Drivers of Fire Anomalies in the Brazilian Amazon: Lessons Learned from the 2019 Fire Crisis

Use of machine learning to identify protective factors for death from COVID-19 in the ICU: a retrospective study

Background

Patients in serious condition due to COVID-19 often require special care in intensive care units (ICUs). This disease has affected over 758 million people and resulted in 6.8 million deaths worldwide. Additionally, the progression of the disease may vary from individual to individual, that is, it is essential to identify the clinical parameters that indicate a good prognosis for the patient. Machine learning (ML) algorithms have been used for analyzing complex medical data and identifying prognostic indicators. However, there is still an urgent need for a model to elucidate the predictors related to patient outcomes. Therefore, this research aimed to verify, through ML, the variables involved in the discharge of patients admitted to the ICU due to COVID-19.

Methods

In this study, 126 variables were collected with information on demography, hospital length stay and outcome, chronic diseases and tumors, comorbidities and risk factors, complications and adverse events, health care, and vital indicators of patients admitted to an ICU in southern Brazil. These variables were filtered and then selected by a ML algorithm known as decision trees to identify the optimal set of variables for predicting patient discharge using logistic regression. Finally, a confusion matrix was performed to evaluate the model's performance for the selected variables.

Results

Of the 532 patients evaluated, 180 were discharged: female (16.92%), with a central venous catheter (23.68%), with a bladder catheter (26.13%), and with an average of 8.46- and 23.65-days using bladder catheter and submitted to mechanical ventilation, respectively. In addition, the chances of discharge increase by 14% for each additional day in the hospital, by 136% for female patients, 716% when there is no bladder catheter, and 737% when no central venous catheter is used. However, the chances of discharge decrease by 3% for each additional year of age and by 9% for each other day of mechanical ventilation. The performance of the training data presented a balanced accuracy of 0.81, sensitivity of 0.74, specificity of 0.88, and the kappa value was 0.64. The test performance had a balanced accuracy of 0.85, sensitivity 0.75, specificity 0.95, and kappa value of 0.73. The McNemar test found that there were no significant differences in the error rates in the training and test data, suggesting good classification. This work showed that female, the absence of a central venous catheter and bladder catheter, shorter mechanical ventilation, and bladder catheter duration were associated with a greater chance of hospital discharge. These results may help develop measures that lead to a good prognosis for the patient.

The potential impact of PM2.5 on the covid-19 crisis in the Brazilian Amazon region

OBJECTIVE

This study aims to assess covid-19 morbidity, mortality, and severity from 2020 to 2021 in five Brazilian Amazon states with the highest records of wildfires.

METHODS

A distributed lag non-linear model was applied to estimate the potential exposure risk association with particulate matter smaller than 2.5-µm in diameter (PM2.5). Daily mean temperature, relative humidity, percentual of community mobility, number of hospital beds, days of the week, and holidays were considered in the final models for controlling the confounding factors.

RESULTS

The states of Para, Mato Grosso, and Amazonas have reported the highest values of overall cases, deaths, and severe cases of covid-19. The worrying growth in the percentual rates in 2020/2021 for the incidence, severity, and mortality were highlighted in Rondônia and Mato Grosso. The growth in 2020/2021 in the estimations of PM2.5 concentrations was higher in Mato Grosso, with an increase of 24.4%, followed by Rondônia (14.9%).

CONCLUSION

This study establishes an association between wildfire-generated PM2.5 and increasing covid-19 incidence, mortality, and severity within the studied area. The findings showed that the risk of covid-19 morbidity and mortality is nearly two times higher among individuals exposed to high concentrations of PM2.5. The attributable fraction to PM2.5 in the studied area represents an important role in the risk associated with covid-19 in the Brazilian Amazon region.

Human impacts outpace natural processes in the Amazon

Amazonian environments are being degraded by modern industrial and agricultural activities at a pace far above anything previously known, imperiling its vast biodiversity reserves and globally important ecosystem services. The most substantial threats come from regional deforestation, because of export market demands, and global climate change. The Amazon is currently perched to transition rapidly from a largely forested to a nonforested landscape. These changes are happening much too rapidly for Amazonian species, peoples, and ecosystems to respond adaptively. Policies to prevent the worst outcomes are known and must be enacted immediately. We now need political will and leadership to act on this information. To fail the Amazon is to fail the biosphere, and we fail to act at our peril.

Achieving Brazil's Deforestation Target Will Reduce Fire and Deliver Air Quality and Public Health Benefits

Climate, deforestation, and forest fires are closely coupled in the Amazon, but models of fire that include these interactions are lacking. We trained machine learning models on temperature, rainfall, deforestation, land-use, and fire data to show that spatial and temporal patterns of fire in the Amazon are strongly modified by deforestation. We find that fire count across the Brazilian Amazon increases by 0.44 percentage points for each percentage point increase in deforestation rate. We used the model to predict that the increased deforestation rate in the Brazilian Amazon from 2013 to 2020 caused a 42% increase in fire counts in 2020. We predict that if Brazil had achieved the deforestation target under the National Policy on Climate Change, there would have been 32% fewer fire counts across the Brazilian Amazon in 2020. Using a regional chemistry-climate model and exposure-response associations, we estimate that the improved air quality due to reduced smoke emission under this scenario would have resulted in 2,300 fewer deaths due to reduced exposure to fine particulate matter. Our analysis demonstrates the air quality and public health benefits that would accrue from reducing deforestation in the Brazilian Amazon.

Dynamics of Fire Foci in the Amazon Rainforest and Their Consequences on Environmental Degradation

Burns are common practices in Brazil and cause major fires, especially in the Legal Amazon. This study evaluated the dynamics of the fire foci in the Legal Amazon in Brazil and their consequences on environmental degradation, particularly in the transformation of the forest into pasture, in livestock and agriculture areas, mining activities and urbanization. The fire foci data were obtained from the reference satellites of the BDQueimadas of the CPTEC/INPE for the period June 1998–May 2022. The data obtained were subjected to descriptive and exploratory statistical analysis, followed by a comparison with the PRODES data during 2004–2021, the DETER data (2016–2019) and the ENSO phases during the ONI index for the study area. Biophysical parameters were used in the assessment of environmental degradation. The results showed that El Niño’s years of activity and the years of extreme droughts (2005, 2010 and 2015) stand out with respect to significant increase in fire foci. Moreover, the significant numbers of fire foci indices during August, September, October and November were recorded as 23.28%, 30.91%, 15.64% and 10.34%, respectively, and these were even more intensified by the El Niño episodes. Biophysical parameters maps showed the variability of the fire foci, mainly in the south and west part of the Amazon basin referring to the Arc of Deforestation. Similarly, the states of Mato Grosso, Pará and Amazonas had the highest alerts from PRODES and DETER, and in the case of DETER, primarily mining and deforestation (94.3%) increased the environmental degradation. The use of burns for agriculture and livestock, followed by mining and wood extraction, caused the degradation of the Amazon biome.

Highway Network and Fire Occurrence in Amazonian Indigenous Lands

The construction and expansion of highways aiming to improve the integration of the most isolated regions in Brazil facilitated the access to many inhabited areas in the Amazon biome, but had as a consequence assisted the degradation of many of these regions. Over the last two decades, we have observed in this biome a gradual diversification and intensification of land uses through vegetation loss and an increase in fire associated with deforestation and an increase in grazing areas. We used data from several active fires products derived from 14 different satellites, available on the Brazilian National Institute for Space Research (INPE). We evaluated the influence of highway infrastructure on fire occurrence inside and around Indigenous Lands (IL) located in the Brazilian Amazon biome, from 2008 to 2021. We classified 332 ILs into “cut by highways”, “without highways”, and “with highways in a 10 km buffer”. We performed: (a) the descriptive statistics of the fire occurrence by state, by season, and by type of land use and land cover (LULC) affected by fire; (b) the spatial distribution of the active fire density; and (c) a simple linear regression model between the fire occurrence and the IL area. Our results showed that in total, 16–46% of the fires occurred within the IL in most of the states, while the 10 km buffer was the region most affected by fire. We confirmed that in the last three years there was a significant increase in the number of active fires, representing anomalies in fire occurrence across the studied period. We discussed the result implications and the role of the highway network in environmental degradation inside and around the ILs located in the Brazilian Amazon.

Characterizing Spatial Patterns of Amazon Rainforest Wildfires and Driving Factors by Using Remote Sensing and GIS Geospatial Technologies

Known as the “lung of the planet,” the Amazon rainforest produces more than 20% of the Earth’s oxygen. Once a carbon pool for mitigating climate change, the Brazilian Amazônia Biome recently has become a significant carbon emitter due to increasingly frequent wildfires. Therefore, it is of crucial importance for authorities to understand wildfire dynamics to manage them safely and effectively. This study incorporated remote sensing and spatial statistics to study both the spatial distribution of wildfires during 2019 and their relationships to 15 environmental and anthropogenic factors. First, broad-scale spatial patterns of wildfire occurrence were explored using kernel density estimation, Moran’s I, Getis-Ord Gi*, and optimized hot spot analysis (OHSA). Second, the relationships between wildfire occurrence and the environmental and anthropogenic factors were explored using several regression models, including Ordinary Least Squares (OLS), global (quasi) Poisson, Geographically-weighted Gaussian Regression (GWGR), and Geographically-weighted Poisson Regression (GWPR). The spatial analysis results indicate that wildfires exhibited pronounced regional differences in spatial patterns in the vast and heterogeneous territory of the Amazônia Biome. The GWPR model outperformed the other regression models and explained the distribution and frequency of wildfires in the Amazônia Biome as a function of topographic, meteorologic, and environmental variables. Environmental factors like elevation, slope, relative humidity, and temperature were significant factors in explaining fire frequency in localized hotspots, while factors related to deforestation (forest loss, forest fragmentation measures, agriculture) explained wildfire activity over much of the region. Therefore, this study could improve a comprehensive study on, and understanding of, wildfire patterns and spatial variation in the target areas to support agencies as they prepare and plan for wildfire and land management activities in the Amazônia Biome.

Three Decades after: Landscape Dynamics in Different Colonisation Models Implemented in the Brazilian Legal Amazon

Several colonisation projects were implemented in the Brazilian Legal Amazon in the 1970s and 1980s. Among these colonisation projects, the most prominent were those with the “fishbone” and “topographic” models. Within this scope, the settlements known as Anari and Machadinho stand out because they are contiguous areas with different models and structures of occupation and colonisation. The main objective of this work was to evaluate the dynamics of Land-Use and Land-Cover (LULC) in two different colonisation models, implanted in the State of Rondônia in the 1980s. The fishbone and topographic or Disorganised Multidirectional models were implemented in the Anari and Machadinho settlements, respectively. A 36-year time series of Landsat images (1984–2020) was used to evaluate the rates and trends in the LULC process in the different colonisation models. In the analysed models, a rapid loss of primary and secondary forests (anthropized areas) was observed, mainly due to the dynamics of its use, established by the Agriculture/Pasture relation with a heavy dependence on road construction. Understanding these two forms of occupation can help the future programs and guidelines of the Brazilian Legal Amazon and any tropical rainforest across the globe.

Analysis of how the spatial and temporal patterns of fire and their bioclimatic and anthropogenic drivers vary across the Amazon rainforest in El Niño and non-El Niño years

In the past two decades, Amazon rainforest countries (Brazil, Bolivia, Colombia, Ecuador, Guyana, Peru and Venezuela) have experienced a substantial increase in fire frequency due to the changes in the patterns of different anthropogenic and climatic drivers. This study examines how both fire dynamics and bioclimatic factors varied based on the season (wet season and dry season) El Niño years across the different countries and ecosystems within the Amazon rainforest. Data from publicly available databases on forest fires (Global Fire Atlas) and bioclimatic, topographic and anthropogenic variables were employed in the analysis. Linear mixed-effect models discovered that year type (El Niño vs. non-El Niño), seasonality (dry vs. wet), land cover and forest strata (in terms of canopy cover and intactness) and their interactions varied across the Amazonian countries (and the different ecosystems) under consideration. A machine learning model, Multivariate Adaptive Regression Spline (MARS), was utilized to determine the relative importance of climatic, topographic, forest structure and human modification variables on fire dynamics across wet and dry seasons, both in El Niño and non-El Niño years. The findings of this study make clear that declining precipitation and increased temperatures have strong impact on fire dynamics (size, duration, expansion and speed) for El Niño years. El Niño years also saw greater fire sizes and speeds as compared to non-El Niño years. Dense and relatively undisturbed forests were found to have the lowest fire activity and increased human impact on a landscape was associated with exacerbated fire dynamics, especially in the El Niño years. Additionally, the presence of grass-dominated ecosystems such as grasslands also acted as a driver of fire in both El Niño and non-El Niño years. Hence, from a conservation perspective, increased interventions during the El Niño periods should be considered.

Forest Canopy Changes in the Southern Amazon during the 2019 Fire Season Based on Passive Microwave and Optical Satellite Observations

Canopy dynamics associated with fires in tropical forests play a critical role in the terrestrial carbon cycle and climate feedbacks. The aim of this study was to characterize forest canopy dynamics in the southern Amazon during the 2019 fire season (July–October) using passive microwave-based vegetation optical depth (VOD) and three optical-based indices. First, we found that precipitation during July–October 2019 was close to the climatic means, suggesting that there were no extreme hydrometeorological events in 2019 and that fire was the dominant factor causing forest canopy anomalies. Second, based on the active fire product (MCD14ML), the total number of active fires over each grid cell was calculated for each month. The number of active fires during the fire season in 2019 was above average, particularly in August and September. Third, we compared the anomalies of VOD and optical-based indices (the normalized difference vegetation index (NDVI), the enhanced vegetation index (EVI), and the normalized burn ratio (NBR)) against the spatiotemporal distribution of fires during July–October 2019. Spatially, the location with a concentrated distribution of significant negative VOD anomalies was matched with the grid cells with fire activities, whereas the concentrated distribution of strong negative anomalies in optical-based indices were found in both burned and unburned grid cells. When we focused on the temporal pattern over the grid cells with fire activity, the VOD and the optical-based indices behaved similarly from July to October 2019, i.e., the magnitude of negative anomalies became stronger with increased fire occurrences and reached the peak of negative anomalies in September before decreasing in October. A discrepancy was observed in the magnitude of negative anomalies of the optical-based indices and the VOD; the magnitude of optical-based indices was larger than the VOD in August–September and recovered much faster than the VOD over the grid cells with relatively low fire activity in October. The most likely reason for their different responses is that the VOD represents the dynamics of both photosynthetic (leaf) and nonphotosynthetic (branches) biomass, whereas optical-based indices are only sensitive to photosynthetic (leaf) active biomass, which recovers faster. Our results demonstrate that VOD can detect the spatiotemporal of canopy dynamics caused by fire and postfire canopy biomass recovery over high-biomass rainforest, which enables more comprehensive assessments, together with classic optical remote sensing approaches.

Drought-driven wildfire impacts on structure and dynamics in a wet Central Amazonian forest

While the climate and human-induced forest degradation is increasing in the Amazon, fire impacts on forest dynamics remain understudied in the wetter regions of the basin, which are susceptible to large wildfires only during extreme droughts. To address this gap, we installed burned and unburned plots immediately after a wildfire in the northern Purus-Madeira (Central Amazon) during the 2015 El-Niño. We measured all individuals with diameter of 10 cm or more at breast height and conducted recensuses to track the demographic drivers of biomass change over 3 years. We also assessed how stem-level growth and mortality were influenced by fire intensity (proxied by char height) and tree morphological traits (size and wood density). Overall, the burned forest lost 27.3% of stem density and 12.8% of biomass, concentrated in small and medium trees. Mortality drove these losses in the first 2 years and recruitment decreased in the third year. The fire increased growth in lower wood density and larger sized trees, while char height had transitory strong effects increasing tree mortality. Our findings suggest that fire impacts are weaker in the wetter Amazon. Here, trees of greater sizes and higher wood densities may confer a margin of fire resistance; however, this may not extend to higher intensity fires arising from climate change.

Roads in the Southwestern Amazon, State of Acre, between 2007 and 2019

Over the past 40 years, roads have been the main driver behind the State of Acre’s occupation and development. However, the expansion of roads, has often been associated with the advance of deforestation, habitat fragmentation, and social conflicts. There are no up-to-date data available on the current extent of Acre’s road network nor its environmental and socioenvironmental impacts. In this study, we updated the State of Acre’s road network map for the period 2007 to 2019 through the visual interpretation of 153 Landsat images (5, 7, and 8) at a scale of 1:50,000. To estimate the impact of roads, we measured the distribution of roads in municipalities and in different land tenure categories and calculated the correlation between roads and annual deforestation. Up to 2019, we estimated 19,620 km of roads, of which 92% were unofficial roads, 6% federal roads, and 2% state roads. The roads increased at an average annual rate of 590 km year−1. The most significant advance in road length between 2007 and 2019 was in protected areas (240%), followed by public lands (68%) and settlement projects (66%). We recommend monitoring of the road network to understand the landscape’s evolution and support actions against illicit environmental and socioenvironmental impacts.