A Systematic Review of Passive Cooling Methods in Hot and Humid Climates Using a Text Mining-Based Bibliometric Approach

Abstract

The number of scientific papers has dramatically increased in recent years; however, such a huge number of papers often become difficult to review qualitatively because of limited time and cost. A text-mining-based bibliometric analysis method is developed to provide a comprehensive overview of passive cooling methods used in buildings in hot and humid climates. First, a comprehensive overview analysis is conducted to clarify the trends in studies on passive cooling methods between 1970 and 2022, using 39,604 publications. Second, 15 coding rules are constructed to perform a detailed analysis, and possible combinations of passive cooling methods are discussed. The detailed analysis of the co-occurrence network based on the comprehensive overview and 15 coding rules shows that the recent studies on thermal energy storage have mainly focused on phase change material (PCM), which is a latent heat storage material, rather than sensible heat storage materials such as concrete. The analysis of the co-occurrence network indicates that natural ventilation maintains the vital role of passive cooling methods by improving thermal comfort in hot climates. The constructed method and possible combinations of passive cooling methods for hot and humid climates will help engineers find effective combinations in the planning stage.

1. Introduction

Currently, hot environments caused by climate change and heat island effects have become crucial in various climatic regions, such as temperate and hot climates. According to the Intergovernmental Panel on Climate Change (IPCC), there is a high possibility that the global rise in temperature between 2021 and 2040 will exceed 1.5 °C, even if the level of greenhouse gas emissions in 2019 is reduced by 50% by 2030 [1]. Because of such deterioration of the thermal environment, the International Energy Agency (IEA) [2] reported that space cooling accounted for nearly 16% of the global final electricity consumption in the building sector in 2020 (approximately 1885 TWh). According to the 2022 World Population Prospects [3], more than 50% of the projected increase in the global population is concentrated in tropical Africa and Asia, where space cooling is frequently demanded throughout the year. Accordingly, the global electricity demand for space cooling is expected to more than triple by 2050 [4]. Passive cooling methods utilize natural resources by means of architectural elements, thereby reducing cooling loads in buildings [5,6]. The determination of effective passive cooling methods for hot and humid climates can significantly contribute to reducing the energy consumption of space cooling while maintaining thermal comfort.

Table 1 presents the summary of review papers on passive cooling methods for buildings. Several studies have comprehensively reviewed individual passive cooling methods under different climatic conditions, including hot and humid climates [6,7,8]. Moreover, researchers have published various review papers on passive cooling methods that can adapt to hot and humid conditions, such as building envelopes [9,10,11,12], shading devices [10,11,12,13,14,15], radiant cooling [16,17], evaporative cooling [17,18,19,20], and natural ventilation [17,20,21,22,23], and each passive cooling method was systematically organized into subcategories based on its characteristics and principles (Table 1). For instance, regarding natural ventilation, previous studies have widely investigated both methods and principles of comfort ventilation [23], night ventilation [24], wind-driven ventilation [25], buoyancy ventilation [26,27], opening systems that assist in improving ventilation performance [23,28], and wind catchers [29,30]. A tremendous number of studies on passive cooling methods have been published in both reviews and original papers. Santamouris and Kolokotsa concluded that passive cooling methods are reaching a phase of maturity [20]. However, few comprehensive and systematic reviews of passive cooling methods focused on hot and humid climates, although literature reviews of each passive cooling method alone were frequently conducted. In hot and humid climates, such as Southeast Asia, it is difficult to achieve the thermal comfort of occupants using a single passive cooling method alone because the daily maximum air temperature and relative humidity usually exceed 30 °C and 70%, respectively [31,32]. Previous studies recommended the use of a combination of several passive cooling methods to improve the cooling effect [16,33,34,35]. Therefore, it is particularly important to find an optimum combination of passive cooling methods to achieve thermal comfort in hot and humid climatic regions.

Comprehensive reviews are one of the methods used to find an optimum combination of passive cooling methods. Passive cooling methods are often constructed based on traditional knowledge and experience [36]. Thus, it is important to analyze long-term references and their transitions until recent years. In general, passive cooling methods aim to (1) prevent heat gain, (2) modulate heat, and (3) dissipate internal heat [37]. Based on this principle, passive cooling methods have been classified, and researchers have often examined the optimum combination of passive cooling methods based on this classification [6,37,38,39]. Historically, such comprehensive reviews have also been conducted using qualitative approaches. In 1992, as a pioneering attempt, Antinucci et al. [40] reviewed 119 references on heat dissipation and protection design manually. Tejero-González et al. [8] conducted a literature review of 124 studies in 2016. In 2017, Panchabikesan et al. [7] investigated the potential of passive cooling methods by reviewing 127 documents. Bhamare et al. [6] reviewed 255 studies to provide an overview of passive cooling methods. However, Miranda et al. [41] indicated that the limitations of the previous qualitative literature reviews were time and cost, and a potential bias in selecting publications cannot be undeniable. In general, the number of publications has dramatically increased in recent years; however, the number of reviewed qualitative literature papers (i.e., the sample size) cannot dramatically increase because of the limited time and cost. A qualitative literature review based on sophisticated researcher knowledge and experience is undoubtedly valuable (Table 1). Nevertheless, a quantitative and systematic literature review method is required to capture comprehensive overviews of the research trends in response to the recent rapid increase in publications.

Bibliometric analysis using text mining methods can be an effective approach for conducting quantitative and systematic literature reviews in a wide range of research fields. Nie and Sun [42] used a text mining method to identify trends in research design, such as product and information designs, over the last 12 years. They concluded that text mining can help researchers obtain a comprehensive understanding of the knowledge in a certain field that is hidden in many studies [42]. Donthu et al. [43] handled 5344 documents published in the Journal of Business Research between 1973 and 2017 to analyze prominent topics and prolific authors [43]. Therefore, a bibliometric analysis using text mining is suitable for providing an overview of wide- and long-term information without certain bias. As text mining can organize a considerable amount of text information, it has helped doctors to understand medical research comprehensively [44], executives to decide on international strategic management [45], and marketers to gain insight into market advantages regarding consumer behavior [46]. Nevertheless, few studies in the building field have been conducted on bibliometric analyses using text mining [47], and an effective process for bibliometric analysis that can be applied to passive cooling methods has yet to be established. Previous bibliometric analyses categorized publications based on the author names, journal names, and existing categories in search engines, such as medical science, business science, and building technology [41,48]. Vanhala et al. [46] used single terms in a text mining-based bibliometric analysis to clarify latent topics in business research. However, a single term cannot express passive cooling methods, such as ventilation, and most methods consist of multiple terms, such as comfort ventilation and night ventilation. It is necessary to determine the type of bibliometric analysis that has a combination of multiple terms that is effective in passive cooling fields.

This study aims to provide a comprehensive overview of passive cooling methods for buildings that adapt to hot and humid climates through the text-mining-based bibliometric analysis. First, the bibliometric analysis method using text mining was constructed to quantitatively handle the considerable number of publications on passive cooling methods for buildings. The novelty of the constructed method includes the following: (1) the bibliometric analysis is conducted using the compound words based on coding rules; and (2) the selection bias is minimized through the quantitative analysis. This comprehensive review analyzes the overall trends in passive cooling methods that have been investigated since the 1970s. Secondly, coding rules consisting of compound words were built based on the quantitative overview analysis to find possible combinations of passive cooling methods for hot and humid climates. The constructed bibliometric analysis method would be an effective review method for drawing a comprehensive overview and trends in passive cooling methods in response to the recent increase in publications in the building field. Moreover, this method will help to find potential combinations of passive cooling methods for hot and humid climates based on research trends and the relationships between words in previous publications, which are worth investigating in the future.

In the first part of this paper, the details of the constructed bibliometric analysis method using text mining are described, followed by an introduction of feasible indices, such as the Jaccard coefficient (C_Jac) and collocation score (SC_Col), to evaluate the publications. Second, a comprehensive overview of passive cooling methods is conducted using 39,604 documents. Based on the comprehensive overview, 15 coding rules for passive cooling methods are introduced. Possible combinations of passive cooling methods for hot and humid climates are also discussed in the detailed analysis. The limitations of the proposed method are discussed in the following section.

Table 1. Summary of review papers on passive cooling methods for buildings.

Author	Sample Size	Passive Cooling Methods	Description
Sadineni et al. (2011) [9]	89	Building envelope (wall, fenestrations, roof), thermal mass, and PCM	Various building envelope components were reviewed concerning energy efficiency, and the cost benefits of envelope component technology were clarified.
Mirrahimi et al. (2016) [10]	118	Building envelope (wall, roof, glazing), shading device, and building form	The effects of building envelopes on energy savings and thermal comfort for high-rise buildings in the hot-humid climate, particularly in Malaysia, were clarified.
Bachrun et al. (2019) [11]	92	Building envelope (window, glazing, wall, roof) and shading device	The principles of passive design in the building envelope significantly influence the thermal comfort level of buildings.
Azmi and Ibrahim (2020) [12]	88	Building envelope (wall, roof, window) and shading device	The influence of building envelope design and parameters on the thermal performance of mosque buildings were discussed.
Yusoff et al. (2022) [13]	84	Shading device	The combined effects of the shading device on reducing solar radiation and enhancing natural ventilation are essential to achieving optimal performance.
Al-Masrani et al. (2018) [14]	167	Shading device	Egg-crate device is the best device to improve daylight and thermal performance in the tropics.
Zulkarnain et al. (2021) [15]	29	Shading device	The thermal-daylighting balance of the shading device was investigated, and adjustable shading control can achieve the optimal balance.
Suhendri et al. (2020) [16]	121	Radiant cooling	Most studies on radiant cooling focused on reducing cooling energy. Radiant cooling may lead to healthy and comfortable buildings. The investigation of the combination of radiant cooling with other passive cooling methods remains outstanding.
Samuel et al. (2013) [17]	112	Radiant cooling, ventilative cooling, evaporative cooling, thermal mass, and PCM	Radiant cooling provides better thermal comfort because of the direct treatment of radiation loads, low draft, and low vertical temperature gradient.
Tejero-González et al. (2021) [18]	100	Evaporative cooling	A critical view of the optimal operating conditions for wetted-surface evaporation coolers was conducted.
Cuce and Riffat (2016) [19]	48	Evaporative cooling	Evaporative cooling has great potential to achieve energy savings in hot and arid climates and is very cost-effective compared with alternative air-conditioning applications.
Santamouris and Kolokotsa (2013) [20]	237	Evaporative cooling and natural ventilation	Expected energy savings using heat dissipation techniques such as evaporative cooling and ventilation may reach 70% compared with a conventional air conditioning building.
Aflaki et al. (2015) [21]	70	Natural ventilation	Ventilation shafts, window-to-wall ratio, window-to-floor ratio, building position, and building orientation are the most important elements to provide effective natural ventilation.
Ahmed et al. (2021) [22]	123	Natural ventilation	Although high ventilation rates are effective in improving thermal comfort, they may reduce heatwave resilience and negatively affect indoor air quality due to outdoor air pollution.
Izadyar et al. (2020) [23]	253	Natural ventilation	A literature review on the impacts of building façades on natural ventilation showed that the application of balconies has not been adequately investigated yet.
Bhamare et al. (2019) [6]	255	Microclimate, solar control, thermal mass (PCM), ventilative cooling, evaporative cooling, and radiant cooling	A comprehensive review classified passive cooling methods, and their influence on temperature reduction and energy savings in several climates was further analyzed.
Panchabikesan et al. (2017) [7]	127	Evaporative cooling, radiant cooling, and PCM-based free cooling	The cooling potentials (W/m2) of passive cooling methods were estimated for five selected Indian cities.
Tejero-González et al. (2016) [8]	124	Solar control, natural ventilation, and evaporative cooling	Climate parameters affecting the applicability of passive cooling methods were compiled.
Antinucci et al. (1992) [40]	119	Microclimate, sharing, solar control, thermal mass, ventilative, evaporative, and radiant cooling	The state of the art on passive cooling in 1992 was presented, and the potential and limitations of passive cooling were defined.

Table 1. Summary of review papers on passive cooling methods for buildings.

Author	Sample Size	Passive Cooling Methods	Description
Sadineni et al. (2011) [9]	89	Building envelope (wall, fenestrations, roof), thermal mass, and PCM	Various building envelope components were reviewed concerning energy efficiency, and the cost benefits of envelope component technology were clarified.
Mirrahimi et al. (2016) [10]	118	Building envelope (wall, roof, glazing), shading device, and building form	The effects of building envelopes on energy savings and thermal comfort for high-rise buildings in the hot-humid climate, particularly in Malaysia, were clarified.
Bachrun et al. (2019) [11]	92	Building envelope (window, glazing, wall, roof) and shading device	The principles of passive design in the building envelope significantly influence the thermal comfort level of buildings.
Azmi and Ibrahim (2020) [12]	88	Building envelope (wall, roof, window) and shading device	The influence of building envelope design and parameters on the thermal performance of mosque buildings were discussed.
Yusoff et al. (2022) [13]	84	Shading device	The combined effects of the shading device on reducing solar radiation and enhancing natural ventilation are essential to achieving optimal performance.
Al-Masrani et al. (2018) [14]	167	Shading device	Egg-crate device is the best device to improve daylight and thermal performance in the tropics.
Zulkarnain et al. (2021) [15]	29	Shading device	The thermal-daylighting balance of the shading device was investigated, and adjustable shading control can achieve the optimal balance.
Suhendri et al. (2020) [16]	121	Radiant cooling	Most studies on radiant cooling focused on reducing cooling energy. Radiant cooling may lead to healthy and comfortable buildings. The investigation of the combination of radiant cooling with other passive cooling methods remains outstanding.
Samuel et al. (2013) [17]	112	Radiant cooling, ventilative cooling, evaporative cooling, thermal mass, and PCM	Radiant cooling provides better thermal comfort because of the direct treatment of radiation loads, low draft, and low vertical temperature gradient.
Tejero-González et al. (2021) [18]	100	Evaporative cooling	A critical view of the optimal operating conditions for wetted-surface evaporation coolers was conducted.
Cuce and Riffat (2016) [19]	48	Evaporative cooling	Evaporative cooling has great potential to achieve energy savings in hot and arid climates and is very cost-effective compared with alternative air-conditioning applications.
Santamouris and Kolokotsa (2013) [20]	237	Evaporative cooling and natural ventilation	Expected energy savings using heat dissipation techniques such as evaporative cooling and ventilation may reach 70% compared with a conventional air conditioning building.
Aflaki et al. (2015) [21]	70	Natural ventilation	Ventilation shafts, window-to-wall ratio, window-to-floor ratio, building position, and building orientation are the most important elements to provide effective natural ventilation.
Ahmed et al. (2021) [22]	123	Natural ventilation	Although high ventilation rates are effective in improving thermal comfort, they may reduce heatwave resilience and negatively affect indoor air quality due to outdoor air pollution.
Izadyar et al. (2020) [23]	253	Natural ventilation	A literature review on the impacts of building façades on natural ventilation showed that the application of balconies has not been adequately investigated yet.
Bhamare et al. (2019) [6]	255	Microclimate, solar control, thermal mass (PCM), ventilative cooling, evaporative cooling, and radiant cooling	A comprehensive review classified passive cooling methods, and their influence on temperature reduction and energy savings in several climates was further analyzed.
Panchabikesan et al. (2017) [7]	127	Evaporative cooling, radiant cooling, and PCM-based free cooling	The cooling potentials (W/m2) of passive cooling methods were estimated for five selected Indian cities.
Tejero-González et al. (2016) [8]	124	Solar control, natural ventilation, and evaporative cooling	Climate parameters affecting the applicability of passive cooling methods were compiled.
Antinucci et al. (1992) [40]	119	Microclimate, sharing, solar control, thermal mass, ventilative, evaporative, and radiant cooling	The state of the art on passive cooling in 1992 was presented, and the potential and limitations of passive cooling were defined.

2. Methodology

2.1. Outline

Figure 1 presents the research flow for sampling and screening the publications, which is the first step in the bibliometric analysis. First, the Web of Science (WoS) and Scopus engines were used to find the target publications. These citation indexing platforms are regarded as the largest databases of academic journals and conference proceedings in various academic disciplines. Martín-Martín et al. [49,50] compared Google Scholar, Web of Science, and Scopus to determine the influence of publication coverage on the bibliometric analysis. Although Google Scholar had 1.5–1.8 times more publications than WoS and Scopus in the engineering and computer science fields, more than 60% of the unique publications in Google Scholar comprise dissertations, book chapters, and informally published papers [49,50,51]. Therefore, Scopus and WoS were selected as suitable search engines for obtaining reliable publications. The publications sampled from the two engines were screened and organized using EndNote 20 (Clarivate, Philadelphia, PA, USA), a commercial reference management software. Two researchers reviewed the sampled publications. After screening the publications, KH Coder 3 (Higuchi, Kyoto, Japan), an open-source software for bibliometric analysis, was used for text mining the titles and abstracts of the screened papers and generating the co-occurrence networks and multidimensional scaling map. KH Coder uses the Stanford part-of-speech (PoS) tagger, which has a high accuracy for tagging to analyze English text data and is commonly used in previous studies [52]. To provide a comprehensive overview of passive cooling methods, the titles of the publications were used as the target because using a large number of words would make the analysis difficult if the text mining included the entire article. Moreover, the titles often represent an overview of the publication. To find potential combinations of passive cooling methods in detail, the abstracts of the sampled and screened papers were subjected to text mining.

Figure 2 shows the constructed bibliometric analysis method, which corresponds to the analysis of the titles and abstracts shown in Figure 1. First, text mining was performed on the titles of the sampled publications to provide a comprehensive overview of passive cooling methods. Based on the co-occurrence network and multidimensional scaling analyses, the relationships between the words were obtained, and coding rules were constructed to set the target compound words for the text mining of the abstracts of the sampled publications. Using coding rules, the text mining of the abstracts of the sampled publications (detailed analysis is shown in Figure 2) was performed to discuss the possible combinations of passive cooling methods.

2.2. Sampling and Screening Papers

The papers were sampled and screened based on the systematic review of the guidelines proposed by Pullin and Stewart [53] (Figure 1). Three categories of words (listed in

Table 2. Keywords of the search formula for the sampling publications and publications retrieved from each search engine.

Category	Keywords of the Search Formula (Field Topic)	Publications Retrieved from WoS	Publications Retrieved from Scopus
#1 Architectural object	apartment* OR architect* OR buil* OR dwelling* OR home* OR hous* OR indoor* OR office* OR residen* OR room*	4,414,331	6,567,135
#2 Technology and method	air OR bioclimat* OR climat* OR cool* OR evaporat* OR green OR heat* OR natural* OR passive* OR PCM OR radia* OR shad* OR simulation OR solar OR stor* OR sustainab* OR system* OR technolog* OR thermal OR tradition* OR ventilat* OR vernacular	21,783,521	33,447,968
#3 Climate and region	“hot and humid” OR “hot climat*” OR “hot dry” OR “hot humid” OR “humid climat*” OR “South East Asia” OR “Southeast Asia” OR “Sub tropic” OR “Subtropical climate” OR area* OR subtropics* OR summer OR thermal OR tropic*	5,479,498	7,979,247
Search formula used in the search engines	#1 AND #2 AND #3	485,207	509,962

*: Unlimited flection, “ ”: Phrase search.

), which included possible contextual search words, were set using the Boolean search formulas OR and AND to avoid missing relevant papers. The words commonly used in the previous comprehensive literature reviews on passive cooling methods were used to create the categories. Categories 1, 2, and 3 represent the architectural objects, technologies, methods, climates, and regions, respectively. Miranda et al. [41] reported that creating categories consisting of a search formula is effective for properly collecting publications in the target research field. The search formula consisted of terms from each category to screen the publications on passive cooling methods for hot and humid climates. In the Scopus engine, the maximum number of exports was up to 20,000 publications in one search result. Thus, the papers were exported separately for each year. Using the abovementioned methods, N = 485,207 and N = 509,962 publications from 1970–2022 (until March) were exported from the Web of Science and Scopus, respectively. In addition to the publications from Web of Science and Scopus, publications referring to papers written in the local language and excluded from the two search engines were manually added to the analysis (n = 599) by three reviewers. Although the search formula for passive cooling methods was developed, ineligible research fields such as foods and animals were included in the output publications. The exported papers were screened properly for duplication and the relevance of journal names based on the preferred reporting items for systematic reviews and meta-analyses (PRISMA 2020) guidelines [54,55] (see Supplementary Materials). A total of 936,350 publications were removed, and 59,418 publications (n = 34,008 from WoS and n = 24,846 from Scopus) were combined for EndNote. In EndNote, the title, abstract, and keywords were checked, and 11,664 publications were removed due to a lack of relevancy. Overlapping documents between the Web of Science and Scopus were also eliminated (n = 8150) to avoid redundancy. Finally, 39,604 publications were selected as target publications for the bibliometric analysis.

2.3. Quantitative Evaluation of the Papers

To understand the comprehensive overview of passive cooling methods, the top 200 most frequent nouns, proper nouns, adjectives, and verbs were extracted from the sampled publications using the text mining method (Figure 2). Based on these words, their relationships were quantitatively evaluated using the Jaccard coefficient in the co-occurrence network analysis. The Jaccard coefficient (C_Jac) was calculated as follows (Equation (1)).

$$C_{Jac}=\frac{A\cap B}{A\cup B}=\frac{A\cap B}{A+B-A\cap B}$$

(1)

A high C_Jac value indicates that the connected words appear frequently in the same document. In addition to the co-occurrence network analysis, a multidimensional scaling analysis using Kruskal’s algorithm [56,57] was applied to provide the comprehensive overview of the passive cooling methods. The following relationship (Equation (2)) between the similarity of the words (δ) and the distance between the words (d) must be satisfied by Kruskal’s algorithm to be able to determine the position of words in the multidimensional space:

$${\delta }_{ij}>{\delta }_{rs}\to d_{ij}\le d_{rs}$$

(2)

Although the position of words can be determined based on the monotonic-decreasing relationship between the similarity of the words and the distance between them, the position of words does not always satisfy the monotonic-decreasing relationship due to the relationship between other words. In the multidimensional scaling map, incompatibility stress (S) is used to evaluate the satisfaction with a monotonic relationship. In this study, 1000 iterations were performed using Kruskal’s algorithm to minimize the incompatibility stress (S) when determining the position of words, which is expressed by Equation (3):

$$S=\sqrt{\frac{\sum _{i=1}^{n-1}\sum _{j=2}^n{(d_{ij}-\widehat{d_{ij}})}^2}{\sum _{i=1}^{n-1}\sum _{j=2}^n{d}_{ij}^2}}\hspace{1em}(i<j).$$

(3)

Based on the results of the co-occurrence network and multidimensional scaling analyses, a coding rule was constructed to find possible combinations of the passive cooling methods. The constructed coding rule consisted of two or more words (such as natural ventilation and night ventilation) because it was difficult to represent passive cooling methods using a single term (such as ventilation). The collocation score (SC_Col) proposed by Nakagawa et al. [58] was used to determine the compound words concerning passive cooling methods, which were frequently used. The collocation score (SC_Col) was calculated using the frequency of the target compound word (f(cw)) and the frequency of the concatenated words before (fl) and after (fr) each word (w_i) of the target compound word (cw) (Equation (4)).

$$\begin{gathered} {SC}_{Col}=f(cw)\times {(\prod _{i=1}^k\left(fl\right(w_i)+1)(fr\left(w_i\right)+1))}^{\frac{1}{2k}} \\ cw=w_1, w_2, \cdots , w_k \end{gathered}$$

(4)

Using 15 coding rules (see Section 3.2) that had high collocation scores (SC_Col) and were based on the words in the co-occurrence network and multidimensional scaling map, text mining was applied to the titles and abstracts of the publications. Each coding rule consisted of several compound words.

In the following analysis, the co-occurrence network was expressed using the Jaccard coefficient (C_Jac) and the centrality of the target compound words. Betweenness centrality (BC(n)) illustrates the extent to which a target node (n) is between others and is expressed as follows [59]:

$$BC\left(n\right)={\sum }_{s=1}^N{\sum }_{g=1}^{s-1}\frac{P_{n(s,g)}}{P_{(s,g)}}$$

(5)

where N is the total number of nodes, P_(s,g) is the shortest path from s to g, and P_n(s,g) is the shortest path from s to g via n. The 75 words with the highest C_Jac values were extracted from the top of the target compound words. In previous studies, 75 words were determined for the co-occurrence network to identify clear co-occurrence relationships [60,61]. Co-occurrence networks, including the target compound word, were presented based on the co-occurrence relationships between the 75 words and BC(n).

3. Results

3.1. Comprehensive Overview of the Passive Cooling Methods

Figure 3 shows the co-occurrence network with the C_Jac and multidimensional scaling map for words appearing in the titles of papers for the comprehensive overview of passive cooling methods. The words with a C_Jac of more than 0.08 appeared in the co-occurrence network. Building was the most commonly used, with a frequency of 10,389 in 39,604 references, followed by thermal, with a frequency of 9352. The C_Jac between thermal and comfort was 0.23. The high frequency of thermal and C_Jac of thermal comfort imply that studies on thermal comfort have been frequently conducted. Phase_change had the highest C_Jac of 0.59 among the top 200 words, and both words were strongly connected with material, with the C_Jac of 0.31 and 0.37 for change_material and phase_material, respectively. Phase change material (PCM) had a relationship between heat_storage and thermal_storage, indicating that PCM have been frequently studied owing to their thermal storage effect. In the multidimensional map, PCM and phase change material were placed near envelope, wall, and roof. Natural_ventilation (C_Jac = 0.23), wall_insulation (C_Jac = 0.09), and green_roof (C_Jac = 0.17) were the passive cooling strategies that appeared in the co-occurrence network. In particular, ventilation was located near comfort, hot, and climate in the multidimensional scaling map. Natural ventilation can be regarded as a passive cooling strategy expected to improve thermal comfort in hot climates. Occupant_behavior appeared in the co-occurrence network for passive cooling methods with a C_Jac of 0.10. Because window, ventilate, and room were positioned near behavior, previous studies may have focused mainly on the window-opening behavior in ventilated rooms. The co-occurrence network and multidimensional scaling analyses can be used to clarify the commonly studied passive cooling strategies and analyze the purposes and applications of previous studies.

3.2. Constructing Coding Rules

Using the titles of the 39,604 references, the collocation scores (SC_Col) for 68,127 compound words were calculated. Based on the words in the co-occurrence network and multidimensional scaling maps, the coding rules regarded as having a relatively high SC_Col were constructed (

Table 3. Coding rules of passive cooling methods and their collocation scores (SC_Col).

* cool roof cool + roof (1898) OR cool + roofs (2490)

* evaporative cooling evaporative + cooling (12,489)

* green roof green + roof (16,198) OR green + roofs (12,071)

* green wall green + wall (900) OR green + walls (4459) OR greenwall

* natural ventilation natural + ventilation (152,725)

* night ventilation night + ventilation (5831)

* occupant behavior occupant + behavior (4108)

* phase change material phase + change + material (99,009) OR phase + change + materials (98,214) OR pcm

* radiant cooling radiant + cooling (5153) OR radiative cooling (1866)

* solar shading solar + shading (3097)

* thermal behavior thermal + behavior (92,822) OR thermal + behaviour (61,084)

* thermal comfort thermal + comfort (2,676,805)

* thermal insulation thermal + insulation (221,109)

* thermal storage thermal + storage (149,724) OR heat + storage (14,497) OR thermal + energy + storage (768,985)

* tropical climate tropical + climate (28,031) OR hot + humid (4866)

*: Classification, +: Collocation.

). Thermal comfort had the highest SC_Col of 2,676,805 among the 68,127 compound words because the frequencies of thermal and comfort were 9352 and 2737, respectively. The effects of passive cooling methods have frequently been evaluated in terms of thermal comfort [20]. Thus, the SC_Col of thermal comfort was high. Regarding passive cooling methods, *thermal energy storage (* indicates category), including thermal storage (SC_Col = 149,724), heat storage (SC_Col = 14,497), and thermal energy storage (SC_Col = 768,985), had the fifth highest SC_Col in the high-coding category. In the co-occurrence network, the word storage was connected to phase_change_material, which has relatively high latent heat storage. The collocation scores of phase change material and phase change materials were 99,009 and 98,214, respectively. Natural ventilation (SC_Col = 152,725) and green roof (SC_Col = 16,198), which appeared in the co-occurrence network due to the high C_Jac, also appeared in the top 200 compound words. If words appear in the co-occurrence network due to a high C_Jac, the SC_Col tends to be high. The high C_Jac can be a criterion for making a suitable coding rule; thus, a C_Jac of more than 0.08 was used in this study.

*thermal insulation, *green wall, *cool roof, and *night ventilation were added to the coding rules as the derivative words for the above-mentioned compound words. Although wall and insulation were connected to each other with a C_Jac of 0.08, thermal insulation, whose SC_Col was the twelfth highest (SC_Col = 221,190) among the 68,127 compound words, was a more frequently used compound word. Therefore, thermal insulation was added to the coding rule instead of wall insulation. Green roof was shown in the co-occurrence network, and roof was positioned near wall in the multidimensional scaling map; thus, the green wall was selected as a coding rule for further analysis. Among the 15 coding rules, the cool roof had a relatively low SC_Col, with a maximum SC_Col of 2490 for *cool roof. Nevertheless, it was in the top 2.2% of the compound words. Therefore, cool roofs can be regarded as representative passive cooling methods.

*evaporative cooling and *radiant cooling did not appear in both the co-occurrence network and multidimensional scaling maps. Both passive cooling methods considered materials, structures, and installation positions [16,17,18,19,20,62,63,64]; thus, it is difficult to connect specific words in the co-occurrence network. Interestingly, radiant and evaporative were among the 200 most frequently used words. Nevertheless, evaporative cooling and radiant cooling had relatively high SC_Col values of 12,489 and 5153, respectively. As discussed above, it is difficult for a single term to represent passive cooling methods. Therefore, evaporative and radiant were employed as compound terms using two or more words rather than a single term because neither word appeared in the co-occurrence network or multidimensional scaling maps, despite the relatively high SC_Col. Therefore, *evaporative cooling and *radiative cooling were added to the coding rules. To construct effective coding rules based on the quantitative literature review that can represent passive cooling methods, referring the collocation score (SC_Col) is recommended as well as the Jaccard coefficient (C_Jac) in the co-occurrence network and the position of words in the multidimensional scaling map. The 15 coding rules listed in Table 3 were applied for the detailed analysis.

3.3. Trends in Passive Cooling Methods

In the following sections, the list of the publications used to create the co-occurrence network is referred in the discussion of the features of the respective passive cooling applications. Figure 4a shows the number of publications each year from 1970 to 2021 using keywords from the search formula. Here, the trend in the sampled publications was compared with the number of publications registered in the “Construction & Building technology” category, which includes heating and air conditioning, energy systems, and indoor air quality research as the referential data in the Web of Science (WoS) database. The number of referential publications on Construction & Building technology in the WoS increased by 624% between 1990 and 2020. The ratio of the number of sampled publications to that of the referential publications ranged from 2.8–7.3% from 1991–2000 and increased to 8.1–13.8% and 8.2–16.9% from 2001–2010 and 2011–2020, respectively. Cañas-Guerrero et al. [48] found a growth in words related to the indoor environment and energy efficiency, such as simulations, temperature, energy, environment, and thermal comfort, from 2006–2011. This tendency can affect the recent increase in the number of publications on passive cooling methods, which improve the indoor thermal environment and reduce building energy consumption.

Figure 4b presents the number of publications between 1970 and 2021 containing the compound words in their titles based on the 15 coding rules. The publications obtained under the 15 coding rules accounted for 1.5–9.8% of the sampled publications between 1970 and 1999. This rises to 11.2–17.2% from 2000–2020, mainly owing to the rapid increase in the publications related to phase change material (PCM). The number of publications on PCM increased by 8150% between 2002 and 2021. Owing to its light weight and high energy density, PCM has been installed on the walls [65,66,67], floors [68,69], roofs [70], and ceilings [71] of buildings in hot and humid climate regions where the diurnal temperature range is less than 10 °C. Souayfane et al. [72] found that the required PCM ambient temperature range to employ its cooling effect is less than 5 °C. Previous studies have investigated manufacturing methods for PCMs, such as immersion, powder, microencapsulation, and macroencapsulation [73]. The manufacturing methods and installation positions of PCMs in buildings have led to various specifications. The wide applicability of PCMs in buildings may have contributed to their increasing occurrence in recent studies.

Among the 15 coding rules, studies on thermal comfort were most frequently conducted between 1990 and 2022 (March). The publications on thermal comfort accounted for 3.0–5.9% of the sampled publications. The purpose of the previous studies on thermal comfort can be divided into two large categories: developing the thermal comfort ranges of the occupants using thermal comfort indices, such as the operative temperature (OT), predicted mean vote (PMV), and standard effective temperature (SET*) [74,75,76]; and evaluating the cooling effect of each method and technique in terms of the indices [31,32,33,34,35]. The broad scope of the research on thermal comfort has led to an increase in the publications.

Natural ventilation is one of the bases of passive cooling. From 1995 to 2009, the number of papers with natural ventilation in their titles was the second largest among those analyzed under the 15 coding rules. As the number of publications on natural ventilation has steadily increased since the 1980s, natural ventilation can be regarded as a common passive cooling method. Previous studies have shown that natural ventilation can be combined with building elements such as windows, wind catchers, balconies, and wing walls, as well as other cooling methods such as evaporative and radiant cooling [21,22,23].

Figure 5 depicts the co-occurrence network that connects each decade to clarify the trend in the research on passive cooling methods, considering all the sampled publications (N = 39,604). Construction, materials, and structures, which are regarded as the fundamental elements of architecture, were discovered in the 1970s and 1980s. Meanwhile, the term material was also obtained in the 2000s and 2020s. Cañas-Guerrero et al. [48] reported that materials and structures such as cement, steel, composites, and fibers stood out and continued to be important subjects because they were related to the energy, thermal comfort, and ventilation of buildings. Tabatabaei and Fayaz [77] investigated the effect of 20 façade materials on urban heat island and indicated that the appropriate choice of material and color of urban vertical surfaces can effectively reduce urban heat island. Recycled materials have recently drawn attention toward sustainable development. For instance, asphalt concrete using waste glass as aggregates absorbed less heat during the day and emitted less heat at night than that with limestone aggregates, thus mitigating the urban heat island effect [78]. Moreover, the number of studies on PCM has rapidly increased since the 2000s. Consequently, the PCM was found in the 2010s and 2020s. Materials may play an important role in passive cooling methods throughout the entire period. Wind, natural ventilation, ventilate, and ventilation, which are associated with ventilation, were linked in the 1970s and 1990s–2020s. The number of publications with natural ventilation in their titles accounted for 4.4% of the publications obtained in the 2020s under the 15 coding rules (Table 4). Nevertheless, ventilation is still frequently investigated.

Regarding the tools used for evaluating passive cooling methods, the word numerical was obtained in the 2000s and 2010s in the co-occurrence network. Computational fluid dynamics (CFD) was first introduced in the ventilation industry in the 1970s [79]. In 1980, a numerical calculation of the flow related to ventilation over an aerofoil took half an hour at a cost of 1000 USD [80]. Nowadays, the calculation time and cost are minimal. Regarding building energy simulation programs, DOE-2, BLAST, and ESP-r were found in the 1970s [81]. EnergyPlus, which combines the advantages of BLAST and DOE-2, was developed in 2001 [82]. Strachan et al. [83] reported that the errors in the thermal load calculated by simulation programs, including ESP-r, were approximately 25% of the daily values in the 1970s. Approximately 30 years later, Fumo et al. [84] reported that the error in calculating the energy consumption was mostly less than 10% using the EnergyPlus benchmark models. Owing to the improvement and dissemination of computers, many researchers can easily conduct numerical simulations. Thus, the number of studies using numerical simulations has increased rapidly since the 2000s. Optimization has been used since the 2010s. Gassar et al. [85] conducted a comprehensive literature review of optimization studies for use in the building field. They excluded studies published before 2000 because the amount of literature on building optimization was not significant [86]. Previously, optimization was performed based on only one objective, which was used to determine the optimal solution for maximizing or minimizing the objective function [87]. However, multi-objective optimization problems need to be solved using real-world designs [88]. The advantage of numerical simulation is the investigation of several parameters under limited time and cost, owing to its improved computing capacity. For these reasons, multi-objective research, such as evaluating the trade-off relationship between thermal and daylighting environments for shading devices and windows [88,89], was frequently conducted in the 2010s and the 2020s. The prevalence of reliable numerical simulations with reduced time and cost could be the cause for the increase in the number of sampled publications since the 2000s.

Overall, the text mining method for the titles of papers and co-occurrence networks can be effective in clarifying the trends in the studies on passive cooling methods. In particular, the 15 coding rules can be used as representative research topics on passive cooling methods because the co-occurrence network for each decade considered all the sampled publications (N = 39,604). Most words that appeared in the co-occurrence network were related to the 15 coding rules because of their similarities. Although the number of sampled publications has dramatically increased since the 2000s, the coding rules can capture the recent rapid increase and clarify the research trends based on a large number of publications.

3.4. Co-Occurrence Network Analysis Based on the Coding Rules

Using the 15 coding rules, a co-occurrence network analysis, which refers to co-occurrence relationships, was conducted on the abstracts to find the purposes of the application as well as the strengths and weaknesses of each passive cooling method. In particular, the potential combination of passive cooling methods for hot and humid climates was uncovered, including the recent rapid increase in the number of publications. As the abstracts of the publications were the objects of text mining, unnecessary terms based on parts of speech were screened (Figure 2).

3.4.1. Evaporative Cooling, Green Roof, Green Wall, and Tropical Climate

Figure 6 depicts the co-occurrence network of evaporative cooling, green roofs, green walls, and tropical climate with a minimum-spanning tree that minimizes the length of the edges of the tree. A relatively high SC_Col and the occurrence of the word evaporative, which was connected to climate_hot_humid, proved that evaporative cooling can be regarded as a common passive cooling method in hot and humid climates. Previous studies have suggested that a relative humidity of less than 70% does not generate thermal discomfort and that evaporative cooling can be used in hot and humid climates [90,91]. Nevertheless, terms related to dehumidification that were linked to humid_hot_climate stood out in the co-occurrence network of evaporative cooling, and dehumidification appeared in tropical climates. Passive dehumidification systems that incorporate building envelopes such as roofs and walls were proposed in refs. [92,93]. Passive dehumidification and evaporative cooling systems can be a potential combination of passive cooling methods that can reduce the risk of thermal discomfort. For green roofs and walls, humidification of buildings through evapotranspiration may not be a problem, unlike evaporative cooling, because of the relationship between the effect and scale. Previous studies indicated that green roofs and walls exhibit less influence on transpiration under the local microclimate, such as at the building scale; however, they had a greater significance on a larger scale if the green coverage ratio was high [94,95]. Vegetation on green roofs and walls prevents the warming and overheating of the underlying surfaces due to incoming radiation and mitigates the radiation environment [96], which is associated with short-wave, thermo-radiative, and radiation in the co-occurrence network. Therefore, urban_island, which indicates the urban heat island effect, was connected to mitigation_cool (C_Jac = 0.39) in the co-occurrence network of green roofs. Green walls and roofs are suitable passive cooling methods in hot and humid regions because the exterior walls of residential buildings in tropical regions are generally composed of concrete without insulation [97].

3.4.2. Natural Ventilation, Night Ventilation, Thermal Comfort, and Occupant Behavior

Figure 7 shows the co-occurrence network of natural ventilation, night ventilation, thermal comfort, and occupant behavior with the minimum spanning tree. In the co-occurrence network of the tropical climate (Figure 6d), passive_natural_ventilation and comfort, which had relatively high centrality and frequency, were observed. Similarly, natural_ventilation was linked to thermal_comfort, with a C_Jac of 0.66 in the co-occurrence network of natural ventilation. No words related to energy consumption were found. Sakiyama et al. [98] reported that 43% of the reviewed papers on naturally ventilated buildings are in hot-humid, tropical, and subtropical climates and concluded that thermal comfort is a vital aspect of natural ventilation performance in their comprehensive literature review of natural ventilation. Natural ventilation can improve the thermal comfort of occupants through comfort ventilation and dissipate the heat from building structures. Therefore, natural ventilation can reduce the energy consumption for cooling. Nevertheless, previous studies on natural ventilation in hot and humid climates focused on improving thermal comfort rather than reducing the energy consumption of buildings. In general, the ratios of heat loss from the human body in indoor spaces are approximately 50% through radiation, 30% through convection, and the rest through evaporation [99,100,101]. Evaporative heat loss from the human body accounts for more than 50% of the total heat loss in warm environments with a relative humidity of approximately 60% [102,103]. Previous studies showed that the convective heat transfer coefficient of the human body, which affects the evaporative heat transfer coefficient as well, increases significantly when the airspeed exceeds 0.3 m/s [104,105]. Under hot and humid conditions in naturally ventilated buildings in Indonesia, the results of a questionnaire survey indicated that the occupants generally prefer high wind speeds [106]. When the operative temperature was 30 °C, the minimum preferred wind speed was 0.9 m/s in a hot-humid climate [107]. Passive cooling methods that employ natural ventilation under hot and humid conditions meet these criteria to improve thermal comfort. The combination of natural ventilation with other passive cooling methods, such as roof covers, ceiling insulation [34], and radiant cooling [35], increases thermal comfort. In particular, the use of horizontal pivot windows, which is a preferable window type for comfort ventilation and radiant cooling, increased the thermal comfort period up to 85% during the daytime in Indonesia [35]. Accordingly, improving thermal comfort contributes to a reduction in energy consumption [69]. The C_Jac of comfort_energy and energy_consumption were 0.58 and 0.43, respectively.

In the co-occurrence network of the occupant behavior, behavior_window_opening_closing, which is related to natural ventilation, had a higher centrality than behavior_pattern_ac_on/off, which is related to energy consumption. This supports the above-mentioned implication for the preference of thermal comfort in natural ventilation studies. Nevertheless, window-opening patterns in hot and humid climates are affected by the household size, age of the respondent, household income, and concerns about insects more than the thermal sensation of the occupants [108].

For night ventilation, overhung, blind, and PCM were observed in the co-occurrence network. Toe and Kubota [109] predicted the cooling effect of passive cooling methods for Malaysian terraced houses and recommended the combination of night ventilation with the roof and ceiling insulations and continuous low-roof eaves at approximately the window height level (650–750 mm depth) to increase the thermal comfort period. Kitagawa et al. [68,69] used PCM to increase the thermal storage of Indonesian apartments, which are often constructed from materials with a low thermal mass, thus improving the cooling effect through night ventilation in hot and humid climates. Ran and Tang [110] proposed the use of a combination of a green roof and night ventilation for controlling the indoor air temperature in hot and humid regions, which can reduce the temperatures by up to 2.3 °C compared with the building’s combined night ventilation and wall insulation. Although the cooling potential of night ventilation in hot and humid climates was considered low compared with that in other climatic regions [111], thermal comfort can be improved by combining night ventilation with other passive cooling methods.

3.4.3. Thermal Energy Storage, Phase Change Material, Thermal Behavior, and Radiant Cooling

Figure 8 shows the co-occurrence network of thermal energy storage, phase change material, thermal behavior, and radiant cooling with the minimum spanning tree. In the co-occurrence network of thermal energy storage, material_phase_change, and pcm, whose frequencies were 587–801 and 1277–5440 in the conditional and unconditional occurrences in the network, respectively, are shown. The conditional refers to only the co-occurrence, and the unconditional indicates all occurrences in the detailed analysis. Previous studies on thermal energy storage have mainly focused on PCM, which is a latent heat storage material, and there has been a rapid increase in research since the 2010s. Moreover, sensible heat storage materials such as concrete have not been frequently investigated, although thermal energy storage has been investigated since the 1970s. Devax et al. [112] reported that a latent heat thermal storage material, i.e., PCM, stored and released heat per unit volume 5–14 times as much as a sensible heat thermal storage material such as concrete for building applications. Therefore, PCM is a suitable material for thermal storage through night ventilation in hot and humid climate regions, where the diurnal temperature range tends to be small.

Material_phase_change and pcm appeared in the co-occurrence network of thermal behavior. Because of the complex phenomena in the phase changes of PCMs, previous studies have investigated the unique thermal behavior of PCM (paraffin) during phase-change processes. Liu et al. [113] found that the melting time of PCMs (paraffin) exposed to a fixed temperature was shorter than that of employing a fluctuating inflow temperature, although the average exposure temperatures for both cases were the same. Iten et al. [114] reported that the latent heat of PCMs (paraffin) was underestimated under a heating rate of 10 °C/min, compared to the slower heating rate of 0.2 °C/min in differential scanning calorimetry (DSC) testing. The differential_calorimetry, which is a common method for measuring the thermal properties of PCM, appeared in the co-occurrence networks of the thermal energy storage and phase change material. Although PCMs have been employed in radiant cooling systems for buildings in hot and humid climates, such as floors [68,69] and ceilings [115], by lowering the surface temperature of the building surface, PCM was not observed in the co-occurrence network in radiant cooling.

Condensation was frequently observed in the co-occurrence network, with a frequency of 36 conditional occurrences. When active radiant cooling systems employing chilled water with a supply temperature of 17–22 °C are applied to buildings in hot and humid climates, the condensation that forms on the panels due to the high humidity, which has a C_Jac of 0.35 in the co-occurrence network, can be a problem [116,117]. The annual average dew point temperature in the hot and humid climate of Jakarta, Indonesia, was 25.0 °C [118]; thus, the active radiant cooling system had a high chance of suffering from condensation. Meanwhile, the floor surface temperature was rarely lower than the dew point temperature when a passive radiant cooling system using PCMs was applied to a naturally ventilated building in the hot and humid climate of Indonesia [119].

3.4.4. Cool Roof, Thermal Insulation, and Solar Shading

Figure 9 illustrates the co-occurrence network of cool roofs, thermal insulation, and solar shading, which prevent heat gain with the minimum spanning tree. Cool roofs include green and high-solar-reflectance roofs [120]. Therefore, reflective_coating, green, and vegetation appeared in the co-occurrence network of cool roofs. In general, roofs contribute nearly 50–60% of the total cooling load of low-rise buildings in hot and humid climate regions [121]. Although the number of publications on cool roofs accounted for only 0.2% of the sampled publications in this study, cool roofs can be recommended as a passive cooling method in hot and humid climates. Cool roofs reduce the surface temperature during the daytime and heat gain from solar radiation [122]. Island_phenomena (C_Jac = 0.51) and island_mitigation (C_Jac = 0.45) were associated with the urban heat island effect, and saving_benefit (C_Jac = 0.27) and saving_annual (C_Jac = 0.32) implied that cool roofs influenced the annual energy savings. Previous studies [122,123] recommended that passive cooling methods can be applied not only to individual buildings but also to large scales, such as the city scale, to further increase the cooling effect because the surrounding environments strongly influence the effect of passive cooling methods.

Similar to cool roofs, deciduous and planting, which are associated with green walls, were observed in the co-occurrence network of solar shading because shading is often considered the most important factor for cooling using vegetation, where deciduous trees are effective for controlling solar shading during different seasons [93,124,125]. The co-occurrence network of solar shading indicates that the shading devices attached to the building façade, such as blind and green walls, have been investigated more than eaves and overhangs. Nevertheless, external shading devices (such as louvers, shutters, and blinds) installed in front of openings on building façades may not be recommended for naturally ventilated houses in hot and humid climates because the influence of natural ventilation is greater than that of shading on the thermal comfort of occupants, which leads to a wind blockage effect that reduces inflow due to the obstruction of airflow [126,127]. There is a trade-off between shading and ventilation effects, daylighting, and visibility. Gassar et al. [84] reported that the trade-off relationship between daylighting and the cooling load is often investigated in optimization studies. Movable_shade (C_Jac = 0.25) and Movable_slat_blind (C_Jac = 0.25, 0.40, respectively) can be solutions for the trade-off relationship in the optimization process. Thermal insulation, conductivity_material (C_Jac = 0.30), property_material (C_Jac = 0.33), and concrete, which are related to the material, have often been investigated. Although the temperature difference between indoor and outdoor environments is relatively small in hot and humid climates compared to other climatic regions, a cooling load occurs throughout the year. Roof and ceiling insulations are strongly recommended to decrease the heat_transfer (C_Jac = 0.35) from the roof to reduce the energy_consumption (C_Jac = 0.39) and cooling loads in buildings, even in hot and humid climates [34].

4. Discussion

4.1. Co-Occurrence Relationship and Word Frequency in the Co-Occurrence Network

To perform the detailed analysis, 75 words that had a high C_Jac with the target compound word were outputted from the text mining on the abstracts of the sampled publications, and the co-occurrence relationships between the words whose C_Jac was more than 0.2 are shown in the co-occurrence networks. Blanchet et al. [128] reported that ecological research on reliable co-occurrence networks requires as many samples as possible. However, as the word frequency increased, the number of co-occurrence relationships with a C_Jac of more than 0.2 decreased. Because high-frequency words such as building and energy appeared in the co-occurrence network, it was difficult to find the characteristics of each passive cooling method and a concrete combination of passive cooling methods. Romesburg [129] indicated that high frequencies increase the denominator of C_Jac and therefore complicate co-occurrence relationships. In this study, when the maximum frequencies of conditional and unconditional occurrences in the network exceeded 500 and 10,000 times, respectively, only approximately 25% of the outputted words were used for the co-occurrence network because of the decrease in C_Jac. Therefore, a large sample size (i.e., a word frequency of more than 10,000 times) may not be suitable for the detailed analysis to clarify the characteristics of each passive cooling method.

In the case of a medium sample size, the co-occurrence network expressed the purpose of the application, strengths and weaknesses of each passive cooling method, and potential combinations with other passive cooling methods. Approximately more than 50% of the outputted words appeared in the co-occurrence network with a C_Jac of more than 0.2, and the maximum word frequencies of conditional and unconditional occurrences in the network were 35–118 times and 1043–4081 times, respectively. Garg and Kumer [130] conducted a co-occurrence network analysis of Twitter using a sample size greater than 100 tweets to ensure reliable analysis.

The constructed method was applied to the comprehensive overview employing a considerable number of the publications (N = 39,604). Nevertheless, the frequency of each word of less than 4000 is recommended for a detailed analysis. Moreover, the frequency of words in the sampled publications, which is the frequency of unconditional occurrences, needs to be at least 100 times to conduct a reliable network analysis.

4.2. Evaluation of the Constructed Bibliometric Analysis Methods

Miranda et al. [41] reported that the largest sample size of reviewed research papers on passive cooling technology contained a maximum number of 2859 samples. Nevertheless, they referred to publications written by authors who published a large number of publications only and concluded that the exclusion of a substantial number of other studies was a limitation of their study [41]. Cañas-Guerrero et al. [48] classified the publications on Construction & Building technology in the WoS based on the existing categories, such as research institutes and journals, to reduce the bias in selecting papers. However, this method cannot analyze a specific topic because of limitations in the categorization functions of the WoS [48]. In this study, passive cooling was selected as the target topic, and the use of the coding rules composed of the compound words enabled the limitations of previous studies to be overcome.

From the titles of the sampled publications (N = 39,604), 68,127 compound words were detected in the analysis. Subsequently, 26 compound words were selected based on the C_Jac, that is, the relationship between the position (d) and similarity (δ) of each word, and the collocation score (SC_Col), to create the 15 coding rules. Nakagawa et al. [58] indicated that expressing technical terms using a single term is difficult. In this study, we found that some words, such as evaporative and radiant, were employed as compound words rather than as single terms. Therefore, calculating SC_Col, which evaluates the importance of compound words, is recommended to construct effective coding rules. Based on the results of SC_Col, each coding rule needs to include compound words with a minimum SC_Col of approximately more than 2500 to construct effective coding rules. Fifteen coding rules were employed for 26 out of the 68,127 compound words that accounted for approximately 15% (n = 5961) of the sampled publications (N = 39,604) between 1970 and 2022.

Table 4 presents the number of publications on passive cooling methods based on the coding rule and references qualitatively reviewed by the present authors based on the co-occurrence networks constructed from a considerable number of publications. The reviewed references helped in the investigation of the detailed features of each passive cooling method, including strengths and weaknesses, in addition to the co-occurrence network analysis. Compared with the previous review papers (see Table 1), a comparable number of publications for each passive cooling method (i.e., 26–1922 publications) was efficiently sampled in this study, and the constructed method was able to analyze the trends of the passive cooling methods quantitatively, as shown in Table 4. Although the number of qualitatively reviewed papers for each passive cooling method in this study can be less than that of the previous review papers, the constructed method can draw its comprehensive overviews through the co-occurrence networks comprising a large number of samples.

Table 4. Number of publications on passive cooling methods for each period and reviewed publications based on the co-occurrence networks discussed in Section 3.4.

	1970~ 1974	1975~ 1979	1980~ 1984	1985~ 1989	1990~ 1994	1995~ 1999	2000~ 2004	2005~ 2009	2010~ 2014	2015~ 2019	2020~2022 *	Total	Reviewed Publications and Reference
Sampled publications	332	748	1540	1476	925	971	1598	3918	6681	13,537	7878	39,604
* cool roof	0	0	0	0	0	0	0	2	26	38	17	83	4	[120,121,122,123]
* evaporative cooling	0	0	2	0	2	2	11	17	20	54	24	132	4	[90,91,92,93]
* green roof	0	0	0	0	0	1	2	21	61	97	44	226	5	[94,95,96,97,110]
* green wall	0	0	0	0	0	0	0	4	3	17	9	33	8	[93,94,95,96,97,125,126,127]
* natural ventilation	0	1	7	2	2	20	32	97	100	132	61	454	12	[34,35,98,99,100,101,102,103,104,105,106,107]
* night ventilation	0	0	1	0	0	8	1	13	10	12	8	53	3	[109,110,111]
* occupant behavior	0	0	0	0	0	0	0	1	14	42	10	67	1	[108]
* phase change material	0	1	2	0	0	0	4	59	222	546	363	1197	9	[68,69,112,113,114,115,119]
* radiant cooling	0	0	0	1	0	3	3	10	17	31	40	105	5	[35,116,117,118,119]
* solar shading	0	0	0	0	1	0	2	0	8	10	5	26	7	[84,93,109,124,125,126,127]
* thermal behavior	1	1	6	4	3	10	11	34	45	95	41	251	2	[113,114]
* thermal comfort	1	3	15	12	28	35	71	241	313	736	467	1922	10	[69,99,100,101,102,103,104,105,106,107]
* thermal insulation	3	2	17	2	4	2	11	32	91	112	71	347	2	[34,109]
* thermal storage	0	15	19	14	8	11	25	52	178	351	200	873	5	[68,69,112]
* tropical climate	0	0	13	3	7	3	6	29	32	58	41	192	2	[92,93]

* until March 2022.

Table 4. Number of publications on passive cooling methods for each period and reviewed publications based on the co-occurrence networks discussed in Section 3.4.

	1970~ 1974	1975~ 1979	1980~ 1984	1985~ 1989	1990~ 1994	1995~ 1999	2000~ 2004	2005~ 2009	2010~ 2014	2015~ 2019	2020~2022 *	Total	Reviewed Publications and Reference
Sampled publications	332	748	1540	1476	925	971	1598	3918	6681	13,537	7878	39,604
* cool roof	0	0	0	0	0	0	0	2	26	38	17	83	4	[120,121,122,123]
* evaporative cooling	0	0	2	0	2	2	11	17	20	54	24	132	4	[90,91,92,93]
* green roof	0	0	0	0	0	1	2	21	61	97	44	226	5	[94,95,96,97,110]
* green wall	0	0	0	0	0	0	0	4	3	17	9	33	8	[93,94,95,96,97,125,126,127]
* natural ventilation	0	1	7	2	2	20	32	97	100	132	61	454	12	[34,35,98,99,100,101,102,103,104,105,106,107]
* night ventilation	0	0	1	0	0	8	1	13	10	12	8	53	3	[109,110,111]
* occupant behavior	0	0	0	0	0	0	0	1	14	42	10	67	1	[108]
* phase change material	0	1	2	0	0	0	4	59	222	546	363	1197	9	[68,69,112,113,114,115,119]
* radiant cooling	0	0	0	1	0	3	3	10	17	31	40	105	5	[35,116,117,118,119]
* solar shading	0	0	0	0	1	0	2	0	8	10	5	26	7	[84,93,109,124,125,126,127]
* thermal behavior	1	1	6	4	3	10	11	34	45	95	41	251	2	[113,114]
* thermal comfort	1	3	15	12	28	35	71	241	313	736	467	1922	10	[69,99,100,101,102,103,104,105,106,107]
* thermal insulation	3	2	17	2	4	2	11	32	91	112	71	347	2	[34,109]
* thermal storage	0	15	19	14	8	11	25	52	178	351	200	873	5	[68,69,112]
* tropical climate	0	0	13	3	7	3	6	29	32	58	41	192	2	[92,93]

* until March 2022.

4.3. Potential Combinations of Passive Cooling Methods for Hot and Humid Climatic Regions

The results of the detailed analysis using the coding rule and further reviews of the publications help find potential combinations of passive cooling methods for hot and humid climatic regions. For example, in a previous study, a strong improvement in the thermal comfort was reported when natural ventilation was applied as comfort ventilation owing to the evaporative and convective heat losses [104,105,131]. Correspondingly, our results show that the C_Jac of natural_ventilation_thermal_comfort is more than 0.5 in the co-occurrence network of natural ventilation. Previous studies have also found that a window system can be employed to increase the influence of natural ventilation [35] in response to the wind speed preferences of occupants in hot and humid climates [106]. Nevertheless, opening windows for natural ventilation during the daytime often increases the heat gain in the tropics. Several previous studies have shown that roof cover and ceiling insulation [33,34] which reduce the additional heat gain from solar radiation were better combined with natural ventilation to reduce the total heat gain of buildings as much as possible. However, previous studies [126,127] argued that although external shading devices such as louvers, shutters, and blinds decrease the heat gain from solar radiation, they are not recommended for naturally ventilated buildings because of the wind-blockage effect.

In combination with night ventilation, the thermal storage effect has been frequently investigated in previous studies. In the present analysis, overhang and blind appeared in the co-occurrence network of night ventilation, suggesting that the reduction in solar heat gain using these shading devices can be effective in maintaining the coolness obtained through night ventilation during the daytime. Similarly, in previous studies, green roofs and wall insulation were combined with night ventilation to eliminate the heat gain in indoor spaces [110]. Meanwhile, the high occurrence of the words related to PCM in the night ventilation and thermal energy storage co-occurrence networks indicates that there is a high potential for the use of latent heat thermal storage systems employing PCMs to increase the thermal storage effect through night ventilation in hot and humid climates. For example, Kitagawa et al. [119] reported that the thermal storage effect of PCMs increased the thermal comfort period in Indonesia, regardless of outdoor conditions. Therefore, PCMs may compensate for the disadvantages of natural ventilation, such as fluctuating outdoor wind conditions, and maintain thermal comfort even when the windows are closed [118]. Considering the recent drastic increase in the number of publications on PCMs, this combination is expected to be investigated further in future studies.

4.4. Limitations

When passive cooling methods are applied to buildings, the analysis of climatic conditions is required beforehand to maximize the cooling effect because the effectiveness of each method depends on the climatic characteristics of the building locations and design [132]. For example, Bhamare et al. [133] represented the comfort zones of 18 cities in India using a bioclimatic chart and suggested the cooling potential and appropriate passive cooling methods for each city. Meanwhile, the constructed method makes it difficult to concretely analyze the quantitative effects and specifications of the passive cooling system proposed in the publications because C_Jac and SC_Col are calculated based on the statistical data. Moreover, this study referred to tropical climates in the co-occurrence analysis (see Section 3.4.1). It did not analyze specific countries and regions in detail. Therefore, the influence of countries and regions on the results of the analysis is uncertain. Similar to this study, a previous study [41] did not include the words related to specific countries and regions when searching for documents on passive cooling methods and determining countries or regions based on the affiliation of the authors, which is sometimes different from the target country or region. If the impact of a specific country or region needs to be considered in a study, the part-of-speech (PoS) classification of countries and regions as proper nouns should be conducted in the documents, and the recommended coding rules should be properly set.

5. Conclusions

A bibliometric analysis method using text mining was constructed to provide a comprehensive overview of passive cooling methods for buildings in hot and humid climates. The constructed method was divided into two phases. First, a comprehensive overview analysis was conducted to clarify the trends in the studies on passive cooling methods between 1970 and 2022 (March). Subsequently, coding rules consisting of compound words were built based on the quantitative overview analysis to find possible combinations of passive cooling methods for hot and humid climates. Using the Web of Science and Scopus, 39,604 publications were sampled for this study. The main findings are summarized as follows:

·

The constructed text mining-based bibliometric analysis method statistically and comprehensively scrutinized a significant number of words in the sampled publications. In particular, constructing coding rules based on quantitative data, such as the Jaccard coefficient (C_Jac) and the collocation score (SC_Col), can reduce the bias when selecting publications and efficiently analyze the commonly studied passive cooling methods while extracting the research trends from a large number of publications. Calculating the collocation score (SC_Col), which evaluates the importance of compound words, is recommended when constructing an effective coding rule because some words are employed as compound words rather than as single terms. 15 coding rules using 26 out of the 68,127 compound words covered approximately 15% (n = 5961) of the sampled publications (N = 39,604) between 1970 and 2022.

·

Although the constructed method can be applied to the comprehensive overview that analyzes a considerable number of publications, a frequency of each word less than 4000 may be recommended for the detailed analysis to find potential combinations of passive cooling methods. Meanwhile, the frequency of words in the sampled publications, which is the frequency of unconditional occurrences, needs to be at least 100 times higher to conduct a reliable network analysis.

·

In hot and humid climates, natural ventilation plays a vital role in passive cooling methods to improve thermal comfort. As the number of publications on natural ventilation has steadily increased since the 1980s, natural ventilation is regarded as a common passive cooling method. Meanwhile, the number of publications on PCM has rapidly increased since the 2010s. The recent rapid increase in the number of studies implies that PCMs can be regarded as a suitable passive cooling method for thermal energy storage in hot and humid regions where the diurnal temperature range tends to be small. The constructed method and possible combinations of passive cooling methods will help engineers find effective combinations to improve thermal comfort in the planning stage of buildings. In future studies, qualitative literature reviews of passive cooling methods referring to the results of this study can be recommended to determine effective passive cooling methods.