Introduction

Drought is one of the costliest natural disasters1 and exists on a range of spatiotemporal scales2,3,4. As a drought gets more severe and persists longer, their adverse effects are catastrophic5,6. Droughts have long been a concern for policymakers, researchers, and communities, given their capacity to disrupt livelihoods, induce food insecurity, and trigger environmental degradation7. Over 1951–2010, droughts have occurred in different regions of the world6,8,9, encompassing Africa10,11, East Asia5, Central America12, South America13, and the Mediterranean14. The consequences of a drought can differ across countries, depending on the implementation of preparation and response policies and strategies15. A simple drought index from observations and models revealed a significant increasing trend of meteorological droughts globally9,16,17. However, a complex drought index based on the Penman-Monteith (PM) equation considers the underlying physical principles including not only changes in temperature but also changes in other meteorological conditions, such as available energy, humidity, and wind speed. The PM-based drought index showed a little change in drought frequencies over the globe since 195016. With uncertainty in drought hazard estimation, state-of-the-art climate forecast models still showed a limited prediction skill of seasonal and multi-year droughts17.

Drought is a creeping phenomenon18. The onset of a drought is typically slow and is difficult for the public to be aware of its occurrence until it causes adverse impacts on natural habitats and society15. Droughts can trigger a social response beyond the affected areas by seeking and sharing information about the ongoing drought via search engine portals and social networks19,20. While the limited skill of drought forecasts makes it difficult to provide actionable information for proactive drought mitigation plans21, the connected global community via internet can improve their drought awareness in a timely manner through seeking and sharing relevant information about the ongoing drought22. The enhanced drought awareness can help raise the funds and supports from non-exposed communities over the globe, and eventually improve the resilience of the global community to drought. For effective international drought response and recovery plans, understanding of the dynamics of global drought awareness should be in multiple spatial dimensions (herein, local-, remote-, and global-level awareness), but remains limited.

Social monitoring data has been popular in recent years as a means of improving the awareness level of natural disasters and understanding possible causes of changes in social responses to natural disasters, such as floods, drought, and earthquake23,24,25,26. Social monitoring data is categorized into active and passive sources18. In the initial stages of an emerging natural disaster, active social monitoring services, such as Twitter/X and Facebook, are commonly employed for rapid risk communication; however, they can potentially capture inaccurate or misleading information via the social network7,26. In contrast, passive social monitoring systems offer observational data that can provide a more precise representation of the awareness patterns among individuals acting independently during a disaster emergency. These social monitoring data provide multidisciplinary research opportunities to assess the impact of droughts on regional and global communities, which can advance our understanding of the global challenge for water security and sustainable development in the coupled natural-social systems27.

Here we offer a multi-dimension perspective on global drought hazard and awareness across nations and continents. For drought hazard analysis, we focus on 70 countries with sufficient relevant online search activity volumes for drought awareness assessment and 63 countries with greater than 0.5 mm/day of the annual average precipitation on the 30-year average for drought hazard analysis (see Methods). We assess drought hazard based on the 12-month Standardized Precipitation Index (SPI-12), which is widely recognized as an agricultural drought index (see Methods). For drought awareness analysis, we use the Google Trends data to monitor the public awareness22 (search volume) of the global community on drought. Here, we introduce multiple spatial dimensions to assess changes in social response to drought in a comprehensive manner: local-, remote-, and global-level awareness. Local-level awareness (LA) is related to the online seeking activity volumes (herein, the awareness/interest level) of an exposed nation to a drought. Remote-level awareness (RA) is measured by the online seeking activity volumes of a country of interest relevant to droughts in other countries with no occurrence of local droughts. Global-level awareness (GA) is defined as the online seeking activities of global community, excluding the exposed nation to a local drought. The ratio of RA to LA (RA/LA) is computed to compare relevant search activity volumes to remote droughts with those to local droughts. The ratios of GA to LA (GA/LA) and RA/LA are also calculated to compare the level of the global community’s interest in a drought with that of the public’s interest of the affected nation. We conduct a multi-dimensional analysis of global drought awareness by categorizing the local-, remote-, and global-level awareness on drought (see Methods). This multi-dimensional analysis of global drought awareness will unveil the dynamics of drought-related online search activities of the global community by answering the following questions: (a) how does the spatiotemporal variability of drought events correlate with changes in drought awareness of the global community? and (b) what are potential factors that influence the local-, remote-, and global-level awareness of the global community on drought? By amalgamating scientific insights with observational and passive monitoring of drought, this study will envision a more comprehensive and proactive approach to addressing the multifaceted challenges posed by drought.

Results

Global drought hazard in the recent decade

Since 2010, moderate (D1) and exceptional (D4) droughts have affected about 20% (less than 5%) of the 70 nations (Fig. 1). Over the globe, the areal extent of droughts has fluctuated in two to three-year intervals. In mid and late 2010s, a majority of the 70 nations have experienced at least moderate droughts.

Fig. 1: Characteristics of global drought events.
figure 1

a Areal fraction under drought conditions of the 63 countries. b Drought Intensity from 2010 through 2021 over 63 countries globally. c Drought duration in months from 2010 to 2021 over 63 countries globally. D1, D2, D3, and D4 are the drought categories having SPI12 ≤ −0.5, −1, −1.5, and −2 respectively.

Full size image

This study compared the characteristics of recent droughts (2010–2021) with those of historical droughts (1901–2009) to assess the changes in drought hazard at the continental level (Fig. 2 and Supplementary Fig. 1). The long-lasting drought persistence is observed across North America, Europe, Asia, and Oceania. In Asia, a subtle alteration in the duration of drought events is attributed to the long-lasting Kazakhstan drought condition in the recent decade. Two severe droughts with a longer duration are observed in Africa since 2010. This result underscores a possible temporal shift to a long drought duration of moderate and severe droughts in Africa.

Fig. 2: Continental temporal scale of drought patterns.
figure 2

The blue shade represents the Intensity vs Duration data from 1901 through 2009, and the red dots represent the Intensity vs Duration data from 2010 through 2021.

Full size image

Over 2010–2021, 32 nations experienced extreme drought conditions (−2 of SPI12 or below). Brazil, Chile, Kazakhstan, Colombia, Turkey, and Israel reported an alarming drought intensity at −3 of SPI12 or below. In contrast, seven nations, Canada, Kenya, Norway, Puerto Rico, Russia, the Philippines, and the United States, exhibited moderate drought conditions (−1 of SPI12 or above). Russia exhibited an absence of meteorological drought conditions from 2010 to 2021, yielding no drought occurrence during our study period.

Chile and Kazakhstan endured long-lasting drought conditions in the last decade. Chile demonstrated a comparable drought intensity to those in Iran and Brazil, but Kazakhstan exhibited comparatively lower intensity. Furthermore, China manifested the drought intensity with a relatively short duration, indicating a possible change in the drought characteristics, such as flash droughts with a rapid onset28. South Africa endured a long-lasting drought period, but with a low level of drought intensity. It is worth noting that we conducted the national-level drought hazard based on the national average precipitation-based drought index, which might be improper for a regional-scale drought hazard assessment, such as the western U.S. and South Africa’s Cape Town droughts in the recent decade.

In the last decade, Chile, Kazakhstan, South Africa, Bangladesh, Iran, Venezuela, El Salvador, Peru, and Brazil exposed a high drought hazard. They have endured a prolonged drought (D1 to D4) period exceeding nine years from 2010 to 2021. Out of the 70 countries, 22 countries experienced drought periods below four years and 32 countries fell within the intermediate range (four to eight years) of drought duration. This comprehensive assessment indicates the varying degrees of drought severity across the globe, emphasizing the need to understand multi-dimensional awareness of the global community on drought.

Global drought awareness in the recent decade

The Google Trends data showed a persistent increase in internet search activities of the global community about drought in the recent decade (Fig. 3). The multi-dimensional analysis of global drought awareness demonstrated a less-than-one ratio of RA/LA and a greater-than-one ratio of the level of GA/LA in a majority of the 70 countries, indicating that these countries are more interested in the hazard of local droughts than that of remote droughts, but the local droughts in these countries were aware by the global community. 27 countries (38%) showed a pronounced prevalence of local awareness exceeding global awareness (the less-than-one GA/LA ratio). 16 out of these 27 countries (23%), the level of local awareness surpassed both the levels of remote awareness, indicating a relatively high level of local drought awareness compared with the awareness of remote drought conditions. It is worth noting that national-level internet search activity data from Google Trends might miss some of individuals who might be interested in the local/remote drought conditions, but not search them on Google due to internet access or fulfilling their basic needs.

Fig. 3: Spatiotemporal patterns of drought awareness in multi-dimensions.
figure 3

a Monthly relative online search activity volumes from Google Trends worldwide. A solid black line in (a) denotes the median of monthly relative online search activity volumes over 2010–2021. b A schematic diagram for local, remote, and global drought awareness. c Ratios of remote to local drought awareness. d Ratios of global to local awareness. Red hatch lines in (c) and (d) depict the countries with significant uncertainty in the Google trends data (see Methods).

Full size image

We found diversity in social response to past droughts in the LA-RA-GA dimensions. Chile, Kazakhstan, and Russia showed a comparable level of local and remote drought awareness. Chile and Kazakhstan have experienced a long-lasting drought, but Russia has experienced no drought occurrence over the study period. These results indicated that drought hazard is not the sole driver of interactions between local and remote drought awareness. Saudi Arabia, the United Arab Emirates, Algeria, Morocco, and Egypt characterized as deserts (<0.5 mm/day on the 10-year average) showed that the ratio is equal to one, reflecting equivalence in the level of local and remote drought awareness. 19 countries exhibited a greater-than-one ratio of remote drought awareness to local drought awareness, which included seven European (Sweden, the Czech Republic, Austria, Romania, Ireland, Finland, and Norway), five American (the United States of America, Puerto Rico, Costa Rica, Bolivia, and Ecuador), four Asian (India, Pakistan, Vietnam, Hong Kong), and three African (Nigeria, Mozambique, Zimbabwe) countries. These countries were a potential key community in monitoring the global drought conditions and disseminating the relevant information, such as sharing the news, blogs, and pictures from mass and social media.

Multi-dimensional impacts of GDP and drought duration on drought awareness

The impact of drought severity and Gross Domestic Product (GDP) per capita on global drought awareness were examined in the local-remote-global awareness dimensions (Fig. 4). Our results indicated that remote awareness predominated in high-GDP countries. The countries that experienced severe drought showed an elevated level of both local and global awareness. Specifically, 50% of the 70 countries surpassed the median remote level (dotted lines in Fig. 4), with a notable contribution from Europe (21%), and followed by Asia (14%), North America (7%), Africa (3%), South America (3%), and Oceania (1%). 47% of the 70 countries exceeded the median global awareness level, with noteworthy contributions from Asia (14.28%), Europe (10%), Africa (8.57%), South America (8.57%).

Fig. 4: Roles of drought severity and GDP on drought awareness in multi-dimensions.
figure 4

Marker size in (a) and (b) indicates the GDP of a country and the duration of the longest drought events over 2012–2021. c Global distribution of the six-category countries. White inclined lines indicate countries with inconsistency among the 10 Google Trends data (see Methods).

Full size image

Furthermore, 54% of countries exceeded the median local awareness level, with contributions from Asia (22.8%), Europe (10%), Africa (8.57%), South America (8.57%), and North America (4.28%). These results underscore a heightened remote awareness in Europe, which was attributed to a combination of low drought severity and high GDP, indicating Europe’s concern for drought conditions in not only their countries but also other neighbouring countries. We also observed a similar role of drought duration in the LA-RA-GA relationship, indicating that a more severe drought heightens local and global awareness (Supplementary Fig. 2).

Discussion

Our multi-dimensional analysis showed that drought awareness of the global community is not always contingent upon the drought hazard and characteristics. In the recent decade, certain countries like Kazakhstan, South Africa, and Chile experienced frequent and severe droughts yet exhibit comparatively a low level of both local and remote awareness. Other countries with limited drought occurrences demonstrated heightened remote awareness. Our investigation succeeded in revealing instances where the level of remote awareness surpasses that of local awareness. This discrepancy does not imply a lack of concern among the populace for drought within their respective countries; rather, it underscores the influence of short-lived drought periods on online information search behaviour of the global community.

Individuals worldwide express varying degrees of concern about drought, irrespective of its occurrence in their nations or localities, leading to a discernible pattern of elevated global awareness compared to local awareness. Netherlands, Denmark, and Belgium are less susceptible to drought since they have a low level of local and remote awareness. We speculate that a low drought hazard can lead to a low level of awareness and interest of local and remote droughts at the national level. However, the Google Trends data used in this study are at the aggregate-level awareness, indicating a limited explanatory power of actual causes of a low level of drought awareness in these countries. To investigate the actual cause of a low-level of drought awareness, other data, such as survey and interview data, are required.

This study used relevant search activity volumes to drought (a search topic of disaster type) as a proxy of national-level drought awareness. To avoid irrelevant search activity volumes, this study focused on 70 nations with a high search activity volume. While this study used “drought” as a search topic, using other relevant search terms, such as “water storage” and “precipitation deficits”, can be used to explore how the public’s interest and received risk are changed over drought propagations.

Wikipedia and newspapers might be other data source to examine how online information seeking patterns changes during an emerging drought. Daily views of the Wikipedia page have been compared with the Google Trends data26. Daily views of the “Drought” page of Wikipedia were retrieved. Wikipedia has the “Drought” page in 103 languages since July 2015. Daily views of the “Drought” page are available in 87 languages. For temporal resolution matching with the Google Trends data, first the daily views of 87 languages of “Drought” were summed and then the total daily views were averaged over each month. Lastly, the monthly averages of the total daily views were divided by the maximum view from July 2015 through December 2021. The correlation of the daily views and Google Trends showed a low correlation (r = 0.15; Supplementary Fig. 3). This is different from the findings of global earthquake awareness that showed a high correlation between the Google Trends and daily views of the Wikipedia “Earthquake” page. This result suggested diversity in online seeking information activity patterns depending on the characteristics of a disaster of interest.

In this study, relevant search activity volume data were downloaded 10 times on different dates and applied the principal component analysis to the 10 sample data to assess consistency among the data. 18 nations showed inconsistency among the 10 sample data, indicating high uncertainty in the search activity volume data (hatched in Fig. 4c).

The findings of this study were based on the aggregate-level drought awareness, which might have a limited explanatory power of the cause-effect processes between drought hazard and awareness.

Individual- and aggregate-level drought awareness have different patterns due to various socioeconomic factors. Survey and interview data can provide an opportunity to investigate how individuals perceive the risk of local and remote droughts, which can synthesize the findings of this study to advance our understanding of interactions between individual- and aggregative-level awareness. These survey and interview data however are not available over the globe and in real time/continuously because they are costly and time consuming. The findings of this study from the national-level search activity volume data highlight a need for development of the global survey and interview data for drought awareness.

This study proposed a multiple spatial dimensional analysis of drought awareness, introducing local, remote, and global-level awareness. This multi-dimension analysis focused on country-level awareness, which can affect drought awareness at multiple levels. When the averages of precipitation over a large country like Canada, China, Russia, and U.S.A, are used to calculate SPI, precipitation deficits should be severe enough to detect country-level droughts. If droughts are extensive and long-lasting, then people are more likely to seek search relevant information to the ongoing drought. Besides, the Google Trends data provide relative search activity volumes to the maximum search activity volume during the search period (herein 2010–2021). During the onset of a local drought, relevant search activity volumes would be relatively small. Therefore, using the country-level SPI and Google Trends for larger countries guarantees the severity of drought and the significance of relevant search activity volumes, which can reduce uncertainty in measuring the level of local awareness.

On the other hand, it is expected that there are less droughts in larger countries than smaller countries, simply because intrinsic (spatial) variation of precipitation within a large country. For example, if a small region within a larger country experiences a local drought, the proposed methodology would not identify any drought at a national level. Then, any searches within that region would be classified as remote awareness for the country when they experience the local drought. The proposed methods can cause uncertainties of remote awareness estimates, particularly large countries. We further compared relevant search activity volumes during drought with those during non-drought periods, that is, before and after drought events (Supplementary Fig. 4). We found those search activity volumes before and after drought events are relatively low compared with those drought events, indicating that the proposed methods are still a reliable approach to conduct a multi-dimensional study of local and remote awareness.

The comparison between the characteristics of historical droughts and recent occurrences reveals notable similarities at the continental scale except for South America. Despite a high drought hazard, the African countries exhibit comparatively low drought awareness levels. It is worth noting that six African nations (Algeria, Egypt, Ethiopia, Mozambique, Nigeria, and Zimbabwe) showed inconsistency among the 10 sample data, indicating that additional data, such as survey and interview data, are required for these countries, to determine whether the actual level of drought awareness is low or the actual awareness is high but other socioeconomic factors, such as the limited internet access and educational opportunities, cause a low search activity volume. Our analysis unveiled a lack of a robust correlation between drought vulnerability and awareness, suggesting the influence of additional socioeconomic factors on shaping public awareness. This insight highlights the complexity of the relationship between drought events and awareness levels, urging further exploration of these factors for a more comprehensive understanding.

Our investigation further revealed that countries like Greece, Malaysia, and El Salvador with low susceptibility to drought exhibit a noteworthy interest in searching drought-related queries on the Google platform. This observation aligns seamlessly with existing literature that delves into the impact of a “no-loss” experience, where even nations with minimal drought occurrences like Russia demonstrate heightened awareness and concern. Since 1901, precipitation has increased over Russia (Supplementary Fig. 5).

We observed that after the drought recovery, the search activity volumes about drought decreased but maintained relatively a certain level, contributing to sustained global interest in drought-related topics and resulting in a higher level of remote awareness than before the drought occurrence. This intriguing temporal pattern emerges, indicating that awareness tends to elevate notably in the aftermath of a 3- to 6-month drought recovery period. This temporal lag contributes to the phenomenon wherein certain vulnerable countries display higher remote awareness than local awareness, as information-seeking behaviour intensifies after this recovery interval when the adverse effects are measurable.

Our findings offered valuable insights for crafting targeted interventions in countries characterized by both severe drought conditions and low public awareness. By leveraging our results, policymakers can formulate effective strategies to mitigate the adverse impacts of an emerging drought. Firstly, we recommend that the global community should build multiple channels that can report and share an emerging drought. Citizen participatory surveillance systems are costly efficient and reliable, particularly over nations with a low level of local awareness, particularly in the Asian and African countries. These citizen participatory surveillance system can calibrate the public’s perceived risk of the ongoing drought efficiently. Secondly, the global community should revitalize international exchanges programs of researchers and administrative staffs between the European countries and countries in other continents. We found that most of the European countries had a balance between local and remote awareness, indicating their continuing role in enhancing global drought awareness. Lastly, transdisciplinary research efforts should be encouraged to advance our understanding of social response to drought across nations and investigate new drought risk management plans are available with artificial intelligence and big data. These policy recommendations are crucial for sustainable development in a changing climate.

This study not only provides empirical evidence of global disparities in public interest regarding drought but also underscores the significant role developed nations play in shaping information-seeking behaviour on a global scale, drawing from diverse data sources. The latter discovery regarding the influence of developed nations can be attributed to long-standing conditions stemming from socioeconomic development. In societies with less economic advancement, immediate basic needs often take precedence, constraining attention to knowledge-seeking behaviours26. Conversely, residents of more prosperous nations possess greater opportunities for self-actualization through knowledge pursuit. Nonetheless, our results unveil nuances within developing countries, demonstrating that the social response to drought hazard varies among nations with lower socio-economic development.

Methods

Google Trends (GT) data

The Google Trends platform provides relative search activity volumes relevant to a term/topic of interest since 2004. A keyword can be chosen as a search term or search topic. The former option provides relevant search activity volumes including only the exact word (herein, “drought”) and the latter option provides relevant search activity volumes to a search topic in not only a certain language but also other languages19. This study used “drought” as a keyword with the search topic option (e.g., the disaster type of “drought” in not only English but also other languages). Depending on the length of the search period, the temporal resolution of relevant search activity volumes is determined whether it is hourly (highest), daily, weekly, or monthly (the coarsest). The spatial resolution can be chosen by users. It is worth noting that the Google Trends data are reliable when the search activity volume is sufficient. In this study, nations with a low search volume were excluded and the start and end year was chosen as 2010 and 2021, respectively.

We also used other keywords, such as “water scarcity” and “water shortage”, to evaluate the consistency of relevant search activity volumes with the keyword, “drought”. We found that the Google Trends platform didn’t provide a search topic option for “water shortage” and “precipitation deficit” mainly due to a relatively low volume of search activities. Singapore showed 0.33 of Pearson correlation coefficient of search activity volumes between “drought” and “water shortage. We found that search activity volume related to “drought” are dominant but some countries, such as Japan (Pearson correlation coefficient = 0.107), Mexico (0.542), Pakistan (0.086), Peru (0.73), and Vietnam (−0.186), showed comparable activity volumes related to “drought” and “water scarcity.” (Supplementary Table 1). Mexico and Peru experienced a severe drought in the recent years, which lead to a high temporal correlation coefficient10.

This study extracted monthly relative search activity volumes of 70 nations from January 2010 to December 2021. The relative search activity index is calculated by dividing the total number of searches within a month by the maximum number during the selected period, with values ranging from 0 to 100. To assess uncertainties in the Google Trends data, first we retrieved relative search activity volumes of the 70 nations on ten different dates using the same search topic and period. Then, we applied the principal component analysis to the 10 Google Trends sample data for each country. The first principal component mode is used to examine the consistency/uncertainty with the 10 sample data (Supplementary Fig. 6). 52 out of 70 countries showed that their county-level Google Trend data are consistent to each other (less uncertain); however, the remaining 18 countries (hatched area in Fig. 3; Supplementary Figs. 7 and 8) showed inconsistency among the 10 sample data, which means that relative search activity volume data are sensitive to the retrieval date. These 18 countries include Egypt, Dominican Republic, Hungary, Bolivia, El Salvador, UAE, Algeria, Ecuador, Guatemala, Ethiopia, Greece, Mozambique, Sri Lanka, Pakistan, Nigeria, Bangladesh, Kazakhstan, and Zimbabwe.

Global precipitation data

This study utilized the country-level spatial averages of Climate Research Unit version 4.06 monthly precipitation data over 1901–202129. For validation, the spatial averages of the Global Precipitation Climatology Centre (GPCC) monthly precipitation data were calculated for the selected 70 countries over 1901–2019, using the Global Administrative Areas (GADM) version 3.6 (https://gadm.org/download_country_v3.html). For cross-validation, a correlation analysis of the country-level CRU and GPCC spatial averages was conducted (Supplementary Fig. 9). All countries show a positive correlation between the CRU and GPCC spatial averages. Puerto Rico showed the lowest correlation coefficient (0.377). It is worth noting that six countries in Middle East and Africa (Algeria, Egypt, Morocco, Mozambique, Saudi Arabia, and U.A.E.) were excluded because they have annual precipitation was 0.5 mm/day or below on the 30-year average. This low precipitation threshold value have often used to determine desert regions30.

Identification and characterization of drought events

In this study, the 12-month normalized precipitation index (SPI12) was calculated for drought detection and characterization, following the method10. The threshold level approach is employed to ascertain the onset and recovery of a drought event1,1322,23. In this study, a drought event is detected when the SPI12 value is −0.5 or below for drought onset and the SPI12 value is 0.5 or above for drought recovery with at least a three-month drought duration (Supplementary Fig. 10). The drought duration can be measured from the onset month through the recovery month13. Drought intensity is defined as the minimum of the SPI12 values during the drought duration21,30. Based on the intensity, a drought event is classified into four levels: moderate (D1;−0.5 of SPI12 or below and −1.0 of SPI12 or above), severe (D2; below −1.0 of SPI12 and −1.5 of SPI12 or above), extreme (D3; below −1.5 of SPI12 and −2.0 of SPI12 or above), and exceptional drought (D4; below −2.0 of SPI12). The number of months from the onset to the recovery month can be defined as drought duration. We applied this method to calculate the drought events using historical monthly average precipitation data from 1901 to 2021 for 63 nations. Exclusion of seven countries from SPI analysis due to consistently low monthly precipitation (<0.5 mm per day on the 10-year average).