure they will be able to participate until the last round.
The levels of importance of the indicators in this research were verified to validate the indicators and to choose the significant indicators for Cambodia. Therefore, these levels of importance are not recommended to use for ranking or prioritizing the 32 consensus indicators. By the way, based on the method explained in Section 2.1, this research suggests future research to prioritize these indicators by using the ANP (analytic network process) or AHP (analytic hierarchy process) method. These two methods use pairwise comparison and generally require a number of respondents from relevant fields at least 100. The fields related to consensus indicators' categories are recommended. Based on the consensus indicators and their categories, this research discussed the sectoral policies and responsibilities of the government agencies for the development and assessment of sustainable cities in Cambodia. Consequently, we can see how aligning the consensus indicators into the existing urban and relevant policies is necessary, but beyond the scope of this research. Therefore, future research exploring this point would contribute to an inclusive policy alignment.

Conflicts of Interest:
The authors declare no conflict of interest. Even though the General Secretariat of the National Council for Sustainable Development helped invite the panelists and the Ministry of Environment provided the place for the panels, they had no role in the design of this research, interpretation of the data, discussion of the results, and preparation and revision of the manuscript. Table A1. Summary of Delphi technique.

Background
Delphi technique was developed during the 1950s by workers at the RAND Corporation while involved in a US Air Force sponsored project [56,57] and it is used as a systematic procedure to evoke expert opinion and usually conducted through a series of questionnaires [58][59][60][61]. It has become a widely used tool for measuring and aiding forecasting and decision making in a variety of disciplines [57,62].

Purposes
Delphi method had been used in many areas and can be used for any purpose requiring the use of committees [58][59][60][61][62]. Based on many studies, Sourani and Sohail (2015) presented Delphi as useful to (a) obtain accurate information that is unavailable, (b) handle complex problems that require more judgmental analysis, (c) define areas where there is considerable uncertainty and/or a lack of agreed knowledge or disagreement, (d) allow for combining fragmentary perspectives into a collective understanding, (e) model a real-world phenomenon involving a range of viewpoints and for which there is little established quantitative evidence, and (f) highlight topics of concern and assess uncertainty in a quantitative manner [63][64][65][66][67][68][69][70][71][72][73][74][75][76][77].

Characteristics
Delphi technique has three main characteristics: (1) Anonymity allows individuals to provide their opinions and judgments through questionnaires; (2) iteration gives opportunity to individuals to change their opinions and judgments; and (3) statistical response provides the previous round opinions and judgments through simple statistical summary (a mean or median value) [56][57][58][59][60][61][62][63]. These characteristics are necessary to define the attributes of the Delphi procedure, although there are numerous ways in which they may be applied [57]. Furthermore, the processes continues for a pre-determined number of rounds or until some predetermined criterion has been met, e.g., reaching consensus (see Figure A1) [63].

Selection of panelists
Delphi technique generally uses expertise panel. Martino (1983) argued that an expert may be defined in broad terms as 'someone who has special knowledge about the specific subject' [58] (p. 27). Furthermore, key criteria for selecting the experts to participate in a Delphi process of panel surveys are knowledge; however, criteria of willingness and availability to participate are also important to consider [58,61,78]. This attribute may be considered secondary to the knowledge or degree of expertness [58,63].

Number of panelists
In general, Delphi studies use different sizes of panels [63]. Weidman and colleagues (2011) did not specify the number of experts needed for a Delphi study; however, it is recognized that a minimum appropriate size would include 7 or 8 experts [79]. Mitchell and McGoldrick (1994) argued that the size should be no less than 8 to 10 members [38]. Hallowell and Gambatese (2010) mentioned that most studies incorporated 8 to 16 experts and suggested a minimum of 8 and they also argued that the specific number should be determined by the research characteristics [63,78,80].

Number of rounds
Number of rounds in Delphi studies is varied. Chong and Zin (2010) argued that the number of rounds should be based on the objectives of research [63,81]. Gunhan and Arditi (2005) mentioned that most changes in responses took place in the first two rounds and that little was gained after that [63,82]. Hallowell and Gambatese (2010) presented the advantage of having a three-round Delphi, which facilitates obtaining reasons for outlying responses from Round 2 and reporting these in Round 3; such a process could facilitate the consideration of all options and the attainment of a consensus about the correct value instead of conforming to an incorrect opinion [63,78].

Time consuming
By using a Delphi method to develop the agreed set of economic sustainability criteria that should be addressed in a procumbent strategy, Sourani and Sohail (2015) argued that Delphi could be time-consuming for participating experts because the experts are busy people. With the need for them to respond to several rounds, there is a considerable risk of drop out by some of them. A sufficient number of experts should, therefore, be appointed.
Measures to reduce possible fatigue should be considered, including proper research design [63].

Consensus
From a review of a wide range of Delphi studies, it is shown that authors measured consensus in different ways. Jones and Hunter [83] and Hughes [84] argued that the purpose of measuring consensus is to determine the extent to which experts agree about a given issue. Quite commonly, the extent of consensus was measured by the percentages of respondents agreeing on certain answers; for example, Sourani and Sohail [63] argued that based on a 5-point Likert-type scale, the certain criterion can be established based on having 75% or more of the respondents agreeing on giving a ranking that is equal to or above 3.

Consensus
consensus in different ways. Jones and Hunter [83] and Hughes [84] argued that the purpose of measuring consensus is to determine the extent to which experts agree about a given issue. Quite commonly, the extent of consensus was measured by the percentages of respondents agreeing on certain answers; for example, Sourani and Sohail [63] argued that based on a 5-point Likert-type scale, the certain criterion can be established based on having 75% or more of the respondents agreeing on giving a ranking that is equal to or above 3. Figure A1. Delphi processes. Source: Sourani and Sohail (2015). Figure A1. Delphi processes. Source: Sourani and Sohail (2015).  √ " refers to 'confirmed level of importance and reached consensus'; "x" refers to 'did not confirm level of importance' and/or 'did not reach consensus'. This indicator assesses the total population in the labor force both the employed and unemployed population, excluding the housewives and the jobless population who are not looking for work, such as stay-at-home moms, retirees, and students. This indicator assesses the quality of residential buildings by the percentage of new buildings (aged less than 30 years old) compared to the total residential buildings in the city. It is the opposite indicator of aging residential buildings. If new buildings are much less than old buildings, redevelopment or regeneration is needed in order to improve the housing quality, environment, and image of the city.

Category Indicator Explanation
Transport SCI23: Public transport sharing rate This indicator assesses public transport by the percentage of public transport means compared to the total transport means in the city. (Public transport helps reduce the uses of individual cars which is very significant in reducing traffic congestion and greenhouse gas.) This indicator assesses the insurance used by the residents in the city, including the services and quality. It measures the percentage of residents using insurance services compared to total population in the city. It also measures the sufficiency of the insurance companies providing services in the city.
Water Use SCI38: Percentage of households with access to potable water infrastructure This indicator assesses the potable water supply accessibility in the city. It measures the percentage of households with accessing to potable water supply infrastructure compared to total households in the city. (Potable water is considered in terms of both water security and sanitation.) SCI41: Average of the water consumption rate per person or household This indicator assesses the level of water consumption in the city. It measures the water consumption rate per person or household per day on average. (It is also to understand the ratio of water consumption compared to the total water supply in the city.) SCI40: Number of natural or artificial reservoirs in or nearby the city This indicator assesses the availability of fresh water sources in the city. It measures the number and size of natural or artificial reservoirs in or nearby the city. It is significant to understand the situation of water supply in the city, especially in the dry season (sufficiency or not).

Category Indicator Explanation
Waste Management SCI45: Percentage of solid waste regularly collected This indicator assesses the public organizations in place, especially the city government's efforts in collecting solid waste. It measures the percentage of households linked to the network that disposes of solid waste compared to the total household in the city.