Counting the Dead: 17 Million Vertebrates Directly Killed by the 2020’s Wildres in the Pantanal Wetland, Brazil

Anthropogenic factors have signicantly inuenced the frequency, duration, and intensity of meteorological drought in many regions of the globe, and the increased frequency of wildres is among the most visible consequences of human-induced climate change. Despite its role in determining biodiversity outcomes in different ecosystems, wildres can cause negative impacts on wildlife. We conducted ground surveys along line transects to estimate the rst-order impact of the 2020 wildres on vertebrates in the Pantanal wetland, Brazil. We adopted the distance sampling technique to estimate the densities and the number of dead vertebrates in the 39,030 square kilometers affected by re. Our estimates indicate that at least 16.952 million vertebrates were killed immediately by the res in the Pantanal, demonstrating the impact of such an event in wet savanna ecosystems. The Pantanal case also reminds us that the cumulative impact of widespread burning would be catastrophic, as re recurrence may lead to the impoverishment of ecosystems and the disruption of their functioning. To overcome this unsustainable scenario, it is necessary to establish proper biomass fuel management to avoid cumulative impacts caused by re over biodiversity and ecosystem services.


Introduction
Anthropogenic factors have signi cantly in uenced the increasing frequency, duration, and intensity of meteorological drought in many regions of the globe 1,2 , challenging the sustainability of life on Earth.
Extreme droughts caused by human-induced climate change have been pointed out as the global-scale determinant of the observed increasing wild re occurrence 3,4 . In fact, during the last few years we have been witnessing an astonishing increase in intensity and frequency of wild res, leading to a globally unprecedented amount of burnt area 3,5,6 . However, the impacts of wild re on wildlife are still poorly known, preventing our better understanding of the cumulative effects on the ecosystem functioning.
Besides the climate change implications for the wild re increased frequency, many ongoing drivers at regional scales are to blame as well, including short to long-term anthropogenic activities such as deforestation, incorrect ignition and use of re, absence of or inadequate landscape management strategies, vegetation encroachment, increased need of re as management tool, and release of greenhouse gases which, in turn, contributes to climate change 7,8,9,10,11,12 . In fact, during the last few years we have been witnessing an astonishing increase in intensity and frequency of wild res, leading to a globally unprecedented amount of burnt area 3,5,6 .
Seasonal wet-dry ecosystems, such as wet savannas, are particularly vulnerable to burning due to the higher vegetational load sustained by ood-induced fertility that makes these areas susceptible to burning during dry season 6, 13,14 . Recently, among the most dramatic wild re events, stands out the recorded widespread burning that hit the 170,000 km 2 Pantanal wetland, in the center of South America, which may be characterized as an extreme wild re event (EWE) 15 . While res burned 16,210 km 2 of the Brazilian portion of the Pantanal in 2019, astonishing 39,030 km 2 burned in 2020 16 .The large quantities of organic matter accumulated in lower, long-lasting ood areas covered by dense aquatic plant communities and oating mats, as well as in the open grasslands subjected to vegetation encroachment, created the scenario for the catastrophic 2020 wild re in the Pantanal 17,18 .
The scenarios of climate change for the Pantanal region indicate decrease in the amount of rainfall, higher temperatures, and higher frequency of extreme climate events 19 . Fire is an evolutionary driver that shapes landscapes, biodiversity, human behavior, and the dynamics of socio-ecological systems worldwide 20,21 , and a natural and important component in savanna ecosystems worldwide 22,23 . Among these savannas, the Brazilian Cerrado and Pantanal wetland are in uenced by re since before human arrival 24,25 , and are considered re-dependent ecosystems 20 . However, we still need a better understanding of the impact caused by wild res on the ecological functions and ecosystem services provided by wildlife. Additionally, this understanding is relevant for the process of sensitization of landowners, decision-makers, and the whole society towards the necessity of an integrated re management strategy for the region.
The effects of re on wildlife populations are generally classi ed as rst-order (direct or immediate), second-order (indirect), and evolutionary effects from re history 26,27 . Experiences on estimating impacts of large-scale wild re on wildlife are still scarce, and often do not separate mortality from other outcomes from this type of disturbance. Usually, estimates have been assessed based on known population densities or estimated biomass, depending on the available knowledge on speci c species or species groups 28, 29,30 . Although these assessments may be considered valid attempts of estimating the number of animals killed by re, they are not based on direct carcass counts. This paper focuses on estimating the deaths among vertebrates directly caused by wild res, based on data collected in the eld by accounting for carcasses up to 48 hours after re events in the Pantanal wetland, Brazil. With this report we want to contribute to the increase of awareness in society regarding the impact of such events on wildlife, as the scenario of climate change poses to humanity perhaps its major challenge in history. The estimates we present, besides its direct accounting in the eld, may help to evaluate the potential cumulative impact of eventually repeated wild res in ecosystems, as this is a plausible scenario posed by climate change. Survey protocol and data analysis -We conducted post re line transects to count vertebrates, using the distance sampling technique 41 to estimate the number of deaths in the 39,030 km 2 burned region of the Pantanal wetland during 2020 wild res 17 (Fig. 1). The transects were placed opportunistically, according to the re events at different locations between August 1st and November 17th, 2020, as well as timely accessibility and reduction of risks to the eld staff. To avoid removal of carcasses by scavengers, samplings were conducted within 72 hours after burning, but mostly within a 24-48 hours period. All eld biologists and technicians covering the burned region employed the same standardized sampling protocol. Distance sampling requires the measurement of the perpendicular distance between detected target objects and the transect line, as well as the transect length, to estimate densities based on the curve of detection probability; the premise is that detection probability decreases as the distance from the transect line increases 41 . All dead vertebrates detected along the transect line were identi ed at least to the Order level, and to lower taxonomic level whenever possible. We used the Distance 7.3 software 68 to estimate densities and the total number of dead animals, separating them into two groups regarding body size: small vertebrates (less than 2 kg) and medium to large vertebrates (2 kg and over). During the analysis process, we adopted these two groups as strata in the study area, obtaining separate estimates for small and medium to large vertebrates, as well as a pooled estimate for the entire area burned in the Pantanal.

Methods
Based on group and on pooled estimates, we estimated the number of dead vertebrates in subgroups by using the percentage of records of each sub-group in relation to the total number of carcasses recorded in the eld, as well as to the number of records by group. The small vertebrate subgroups correspond to amphibians, small lizards, small snakes, small birds, small rodents, and marsupials. Subgroups within the medium to large vertebrates were chelonians, large lizards, anacondas, caimans, medium-large birds, anteaters, armadillos, medium-large rodents, ungulates, and primates, which comprehend the vertebrate species detected in our surveys.

Results
Our survey in the Pantanal is the rst found in the literature that apply distance sampling to account for animals killed by wild res. We covered 126 line transects distributed from northern to southern Pantanal wetland ( Fig. 1), totalizing 114.43 km, from which 302 records of dead vertebrates were obtained. The effectively sampled strip along transects were estimated to be 2.72 ± 0.21 m wide for small vertebrates (CV = 7.69) and 7.28 ± 1.33 m wide for medium to large vertebrates (CV = 18.26), and the overall detection probability was 0.108 ± 0.084 and 0.29 ± 0.053, respectively. The tted probability model of the distribution of both small vertebrates and medium to large vertebrates was a Negative Exponential key, k(y) = Exp(-y/A(1)) with simple polynomial adjustments of order 2, 4, while for large vertebrates it was k(y) = Exp(-y/A(2)). As a result, we estimated that 16,009,000 ± 2,802,300 small vertebrates were killed by the wild res, as well as 943,830 ± 252,740 medium to large vertebrates, in the 39,030 km 2 burnt area from January to November 2020 (Table 1). In total, a pooled estimate of 16,952,000 (217.17 vertebrates per km 2 ) vertebrates died due to direct effect of the 2020 wild res in the wetland ( Fig. 1; Table 1,). The dead animals most frequently detected in our surveys pertain to the subgroups of small snakes, small birds, medium-large birds, and small rodents, when considering the pooled estimate for the deaths caused by the wild res (Table 2). However, when within groups are examined, call to the attention the number of large lizards, artiodactyls, and primates (Table 2). Among the specimens identi ed to Species or Genus level we found 1 tortoise, 5 amphibian, 3 small lizard, 2 large lizard, 1 anaconda, 9 small snake, 1 caiman, 11 small bird, 7 medium to large bird, 2 small marsupial, 2 armadillo, 1 anteater, 3 primate, 2 small rodent, 3 medium to large rodent, and 2 ungulate species; we identi ed the carcasses as pertaining to 55 different taxonomic entities at least to the Family level, from which 53 were identi ed at least to the Genus level. ( Table 3, in the Supplementary Information).

Discussion
The estimated numbers indicate an astonishing immediate impact of the Pantanal's 2020 wild res in the vertebrate communities, even considering that the estimates do not re ect the complete gure of mortality, as hidden (e.g., underground), delayed or second order effects certainly caused an unknown number of deaths. In fact, late mortality may be caused not only by body burns but also due to changes in the vegetation and the consequent impacts on resource quality, availability, and productivity at every trophic level, ultimately leading survivors to starvation 31,32,33,34,35 . Mortality also may occur due to increased predation during displacement from affected home ranges 31,36 . The negative consequences may be stronger for small populations or species that require more time to recover 36,37,38) . Thus, the overall impact of the catastrophic wild re that hit the Pantanal in 2020 on the vertebrate communities should be considered as substantially higher than our estimates of the direct mortality by the re.
Our ndings did not include several species killed by wild res in the Pantanal, such as those animals gured in the news media and animals found out of our standardized survey protocol by the eld personnel, collaborators, and re ghters. Among these species were large-bodied animals usually living at relatively low densities, such as the jaguar (Panthera onca), the puma (Puma concolor), the lowland tapir (Tapirus terrestris), as well as species such as the red-brocket deer (Mazama americana), the giant anteater (Myrmecophaga tridactyla), the marsh deer (Blastocerus dichotomus), the pampas deer (Ozotoceros bezoarticus), the collared peccary (Pecari tajacu). Other common carnivores not detected in our surveys were the crab-eating fox (Cerdocyon thous), the racoon (Procyon cancrivorus), the maned wolf (Chrysocyon brachyurus), the tayra (Eira barbara), ocelot (Leopardus pardalis), the coati (Nasua nasua), and the jaguarundi (Herpailurus yagouaroundi), among other species known to occur in the Pantanal 39,40 . Large-bodied vertebrates have been among the injured animals frequently found alive by rescuers after the res in the Pantanal, indicating that they may be less prone to die immediately. Large animals may die days and weeks after the re due to burnings, and this may explain the fact that species such as tapir, marsh deer, and jaguars were not detected in our surveys but were found dead elsewhere in the region. Medium to large animals were detected at larger distances from the transect line if compared with smaller animals, indicating that our decision on splitting the data set in these two groups was correct. Conceptually, the effective sample trip is de ned by the distance from the transect line in which the number of animals detected beyond such distance equals the number of animals missed within this distance, and it is used to estimate densities 41 .
Indirect estimates for the Australian savannas suggest that nearly 15,780 vertebrates per km 2 were affected by the 2019/2020 bush res, due to direct mortality and displacement, starvation, habitat loss and impoverishment, among other indirect effects of re 28 . Similar exercise conducted in the Pantanal for the 2020 wild re resulted in approximately 1,710 affected vertebrates per km 2 (at least 65 million native vertebrates) plus 4 billion invertebrates in 38,000 square kilometers 30 . In Bolivia, researchers adopted an expert-based approach to determine mortality rates due to re and theoretical estimates of mammal population densities, suggesting that 295.7 mammals per km 2 (5.9 million individuals) were killed by the 2019 wild res that affected nearly 20,000 km 2 of the Chiquitano Dry Forest 29 . Although these assessments may be considered valid attempts of estimating the impact of wild res, they are not based on direct carcass counts, they do not account for the extremely variable landscape composition, vegetation biomass, and ammability among regions hit by these events. Comparisons among these numbers and the results we present in this article may be virtually impossible as methods also varied among these studies. It is important to highlight that our estimates for the Pantanal are certainly underestimates for some taxonomic groups. Among the expected undetected dead animals in the surveyed areas in the Pantanal we may include especially amphibians, snakes, rodents, and armadillos, among other animal groups composed of fossorial animals and wood hollow users. Many small-bodied species may have died in places where they can not be accounted for, or their bodies may have been completely calcined or covered by ash. However, as our estimates are based on direct accounting of carcasses, they may be considered a better approximation of the reality when compared with theoretically based estimates of densities and/or mortality due to wild re.
Estimating the number of deaths among wildlife species due to re is relevant to contribute to the discussion on the potential cumulative impact of recurrent wild res on ecosystems, as extensive res compose a plausible scenario under climate change worldwide 1,2,3,4,42 . Indeed, the climate change scenarios for the Pantanal region indicate a 30% rainfall decrease in relation to the average precipitation between 2070 and 2100, as well as an increase in temperature and frequency of the extreme climatic events 20 . However, there was already a 40% shortage in rainfall in the region in 2020 19 , as well as an increase of 2°C in the average temperature since 1980 18 , creating the ideal conditions that contributed to the 2020's wild res in the Pantanal wetland. In fact, there is a trend occurring in the Pantanal, as a 376% increase in annual average burned area has been registered for the last two decades, with 43% of the area not being previously burned during this period 6 . Drought variability in the Pantanal region is closely related to teleconnection patterns associated with sea surface temperature anomalies in the Atlantic and Paci c oceans 43 . Despite the extreme droughts in the Pantanal region seem to be linked to a temperature anomaly in the tropical Atlantic ocean 19,43,44 , there is a negative synergy between the extreme climatic event with locally inadequate human factors in causing uncontrolled re 15,45 . The complicated perspective for the future of the Pantanal wetland also relies on the fact that moisture from the Amazonian rainforest plays a signi cant role in controlling summer rainfall in central-southern Brazil, including Pantanal 46 . Recent studies show that deforestation in the rainforest is connected to lower transfer of moisture for the Pantanal wetland 47 . Since deforestation and re are increasing again in the Amazon rainforest 15 , a challenging scenario for the Pantanal also includes other contributing factors such as hydrological changes due to river damming 48,49,50,51 , soil erosion and deforestation in surrounding plateaus 52,53 , which can cause wetland area losses, exposing more areas to the risk of re.
Pantanal is already the Brazilian ecoregion with the highest average re foci per square kilometer 54 . Thus, the perspective of an increased frequency and extension of re in the Pantanal and other tropical ecosystems poses a serious threat to the conservation of biodiversity and ecosystem services, as the cumulative effects may be considerable in a long-term perspective. In fact, the 2020 wild res were not limited to the Pantanal wetland, as exceptionally large burning areas were also registered further south, from the vicinity of Buenos Aires in Argentina, crossing into the Chaco region of Argentina, Paraguay and Bolivia and heavily hitting the Chiquitano dry forests of Bolivia, as well as large portions of the Amazon rainforest and the Cerrado savanna in Brazil. Under this perspective, strategies capable of preventing wild re disasters are key to avoid ecosystem degradation and economic losses, as well as increased emission of greenhouse gases, considering the climate change scenarios. One relevant instrument for such a strategy is the implementation of proper public policies. In Pantanal context, it is important to enact the Integrated Fire Management (IFM) National Bill n. 11276/2018, which is still under discussion in August 2021 in the Brazilian Chamber of Deputies, drafted with the participation of state agencies, scientists, and traditional and indigenous communities. Also, state-level legislation is relevant as it is more detailed and focused on regional socio-economic, ecological, and cultural context and nuances. In fact, Mato Grosso do Sul state approved a speci c policy to regulate the use of re 55 , and Mato Grossothe other Brazilian state where the Pantanal wetlands are located-is currently developing its own policy. Both pieces of legislation should serve as the basis for the proper use of re in the Pantanal, but awareness and training are highly necessary. The basis for these policies is fuel management, which has proven to be an effective way to prevent wild res and even reduce the risks posed by the combination of biomass, weather, temperature, and drought 56,57,58,59 .
As wild res pose a worldwide threat to ecosystem resilience and overall sustainability, it's worth reinforcing the need for (i) continuous monitoring for early detection of re risk and re events; (ii) the establishment of re ghter brigades in strategic locations with continuous operation; (iii) the improvement of logistic capabilities to allow effective access to distant and marshy areas in the oodplain, (iv) community education programs focused on proper re use for biomass management purposes, (v) effective enforcement of re policies; and (vi) implementation of wildlife rescue and rehabilitation centers. In the speci c case of the Pantanal wetland, for instance, it would be necessary to effectively implement such strategies for re management in connection with economy, biodiversity conservation, ecosystem management, and public policy, as it has been proposed by sustainability agendas 48 . The case of the Pantanal reminds us that integrated re management, as well as the implementation of sustainable land use and restoration to mitigate the inevitable impact of climate change are a crucial part of our survival strategy, given our dependency on ecosystems, their biodiversity, and services.