Assessing the role of compound drought and heatwave events on unprecedented 2020 wildfires in the Pantanal

Burned area (BA) was obtained from two main sources. Monthly values were obtained from the MCD64A1 collection 6 derived from the MODIS (moderate resolution imaging spectroradiometer) sensor at 500 m spatial resolution from 2001 to 2020 [28]. For improved accuracy on day-to-day variability of BA [29], daily values for 2020 were obtained through the ALARMES dataset with a 500 m spatial resolution using images from the visible infrared imaging suite imager sensor [29].

Meteorological parameters, including maximum temperature (Tmax), precipitation, surface net solar radiation, geopotential height and temperature at several levels of the atmosphere were extracted, at daily scale, from the European Centre of Medium-range Weather Forecast ERA-5 reanalysis dataset [30]. Soil moisture, evaporation and potential evaporation, at daily scale, were obtained from the Global Land Evaporation Amsterdam Model (GLEAM v3.5a) [31, 32]. All variables were retrieved at a gridded 0.25° × 0.25° spatial resolution and the composite anomalies were computed with respect to the climatological seasonal cycle (1981–2010).

Surface meteorological fire danger conditions were evaluated using the fire weather index (FWI) [33], allowing summarizing the chances of a fire to ignite and propagate and to foresee hazardous fire conditions [34]. The FWI product is provided by the Copernicus Emergency Management Service [35], computed with meteorological fields from the ERA5 reanalysis [36]. Daily values were obtained for the historical period (1980–2020) on a regular grid of 0.25° × 0.25° resolution [37]. All analyses were carried out for the Brazilian sector of the Pantanal wetland.

2.2.1. Fire analysis

To assess the exceptionality of the P20Fs we considered the ratio between the total BA in 2020 and the respective mean BA for the 2001–2019 period. We also estimated the fire return period, defined as the ratio between the 20 years that encompass our study period (2001–2020) and the annual recurrence. Finally, we computed the 75th percentile (P75) of the 2001–2019 period and the percentage of the 2020 BA with no fire and low recurrence (1–2 years). The above-mentioned metrics were computed for each of the nine hydrological subregions of Pantanal [38], to evaluate regional discrepancies within the biome (figure 1(a)).

Zoom In Zoom Out Reset image size

2.2.2. Heat wave identification

HW was defined as a period of three or more consecutive days with daily Tmax values above predefined climatological (1981–2010 base period) percentiles (80th, 90th and 95th) of Tmax for each calendar day (on a 15 day moving window). Based on this definition, a secondary metric was computed: the percentage of the Pantanal domain under HW conditions ($\% {\text{ Pantana}}{{\text{l}}_{{\text{HW }}}}$). This method was already used in previous studies conducted for the USA [39] and Brazil [27] and consists of determining the yearly percentage of the total Pantanal cells (${\text{cellsPA}}{{\text{N}}_{{\text{total}}}}$) that experienced HW conditions:

$$\begin{align}\% {\text{Pantana}}{{\text{l}}_{{\text{HW }}}} = { }\frac{{{\text{cellsPA}}{{\text{N}}_{{\text{HW}}}}}}{{{\text{cellsPA}}{{\text{N}}_{{\text{total}}}}}} \times 100{ }{\text{}}.\end{align} \tag{ 1 }$$

Per year, the number of total cells (${\text{cellsPA}}{{\text{N}}_{{\text{total}}}}$) is obtained by considering the total number of grid-points within the region (${\text{cellsPA}}{{\text{N}}_{{\text{region}}}}$) and the hypothetical total number of days that could experience HW conditions (${\text{cellsPA}}{{\text{N}}_{{\text{time}}}}$):

$$\begin{align}{\text{cellsPA}}{{\text{N}}_{{\text{total}}}} = {\text{cellsPA}}{{\text{N}}_{{\text{region}}}} \times {\text{cellsPA}}{{\text{N}}_{{\text{time}}}}.\end{align} \tag{ 2 }$$

In our particular case, the ${\text{cellsPA}}{{\text{N}}_{{\text{time}}}}$ corresponds to the total number of days of the fire season in the Pantanal (July to October [3]). For instance, for a particular year, a percentage of 100% indicates that all the Pantanal experienced HW conditions during all the fire season days.

2.2.3. Drought conditions

2.2.4. Relating fires with the heatwave/drought conditions

The P20Fs show an increase in BA for almost all subregions ranging from ∼60% to 1190% of the historical mean value (figure 1(b)). Higher ratios are found in the northern subregions, namely São Lourenço (II) and Cuiabá (IV), which burned ∼65% and 55% of their area in 2020 (table S1 available online at stacks.iop.org/ERL/17/015005/mmedia), respectively. These values were absolute outliers within the historical series (figure 1(c)), as so far these subregions had burned a yearly average of ∼5.1% and 5.8% (table S1), respectively. In the P20Fs only one subregion burned less than its annual average over the 2001–2019 period: Negro de Mato Grosso do Sul (VIII); which, along with Miranda (III) and Baixo Paraguai (V), obtained the lowest ratios to historical mean values (figure 1(b)). Historically, the northern regions are characterized by lower return periods, whereas the southern regions burn more regularly (figure 1(c)). However, this historical tendency was reversed in 2020, when most of the BA was in forested areas of northern Pantanal. Conversely, southern and south-eastern subregions, characterized by large extents of pasture and grasslands (figure 1(a)), burned considerably in 2020 but did not reach record levels. Nevertheless, with the exception of Negro de Mato Grosso do Sul (VIII), the BA from the P20Fs went above P75 of the historical time series for all southern subregions (table S1).

Most subregions in the Pantanal burn within a 4 month period from July to October (figure S1) and, in this regard, 2020 kept as expected: a steady BA increase from July to September is seen in Pantanal, with a peak on 12 September (116 605 ha) and a secondary observed on 27 September (95 478 ha; figure S1). Médio Paraguai (VII) and Taquari (IX) showed the earliest signs of burning in July, while the remaining subregions burned over August to October, and solely Baixo Paraguai (V) and Médio Paraguai (VII) showed considerable BA in the earlier weeks of November. The latter subregion burned consistently over a period of 5 months, severely contrasting with its historical series where BAs mainly occur in September and October. It is also worth noting how Médio Paraguai (VII) burned very little in previous years (2016–2018; figure 1(c)).

Around a third of the BAs in the P20Fs had been undisturbed since 2001, and another 31% burned only once or twice over the entire study period (table S1). Of the entire P20Fs, 64% of BAs were areas not accustomed to regular and systematic burning. Noteworthy are the cases of Cuiabá (IV) and Médio Paraguai (VII) with ∼18% and 19%, respectively, of areas that had not or barely burned within the last 19 years.

Results show unprecedented extreme heat conditions, with Tmax anomalies for the last two fire seasons over the Pantanal (2019 and 2020) positioned in the high-end tail of the empirical distribution of average Tmax anomalies (figure 2(a)). By contrast, the years 1992, 1990, 1984 are in the low-end tail, as in general, the years within the first half of the analysis period. The time series of Tmax (figure 2(b)) is characterized by a pronounced and statistically significant positive trend of 0.76 °C per decade, responsible for warming throughout the last four decades of ∼3 °C. Accordingly, the spatially averaged Tmax level during the P20F season was 34 °C, roughly 4 °C higher than the average for the first decade in the 1980s. The percentage of the Pantanal under HW conditions (figure 2(c)) followed, closely, the Tmax evolution (figure 2(b)). Because of this sharp warming trend, the spatial and temporal signature of HWs had marked increase, with unprecedented extreme heat conditions in 2020 as well. Analyzing the monthly SPI-6 values from 1980 and 2020 (figure 2(d)), one concludes that during the 21st century most of the fire seasons were preceded by the occurrence of precipitation deficits. As previously described, this period also marks a sharp increase in the Pantanal under HW conditions (figure 2(b)), indicating that after the turn of the century the CDHW conditions became more frequent, in particular for 2020. Accordingly, 2020 was also marked by record fire danger (figure 2(e)): fire season averaged FWI reached values above 30 for the second year in a row. Previously, 2010 held the highest value, consistent with widespread drought conditions in neighboring biomes [41, 42]. Higher fire danger values over the last two decades strongly contrast with those of the 20th century, with a significant positive trend over the last 40 years.

Zoom In Zoom Out Reset image size

In general, 2020 was marked by the occurrence of numerous HW episodes over the Pantanal when the daily area-averaged Tmax values were considerably above the expected levels for several periods of three or more consecutive days (figures 3(a) and S2). Thus, several HPs were also observed, particularly during the fire season. The first HP occurred from 26 August to 1 September, the second from 5 to 20 September and the third from 25 September to 15 October (red boxes in figure 3(a)).

Zoom In Zoom Out Reset image size

Throughout 2020, a temporal match between the occurrence of HPs and increasing values of BA (figures 3(a) and S2) was observed. However, it was during the austral winter and the three considered HPs that this temporal correspondence was more pronounced, indicating a close relationship between the induced atmospheric heat-stress conditions and the occurrence of fires. On average, the Tmax value for the three HPs was 38.5 °C, representing a staggering temperature anomaly of about 5.8 °C. In fact, on 1 October (the 6th day of the third HP) the mean Tmax value reached 41 °C, establishing a new record-breaking level for the region. A very similar value was observed nine days later on 10 October, defining this as a period of outstanding extreme heat stress conditions. During this HP of 21 d, the Tmax values were on average 6.5 °C higher than the expected mean levels and a total of 983 900 ha burned, a value that accounts for 25% of the total BA recorded during 2020 in Pantanal. The BA recorded over the entire Pantanal during these three massive HPs accounted for 55% (60%) of the total 2020 (fire season) BA. In all subregions, with the exception of Baixo Paraguai, the BA observed during the three HPs accounted for more than 50% of the amount from the fire season. Moreover, in six of the nine subregions, this BA amount corresponds to more than two-thirds of the fire season, reaching 95% in Miranda (figure S2).

The months preceding the 2020 fire season were marked by large deficits in precipitation (figure 2(d)), within the drought period. During the P20F season, precipitation levels were lower than expected, reaching zero or near-zero values for most of the days (figure 3(b)). Thus, the drought pattern and soil desiccation that initiated during the first months due to a drier wet season substantially amplified throughout the following months, leading to extreme negative anomalies of accumulated soil moisture (figure 3(c)). These precipitation deficits combined with clear sky conditions that were linked to large amounts of incoming shortwave radiative energy at the surface and enhanced diabatic processes (figure 3(a)), induced large evaporation rates from the surface to satisfy the high atmospheric demand for water. This combined process was crucial for the establishment of the pronounced soil moisture deficits and evaporative stress observed during the P20F season.

Concurring warm and dry conditions controlled the partitioning of water and energy fluxes at the surface. The evaporative fraction observed during 2020 followed very closely the precipitation and temperature regimes (figures 3(c) and S2). Several periods marked by a sharp decrease in the evaporative fraction values were clearly paired with dry episodes combined with extremely hot conditions. Thus, negative anomalies of the evaporative fraction were a constant presence during 2020 (figures 3(c) and S2). However, it was during the fire season that the values reached their minima indicating the presence of a strong soil moisture-temperature coupling regime (water-limited) in which disproportional surface losses in the incoming shortwave radiation through upward sensible heat flux allowed a re-amplification of the near-surface (air) temperatures. The atmospheric cooling through latent heat flux was then suppressed as well as the capacity of the surface to mitigate the low atmospheric humidity levels.

Finally, we evaluate the synoptic conditions that triggered the development of such CDHW events (figure 4). The spatial pattern of the 500 hPa geopotential height anomaly field indicates the presence of concentric positive anomalies during the second and third HPs over the Pantanal (figures 4 (b) and (c)). During the first HP, positive anomalies were also observed. However, they resulted from a northwest extension of the high-pressure system located over the South Atlantic Ocean (figure 4(a)). Exceptional low-tropospheric heating was also recorded as it can be observed by analyzing the 850 hPa temperature anomaly field. These conditions represent an enhanced anomalous anticyclonic circulation pattern over the Pantanal. This continental high-pressure anomaly was widespread and responsible for the air subsidence, causing pronounced adiabatic heating at the surface, through air compression, as well as the persistent clear sky conditions that promoted enhanced diabatic heating at surface (figure S3), low levels of humidity and the absence of precipitation episodes. Therefore, the ideal synoptic conditions for strong atmospheric heating and large evaporation rates were present throughout the P20F season, in particular during the third HP, when the Tmax values were, on average, 6 °C above the expected levels (figure 4(f)). Changes in the low tropospheric wind configuration were also observed, showing the signature, close to the surface, of this anticyclonic circulation pattern. During the two first HPs, it can be observed that the wind pattern presented a higher-than-normal northeast–southwest orientation (figures 4(d) and (e)). This anomalous wind pattern was marked by a confluence throughout a north-south oriented asymptote towards south Paraguay (during the first HP), and throughout a northwest-southeast oriented asymptote towards southeastern Brazil (during the second HP). In fact, by analyzing the mean of the observed wind configuration recorded during these two periods (figures S3(a) and (b)) one may conclude that air masses predominantly from the northeastern regions moved towards the Pantanal. During the third HP the 925 hPa wind pattern was substantially different (figures 4(f) and S3), showing an anomalous northwest-southeast orientation over the Pantanal. Nevertheless, a pronounced confluence similar to the one observed during the second HP was present. In fact, the asymptotes marking these regions of strong confluence were, for all the analyzed HP's, oriented towards the regions where the anomalies of Tmax were higher. This could indicate that the intense daytime heating in the low troposphere over these regions caused the lifting of air, imposing pronounced changes in the normal near-surface wind configuration.

Zoom In Zoom Out Reset image size

Therefore, during three HPs, the ideal synoptic conditions, triggering high rates of potential evaporation from the occurrence of clear sky conditions linked to atmospheric subsidence (figure S3), were observed over central SA, particularly in the Pantanal. However, due to the desiccated soil already observed at the time (figure 3(b)), the surface could not meet such atmospheric water demand. This led to low rates of actual evaporation and, consequently, to pronounced evaporative stress in the region (figure 3(c)) when extreme low levels of evaporative fraction were observed during these periods. The spatial pattern of the SPI-6 values, computed from the months when these HPs occurred, confirms severe meteorological drought conditions (figures 4(g)–(i)). An approximately northwest-southeast oriented broad region extending from northern Bolivia to southeastern Brazil, with Pantanal in its center, endured pronounced negative SPI-6 levels from August to October (ranging from −1 to −4). The soil moisture deficits during the three HPs (figures 4(g)–(i)) confirm this situation and are spatially consistent with the analysis of figure S2 by showing the high potential of soil desiccation in inducing low levels of evaporative fraction. A similar situation was also observed southwards, particularly over southern Paraguay and over northern Argentina. It is noteworthy the spatial match between the regions with strong positive Tmax anomalies and areas with negative soil moisture anomalies, emphasizing CDHW conditions, unequivocally associated with the land-atmosphere feedbacks over these SA regions and particularly over all subregions of the Pantanal (figure S2).