1. Introduction
Mozambique is one of the countries most affected by hunger and natural hazards [
1]. It is also affected by civil conflict and war, epidemics, and diseases such as malaria or HIV [
2]. Rural livelihoods predominate the area, and due to climatic conditions, droughts, and other processes, malnutrition and food insecurity are prevailing [
3,
4]. Despite tremendous efforts by international organizations to provide food and development aid, the conditions have improved, but malnutrition and high death rates due to diseases or natural hazards prevail [
5]. It is, therefore, still important to conduct scientific research into the reasons and drivers of risk in Mozambique. Hence, a literature analysis of major risks and disasters related to rural livelihoods and food security has been conducted. The literature analysis assesses which studies and topics have predominated research so far. Based on these results, future directions and needs for further research are derived from the results of this article. To understand what is necessary to research, the body of literature is analyzed regarding which already substantial work exists to permit literature analysis. On the other hand, such results can indicate gaps in the current research.
Currently, there is an increasing international trend in analyzing how risks are interconnected and how compounding events and creeping processes create interdependencies and dynamics that provide future challenges [
6,
7,
8]. The debate about future development pathways is especially strong within the scientific fields of climate change research and sustainable development [
9]. Mozambique experienced high death tolls due to natural hazards, and rural livelihoods are closely dependent on climate [
10,
11]. Therefore, sustainable development and rural livelihoods have become major research interests [
12]. Recent research analyzes the interrelations between climate change and armed conflicts and how far international law already covers both [
13]. This is not only overlapping the drivers of risk, but also the compounding drivers of the impact that are increasingly being analyzed. For instance, increasing urbanization drives marginalization and segregation, contributing to higher climate change risks [
14]. Urban and rural areas, on the other hand, are analyzed in terms of farmers’ perceptions of climate change and their information-sharing networks using mobile phones [
15]. Contributions of climate change to natural hazards’ compounding events, such as rainfall and tidal waves, are analyzed for coastal cities such as Maputo [
16]. Research on climate change in Mozambique covers sustainable livelihoods, especially vulnerable populations, agriculture, droughts, storms, and floods [
17]. Daily practices in dealing with hazards, such as floods and participatory processes at the local level to engage with climate change actions and participatory urban planning, are adaptation actions and ‘everyday realities’ that are analyzed in various studies [
18,
19,
20]. Agricultural investments and road infrastructure are drivers of adaptation and development under the uncertainty of climate change scenarios, as is the case for neighboring countries [
21,
22].
Concerning natural hazards, the socio-economic development of least developed countries is of special interest to studies using climate change development scenarios [
23]. Related research covered the loss and damage debate influencing political discourse on climate risk and compensation financing [
24]. Similar research covers Disaster Risk Reduction policies in Mozambique, exposing financial and coordination gaps [
25]. There is relatively little research about natural hazards in Mozambique comparing natural hazard types. Most research focuses on specific natural hazard types, such as flood risks and related management, governance, or analytical aspects [
26]. Major cyclones and their impacts are covered in scientific studies and media based on huge damages and losses [
27]. Supporting humanitarian preparedness and response through early warning systems is related to cyclone research [
28]. Droughts are covered in rainfall patterns, agriculture, and related famine and child mortality risks [
29,
30]. Recent studies cover analytical methods such as deep learning or machine learning for flood hazard mapping [
31,
32]. Resilience is analyzed for farmers coping with natural hazards with a financial and risk management perspective [
33].
Concerning food security, malnutrition is a major determinant in Mozambique and urban areas [
34]. However, most research covers agricultural areas and farmers’ strategies and resources dealing with food production and food security [
35]. Land holdings are a major determinant of income and food security in war-affected northern Mozambique [
36]. Biofuel feedstock is a special topic of attention in research and analyzes the smaller production rates of small-scale farmers versus agro-industrial projects, illustrating an ambiguous role of sustainability principles [
37,
38]. The income and employment of smallholder farmers remain the focus of recent research on food security [
39,
40].
Studies analyzing rural–urban relations to livelihoods, climate change, or natural hazards are scarce. One study investigates migration decisions due to environmental stress and finds perceptions of weather events that are highly correlated with economic situations in the coastal city of Beira [
41]. New economic niches are created by female vendors in urban or peri-urban areas selling small-scale agricultural products [
42]. Urbanization in Mozambique or Angola was driven by rural farmers’ demands to diversify income in urban informal sectors, limited market opportunities, and a lack of infrastructure [
43].
This article explores how many studies have looked into climate change, natural hazards, disaster risks, diseases, daily risks, sustainable development, rural livelihoods, and connected fields. It is interesting to analyze how far interrelations between those fields already exist. These interrelations are analyzed using a bibliometric analysis. The co-occurrence of keywords in scientific articles is analyzed using a bibliometric clustering tool. The results are intended to be helpful for science and non-governmental organizations (NGOs), as well as development planning and governance. The main hypothesis guiding this article is that multi-risk analysis studies are still underrepresented in Mozambique.
2. Materials and Methods
Mozambique is bordered in the south by South Africa and Swaziland, in the west by Zimbabwe and Zambia, and the north by Malawi and Tanzania. To the east, Mozambique shares a long coast with the Indian Ocean. It is predominantly low-lying terrain, with vast agricultural and even undeveloped stretches of land. The climate is tropical to subtropical, with more rainfall along the coast due to the regular seasonal influence of the Indian Ocean monsoon grains. The south of Mozambique is dryer than the north. Temperature averages range from 20° to 23° in winter and 24° to 27° in summer [
44]. Mozambique experienced a Civil War for 16 years that ended in 1992, with 5 million people being forced into migration [
11]. Several disastrous floods have affected the country and displaced hundreds of thousands. Cyclones and grouts are other recurrent hazards.
An initial online literature search guided the search strings and selection of the main focal problem keywords. Similarly to a snowball principle [
45], the conceptual focus was informed by the information provided by a conference [
46] to which the author was invited. The conference in Mozambique, 14–15 February 2024, focused on rural food environments and on the invited session on the relation of rural livelihoods to resilience and of food security to disaster risks. The conference marked the end of a three-year research project focusing on rural livelihoods [
47].
In addition to that information and the online literature, the search first focused on reports of international and local NGOs and their findings about needs assessments connected to the overall topic of rural livelihood, risks, sustainable development, disaster events, food security, and its related aspects. Additional fields of interest also informed the further selection, such as whether there are differences between rural and urban populations, vulnerabilities, livelihoods, and risks. This information collection then informed the selection of the main research questions that later guided the selection of the literature analysis, focusing on scientific literature only.
The method in this study consists of a literature analysis based on a Scopus database search with search strings, including abstract, title, and keyword information. The Web of Science and Google Scholar results were also screened. Still, since they mostly produced overlaps, the study selected only one platform, Scopus, to make it easier for comparison studies to use the same search strings and samples. The temporal coverage of the Scopus database varies according to the search term combination. For example, the terms Mozambique AND war go back to the earliest publications from the 1960s. Publications about the Mozambican Civil War that ended in 1992 are covered. However, since all the publications may not be digitized or registered, it could imply a representation bias. Language is another bias observed in the publication results. For the search term combination Mozambique AND war, from around 800 results (title, abstract, and keywords), only 25 were in Portuguese and 12 were in French, while English dominated as a language.
The files of 19 topical search strings were exported as CSV files. A bibliometric cluster analysis of co-occurrences of keywords was conducted with the help of the VOSviewer tool, version 1.6.20. Parts of the functionality of the open-source tool have been used, but it was mainly used to generate systematic keyword occurrence data and was analyzed further using Excel, version 16.66.1.
The guiding research questions for this article are as follows:
Which are the predominant risks in Mozambique, and which relations of risks can be observed? This main research question is further broken down into more specific research questions to narrow down the focus:
RQ1. Regarding food security and rural–urban livelihoods, which are the predominant risks in Mozambique that are considered in research?
RQ2. Which of those risks are interconnected in terms of hazard–vulnerability relations?
RQ3. Which of those risks are daily or disaster risks, and which are connected to which type of vulnerable populations?
RQ4. How can those interconnections be integrated into a conceptual framework?
2.1. Context and Relevance of Hazards and Risk in Mozambique
Mozambique has high population growth rates and young segments of society [
48]. Population and related urban growth are to be expected, as well as a transition from rural to urban livelihoods. This transition will differentiate the population into rural, small, provincial, and major urban centers. It is, therefore, an important time to take stock of the conditions and different risks within rural and already existing urban areas. This can help to better plan for future development and improve livelihoods in a way that is customized to the local context. At the same time, social and cultural contexts also have to be observed. It is therefore important to also analyze which population segments, age groups, gender, and other factors already shape risks in Mozambique.
Diseases and health-related risks will be analyzed since they take high death tolls every year. This will be compared with death tolls from disaster events, mainly from natural hazards. The term ‘daily risk’ is used here for those health-related risks that occur regularly, are captured in annual statistics, and are relatively similar each year. For example, the number of people dying from HIV ranges from around 40,000 per year [
49], malaria to around 15,000 [
50], and tuberculosis from around 12,000 [
51].
The EM-DAT [
52] disaster database shows that Mozambique has continuously experienced different types of disasters, with deaths of over 100 people per year in the past decades (
Table A1 in the
Appendix A). The highest death toll has been a drought from 1979 to 1981, with an estimated 100,000 deaths. This is followed by 800 people dying in the riverine flood in the year 2000. The types of disasters range from floods, cyclones, and drought on the natural hazard side to cholera outbreaks, road and rail accidents, and explosions on the artificial side.
Comparing this number of events with the highest number of people affected is shown in
Table A2 in the
Appendix A. A staggering number of millions of people were affected due to droughts, floods, and tropical cyclones. Many of these cyclone events happened quite recently, such as cyclones in 2019, 2021, or 2022, and also the 2020 and 2021 droughts.
2.2. Bibliometric Analysis
Bibliometric analysis is an established method in many fields of science [
53,
54,
55]. It analyzes different sources of literature systematically and uses graph theory or statistics to indicate interrelations of keywords, authorships, etc. [
54,
56]. Research trends and developments can be analyzed by tracking keyword occurrences over time [
57]. Different tools exist to support extracting keywords or other information, analyzing their statistical occurrences, and weighting similarities using graph theory. The VOSviewer is one of the other tools that are open and not restricted by a paywall. It is also applied to analyze standard science literature platforms such as Web of Science or Scopus [
58,
59].
The settings are described to make them comparable to follow-up studies or studies in neighboring countries. The import settings in the VOSviewer were set to a bibliographic data map and co-occurrence—all keywords. Import thresholds were set with a target of max. 50 keywords in the first run; keywords that were not relevant, such as “article” or “Mozambique” or neighboring country names, were removed. The first run mainly enabled a first visual interpretation of the main keywords occurring. This aided in the identification of keywords to select for further analysis. The visualization settings were set to maximize line sizes and labels. Several runs were conducted with the same files, excluding keywords not connected to the original research focus of the guiding research questions.
In the second run, keywords, such as area or methods, were excluded, and hazard, risk type, vulnerability, or context-related keywords were kept. In the third run, all keywords were retained, except those without any link to another keyword in the VOSviewer (the tool automatically offered this option in the final exporting step). It was exported as a JSON file, and then the number of keywords that occurred was analyzed in Excel.
The exact phrasing of the search terms and identified keywords, including plural or alternative expressions, is captured in the representation of the results further below. In particular, population characteristics such as age groups or rural/urban contexts related to livelihoods were captured.
The result table (Tables 1–7) shows the keywords per item. The top three items per focal problem and the total sums of the columns and lines are highlighted in gray for better readability and visualization. In case there are equal numbers, a max. of three are highlighted in gray. Keywords that were the same as the term used in the focal problem, such as climate change, malnutrition, resilience, or vulnerability, were excluded from highlighting. Only five or more keyword occurrences are selected for the top three rankings since lower numbers are considered insufficient to conclude. The results show the absolute numbers of keywords per focal problem to identify the suitable keywords for further research and the related data sets from the search strings. In addition, the data were normalized as keywords per number of articles and included in the
Appendix A for comparison.
2.3. Data Set
The data set consists of results from Scopus conducted for 19 search strings with a sufficiently high number of over 7000 resulting articles and over 60,000 keywords. Search string results with less than 100 articles were excluded since they produced less than 1000 keywords as data sets and were therefore regarded as insufficient for further cluster analysis. Both thresholds are based on experience with previous analyses and publications that showed that sampling below a certain threshold does not produce sufficient entries in title or keywords to generate clusters in the bibliometric analysis. This rough value was gathered through previous experience working with the VOSviewer and related publications. The search was conducted for all search strings on the same day, 19 December 2023, to enable a consistent data set. In previous studies, it had been observed by the author that even within a few days, new entries of publications can enter a database such as Scopus.
Such strings consistently included the country name Mozambique as connected with the focal problem keywords. The main focal problem keywords were selected according to the most predominant natural hazards and also the most predominant types of diseases in Mozambique. Human-caused hazards of other types, such as road, rail, or industrial accidents, were excluded due to the relatively lower death tolls and to narrow the focus to rural livelihood conditions.
Additional search terms included urban and rural contexts. Case study regions of the related conference project were tried. Still, the number of results in Scopus regarding the available scientific articles needed to be higher to enable a more systematic bibliometric analysis. As an alternative for an urban center, the city of Maputo was retained and produced over 270 articles. Finally, a few selected additional search terms were included in light of current research interests in disaster risk, such as transformation, infrastructure, and resilience. The overall search terms risk, disaster, security, and safety were also included to cover the main research question and the focal problem of risk.
The number of articles that were found already revealed heterogeneity in the findings. Most articles were found on the topic of risk connected to Mozambique. Articles on food follow this, but the further connection and limitation to food and security have already resulted in a much lower number of findings. The search term food was retained to enable a later comparison to one of the focal problems and research questions centering around food security. The VOSviewer tool analysis identified different keywords from those articles (
Table A3 in the
Appendix A).
Interestingly, the ratio of keywords per article also varies. That ratio (last column in
Table A3 in the
Appendix A) may indicate how much the topics of the focal problems and the research intention are also reflected in the results. While the number of keywords and articles is highest for the term risk, the ratio of keywords per the number of articles is relatively low compared to the other search string combinations. For example, the search string Mozambique AND safety produces the highest ratio of keywords per article. As a first assumption, the data set of Mozambique AND safety could provide more keywords into connections that are relevant to this study’s scope. This assumption will be tested by analyzing the data in more detail in the results section.
More search string combinations were tried, but articles were excluded due to the low number of results in Scopus, as shown in
Table A4. Therefore, analyzing the originally intended research question about multiple risks took time. Different expressions and combinations of the terms multi AND risk were tried. Also, in the more detailed analysis, the findings of individual keywords within articles around multi-risk were analyzed to try to identify explicit research on this topic of interrelated risks. The overall number of findings of articles in research about Mozambique, with over 15,000 articles, indicates how prevalent topics such as risks are. With a finding of 1700 articles from 15,000 about risks, the topic of risk is a predominant topic for Mozambique overall. This underlines the scope and relevance of the research.
3. Results
The results from the literature analysis are structured in the following according to different topics. These topics emerged during the literature analysis of the raw data and the clusters provided by the VOSviewer. To better organize the results and improve the readability, many topics were organized into focal problems defining the tables’ first column. The next topics in the header row of the tables are keywords such as cyclone, drought, etc., that emerged from the VOSviewer analysis but were also guided by the research question focus. Finally, many column-defining keywords were reorganized into different tables according to hazards, disease, types, population, and livelihood tables, and research paradigms were separated into different tables.
Table 5 shows the first topic of hazards. From the range of hazards, climate change consistently has the highest number of keyword occurrences, except for articles mainly related to civil war or safety. The findings for rural AND risk are also comparatively lower than for other focal problems or hazards in
Table 1.
Rain, war, and pandemic have the lowest keyword occurrences from all hazards analyzed. On the other hand, droughts and floods have high numbers, as well as cyclones and epidemics. This shows many articles within those hazard topics, resulting in many keywords related to our research questions. The results mainly indicate keywords that are promising for further literature analysis. But, at the same time, it is interesting to identify very low numbers of keywords, lower than five, for example. These keywords and focal problems can help to identify research gaps or areas for further research.
The results of the table can also be analyzed line by line for each focal problem. Since a detailed description would exceed the manuscript’s length, only examples per table are discussed. For example, the focal risk problem mainly relates to epidemics, climate change, and floods. Droughts have lower keyword coverages within the risk data set by comparison. This is interesting since droughts have produced the highest numbers of deaths and also people affected by far in Mozambique. However, it might be another limitation that droughts are only sometimes identified as disasters. It is a matter of definition and, therefore, also terminology as to whether droughts are regarded as disasters, ‘Slow-onset events’, or slow and creeping processes. This could add another bias to the search results and reveal a gap in identifying droughts as disasters in this context. This indicates that traditional risk research and the risk topics in Mozambique are broader and therefore need further analysis.
The following focal problems have been selected to provide different data sets and additional information to identify better which risks are relevant in research on Mozambique. Regarding the highest number of keywords per focal problem, the data set of articles around climate change produces even higher numbers than the data set about risk despite a lower overall number of articles and keywords. Research keywords around climate change center mostly around floods, followed by droughts and then cyclones. Floods are the third highest number of keywords related to the hazards in
Table 1. Floods are related mostly to climate change, cyclones, and droughts and, therefore, are similar to the climate change data set.
As a major research interest of this article, food security produces the relatively highest number of keywords for climate change, droughts, and epidemics. Analyzing the urban or rural focal problem rows, the overall keywords are lower than climate change or risk. Interestingly, however, rural and urban risk data sets have high numbers of keywords for epidemics, followed by floods. Climate change is also prevalent in both data sets but has a higher coverage in urban risk articles.
The next collection centers around diseases, as shown in
Table 2. The highest numbers of keyword occurrences per disease type are for HIV infections and malaria, followed by diarrhea. This is consistent with the findings of the most predominant occurrences in the VOSviewer analysis of the data sets. Although the main focus was on natural hazards with the highest death tolls, such as droughts or floods, within the focal problem of risk, HIV and malaria seem to be predominating and have much higher overall keyword numbers, over 300, as compared to the hazards table, for example. This underlines our research assumption that diseases and what are termed here as daily risks are just as important, if not more important, to compose the overall risk, at least in research about Mozambique.
The focal problem of risk has the highest number of keywords, followed by malnutrition and the combination of rural AND risk. Food security has a very low number of keywords related to disease topics.
Analyzing a few selected focal problems, risk and malnutrition share the same types of diseases with the highest numbers of such terms. But also, rural as well as urban focal problem sets confirm that HIV and malaria are the most prevalent diseases covered in studies. Other diseases, such as cholera and tuberculosis, are not within the top three highest numbers of keywords, but still, they show a high coverage in the area of focal problems, such as risk, malnutrition, food, safety, or disaster. And, especially for the focal problem of rural AND risk, tuberculosis also has a relatively high number of keywords.
Table 3 shows the keywords for the different population age groups. This can help identify which population groups have different risks covered in scientific studies. Adults have the highest number of keywords, followed by children and infants. Aged people are relatively low by comparison. The focal problems with the highest keywords are risk, followed by rural AND risk and Maputo AND risk.
The gray highlighting shows that children and adults consistently comprise those age groups with the highest numbers of keywords and the most focal problems. Food security only has adults with more than five keywords, which may indicate that this topic has yet to be covered regarding age, profile, and population differences. This also applies to other focal problems, such as cyclones or resilience. This might indicate future areas of research that could still be emerging, or it could, on the other hand, show that population and age factors have not yet had such an interest and maybe thereby relevance. Rural AND risk topics seem to cover adults as well as infants and children, and there is no major discrepancy to urban AND risk topics.
Table 4 investigates further population characteristics related to gender, social and economic conditions, and livelihoods. The highest numbers of keywords are captured for female and male characteristics, followed by pregnancy. Females and males are also consistent for all focal problems. Other items, such as education, income, livelihood, refugees, or vulnerable populations, only have selected areas with high numbers of keywords. Or, in the case of vulnerable populations, they are not well covered, except for the data set about risk or maybe rural AND risk and vulnerability.
The focal problems with the highest number of keywords are risk, rural AND risk, but also Maputo AND risk.
Table 5 summarizes items around mortality and morbidity. Mortality is most related to risk studies, but also to morbidity, malnutrition, and rural and risk data sets. Mortality is further broken down into different age groups. However, they show low numbers by comparison. Selected health-related problems, such as malnutrition or stunting, were added to enable a comparison of death rates. Stunting or lower than average growth of children and adults is important in food security and malnutrition research and has therefore been added. Obesity has been added to compare urban, rural, developed, and underdeveloped areas.
The highest number of keywords appear for risk, followed by malnutrition and rural AND risk. Mortality and morbidity also have the highest numbers of keywords for all focal problems. Interestingly, rural and urban risks again share the same high occurrences of keywords covered in scientific articles for mortality and obesity. Stunting is related to risk and malnutrition, but not so much in studies about urban or rural risk.
Table 6 shows items related to local and regional contexts. The highest number of keywords exist for rural areas, followed by urban areas. The results show that rural areas are predominant in more focal problem data sets than urban areas and are highly related to our research scope. The highest numbers of keywords were captured for focal problems or risk data sets, followed by rural AND risk and food. Overall, this shows that rural context and food security are highly interrelated with risk topics. For the city of Maputo, hospitals have relatively higher coverage and studies than other focal problems, except for risk.
Again, it is interesting to analyze those focal problems that have a low number of keywords. Cyclones, for example, have not been covered in scientific studies regarding rural or urban contexts very much so far. It is also less the case for drought as a topic or even climate change. This could be related to these hazards having large regional coverage. The low numbers of keywords for schools or primary schools mainly show that these are insufficient for further analysis. Schools and hospitals would be interesting local hubs to connect studies to.
Finally,
Table 7 shows various scientific terms that drive current research interest. The most keywords are captured for sustainable development, vulnerability, and poverty. Poverty had been expected to play a dominant role related to food security risks and rural livelihoods. Interestingly, resilience is a paradigm that has yet to be covered by high occurrences in articles, except for climate change. Multi-risk is not covered at all by keywords. The table also shows that using different search term variations was useful. Sustainable development produced many more results than sustainability, for example. This did not work for multi-risk. Similar combinations with development were not found to produce more results for terms such as vulnerability. For resilience, it has to be said that some focal problems show zero results here. However, in combination with other terms, it often produced one result, for example, for ecological resilience. For reasons of consistency, however, they have not been added; they have also been checked for every data set, and they have never produced more than one result.
The highest number of keywords per focal problem is retained for risk, food, and vulnerability. Overall, this table shows that traditional paradigms, such as vulnerability and poverty, prevail, while novel topics, such as resilience, multi-risk, or extreme events, are still rather low by comparison.
To summarize the results differently, all focal problem data sets were merged into one data set, which was then analyzed in the VOSviewer.
Figure 1 shows the results of all those combined key terms used in the detailed literature analysis above. Three classes are shown for only keywords with many occurrences in all focal problem data sets. The blue cluster is around malaria and is related to infants, newborns, children, and pregnancy. The green cluster is based on HIV infections and tuberculosis. These diseases are mainly related to gender and age characteristics, mainly to adults or aged persons. The last red cluster is related to climate change, droughts, epidemics, and diarrhea. These topics are related to the contexts of rural and urban areas, poverty, and education. The figure shows that, overall, research studies focus on certain types of diseases and population groups. While malaria seems to be more prevalent among younger persons, HIV infections are more related to adults. Studies also consider rural or urban contexts, and other vulnerability characteristics, such as education or poverty, prevail in studies covering climate change and droughts.
