This section is organized into two main parts. First, results of data analysis of drought indicators/indices (DIs) are presented. Second, the results of the data analysis of DR and DF are presented.
3.1. Drought Indices
Figure 2 presents the proportions of DIs for five different categories: meteorological, soil moisture (agricultural), hydrological, remote sensing, and composite (or modeled). The tree map of DIs (
Figure 2) clearly illustrates that the most published documents belong to the meteorological category, whereas the least published documents were reported for composite or modeled DIs. The percentages of documents (out of 11,291 publications) were 57%, 8%, 4%, 29%, and 2% related to meteorological, soil moisture, hydrological, remote sensing, and composite/modeled DIs, respectively. The most notable DIs in terms of the number of published documents were the Standardized Precipitation Index (SPI), the Normalized Difference Vegetation Index (NDVI), the Palmer Drought Severity Index (PDSI), and the Standardized Precipitation Evapotranspiration Index (SPEI), with 2483 (22%), 2023 (17.9%), 1245 (11%), and 1207 (10.7%) publications (percentages), respectively. The total number of documents on hydrological DIs was 424, which was 4% of total publications in terms of all types of DIs. There were fewer applications of hydrological DIs in Australia–Oceania, Middle-East and North Africa, and Sub-Saharan Africa than in other regions [
16]. This could be due to a lack of interest by researchers rather than an absence of drought in these regions. The most popular indicator in the hydrological category was the Streamflow Drought Index (SDI), followed by the Standardized Streamflow Index (SSFI), with 185 and 89 publications, respectively. The most popular soil moisture indicator was Soil Water Storage (SWS) which captured 680 publications across 62 different countries (
Table S1). In terms of soil moisture, Australia–Oceania was the leading region [
16]. Composite/modeled DIs, which became popular in the mid-2000s, captured the least number of publications (239 documents) compared to the other four categories. The most widely cited indices—DIs, SPI, NDVI, SPEI, SWS, PDSI, the Aridity Index (AI), the Rainfall Anomaly Index (RAI), SDI, and the Vegetation Condition Index (VCI)—were used in 85, 76, 68, 62, 57, 43, 39, 39, and 39 countries around the world, respectively. The DI studies showed an increase after the 1990s, for example, 35 out of 50 DIs were published after the 1990s.
There were several factors that affected the selection and use of DIs, such as data availability, missing data, availability of a code or program to use the index, multiple inputs for calculation, and complexity of calculation. The most complex indices/indicators were not the best ones to apply in drought studies. Visualization of published DIs in terms of the types of DIs and ease of use with their document numbers can help researchers compare them. A Sankey diagram (
Figure 3) was prepared to illustrate the most used DIs from each category and their ‘ease of use’. Details of the method of classification based on the ‘ease of use’ can be found in the WMO and GWP (2016). The top three DIs, which captured the highest number of publications, were selected from meteorological, soil moisture (2 indicators selected), hydrological, remote sensing, and composite/modeled indicators to prepare the ‘alluvial diagram’.
Figure 3 shows that nodes of ‘ease of use’ are decreasing while the difficulty level of DIs is increasing. The total number of publications was 5280, 2923, and 680 at levels 1, 2, and 3 ease of use, respectively (based on the selected indicators in
Figure 3). This result shows that the ‘ease of use’ of a drought index is one of the main criteria in choosing an index in addition to its reliability. It was noted that the simplicity of an index did not mean it was the best index to use.
The most popular DIs—SPI, PDSI, and SPEI—covered 38.5%, 19.5%, and 18.7%, respectively, of meteorological drought studies in terms of total publications. The NDVI was the most dominant index, which consisted of 60.9% of remote sensing studies.
Figure 4 presents the published documents by year for the most-used DIs based on more than 1000 publications.
The PDSI that had the longest records of meteorological DIs did not show a notable increasing trend over time. The highest number of documents were published in 2018, 2019, and 2020, being 105, 101, and 107 publications, respectively. The PDSI uses precipitation, temperature, and available water content as input data [
26].
Figure 4 shows that the PDSI was not preferred compared to other meteorological DIs due to its complexity of calculation and several other limitations, including (i) the need for serially completed data, (ii) the PDSI being slow to respond to developing and diminishing droughts, (iii) not being useful for identifying a rapid emerging drought due to its lag time, (iv) drought classifications (e.g., “extreme” or “severe”) of PDSI values varying widely from one location to another [
27,
28,
29,
30,
31,
32]. On the other hand, the popularity of SPI, SPEI, and NDVI increased over the period. The SPEI was introduced in 2010 by Vicente-Serrano et al. [
33] to overcome the limitations of current DIs. Precipitation and temperature were input data for the SPEI. One of the main advantages of the SPEI is that it is applicable in future climate models under various future scenarios. The SPEI showed the steepest increasing trend when compared to other indices (
Figure 4). The NDVI [
34] uses satellite data to identify and monitor the effects of drought on agriculture. The SPI was proposed by McKee et al. [
35] to monitor meteorological drought by using precipitation data as input. Similar to other DIs, the SPI has both advantages and disadvantages, and more details on this can be found in Mishra and Singh [
31]. An abrupt change was detected in 2010 for SPI-related studies. An increasing trend can be seen after 2010 for the SPI (
Figure 4). National meteorological and hydrological services (NMHSs) were encouraged to use the SPI to characterize meteorological drought. It was recommended by the WMO in 2009 to take the necessary steps in drought management globally using the SPI [
36]. The average number of published documents by year for the SPI, SPEI, NDVI, and PDSI was found to be 104, 101, 56, and 28, respectively.
Figure 5 illustrates the word cloud of frequently occurring keywords plus in DI studies. Keywords plus are index terms obtained by a computer algorithm based on frequent words (more than once) in the titles and the reference list of documents. The use of keywords plus helps capture a document’s content with greater variety and depth in bibliometric analysis [
37] since keywords plus comprise the majority of author keywords [
38]. Keywords plus have been widely used to identify gaps or research trends in different scientific studies [
14,
39,
40].
Bigger key word size and bold font emphasize the frequency and strength of DI sources (
Figure 5). The word cloud represents visually the most frequent words used in relevant studies and helps identify the more (or less) important ones. The most dominant keywords plus were drought (
n = 7137), climate change (
n = 2202), remote sensing (
n = 1612), soil moisture (
n = 1370), evapotranspiration (
n = 1237), vegetation (
n = 1219), China (
n = 1181), precipitation (climatology) (
n = 1046), NDVI (
n = 1038), and standardized precipitation index (
n = 941) in the DI-related studies.
Figure 6 and
Table 1 present the ranking of the top-20 countries in total citations and the top publication category for DI studies, respectively. The quantity of publications is an essential indicator to assess the development trend in a specific field [
17]. Therefore, a country’s research strength in a particular field to a certain extent can be reflected by its number of publications; however, it does not indicate the frequency and intensity of droughts in the given country. It can be seen that the USA had the highest total citations of 62,011 (
Figure 6), whereas China was the first-ranked country in terms of the top publications, with 1694 published documents (
Table 1). According to the total citations, China (
n = 28,268), Spain (
n = 13,808), India (
n = 6229), and Australia (
n = 5788) were the countries that followed the USA, respectively (
Figure 6). In terms of top publications, the USA (
n = 1372), India (
n = 393), Iran (
n = 323) and Spain (
n = 290) were recorded as the highest after China, in that order. On the other hand, average article citations exhibited different consequences. For instance, Switzerland (
= 80.5) was the first-ranked country in terms of average article citation, whereas China was placed fifteenth, with 16.69 average article citations. In accordance with average article citations, Japan (
= 57.22), Spain (
= 47.61), the USA (
= 45.20), and the United Kingdom (
= 44.47) were the countries after Switzerland, respectively. There is no doubt that the number of publications has a direct link with the average number of article citations. All in all,
Figure 6 and
Table 1 show that the USA and China were two leading countries in terms of total citations, number of publications, frequency of publications, single country publications (SCPs), and multiple country publications (MCPs).
SCPs and MCPs refer to the international collaboration of authors in the DI studies. SCP represents publications done by authors who belonged to the same country, whereas MCP represents that the publications that were written by authors belonging to different countries. Academic collaboration on an inter-country/international collaboration level is an essential element to evaluate academic development in any specific field.
The most related journals, with their Hirsch-index (H-index), g-index, number of publications, publication ratio (PR), total citations, and citations per publication (CPP), are presented in
Table 2. The H-index was developed by Hirsch [
41] to quantify and evaluate academic achievements. Simply, the H-index is a journal’s (or author’s) number of publications (h), each of which has been cited in other papers at least H times. A higher H-index shows a greater academic impact. Hence, the H-index is an important parameter to evaluate the quality and quantity of academic studies in any bibliometric analysis [
17,
42]. The g-index was introduced as an improvement of the H-index by Egghe [
43] in order to measure the global citation performance of a set of articles. Costas and Bordons [
44] noted that the g-index was more sensitive than the H-index and that these indices complement each other.
From the analysis of 1424 journals, the most productive journals, in the top 20, based on the DI studies, are shown in
Table 2. The number of publications listed in the top 20 journals was 31.7% of the total. The International Journal of Climatology was the leading journal out of 1424 journals. It had published 261 articles and accounted for 3.15% of the total publications. In terms of CPP, the Journal of Climate’s ranking was much higher than the rest of the journals. Moreover, it had the highest total citation and the second-highest g-index of 99 articles. Although Remote Sensing was the second most productive journal in terms of the number of publications, its total citations and CPP were below the average of the top 20 journals because of its late first issue (2010).
Factorial analysis of the co-occurrences of keywords plus was performed through the multiple correspondence analysis (MCA) [
45,
46,
47] to obtain a conceptual structure map [
48] of the DI studies (
Figure 7). The more similar the words in distribution, the closer to each other they are mapped in a two-dimensional space based on the relative positions of points (keywords). Three clusters were generated: Cluster 1—in green—highlights the meteorological drought, drought indices, and streamflow; Cluster 2—in blue—identifies remote-sensing-related studies; and Cluster 3—in red—highlights the other frequent keywords plus, such as water supply, climate change, water management, and climate models.
There are several science- or citation-based mapping methods, such as direct citation, bibliographic coupling, co-citation clustering, and co-citation analysis. These methods differ from each other, and the details of their relative accuracy can be found in [
49]. Moreover, there are different methods to evaluate citation relations of scientific studies. For example, Šubelj et al. [
50] compared different representative methods such as Louvain (modularity optimization method), Walktrap (random walks-dynamical process), and Infomap (map equation algorithms). The modularity optimization method (Louvain) was reported to be the fastest clustering algorithm by Šubelj et al. [
50]. Citation analysis is the most common analysis in bibliometrics, and many studies have been used in different fields for author co-citation analysis [
51,
52,
53,
54]. Therefore, in our study, author co-citation analysis was done with the Louvain clustering algorithm. Co-citation of two documents occurred when two documents were cited together in a third document.
Figure 8 presents the findings of frequently cited authors in terms of co-citation. It should be noted that only the first author’s name was considered in the analysis. We can interpret the ranking of co-citation analysis in terms of three different algorithms, which are betweenness centrality, closeness centrality, and PageRank (not shown in here). PageRank is one of the complementary methods in citation analysis, which allows us to identify publications referenced by highly cited articles [
55]. Betweenness centrality measures the number of times an author acts as a bridge or the shortest path between two other authors [
56], whereas closeness centrality measures the distance of a vertex to all others in the network [
55,
57]. Wang, Zhan, Mckee, Vicente-Serrano, and Li were found to be the top-ranking authors based on betweenness centrality. The closeness centrality outcome showed that the top 5 authors were Wang, Zhan, Li, Liu, and Chen. According to the PageRank algorithm, Mckee, Vicente-Serrano, Palmer, Dai, and Mishra were the top five authors who were referenced by highly cited publications.
Table 3 summarizes the top 20 DIs and the most applied countries. The top 15 countries were selected based on a higher number of publications, citations, and affiliations in the DI studies, whereas the DIs were selected based on the number of publications and frequency of use by countries. Countries are listed in alphabetical order. DIs were categorized as meteorological, soil moisture (agricultural), hydrological, remote sensing, and composite (or modeled). Then, the DIs were listed in alphabetical order in each category.
Table 3 presents that the USA and China are the two leading countries in the category of DI publications, whereas China is number one in terms of the use of various DIs. The Effective Drought Index (EDI) was the most used drought index in Korea, among other countries, which is the origin of the EDI. The Drought Reconnaissance Index (DRI) and SDI were the most applied indices in Iran compared to other countries.
3.2. Drought Risk and Forecast
The word cloud of frequently occurring keywords plus in the DR_DF studies is shown in
Figure 9. The most appeared keywords plus were drought (
n = 1739), risk assessment (
n = 501), climate change (
n = 461), weather forecasting (
n = 291), forecasting (
n = 261), China (
n = 190), prediction (
n = 160), soil moisture (
n = 160), and United States (
n = 154) in the DR_DF studies.
Figure 10 and
Figure 11 present the spatial distribution of the top countries in terms of the total number of publications and the SCP and MCP for the DR and DF studies, respectively. China was the most dominant country in the DR field in terms of total publications (
n = 175), SCP (
n = 139), and MCP (
n = 36), which had double the figure of the USA records. The USA (
n = 86), Australia (
n = 31), Germany (
n = 27), and the United Kingdom (
n = 25) followed China in the DR field. On the other hand, the USA was in the first rank with 192 publications and 152 SCPs, whereas China was leading in the international collaboration level with 50 MCPs on DF studies.
Greece and Mexico were not in the top 20 rankings in the DR and DF studies; however, they were in the 18th and 19th places in the DI studies. Similarly, Brazil, Portugal, and Turkey were not on the top-ranking list for DR, but they had adequate publications in the DI and DF field. South Africa was the only African country that ranked in the top 20 in the DI, DR, and DF studies.
Table 4 shows the ranking of the top 20 countries’ total and average article citations in the DR and DF studies. The highest total citation was recorded in the USA, whereas China was the second in both fields. On the other hand, the average article citation of China was lower than most of the countries in the top-ranking list. Some other countries, such as Egypt, Slovakia, and Finland, can be seen in
Table 4, yet they were not in the ranking of the top 20 in terms of the number of publications. Australia and United Kingdom showed better performance in the DR and DF fields than in the DI field. For instance, Australia was ranked in the third and fifth places based on the number of publications, and its position was fourth and third in the DR and DF fields, respectively. Moreover, the number of MCPs shows that Australia and the United Kingdom had good inter-country collaboration, followed by the USA and China.
Figure 12 and
Figure 13 display international collaborations between countries with the selected minimum productivity of publications and citations. Co-authorship analysis for a scientific collaboration has been used in several studies [
58,
59,
60,
61]. Network visualization maps illustrate the extent and strength link among countries. The size of the circles depicts the total strength of the country, and the thickness of the lines represents the strength of collaboration between any two countries.
Co-authorship analysis between countries with the full counting method [
21] shows that there were 124 links between 25 out of 90 countries in 6 clusters in the DR field (
Figure 12). The following pairs of countries were found to be having strong collaborations: USA–China (link strength = 20), USA–UK (link strength = 14), China–UK (link strength = 9), USA–Italy, and USA–France (link strength = 7). Number of co-authorships was the highest for the USA, followed by China and the UK, based on the DR studies.
Figure 13 depicts 153 links between 25 out of 98 countries in 4 clusters in the DF field. The strong collaborations were found among the following countries: USA–China (link strength = 29), USA–Australia (link strength = 10), UK–Netherlands (link strength = 10), USA–UK (link strength = 9), USA–Italy, and China–UK (link strength = 9). The number of co-authorships was the highest for the USA, followed by China and the UK, based on the DF studies.
Figure 14 presents the most relevant sources for the DR_DF studies. Results from the study disclosed the top 20 sources with the most published research articles on DR_DF-related research. Natural Hazards was ranked first, with 56 articles, 19 H-index, and 1258 total citations. The Journal of Hydrology was ranked second, with 45 articles, 23 H-index (the highest), and 1801 total citations, followed by Theoretical and Applied Climatology (38 articles, 16 H-index, 539 total citations). The Journal of Hydrology had the highest citations (1801 total citations), followed by the Journal of Hydrometeorology, with 1399 total citations.
Trend topics (TTs) on DR_DF research between 1990 and June 2021 are shown in
Figure 15. TTs were created based on the most frequent keywords plus, similar to the word cloud (
Figure 9). However, TTs reveal the most frequent words with their occurrence by time. The size of the dots represents the frequency of words, whereas the horizontal line depicts the time frame of frequency of occurrence. For instance, the term ‘New South Wales’ occurred 16 times between 2019 and 2021 (June included). Why had ‘New South Wales’ been in the TTs for the last 2 years? This was because New South Wales (NSW) has had exceptional droughts since mid-2017 [
62,
63,
64] and extraordinary wildfires in 2019/2020 as a cascading effect of drought [
65]. Therefore, DR_DF-related studies on the ‘New South Wales’ term have been trending since 2019. Furthermore, the frequency of the term ‘Australia’ was recorded 70 times between 2008 and 2019. This result clearly shows that the number of studies on DR_DF in Australia was focused after one of the worst droughts, the ‘Millennium Drought’ [
66,
67]. Similarly, other TTs allow researchers to interpret how studies have evolved over time and what trends they show.
We determined research themes that allowed for the superior interpretation of the results.
Figure 16 presents thematic maps of author keywords in the DR_DF-related studies. The
x-axis represents the centrality (relevance degree), which measures the importance of the selected theme, and the
y-axis represents the density (development degree), which measures the development of the chosen theme [
68,
69]. The graph was divided into four parts as follows: motor themes (the upper right quadrant), niche themes (the upper left quadrant), emerging or declining themes (the lower left quadrant), and basic themes (the lower right quadrant). Motor themes are well developed and important for the structuring of the research field. The cluster placed in the motor themes shows a strong relevance degree and high density. Niche themes are well developed but isolated and hence are of only marginal importance for the field. Niche themes have high density but lower centrality. Emerging or declining themes are weakly developed and marginal. The lower left quadrant of the thematic map has both low density and low centrality. Basic themes are important for a research field not yet well developed. This theme has high centrality but lower density.
We used the parameters to create the thematic map (
Figure 16), as follows: the top 430 keywords; however, the items placed in the clusters are set to the minimum cluster frequency of 3, one representative label for each cluster. Other relevant details about cluster representation, themes, and keywords in clusters are given in
Table S2 in the Supplementary section.
In total, 10 clusters were generated in the thematic map of keywords. ‘Drought’, ‘drought risk’, and ‘SPI’ clusters were placed in basic themes, and the topics under these clusters were related to climate change, risk assessment, agricultural drought, drought forecasting, and vulnerability. Some part of the ‘drought risk’ theme was in motor themes, and many parts were in the basic theme. The clusters placed in the basic or transversal theme were important for research; however, more research is required. ‘Risk management’ and ‘drought prediction’ clusters were positioned in motor themes. ‘Drought prediction’ showed higher centrality and density than ‘risk management’. Some topics under these clusters were related to hydrological drought, climate variability, water management, uncertainty, and meteorological drought. It may be noted that more research is needed in the ‘risk management’ category, with water management, drought management, and climate variability topics. ‘Drought stress’, ‘cloud computing’, and ‘Markov chain’ were placed in the niche theme. These clusters (or themes) were considered highly developed but isolated. The density of ‘cloud computing’ and ‘Markov chain’ was high, yet centrality was low. ‘Drought stress’, on the other hand, had higher centrality/lower density than ‘cloud computing’ and ‘Markov chain’. Drought tolerance, water deficit, grain yield, and wheat were related to the ‘drought stress’ theme. Emerging or declining themes involved ‘prediction’ and ‘remote sensing’ clusters. These two marginal themes had low density and low centrality, and they were weakly developed. Some topics under these clusters were modeling, land surface model, GIS, and downscaling which indicate that more research is required in this field [
20].
Table 5 presents the top-20 most productive authors on DR- and DF-related studies. The number of an academic researcher’s output can reveal the strength of their research and their effectiveness in carrying out an adequate academic study [
17]. Hence, to weigh an author’s impact on a particular field, the quantity of a researcher’s academic papers in a field can be inferred as an important index. The results showed that the top authors were from China, especially based on DR-related studies. The most productive author in the DR field was Zhang Q, who had 20 articles. 4.31 article fractionalized, and 11 H-index. Wilhite DA, from the USA, had the highest total number of citations in this field: he published 9 articles in the field and was cited 448 times. Singh VP, from the USA, was the most productive author in the area of DF with 15 articles, 700 total citations, and 12 H-index.