UNIVERSIDADE DO ESTADO DO AMAZONAS - UEA
PROGRAMA DE PÓS-GRADUAÇÃO EM CLIMA E AMBIENTE - CLIAMB
EVALUATION OF DOWNSCALED PROJECTIONS OF PRESENT AND FUTURE FIRE RISK IN THE AMAZON BASIN
JOSIVALDO LUCAS GALVÃO SILVA
Manaus, Amazonas July, 2022
JOSIVALDO LUCAS GALVÃO SILVA
EVALUATION OF DOWNSCALED PROJECTIONS OF PRESENT AND FUTURE FIRE RISK IN THE AMAZON BASIN
Advisor: Dr. Vinícius Buscioli Capistrano Co-advisor: Dr. José Augusto Paixão Veiga
Manaus, Amazonas July, 2022
Dissertação apresentada ao Instituto Nacional de Pesquisas da Amazônia e à Universidade do Estado do Amazonas como parte dos requisitos para obtenção do título de Mestre em Clima e Ambiente, área de concentração Interações Clima- Biosfera da Amazônia
Programa de Pós-Graduação do Instituto Nacional de Pesquisas da Amazônia – PPG-INPA Programa de Pós-Graduação em Clima e Ambiente – PPG-CLIAMB
Av. André Araújo, 2.936 - Petrópolis - CEP 69067-375 - Manaus -AM, Brasil Fone/Fax:
55 92 3643-3755 E-mail: coordenacaocliamb@gmail.com
ATA DE DEFESA PÚBLICA DE DISSERTAÇÃO DE MESTRADO DO PROGRAMA DE PÓS-GRADUAÇÃO EM CLIMA E AMBIENTE DO INSTITUTO NACIONAL DE PESQUISAS DA AMAZÔNIA E UNIVERSIDADE DO ESTADO DO AMAZONAS.
Aos 25 dias do mês de julho do ano de 2022, às 14:00 horas, de forma síncrona e remota, reuniu-se a Comissão Julgadora de Defesa Pública, composta pelos seguintes membros: Prof. Dr. Vinicius Buscioli Capistrano (Presidente), o Prof. Dr. Luiz Antônio Candido (INPA), o Prof. Dr. José Guilherme Martins dos Santos (INPE) e o Prof. Dr. Marcus Bottino (INPE), tendo como suplentes o Prof. Dr. Lincoln Alves (INPE), sob a presidência do primeiro, a fim de proceder a avaliação pública do trabalho de Dissertação de Mestrado do Josivaldo Lucas Galvão Silva intitulado: “Evaluation of Downscaled Projections of Presente and future fire risk in the Amazon Basin”, orientado pelo Prof. Dr. Vinicius Buscioli Capistrano (UFMS) e coorientado pelo Prof. Dr. José Augusto Paixão Veiga (UEA).
O Presidente da Comissão Julgadora deu início à seção e informou os procedimentos do exame. O aluno fez a exposição do seu trabalho que foi avaliado pelos membros da Comissão Julgadora de Defesa Pública. A Comissão Julgadora, então, se reuniu e os membros emitiram os seguintes pareceres:
Nome Parecer Assinatura
Dr. Vinicius Buscioli Capistrano ( ) Aprovado ( ) Reprovado _________________________
Dr. Luiz Antônio Candido ( ) Aprovado ( ) Reprovado _________________________
Dr. José Guilherme M. dos Santos ( ) Aprovado ( )Reprovado__________________________
Dr. Marcus Bottino ( ) Aprovado ( ) Reprovado__________________________
Nada mais havendo a relatar, foi lavrada a presente ata que, após lida e aprovada, foi assinada pelos membros da Comissão Julgadora.
x
Catalogação na Publicação (CIP-Brasil)
S586e Silva, Josivaldo Lucas Galvão
Evaluation of downscaled projections of present and future fire risk in the Amazon Basin / Josivaldo Lucas Galvão Silva; orientador Vinicius Buscioli Capistrano; coorientador José Augusto Paixão Veiga. - Manaus:[s. l.], 2022.
7.6 KB
105 p. : il. color.
Dissertação (Mestrado - Programa de Pós-Graduação em Clima e Ambiente - Cliamb.) - Coordenação do Programa de Pós-Graduação, INPA, 2022.
1. Modelagem climática. 2. Eta Model. 3. Mudanças climáticas. I.
Capistrano, Vinicius Buscioli. II. Veiga, José Augusto Paixão. III. Título CDD 551.6
Sinopse:
Avaliou-se os impactos provocados pelo aumento de gases de efeito estufa e desmatamento no risco de fogo na bacia Amazônica, por meio de modelagem climática regional.
Palavras-Chave: Modelagem climática, Eta Model, mudanças climáticas, desmatamento, KBDI
To my mother, Claudia Galvão da Fonseca, who never spared any effort for me to have an education.
ACKNOWLEDGMENTS
I would like to express my thanks to everyone who contributed, directly and indirectly, to the completion of this dissertation.
To INPA, UEA, CLIAMB, all professors, employees and colleagues from the post-graduation program.
To the National Council for Scientific and Technological Development– CNPq, for the grant who made this research possible.
To my advisor, professor Dr. Vinicius Buscioli Capistrano, for his patience, help, dedication, and for all that he taught me, whose example I will follow for all my professional life.
To my co-advisor, professor Dr. José Augusto Paixão Veiga, for his help, support, empathy and for all the encouragement words in hard moments during this research.
To my friends, Thalita Raquel, Tallyson Tavares, Leonardo Bicalho, Ranyelli Figueiredo, Everton Gonçalves, Lincoln Ferreira and Nikoly Negreiros for the immense support, and for reminding me, daily, that friendship is stronger than any distance.
To the MCA, whose meetings were essential to my professional growth and mental health throughout the pandemic and home office. Specially to Adriane Brito, whose work was essential for the conclusion of this dissertation.
And finally, to my family, who love and support me unconditionally.
RESUMO
As mudanças climáticas e o desmatamento são dois grandes obstáculos que a floresta amazônica está enfrentando no século XXI. Consequentemente, estes dois fatores contribuem para um cenário ideal onde a ocorrência de fogo se torne mais frequente e intensa. Portanto, este trabalho teve como objetivo avaliar os efeitos do aumento dos gases de efeito estufa e do desmatamento na precipitação, temperatura e risco de fogo na bacia amazônica. Para isto, foram utilizadas simulações numéricas do modelo Eta forçado por três modelos de circulação global (BESM, HADGEM-ES2 e MIROC5). Estas simulações resultaram em três experimentos diferentes, sendo eles o experimento de controle, que caracteriza o clima presente (1979-2005) e dois experimentos de sensibilidade que caracterizam o clima futuro (2070-2100). As simulações do clima futuro tiveram como base dois cenários, o primeiro apenas com o aumento da concentração de CO2 atmosférico (RCP8.5_DEF15) e o segundo com a combinação de alta emissão de CO2 e desmatamento (RCP8.5_DEFTOT). O risco de fogo foi calculado com as simulações de precipitação e temperatura máxima através do Keetch-Byram Drought Index (KBDI). Para melhor entender os resultados das simulações, foi realizada a validação do experimento de controle para o período seco (Julho, Agosto e Setembro). De maneira geral, a temperatura máxima e a temperatura média foram subestimadas nas simulações em toda a bacia e a precipitação, por sua vez, foi superestimada.
Para o KBDI, o uso do método de ensemble multi-modelos apresentou vantagem sobre os membros individuais, e apesar dos vieses encontrados nas variáveis anteriores, o KBDI tendeu a não apresentar valores extremos. Para o clima futuro, utilizando cenários de emissão de gases de efeito estufa e desmatamento, a precipitação foi reduzida e a temperatura máxima aumentou tanto no período seco quanto no período chuvoso. No entanto, os maiores impactos foram observados durante o período seco, e principalmente no cenário RCP8.5-DEFTOT.
Para o KBDI, no experimento RCP8.5-DEF15, foi encontrado que seus valores aumentam nos dois períodos, se concentrando ao norte no período chuvoso e generalizado na área da bacia no período seco. Por outro lado, para o RCP8.5-DEFTOT, a anomalia do KBDI é maior, evidenciando que em um cenário de mudanças climáticas e desmatamento, a floresta perde a capacidade de suprimir condições favoráveis ao fogo. Além disto, em anos extremos dos experimentos RCP8.5-DEF15 e RCP8.5-DEFTOT, 40% e 58% da área da bacia apresenta risco de fogo extremo no período seco, respectivamente. Portanto, os resultados desta pesquisa reafirmam a importância da redução de gases efeito estufa e do desmatamento, para que a ocorrência de fogo não seja uma ameaça a biodiversidade e aos serviços ecossistêmicos da maior floresta tropical do mundo.
ABSTRACT
Climate change and deforestation are two of the major obstacles that the Amazon Forest is facing in the 21st century. Consequently, these two factors contribute to an ideal scenario where fire occurrence becomes more frequent and more intense. Therefore, this work aimed to evaluate the effects of increasing greenhouse gases and deforestation on precipitation, temperature, and fire risk in the Amazon basin. To do this, Eta model simulations forced by three global circulation models (BESM, HADGEM-ES2 and MIROC5) were used. These simulations resulted in three different experiments, with the control experiment characterizing the current climate (1979-2005) and two experiments characterizing the future climate (2070- 2100). The simulations of the future climate were based on two scenarios, the first with only increasing atmospheric CO2 concentrations (RCP8.5_DEF15) and the second with the combination of high CO2 emissions and deforestation (RCP8.5_DEFTOT). Fire risk was calculated via the simulations of precipitation and maximum temperature using the Keetch- Byram drought index (KBDI). To better understand the behavior of the simulations, validation of the control experiment was performed for the dry period (July, August and September). In general, the maximum temperature and the temperature at 2 meters were underestimated in the simulations throughout the basin, and precipitation was overestimated. For the KBDI, the use of the multi-model ensemble method showed an advantage over individual members and, despite the biases found in the previous variables, the KBDI did not show a tendency to reach extreme values. For the future climate, using greenhouse gas emissions and deforestation scenarios, precipitation was reduced and the maximum temperature increased in both the dry and rainy periods. However, the largest impacts were observed during the dry period, and especially in the RCP8.5-DEFTOT scenario. For the KBDI, in the RCP8.5-DEF15 experiment, it was found that its values increase in both periods, and are concentrated to the north in the rainy period and are generalized in the basin area in the dry period. For RCP8.5- DEFTOT, the KBDI anomaly is greater, and shows that in a scenario of climate change and deforestation, the forest loses its ability to suppress conditions that are favorable to fire.
Furthermore, in the extreme years of the RCP8.5-DEF15 and RCP8.5-DEFTOT experiments, respectively, 40% and 58% of the basin area present a risk of extreme fire in the dry period.
Therefore, the results of this research reaffirm the importance of reducing greenhouse gases and deforestation so that fire occurrence is not a threat to the biodiversity and ecosystem services of the world’s largest tropical forest.
SUMMARY
LIST OF FIGURES ... ix
LIST OF TABLES ... xii
LIST OF ACRONYMS AND ABBREVIATIONS ... xiii
INTRODUCTION... 14
OBJECTIVES ... 16
LITERATURE REVIEW ... 17
Effect of climate change on fires ...17
Effect of deforestation on fires ...20
The joint effect of global climate change and deforestation on forest fires ...23
CHAPTER 1 ... 27
Abstract ...28
Resumo ...28
INTRODUCTION... 29
MATHERIALS & METHODS ... 30
Study region ...30
Model description and simulation strategy ...31
Regional circulation model ...31
Earth system models ...31
Fire risk index ...33
Observed data and model validation ...33
RESULTS ... 34
Mean temperature...34
Maximum temperature ...35
Keetch-Byram drought index ...37
DISCUSSION ... 39
CONCLUSION ... 41
ACKNOWLEDGMENTS... 41
REFERENCES ... 42
CHAPTER 2 ... 50
Abstract ...51
INTRODUCTION... 52
MATERIALS AND METHODS ... 53
Experimental design ...53
Keetch-Byram Drought Index (KBDI) ...54
Projections and KBDI analyses ...55
RESULTS ... 56
Increased CO2 scenario (RCP8.5-DEF15) ...56
Increased CO2 combined with a full deforestation scenario (RCP8.5-DEFTOT) ...58
DISCUSSION ... 62
CONCLUSION ... 66
REFERENCES ... 66
SYNTHESIS ... 78
REFERENCES ... 79
A – Supplementary Material of the manuscript: Climate downscaling in the amazon basin: evaluating fire risk ... 98
B – Noah Vegetation Parameters ... 105
LIST OF FIGURES
Figure 1- Air temperature (°C) and change in precipitation (%) in the annual cycle for the eastern and southern Amazon under the RCP8.5 scenario. Reference period 1961-2005.
Adapted from Marengo et al. (2018). ... 18 Figure 2- Diagram demonstrating the synergy between deforestation and climate change in the Amazon. Adapted from Staal et al. 2020. ... 20 Figure 3- Flowchart demonstrating the interactions between the environment and the political sphere and the positive and negative feedback on Amazon wildfires and policies, as well as the impacts on carbon emissions and fire probability. Adapted from Aragão et al. (2018). .... 24 Figure 4- Simulated relative fire probability (RFP) for the month of July (2003 to 2015).
Black crosses mark grid cells in which 13 or more hotpixels were detected. Adapted from Fonseca et al. (2019). ... 26
CHAPTER 1 - CLIMATE DOWNSCALING IN THE AMAZON BASIN:
EVALUATING FIRE RISK.
Figure 1. (A) Mean temperature bias (°C) from ERA5 versus the ensemble for the dry period (July, August and September) in the Amazon basin from 1979 to 2005. Dotted regions are statistically significant at a 95% confidence level. (B) Taylor diagram of the mean temperature (°C) for the dry period (July, August and September) in the Amazon basin from 1979 to 2005.
... 35 Figure 2. (A) Maximum temperature bias (°C) from CPC versus the ensemble, for the dry period (July, August and September) in the Amazon basin from 1979 to 2005. Dotted regions are statistically significant at a 95% confidence level. (B) Taylor diagram of the maximum temperature (°C) for the dry period (July, August and September) in the Amazon basin from 1979 to 2005. ... 36 Figure 3. (A) Precipitation bias (mm d-1) from CPC versus ensemble, for the dry period (July, August and September) in the Amazon basin from 1979 to 2005. Dotted regions are statistically significant at a 95% confidence level. (B) Taylor diagram of the precipitation (mm d-1) for the dry period (July, August and September) in the Amazon basin from 1979 to 2005. 3
Figure 4. Keetch-Byram drought index calculated from the CPC dataset (A) and Keetch- Byram drought index bias calculated from the CPC versus the ensemble data, for the dry period (July, August and September) in the Amazon basin from 1979 to 2005 (B). Dotted regions are statistically significant at a 95% confidence level. ... 37 Figure 5. (A) Consecutive dry days bias from the CPC versus ensemble, for the dry period in the Amazon basin (July, August and September) from 1979 to 2005. (B) Taylor diagram of the Keetch-Byram drought index for the dry period in the Amazon basin (July, August and September) from 1979 to 2005. ... 38 CHAPTER 2 - IMPACTS OF CLIMATE CHANGE AND DEFORESTATION ON FIRE RISK IN THE AMAZON BASIN.
Figure 1 – Vegetation map used by the Eta simulations: A) Vegetation coverage and deforestation for the reference year of 2015 and B) Vegetation coverage for a hypothetical scenario of total deforestation of the Amazon Forest. ... 54 Figure 2 – Maximum temperature anomalies (oC) relative to the increased CO2 concentration for the periods of DJF (A) and JAS (B). Dashed regions are statistically significant at a 95%
confidence level. ... 56 Figure 3 – Precipitation anomalies (mm d-1) relative to the increased CO2 concentration for DJF (a) and JAS (b). Dashed regions are statistically significant at a 95% confidence level. . 57 Figure 4 – KBDI anomalies relative to the increased CO2 concentration for DJF (A) and JAS (B). Dashed regions are statistically significant at a 95% confidence level. ... 58 Figure 5 – Maximum temperature anomalies (oC) relative to the increased CO2 concentration and total deforestation scenario for DJF (A) and JAS (B). Dashed regions are statistically significant at a 95% confidence level. ... 59 Figure 6 – Mean precipitation anomalies (mm d-1) relative to the increased CO2 concentration and total deforestation scenario for DJF (A) and JAS (B). Dashed regions are statistically significant at a 95% confidence level ... 59
Figure 7 – KBDI anomalies relative to the increased CO2 concentration and total deforestation scenario for DJF (A) and JAS (B). Dashed regions are statistically significant at a 95%
confidence level. ... 60 Figure 8 – Areas for which KBDI is classified between 400 to 800 (high and extreme risk of fire) in the JAS period. A) Control experiment is in yellow, RCP8.5-DEF15 is in red for the 30-year period. B) Control experiment is in yellow, RCP8.5-DEFTOT is in red for the 30-year period. C) Control experiment is in yellow, RCP8.5-DEF15 is in red for the 5 highest KBDI years and D) Control experiment is in yellow, RCP8.5-DEFTOT is in red for the 5 highest KBDI years. ... 62
LIST OF TABLES
Table 1 - Physical parameterizations used in Eta-Model. ... 533 Table 2 – Mean anomalies of sensible heat, latent heat, Bowen ratio and evapotranspiration over the area of the Amazon basin for DJF and JAS periods. ... 611
LIST OF ACRONYMS AND ABBREVIATIONS
BAM - Brazilian Atmospheric Model BESM - Brazilian Earth System Model CFWI - Canadian Fire Weather Index
CMIP - Coupled Model Intercomparison Project CPC - Climate Prediction Center
ECMWF - European Centre for Medium-Range Weather Forecasts ESM - Earth System Model
FFDI - McArthur Forest Fire Danger Index GCM - Global Climate Model
GHG - Greenhouse Gases
HadGEM-ES - Hadley Centre Global Environmental Model IPCC - Intergovernmental Panel on Climate Change KBDI - Keeth-Byram Drought Index
MIROC - Model for Interdisciplinary Research on Climate
PRODES - Program for Monitoring Deforestation of the Brazilian Amazon Forest by Satellite
RCM - Regional Climate Model
RCP - Representative Concentration Pathway SSiB - Simplified Simple Biosphere Model SSP - Shared Socioeconomic Pathways SST - Sea Surface Temperature
TRIFFiD - Top-down Representation of Interactive Foliage Including Dynamics UKCA - United Kingdom Chemistry and Aerosol model
INTRODUCTION
The fires in the Amazon have a great impact on the environment, the economy and human health. Fires in this region cause tree mortality, increase the vulnerability of the forest to new fires, and emit large amounts of greenhouse gases (GHGs) into the atmosphere (Barbosa and Fearnside, 1999; Nepstad, 1999; Alencar et al., 2006). Naturally, the Amazon Forest is highly resilient to fire, so only under specific climatic conditions (warm and dry) and land use changes do fires occur in the Amazon (Nepstad, 1999; Leisenfeld et al., 2016). The advance of human activities, which are sources of ignition, and global climate changes related to the increased concentration of GHGs in the atmosphere may promote favorable conditions for increases in the occurrence and severity of fires in the Amazon (Marengo et al., 2018).
Therefore, it is imperative to study both the individual and combined impacts of deforestation and climate change for the future of the Amazon.
Since the industrial revolution, anthropic activities have altered the chemical composition of the atmosphere (IPCC, 2013). These changes are aggravating the greenhouse effect, which is naturally essential for the maintenance of life on Earth, though its intensification causes an increase in the global average temperature (Reboita and Ambrizzi, 2022). Taking into account a pessimistic emission scenario for the end of the 21st century, climate models project an increase in the average global temperature of up to 5.7 ºC (IPCC, 2021). The impacts of climate change are diverse, both on human activities and on meteorological events. Trends indicate that warmer temperatures and increased aridity due to climate change may be driving more frequent and more intense forest fires (Williams et al., 2019; Gannon et al., 2021). In all emission scenarios, for example, a trend of increasing extreme drought events is observed (Sheffilld and Wood, 2008), thus, implying that in the future the Amazon Forest may present climatic conditions compatible with a more frequent occurrence of fires. Projections show that, under dire scenarios of climate change, the Amazon Forest could lose its natural barriers against fire, as nights become warmer and consecutive dry days become more frequent (Balch et al., 2022; Reboita et al., 2022). The most recent IPCC report also affirms that the Amazon will have one of the highest increases in fire weather risk in the world (Ranasinghe et al., 2021).
Fires play an important role in the discussion on climate change because they have a potential positive feedback effect. During the 1998 drought, which affected the northern, central, and eastern Amazon and which was caused by El Niño (Borma and Nobre, 2013),
about 40,000 km² of the Brazilian Amazon Forest were lost to fire, which resulted in a carbon emissions of around 0.049 to 0.329 PgC (Alencar et al., 2006). The Amazon Forest has a substantial amount of carbon stored in its biomass, between 70 to 120 PgC (Ometto et al., 2014), which under climate change conditions may be more susceptible to fires and to being emitted into the atmosphere in the form of GHGs.
Aside from climate change, land use change is one of the most important sources of fires in the Amazon basin (Leisenfeld et al., 2016). The land use change in the Amazon occurs mainly through slash-and-burn agriculture, the opening of areas for pasture, and mechanized agriculture (Cochrane, 2009). Since the Amazon Forest presents low risk of fire due to its climatic characteristics, slash-and-burn agriculture is one of the main ignition sources that cause wildfires (Cochrane and Laurance, 2008). The strategy used is to cut large areas of forest, wait for the dry season, which dries the vegetation and, finally, use controlled fire to clear the area, thus allowing the area to be used for pasture or agricultural activities. However, forest fragmentation and intensification of the edge effect promote ideal conditions for the spread of fires in mature forests (Numata et al., 2017).
This dynamic between activities that involve ignition sources and fire is difficult to predict, since it involves a substantial socioeconomic variable for its occurrence. Indeed, land use and ignition sources are determinant factors for fire occurrence (Bowman et al., 2009);
however, changes in climate have increasingly been identified as a major cause for fire occurrence (Le Page et al., 2017; Williams et al., 2019). As the prediction of fire occurrence is limited by the human variable (Gannon et al., 2021), one strategy for prediction is the use of fire danger indices such as the McArthur Forest Fire Danger Index (FFDI) (McArthur, 1963), the Canadian Fire Weather Index (CFWI) (Wagner, 1987), and also, but not limited to, the Keeth Byram Drought Index (KBDI). These indices describe the conditions necessary for forest fires to occur, rather than predict the occurrence of the fire itself (Liu et al., 2010). One of the advantages of using these indices is the compatibility of the meteorological variables that are needed for the calculations and the variables simulated by atmospheric models.
Given the above, studies that seek to understand the dynamics between fire, climate, and deforestation in the Amazon are extremely important. They not only increase the knowledge on the matter, since there is a lack of studies that evaluate the potential fire risk using mesoscale models (Liu et al., 2010), but they also serve as a basis for policymakers so that they can evaluate the scenarios and prepare pathways for mitigating the impacts of future fires in the Amazon Forest.
Therefore, recognizing the need to contribute to scientific knowledge in regards to how climate change and land use change impact fires in the Amazon, the present study raises the following questions: How does the increase in atmospheric CO2 concentration modify the fire risk potential in the Amazon? And, how does the joint effect of increasing CO2
concentrations and deforestation affect fire risk potential?
OBJECTIVES
GENERAL OBJECTIVE:
The general objective of this study was to evaluate the impacts caused by the increase in the concentration of greenhouse gases and deforestation on the fire risk in the Amazon basin.
SPECIFIC OBJECTIVES:
a) To validate the historical Eta-model projections of precipitation, temperature and fire risk;
b) To assess the impact of the high greenhouse gas emission scenario on the seasonality and spatial distribution of fire risk in the Amazon Forest;
c) To assess the joint effects of the high greenhouse gas emission scenario and large-scale deforestation on the seasonality and spatial distribution of fire risk in the Amazon Forest.
LITERATURE REVIEW
Effect of climate change on fires
For a fire to occur, fuel, dry weather, and an ignition source are always needed (Nepstad et al., 1999). There are highly adapted ecosystems that are dependent on fires such as savannas and grasslands, where it acts as an essential service (Keeley and Pausas, 2019);
however, natural fires are extremely rare in tropical forests. Sanford et al. (1985), through historical reconstruction of fires in the Amazon, suggest that fires naturally occur at a secular frequency. This is due to the natural fire resistance of the forests, mainly the biomass concentrated in the tree trunks and the humid microclimate inside the forests (Nepstad et al., 1999). Therefore, in a fire-resistant ecosystem, the increasingly frequent presence of fires is due mostly to anthropogenic factors (Cochrane, 2003; Aragão et al., 2008; Alencar et al., 2015; Tasker and Arima, 2016; Aragão et al., 2018).
Although human activities can provide sources of ignition, fires are more intense and spread more easily under favorable weather conditions (Tasker and Arima, 2016). In fact, Marengo et al. (2008) report that during the 2005 drought, reduced precipitation, humidity, and increased temperature were significant factors for the occurrence of wildfires in the southwestern part of the Amazon rainforest. In addition, there is evidence that changes in temperatures of the Pacific and Atlantic Oceans may be related to increased fire risk (Alencar et al., 2006; Morton et al., 2013; Alencar et al., 2015; Da Silva Junior et al., 2019). Therefore, evidence of a possible increase in the length of the dry season in the southern Amazon region (Marengo et al., 2018) may directly alter the fire dynamic in the region.
For the Amazon, climate models mostly project an increase in temperature and decrease in precipitation (Shukla et al., 1990; Nobre et al., 2009; Marengo et al., 2011;
Marengo et al., 2018; Gomes et al., 2021). On a regional scale, Marengo et al. (2018) show that, based on Coupled Model Intercomparison Project 5 (CMIP5) models for the RCP8.5 (Representative Concentration Pathway) scenario, temperatures may increase up to 7°C in the southern part of the forest (Figure 1). Additionally, for precipitation, the authors report that the models project a decrease of up to 55% in monthly averages during the period of transition between the dry and rainy seasons, which are conditions that would potentially increase fire activity. Despite the uncertainties, climate change may have a direct impact on
wildfires in the Amazon by favoring the ideal conditions necessary for fire to occur (Bush et al., 2008; Cochrane and Laurance, 2008; Marengo et al., 2018). As such, several studies have been conducted to assess the impact of climate change on forest fires (Golding and Betts, 2008; Liu et al., 2010; Sun et al., 2019; Gannon et al., 2021).
Figure 1- Air temperature (°C) and change in precipitation (%) in the annual cycle for the eastern and southern Amazon under the RCP8.5 scenario. Reference period 1961-2005. Adapted from Marengo et al. (2018).
The potential interaction between deforestation and climate change on fire risk in the Amazon was studied by Golding and Betts (2008). The FFDI was used together with the HadCM3 global circulation model, which was forced by the A1B scenario and the deforestation scenario presented in the Special Report on Emission Scenarios (SRES). The largest increases in wildfire risk occurred in the central portion of the basin; however, the entire Amazon basin showed a “moderate” risk of wildfire risk. The authors found that a
“high” risk of fire in 2080 was present in approximately 50% of the area. Furthermore, the areas projected to have the greatest potential to catching fire were those in which the
simulations showed the greatest impacts from deforestation, thus evidencing the connection between fire potential and deforestation.
Liu et al. (2010) used the KBDI to evaluate trends in fire potential under greenhouse gas emission scenarios. Four global circulation models (HadCM3, CGCM2, CSIRO and NIES) were used to simulate atmospheric conditions for the period 2070 to 2100 under the SRES. The authors found that the models, together with the KBDI, are able to represent the global distribution of fire risk. For South America, the high fire potential period is projected to increase to 10 months (May to December). The study also highlights the sensitivity of the KBDI for the different models; for example, the A1 scenario projected by HadCM3 simulated the largest increase in the KBDI among the models used. Liu et al. (2010) finally concluded that the KBDI is both model and scenario sensitive.
In order to assess the impacts of temperature increase on fires, Sun et al. (2019) used five global atmospheric models (GFDL-ESM2M, HadGEM2-ES, IPSLCM5A-LR, MIROC- ESM-CHEM and NorESM1-M). For the wildfire indicator (FFDI), the authors found that under the scenarios studied, the duration of wildfires could increase for the Amazon, northern China, central Asia, Australia, and regions of the western United States. The increase in duration, according to Sun et al. (2019) is due to the projected increase of 1.5 to 2.0 °C in air temperature in the mentioned regions. However, the authors found no evidence that the Amazon suffers from more frequent fires. Despite this, they point to Brazil, along with India, the United States and Russia as the countries that showed the greatest increase in the risk of wildfires.
Gannon et al. (2021) used precipitation and temperatures from the CMIP5 to calculate the KBDI. The objective of the work was to analyze the frequency and severity of fire risk using the RCP8.5 scenario to calculate the KBDI on a global scale. The authors adapted the KBDI to more faithfully demonstrate the conditions of the forests, and applied a flammability factor, which was inherent to each type of vegetation classified in the study. As a result, they found that in all regions of the globe, the projections indicate an increase in the severity and frequency of fires. More specifically, the Amazon region presents the greatest change in the maximum KBDI compared to the base period. In addition, the authors found that the number of days with high fire potential exceeds 60 days, a significant increase compared to the historical period.
Effect of deforestation on fires
The deforestation of the Amazon is directly linked to anthropogenic activities. The increase in population density and improvement of road infrastructure, in addition to the advance of agricultural activities, have proven to be determining factors for deforestation in the Amazon Forest (Laurance et al., 2002; Margulis, 2003; Fearnside, 2006). As recently as the 1980s, much of the deforested area in the Amazon was used for pasture (Fearnside, 1983).
However, since the 2000s, mechanized agriculture has taken a crucial role in Amazonian deforestation, especially in the “Arc of Deforestation” (Morton et al., 2006).
Between the years 2005 and 2010, the deforestation rate for the entire Amazon was 47,296 km², with Brazil being the country that contributed the largest parcel of deforested area to this rate (RAISG, 2015). These high rates of deforestation can cause irreversible changes to the forest. Nobre and Borma (2009) warn that deforestation of more than 40% of the Amazon Forest could lead the region to an inflection point or tipping point, which would cause the system to change its state of equilibrium.
The impact of deforestation on Amazonian climate has been extensively studied (Salati and Nobre, 1991; Dickinson and Kennedy, 1992; Eltahir and Bras, 1994; Roy and Avissar, 2002). Its effects on climate can be assessed at both regional and global scales, with a high correlation between deforestation and observed changes (Werth and Avissar, 2002).
Deforestation affects climate by altering energy fluxes (Zhang et al., 1996), and, regionally, one of the most important impacts of deforestation on forest climate is decreased evapotranspiration, altered moisture convergence over the forest and, consequently, decreased precipitation (Figure 2) (Dickinson and Kennedy, 1992; Zhang et al., 1996). Thus, deforestation increases the potential for fires, both by altering the regional climate and by weakening the forest (Aragão et al., 2008).
Figure 2- Diagram demonstrating the synergy between deforestation and climate change in the Amazon. Adapted from Staal et al. 2020.
Changes in the fire regime in the Amazon have been observed in areas that have already been deforested and where there are activities that provide sources of ignition (Alencar et al., 2015). These activities are basically slash and burn of primary and secondary forest for agriculture or grazing, or pasture burning activities to suppress some invasive species (Nepstad, 1999). Combining deforestation with these ignition sources, accidents in which fires related to anthropic activities get out of control and hit the forest are becoming increasingly common (Holdsworth and Uhl, 1997; Nepstad et al., 1999). According to the characteristics of Amazonian forest fires, they can be divided into three types: slash-and-burn fires, lowland fires, and fires in previously cleared areas (Nepstad et al., 1999).
Although intense fires occur in different parts of the Amazon basin, most of them are concentrated in the Arc of Deforestation, in the southern and eastern portions (De Faria et al., 2017). Alencar et al. (2006) showed that in the year 1995, without a drought event, transition forests located in the south of the forest showed greater burned area than dense forests, which evidences the role of deforestation in the impact fires can have. Da Silva Junior et al. (2019) found that for the west of the Amazon Forest, fires are concentrated in areas where agricultural activity, high forest fragmentation and the presence of roads prevail.
Forests that have already undergone modifications related to anthropic activities are more flammable, which facilitates the spread of wildfires in different forest types (Alencar et al., 2006; Aragão et al., 2018). A less dense canopy allows solar radiation to pass through, causing the forest interior to become warmer and drying out the understory (Brando et al., 2020). In addition, when compared to non-disturbed forests, those that have been burned are more flammable (Cochrane and Laurance, 2008). Thus, fire in the Amazon Forest presents a cyclical feedback loop in which fire occurs, and large trees die and fall, clearings are opened and the amount of flammable biomass in the forest increases (Cochrane and Schulze, 1999;
Nepstad et al., 1999).
Fire not only makes the forest more vulnerable, but also has the potential to alter the local climate. According to Tosca et al. (2013), aerosols released from burning forests can affect precipitation in the region by modifying the dynamics of cloud formation. Therefore, preventing deforestation in the Amazon Forest is essential in order to prevent fires from permanently affecting the biome.
The high frequency of fires in the Amazon may be outstripping the forest’s ability to recover (De Faria et al., 2020). However, a decrease in deforestation in the Amazon does not necessarily imply a decrease in outbreaks of fire (Aragão et al., 2018). Aragão et al. (2018) report that even with deforestation rates decreasing between the period 2003 to 2015, fire
incidence increased significantly during the 2015 drought. Therefore, climate, deforestation, and socioeconomic factors need to be taken into consideration in order to obtain a good assessment of fire dynamics (Morton et al., 2013). Given such an outlook, the scientific community has produced several studies that measure the effects of deforestation on the occurrence of fires in the Amazon (Carvalho et al., 2010; Lima et al., 2012; Van Marle et al., 2016; Silva et al., 2018; Silva et al., 2021).
Carvalho et al. (2010) conducted an experiment to estimate fire spread and tree mortality. The experiment took place in the area near Alta Floresta, in the Brazilian midwest, in the Arc of Deforestation, in 2001, and tree mortality was estimated two years later. The authors found that the experimental area closest to a previously burned site had the highest tree mortality rate (56%). According to the authors, this shows that forest areas near deforested and burned areas are more vulnerable to fire.
Lima et al. (2012) studied the spatial relationship between fire use and deforestation in the Amazon in the states of Rondônia and Mato Grosso. Satellite data were used to create forest scar maps, and the deforestation map was obtained through the Program for Monitoring the Deforestation of the Brazilian Amazon Forest by Satellite (PRODES). The results of the research indicate that the burned area depends to some extent on the deforested area, since 90% of the fires identified in the period studied occurred in regions that were deforested.
However, most of the burn scars found in the year 2005 were in areas that were deforested before 2002. The authors argue that this is counter intuitive, since it is assumed that the increase in burns would be related to more recent deforestation. The use of fire for management of pasture areas or for “clearing” vegetation in the process of regeneration for a new crop or pasture could be one explanation for this phenomenon.
Van Marle et al. (2016) studied fire and deforestation dynamics in the Amazon for the period between 1973 and 2014. Both observed data and satellite data were used. To verify the relationship between fire and deforestation, they compared emissions based on satellite- measured visibility with two deforestation databases. The authors found that, despite the limitations of the applied methodology, visibility-based estimates are good at predicting fire.
The best correlation was found in the western part of the basin, a region where the Andes acts like a barrier to wind and accumulates smoke from fires. For the period between 1990 and 2014, the satellite data indicates that 31% of the fire identified in the study was related to deforestation. The authors concluded that deforestation and forest fires have complex dynamics, although they are related, and the choice of deforestation database can influence the outcome.
Silva et al. (2018) investigated fire dynamics for the southwestern portion of the Brazilian Amazon. The study aimed to understand the extent of burned areas over 33 years (1984 to 2016) and its relationship with extreme drought events, deforestation, and spatial distribution. Satellite data, deforestation data from PRODES, and data from the Brazilian Institute of Geography and Statistics were used, and the authors found that during the 33 years studied, 525,426 ha were classified as burn scars, with the years 2005 and 2010 being those that accounted for 90% of the total burned area. About 88% of the fires that occurred in the studied area was in an area with large deforestation and large forest fragmentation. Another important result found is that, of the total area where fire and logging occurred, 37% of the area was logged after burning. The authors highlight the importance of studying fires not only as a degradation factor, but also as a possible facilitator of deforestation.
Silva et al. (2021) studied the relationship between deforestation, GHG emissions, and fire occurrence. Because it covers the period from 2006 to 2019, the study includes the new forest code that came into effect in 2012. The region studied was Novo Progresso, in southwest of Pará state, and which is located in the Arc of Deforestation. PRODES and satellite data were used for analysis, and GHG emissions were estimated by in situ experiments. The significant relationship between the occurrence of fire and areas that suffered degradation by slash and burn activity was one of the main results of the research.
The rate of occurrence of fire was greater than the deforestation rate. The authors argue that this result demonstrates that even with reduced deforestation of primary forest, fire occurs in degraded forests precisely because they are more flammable.
The joint effect of global climate change and deforestation on forest fires
Global climate model experiments involving future climate are improved when land use is taken into account (Feddema et al., 2005). For the future, land use is directly related to socioeconomic factors (Marengo et al., 2018). In a large-scale deforestation scenario, for example, precipitation reduction of up to 20% is projected for the Amazon (Moore et al., 2007).
In the form of a flowchart, Figure 3 presents the complex relationship between climate, land use, forests, and public policies in fire dynamics (Aragão et al., 2018). In the climate sphere, climate change and droughts cause positive feedback on Amazonian wildfires.
Climate affects both human-modified forests and intact forests, the latter with high potential carbon emissions to the atmosphere. Land use also affects both forests and climate through deforestation and grazing management. In the political sphere, actions such as a robust forest
code, enforcement, sustainable development, alternatives to fire usage as a management tool, and global initiatives, such as the IPCC, can act to reduce feedback in all three spheres of the environment. Since climate variation contributes to fire severity (Liu et al., 2010), several studies have aimed to understand the relationship between global climate change, deforestation and wildfires (Le Page et al., 2010; Silvestrini et al., 2011; De Faria et al., 2017;
Le Page et al., 2017; Fonseca et al., 2019).
Figure 3- Flowchart demonstrating the interactions between the environment and the political sphere and the positive and negative feedback on Amazon wildfires and policies, as well as the impacts on carbon emissions and fire probability. Adapted from Aragão et al. (2018).
A comparison between present and future fire vulnerability was carried out by Le Page et al. (2010) using both satellite data and climate simulations from CMIP3 to map active fire hotspots. The authors found that, under present climate conditions, 58% of the Amazon is fire resistant. Under a scenario of climate change and land use change, 63% of the forest showed vulnerability to fire-related deforestation. The authors also found a moderate reduction of precipitation (200 mm) during the dry season, which explains the increased risk of fires in the region.
Silvestrini et al. (2011) used a statistical model to simulate fire risk as a result of anthropogenic actions. The IPCC A2 scenario, which has the highest CO2 emission rate, was
used to force the HadCM3 model and the Business-as-Usual deforestation scenario.
Considering only climate change, it is observed that there is a risk of fire in the northwestern portion in the Amazon, which under current conditions is fire resistant. Separately, deforestation and road expansion are responsible for 19% increase in hotpixels and climate change 12%. However, when studied together, they represent a 49% increase, which is a result that indicates a synergy between the two factors for fire occurrence in the region.
De Faria et al. (2017) adapted the Carbon and Land Use Change model to include effects of droughts on fire fuel, behavior and intensity. Using meteorological variables from CMIP5, the authors used the RCP2.6 and RCP8.5 emissions scenarios to evaluate the possible effects of climate change on fire dynamics. The simulations suggest that, in a more pessimistic scenario (RCP8.5), fire risk could double carbon emissions per area burned. In addition, the authors found that, for the RCP2.6 scenario, the area where presents a
“dangerous” fire intensity classification is reduced by 68% when compared to the area in the RCP8.5 scenario. This evidences that effective reduction in GHG emission rates can mitigate these effects on fire severity.
Similarly, Le Page et al. (2017) also studied the effect of deforestation and RCPs 4.5 and 8.5 on Amazon wildfires. The model used to simulate the risk of wildfires was the HESFIRE model, which was specifically calibrated for the Amazon. The land use change scenarios used were those described in the RCP and the climate change scenarios were obtained through RCMIP5 simulations. The authors reported that climate change alone is capable of significantly affecting fire risk in the Amazon. In the RCP8.5 scenario, there was a 48% increase in burned area and a 75% increase in burnt understory when compared to the current climate. In the experiment that maintained the fixed GHG concentration of the present climate and only changed the land use, the increase would be 18% of the burned area and 47% of fire in the forest litter. The authors point out that the level of interaction between climate and land use depends on the severity of the scenarios and the magnitude of change for each climate model.
Fonseca et al. (2019) also studied the relative and joint effects of climate change and land use change on fire risk. The maximum entropy model was used to simulate the relative fire probability, with CMIP5 data for climate change (RCPs 4.5 and 8.5) and the scenarios described by the IPCC Shared Socioeconomic Pathways (SSP) for land use. The authors found a strong correlation between areas with high fire risk and areas where satellite hotpixels were identified (Figure 4). Looking only at the RCP4.5 scenario, fire risk increases for almost the entire region and, although it is a moderate increase, it does not affect the northwestern
part of the basin. In the RCP8.5 scenario, the change is more significant; however, the extreme northwest of the Amazon basin does not show much change. The authors point out that, for this region, there are few ignition sources, sparse land use, and little change in precipitation and evaporation under the climate change scenarios. The study highlights the importance of land use change on future wildfires, as the authors found a greater increase in fire risk under land use change scenarios than climate change scenarios. Looking at the combined effect, the most pessimistic scenario suggests that the fire season may increase and reach 6 months of the year (June to December).
Figure 4- Simulated relative fire probability (RFP) for the month of July (2003 to 2015). Black crosses mark grid cells in which 13 or more hotpixels were detected. Adapted from Fonseca et al. (2019).
CHAPTER 1
Silva, J.L.G.; Capistrano, V.B.; Veiga, J.A.P; Brito, A.L.
2022. Climate downscaling in the amazon basin:
evaluating fire risk. Manuscript formatted for Acta Amazonica.
Climate downscaling in the amazon basin: evaluating fire risk.
Abstract:
Studies regarding deforestation, the hydrological cycle, climate change and fire weather can benefit from the more detailed simulations that can be provided by Regional Climate Models (RCM). However, while fire activity in the Amazon basin has been a topic of great attention, few studies have used RCM runs to assess fire weather in the region. Therefore, herein, we evaluate the precipitation, temperature and a fire risk index from the ensemble of Eta model simulations coupled with three different global climate models for the Amazon basin. The RCM runs were compared to reanalysis data for the 1979 to 2005 dry season period. The maximum and mean temperature fields were underestimated over the entire study area;
however, they did show a high spatial correlation with the reference data. The precipitation is overestimated over the Amazon region, which is in accordance with the major sources of moisture analyzed. The Keetch-Byram drought index (KBDI) was not greatly affected by the bias found in the previous variables, and the ensemble showed improvements when compared to the individual members on their own. Overall, for the KBDI, there are advantages to performing an ensemble of model simulation and that there are some limitations in the Eta model. Furthermore, the KBDI can also be used in future climate scenarios runs using the Eta model in order to understand how fire activity may be affected by climate change in the Amazon region.
Keywords: KBDI, Amazon Basin, Model Evaluation, Eta Model, Climate Modeling.
Downscaling climático na Bacia Amazônica: Avaliando o risco de fogo Resumo:
Estudos sobre desmatamento, ciclo hidrológico, mudanças climáticas e fogo podem se beneficiar de simulações mais detalhadas que podem ser fornecidas pelos Modelos Climáticos Regionais (RCM). No entanto, embora a atividade de fogo na bacia amazônica tenha sido um tópico de grande atenção, poucos estudos usaram simulações de RCM para avaliar as condições favoráveis ao fogo na região. Portanto, aqui, avaliamos a precipitação, temperatura e índice de risco de fogo a partir de um conjunto de simulações do modelo Eta forçado por resultados de três modelos climáticos globais diferentes para a bacia Amazônica. As simulações do RCM foram comparadas com dados de reanálise para o período da estação seca de 1979 a 2005. Os campos de temperatura máxima e média foram subestimados em toda a área de estudo; no entanto, eles mostraram uma alta correlação espacial com os dados de referência. A precipitação é superestimada na região amazônica, o que está de acordo com as principais fontes de umidade analisadas. O índice de Keetch-Byram (KBDI) não foi muito afetado pelo viés encontrado nas variáveis anteriores, e o conjunto apresentou melhor estatística quando comparado aos membros individuais. No geral, para o KBDI, há vantagens em realizar um conjunto de simulação de modelos, entretanto o KBDI usando dados do modelo Eta apresenta limitações. Além disso, o KBDI pode ser calculado em cenários climáticos futuros usando o modelo Eta para entender como o risco potencial de fogo pode ser afetado pelas mudanças climáticas na região amazônica.
Palavras-chave: KBDI, Bacia Amazônica, Avaliação de modelo, Modelo Eta, Modelagem climática.
INTRODUCTION
Global climate models (GCM) are a powerful tool and are capable of simulating the present and future climate. In order to function on a global scale, GCMs generally simulate the climate on a horizontal grid of 100-200 km, which makes them unable to represent some land features, land-use and land-ocean details, urban areas, and many others (Ambrizzi et al.
2019). Thus, a GCM has some limitations, such as the inability to simulate at a high level of detail, as well as the lack of representation of small-scale processes and detailed near-surface variables (Cabré et al. 2014). One way of surpassing this limitation is to use a Regional Climate Model (RCM) with increased spatial resolution over a limited and smaller domain, and forcing it with lateral boundary conditions that are given by GCMs (Dickinson et al.
1989). Therefore, RCMs can be used as a tool for studies that need a higher horizontal resolution, which may be for many different purposes (Xuejie et al. 2001; Campbell et al.
2011; Karmalkar et al. 2011; Van Oldenborgh et al. 2013; De Jong et al. 2019).
In the Amazon, regional models have already been used to validate the present climate (Chou et al. 2012; Builes-Jaramillo & Pántano, 2021), to access the effects of deforestation on regional circulation (Ruiz-Vásquez et al. 2020; De Sales et al. 2020) and on the hydrological cycle (Gomes et al. 2020), as well as in studies on climate change (Llopart et al. 2014; Rocha et al. 2015). However, for fire weather and fire risk, while some studies on a global scale have been carried out (Liu et al. 2010; Fonseca et al. 2019; Gannon & Steinberg, 2021), there is a lack of studies for the Amazon region, i.e., on a regional scale. As shown in previous research, in the face of climate change scenarios, the Amazon basin may present favorable conditions for increased fire occurrence (Marengo et al. 2018; Vogel et al. 2020). This is particularly important since the Amazon Rainforest has around 86 Pg C, and almost 80% of this biomass is above ground (Saatchi et al. 2007). When burned, the Amazon Rainforest can become a significant source of carbon for the atmosphere. Thus, it is essential to understand the relationship between climate conditions and fire in the Amazon.
Many efforts have already been made to study the relationship between deforestation, drought and fires in the Amazon Forest. The forest’s vulnerability to fires is enhanced by drought events (Aragão et al. 2014), and this impact has been exacerbated over the years (Anderson et al. 2018). These drought events have a direct impact on carbon emission, which can be seen in the 2010, 2015 and 2016 droughts, when the anomalous fires in Amazon Rainforest were responsible for a combined emission of 0.74 PgC (Silva Junior et al. 2019).
Moreover, since the Amazon region has high spatial variability in the drought season,
recently, there have been efforts to understand fire seasonality (Carvalho et al. 2021).
Therefore, as fires in the Amazon region and climate change are receiving more attention from the scientific community, it becomes necessary not only to understand the dynamics of fire and how past events have occurred, but also investigate how future possible climate scenarios may affect fire weather in the Amazon.
Herein, we investigate the errors and uncertainties of simulations of the present climate for the dry period of the region (July, August and September), as well as the model’s ability to represent the fire risk index. The evaluation was done with the Eta model nested with the Brazilian Earth System Model (BESM), the UK Met Office Hadley Centre Global Environmental Model (HadGEM2-ES) and the Model for Interdisciplinary Research on Climate (MIROC5), for the period 1979 to 2005. Via this, we aim to understand the physical processes related to optimal meteorological conditions for fire occurrence in the Amazon basin and also to evaluate the applicability of the fire index for climate change studies from this Eta model configuration.
MATHERIALS & METHODS Study region
The Amazon basin is the largest hydrographic basin in the world, and is named after its main river, the Amazon River. It covers an area of approximately 6.2 million square kilometers and encompasses the Brazilian states of Acre, Amapá, Amazonas, Pará, Roraima, Rondônia, Mato Grosso, Maranhão and Tocantins, as well as the countries of Bolivia, Colombia, Ecuador, Guyana, French Guiana, Peru, Suriname and Venezuela. The average accumulated rainfall in the region is approximately 2,300 mm year-1 and the average temperature values vary between 24 °C and 26 °C, with low thermal amplitude throughout the year (Fisch et al. 1998). One of the defining characteristics of the region is the high spatial and temporal variability of rainfall (Sombroek, 2001; Espinoza-Villar et al. 2009). The spatial and temporal variability over the Amazon Rainforest occurs due to the systems that act over the region, such as the Intertropical Convergence Zone (Mehta, 1998; Wang & Fu, 2007), the South Atlantic Convergence Zone (Kodama, 1992), the Bolivian high (Lenters & Cook, 1997), the Pacific Decadal Oscillation and the El Nino-Southern Oscillation (ENSO) (Marengo, 2004; Espinoza-Villar et al. 2009).
Model description and simulation strategy
In this study, the Eta regional climate model, nested with three different Earth System Models (ESMs), was used to simulate the present climate. The three runs were performed for the period 1960 to 2005 using the initial and boundary conditions of the ESMs BESM, HadGEM2-ES and MIROC5. The experiment was performed continuously throughout the 45 years, initiating at 00:00 GMT of January 1st, 1960. Levels of CO2 were fixed at 330 ppm.
The vegetation map used for by the Eta model was obtained from the ProVeg project (Sestini et al. 2002), and updated with deforestation data for the base year of 2015 from Projeto de Monitoramento do Desmatamento da Floresta Amazônica Brasileira por Satélite - PRODES.
Regional circulation model
The regional climate model used for this study was the Eta model. One of the main features of the Eta regional model is the Eta vertical coordinate (η), defined by Mesinger (1984), which has the benefit of reducing the error in calculations near steep surfaces of variables such as the strength of the pressure gradient, advective processes and horizontal diffusion (Dereczynski et al. 2000). The Eta model uses Arakawa’s E-type horizontal grid (Arakawa & Lamb, 1977) and, for this study, it was configured with a horizontal resolution of 20 km and 38 vertical levels.
For the simulations, the Eta model used the parametrization of turbulent diffusion in the planetary boundary layer proposed by Mellor & Yamada (1974). Shortwave radiation (Lacis & Hansen, 1974) and longwave radiation (Fels & Schwarzkopf, 1975) parametrizations are present. In addition, cumulus parametrization, using the Betts-Miller- Janjic scheme (Janjić, 1994), and cloud microphysics (Zhao et al. 1997) are used to simulate precipitation. Land processes are represented by the NOAH land model (Ek et al. 2003).
To perform these simulations, the Eta regional model was subjected to two adaptations. The first adaptation occurred in order for the regional model to use the monthly averages of sea surface temperature (SST) provided by the Earth System Models (BESM, HadGEM2-ES and MIROC5). The second was a calendar modification, giving a 360-day year, in order for the Eta model to be compatible with the HadGEM2-ES lateral boundary conditions.
Earth system models
The Brazilian Earth System Model (BESM) is an earth system model built from a national effort aimed at understanding global climate change, its causes, effects, and impacts
on society (Nobre et al. 2013; Capistrano et al. 2020). It was set up at a horizontal resolution of approximately 200 km and 28 vertical levels. In order to represent atmospheric and land surface processes, BESM uses the Brazilian Atmospheric Model (BAM) (Figueroa et al.
2016) and the Simplified Simple Biosphere Model - SSiB (Xue et al. 2001), respectively. The shortwave radiation calculation is based on the CLIRAD-SW-M model (Tarasova et al. 2007), and the Harshvardhan et al. (1987) scheme calculates the longwave radiation. The cloud interaction is based on the scheme described in Slingo (1987), Hou (1990) and Kinter et al.
(1997). The BESM show significant bias when simulating the ITCZ over both Atlantic and Pacific oceans, but has a good representation of rainfall over the South America, due to the representation of South Atlantic Convergence Zone (Veiga et al. 2019).
The UK Met Office Hadley Centre Global Environmental Model, version 2 (HadGEM2-ES) is an earth system model (Collins et al. 2011). The horizontal resolution of the atmospheric component is N96, approximately 1.875° x 1.25°, with 38 vertical levels. The global dynamic vegetation model TRIFFID (Top-down Representation of Interactive Foliage Including Dynamics) (Cox, 2001) was used to describe the terrestrial vegetation and carbon cycle. In regards to the ocean, its biological and chemical processes were represented by the DiatHadOCC model. The UKCA model (United Kingdom Chemistry and Aerosol model) was used to calculate the chemistry of the troposphere. The HadGEM2-ES shows a positive bias in the dry period over the amazon, however, over South America the climate model is able to represent the annual variability of rainfall (Cavalcanti and Shimizu, 2012).
The Model for Interdisciplinary Research on Climate (MIROC) version 5 is described in detail by Watanabe et al. (2010). The atmospheric spectral component of the model has T85 resolution, which corresponds to approximately 150 km horizontally and has 40 vertical levels. The ocean coupling was performed using the COCO 4.5 model, which has 1° of resolution horizontally and 50 depth levels. The radiative transfers are calculated using the k distribution scheme (Sekiguchi & Nakajima, 2008). The model has a cloud microphysics scheme that is coupled with the radiation scheme, and is called SPRINTARS. To represent surface processes MIROC5 uses the MATSIRO scheme (Takata et al. 2003). Overall, the MIROC5 has a good representation of precipitation over the Amazon region (Watanabe et al.
2010) and is able to simulate better ITZC position in comparison with previous versions (Hirota et al. 2011)
Fire risk index
The Keetch-Byram drought index (KBDI) was used to estimate soil moisture through precipitation and maximum daily temperature. Since soil moisture deficiency and droughts can influence the flammability of vegetation, the KBDI is used as a tool for identifying dry areas that are susceptible to the occurrence of wildfires (Keetch & Byram, 1968). The KBDI has been applied in tropical areas (Dolling et al. 2005; Taufik et al. 2015) and specifically in the Amazon biome (Nogueira et al. 2017; Cavalcante et al. 2021), with good results for correlating the potential with occurrences of fires. The main advantage of using this index is that it only requires two meteorological variables (daily maximum temperature and daily precipitation). It gives values ranging from 0, when there is no soil moisture deficiency, to 800, which denotes absolute soil moisture deficiency (Heim, 2002). KBDI values correlates to fire risk in a scale where values from 0 to 200 are considered low fire risk, from 200 to 400 the risk is moderate, 400 to 600 the fire risk is high, and 600 to 800 is considered a very high risk of fire (Liu et al. 2010).
The KBDI, with variables expressed in terms of the international system unit, can be calculated in a daily scale from equations 1 and 2 (Crane 1982):
𝐾𝐵𝐷𝐼 = 𝑄𝑡−1+ dQ − dP (1)
𝑑𝑄 =[203.2 − 𝑄][0.968𝑒𝑥𝑝 (0.0875𝑇 + 1.5552) − 8.30]𝑑𝑡
1 + 10.88𝑒𝑥𝑝 (−0.001736𝑅) x10−3 (2) where dP is the daily precipitation (mm), dQ is the drought factor (mm), the KBDI is the moisture deficiency in mm, T is the maximum daily temperature (°C), and R is the average annual precipitation in mm.
Observed data and model validation
The NOAA Climate Prediction Center’s (CPC) precipitation and maximum temperature dataset (Xie et al. 2010) is used here to assess the ensemble of the Eta simulations for the years 1979-2005. The CPC dataset is available at a regular spatial resolution of 0.50 x 0.50 degrees and ranges from January 1979 to the present day. The KBDI, which was calculated by precipitation and maximum temperature simulated by the Eta model, was compared against the KBDI calculated by the same variables from the CPC dataset. Similarly, the simulated mean temperature was evaluated using the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 monthly reanalysis data at 0.25 degrees (Hersbach et al.
2020).
The validation took place over the dry period (July, August and September), using a three-member ensemble (Eta-BESM, Eta-HadGEM2-ES, Eta-MIROC5). Statistical significance of the bias is calculated by the two-sample t test (Wilks, 2011) with 95%
confidence. For comparison, the bias method was used on mean values and, in order to evaluate the pattern between the reference and simulations, the Taylor diagram (Taylor, 2005) was used.
RESULTS
Mean temperature
The mean temperature bias is shown in Figure 1a. The southern part of the Amazon basin is the area where the models simulated the warmest mean temperature. However, the ensemble models also show a cold bias in the northern and western areas of the basin. The positive bias was mainly due to the Eta-BESM run (shown in Figure S1), which overestimated the mean temperature in the entire basin with the exception of the Andes region. Removing the Andes region on the analysis did not change significantly the results as the error persists over its borders. On the other hand, Eta-MIROC5 presented the greatest cold bias between the members of the ensemble, although Eta-HadGEM2-ES also demonstrates a cold bias for the northern, western and central regions of the Amazon basin (Figures S2 and S3, respectively). The Taylor diagram shows the comparison between the reference and the Eta model simulations (Figure 1b). The points from 1 to 4 are the different climate simulations and the red star is the reference data. In other words, the closer the points are to the reference, the better the simulation was able to represent the temperature in the time window of the study. Firstly, it is possible to notice that all simulations of the Eta regional model are very close to each other. This indicates that all simulations, even when forced with different initial and boundary conditions, have a very close mean temperature pattern output.
In addition, the simulations showed high spatial correlation with the reference data, with the Eta-HadGEM2-ES simulation having the highest correlation (0.97). The ensemble correlation (0.96) is the second best, and showed an improvement on the Eta-BESM and Eta-MIROC5 runs.
