Scoping Review (SR) via Text Data Mining on Water Scarcity and Climate Change

Abstract

Climate change is causing the risk of weather events and instable water accessibility, making water insufficiency a serious problem. According to the 2022 Intergovernmental Panel on Climate Change (IPCC), 70% of extreme weather events such as droughts and floods have been water-related in the last 15 years. Since the climate change processes are speeding up, this percentage is expected to increase. A plethora of researchers have been working on the correlation between water scarcity and climate change. The purpose of this paper is to examine the published research dealing with water scarcity and climate. Therefore, the study carries out a scoping review (SR) via text data mining and reveals the related topics. Two kinds of analysis were carried out using IRaMuTeQ software: descriptive analysis (TTR, Giraud index, Herdan index and Zipf’s curve) and cluster analysis (Reinert’s method). The results show that the topic of water scarcity refers to the direct and indirect economic impacts on its availability for irrigation, the willingness to pay more for an irrigation water supply and the role of public institutions in “achieving sustainable development goals”. The conclusion of the paper highlights the role of this analysis for developing future research and identifies implications for theory, practice and policy in order to overcome the current global challenges related to water scarcity and climate change.

1. Introduction

Sustainable development has become a main concern for modern society as the climate change processes are speeding up [1,2] within an environment of uncertainty and financial instability [3,4]. People take more responsibility for nature and the environment [5]. Increasing importance has been given to the rational use of raw materials and especially water. Indeed, water scarcity is the primary issue for today’s world [6]. Especially due to the COVID-19 pandemic, the urgency of clean water has been raised: the world population saw the crucial importance of sanitation for avoiding pathogens, bacteria, infections and related diseases [7]. On the other hand, according to a United Nations report, billions of people—mostly in rural areas—do not have access to clean drinking water and sanitation. Moreover, 40 percent of the global population is affected by water scarcity with an increasing trend, and 1000 children die every day due to water- and sanitation-related diseases [8]. According to Mancosu et al. [6], the problem will be more serious in the future. Certainly, yet in 1999, Seckler and colleagues [9] argued that more than a quarter of the world’s population would deal with the problem of water scarcity in the twenty-first century.

Investigating the reasons for water scarcity, Mianabadi et al. [10] argue that it might be caused not only by natural droughts as it is considered widely but also by anthropogenic droughts. Indeed, Tschand [11] argues that agricultural irrigation wastage is the main factor for the global water shortage epidemic. The author proposes the usage of smart farming tools and automated irrigation systems in order to minimize the impact of agricultural fields on global water scarcity. Similarly, in the example of North China, Zhang et al. [12] argue that over 80% of the total local water consumption is generated by irrigation systems, and the right management of crop watering could be the solution. Mehmood et al. [13] have raised another problem related to the lack of clean water and agricultural activities. Specifically, the authors state that due to water shortages, farmers reuse wastewater for crops which negatively affects the quality of the product and human health. The authors suggest raising the awareness of farmers regarding this problem and strengthening policy in order to avoid such hazardous impacts from agricultural activities. In addition, Mianabadi et al. [10] suggest that human activities and governmental plans linked to the water usage should be revised.

On the other hand, Dang et al. [14] state that the reason for water scarcity in China is the growing population and increased household water usage. The authors argue that revised urban sustainable water management strategies and strengthened state polices, in order to increase water-saving activities and improve the reuse rate, could solve the abovementioned problem.

Cacciuttolo and Valenzuela [15] also highlight the problem of climate change and water scarcity in the field of tailings management in copper mining projects and underline the importance of new technologies and strategies for overcoming this problem. Specifically, the authors suggest the development of water supply technologies or their combination for accessing alternative new water supplies, such as sea water desalinization, the direct use of sea water or water recovery from tailings. Similarly, Vila-Tojo et al. [16] argue that the way for overcoming the problem of water scarcity lays in the development of recycled water systems and raising the awareness of populations so this system is accepted publicly. Indeed, Afsari et al. [17] describe the positive impact of rainwater harvesting (RWH) for ensuring safe water for populations. Specifically, the authors state that thanks to RWH, the investigated region showed improved livelihood potential and security against climate change. However, the restricted financial resources, absence of technical knowledge, unawareness of hygiene and lack of training programs obstructed RWH technology’s adoption. Numerous international entities, such as the United Nations (UN), Food and Agriculture Organization (FAO), United Nations International Children’s Emergency Fund (UNICEF), World Bank and European Commission (EC) highlight the importance of the issues related to water scarcity, identifying some important facts [18,19,20,21,22]. Specifically, according to a UN report [18], despite the world progressing steadily, many people will lack essential services for human health, such as safely managed drinking water, sanitation and hygiene services, in 2030. On the other hand, the FAO [19] deals with the problem of water scarcity in the agricultural sector and argues that about 3.2 billion people worldwide live in agricultural areas with high to very high-water shortages. UNICEF [20] underlines the importance of the living conditions for children and highlights the fact that 450 million children suffer from high or extremely high-water vulnerability. The World Bank [21] research outputs, which consider the correlation between water scarcity and climate change, argue that water must be at the core of adaptation strategies in order to reach sustainable development goals (SDGs). Additionally, the EC [22] advises on safe water reuse by developing guidelines and relevant policies regarding the global issue of water scarcity. USAID defines water security as ”having safe use of an“ adequate, reliable and resilient quantity and quality of water for health, livelihoods, ecosystems and productive economies [23]. Considering the global importance of water scarcity and climate change issues, as well as related economic and human wellbeing problems, the aim of this paper is to investigate and review the published academic works dealing with the abovementioned problems, analyze the trend of publications over the years and reveal the related topics. The results can draw new challenges, develop future research plans and identify implications for theory and practice, as well as for policy [24]. Indeed, our findings can push up and highlight the crucial role of strategies for ensuring environmental sustainability and competitiveness. The handling and protection of global water resources appear the main cornerstones of sustainable goals. In some way, the novelty of this work can be traced both in the significant role played by the findings and in the scoping review method that integrates quantitative and qualitative analyses. The next session of the article describes the materials and the methods of this study. That is followed by the results and discussion. Finally, the conclusion and implications conclude the paper.

2. Materials and Methods

2.1. Research Approach

The research was divided into three phases: the first phase concerned the creation of the database, based on the keywords “water scarcity” and “climate change”; the second involved the analysis of the data through lexicometric analysis; and finally, the third produced a cluster analysis according to Reinert’s method [25,26]. This methodology is framed by the scoping review (SR) of literature via text data mining, and it allows us to detect, select, classify and analyze the relevant literature in the field [27] and reliably reunites the results [28]. In this way, it is possible, through the synthesis of the results of the previous research, to create the state of the art of the topic [29], giving the possibility of mapping the literature and [30] creating a frequency ranking of words and their combinations. As the literature is growing, it became difficult for researchers to identify and investigate the research objective [31]. A manually performed literature review is a complex, multistage and time-consuming process where the novel techniques of text data mining can be a great support in minimizing the need of researchers’ effort and expertise for literature review activities [32]. Indeed, in the given circumstances, text mining that automizes some screening processes is offered as an alternative solution for researchers in order to investigate a matter more efficiently and effectively through literature review [33].

2.2. Data Collection and Analysis

2.2.1. Dataset

The papers that were considered for our data set were extracted from Web of Science (WoS), a database from Clarivate Analytics. WoS is one of the world’s leading databases recording over 12,000 journals and 160,000 international conferences [34]. It is the most consistent database according to Zhao and Strotmann [35] and widely used for citations for scientific purposes [36].

All the articles were extracted from WoS through the query “water scarcity” AND “climate change” without specifying an initial period and up to the search date which was September 2022. The research focused only on journals in the economic sector to ensure homogeneity between corpuses and avoid nonrobust results in textual analysis. Furthermore, only articles written in English were identified [37].

2.2.2. Lexicometric Analysis

The lexicometric analysis is based on a precise linguistic glossary, the main terms of which are listed below. By corpus or textual corpus, we mean a linguistic resource made up of a large and structured set of texts. The term token or occurrence refers to the total number of words in the corpus, etc., regardless of how often they are repeated. The term type or form refers to the number of distinct words in a corpus, etc. The term hapax is a word that occurs only once in a corpus. Finally, by co-occurrence, we mean a term that expresses the frequency by which two terms of a corpus occur next to each other in a certain order. It usually indicates words that together create a new meaning.

Lexicometric analysis was performed using the software IRaMuTeQ 0.7 alpha 2 (http://www.iramuteq.org/ (accessed on 15 October 2022)). IRaMuTeQ (Interface de R pour les Analyzes Multidimensionnelles de Textes et de Questionnaires) [38] was developed by the French programmer Pierre Ratinaud and is an open source, multifunctional and statistical software for the analysis of textual data [39] which supported the French language until 2009 and currently includes English and some other languages [40]. Its algorithms are based on Python programming language and statistical software R [41]. With statistical tools and through a recursive approach, IRaMuTeQ 0.7 performs text analysis such as similarity analysis, specificity analysis and lexical correspondence analysis (LCA) for text data. Therefore, IRaMuTeQ was chosen for the textual analysis because, compared to other software, it shows a high processing capacity and allows us to arrive at detailed reports with quality and highly reliable results. In addition, unlike other software of the same quality, it is open source while maintaining the same standards.

The analysis phases were as follows: (a) subject the corpus to cluster analysis, (b) identification of complex and compound lexias, (c) representation in the corpus of lexias and (d) application of IraMuTeQ’s unsupervised learning techniques through the construction of lexicometric measures and cluster analysis using Reinert’s method.

The analysis also provided several statistical analyses that together determine the descriptive aspects of the corpus, such as:

Type Token Ratio (TTR): This is a measure of the variation in vocabulary within a text, given by:

$$TTR=\frac{V}{N}\ast 100$$

That is to say, the ratio between V = number of different words (types), and N = total number of words (occurrences or tokens). High values of TTR indicate that there is a high degree of lexical variability in the text, whereas low values indicate the opposite. The type token ratio is therefore a relationship between the width of the vocabulary (V) and the size of the corpus (N).

If its value is less than 20%, the corpus is considered adequate for a lexicometric type treatment.

Hapax%: starting from the assumption that a hapax is a word that occurs only once in a single text or corpus, Hapax% is given by:

$$\frac{ V_1}{V}\%$$

i.e., the ratio between the total number of words that occur only once (V₁) and the total number of types (V) expressed in percentage terms. If it is less than the 50% cutoff, it indicates a corpus size corrected for statistical tractability.

Guiraud index: This is a measure of the lexical richness that aims to quantify the variety of vocabulary used, given by:

$$Guiraud index=\frac{V}{\sqrt{N}}$$

i.e., the ratio between V = number of types, and the square root of N = total number of occurrences. Compared to similar TTR, the square root has a mitigating effect on the impact of the number of tokens. It indicates an acceptable lexical richness of the corpus if it exceeds the minimum limit of 22 in absolute terms.

Herdan index: This is another index of lexical wealth given by:

$$Herdan index=\frac{Log V}{Log N}$$

That is to say, the ratio between the logarithm of types and the logarithm of tokens.

Zipf’s law: the empirical law discovered by Zipf (Zipf, 1949) reveals some frequency distributions of words in human language. In fact, according to the principle of least effort, people tend to use little vocabulary; therefore, there are a few words with high occurrence, and most of the words have low frequency. It is, therefore, possible to measure the greater or lesser discrepancy with respect to this empirical law:

$$f_r=\frac{C}{{\left(\mathrm{r}+\mathsf{\alpha }\right)}^{\beta }}$$

Here, f represents the word frequency, r stands for word rank in the list, C is a constant, and α and β are both constants for the corpus being analyzed. To examine whether the word frequency distribution satisfies Zipf’s law, Zipfian curve drawn on doubly logarithmic axes is a general means. If the distribution curve on doubly logarithmic axes is close to a straight line with slope −1.3, i.e., an absolute value 1.3, the frequency distribution of words in the corpus presents Zipf’s law. The Zipf’s curve slope is given by:

$$Zip{f}^{\prime }s Curve Slope=\frac{logN}{logV}$$

Here, V = number of types, and N = total number of tokens.

Pareto chart: This is a bar chart sorted in descending order from the highest frequency to the lowest frequency from left to right. In lexicometric analysis, the height of the bars reflects the frequency of words. Specifically, we used a modified Pareto chart with dual Y axis, containing a cumulative percentage line.

2.2.3. Cluster Analysis using Reinert’s Method

A cluster analysis was then performed using Reinert’s method (i.e., descendant hierarchical clustering) for textual data analysis, obtaining a dendrogram that highlights the most represented lemmas and forming homogeneous and heterogeneous clusters within them. In particular, IRaMuTeQ produces figures in which the width of the word is proportional to the number of text segments (i.e., the greater the width, the greater the number of text segments), whereas the height of the clusters (indicated by different colours) represents the frequency of text segments within the clusters.

3. Results

3.1. Descriptive Results

The corpus is made up of 127 variables: titles and abstracts extracted from the Web of Science Clarivate Analytics in the field of business journals.

Table 1. Characteristics of the corpus.

Characteristics	Frequencies
Number of texts	127
Number of text segments	715
Number of type	25,719
Number of forms	3786
Number of hapax	1683

shows 127 partition variables (i.e., journal article titles and abstracts). The software divided the file into 715 text segments. The number of occurrences, also called tokens, is 25,719; they constitute the total frequency of the corpus and are indicated by N. The number of different words (“forms” or “type”) is 3786 units, whereas the hapax (words with unit frequency) amount to 1683. The descriptive statistics depicted in Table 1 prepare the construction of the lexicometric indices for the validation of the corpus.

The contents were saved in text format (UTF-8 coding) for the automatic analysis. Each text was accompanied by the partition key variables indicated below:

(a)

identifier: **** number

(b)

author: * author_author name

(c)

year: * year_number

(d)

source: * source_journal name

Example of a trace:

**** 0001 * author_koopman * year_2015 * source_climatechangeeconomics

Subsequently, the lexicalization (

Table 2. Co-occurrences.

Terms	Frequency
climate change	57
water scarcity	52
water resources	18
water supply	17
water availability	16
global food	13
food security	12
population growth	12
irrigation water	10
water consumption	9
long term	8
water demand	8
water requirements	8
economic growth	7
economic impacts	7
impact of water	7
water efficiency	6
water intensive	6
crop production	5
energy production	4

), i.e., the identification of words that together take on a different meaning, mainly associates the word water with the terms “resources”, “supply”, “availability”, “irrigation”, “consumption”, “demand”, “requirements”, “efficiency”, and “intensive” and sheds light on the importance of water as a resource for economic and personal use, which is a relevant theme in the long term and inextricably linked to the growth population.

The first lexicometric measure (

Table 3. Lexicometric measure.

Measure	Equation	Value
TTR	$\frac{V}{N}\%$	$\frac{3786}{25,719}\%=14.72\%$
Hapax%	$\frac{ V_1}{V}\%$	$\frac{1683}{3786}\%=44.45\%$
Giraud index	$\frac{V}{\sqrt{N}}$	$\frac{3786}{\sqrt{25,719}}=23.6$
Herdan index	$\frac{log V}{log N}$	$\frac{log 3786}{log 25,719}=0.80$
Zipf’s slope	$\frac{logN}{logV}$	$\frac{log 25,719}{log 3786}=1.23$

) is the TTR (type token ratio) given by the ratio between the number of different words (types), equal to 3786, and the total number of words (occurrences or tokens) of 25,719 lexical units, i.e., 14.72 expressed in percentage terms; it is a sign of a valid lexicographic measure that allows us to consider whether the corpus is adequate for automatic or semiautomatic treatment.

The absolute number of hapax is 1683 units, corresponding to 44.45% of the forms (Table 3). It is below the limit threshold of 50%; this measure argues in favor of the correct size and breadth of the corpus as a whole.

The Guiraud index, given by the ratio between the number of forms and the square root of the occurrences, is 23.6 (Table 3), i.e., it indicates an acceptable lexical richness of the corpus as it exceeds the minimum limit of 22 in absolute terms.

The Herdan index is 0.8 (Table 3), confirming the considerations relating to lexical richness. The slope of Zipf’s curve (Figure 1) is 1.23 (Table 3) in absolute terms. In this case, it represents a good lexical richness of the corpus, whereby lexical richness we mean the proportion between different words.

The study of the distribution of words in the text (see Figure 2), as they are processed by the software, takes place by opting for two tools: a table ordered according to decreasing frequencies and a Pareto chart. According to Figure 2, the first twenty words of the term document matrix, represented in absolute frequencies, include referential words that are linked to the names “model”, “impact”, “resource”, “policy”, “agricultural” and “conflict”.

The words with the most representation, organized according to the descending order of the frequency, indicate that the first five items (excluding “water”, “water scarcity” and “climate change” deriving from the search query) are “irrigation”, “model”, “increase”, “economic impact”, “agricultural” and “conflict”.

The parts of speech (POS) of the table are made up of 14 nouns (nom), 5 unrecognized forms (nr) and 1 verb (ver); they do not present interpretative ambiguities as they have a well-identified semantic area.

The theme of water scarcity refers to the direct and indirect economic impacts on its availability for irrigation [42], the willingness to pay more for an irrigation water supply [43] and the role of public institutions in “achieving sustainable development goals” [44]. In the aspect of agricultural production, there is a “growing demand and an inefficient management of water and related infrastructure” [45].

The literature also argues its potential to provoke social conflicts [46] due to the water mass decline and its importance as a factor for production “across all sectors and regions of the global economy”. The conflicts that can arise from water scarcity require diversification in favor of “lesser water intensive sectors combined with investment in water saving infrastructure and improved irrigation efficiency” [47]. The consequences of the reduced availability of water for people, society, ecological systems and economic growth [48] determine the need for rational and efficient use.

Water resource management planning constitutes a crucial concern [49], and it has implications for the energy sector [50]; in fact, it “has an overall negative effect on the level of energy production” [51].

It is necessary to take into account both the “spatial and temporal variations of water availability as well as the socioeconomic and policy conditions that affect international trade” [52], increasing the competition for the acquisition of this resource [53].

The drivers that condition water scarcity are to be found in economic growth and will be crucial for the “agricultural economies in many areas of the world” [54] and for evaluating the impact on agricultural production and the “magnitude subsequent on global food security” [55].

The macroeconomic impact of water scarcity requires an assessment through the use of scenario analysis [56] “as described in the shared socioeconomic pathways” (SSPs); input–output models based on IPCC, RCP and OECD scenarios have also been used in the literature on SSPs [57]. The analysis of future water scarcity often takes place using the “Computable General Equilibrium Model” [42,58,59], and it is penalized by the “absence of standardized data” [59] which makes it difficult to model water demand or water supply; in addition, some scholars highlight a “variable data quality provided by numerous source and estimated values” [60] and, consequently, the need for “national, international and global water databases”.

Researchers also use “sequential modeling methodology” [61], “multiregional input-output optimization model to assess impact of water supply”, spatial stochastic frontier models applied to irrigation [62] and the typical dynamic stochastic general equilibrium model (DSGE) as quantitative and stochastic instruments of traditional economics [63].

3.2. Cluster Analysis Using Reinert’s Method

With respect to the results of the cluster analysis, Figure 3 shows three clusters generated using the agglomerative hierarchical method with IRaMuTeQ software, showing a very good segment classification result of 86.29%, which is well beyond the 75% minimum threshold. Furthermore, 617 segments were classified out of a total of 715 as well as 1092 lexical units with active forms greater than three.

In summary:

• Number of texts: 127
• Number of text segments: 715
• Number of modules: 3786
• Number of occurrences: 25,719
• Number of lemmas: 2.539
• Number of supplementary modules: 397
• Number of active forms with frequency ≥ 3: 1092
• Average of shapes per segment: 35.970629
• 617 classified segments out of 715 (86.29%)

In particular:

• Cluster 1 classified 251 text segments out of a total of 617 (40.68%)
• Cluster 2 classified 175 text segments out of a total of 617 (28.36%)
• Cluster 3 classified 191 text segments out of a total of 617 (30.96%)

Cluster 1 (red) groups political–institutional issues related to water scarcity, including institutional and governance issues, security and the potential generation of conflicts, together with the need for reforms for the community.

Cluster 2 (green) constitutes a universe of meanings linked to the analysis tools used by researchers such as scenario analysis and general economic equilibrium models; it also evaluates the aspects linked to the demand for the resource in relation to the various economic sectors, countries, regions and, from a global perspective, with reference to a global dimension.

Cluster 3 (blue) assumes significant importance for the relationship between water scarcity and the agricultural sector, the issue of irrigation, the effects on farms and the value of water resources in this specific economic sector.

In summary, three relevant strands dominate the investigated dataset: political–institutional issues (Cluster 1), the analysis tools used by researchers such as scenario analysis and general economic equilibrium models (Cluster 2) and the relationship between water scarcity and the agricultural sector (Cluster 3).

4. Discussion and Conclusions

The study results express the relationship between the topics of “water scarcity” and “climate change” and several relevant areas. Specifically, the problem of water scarcity is related to the economic condition of the country. The implementation of novel systems that give the possibility of reusing water resources in a correct way requires investment in the field; on the other hand, reusing the wastewater in the agriculture might impact product quality and human health [13,42,43]. Water scarcity has a negative effect on energy production [51] and international trade as well [52]. Therefore, water scarcity is strongly related to economic growth [54] and governmental intervention for developing the infrastructure and a system as a whole in order to use water in compliance with sustainable development goals. This goal is becoming increasingly important [44].

Moreover, this study contributes this theory by applying research methodology which has not been widely used but offers text analysis in a more effective and efficient way than manual analysis [33]. In line with other research [64], it is possible to highlight that climate change is likely to negatively affect either crop production or water availability, pushing for further investigation to improve irrigation strategies. Therefore, a scientific approach is recommended to decrease ecology, sustainability, and environment impacts as also stated by Zy Harifidy et al. [65]. In addition, according to Pizarro et al. [66], human actions and overuse are also quickly exhausting water resources, in addition to changing the climate and affecting precipitations. Encouraging sustainable water management (SWM) practices seems crucial in order to safeguard water sustainability by pushing policymakers to reformulate strategies and policies to combat water scarcity [67].

During and after the pandemic, there has also been a major focus on the importance of clean water for sidestepping diseases [7]. Therefore, by considering the prominence of the link between water scarcity and climate change issues, as well as the correlated economic and quality of life problems, this paper can promote the development of future research by identifying paths for theory, practice and policy [24].

Consequently, the outcomes of this study have huge implications for theory and practice. Certainly, they might be used as a guideline for governments and policy-makers in order to properly plan the system of water usage in different industries for different purposes, advance the techniques for the safe reuse of water resources and increase the availability of clean water for society and businesses.

As claimed by all the international supra-governmental entities, such as the World Bank, WHO, FAO and UNICEF, the issues related to water scarcity are complex and change across countries and areas, therefore requiring specific and efficient technologies for specific contexts (remote sensing and geophysical tools), and methodologies appear crucial. The purpose of the research in this topic is to highlight the necessity to try and build and collaboratively handle water by means of different paths: developing water structures–networks; reinforcing green solutions and infrastructures; and promoting information and knowledge transfer aimed at changing businesses, people’s behavior and policy.

The investigating topic of this article is of crucial importance for current society. Water scarcity limits the availability and accessibility of water in households and workplaces which affects a lot of industries, as well as economies of scale. In addition, water scarcity results in difficulties in maintaining hygiene which causes an increased threat of contracting diseases. The COVID-19 pandemic clearly illustrated the correlation between hygiene and virus transmission. Therefore, the study outputs have a significant contribution in practice, as well as in theory in order to further develop future research in the field.

Moreover, the article uses the scoping review together with the text data mining techniques through IRaMuTeQ software as a methodology. Although it is not a widely used method, it generates new knowledge in the field by mapping the literature on the topic, making relevant statistical analysis and identifying key concepts, theories and sources of evidence, and, therefore, it engages theory and practice.

The limitation of the research might be due to the field of the collected articles, i.e., economics: this surely reduces the quantity of the reviewed works. On the other hand, the huge quantity of articles from different fields could make both the one-by-one analysis of the articles for the authors and the analysis for the software more difficult. Then, the large heterogeneity of the clusters within them should be further reduced with the identification of subclusters in order to reduce the internal variance.

It is possible to define possible future directions. Firstly, in addition to the Web of Science, the Scopus dataset can be included in the analysis. Secondly, the relationships among water scarcity types can be investigated more. In addition, we intend to adopt further data mining techniques, both to complete the analysis framework already started and make useful comparisons between the methodologies used in order to verify which of them provides the most reliable and effective classifiers. These additional techniques include support vector machines (SVM), linear discriminant analysis (LDA), random forest (RF) and multilayer perceptron (MLP). Furthermore, in our future lines of research, we will use the Shannon entropy indicator in association with data mining and text search techniques.