Amplified temperature sensitivity of extreme precipitation events following heat stress

Extreme precipitation events (EPEs) are among the most destructive natural hazards, causing major societal, economic, and environmental impacts. Reports from the WMO (1970–2021) identify heavy precipitation–flood disasters as the most prevalent and most economically damaging among weather-related disasters, and IPCC AR6 notes more than 700 million people experiencing more intense EPEs than in the 1950s. Increasing attention has focused on compound events that link heat and precipitation. Compound heat stress and heavy precipitation events (CHPEs)—where extreme precipitation closely follows heat stress—often cause greater damage than single EPEs. Despite growing research on CHPEs, comparative analyses of the extreme precipitation components that occur with preceding heat stress (EPE-Hs) versus those without (EPE-NHs) have been lacking. Understanding these differences is essential for adaptation in regions where a large fraction of extreme precipitation follows heat stress. This study targets the warmest five months locally, compares EPE-Hs and EPE-NHs using reanalysis and CMIP6 models, and addresses: (1) where and how spatiotemporal variations in EPEs differ between EPE-Hs and EPE-NHs in historical and future periods; and (2) how these EPE types respond to temperature under different future scenarios.

Prior studies using CMIP6 multi-model ensembles project increased future EPE frequency and intensity globally, especially under high-emission scenarios. Observed and modeled precipitation–temperature (P–T) scaling ranges from super-Clausius–Clapeyron (C–C) in mid-latitude and dry regions to sub-C–C or negative in tropical/wet regions, often exhibiting a hook structure (increase to a peak then decrease) at mid-latitudes. High-resolution modeling suggests sharper future hook behavior and increased peak intensities. Event-type dependence also matters: atmospheric river-induced extremes show higher scaling rates than non-AR events in California. Frequency of CHPEs has risen since the 1960s and is projected to continue rising under high emissions. These works underscore regional disparities, scenario dependence, and event-type specificity in extreme precipitation responses to warming, motivating a split analysis of EPE-Hs versus EPE-NHs.

Study domain: global land excluding deserts (annual precipitation <200 mm). Warm season defined as the local hottest five months. Data: Observations include ERA5 (hourly, 0.25°) 2 m air temperature, 2 m dew point temperature, CAPE, SSHF, VIMC, TCWV, and CPC daily precipitation (0.5°), for 1979–2014. Both datasets were converted to daily and regridded to 1°×1°. Models: Six CMIP6 ESMs: CNRM-CM6-1, EC-Earth3-Veg, KACE-1-0-G, MPI-ESM1-2-HR, MRI-ESM2-0, NorESM2-MM for baseline (1979–2014) and future (2015–2099) under four SSP–RCPs: SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5. Model data interpolated to 1°×1°. Bias correction via Quantile Delta Mapping (QDM) preserving modeled relative quantile changes while correcting biases. Event identification: At each grid, the 90th percentile of daily precipitation in the warm season (1979–2014) defines the EPE threshold. An EPE occurs when daily precipitation exceeds this threshold for ≥1 day. Heat stress events are defined using lethal heat stress temperature Th (derived from wet bulb Tw and RH) exceeding its warm-season 90th percentile for ≥3 consecutive days. CHPE is defined as a heat stress event followed by an EPE within a 3-day interval; such EPEs are EPE-Hs. EPEs not preceded by qualifying heat stress within this window are EPE-NHs. Robustness checks use 95th percentile thresholds and 1- and 7-day intervals. Metrics: For each grid and year, compute EPE frequency (events/year), duration (average days/event), and magnitude M, defined as the average normalized daily precipitation anomaly within events: M = Σ(P_rd − P_r25th)/(P_r75th − P_r25th)/D, where D is event duration and P_r25th/P_r75th are 25th/75th percentiles of wet-day precipitation in the historical period. Trends: Linear trends via least squares; significance via Mann–Kendall test (α=0.05). Temperature sensitivity: For each grid, compute the relative change (%) in the total precipitation amount associated with EPEs between baseline and future (near-term: 2015–2040, mid-term: 2041–2070, long-term: 2071–2099). Divide by the change in global surface air temperature (GSAT, °C) between baseline and future to obtain sensitivity (%/°C). Summarize via spatial maps, latitude means, regional means (22 IPCC-like land regions), and global means. P–T scaling: Compute annual EPE precipitation amounts (global/regional area-weighted means). Smooth P–T relations using LOWESS and estimate scaling rates by linear regression in log space: log(P) = α + β T, with scaling rate α_s = 100 (e^β −1) %/°C. For global scaling, use GSAT; for regional scaling, use regional SAT (RSAT). Analyze structures for all EPEs, EPE-Hs, and EPE-NHs, in baseline and under all scenarios. Atmospheric condition analysis: Composite anomalies one day before EPE onset for CAPE, specific humidity (SH), surface sensible heat flux (SSHF), vertically integrated moisture convergence (VIMC), and total column water vapor (TCWV). Compare EPE-Hs vs EPE-NHs with t-tests (p<0.05).

• Historical contrasts (1979–2014):
  • EPE-Hs are globally less frequent than EPE-NHs but have longer duration and greater magnitude, especially north of 45°N, the central U.S., and the southern Tibetan Plateau.
  • EPE-NHs occur more often in Southeast Asia, Amazon, northeastern America and Asia; high-magnitude EPE-NHs occur in central North America, NE Brazil, S. Argentina, Middle East, NW India, S. Africa, and S. Australia.
  • Trends: Global EPE frequency increases mainly due to rising EPE-H frequency. EPE-H frequency shows increasing trends in all 22 regions (mostly significant), while EPE-NH frequency decreases in WNA, AMZ, WAF, and EAF. Duration shows no global trend overall, but EPE-H duration increases in ALA, CAM, SAF, NAU and decreases in WAF. Magnitude slightly increases globally; EPE-H magnitude increases in ALA, GRL, NEB, NAS, MED, EAF, NAU and decreases in CAM, TIB, SAS; EPE-NH magnitude increases in ALA, CAM, NEB, SSA, EAF.
• Future projections (MME of six CMIP6 models):
  • All EPEs: Slight increases in frequency, duration, and magnitude through 1979–2099, with higher values under higher-forcing scenarios, especially for magnitude in the late 21st century.
  • EPE-Hs vs EPE-NHs diverge strongly:
    • Frequency: EPE-Hs increase from baseline to ~2060s; EPE-NHs tend to decline. Under scenarios, EPE-NH frequency is paradoxically higher under lower forcing.
    • Duration: EPE-Hs increase substantially and persistently; EPE-NHs remain relatively stable.
    • Magnitude: EPE-Hs rise modestly early, then surge after ~2050 under SSP3-7.0/SSP5-8.5; EPE-NHs remain comparatively stable across scenarios.
  • Spatially (long-term, 2071–2099 vs baseline): Magnitude increases almost everywhere for all EPEs, except Mediterranean and southern Australia under SSP1-2.6. Largest increases under SSP5-8.5 in Tibetan Plateau, NW North America, W South America, and E Africa. EPE-H magnitude increases exceed those of EPE-NHs worldwide; EPE-NH magnitude decreases in Mexico, S Brazil, central/southern Africa, central Asia, Pakistan, and Australia.
• Temperature sensitivity (%/°C):
  • EPE-Hs exhibit much higher sensitivity than all EPEs and EPE-NHs under all scenarios; global mean sensitivity for EPE-Hs exceeds 200%/°C, especially pronounced in low latitudes (20°N–20°S).
  • For all EPEs, global average sensitivity decreases from near-term to long-term, indicating stronger near-term response to warming.
  • Spatial patterns: Positive sensitivity for all EPEs across most land; negative sensitivity in Central America, NE South America, Mediterranean, southern Africa, and SE Australia. EPE-NH sensitivity is often negative in mid-latitudes; positive in Southeast Asia, eastern Africa, NW South America, and >60°N.
  • Regional (22 regions): Sensitivity generally higher under low-forcing scenarios for most regions; exceptions (SSA, WAF, EAF, SAF, NAU) show higher sensitivity under high-forcing scenarios. Temporal evolution differs by region group.
• Global P–T scaling structures:
  • All EPEs: Monotonic linear increase with GSAT; global scaling rate ≈ 9.07%/°C (super C–C).
  • EPE-Hs: Convex (increase to peak then decrease); peak-point temperature Tp ≈ 17.19 °C (global GSAT), with peak precipitation Pp ≈ 105.4 mm (global-scale metric reported).
  • EPE-NHs: Concave (decrease then increase) with warming.
• Regional P–T scaling (RSAT):
  • All EPEs increase monotonically with RSAT in most regions; super C–C in ALA, SSA, TIB, EAS, SAS, SEA, EAF; sub C–C elsewhere.
  • EPE-Hs: Two patterns—low-latitude regions (CAM, AMZ, NEB, SEA, WAF, EAF, SAF, NAU) show a peak at region-specific Tp (highest WAF ~30.38 °C; lowest SAF ~25.18 °C; largest peak precipitation in SEA ~270 mm; smallest in EAF ~105 mm). Other regions show continuous increases.
  • EPE-NHs: In the same low-latitude regions, decrease then increase with RSAT; in other regions, monotonic decrease.
• Physical environment one day prior to events:
  • High latitudes show larger CAPE, SH, SSHF, and TCWV before EPE-Hs than before EPE-NHs, consistent with their longer duration and higher magnitude. Preceding heat stress elevates RH and instability, priming extreme precipitation; differences likely reflect thermodynamic dominance with dynamic modulation.

The study addresses where and how EPEs differ when preceded by heat stress versus not, and how each type responds to warming. Empirically, EPE-Hs are rarer but longer-lasting and more intense historically and become more frequent, longer, and more intense under warming, whereas EPE-NHs tend to decline in frequency and show stable duration/magnitude. The pronounced temperature sensitivity of EPE-Hs (often >200%/°C) indicates that a relatively small increase in GSAT can lead to disproportionately large increases in the total EPE-H precipitation, especially in low latitudes. Divergent P–T scaling structures reveal that the presence of antecedent heat stress fundamentally alters how extremes scale with temperature: EPE-Hs peak then decline (convex), while EPE-NHs show concave behavior, underscoring compound-process dependence. Spatial and regional disparities highlight that adaptation and early-warning strategies must be tailored by latitude and region. Composite diagnostics indicate that prior heat stress increases moisture availability and convective instability (higher SH, TCWV, CAPE) and surface fluxes, priming the atmosphere for more severe precipitation extremes. This thermodynamic enhancement explains the amplified sensitivity for EPE-Hs, while reduced RH at very high temperatures may temper EPE-NH scaling in some regions. The results are relevant for risk management, as many areas—particularly in low latitudes and some high-latitude regions—will see stronger changes in EPE characteristics when preceded by heat stress.

This work provides the first global comparison of extreme precipitation events preceded by heat stress (EPE-Hs) versus those without (EPE-NHs) across historical and CMIP6-projected climates focused on local warm seasons. Key contributions include: (1) documenting that EPE-Hs are less frequent but longer and more intense historically, with strong future increases in frequency, duration, and magnitude, unlike EPE-NHs; (2) quantifying much higher temperature sensitivity for EPE-Hs (global averages >200%/°C), especially in low latitudes; (3) revealing distinct global and regional P–T scaling structures: monotonic super C–C scaling for all EPEs, convex scaling with a peak for EPE-Hs, and concave scaling for EPE-NHs; and (4) diagnosing atmospheric precursors (higher CAPE, SH, SSHF, TCWV) that help explain the amplified EPE-H response. These insights support region-specific early-warning, planning, and adaptation strategies under global warming. Future research should deepen the physical attribution with process-resolving and convection-permitting models, explore model uncertainties and event-type classifications, assess sensitivity to thresholds and compounding windows, and examine impacts on hydrological extremes and cascading risks.

• Seasonal focus: Analyses are limited to local warm seasons; scaling behaviors and sensitivities may differ in other seasons.
• Threshold/event definition: Results depend on the 90th (and tested 95th) percentile thresholds and a 3-day (tested 1- and 7-day) CHPE window; alternative definitions could change event attribution between EPE-H and EPE-NH.
• Model ensemble size and resolution: Only six CMIP6 ESMs were used, at coarse resolution; convection-permitting processes are unresolved and may affect extreme precipitation characteristics and scaling.
• Bias correction and regridding: QDM and bilinear interpolation introduce methodological assumptions; while preserving relative changes, residual biases can remain.
• Uncertainty and agreement: Some regions show low inter-model agreement on sensitivity signals; scenario dependence varies by region and time slice.
• Mechanistic attribution: While composites suggest thermodynamic priming, detailed dynamical mechanisms (e.g., circulation regimes, mesoscale organization) are not fully resolved and warrant further study.