The characterization, mechanism, predictability, and impacts of the unprecedented 2023 Southeast Asia heatwave

The frequency, duration, and intensity of heatwaves have risen markedly in recent decades, causing substantial mortality and socioeconomic impacts worldwide. Catastrophic events in Europe (2003), Russia (2010), and the Pacific Northwest (2021) illustrate the growing toll of extreme heat. While many recent studies emphasize unprecedented mid- and high-latitude heatwaves where surface warming trends are pronounced, Southeast Asia (SEA) has received comparatively less attention despite its high vulnerability due to dense population, complex terrain, and developmental challenges. Surrounded by the world’s warmest oceans, SEA is particularly susceptible to extreme heat, and recent research indicates heatwaves in the region are becoming more frequent, longer-lasting, and more intense under global warming. In mid-April and early May 2023, a record-breaking heatwave struck Continental Southeast Asia (CSEA), with exceptional temperatures and widespread impacts. This study aims to systematically characterize the event’s spatiotemporal evolution, elucidate the physical mechanisms (including synoptic controls and tropical wave interactions), assess subseasonal forecast performance, estimate return periods and joint probabilities of key drivers, and evaluate broad impacts, to inform more effective risk management and early warning strategies.

Prior work documents increasing global heatwave trends and associated mortality, with landmark events in Europe (2003), Russia (2010), and the Pacific Northwest (2021). Many studies focus on mid- and high-latitude regions where surface warming is fastest. However, SEA’s exposure to compound heat and humidity, its proximity to warm oceans, and socioeconomic vulnerability suggest significant risk. Recent studies specific to SEA report increasing frequency, duration, and intensity of heatwaves under continued warming and highlight population exposure trends. Literature also emphasizes the role of synoptic-scale high-pressure systems (heat domes), tropical modes (MJO, equatorial Rossby and Kelvin waves) in modulating regional extremes, and the importance of land–atmosphere coupling where soil moisture deficits amplify heat. Forecasting advances at subseasonal-to-seasonal scales have been made, with ECMWF often leading in skill, though skill declines with lead time and land-surface initialization remains a challenge.

Data and study period: The analysis focuses on April–June 2023 over Continental Southeast Asia (CSEA). Temperature data include ERA5 reanalysis daily maximum 2 m temperatures (1950–2023) and 2023 station observations from GHCNd for validation. Synoptic fields (ERA5) include geopotential heights, zonal/meridional/vertical winds, specific and relative humidity at multiple levels, and outgoing longwave radiation (OLR). Anomalies are computed relative to the 1981–2010 daily climatology. Characterization of record-breaking areas: Following previous studies, moving averages of daily maximum temperature with window lengths 1–29 days were computed for each April–June day over 1950–2023. For each calendar day and timescale in April–June 2023, grid cells attaining record highs relative to 1950–2023 were identified, and spatial extents were calculated to track spatiotemporal evolution. Tropical wave decomposition: Intraseasonal OLR anomalies were obtained by removing the mean and first three harmonics of the annual cycle. Wavenumber–frequency spectral filtering (NCL filter_waves) separated contributions from Madden–Julian Oscillation (MJO), equatorial Rossby (ER), Kelvin, and mixed Rossby–gravity (MRG) waves within canonical wavenumber/period bands. MJO phases and amplitudes were analyzed via RMM indices. Land–atmosphere coupling: Soil moisture–temperature coupling strength π was quantified using an energy-balance-based diagnostic, where π = e′ × T and e′ = H′ − H′p = (Rn′ − λE′p) − (Rn′ − λE′). Here primes denote standardized daily anomalies; Rn is surface net radiation, E and Ep are actual and potential evaporation, and λ is latent heat. Positive π indicates stronger coupling where soil moisture deficits increase sensible heat and amplify temperature anomalies. Soil moisture anomalies (0–7 cm) were analyzed from ERA5. Subseasonal forecasts: ECMWF S2S (IFS) ensemble forecasts and 20-year reforecasts (0.5° resolution) were used to evaluate predictability of the 4–7 May peak. Variables analyzed included T2m, Z500, RH925, and 0–7 cm soil moisture (interpolated from 0–20 cm). Forecasts initialized on 1 May, 27 April, 24 April, 20 April, and 17 April (approximately 3–20 day leads) were compared to observations; skill metrics included mean absolute errors and pattern correlation coefficients. Return period and joint probability: A generalized extreme value (GEV) distribution with maximum likelihood estimation was fitted to regional mean 4-day temperatures to estimate the return period of the 4–7 May event. Joint probability densities (JPDs) were computed for pairs of key drivers—Z500 vs RH925, Z500 vs soil moisture (SM), and RH925 vs SM—using April–May data (1950–2023) to assess rarity of combined conditions.

- Event magnitude and records: About 70% of weather stations in Thailand, Laos, Cambodia, Myanmar, and Vietnam exceeded 42 °C. Laos set a national record of 42.9 °C on 19 April; Thailand reached 49 °C in early May; Vietnam set a new national high of 44.1 °C; Myanmar recorded 46 °C on 7 May with 61 reported heat-related deaths. Regional domain-average daily maximum temperature in May 2023 was the highest since 1950. - Spatiotemporal evolution: Four episodes were identified. Episode 1 (14–22 April) had a maximum daily exceedance area of ~266,000 km² (20% of CSEA). Episode 2 (4–7 May) was the peak, reaching ~560,000 km² (42% of domain) and containing most readings >42 °C. Episodes 3 (18–22 May) and 4 (28 May–2 June) were longer but contracted spatially (29% and 19% of domain). - Synoptic drivers: Pre-event, a heat-dome-like configuration featured high Z500 over CSEA, high Z925, northeasterly low-level dry advection, reduced 925-hPa relative humidity (RH925) by about −5%, and limited cloud cover allowing enhanced insolation. During the peak, Z500 intensified over CSEA, partly via southwestward expansion of the Western Pacific Subtropical High, reinforcing subsidence and moisture divergence over CSEA while a strong low over South China diverted moisture northward, further drying CSEA. RH925 decreased by an additional ~7% (from ~70% to ~63%) and cloud cover fell by ~30% relative to pre-event, accelerating surface warming. Post-event, WPSH retreated, a low-level cyclone formed near Vietnam’s east coast, and Cyclone Mocha transported moisture to CSEA, increasing clouds and precipitation and terminating the heatwave. - Tropical wave modulation: ER and Kelvin waves contributed strongly to positive OLR anomalies over CSEA during 4–7 May, consistent with suppressed convection and maximum regional temperatures. MJO transitioned from Phase 4 (from 30 April) to Phase 5 (6–7 May), with amplitude peaking at ~2.6, exceeding 95% of historical records for Phase 4 (and all Phase 4 records in May). This configuration strengthened Z500 and subsidence over CSEA via ER–Kelvin asymmetry and MJO “acceleration” effects. - Land–atmosphere coupling: Soil moisture deficits expanded from pre-event to peak, and coupling strength increased markedly, aligning with hotspots of extreme temperature. Upward sensible heat during the peak was ~−70 W m−2 (about 13% higher than pre-event), indicating strong positive feedback that amplified surface warming. After the event, precipitation increased soil moisture and reduced coupling. - Predictability: ECMWF S2S forecasts captured the broad spatial pattern of the heatwave at ~1-week lead (initialized 1 May) but underestimated intensity by ~1 °C. Skill declined rapidly with longer leads. While enhanced high pressure and negative RH anomalies were reasonably forecast, soil moisture anomalies were poorly predicted even at 1 week, indicating the model failed to capture strong land–atmosphere coupling, contributing to cold bias and underestimation. - Return period and rarity of drivers: The 4-day mean regional temperature exceeding 34.3 °C (4–7 May) corresponds to a return period of ~129 years. Joint probabilities of key drivers were extremely low: high Z500 with low RH925 (~0.15%), Z500 with low SM (~0.30%), and low RH925 with low SM (~0.08%). The conjunction of near-surface drying with soil moisture deficiency—triggering strong positive land–atmosphere feedback—was particularly rare. - Impacts: The heat and aridity led to an unprecedented surge in wildfire hotspots over CSEA (MODIS detections) during the peak episode, significant health impacts including reported fatalities, and adverse effects on agriculture. Rice production in Myanmar, Laos, Thailand, and Cambodia showed sensitivity to the event, with May overlapping key planting/harvesting periods, implying risks to yields and labor productivity under heat stress.

The study resolves the multi-scale drivers of the unprecedented 2023 CSEA heatwave and links them to observed extremes in intensity and extent. Synoptically, an intensified upper-level anticyclone over CSEA, aided by tropical wave interactions (ER and Kelvin) and an unusually strong MJO transitioning from Phase 4 to 5, generated strong subsidence, moisture divergence, and reduced cloud cover, which increased surface solar radiation and diabatic heating. Concurrently, a moisture-attracting low over South China further reduced moisture import into CSEA. These atmospheric conditions, coupled with pre-existing and rapidly intensifying soil moisture deficits, set up a strong positive land–atmosphere feedback in which reduced evaporative cooling and enhanced sensible heat flux amplified near-surface temperatures. A temperature budget analysis at 925 hPa indicated dominant contributions from diabatic heating. Forecast evaluation shows that while dynamic models like ECMWF can anticipate the spatial structure at short leads, deficiencies in simulating soil moisture and coupled land–atmosphere processes lead to systematic cold biases and rapidly diminishing skill with lead time. The extreme rarity of the joint drivers—especially co-occurring near-surface drying and soil moisture deficits—highlights the exceptional nature of the event and signals heightened compound risk in a warming climate. The documented impacts on wildfires, health, and rice agriculture underscore the societal urgency of improving early warnings and resilience.

This work provides an integrated characterization of the 2023 CSEA heatwave, attributing its unprecedented severity to the confluence of enhanced high pressure modulated by tropical waves and MJO, pronounced moisture deficits, and strong positive land–atmosphere coupling. The event’s regional 4-day mean temperature had an estimated 129-year return period, while the joint occurrence of key drivers—especially near-surface drying with soil moisture deficiency—was extraordinarily rare. Forecast diagnosis indicates that limited soil moisture predictability and representation of land–atmosphere coupling in ECMWF contributed to underestimation and rapid loss of skill with lead time. Future research should: (1) improve representation and initialization of soil moisture and land–atmosphere coupling in subseasonal models; (2) deepen process understanding of tropical wave interactions with regional circulation and their model representation; (3) develop statistical and hybrid (physics-informed and machine-learning) forecast systems to enhance lead-time and intensity predictions; and (4) translate advances into operational early warning to mitigate health, wildfire, and agricultural risks in SEA.