Surging compound drought–heatwaves underrated in global soils

Significance

The past decade, particularly the last year, has seen unprecedented heat, threatening ecosystems sustained by soils. Evaluating the impact of droughts and heatwaves on global soils using meteorological variables such as precipitation and air temperature may be inadequate, given the significant differences between soil and air temperatures. We introduce the concept of soil compound drought–heatwaves (SCDHWs) to provide a more objective assessment of concurrent water and heat stresses in global soils. Our analysis reveals that the intensity and growth rate of global SCDHW events are substantially greater and accelerate more quickly than previously estimated from a meteorological perspective. We show that the Southern Hemisphere has been undergoing long-lasting SCDHWs, while northern high latitudes face more intensive SCDHWs.

Abstract

Compound drought–heatwaves (CDHWs) accelerate the warming and drying of soils, triggering soil compound drought–heatwaves (SCDHWs) that jeopardize the health of soil ecosystems. Nevertheless, the behavior of these events worldwide and their responses to climatic warming are underexplored. Here, we show a global escalation in the frequency, duration, peak intensity, and severity of SCDHWs, as well as an increase in affected land area, from 1980 to 2023. The increasing trends, which are particularly prominent since the early 2000 s, and projected to persist throughout this century, are dominated by summertime SCDHWs and enhanced by El Niño. Intensive soil warming as well as climatologically lower soil temperatures compared to air temperatures lead to localized hotspots of escalating SCDHW severity in northern high latitudes, while prolonged duration causes such hotspots in northern South America. Transformation of natural ecosystems, particularly forests and wetlands, to cropland as well as forest degradation substantially enhance the strength of SCDHWs. Global SCDHWs consistently exhibit higher frequencies, longer durations, greater severities, and faster growth rates than CDHWs in all aspects from 1980 to 2023. They are undergoing a critical transition, with droughts replacing heatwaves as the primary constraint. We emphasize the significant intensification of SCDHWs in northern high latitudes as well as the prolonged duration of SCDHWs in the Southern Hemisphere, posing an underrated threat to achieving carbon neutrality and food security goals.

Compound drought–heatwaves (CDHWs), the simultaneous occurrences of water scarcity (in most cases precipitation deficiency) and elevated temperatures, have gained significant attention (1–4), particularly since the mid-2010s. An increasing body of evidence now indicates a widespread increase in the occurrence frequency, duration, intensity, and severity of global CDHWs, which can be attributed to anthropogenic climate warming (5–7). The occurrences of unprecedented CDHWs in Southern China (8) and Europe (9) in 2022 and in South America in 2023 (10) have significantly heightened public awareness regarding their devastating impacts. Soil water deficit during CDHWs drives the allocation of more energy toward sensible heat, resulting in increased atmospheric, land surface, and soil temperatures (11). As a pivotal link in the soil moisture–atmosphere feedback (12, 13), elevated soil temperatures can exert profound impacts on terrestrial ecosystems, triggering vegetation degradation (14), crop failure (15), tree mortality (16), and forest fires (17).

Heatwaves have emerged as a global concern in recent decades, characterized by the occurrence of excessive heat in both land surfaces and the atmosphere (18, 19). Recently, this concept has been extended to encompass diverse geographic entities that are exposed to climate warming, including the oceans (20), glaciers (21), and lakes (22). While atmospheric heat gain has increased significantly in recent decades, global land (excluding the cryosphere) has absorbed approximately six times more heat from 1971 to 2020, owing to its higher heat capacity (23). This disparity underscores the need to define and investigate soil heatwaves or soil heat extremes (12). This definition is more relevant than the standard definition of heatwaves for assessing climate change impacts and risks to the soil carbon budget, particularly concerning soil respiration, which is heavily influenced by variations in soil temperature and moisture content (24, 25).

The impact of temperature changes on the soil carbon budget is commonly assessed in situ using soil temperature measurements, while global assessments typically rely on air temperature due to its broader availability. The differences between air and soil temperatures (26) can lead to inaccurate estimates of the global soil carbon budget, particularly in high-latitude regions (27) and during extreme heatwaves (12). In northern high-latitude regions, where vast amounts of soil carbon are stored, temperature fluctuations have significant effects. During cold seasons, snow and ice can hinder the transfer of sensible heat from the atmosphere to the soil, resulting in slower and delayed soil warming (27). Conversely, in nonfreezing seasons, increasing heatwaves are leading to more frequent and intense soil temperature spikes (12).

Recent years have seen extensive research on soil moisture droughts (28, 29), supported by long-term satellite data and model simulations (30). However, meteorological drought indices, which are typically centered on precipitation, remain prevalent for defining compound events. These indices, such as the standardized precipitation index (SPI), the standardized precipitation and evapotranspiration index (SPEI), and the self-calibrated palmer drought index (sc-PDSI) (4, 9, 31), often focus on precipitation deficits and air temperature extremes, overlooking critical soil conditions. Additionally, current CDHW definitions are based on weekly-to-6-mo timescales, whereas flash droughts can develop within weeks (32, 33). This highlights the urgent need for more frequent monitoring of soil moisture droughts and compound events, facilitated by advanced reanalysis data (34, 35). Addressing these issues necessitates the development of a precise definition for CDHWs that specifically incorporates soil water and heat stress.

Here, we propose the concept of soil compound drought-heatwave (SCDHW), defined using daily measurements of soil moisture and soil temperature, as opposed to CDHW, which relies on daily soil moisture and air temperature measurements. By using soil moisture instead of precipitation to define CDHW, we facilitate a fair comparison with SCDHW. Our study first examines spatiotemporal changes in global SCDHWs, their dependence on soil cover, and contributing factors from 1980 to 2023 using a multimodel ensemble. We further investigate the differences between SCDHWs and CDHWs at both global and regional scales to establish SCDHW as a distinct metric. Our findings reveal a persistent and intensifying trend in SCDHWs over recent decades and extending into the future, which are often underestimated when using the CDHW framework. These results highlight significant risks for global regions vulnerable to CDHWs, underscoring severe threats to carbon neutrality and food security goals.

Results

Crisis in the Soil: Escalating CDHWs.

We observe a progressive increase in the occurrence frequency, duration, peak intensity, and severity of global SCDHWs from 1980 to 2023, particularly since the beginning of this century (Fig. 1). The strongest ever global SCDHW occurrences in our record are observed in 2023. In recent years, SCDHWs have been occurring more than twice as often and lasting over three weeks on average. These events are characterized by a peak intensity exceeding 4 °C and a severity surpassing 40 d °C. The global average anomalies of soil temperature and soil moisture during SCDHWs in 1980–2023 are 1.8 °C and –0.018 cm³ cm^–3, while during isolated soil heatwaves and soil moisture droughts, they are 1.3 °C and –0.013 cm³ cm^–3. SCDHWs have strengthened in all seasons, with the most impactful events observed in summer (SI Appendix, Fig. S1). The severity of SCDHW is primarily attributed to the heightened intensity in the Northern Hemisphere, notably in regions such as Siberia, and the prolonged duration in the Southern Hemisphere, including northern South America (SI Appendix, Fig. S2).

Fig. 1.

Trends (P < 0.001) in the global mean (A) occurrence frequency, (B) duration, (C) peak intensity, and (D) severity of CDHWs and SCDHWs, respectively, during the period from 1980 to 2023. The SCDHW results are presented as mean values with SD based on the nine combinations of soil moisture and temperature data from ERA5, GLDAS, and MERRA2. The CDHW results are presented as mean values with SD based on ERA5 2-m temperature data along with soil moisture data from ERA5, GLDAS, and MERRA2.

The strength of SCDHWs is enhanced during El Niño, further amplified by global warming. Specifically, compared to non-El Niño years in the 1980s, 1990s, 2010s, and 2020s, the strong El Niño years of 1983, 1998, 2015, and 2023 experience a significantly higher SCDHW occurrence frequency by 31.6%, 28.4%, 15.9%, and 12.5% as well as an extended SCDHW duration by 63.2%, 62.4%, 30.9%, and 31.3%. During this century, the strength of SCDHWs has increased while the relative contribution of El Niño has decreased. Sustained global warming fuels major SCDHWs in non-El Niño years, for instance, the unprecedented heatwaves in 2010 that swept across the Northern Hemisphere, including in Europe (36) and Southwestern China (37). Consequently, the global average duration and severity of SCDHWs in 2010 surpass those observed during the El Niño year of 2015 (Fig. 1). The escalation of SCDHWs persisted even during the occurrence of a triple-dip La Niña in 2020–2022.

We observe a rapid expansion of global land affected by SCDHWs since the 2000s, in line with the exacerbation of droughts and heatwaves (Fig. 2 A and B). The rapid increase in hotter soils has substantially expanded SCDHW occurrences this century. In the most recent decade, about two-thirds of global land areas are affected by soil moisture droughts lasting over one week, while half of them experience SCDHWs that persist for longer. SCDHWs in the Southern Hemisphere are predominantly influenced by El Niño. South America experiences extensive areas of SCDHWs in 2015–2016, 2020, and 2023, while Oceania encounters an unprecedentedly large extent of SCDHWs in 2019. Europe has been frequently impacted by SCDHWs since the 2000s. The land areas affected by SCDHWs have remained large in North America, Asia, and Africa during the past decade due to persistent droughts and heatwaves in the western United States, Siberia, and across Africa, respectively (Fig. 2C). Widespread SCDHWs are observed in Latin America in 2023.

Fig. 2.

Total global and continental land areas affected by soil extreme events during the period from 1980 to 2023. (A) Global land areas affected by soil moisture droughts and soil heatwaves lasting more than one week each, as well as SCDHWs lasting more than one, two, and four weeks, respectively. (B) Continental land areas of (B1) North America, (B2) South America, (B3) Europe, (B4) Africa, (B5) Asia, as well as (B6) Oceania affected by SCDHWs lasting more than one, two, and four weeks, respectively. (C) Global SCDHW severity from 2019 to 2023 in (C1−C5).

Soils Suffer More than the Air.

From 1980 to 2023, global SCDHWs consistently exhibit higher frequencies, longer durations, and greater severities than CDHWs (Fig. 3 and SI Appendix, Fig. S3). Despite these differences, they display similar spatial patterns, reflecting a strong correlation between near-surface air temperature and shallow soil temperature (SI Appendix, Fig. S4). Mean CDHW metrics account for approximately 74%, 77%, 84%, and 47% of the spatial variability in the mean SCDHW metrics for frequency, duration, peak intensity, and severity, respectively. However, CDHW metrics are less effective at capturing the full severity of SCDHWs. From 1980 to 2023, SCDHWs have lasted, on average, 3.5 d longer globally than CDHWs, with particularly notable increases in the northern regions of South America and central Africa (Fig. 3A). Although SCDHWs are less intense on average, with a global average of 0.17 °C, they exhibit higher peak intensities in tropical regions and the Northern Hemisphere, especially in Siberia (SI Appendix, Fig. S3C). Consequently, these areas have emerged as hotspots of more severe SCDHWs over recent decades (Fig. 3C). Despite their lower intensity, SCDHWs have a global average severity that exceeds that of CDHWs by 6.91 d °C, due to their higher occurrence frequency and longer duration.

Fig. 3.

Differences in annual mean and the trend of duration (A and B) and severity (C and D) between SCDHWs and CDHWs during the period from 1980 to 2023.

Over the past few decades, there has been a faster increase in the occurrence frequency (0.02 events dec^–1), duration (0.71 d dec^–1), peak intensity (0.1 °C dec^–1), and severity (3.09 d °C dec^–1) of global SCDHWs compared to CDHWs (Fig. 1). Summertime SCDHWs contribute more than 50% of the increase in duration and about two-thirds of the increase in severity (SI Appendix, Fig. S1). The peak intensity of global SCDHWs (CDHWs) has increased more than double (~1.5 times) the rate of global warming since 1980 (38). Hotter soils have become more extreme compared to hotter air. Although the mean metrics of CDHWs and SCDHWs exhibit similar spatial patterns (SI Appendix, Fig. S4), their trends differ across the globe (SI Appendix, Fig. S2). Trends in the CDHW metrics only explain approximately 37%, 42%, 22%, and 25% of the spatial variability in the trends of SCDHW metrics for the occurrence frequency, duration, peak intensity, and severity, respectively. The discrepancies in these trends highlight the long-term differences in CDHWs and SCDHWs and demonstrate the independence of SCDHW from CDHW as an independent metric. Notably, northern South America and Siberia have emerged as hotspots for the accelerated growth of SCDHWs (Fig. 3 and SI Appendix, Fig. S3).

In situ data confirm that SCDHWs exhibit higher frequencies, longer durations, greater severities, and faster growth compared to CDHWs (SI Appendix, Fig. S5). Hot soils can reach more extreme conditions than hot air, especially in croplands where soil temperature anomalies are more pronounced (SI Appendix, Fig. S6 A and C). These extreme soil temperatures typically occur during the later stages of soil drying or in periods of consistently low soil moisture content (comparing negative SM anomalies in SI Appendix, Fig. S6). They are less affected by light precipitation indicated by marginal increases in soil moisture content, especially in forests, and thereby exhibit a higher degree of temporal continuity than hot air (e.g., continuous periods of hot soils since day 100 and day 200 in SI Appendix, Fig. S6B). The reduced sensitivity of soil temperatures to precipitation also leads to a delayed termination of SCDHWs, with extended periods of hot soils persisting despite increases in soil moisture content (SI Appendix, Fig. S6). In contrast, hot air can dissipate quickly, such as after a cold front, and both CDHWs and SCDHWs can end due to the arrival of a cold air mass or storm.

Daily temperature thresholds for defining heatwaves can be lower for soils compared to the air, particularly during spring and summer in the Northern Hemisphere (SI Appendix, Fig. S7). Recent research has shown an increased frequency of instances where soil temperatures exceed air temperatures (12). Even when soil temperatures are slightly lower than or in equilibrium with air temperatures, positive soil temperature anomalies (i.e., soil heat extremes) can still occur independently of positive air temperature anomalies (i.e., air temperature extremes). The higher soil temperatures and lower thresholds for soil heatwaves contribute to the earlier onset and longer duration of SCDHWs compared to CDHWs. This extended temporal range makes SCDHWs more severe in terms of accumulated positive temperature anomalies than CDHWs.

Soil Cover: A Mitigating Factor for Dry-Hot Soils.

Forestlands have lower SCDHW intensities and severities compared to savannas, grasslands, and croplands across the globe (Fig. 4 A and B), although the SCDHW occurrences are occasionally more frequent (negative values in Fig. 4C1) and longer-lasting (negative values in Fig. 4C2). The increased occurrence frequency might be attributed to more precipitation over forests than over the adjacent soil covers, due to more intensive evapotranspiration. Forestlands in general conserve more soil moisture, and the soil temperatures under forest canopy are less susceptive to changes in air temperature. The gradual and marginal increases in soil temperature correspond to lower SCDHW peak intensities (positive values in Fig. 4C3). Similarly, forestland soils also release heat at a slower rate, which might explain the occasionally longer SCDHW durations. Globally, the SCDHW metrics for forestlands are lower than other land surfaces, particularly in terms of their severity (positive values in Fig. 4C4). This result is confirmed by in situ data, which reiterate significantly lower SCDHW severities in forestlands and wetlands (Fig. 4D). Transforming forestlands and wetlands to other soil covers will undoubtedly intensify SCDHWs. More broadly, turning natural vegetation including grasslands to croplands will substantially increase SCDHW severity.

Fig. 4.

Differences in annual mean SCDHW (A) peak intensity and (B) severity between shrublands, grasslands, and croplands as a whole and the adjacent forestlands. The mean differences in occurrence frequency (C1), duration (C2), peak intensity (C3), and severity (C4) are shown along the latitudinal direction, and the land use/land cover dependence of SCDHW severity are shown in (D) based on in situ data. Forests include evergreen needleleaf forests, evergreen broadleaf forests, deciduous needleleaf forests, deciduous broadleaf forests, and mixed forests.

Intensifying and Persistent SCDHWs in the Future.

Droughts are replacing heatwaves as the primary constraint of global SCDHWs. In the past century, the severity of SCDHWs is mainly controlled by the duration of soil heatwaves (SI Appendix, Table S1). While in this century, the duration of soil moisture droughts is becoming increasingly dominant (SI Appendix, Table S1). As the occurrence frequency and duration of soil heatwaves increase much faster than those of soil moisture droughts (SI Appendix, Fig. S8), the severity of future SCDHWs will be dominated by soil moisture droughts and thus by precipitation patterns. Decades of observations have confirmed models predicting more extreme precipitation events in a warming climate (39, 40). These events can lead to soil cooling and the disruption of soil moisture droughts, thereby increasing the occurrence frequency of SCDHWs.

The end of this century will see formidable global SCDHWs averaging 3.5 times over 70 d at a peak intensity exceeding 10 °C under the SSP5-8.5 emission scenario (SI Appendix, Fig. S9). The growth rate in occurrence frequency (~2.5 times from 1980 to 2100, SSP5-8.5) lags that of duration (~5 times from 1980 to 2100, SSP5-8.5), indicating a prolongation of individual SCDHW episodes. The spatial distribution of future SCDHWs varies among models, with notable hotspots observed in the United States, northern South America, Europe, and West Asia (SI Appendix, Fig. S10).

The future occurrence of SCDHWs is projected to be more prolonged and severe compared to CDHWs (Fig. 5). Two opposing trends emerge from the historical period. First, SCDHWs will be more intense than CDHWs because global soil temperatures frequently exceed air temperatures due to net radiation, which increases soil temperature in dry and warm conditions. Second, the difference in occurrence frequency of SCDHWs and CDHWs will diminish in the future. Under the SSP5-8.5 emission scenario, the occurrence frequency of CDHWs is projected to exceed that of SCDHWs in the 2080s. This shift is attributed to the higher resilience of hot soils to precipitation than hot air. In the future, enhanced atmospheric water demand will cause rapid evaporation of light precipitation, resulting in evaporative cooling that primarily affects air temperature while exerting a lesser impact on soil temperature, potentially disrupting CDHWs rather than SCDHWs. The enhanced canopy interception might also play a role.

Fig. 5.

Global mean differences in (A) occurrence frequency, (B) duration, (C) peak intensity, and (D) severity between SCDHWs and CDHWs under varying emission scenarios during the period from 1980 to 2100. The results are presented as mean values with SD based on the CMIP6 EC-Earth3, MIROC6, MPI-ESM1.2-LR, and NorSM2-LM data.

Discussion and Implications

We define the SCDHW exclusively for global soils, which are uniquely impacted by both droughts and heatwaves. In contrast, CDHWs are defined meteorologically based on precipitation and air temperature, which inadequately represent the combined water and heat stresses experienced at the soil level. Soil moisture droughts and soil temperatures differ spatially and temporally from their meteorological counterparts (12, 26, 41, 42), making CDHWs an imprecise proxy for SCDHWs. Our analysis demonstrates that air temperature alone cannot capture the complex dynamics of soil heat, especially during drought conditions. As global warming continues to exacerbate drought frequency and intensity, the growing discrepancies between soil and air temperatures highlight the need to focus on soil-orientated compound events.

Soil temperature, modeled through heat diffusion in vegetation and soils using meteorological forcing data, can vary significantly from air temperature. Relying on air temperature alone to assess CDHWs often underestimates their impact on global soils. Accurate evaluation requires incorporating soil temperature to fully capture the effects of these events. The use of air temperature as a proxy underestimates the severity of past SCDHWs, despite an overestimation of the intensity. Given that soil heat extremes can outpace air temperature extremes (12), the severity of SCDHWs will be substantially underestimated. In the future, soil ecosystems are expected to face more severe conditions than previously anticipated. Currently, thermal responses of soil carbon fluxes are evaluated based on soil temperature, while global heatwaves are assessed based on air temperature. Although global soil warming generally aligns with air warming trends, nonnegligible disparities arise in cold regions (27) and during heatwaves (12). Cold regions serve as hotspots of soil carbon storage, and the exacerbation of heatwaves is making matters worse. Extrapolating site-scale findings to a global scale will likely bias the calculation of soil carbon budget.

SCDHWs are more destructive than CDHWs due to the contrasting heat storage and transfer capacities between soils and the overlying air. Drying of shallow soils intensifies heating in near-surface air and subsoils by increasing the proportion of sensible heat. Additionally, soil temperatures are more reflective of past air temperature anomalies than precipitation signals (43). Although an intrusion of cold air might displace hot air over a short period, the shallow soils will release heat later, thereby sustaining the positive temperature anomalies (43). Soils can not only buffer low air temperatures but also regulate high air temperatures depending on the availability of soil moisture for evapotranspiration. In the future, frequent droughts likely diminish the buffering capability against high air temperatures, resulting in elevated soil temperatures and intensified SCDHWs.

Escalating SCDHWs pose significant threats to terrestrial ecosystems, particularly agricultural systems. Transforming natural ecosystems to croplands is likely to exacerbate SCDHW intensity, making agricultural production increasingly vulnerable to global warming. First, hot soil temperatures are expected to regularly exceed hot air temperatures and persist longer, as net radiation primarily heat soils in dry and warm conditions (12), leading to more intense and severe SCDHWs. These events are predominantly concentrated during the summer growing seasons. Second, the mean duration of individual SCDHWs is projected to increase substantially, reducing recovery periods for soil biota and plant root systems from extreme water and heat stresses. These developments represent serious challenges to global food security.

Northern high latitudes will face continuous threats from more intensive and severe summertime SCDHWs. In summer, soils have a lower mean temperature (26) and lower daily 90th percentile temperature thresholds (SI Appendix, Figs. S7 and S11 A–E) than the air in northern high latitudes. As soil heat extremes escalate at a faster rate than air temperature extremes, positive anomalies in soil temperature will surpass those in air temperature, exerting significant pressures on soil biota and plant root systems that have long adapted to cooler soil environments. The soil temperature thresholds are higher than or close to the air temperature thresholds in northern low-middle latitudes and the Southern Hemisphere (SI Appendix, Fig. S11 F–O) where prolonged droughts might have a greater impact.

Soil temperature will play an important role in soil moisture–atmosphere feedback mechanisms, amplifying droughts, heatwaves, and CDHWs. With ongoing global warming, intensified SCDHWs in the Southern Hemisphere will present major challenges for underdeveloped countries working to achieve food security. The El Niño phenomenon exacerbates this issue by inducing an uneven distribution of record-breaking SCDHW events. Even in non-El Niño years, regional mega SCDHWs can arise from unforeseen atmospheric circulation anomalies. It is imperative that we prepare for increasingly stronger SCDHWs.

Materials and Methods

Gridded Soil Temperature and Moisture Data.

We utilize three reanalysis datasets (1980–2023) to quantify historical soil moisture droughts, soil heatwaves, CDHWs, and SCDHWs. The datasets comprise the European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ERA5) (34), the Global Land Data Assimilation System (GLDAS) Noah v2.0/2.1 (44), and the Modern-Era Retrospective analysis for Research and Applications v2 (MERRA2) (45). We extract the ERA5 2-m air temperature data, and the shallowest (0 to 7 cm for ERA5, 0 to 10 cm for GLDAS, and 0 to 9.88 cm for MERRA2) soil moisture and temperature data from all reanalysis datasets. All data are averaged to daily values and are resampled onto the GLDAS 0.25° × 0.25° grids using a nearest neighboring method. These data remain unadjusted for differences in soil layers, as they do not impact the identification of soil extreme events based on relative magnitudes of data series.

We integrate the GLDAS Noah v2.0 (1980–2014) and v2.1 (2000–2023) datasets by aligning the common data from 2000 to 2014. To address biases in soil temperature, we utilize the grid-by-grid quadratic regression method, while for biases in soil moisture, we utilize the cumulative distribution function (CDF) matching method (46). Specifically, quintic polynomials are employed to match the CDFs of unfrozen (soil temperature > 273.15 K) soil moisture data between the v2.0 and v2.1 datasets. By transforming the v2.0 data from 1980 to 1999, we generate temporally compatible soil temperature and soil moisture data spanning from 1980 to 2023. Please refer to SI Appendix, Fig. S12 for the calibration results.

We utilize four Coupled Model Intercomparison Project Phase 6 (CMIP6) Earth System Model (ESM) datasets (1980–2100) (47) to quantify future CDHWs and SCDHWs. The ESMs encompass the European Community Earth-System Model v3 (EC-Earth3), the Model for Interdisciplinary Research on Climate v6 (MIROC6), the lower-resolved version of the Max Planck Institute for Meteorology Earth System Model v1.2 (MPI-ESM1.2-LR), and the low atmosphere-medium ocean resolution version of the Norwegian Earth System Model v2 (NorESM2-LM). All datasets are derived from the initial ensemble simulation (r1i1p1f1), with at least daily temporal resolutions, covering a full range of shared socioeconomic pathways (SSPs) including SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5. The 2-m air temperature as well as soil temperature and moisture data within the upper portion (0 to 10 cm) are extracted and, if necessary, averaged to obtain daily values.

Global Site Observations.

We collect near-surface (2 m or 1.5 m as per site specifications) air temperature data and shallow-layer (5 cm or 5.08 cm as per network specifications) soil temperature data from the International Soil Moisture Network (ISMN) (48), the FLUXNET, and the Integrated Carbon Observation System (ICOS). Our selection criterion requires a minimum of 15-y soil temperature observations in 1980–2023. The qualified ISMN sites are from the Soil Climate Analysis Network (SCAN), the SNOwpack TELemetry (SNOTEL), the United States Climate Reference Network (USCRN), the Atmospheric Radiation Measurement (ARM), and the Bonanza Creek Long Term Ecological Research (BNZ-LTER). The hourly observations flagged with a quality indicator of “G” (abbreviation for good) are retained and averaged to obtain daily data. Daily averages will not be calculated if there are more than one missing hourly record in a day. We also incorporate the FLUXNET2015 dataset (49) along with the two ICOS datasets released in 2022 (50) and 2023 (51), respectively, for air temperature and the shallowest soil temperature data. Given the ICOS processing pipeline is compliant with the FLUXNET2015, we merge these datasets based on sites' coordinates.

We utilize the ERA5-Land 2-m air temperature and 0 to 7 cm soil temperature data (35) to fill gaps in the observations. First, we compute multiyear mean and SD values for observations on each day of year (DOY) at each site and exclude data points that fall outside the range of mean ± 2 s.d., which may be attributed to temporary sensor malfunctions such as spikes or no responses. Second, we calculate correlation coefficient (r) values between ERA5-Land and observational time series separately for air and soil temperature, and eliminate sites with either r < 0.8 due to long-term sensor degradation or unexpected sensor responses. Finally, we conduct quadratic regressions between ERA5-Land and observational time series to interpolate low-quality or missing data points. Please refer to SI Appendix, Fig. S13 for a spatial distribution map of the final selection comprising 212 ISMN sites and 32 FLUXNET/ICOS sites.

Define and Characterize Extreme Events in Air and Soils.

We adopt Hobday’s method (52) to define extreme events. A soil moisture drought event is defined as a period of ≥5 consecutive days with soil moisture content below the 10th percentile (SM10) on that DOY during the baseline period from 1980 to 2010. To calculate SM10, unfrozen soil moisture data are aggregated within a 5-d window centered on each DOY over the span of 31 y. Gaps of two days or less between two events are considered as part of a continuous event. Similarly, we define a heatwave or soil heatwave event as a period of ≥ 5 consecutive days with air or soil temperature above the 90th percentile (TA90/TS90) on that DOY during the baseline period from 1980 to 2010. We finally define a SCDHW/CDHW event as a period of ≥5 consecutive days suffering both soil moisture drought and heatwave/soil heatwave.

We characterize extreme events based on their occurrence frequency (N), duration (D), peak intensity (I), and severity (S) within a year or season. Peak intensity refers to the maximum deviation in soil moisture from SM10 during soil moisture droughts, the maximum deviation in air temperature from TA90 during heatwaves and CDHWs, and the maximum deviation in soil temperature from TS90 during soil heatwaves and SCDHWs. Severity is the cumulative intensity over time. The mean duration and mean intensity of extreme events are calculated as D/N and S/D, respectively. Absolute values are employed to highlight the adverse impacts associated with all extreme events.

Quantify Spatiotemporal Changes in SCDHWs.

We calculate metrics related to historical soil moisture droughts, soil heatwaves, CDHWs, and SCDHWs from 1980 to 2023 using reanalysis datasets. The ERA5 2-m air temperature is combined with reanalysis soil moisture products to identify CDHWs (3 combinations). All soil moisture and temperature products are used to describe soil moisture droughts and soil heatwaves, respectively, and are matched and cross-matched to identify SCDHWs (3 × 3 combinations). The metrics values are calculated at annual and seasonal scales (spring–MAM, summer–JJA, autumn–SON, winter–DJF). Results are averaged to mitigate uncertainties in datasets and correlative errors between soil temperature and soil moisture within the same dataset. Future CDHWs and SCDHWs are characterized by multi-ESM ensembles, respectively, for the SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5 emission scenarios. All soil moisture droughts, CDHWs, and SCDHWs are considered for unfrozen soils.

The spatiotemporal patterns and driving factors of global SCDHWs are clarified below. First, we calculate the globally mean frequency, duration, peak intensity, and severity from 1980 to 2023. Then, we calculate the metrics values during El Niño years and in summertime to quantify their contributions to changes in SCDHWs. Subsequently, we assess the global land areas affected by SCDHWs for durations of at least one, two, and four weeks, respectively, highlighting SCDHW susceptive regions. Global hotspots of historical SCDHWs (based on multiyear mean) and SCDHW intensification (based on multiyear trend) are identified. We correlate trend in SCDHW severity and trends in soil moisture droughts and heatwaves, revealing the major factors driving changes in SCDHWs across the globe, Northern Hemisphere, and Southern Hemisphere. Finally, we present trends and hotspots of projected future SCDHWs under different emission scenarios.

We calculate metrics values for local CDHWs and SCDHWs based on the in situ air/soil temperature and ERA5-Land 0 to 7 cm soil moisture data. The use of ERA5-Land soil moisture instead of in situ observations is justified by its temporal consistency and higher accuracy. The averaged metrics values across all sites allow for a comparison of magnitudes and trends between CDHWs and SCDHWs at regional scales. Additionally, we present the data anomalies in air temperature, soil temperature, and soil moisture at specific sites to elucidate the distinctions between CDHWs and SCDHWs.

Investigate Land Use/Land Cover Dependence of SCDHWs.

We investigate the land use/land cover (LULC) dependence of SCDHWs at global and regional scales. Globally, we use the 0.25° × 0.25° GLDAS Vegetation Class/Mask data (53) to delineate LULC types. The modified International Geosphere–Biosphere Programme (IGBP) classes are aggregated into four major LULC types. The evergreen needleleaf forest, evergreen broadleaf forest, deciduous needleleaf forest, deciduous broadleaf forest, and mixed forest are categorized into forests. The closed shrublands, open shrublands, woody savannas, and savannas are categorized into shrublands. The cropland and cropland/natural vegetation mosaics are categorized into croplands. We also select the grasslands. Within a 1° × 1° spatial area, we determine the primary and second primary LULC types from the 4 × 4 grids and calculate separately the averages of multiyear mean SCDHW metrics (occurrence frequency, duration, peak intensity, and severity) values. Assuming similar climate conditions within the 1° × 1° spatial area, the differences in metrics values are recognized as LULC dependent. We calculate the differences between forests and the other three LULC types, showing how SCDHWs will evolve if forests are transformed to shrublands, croplands, and grasslands. Regionally, we also categorize all the 244 sites into barren land, shrublands, croplands, grasslands, forests, vegetation mosaics, and wetlands. The multiyear mean SCDHW metrics values are averaged for each LULC type.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The ERA5 and ERA5-Land datasets are from the Copernicus Climate Change Service Climate Date Store (https://cds.climate.copernicus.eu/) (34, 35). The GLDAS Noah datasets are from the NASA Land Data Assimilation System (https://ldas.gsfc.nasa.gov/gldas/) (44). The GLDAS Vegetation Class/Mask is from https://ldas.gsfc.nasa.gov/gldas/vegetation-class-mask (53). The MERRA2 dataset is from the NASA Goddard Earth Sciences Data and Information Services Center (https://gmao.gsfc.nasa.gov/reanalysis/MERRA-2/) (45). The ESM datasets are from the Lawrence Livermore National Laboratory Earth System Grid Federation node (https://esgf-node.llnl.gov/search/cmip6/) (47). Sources of in situ data are included in the article. The code for the identification of extreme events is available at https://github.com/ecjoliver/marineHeatWaves (52).