Disparity between global drought hazard and awareness

The paper addresses how public awareness of drought relates to actual drought hazard across countries and continents. Droughts are costly, occur on multiple scales, and their impacts depend on preparedness and response policies. Prior work shows mixed global trends in drought depending on the index used, with limited forecast skill for seasonal to multi-year droughts. Because droughts develop slowly and their onset is hard for the public to detect, online information seeking and social networks can shape awareness beyond affected areas. The study aims to provide a comprehensive, multi-dimensional perspective on global drought awareness—defined at local, remote, and global levels—using online search activity as a proxy. It asks: (a) how does the spatiotemporal variability of drought events correlate with changes in global community awareness, and (b) what factors influence local-, remote-, and global-level awareness of drought? The purpose is to inform more effective international drought response and resilience-building by understanding awareness dynamics.

The authors synthesize findings on historical drought occurrence across regions (Africa, East Asia, Central and South America, Mediterranean), noting that simple indices suggest increasing global meteorological droughts, whereas physically based PM-derived indices show little change since 1950. Forecast models have limited skill for seasonal to multi-year droughts. Drought’s creeping onset complicates public awareness, but information seeking and sharing via search engines and social networks can expand awareness beyond affected areas. Social monitoring data have been used to study responses to natural hazards; active sources (e.g., Twitter/Facebook) enable rapid communication but may propagate misinformation, while passive sources (e.g., search data) reflect independent information-seeking behavior and can better capture awareness patterns. These insights motivate leveraging Google Trends and complementary sources (Wikipedia, newspapers) to study drought awareness dynamics.

Scope and datasets: The study examines 70 countries with sufficient online search activity for awareness analysis and 63 countries with annual average precipitation >0.5 mm/day (30-year average) for drought hazard analysis. Drought hazard is derived from country-level monthly precipitation (CRU TS v4.06, 1901–2021), with cross-validation using GPCC (1901–2019). Country boundaries from GADM v3.6. Awareness is measured using Google Trends (GT) monthly relative search activity for the search topic “drought” (multi-language topic selection), January 2010–December 2021.

Drought detection and characterization: The 12-month Standardized Precipitation Index (SPI-12) is computed from precipitation. Drought onset is SPI-12 ≤ –0.5, recovery is SPI-12 ≥ 0.5, with minimum duration of 3 months. Intensity is the minimum SPI-12 during the event. Categories: D1 (–0.5 to ≥ –1.0), D2 (< –1.0 to ≥ –1.5), D3 (< –1.5 to ≥ –2.0), D4 (< –2.0). Event duration is months from onset to recovery.

Awareness dimensions and metrics: Local awareness (LA) is the relative online search activity within an exposed nation when it is experiencing a local drought. Remote awareness (RA) is the search activity within a country during droughts occurring in other countries when the focal country has no local drought. Global awareness (GA) is the worldwide search activity excluding the exposed nation during its local drought. Ratios computed include RA/LA and GA/LA to compare relative interest in remote and global awareness versus local awareness. A multi-dimensional classification of countries is performed based on LA, RA, and GA levels and their ratios.

Google Trends processing and uncertainty: GT provides a monthly index scaled 0–100 by month’s searches divided by the period maximum. To assess retrieval-date sensitivity, the authors downloaded data 10 times per country (same topic/period) and applied principal component analysis across the 10 samples. Consistency is inferred from the first principal component; 52/70 countries showed consistent series, while 18 showed inconsistency/high uncertainty. Additional keyword checks (“water scarcity,” “water shortage”) indicated lower availability or mixed correlations with “drought,” supporting the use of the “drought” topic.

Analytical comparisons: The study compares recent (2010–2021) versus historical (1901–2009) drought characteristics (intensity-duration) at continental scales. It evaluates awareness changes over time (2010–2021), RA/LA and GA/LA distributions, and explores associations with drought duration/severity and GDP per capita (marker sizes/colors in figures convey GDP and longest drought duration). Additional checks compare search volumes during droughts versus pre/post (non-drought) periods to validate signal strength.

Exclusions and caveats: Desert or very arid countries (annual precipitation ≤0.5 mm/day) are excluded from SPI analysis. Country-scale SPI and GT focus on national averages, which can under-detect regional droughts in large countries; implications for classification of searches as remote versus local are discussed.

• Global drought hazard vs awareness: Despite little overall change in global drought hazard during 2010–2021 (consistent with PM-based findings), global online search activity about drought increased, indicating elevated global awareness.
• Drought characteristics: Since 2010, about 20% of the 70 nations experienced moderate drought (D1) at a given time and <5% experienced exceptional drought (D4). Areal extent fluctuated on 2–3-year intervals. Long-lasting drought persistence was observed in North America, Europe, Asia, and Oceania; Africa showed a shift toward longer-duration moderate/severe droughts. Over 2010–2021, 32 nations experienced extreme drought (SPI-12 ≤ –2). Brazil, Chile, Kazakhstan, Colombia, Turkey, and Israel reached intensities ≤ –3. Seven nations (Canada, Kenya, Norway, Puerto Rico, Russia, the Philippines, United States) exhibited only moderate conditions (SPI-12 ≥ –1); Russia had no meteorological drought.
• Multi-dimensional awareness patterns: Majority of countries had RA/LA < 1 and GA/LA > 1, meaning people are more interested in local droughts than remote ones, while local droughts garner substantial attention globally. 27 countries (38%) had GA/LA < 1 (local > global awareness); in 16 of these (23% of all), local awareness exceeded both remote and global.
• Countries with RA > LA: 19 countries showed RA/LA > 1, including seven European (Sweden, Czech Republic, Austria, Romania, Ireland, Finland, Norway), five American (USA, Puerto Rico, Costa Rica, Bolivia, Ecuador), four Asian (India, Pakistan, Vietnam, Hong Kong), and three African (Nigeria, Mozambique, Zimbabwe), suggesting key roles in monitoring and disseminating drought information globally.
• GDP and duration effects: Remote awareness predominated in high-GDP countries. Severe/longer droughts elevated both local and global awareness. Proportions exceeding median awareness levels: RA—50% of countries (notably Europe 21%, Asia 14%, North America 7%, Africa 3%, South America 3%, Oceania 1%); GA—47% (Asia ~14.3%, Europe 10%, Africa ~8.6%, South America ~8.6%); LA—54% (Asia ~22.8%, Europe 10%, Africa ~8.6%, South America ~8.6%, North America ~4.3%). Europe exhibited heightened remote awareness despite lower drought severity, consistent with high GDP and interest in neighboring countries’ droughts.
• Regional contrasts: Despite high drought hazard, many African countries showed comparatively low awareness; several had GT data inconsistencies. Chile and Kazakhstan had long-lasting droughts but not uniformly high awareness; Russia had no drought yet showed comparable local and remote awareness, indicating drivers beyond hazard alone.
• Cross-platform signal: Wikipedia “Drought” page views (2015–2021) had low correlation with GT (r = 0.15), contrasting with earthquake studies, implying hazard-specific differences in information-seeking behavior.

Findings indicate drought awareness is not solely determined by drought occurrence or severity. Some countries with frequent and severe droughts (e.g., Kazakhstan, South Africa, Chile) exhibited relatively low local/remote awareness, while countries with limited droughts showed elevated remote awareness. Short-lived droughts can suppress local search interest, contributing to cases where RA > LA. Global awareness often exceeds local awareness, reflecting broad concern irrespective of local exposure. Low-hazard countries (e.g., Netherlands, Denmark, Belgium) showed low local and remote awareness, though aggregate GT data cannot reveal individual-level causes; socioeconomic factors (internet access, education, competing basic needs) likely play roles. Post-drought dynamics show a lagged increase in remote awareness 3–6 months after recovery, potentially explaining periods when vulnerable countries display higher RA than LA. Europe’s balance between local and remote awareness suggests a continuing role in enhancing global drought awareness. Policy implications include establishing multi-channel reporting and citizen participatory surveillance, fostering international exchanges (particularly with European institutions) to disseminate best practices, and promoting transdisciplinary research leveraging AI and big data to improve drought risk management and communication.

The study introduces and applies a multi-dimensional framework for assessing drought awareness—local, remote, and global—using Google Trends as a passive, observational proxy and SPI-12 for drought hazard. It provides empirical evidence of global disparities between hazard and awareness, documents increasing global interest in drought despite little recent change in global hazard, and highlights the influence of drought duration and socioeconomic factors (notably GDP) on awareness patterns. European countries appear pivotal in maintaining balanced local and remote awareness and in amplifying global attention to droughts. Future work should integrate global, real-time survey and interview data to link individual and aggregate awareness, explore additional search terms and media sources to capture different drought phases and impacts, refine methods for large countries and subnational scales, and further investigate socioeconomic drivers to inform targeted, equitable drought communication and resilience strategies.

• Awareness proxy: Google Trends provides aggregate, relative search volumes and cannot identify individual motivations or causality; limited explanatory power for cause–effect processes between hazard and awareness.
• Data reliability: GT series can vary by retrieval date; 18 of 70 countries showed inconsistency across 10 downloads (higher uncertainty). Countries with low search volumes were excluded, potentially biasing coverage.
• Scale effects: Country-level SPI and GT can miss regional droughts in large countries; local events may be underdetected, causing some within-country searches during regional droughts to be classified as remote.
• Climatic exclusions: Very arid/desert countries (annual precipitation ≤0.5 mm/day) were excluded from SPI analysis, limiting generalizability to those regions.
• Socioeconomic constraints: Internet access, education, and competing basic needs can suppress search activity, especially in developing regions (e.g., parts of Africa), confounding awareness estimates.
• Cross-platform generalization: Low correlation between GT and Wikipedia views (r = 0.15) suggests platform-specific behaviors; awareness inferences may not generalize across media.