This study used daily minimum and maximum surface air temperatures from 54 meteorological stations in Brazil, Bolivia, Paraguay, and Argentina. The datasets were obtained from the National Institute of Meteorology in Brazil (INMET),¹ the National Meteorological and Hydrological Service of Bolivia (SENAMHI-senamhi.gob.bo), the National Meteorological Service of Argentina (SMN),² and the Department of Meteorology and Hydrology of Paraguay (DMH).³ As detected by these meteorological services, seven HW episodes occurred during this period in the region. New historical records of maximum temperatures were established during these HW events in 2023, some of which were the warmest in decades. The meteorological services of the region’s countries issued heat stress alerts for HW in cities, which were also reported by local media (newspapers, websites). This data allowed for identification of warm spells and HWs daily across the region of central South America from August 1 to December 31, 2023. The baseline to calculate anomalies is the 1991–2020 long-term mean (LTM) adopted by the WMO. The respective meteorological services that provided the data have their own systems for quality control and data homogenization. The10 representative stations were selected because they do not show missing data.

The Climate Prediction Center CPC Global Unified Temperature data were used to obtain daily global gridded maximum and minimum temperature at 0.5 × 0.5° resolution from 1 January 1979 to the present. CPC data is provided by the NOAA PSL, Boulder, Colorado, USA.⁴ As shown by Satgé et al. (2023) and Kew et al. (2023), while there is a high level of confidence in the NOAA-CPC Global Unified Temperature data used to carry out the analysis, these do not capture very local records, many of which were broken during the heat episodes.

Composites of atmospheric circulation anomalies, used to characterize the dynamical forcing mechanisms responsible for the HWs that affected South America during 2023, were computed using reanalysis data from the European Centre for Medium-Range Weather Forecasts ECMWF Reanalysis Version 5 (ERA5). ERA5 reanalysis provides land/sea and atmospheric fields at hourly time-steps and spatial grids of 0.25° from 1950 to the present (Hersbach et al., 2018, 2020, 2023). We calculated the horizontal wind anomalies at 850 hPa concerning the climatology for each case study to analyze the synoptic-scale circulation during HW events and their precursor conditions. For each identified HW, we averaged the circulation for the days of the event and the 3 days before each event (hereafter “pre-event”). For this purpose, the information of the zonal and meridional wind components at 850 hPa at 0 h GMT from the ERA5 reanalysis was used. ERA5 is a state-of-the-art reanalysis with a spatial resolution of ~30 km and an hourly temporal resolution (Hersbach et al., 2020). Geopotential height data at 500 hPa and air temperature at 850 hPa were used for each case study to complement the view of the atmospheric dynamics. The same methodology was applied to selected surface variables, specifically sensible heat flux, soil moisture in the top 7 cm, and 2 m temperature. Data for these variables were downloaded every 3 h, then averaged daily, with the maximum daily temperature at 2 m also calculated. As additional data we use monthly anomalies for 2023 of 2-meter maximum temperature (°K) and volumetric soil water layer 1 (m³/m³).

In this study, we adopt the definition established by WMO (2024), a definition that has practical applications and has been used widely in several studies worldwide, including those in Brazil (Perkins et al., 2012; Geirinhas et al., 2018; Lavaysse et al., 2018; Marengo et al., 2021; Coronato et al., 2024; Firpo et al., 2012; Feron et al., 2019; Lipfert et al., 2024; Valverde and Barbio, 2023). A HW, as per this definition, is a period of unusually hot weather that persists for at least three consecutive days during the warm period of the year, considering the local climatological conditions. Accordingly, HWs were defined as periods characterized by daily maximum and minimum air temperatures above the corresponding climatological 90th percentile for the month for three or more consecutive days (Marengo et al., 2021). Based on the standard base period 1991–2020, the climatological percentile thresholds are used throughout this study. It is worth noting that the meteorological services of Brazil, Paraguay, Bolivia, and Argentina also use similar definitions for HWs for their operational activities, highlighting our research immediate relevance and usefulness of our research.

Figure 1 shows periods with a sequence of warm spells (<3 days where the maximum and minimum temperature reached the 90th percentile) and HWs (at least three or more consecutive days where the maximum and minimum temperature reached the 90th percentile) for the four countries. This was done using data from 54 stations in Brazil, Paraguay, Bolivia, and Argentina. We applied the criteria explained in Section 2.2 to define an HW. Based on that, we chose a sequence of days when 7 extended HWs affected all countries: Event 1: August 1–12, Event 2: August 19–September 1, Event 3: September 16–30, Event 4: October 3–8, Event 5: October 15–25, Event 6: November 6–21, and Event 7: December 6–19, 2023.

The HW indices used in this study are shown in Table 1 and were defined by Frich et al. (2002) and Dunn et al. (2024). They were computed using daily maximum and minimum temperatures applied in 10 representative locations in central South America, using the CPC NOAA temperature data set extrapolated for the latitude-longitude nearest point of each station.

The calculation for the base period (1991–2020) requires a bootstrap procedure to avoid possible inhomogeneity across the in-base and out-base periods, as described in Zhang et al. (2005). These indices are calculated using the CPC Global Unified Temperature data NOAA data for a grid box nearest to each meteorological station’s location provided by the four countries’ meteorological services. The Theil-Sen estimator is a trend based on the median of slopes of all pairs of lines (Theil, 1950; Sen, 1968) from 1 January 1979 to 31 December 2023 at each location. The Theil-Sen trend is more robust and less sensitive to outliers compared to a simple linear regression (Fernandes and Leblanc, 2005). The Mann-Kendall MK test (Mann, 1945) was used to test the existence of a significant trend (significance level of p < 0.05). The MK test is a non-parametric test appropriate for non-normal distributions.

The HWMId index is defined by Russo et al. (2014) as the maximum magnitude of the HW in a year. The identification of HWs over our region of interest, together with the respective magnitude is based on ERA5 2 m maximum temperature. Following previous works (Perkins et al., 2012; Russo et al., 2014; Xie et al., 2020): first, the 90th percentile of maximum air temperature (CTX90pct) of each day of the year is estimated over a 15-day window centered on the respective calendar day; these daily 90th percentiles are estimated over the 1991–2020 period; HWs occur whenever the daily maximum 2 m temperature exceeds the respective CTX90pct for at least 3 consecutive days. In short, for a given day d, the CTX90pct threshold is the 90th percentile of the set of data Ad defined by Equation 1: (1) A d = 2020 U y = 1991 d + 15 U i = d − 15 T y , i

Where ⋃ denotes the union of sets and Ty, i is the daily Tmax of the da I in the year y.

The HWMId magnitude is then defined in Equation 2 as the sum of the magnitude of the consecutive days composing a HW, with daily magnitude calculated as follow: (2) M d = { T d − T 30 y 25 p T 30 y 75 p − T 30 y 25 p if T d > T 30 y 25 p 0 if T d ≤ T 30 y 25 p

Where Td is the maximum daily temperature on day d of the HW, T30y25p and T30y75p are the 25th and 75th percentile values, respectively, of the time series composed of 30-year annual Tmax within the reference period 1991–2020, estimated as the respective percentiles over dataset A_d (Equation 1). The slope of the Md (Td) is defined at each specific location depending on T30y75p and T30y25p which are different in places with different climates.

A compound weather or climate event results from the combination of multiple drivers and/or hazards that contribute to societal or environmental risk (Seneviratne et al., 2014; Zscheischler et al., 2018; Catto and Dowdy, 2021). Here, a compound drought-heat event was defined as a HW episode that occurs under a month with drought conditions [i.e., a month with an associated Standardized Precipitation Index (SPI < −1)] (Seneviratne et al., 2014; Geirinhas et al., 2021; Zhang et al., 2022). The key to determining occurrences of compound drought-heat events relies on precipitation trends since, in a warmer climate, local warming will be large enough that droughts will always coincide with at least moderately hot extremes (Bevacqua et al., 2022).

We define drought-heat compound events considering drought conditions, as defined by Standardized Precipitation Index SPI < −1 (McKee et al., 1993), and HW events, as identified using the Heatwave Duration Index (HWDI) defined in Table 1. SPI < −1 includes categories extremely dry, severely dry and moderately dry categories. To quantify the precipitation deficit, we used the SPI defined from the rainfall dataset from CPC NOAA. A selected parametric distribution determines its accuracy and is designed to quantify the precipitation deficit for multiple scales, reflecting drought’s impact on the availability of different water resources (WMO, 2024). Drought categories derived from SPI (abnormally, moderate, severe, extreme, and exceptional drought) are selected to be evaluated under HW conditions over time and space.

The compound events here coincide with the drought categories and HW conditions (Calim Costa et al., 2024). The compound is built with SPI-3, which compares the precipitation over a specific 3-month period with the precipitation totals from the same 3-month period for all the years included in the historical record. This was chosen to reflect the short and medium-term moisture conditions and the sum of days with HW (HWDI). The area of the compound is the match between extreme categories of SPI-3 and HW duration above 2 weeks. The 3 months drought memory described by SPI-3 show a greater extent in the area combined with the HW duration, which means a greater extent of combined effect. The seasonal scale of drought reflects better the surface conditions that enable and favor the HW events.

Most northern and central South America were affected by heatwaves, with areas impacted by high temperatures A recently published report from the World Meteorological Organization (WMO, 2024) showed that the annual mean temperature anomalies during 2023 in central South America varied from +1.0°C to +2.5°C relative to 1991–2000. In Marengo et al. (2024), the observed mean warming in 2023 for Paraguay, Peru, Bolivia, and Brazil reached +1.63°C (relative to 1991–2020). In 2023, Brazil, Peru, Bolivia, Paraguay, and Argentina recorded their highest September temperatures after hot winter months. Temperatures have been above 40°C in parts of Brazil during the first week of September (INMET, 2023). From the total of 54 stations, we selected a subgroup of 10 stations representative of each country for the trend analysis. Figure 2 shows a time series of Tmax and Tmin from these 10 stations during August–December 2023. There is good agreement between observed and CPC-NOAA derived maximum and minimum temperatures, except for the Bolivian stations that show slight underestimation. Overall, the CPC-NOAA data captures relatively well the daily temperature variability in those and all stations relatively well. In addition, Table 2 shows the observed highest daily Tmax from August to December 2023 in some stations in Bolivia, Argentina, Paraguay and Brazil, as well as the date of the highest temperature and the respective 1991–2020 LTM.

Information from reports of the meteorological services, press agencies and websites were used together with Figure 2 and Table 2, to identify some episodes of hot spells and HWs in the region. We selected HW episodes from August to December 2023 in the four countries mentioned above. Additional background information comes from the State of the Climate in Latin America and the Caribbean 2023 (WMO, 2024), the work of Perkins-Kirkpatrick et al. (2024). The winter and spring 2023 periods were chosen to capture the HWs dynamics during an El Niño/warm Tropical North Atlantic and because a record drought affected the Amazon basin during that period (Espinoza et al., 2024). In Brazil, the year 2023 oversaw 10 HW. In Brazil and Bolivia, 7 HWs occurred between August and December when El Niño’s warming was more intense (Perkins-Kirkpatrick et al., 2024).

According to WMO (2024) and the meteorological services of Argentina, Brazil, Bolivia, and Paraguay, during the second half of September, countries including Brazil, Peru, Bolivia, Paraguay, and Argentina recorded in 2023 their highest September temperatures. Temperatures in parts of wet central Brazil exceeded 41°C as South America was hit by scorching weather in the middle of the winter, and near-all-time high temperatures were recorded (Cuiabá, in Supplementary Table S1). Spring HWs are relatively common in Brazil (Geirinhas et al., 2017; Monteiro dos Santos et al., 2024b). Before 2023, the most recent and intense HW occurred in the spring of 2020 (Marengo et al., 2021; Libonati et al., 2022). In September 2023, it went closer to 44°C, and west central Brazil recorded the highest temperature in more than 100 years of measurements by the INMET.

Analyzing Table 2 and Figure 2, one can observe that the highest maximum temperatures were recorded in October and November, corresponding to anomalies of 6–8°C above average in some stations, reaching almost 10°C warmer in Belo Horizonte during September 2023. Espinoza et al. (2024) and Marengo et al. (2024) demonstrated that various HWs accentuated the warm conditions that affected central South America, including southwest Amazonia, in August and September 2023, amplifying the effect of the drought affecting the region at that time.

Table 2 shows the maximum temperatures of the warmest days (highest Tmax) between August and December 2023 observed in some stations in the four countries during HW episodes. Tmax in the warmest days in Brazil and Bolivia reached 9°C higher than the 1991–2020 LTM, while in Paraguay and northern Argentina, the hottest temperatures were 10–14°C warmer than the 1991–2020 LTM. It is noticed that the highest maximum temperatures were reached in October and November 2023.

South American countries also reported some other HW even before August 2023. The HW in Argentina from February 28 to March 20 was the most widespread event experienced since the HW observed in 2013 (WMO, 2024). It affected the country’s central and northern coastal zones (SMN). On March 11, Buenos Aires in Argentina recorded its highest-ever temperature for March at 38.6°C, while Mercedes in Uruguay saw the temperature reach a record value of 40.5°C.⁵ One HW event was observed between the 21st and 28 of July 2023, but it was restricted to areas west of Central-West and South Brazil. In July, Brazil was hit by record temperatures of almost 42°C, despite still being in winter.⁶

Figure 3 presents the spatial pattern of the heatwave Magnitude Index (HWMI), which reflects the accumulated magnitude (Md) of the seven selected heatwave (HW) events during 2023 (Section 2.4). The map highlights the spatial distribution and intensity of HWs across South America. Northern and western South America consistently exhibit the most pronounced HWMI values across the seven events. The late August period (Event 2, 19 August–1 September) exhibits the most intense HW activity, with significant HWMI values. HW Event 6 (06–21 of November) stands out for the most widespread extent of HW events (From 5°N to 30°S), with almost all South American countries affected by HWs. Ecuador (positioned near the equator line and the most western country) and Northeast of Brazil (latitude from 0° to 10°S and longitude below 50°W) display consistent Md values across all the selected HW events. Regions with latitudes higher than 40°S do not exhibit any HW occurrence in the selected periods.

Figure 4 presents the daily evolution of the magnitude of the HWMI for 10 selected cities during the HW events between August and December 2023. Most peak Md values across these cities are near 3, indicating that the temperature anomaly with respect to the 25th percentile on a specific day is three times the inter-quartile range (IQR). Notably, Cuiabá, Brazil, recorded the highest Md values, exceeding 3, during the Event 7. San Ignacio de Velasco, Bolivia, is the only city where positive Md values were observed during all HW events, with the highest magnitudes during Events 5 and 6. The Event 6 event impacted all 10 cities, with Md values between 2 and 3 for most locations, except Formosa, Argentina, which remained below 2. Brasilia, Brazil, and Formosa, Argentina, were the cities that have experienced only 3 of the HW events and Las Lomitas, Argentina, showed the minimum number of days with HW events.

Considering the events individually: the Event 1 reached all cities except Brasilia; the Event 2 only five cities (i.e., Manaus, Rio Branco, San Ignacio de Velasco, Cuiabá and São Paulo); the Event 3 did not reach any of the selected cities; the Event 4 reached only the selected cities in the North of South America (i.e., Manaus and Rio Branco); the Event 5 most of the central of South America, corresponding to five cities (i.e., Rio Branco, San Ignacio de Velasco, Cuiabá, S. J. Chiquitos and M. Estigarribia); the Event 6 reached all cities and the seventh only did not impact the selected cities in the North of South America and Formosa (the selected city most far south).

Figures 3, 4 show that the 2 most important HW events were Event 2 (August 19 to September 1) and Event 6 (November 6–21). The former showed the most intense activity and occurred in wintertime, while the latter (Event 6) covered almost all South American countries east of the Andes. According to the meteorological services of Brazil, Argentina, Bolivia and Paraguay the HW Event 2 affected most of tropical South America, leading to temperatures in many areas above 35°C in midwinter with some locations setting all-time heat records (Washington Post, 2023), with maximum temperatures above 35°C in Bolivia, Argentina, and Central Brazil. 12 August saw Rio de Janeiro break a 117-year heat record. Chile saw highs toward 40°C and Bolivia saw temperatures rise badly Asunción saw 33°C (The Guardian, 2023). Monthly records were broken in Villamontes (45°C) and Yacuiba (41.8°C) on August 22, Guajaramirim (39.0°C on August 24), Ascension de Guarayos (39.7°C) and Cobija (38.6°C) on August 31 in Bolivia. Between 22 and 28 August, in Cuiabá, in central-western Brazil, the temperature reached 41.8°C on August 23. The HW also hit Rio de Janeiro and São Paulo, Brazil’s two most populous cities. Rio de Janeiro’s temperature reached 38.7°C on August 22, and in Cuiabá, in central-western Brazil, the highs reached 41.8°C (the year’s highest temperature) on August 23. In São Paulo, thermometers hit a high of 32.3°C on August 23, very close to the highest temperatures registered for August of the order of 33.1°C on August 31, 1952, and 1955. The temperatures above average and rainfall below average favored a raised risk of forest fires (INMET, 2023).

Event 6 was more intense in southwestern Brazil, Paraguay, and Bolivia, with maximum temperatures of 40°C and above. A HW in Brazil started in central and western regions on November 7, when temperatures reached 42°C in Mato Grosso do Sul, and on November 8, 2023, Brazil was hit by another heat wave. Rio de Janeiro had the warmest day of the year, with temperatures reaching 42.5°C on November 12 (INMET, 2023). The record established in Ascension de Guarayos in October was equaled on November 8. Puerto Suarez (43.5°C on November 16). San Borja (41.2°C on November 10), San Jose de Chiquitos (44.1°C on November 8, breaking the record of October) and Yacuiba (44.9°C on November 12) recorded the historical highest temperatures, with several other Bolivian stations recording November highest temperatures. Argentina experienced the hottest summer in 60 years between November 2022 and March 2023, with 10 HW (Collazo et al., 2023). In western Paraguay, on October 23 and 8 and 12 November, the temperature reached 44.5°C in Mariscal Estigarribia and 41.8°C in Puerto Casado (the average monthly maximum is 35.1°C and 33.4°C, respectively) on September 26 and November 11. In Asunción, it measured 41.2°C on November 12. This same HW affected central and southern Brazil, and according to the Brazilian Meteorological Service (INMET, 2023), about 120 stations recorded their highest maximum temperatures on November 12, a record for November. The highest maximum was recorded in Rio de Janeiro, which registered 42.5°C on November 12, followed by Cuiabá with 39.6°C and Teresina with 38.9°C. São Paulo had the highest maximum in the last 9 years, with 37.1°C on November 12 and 37.7°C on November 15. In Rio de Janeiro, temperatures reached 42.5°C on November 12, which became the hottest day of the year. On November 15, INMET thermometers repeatedly registered temperatures above 40°C across the central-west and southeast, with a peak of 43.3°C in Corumbá on the border with Bolivia on that day. With temperatures hitting 37.7°C and 42.6°C on November 15, respectively, São Paulo and Rio de Janeiro broke yearly records. On November 8, maximum temperatures reached 46.0°C in Las Lomitas in Argentina. On November 17, 2023, around a thousand people fainted during a show of American country singer Taylor Swift, held at the Nilton Santos Olympic Stadium, in Rio de Janeiro (Metropolis, 2023), when a person suffered a cardiorespiratory arrest and died during the concert (BBC, 2023). On that day INMET registered maximum temperatures of 39.1°C. On November 18, 11 meteorological stations recorded a maximum temperature ≥40°C. Such extremely high temperatures occurred in a prefrontal situation and at the end of a heatwave that lasted 10 days.

The seven HWs occurring in the second half of 2023 in South America were experienced throughout the continent, as shown in Figure 4. These correspond to events that occurred on August 1–12, August 19–September 1, September 16–30, October 15–25, November 6–21 and December 6–19. Analysis of the atmospheric circulation anomalies allows us to identify that five HWs presented similar conditions on the days before and concurrently with each event. Only Event 2 (on August 19, 2023) presented a different pattern in the synoptic circulation before and during the event.

Figure 5 shows anomalies of the 850 hPa air temperature (shades) and of the 500 hPa geopotential height (contours). Letter H marks positive or high-level pressure anomalies, while L marks negative or low-level ones. The “Pre-event” period considers the 3 days preceding the HW episodes, while “event” considers the HW participating days. Vectors represent 850 hPa wind anomalies. For events 1, 3, 5, 6, and 7, at lower levels of the atmosphere, during the “pre-event” shows positive geopotential anomalies centered over central Chile and Argentina (light green contours in Figure 5). The anomalies are negative in areas of the adjacent Pacific and Atlantic and southernmost Argentina and Chile. However, a broad positive anomaly in the Pacific can be seen at 100°W.

During pre-event 1, a spatial correspondence between the 850 hPa temperature and maximum temperature at 2 m anomalies was evident over the region around 35°S and 70°W (Figure 5 and Supplementary Figure S1). In addition to other mechanisms such as the radiative heating this signature reflects the contribution of subsidence and adiabatic heating processes to the near surface temperature escalation. When analyzing the soil moisture anomalies (Supplementary Figure S2) one can easily note that the areas covered by the highest temperatures were associated to soil moisture values within the expected levels. Accordingly, the sensible heat anomalies were negligible. During event 1, although positive anomalies of 2 m Tmax were witnessed around the same location, they were less intense. An enhanced anticyclonic circulation pattern remained stationary over this area from pre-event 1 to event 1, triggering high evaporation rates. This resulted in a steady and prolonged soil moisture decrease, paving the way for the pronounced negative anomalies observed during event 1 (Supplementary Figure S2). Despite this soil drying, the sensible heat anomalies were still residual indicating that soil moisture was sufficient to meet the atmosphere demand for water.

Overall, the atmospheric and land surface conditions recorded during pre-event 2 were very similar to those observed for pre-event 1. Again, the areas with strong positive anomalies of 2 m Tmax were also characterized by above normal values of 850 hPa temperature. The sensible heat anomalies remained negligible pointing. During event 2, as the high-pressure anomalies moved eastwards, temperature anomalies approached lower positive values with any significant change in the soil moisture and sensible heat anomalies.

The days before events 3, 5, 6, and 7 were characterized by an anomalous low-pressure system located over the southeastern section of the continent that was responsible for cooler near surface temperatures in the region (Supplementary Figure S1). Later, during the HW participating days, this low-pressure system was replaced by an anticlyclonic circulation pattern associated to positive anomalies in the 850 hPa temperature and 2 m Tmax. All the region spanning from north to central South America, covering the Amazon basin and the northeast Brazil, witnessed extremely warm conditions at surface because of this synoptic pattern. The land-atmosphere interactions observed for these four events were quite different from the ones previously described for events 1 and 2. These HW episodes occurred in a later stage of the year, after a long sequence of days marked by large precipitation deficits and evaporation rates after the wintertime dry season. Together, these conditions led to an amplification of the soil desiccation firstly noted during events 1 and 2, resulting in a severe drying during HW events 3, 5, 6, and 7 (Supplementary Figure S2). The co-occurring conditions of soil dry-out and high income of shortwave radiation at surface, promoted by the anomalous anticyclonic circulation patterns, fueled large sensible heat fluxes from surface to the atmosphere (Supplementary Figure S3). This is a clear signature of a strong soil moisture-temperature coupling that contributed for the re-amplification and self-propagation of the pre-established HW and drought conditions in the region.

Regarding the average position of the extratropical high-pressure anomalies, positive temperature anomalies (red colors in Figure 5) are detected in the blocking anticyclone (blocking highs) region, particularly from the north of the Chilean coast to the Ecuadorian region. During the next few days, the “event” situation shows an eastward propagation of the positive geopotential anomalies, reaching the southwest of Brazil and the adjacent Atlantic Ocean (light green contours in Figure 5).

The type of blocking observed in Figure 5 seems to fit the High-over-Low (Diffluent Block) type with a closed high-pressure center to the north of a closed low-pressure center in the south (Detring et al., 2021). A high-pressure anomaly in southeastern South America and southward low-level wind anomalies are detected east of the Andes from Bolivia. Positive temperature anomalies at 850 hPa are found over southeastern South America because of adiabatic warming and subsidence. The persistence of north–south wind conditions at the eastern flank of the Andes for several days during the HW implies that there were no cold winds from the high latitudes to the continent. In addition, the positive temperature and geopotential height anomalies suggest the effect of subsidence anomalies that are commonly associated with the absence of clouds and rain, higher radiation reaching the surface and favoring dry conditions (Geirinhas et al., 2018).

The circulation described during the “pre-events” and “event” situations resembles the South Atlantic Convergence Zone (SACZ) or the South America Low-level jet east of the Andes (SALLJ) episodes (Lanfredi, 2018). The SCAZ and SALLJ play important role on surface warming, and an active SACZ leads to subsidence and anticyclonic circulation over central South America (Alvarez et al., 2019). Previous studies showed that during HWs in central South America in the spring of 2020, southeastern South America in January 2022, and Argentina during the 2022–2023 summertime, the atmospheric circulation pattern had the presence of quasi-stationary extratropical high-pressure anomalies and persistent atmospheric blocking of westerly flow (Marengo et al., 2021; Zhang and Xie, 2023; Suli et al., 2023; Collazo et al., 2023; Perkins-Kirkpatrick et al., 2024). Persistent anticyclones or blocking in the midlevel atmosphere over South America are necessary for summertime continental HWs. These atmospheric systems lead to clear skies and subsidence, facilitating warming together.

Domeisen et al. (2022) and Baldwin et al. (2019) identify various local and remote drivers and feedback influencing HW development. Extratropical atmospheric blocking and global land-atmosphere coupling are the most pertinent. During spring of 2023, while the Amazon region was experiencing a record drought, HWs in this season in central and southern Amazonia were favored by negative anomalies in evapotranspiration and soil moisture, with clear skies and increased temperatures due to the blocking anticyclonic anomalies in central South America. This occurred during the dry-to-wet transition season September–October–November 2023, with warm SSTs in the equatorial Pacific and tropical Atlantic (Marengo et al., 2024; Rodrigues and Woolings, 2017).

At the seasonal level, the SPI-3 better represents the extension of the drought (Figure 6). The extent of drought-stressed areas started to increase in June (above 20%), with maximum values (above 50% of South America) in November. The highest extension of the compound event is detected for SPI-3 values along the September–November season, with more than 40%. For SPI-6, November and December show the largest area. Figure 7 shows the drought-heat compound area in South America during 2023.

We now focus our analysis on Events 2 (August 19–September 1) and 6 (November 6–21) due to their circulation features during winter season and during the driest month in 2023, respectively. In the case of Event 2, drought-heat compound events were identified over northern South America. The SPI-3 shows regions with extreme and exceptionally dry conditions with HW lasting 2–4 weeks over northern and western Amazonia and severe to moderate dry conditions and HW lasting 2–3 weeks in parts of Bolivia, northern Argentina and Paraguay (Figure 7A). The area affected by HWs with a duration of 3 days or more was 82.8%, covering regions with and without drought. The extent of the drought-heat compound during event 2 was 37.9% of South America area.

In November for the Event 6, the extent of the drought-heat compound event was 50.5% of South America (Figure 7B). Extreme and exceptional dry categories occurred mainly over tropical and subtropical South America east of the Andes covering most of tropical Brazil, Bolivia, Paraguay and northern Argentina, associated with HWs duration of 2 or more weeks. From both events, the Event 6 shows the highest intensity, with extreme drought and longer HWs on a larger area when compared to Event 2. The extent of the area with HW duration of 3 days or more is 76.4%, mainly in areas under some level drought condition.

To investigate changes in surface conditions during HW episodes, Supplementary Figure S1 shows 850 hPa Tmax anomalies between 2 and 3°C above normal in central and western Amazonia and in Bolivia, Paraguay and Northern Argentina in August 2023 (Event 2). In Event 6 (November 2023), tropical South America between 10 ^oS and 30 ^oS east of the Andes shows warming between 2 and 3°C, with 3–4°C warner than normal in southern Amazonia and Bolivia. Supplementary Figure S2 shows negative soil moisture anomalies in northern Amazonia from July 2023 onwards, and in Paraguay and northern Argentina in October and November 2023. In November 2023 (Event 6), the entire Amazon region presents negative soil moisture anomalies. This is consistent with Marengo et al. (2024), where the negative latent heat flux anomalies are in line with the detected negative soil moisture shown here as well as with the intense warming in Amazonia during the drought of 2023 in austral spring September–November (Supplementary Figure S4).

Supplementary Figures S5, S6 show the SPI-3 and HWs duration respectively, from January to December 2023. Supplementary Figure S7 shows compound drought-heat events in South America, where the most active season occurs from August to December 2023, coinciding with intense HWs. The largest areas with compound events are November during the Event 6.