Next Article in Journal
Using Count Regression to Investigate Millennial-Scale Vegetation and Fire Response from Multiple Sites Across the Northern Rocky Mountains, USA
Previous Article in Journal
A Cross-Sectional Assessment of the Individual- and Fire Department-Level Factors Affecting Volunteer Firefighter Cardiorespiratory Fitness
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Fire
Volume 8
Issue 8
10.3390/fire8080320
fire-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

A Method for Mapping and Associating Burned Areas with Agricultural Practices Within the Brazilian Cerrado

by
Pâmela Inês de Souza Castro Abreu
1,
George Deroco Martins
1,
Gabriel Henrique de Almeida Pereira
2,
Rodrigo Bezerra de Araujo Gallis
3,
Jorge Luis Silva Brito
3,
Carlos Alberto Matias de Abreu Júnior
1,
Laura Cristina Moura Xavier
3,* and
João Vitor Meza Bravo
3,*
1
Institute of Agricultural Sciences, Federal University of Uberlândia, Uberlândia 38400-902, MG, Brazil
2
Paraná Environmental Monitoring and Technology System, Federal University of Parana, Curitiba 81530-900, PR, Brazil
3
Institute of Geography, Geosciences and Public Health, Federal University of Uberlândia, Uberlândia 38400-902, MG, Brazil
*
Authors to whom correspondence should be addressed.
Fire 2025, 8(8), 320; https://doi.org/10.3390/fire8080320
Submission received: 16 June 2025 / Revised: 25 July 2025 / Accepted: 4 August 2025 / Published: 13 August 2025
Browse Figures
Versions Notes

Abstract

Fire occurs naturally and anthropogenically in the Cerrado biome, influenced by hydrology, climate, topography, and land use. Mapping burned areas is essential for understanding the causes of fire and improving prevention and regulation. However, fire scars are often confused with bare soil in agricultural regions. This study presents a method for mapping burned areas using spectral indices and artificial neural networks (ANN). We evaluated the accuracy of these techniques and identified the best input variables for scar detection. Using Sentinel-2 images from 2018 to 2021 during dry periods, we applied NDVI, SAVI, NBR, and CSI indices. The study included two stages: first, finding optimal classification configurations for fire scars, and second, mapping land use and cover with fire scars and crops. Results showed that using all Sentinel-2 bands and the four indices post-fire achieved over 93.7% accuracy and a kappa index of 0.92. Fire scars were mainly located in areas with temporary crops like soybean, sugarcane, rice, and cotton. This low-cost method allows for effective monitoring of fire scars, underscoring the need to regulate agricultural practices in the Cerrado, where burning poses environmental and health risks.

1. Introduction

Land use and land cover (LULC) are constantly being modified, especially concerning vegetation cover. When vegetation suffers from any change, there is also a modification in ecosystems, which can, within the context of environmental devastation, reduce resources in soil, fauna, and flora [1]. One factor contributing to modifying vegetation cover is forest fires, which can occur through natural or human causes [2,3]. Forest fires frequently occur in regions with a dry climate, wind, available dry vegetation, and favorable topographical conditions [4,5,6,7].
With ever-changing land cover, especially in the context of climate change, forest fires destroy thousands of hectares yearly, threatening the ecological environment and human life [8]. A worrying issue presented by forest fires is how quickly they spread, making them difficult to contain [9]. Detecting forest fires and wildfires is gaining more attention in research worldwide [10,11].
In the Brazilian context, the Cerrado biome presented the highest number of hotspots related to fires and wildfires in 2021 (n = 137,631) [12]. Some research has linked these fires to uncontrolled or unsustainable agricultural practices [13,14]. Fire can cause several adverse effects, such as problems related to human health [15] and climate change at regional and global levels [16]. Thus, the study of fires is critical since there is an ever-increasing need to preserve and recover natural resources such as vegetation [17,18].
The Cerrado biome is the savannah that houses the most extensive biodiversity in the world [19,20], occupying 2,036,448 km2, equivalent to approximately 23.9% of the Brazilian territory [20]. According to MMA, 2021, the Cerrado biome has about 11,627 cataloged native plant species [21]. The Cerrado biome combines characteristics conducive to fires and wildfires. The dry period occurs from May to September [22], and the prevailing climate is seasonal tropical [23]. The vegetation predominantly comprises forest, savannah, grassland formations, and terrain with topography ranging from flat to smooth and rounded [24]. Remarkably, physiographic characteristics favor agricultural production in this region [25]. Here, the focal point is that the Cerrado biome brings together a set of characteristics conducive to spreading fire [26].
Specifically, with the growing demand for agricultural land in the Cerrado, management practices occur that, when incorrect, increase forest fire risk [27]. For example, a common practice is using fire to clean and prepare pastures for planting, which can result in uncontrolled fire, a significant threat to biodiversity conservation [2].
Considering the importance of preserving the environment and avoiding continuous degradation, it becomes necessary to identify the regions that suffer from fires more frequently [28]. For this, information should be collected regarding the locality of fires, aiming to implement strategies to minimize their occurrence [29]. The research investigations examining forest fires usually observe the precise location of occurrence to obtain accurate information about the ignition point [30]. In contrast, field research demands time and resources, making it challenging to obtain data on the historical basis of the event, limiting fire prediction [31,32].
One solution to overcome such limitations is using remote sensing methods, making it possible to remotely study the events occurring in each region with available time series data [33]. In addition, in Brazil, records of hot spots are accessible on online platforms, such as the burning program of the National Institute for Space Research—INPE—which makes the data available for free.
Spectral indexes are one of the tools used to distinguish fire-related features in vegetation and regions with bared soil [34]. Spectral indexes are mathematical expressions that combine spectral bands to highlight information on the surface [35]. Each spectral index can highlight specific information while discriminating it from other objects on the surface [36]. The Normalized Burn Ratio (NBR) is among the most widely used indexes for mapping the severity of forest fires and fire scars [37]. The NBR index focuses on information regarding ash, coal, soil, and surface moisture [38]. The NDVI is most often used to analyze vegetative vigor on the earth’s surface [39]. The SAVI index was created to minimize soil’s effects on vegetation detection, and the factor L varies from 0 to 1, according to the type of vegetation cover [40]. Regions with high vegetation cover receive L = 0, with moderate vegetation L = 0.5 and low vegetation L = 1 [40]. The CSI index detects recently or severely burned areas [41].
Agriculture, especially in regions with a predominance of temporary crops, can hide the detection of fire scars since these crops have phenological stages [42] that modify the spectral response [43,44]. Thus, a given temporary crop region, for example, has different spectral responses during the crop cycle since there will be modifications in the physiognomic and physiological structure of the plant.
Because spectral indexes are mathematical expressions that work with reflectance values, it is natural that the detection of phenomena such as fire scars encounters barriers. For example, ref. [45] indicate that the most significant difficulty lies in confusion with the class “bared soil.” Ref. [46] explain that in regions with urban areas or bared soil, there may be confusion with the burning class due to the sensitivity of the indexes in correctly discriminating land cover information. Ref. [47] points out that the confusion occurs due to highly dark pixels, which can be confused with areas of bared soil or natural/artificial dark spots. Therefore, it is necessary to fill this knowledge gap using robust tools such as machine learning. According to [48], ML algorithms can work with various data without making previous assumptions regarding their distribution. Machine learning (ML) seeks to extend computational functions to improve performance automatically through its own experience [49]. ML is increasingly being used in remote sensing [50].
Considering these problems, we answered the following questions: What aspects hinder the detection of burned areas when applying spectral indexes to map the occurrence of fire scars? Have agricultural practices in the Cerrado favored the occurrence of fires? Thus, here we developed an accurate and low-cost solution to detect fire scars based on artificial neural networks and satellite images, overcoming the aspects that hinder those scars from detection; we also analyzed the association between the areas with fire scars and the types of crops within a representative region inside the Cerrado.

2. Material and Methods

2.1. Study Area

The microregion of Patrocínio is in the southeast region of Brazil, in the state of Minas Gerais, at the intersection of latitudes ranging from 18°40′44.004″ S to 19°03′58.674″ S, and longitudes 47°09′51.546″ W to 46°32′46.413″ W (Figure 1). The study region has an area of approximately 11,986.16 km2, which covers eleven municipalities, namely: Abadia dos Dourados, Coromandel, Cruzeiro da Fortaleza, Douradoquara, Estrela do Sul, Grupiara, Iraí de Minas, Monte Carmelo, Patrocínio, Romaria, and Serra do Salitre.
The microregion of Patrocínio is known for its agro-industrial activities [51]. The main crops are corn, sugarcane, soybeans, coffee, and animal husbandry [52]. According to [53], agriculture is the main economic activity in this geographic region because of the uneven relief, favorable hydroclimatic conditions, and fertile soils. These conditions save expenses on soil correction inputs and enable agricultural machinery to maximize productive potential at lower costs [54,55].
The study area was chosen because it is in the Cerrado of Minas Gerais, which has several characteristics that make it prone to the occurrence of fires, supporting the hypothesis tested in this article. In addition, its economic importance and potential for supplying natural resources make this area a study priority in the region.
According to data from the National Institute of Meteorology, NMET, the period of scarce rainfall in the region starts around April/May and continues until September/October. Figure 2 shows the rainfall record in the years of study, data which helped us define the period of interest and data collection.
Figure 3 presents the graph of the fire spots detected monthly each year. Figure 4 shows the distribution of fire spots, and Figure 5 illustrates the occurrence of fires per km2 in the municipalities of the Patrocínio microregion from 2018 to 2021 (in dry months). On average, the dry months in the region range from April to October.
We divided the research into two experiments. The objective of experiment 1 was to evaluate the band’s relationship configuration to allow the best detection of burned areas in the image classification process. The aim of experiment 2 was, based on experiment 1, to analyze the association of agricultural practices with the occurrence of fire spots. Thus, in the following items, we describe each procedure in detail.

2.2. Data Acquisition and Pre-Processing

Data on hotspots in the study area were obtained from the INPE website. We searched for the dry season interval for each year of interest. The dry season was established by rainfall data from stations located in the region, ranging from April to October in the years of study. Pre-fire images are used to verify what was in the region before the dry season. Thus, based on the definition of dry spells, images from 2018 to 2021 were acquired, as shown in Table 1. Sentinel-2 sensor MSI images were acquired through the official ESA website (https://dataspace.copernicus.eu/explore-data/data-collections/sentinel-data/sentinel-2, accessed on 15 March 2022), all six referring to the same data.
Considering the presence of particles in the atmosphere, such as water, nitrogen, oxygen, and aerosols, it was necessary to perform atmospheric correction to reduce the particles’ influence on the image [56]. The atmospheric correction of the images was performed using the Semi-Automatic Classification Plugin (SCP) in the QGIS 3.16.4 software. The SCP plugin performs atmospheric correction using the Dark Object Subtraction (DOS) method [57]. The DOS method estimates atmospheric intervention based on the satellite image’s digital numbers (DNs). According to [58] shaded regions should present low DNs; however, due to atmospheric scattering, they have DNs values, and thus these pixels are used as a correction parameter. Ref. [59] states that the analysis is conducted by choosing the shortest wavelength band. From the study of the frequency histogram of the DNs, the value of the darkest pixel is identified. Thus, the correction is made uniformly, subtracting the scattering value of the entire scene [59].
In addition, all the Sentinel-2 satellite bands were resampled to the spatial resolution of 10 m using SNAP 6.0 software. The nearest neighbor method (NNM) was used to perform the resampling, which uses the gray level values closest to the inverse mapping [59].

2.3. Mosaic Generation

Due to the size of the study area, it was necessary to use six scenes to compose its limits, the scenes acquired being KHE, KKA, KKU, KKV, KLU, and KLV. The mosaic was made in the pre- and post-fire periods for each year. We selected only images with 30% or less cloud cover.

2.4. Mapping Fire Scars

For the mapping of the fire scars, the indexes present in Table 2 were used, using NIR (near infrared), SWIR2 (shortwave infrared), and R (red) bands.

2.5. First Experiment—Bands and Indexes Evaluation

In the first experiment, we evaluated the band configuration that allows better detection of burned areas in the image classification process. Thus, six tests were used each year (2018 to 2021), divided into two groups: post-fire period and variation between pre-fire and post-fire period. In the analysis of the methods used for the post-fire period and the variation between the pre-fire and post-fire period, classification performance was tested: (I) with the use of all bands of the Sentinel-2 sensor and the four indexes; (II) only with the NBR index, and (III) only with the CSI index. The pre- and post-fire period variation refers to the modifications in land cover during the dry season, calculated by a mathematical subtraction between pre- and post-fire data.
For this experiment, Region of Interest (ROI) samples were collected for fire scars, bare soil, native vegetation, reforestation, water, and crops. The other classes were acquired only by visual analysis. The fire samples were identified using INPE fire spot data and visual analysis on the Sentinel-2 satellite’s true color composition (RGB). After collection, the vector files were used in the Envi 5.1 software, which extracted the surface reflectance values for each pixel, band, and index.
The samples were divided into two groups, 80% for training and 20% for validation, as shown in Figure 6.
The classes used in this study were as follows: 1: fire scars, 2: exposed soil, 3: native vegetation, 4: reforestation, 5: water, and 6: crops. We used WEKA 3.8.5 software in this study, which produces metrics capable of analyzing the available dataset and estimating the subset with the highest possibility of predicting to which class each pixel belongs. The multilayer perceptron (MLP) neural network was used as a classifier, and the defined hyperparameters were learning rate 0.3, momentum 0.2, and training time 500; the number of neurons in the hidden layers was defined by the arithmetic mean between the number of input variables and the number of output classes. As a result of the processing, the overall accuracy was obtained by computing the kappa index, which analyzes the classification according to Table 3, proposed by [62]. We also used the confusion matrix, which represents the classification quality based on the correlation between the reference and classified information.
In addition to these evaluation parameters, omission error analysis was performed, through which it is possible to verify the percentage of pixels omitted from the classification, as presented in Equation (1).
O E = 100 − V n ∑ r n
where OE: omission error; V: the amount of correctly classified pixels; n: class to be calculated; and r: distribution of reference data. In addition, the inclusion error was analyzed, which analyzes and reports the number of pixels that were wrongly included in each class, as presented in Equation (2).
I E = 100 − V n ∑ c n
where IE: inclusion error; V: the amount of correctly classified pixels; n: class to be calculated; and c: distribution of classified data.

2.6. Second Experiment—Scars Classification

According to the analysis performed in the first experiment, it was possible to determine the best input configuration of the ANN algorithm for classifying burned areas. Thus, the classification was performed using only the images in the post-fire period, together with the use of all bands of the Sentinel-2 satellite and the four indexes we studied.
In this experiment, we used the fire data and the NBR index to refine the samples collected for classification, aiming to improve computational and classifier performance. Thus, we obtained the pixels and their respective surface reflectance values for the burned areas using the ENVI 5.1 software.
Then, WEKA 3.8.5 indicates the classes classified based on the training data as output. In addition, it presents its prediction error, which means the probability that the pixel has been correctly classified. Then, only the areas understood as burned were extracted for each of the years under study. The burning classes were refined using hot spot point data and the NBR index, which allowed the elimination of areas incorrectly classified by WEKA 3.8.5.
To analyze the influence and association of burned areas with the agricultural practices present in the study area, we used raster data from MapBiomas [63], through which we identified information on land use and cover, emphasizing temporary and perennial crops. Thus, the land use and land cover maps were made, allowing an analysis of the presence of fire scars and the type of crops and plantations present near these burned areas.

3. Results and Discussion

3.1. Results and Discussion: First Experiment

The classification process using machine learning showed that using all Sentinel-2 satellite bands combined with the NDVI, SAVI, NBR, and CSI indexes was the most effective model for minimizing classification errors. Both the classification of the post-fire period and the variation between the pre-fire and post-fire period presented a kappa index greater than 0.86; therefore, these classifications were considered near perfect. The comparative results for the different classification configurations tested over the years are summarized in Table 4, which highlights the superior performance of the “all bands + index”configuration compared to the use of isolated indices such as NBR and CSI, in both Kappa and Global Accuaracy.
Table 5 focuses on the results for 2018, and the inclusion and omission of pixels of the burning class were observed. There was the inclusion of 26.731% of pixels, in which 6.485% were classified as reforestation and 20.245% as crops. The omission presented was 0.476%, in which the pixels were classified as native vegetation. Based on the overall accuracy of 97.876% and the kappa index of 0.974, it can be considered that the classification containing all bands of the Sentinel-2 satellite and the four indexes in the post-fire period of the year 2018 presented satisfactory performance, meeting the quality parameter established in Table 3.
Based on Table 6, the pixel inclusion error of approximately 9.328% was observed in the burned class, with about 0.433% being found in the native vegetation class and 8.893% in the reforestation class. The omission error for the burned class was only 0.476%, with these occurring in the water class. According to the overall accuracy of 94.940% and the kappa index of 0.939, the classification of all Sentinel-2 satellites’ bands and indexes in the 2019 post-fire period showed almost perfect quality, as presented in Table 3. These results are similar to those ones found by [64,65], indicating that ML classifiers can reach high accuracy in mapping fire scars.
Compared to other multispectral sensors, we understand these results occurred since the Sentinel-2 satellite bands offer considerably higher accuracy regarding crop biophysical variables [66]. Furthermore, the spectral indexes NDVI, SAVI, NBR, and CSI possess characteristics capable of highlighting regions of fire scars. Therefore, in addition to the surface reflectance values from the sensor bands, the indexes assisted in highlighting the biophysical characteristics of the site [66]. Thus, the combination of the Sentinel-2 bands and the indexes mentioned above resulted in samples with greater information variability. This factor probably contributed to the classification performance since each class could be better characterized.
This behavior is reflected in the results for the years 2020 and 2021. In 2020, the classification presented discrimination among classes, with omission and inclusion errors close to zero. Fire scars were classified with 0% omission and inclusion errors, and the overall accuracy reached 99.841%, with a kappa index of 0.998 (Table 7). In 2021, the model showed the following results: overall accuracy of 97.400% and kappa of 0.968, although some confusion between native vegetation and reforestation was observed, as detailed in Table 8.
In the step of assigning weights in the backpropagation algorithm of a multilayer perceptron neural network, when the input classes are well discriminated, the values assigned to the weights tend to converge to their optimal values with fewer iterations [67]. Moreover, since the error propagated between layers will be smaller, the classification performance tends to present more satisfactory results [68].
The results show classifications with overall accuracies greater than 93%, which can be found in similar research [65]. Additionally, studies like that of [64] reinforce that the proposed methodology effectively maps fire scars. Based on the algorithm used here, it is intended that the post-fire period will make it possible to aid in processing time. Furthermore, according to [69], the vegetation of a given region can be removed due to causes other than fires, which can cause incorrect classification of burned and unburned areas.

3.2. Results and Discussion: Second Experiment

The spatial classification of the fire class was obtained using the ANN algorithm, based on the dataset that presented the best performance and enabled the identification of the fire scars in 2018 to 2021. In addition, the types of crops (annual and temporary) were presented to assist in identifying and associating the areas with fire scars and the type of crops located near them. Figure 7 shows the land use and cover map obtained through the classification methodology.
Based on Figure 7, it was possible to observe that in the four years under study, the fire scars were located near agricultural areas, mainly nearby temporary crops, such as the representative case of soybeans. Furthermore, it was possible to observe an increase in fire scars between 2020 and 2021. According to [64], temporary crops consist mainly of soybeans, sugarcane, rice, and cotton.
According to [60], temporary crop plants undergo several stress factors related to environmental conditions, such as high temperature and lack of rain, resulting in a moisture deficit. The abovementioned crops experience a dehydration process in advanced phenological stages, since the moisture deficit increases throughout the plant’s growth, and it is common for them to be dry during the harvest period [70]. The phenological dynamics of temporary crops and the climatic constraints and characteristics of the Cerrado, such as lack of rainfall, can be considered factors contributing to fires [71]. In addition, in some cases, farmers use burning practices to save costs when cleaning the crops [72].
Accordingly, ref. [73] investigated the effect of soybean post-harvest residues and wind on fire occurrence and pointed out that there is less chance of fire spread in small grain crops since the harvest is close to the soil surface, leaving few residues accessible for burning to occur outside of each bed. Burning residues is legal; however, when performed incorrectly, it can cause environmental damage, as presented in the study by [74]. In addition, the authors associated the burning of waste with soil damage.
According to [75], regions with sugarcane and rice plantations pose environmental concerns since, in dry periods, the husks and straws of these crops are more susceptible to burning, which can ignite fires. In the study by [76], the authors found fire in regions where rice, corn, and cotton were cultivated, with a higher occurrence in the last two crops mentioned. These findings explain what we found in our results, showing that burned areas are more often associated with temporary crops.
According to [64], perennial crops mainly consist of coffee and citrus. As shown in Figure 7, the study area had a large coffee plantation site. Ref. [77] reinforce that precipitation and temperature variations can cause crop displacement to areas closer to forests. Arabica coffee, for example, is a crop that requires lower temperatures, varying around 18 °C to 21 °C, approximately [78]. In regions and seasons of strong wind, its leaf area can be reduced [79]. According to [80], when exposed to prolonged dry seasons, Arabica coffee can have its growth and development impaired. According to [61], the widespread production of coffee results in many residues, such as bark, dry leaf, and defective beans, which are combustible fire materials. Thus, [81] emphasize the importance of raising awareness about coffee cultivation in forest areas since this practice has the potential to cause deforestation and degrade forests.

3.3. Advances and Limitations

The proposed method for mapping burn scars, which integrates spectral indices from Sentinel-2 images with artificial neural networks (ANNs), stands out for its performance, evidenced by a global accuracy exceeding 93% and a kappa index above 0.92 across all analyzed periods. This low-cost approach, based on free satellite data and open-source algorithms, ensures its replicability and applicability in diverse regions, and overcomes the intrinsic limitations of methods that exclusively employ traditional spectral indices. The ANNs’ ability to discern fire scars even in complex agricultural areas or where there is confusion with bare soil and dark pixels is a key differentiator, optimizing classification accuracy and enabling effective large-scale monitoring with higher temporal frequency and lower operational cost compared to field surveys.
However, its application requires prior knowledge in digital image processing and neural network configuration, which can constitute an initial technical barrier. Although the combination of bands and spectral indices minimizes confusion with bare soil and temporary crops, there is still the possibility of incorrect classifications in areas with high land use heterogeneity or during sharp phenological transitions. Additionally, the method’s accuracy is subject to atmospheric conditions and the availability of cloud-free images.

4. Conclusions

Here, we could accurately observe the fixed presence of fire scars across the studied geographic region and time period. This reinforces the need to monitor the agricultural practices in the Cerrado, as burning occurs with undesirable frequency, posing several risks to the environment and human health.
This study also presented an accurate, low-cost and innovative method for the ANN classification of burned areas, with the best performances found when all bands of the Sentinel-2 satellite and the four indexes (NDVI, SAVI, NBR, and CSI) were used. The first experiment was necessary because it showed results that aimed to facilitate future research, such as using images only in the post-fire period. Furthermore, classifications were obtained by the abovementioned method with a global accuracy higher than 93.7% and a kappa index higher than 0.92.
Based on the attributes that presented the best performance in the ANN classification, land use and land cover maps were generated, making it possible to analyze the higher incidence of fires near temporary crop areas, such as soybeans, corn, and beans. Although burning is legalized, the agencies responsible and committed to the surveillance of this phenomenon must be attentive to monitoring and investigating the facts.
Based on the land use and land cover maps, it was possible to observe the fixed presence of fire scars in the four years we studied, highlighting the importance of monitoring these areas, because burning, when occurring frequently, can pose several risks to the environment and human health. The methodology used in this study can be integrated into a system for monitoring burned areas to help the responsible agencies monitor burning activities.

Author Contributions

Conceptualization, P.I.d.S.C.A. and J.V.M.B.; Methodology, P.I.d.S.C.A., G.D.M., G.H.d.A.P., R.B.d.A.G., J.L.S.B., C.A.M.d.A.J. and L.C.M.X.; Software, P.I.d.S.C.A. and G.D.M.; Validation, P.I.d.S.C.A., G.D.M., G.H.d.A.P., R.B.d.A.G., J.L.S.B., C.A.M.d.A.J. and L.C.M.X.; Formal analysis, P.I.d.S.C.A., G.D.M. and C.A.M.d.A.J.; Investigation, P.I.d.S.C.A.; Resources, P.I.d.S.C.A.; Data curation, P.I.d.S.C.A.; Writing—original draft, P.I.d.S.C.A. and J.V.M.B.; Writing—review & editing, P.I.d.S.C.A., G.D.M., G.H.d.A.P., R.B.d.A.G., J.L.S.B., C.A.M.d.A.J., L.C.M.X. and J.V.M.B.; Visualization, P.I.d.S.C.A. and J.V.M.B.; Supervision, J.V.M.B.; Project administration, P.I.d.S.C.A., G.D.M. and J.V.M.B. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by FAPEMIG grant number 13710.

Data Availability Statement

The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding authors.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Singh, B.; Venkatramanan, V.; Deshmukh, B. Monitoring of land use land cover dynamics and prediction of urban growth using Land Change Modeler in Delhi and its environs, India. Environ. Sci. Pollut. Res. 2022, 29, 33540–33555. [Google Scholar] [CrossRef]
  2. Bowman, D.M.; Kolden, C.A.; Abatzoglou, J.T.; Johnston, F.H.; van der Werf, G.R.; Flannigan, M. Vegetation fires in the Anthropocene. Nat. Rev. Earth Environ. 2020, 1, 500–515. [Google Scholar] [CrossRef]
  3. Jones, M.W.; Abatzoglou, J.T.; Veraverbeke, S.; Andela, N.; Lasslop, G.; Forkel, M.; Smith, A.J.P.; Burton, C.; Betts, R.A.; van der Werf, G.R. Global and regional trends and drivers of fire under climate change. Rev. Geophys. 2022, 60, e2020RG000726. [Google Scholar] [CrossRef]
  4. Erdős, L.; Török, P.; Veldman, J.W.; Bátori, Z.; Bede-Fazekas, Á.; Magnes, M.; Kröel-Dulay, G.; Tölgyesi, C. How climate, topography, soils, herbivores, and fire control forest–grassland coexistence in the Eurasian forest-steppe. Biol. Rev. 2022, 97, 2195–2208. [Google Scholar] [CrossRef]
  5. Keeley, J.E.; Syphard, A.D. Twenty-first century California, USA, wildfires: Fuel-dominated vs. wind-dominated fires. Fire Ecol. 2019, 15, 1–15. [Google Scholar] [CrossRef]
  6. Tošić, I.; Živanović, S.; Tošić, M. Influence of extreme climate conditions on the forest fire risk in the Timočka Krajina region (northeastern Serbia). Időjárás Q. J. Hung. Meteorol. Serv. 2020, 124, 331–347. [Google Scholar] [CrossRef]
  7. Yu, P.; Xu, R.; Abramson, M.J.; Li, S.; Guo, Y. Bushfires in Australia: A serious health emergency under climate change. Lancet Planet. Health 2021, 4, e7–e8. [Google Scholar] [CrossRef]
  8. Xu, R.; Lin, H.; Lu, K.; Cao, L.; Liu, Y. A forest fire detection system based on ensemble learning. Forests 2021, 12, 217. [Google Scholar] [CrossRef]
  9. Scott, A.C.; Bowman, D.M.J.S.; Bond, W.J.; Pyne, S.J.; Alexander, M.E. Fire on Earth: An Introduction; Wiley-Blackwell: Chichester, UK, 2014; 434p, ISBN 978-1-119-95357-9. Available online: https://www.researchgate.net/publication/299510875_Fire_on_Earth_An_Introduction (accessed on 24 May 2021).
  10. Mahmoud, M.A.; Ren, H. Forest fire detection using a rule-based image processing algorithm and temporal variation. Math. Probl. Eng. 2018, 2018. [Google Scholar] [CrossRef]
  11. Nolan, R.H.; Collins, L.; Leigh, A.; Ooi, M.K.J.; Curran, T.J.; Fairman, T.A.; Dios, V.R.; Bradstock, R. Limits to post-fire vegetation recovery under climate change. Plant Cell Environ. 2021, 44, 3471–3489. [Google Scholar] [CrossRef]
  12. Instituto Nacional de Pesquisas Espaciais (INPE). Queimadas. 2021. Available online: https://terrabrasilis.dpi.inpe.br/queimadas/bdqueimadas/ (accessed on 20 March 2021).
  13. Garcia, L.C.; Szabo, J.K.; de Oliveira Roque, F.; Pereira, A.D.M.M.; da Cunha, C.N.; Damasceno-Júnior, G.A.; Morato, R.G.; Tomas, W.M.; Libonati, R.; Ribeiro, D.B. Record-breaking wildfires in the world’s largest continuous tropical wetland: Integrative fire management is urgently needed for both biodiversity and humans. J. Environ. Manag. 2021, 293, 112870. [Google Scholar] [CrossRef] [PubMed]
  14. Pivello, V.R.; Vieira, I.; Christianini, A.V.; Ribeiro, D.B.; Menezes, L.S.; Berlinck, C.N.; Melo, P.P.L.; Marengo, J.A.; Tornquist, C.G.; Tomas, W.M.; et al. Understanding Brazil’s catastrophic fires: Causes, consequences and policy needed to prevent future tragedies. Perspect. Ecol. Conserv. 2021, 19, 233–255. [Google Scholar] [CrossRef]
  15. Xu, R.; Yu, P.; Abramson, M.J.; Johnston, F.H.; Samet, J.M.; Bell, M.L.; Haines, A.; Ebi, K.L.; Li, S.; Guo, Y. Wildfires, global climate change, and human health. N. Engl. J. Med. 2020, 383, 2173–2181. [Google Scholar] [CrossRef]
  16. Clarke, H.; Nolan, R.H.; De Dios, V.R.; Bradstock, R.; Griebel, A.; Khanal, S.; Boer, M.M. Forest fire threatens global carbon sinks and population centres under rising atmospheric water demand. Nat. Commun. 2022, 13, 7161. [Google Scholar] [CrossRef]
  17. Seddon, N.; Smith, A.; Smith, P.; Key, I.; Chausson, A.; Girardin, C.; House, J.; Srivastava, S.; Turner, B. Getting the message right on nature-based solutions to climate change. Glob. Change Biol. 2021, 27, 1518–1546. [Google Scholar] [CrossRef]
  18. Zittis, G.; Almazrouri, M.; Alpert, P.; Ciais, P.; Cramer, W.; Dahdal, Y.; Fnais, M.; Francis, D.; Hadjinicolaou, P.; Howari, F.; et al. Climate change and weather extremes in the Eastern Mediterranean and Middle East. Rev. Geophys. 2022, 60, e2021RG000762. [Google Scholar] [CrossRef]
  19. Colli, G.R.; Vieira, C.R.; Dianese, J.C. Biodiversity and conservation of the Cerrado: Recent advances and old challenges. Biodivers. Conserv. 2020, 29, 1465–1475. [Google Scholar] [CrossRef]
  20. Instituto Brasileiro de Geografia e Estatística (IBGE). Território. 2004. Available online: https://brasilemsintese.ibge.gov.br/territorio.html (accessed on 15 May 2021).
  21. Ministério do Meio Ambiente (MMA). O Bioma Cerrado. Available online: https://antigo.mma.gov.br/publicacoes/biomas/category/62-cerrado.html (accessed on 18 May 2021).
  22. Embrapa. Bioma Cerrado. 2022. Available online: https://www.embrapa.br/contando-ciencia/bioma-cerrado (accessed on 23 January 2022).
  23. Horák-Terra, I.; Cortizas, A.M.; Da Luz, C.F.P.; Silva, A.C.; Mighall, T.; De Camargo, P.B.; Mendonça-Filho, C.V.; Oliveira, P.E.; Cruz, F.W.; Vidal-Torrado, P. Late Quaternary vegetation and climate dynamics in central-eastern Brazil: Insights from a 35k cal a BP peat record in the Cerrado biome. J. Quat. Sci. 2020, 35, 664–676. [Google Scholar] [CrossRef]
  24. Gonçalves, R.V.S.; Cardoso, J.C.F.; Oliveira, P.E.; Raymundo, D.; De Oliveira, D.C. The role of topography, climate, soil, and the surrounding matrix in the distribution of Veredas wetlands in central Brazil. Wetl. Ecol. Manag. 2022, 30, 1261–1279. [Google Scholar] [CrossRef]
  25. Reis, J.C.; Rodrigues, G.S.; Barros, I.; Rodrigues, R.A.R.; Garret, R.D.; Valentim, J.F.; Kamoi, M.Y.T.; Michetti, M.; Wruck, F.J.; Rodrigues-Filho, S.; et al. Integrated crop-livestock systems: A sustainable land-use alternative for food production in the Brazilian Cerrado and Amazon. J. Clean. Prod. 2021, 283, 124580. [Google Scholar] [CrossRef]
  26. Ratter, J.A.R.; Ribeiro, J.F.; Bridgewater, S. The Brazilian cerrado vegetation and threats to its biodiversity. Ann. Bot. 1997, 80, 223–230. [Google Scholar] [CrossRef]
  27. Segura-Garcia, C.; Bauman, D.; Arruda, V.L.S.; Alencar, A.A.C.; Oliveras Menor, I. Human land occupation regulates the effect of the climate on the burned area of the Brazilian Cerrado. Commun. Earth Environ. 2024, 5, 361. [Google Scholar] [CrossRef]
  28. Hua, L.; Shao, G. The progress of operational forest fire monitoring with infrared remote sensing. J. For. Res. 2017, 28, 215–229. [Google Scholar] [CrossRef]
  29. Goyal, S.; Shagill, M.; Kaur, A.; Vohra, H.; Singh, A. A yolo based technique for early forest fire detection. Int. J. Innov. Technol. Explor. Eng. 2020, 9, 1357–1362. [Google Scholar] [CrossRef]
  30. Barmpoutis, P.; Papaioannou, P.; Dimitropoulos, K.; Grammalidis, N. A review on early forest fire detection systems using optical remote sensing. Sensors 2020, 20, 6442. [Google Scholar] [CrossRef] [PubMed]
  31. Tien Bui, D.; Le, K.T.T.; Nguyen, V.C.; Le, H.D.; Revhaug, I. Tropical forest fire susceptibility mapping at the Cat Ba National Park Area, Hai Phong City, Vietnam, using GIS-based kernel logistic regression. Remote Sens. 2016, 8, 347. [Google Scholar] [CrossRef]
  32. Kharuk, V.I.; Dvinskaya, M.L.; Petrov, I.A.; Im, S.T.; Ranson, K.J. Larch forests of Middle Siberia: Long-term trends in fire return intervals. Reg. Environ. Chang. 2016, 16, 2389–2397. [Google Scholar] [CrossRef]
  33. Lim, C.H.; Kim, Y.S.; Won, M.; Kim, S.J.; Lee, W.K. Can satellite-based data substitute for surveyed data to predict the spatial probability of forest fire? A geostatistical approach to forest fire in the Republic of Korea. Geomat. Nat. Hazards Risk 2019, 10, 719–739. [Google Scholar] [CrossRef]
  34. Martins, M.A.; Lima, A.M.N.; Carvalho, G.A.L.; Costa, M.A. A review of indoor positioning systems: Methods, approaches, and technologies. J. Netw. Comput. Appl. 2020, 166, 102706. Available online: https://www.sciencedirect.com/science/article/abs/pii/S0379711221000588 (accessed on 11 December 2022).
  35. Morlin Carneiro, F.; Angeli Furlani, C.E.; Zerbato, C.; Candida de Menezes, P.; da Silva Gírio, L.A.; Freire de Oliveira, M. Comparison between vegetation indexes for detecting spatial and temporal variabilities in soybean crop using canopy sensors. Precis. Agric. 2020, 21, 979–1007. [Google Scholar] [CrossRef]
  36. Bijeesh, T.V.; Narasimhamurthy, K.N. Surface water detection and delineation using remote sensing images: A review of methods and algorithms. Sustain. Water Resour. Manag. 2020, 6, 68. [Google Scholar] [CrossRef]
  37. Chen, Y.; Lara, M.J.; Hu, F.S. A robust visible near-infrared index for fire severity mapping in Arctic tundra ecosystems. ISPRS J. Photogramm. Remote Sens. 2020, 159, 101–113. [Google Scholar] [CrossRef]
  38. Key, C.H.; Benson, N.C. The Normalized Burn Ratio (NBR): A Landsat TM Radiometric Measure of Burn Severity; U.S. Department of the Interior, Northern Rocky Mountain Science Centre: Bozeman, MT, USA, 1999. [Google Scholar]
  39. Rouse, J.W.; Haas, R.H.; Schell, J.A.; Deering, D.W. Monitoring vegetation systems in the Great Plains with ERTS. NASA 1973, 1, 309–317. [Google Scholar]
  40. Huete, A.R. A soil-adjusted vegetation index (SAVI). Remote Sens. Environ. 1988, 25, 295–309. [Google Scholar] [CrossRef]
  41. Liu, J.; Maeda, E.E.; Wang, D.; Heiskanen, J. Sensitivity of Spectral Indices on Burned Area Detection using Landsat Time Series in Savannas of Southern Burkina Faso. Remote Sens. 2021, 13, 2492. [Google Scholar] [CrossRef]
  42. Spano, D.; Cesaraccio, C.; Duce, P.; Snyder, R.L. Phenological stages of natural species and their use as climate indicators. Int. J. Biometeorol. 1999, 42, 124–133. [Google Scholar] [CrossRef]
  43. Souza, C.H.W.; Teodoro, A.C.; Marques, J.E.; Formaggio, A.R. Recent applications of Landsat 8/OLI and Sentinel-2/MSI for land use and land cover mapping: A systematic review. Remote Sens. 2020, 12, 3062. [Google Scholar] [CrossRef]
  44. Zeng, L.; Wardlow, B.D.; Xiang, D.; Hu, S.; Li, D. A review of vegetation phenological metrics extraction using time-series, multispectral satellite data. Remote Sens. Environ. 2020, 237, 111511. [Google Scholar] [CrossRef]
  45. Santos, S.M.B.; Duverger, S.G.; Bento-Gonçalves, A.; Franca-Rocha, W.; Vieira, A.; Teixeira, G. Remote Sensing Applications for Mapping Large Wildfires Based on Machine Learning and Time Series in Northwestern Portugal. Fire 2023, 6, 43. [Google Scholar] [CrossRef]
  46. Brand, A.K.; Manandhar, A. Semantic segmentation of burned areas in satellite images using a U-net-based convolutional neural network. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2021, 43, 47–53. [Google Scholar] [CrossRef]
  47. Filipponi, F. Exploitation of Sentinel-2 Time Series to Map Burned Areas at the National Level: A Case Study on the 2017 Italy Wildfires. Remote Sens. 2019, 11, 622. [Google Scholar] [CrossRef]
  48. Maxwell, A.E.; Warner, T.A.; Fang, F. Implementation of machine-learning classification in remote sensing: An applied review. Int. J. Remote Sens. 2018, 39, 2784–2817. [Google Scholar] [CrossRef]
  49. Mitchell, T.; Buchanan, B.; Dejong, G.; Dietterich, T.; Rosenbloom, P.; Waibel, A. Machine learning. Annu. Rev. Comput. Sci. 1990, 4, 417–433. [Google Scholar] [CrossRef]
  50. Zhang, K.; Li, P.; Wang, J. A Review of Deep Learning-Based Remote Sensing Image Caption: Methods, Models, Comparisons and Future Directions. Remote Sens. Environ. 2024, 16, 4113. [Google Scholar] [CrossRef]
  51. Oliveira, P.S.; Matos, V.A. Perfil demográfico da microrregião Patrocínio-MG. Rev. Bras. De Gestão E Desenvolv. Reg. 2009, 5. Available online: https://www.rbgdr.net/revista/index.php/rbgdr/article/view/244 (accessed on 22 January 2022).
  52. Santana, C.A.M.; Campos, S.K.; Marra, R.; Aragão, A.A.; Cerrado: Pilar da Agricultura Brasileira. Área de Informação da Sede-Capítulo em Livro Científico (ALICE). Available online: https://www.alice.cnptia.embrapa.br/handle/doc/1121866 (accessed on 25 June 2021).
  53. Franco, C.; Rodrigues, G.S.S.C. Análise das principais condicionantes para inserção do eucalipto no Triângulo Mineiro e Alto Paranaíba durante a década de 1970. In ENANPEGE, A Geografia Brasileira Na Ciência-Mundo: Produção, Circulação E Apropriação Do Conhecimento; São Paulo, Brazil, 2019; p. 159. Available online: https://pt.scribd.com/document/437436768/Xiii-Enanpege-2019-caderno-de-Programacao-atualizado (accessed on 6 April 2022).
  54. Campos, S.A.C.; Coelho, A.B.; Gomes, A.P. Influência das condições ambientais e ação antrópica sobre a eficiência produtiva agropecuária em Minas Gerais. Rev. De Econ. E Sociol. Rural. 2012, 50, 563–576. [Google Scholar] [CrossRef]
  55. Oliveira, M.V.M.; Teixeira, M.M.; Fernandes, H.C.; Silva, A.C.; Furtado Júnior, M.R. Declividade máxima permitida para uma colhedora de café trabalhando em regiões montanhosas. In Anais Do VIII Simpósio De Pesquisa Dos Cafés Do Brasil; Salvador, Brazil, 2013; Available online: http://www.sapc.embrapa.br/arquivos/consorcio/spcb_anais/simposio8/291.pdf (accessed on 18 May 2021).
  56. Gao, Y.; Skutsch, M.; Paneque-Gálvez, J.; Ghilardi, A. Remote sensing of forest degradation: A review. Environ. Res. Lett. 2020, 15, 103001. [Google Scholar] [CrossRef]
  57. Congedo, L. Semi-automatic classification plugin documentation. Release 2016, 4, 29. Available online: https://semiautomaticclassificationmanual.readthedocs.io/en/latest/ (accessed on 28 April 2022).
  58. Chavez, P.S., Jr. An improved dark-object subtraction technique for atmospheric scattering correction of multispectral data. Remote Sens. Environ. 1988, 24, 459–479. [Google Scholar] [CrossRef]
  59. Parker, J.A.; Kenyon, R.V.; Troxel, D.E. Comparison of interpolating methods for image resampling. IEEE Trans. Med. Imaging 1983, 2, 31–39. [Google Scholar] [CrossRef]
  60. Mapbiomas. Mapas e Dados. Available online: https://mapbiomas.org/ (accessed on 11 December 2022).
  61. Martinez, C.L.M.; Saari, J.; Melo, Y.; Cardoso, M.; de Almeida, G.M.; Vakkilainen, E. Evaluation of thermochemical routes for the valorization of solid coffee residues to produce biofuels: A Brazilian case. Renew. Sustain. Energy Rev. 2021, 137, 110585. [Google Scholar] [CrossRef]
  62. Landis, J.R.; Koch, G.G. The measurement of observer agreement for categorical data. Biometrics 1977, 33, 159–174. Available online: https://2024.sci-hub.st/2153/7a79ed8d1ae20779cbcdf5ab36348bc2/landis1977.pdf (accessed on 18 March 2022). [CrossRef] [PubMed]
  63. Souza, C.M.; Shimbo, J.Z.; Rosa, M.R.; Parente, L.L.; Alencar, A.A.; Rudorff, B.F.T.; Hasenack, H.; Matsumoto, M.; Ferreira, L.G.; Souza-Filho, P.W.M.; et al. Reconstructing three decades of land use and land cover changes in brazilian biomes with landsat archive and earth engine. Remote Sens. 2020, 12, 2735. [Google Scholar] [CrossRef]
  64. Seydi, S.T.; Akhoondzadeh, M.; Amani, M.; Mahdavi, S. Wildfire damage assessment over Australia using sentinel-2 imagery and MODIS land cover product within the google earth engine cloud platform. Remote Sens. 2021, 13, 220. [Google Scholar] [CrossRef]
  65. Syifa, M.; Panahi, M.; Lee, C. Mapping of post-wildfire burned area using a hybrid algorithm and satellite data: The case of the camp fire wildfire in California, USA. Remote Sens. 2020, 12, 623. [Google Scholar] [CrossRef]
  66. Xie, Q.; Dash, J.; Huete, A.; Jiang, A.; Yin, G.; Ding, Y.; Peng, D.; Hall, C.C.; Brown, L.; Shi, Y.; et al. Retrieval of crop biophysical parameters from Sentinel-2 remote sensing imagery. Int. J. Appl. Earth Obs. Geoinf. 2019, 80, 187–195. [Google Scholar] [CrossRef]
  67. Hecht-Nielsen, R. Theory of the backpropagation neural network. In Neural Networks for Perception; Academic Press: Cambridge, MA, USA, 1992; pp. 65–93. Available online: https://sci-hub.se/10.1016/b978-0-12-741252-8.50010-8 (accessed on 30 October 2021).
  68. Paola, J.D.; Schowengerdt, R.A. A detailed comparison of backpropagation neural network and maximum-likelihood classifiers for urban land use classification. IEEE Trans. Geosci. Remote Sens. 1995, 33, 981–996. [Google Scholar] [CrossRef]
  69. Engelbrecht, J. A simple normalized difference approach for burnt area mapping using multipolarized C-band SAR. Remote Sens. 2017, 9, 764. [Google Scholar] [CrossRef]
  70. Patanè, C.; Corinzia, S.A.; Testa, G.; Scordia, D.; Consentino, S.L. Physiological and Agronomic Responses of Processing Tomatoes to Deficit Irrigation at Critical Stages in a Semi-Arid Environment. Agronomy 2020, 10, 800. [Google Scholar] [CrossRef]
  71. Kiniry, J.R. Maize phasic development. Model. Plant Soil Syst. 1991, 31, 55–70. [Google Scholar]
  72. Jakimow, B.; Griffiths, P.; van der Linden, S.; Hostert, P. Mapping pasture management in the Brazilian Amazon from dense Landsat time series. Remote Sens. Environ. 2018, 205, 453–468. [Google Scholar] [CrossRef]
  73. Norsworthy, J.K.; Green, J.K.; Barber, T.; Roberts, T.L.; Walsh, M.J. Seed destruction of weeds in southern US crops using heat and narrow-windrow burning. Weed Technol. 2020, 34, 589–596. [Google Scholar] [CrossRef]
  74. Amorim, H.C.; Ashworth, A.J.; Brye, K.R.; Wienhold, B.J.; Savin, M.C.; Owens, P.R.; Silva, S.H. Soil quality indexes as affected by long-term burning, irrigation, tillage, and fertility management. Soil Sci. Soc. Am. J. 2021, 85, 379–395. [Google Scholar] [CrossRef]
  75. Phairuang, W.; Suwattiga, P.; Chetiyanukornkul, T.; Hongtieab, S.; Limpaseni, W.; Ikemori, F.; Hata, M.; Furuuchi, M. The influence of the open burning of agricultural biomass and forest fires in Thailand on the carbonaceous components in size-fractionated particles. Environ. Pollut. 2019, 247, 238–247. [Google Scholar] [CrossRef]
  76. Shen, Y.; Jiang, C.; Chan, K.L.; Hu, C.; Yao, L. Estimation of field-level NOx emissions from crop residue burning using remote sensing data: A case study in Hubei, China. Remote Sens. 2021, 13, 404. [Google Scholar] [CrossRef]
  77. Johnson, M.A.; Ruiz-Diaz, C.P.; Manoukis, N.C.; Verle Rodrigues, J.C. Coffee Berry Borer (Hypothenemus hampei), a Global Pest of Coffee: Perspectives from Historical and Recent Invasions, and Future Priorities. Insects 2020, 11, 882. [Google Scholar] [CrossRef]
  78. Alègre, C. Climates et caféiers d’Arabie. Agron. Trop 1959, 14, 23–58. [Google Scholar]
  79. Caramori, L.P.C.; Caramori, P.H.; Manetti Filho, J. Effect of leaf Water potential on cold tolerance of Coffea arabica L. Braz. Arch. Biol. Technol. 2002, 45, 439–443. [Google Scholar] [CrossRef]
  80. Melke, A.; Fetene, M. Eco-physiological basis of drought stress in coffee (Coffea arabica, L.) in Ethiopia. Theor. Exp. Plant Physiol. 2014, 26, 225–239. [Google Scholar] [CrossRef]
  81. Arai, Y.; Hundera, K.; Yoshikura, T. Challenges in conserving forest ecosystems through coffee certification: A case study from southwestern Ethiopia. Front. Environ. Sci. 2023, 11, 1193242. [Google Scholar] [CrossRef]
Fire 08 00320 g001
Figure 1. The geographic location of the Patrocínio microregion: (a) Brazil, (b) Minas Gerais, and (c) Patrocínio microregion.
Figure 1. The geographic location of the Patrocínio microregion: (a) Brazil, (b) Minas Gerais, and (c) Patrocínio microregion.
Fire 08 00320 g001
Fire 08 00320 g002
Figure 2. Rainfall over the years in the study area.
Figure 2. Rainfall over the years in the study area.
Fire 08 00320 g002
Fire 08 00320 g003
Figure 3. Fire spot distribution over the study area.
Figure 3. Fire spot distribution over the study area.
Fire 08 00320 g003
Fire 08 00320 g004
Figure 4. Fire spots in (a) 2018, (b) 2019, (c) 2020, and (d) 2021.
Figure 4. Fire spots in (a) 2018, (b) 2019, (c) 2020, and (d) 2021.
Fire 08 00320 g004
Fire 08 00320 g005
Figure 5. Heatmap by municipalities in 2018, 2019, 2020, and 2021.
Figure 5. Heatmap by municipalities in 2018, 2019, 2020, and 2021.
Fire 08 00320 g005
Fire 08 00320 g006
Figure 6. The geographical arrangement of the training and validation samples from the four years under study.
Figure 6. The geographical arrangement of the training and validation samples from the four years under study.
Fire 08 00320 g006
Fire 08 00320 g007
Figure 7. Map of fire scars and types of crops in the Patrocínio microregion in 2018, 2019, 2020, and 2021.
Figure 7. Map of fire scars and types of crops in the Patrocínio microregion in 2018, 2019, 2020, and 2021.
Fire 08 00320 g007
Table 1. Date of the images for the 2018, 2019, 2020, and 2021 dry spells.
Table 1. Date of the images for the 2018, 2019, 2020, and 2021 dry spells.
Year/Period2018201920202021
Pre-fire05/0504/2005/2904/24
Post-fire09/2710/1710/0109/06
Table 2. Spectral indexes used for the identification of fire scars.
Table 2. Spectral indexes used for the identification of fire scars.
Spectral IndexesAcronymFormulaReference
Normalized Burn RatioNBR(NIR − SWIR2)/(NIR + SWIR2)[60]
Normalized Difference Vegetation IndexNDVI(NIR − R)/(NIR + R)[61]
Soil Adjusted Vegetation IndexSAVI(1 + L) * [(NIR − R)/(NIR + R + L)], where L = 1[34]
Char Soil IndexCSINIR/SWIR2[48]
Table 3. Kappa coefficient values for the classification analysis.
Table 3. Kappa coefficient values for the classification analysis.
Kappa CoefficientQuality of Rating
<0.00Insignificant
0.00 to 0.20Weak
0.21 to 0.40Fair
0.41 to 0.60Moderate
0.61 to 0.80Strong
0.81 to 1.00Near Perfect
Adapted from [62].
Table 4. Overall accuracy and kappa index for the tested configurations.
Table 4. Overall accuracy and kappa index for the tested configurations.
ValidationConfigurationVariation Between Pre-Fire and Post-Fire PeriodPost-Fire
20182019202020212018201920202021
KappaAll bands + index0.9740.8680.9490.9810.9250.9390.9980.968
NBR only0.5750.5270.6220.4680.6330.5810.7060.502
CSI only0.3960.4680.6100.3680.4970.4890.4370.255
Global Accuracy (%)All bands + index97.87689.00795.77398.43293.73094.94099.84197.400
NBR only64.63560.65468.51155.71469.48465.11975.57558.492
CSI only49.71255.73467.55947.63158.09557.48053.11537.936
Table 5. Confusion matrix of the classification by neural networks of all Sentinel-2 bands and the NDVI, SAVI, NBR and CSI indexes in the 2018 post-fire period.
Table 5. Confusion matrix of the classification by neural networks of all Sentinel-2 bands and the NDVI, SAVI, NBR and CSI indexes in the 2018 post-fire period.
ClassFire ScarsExposed SoilNative VegetationReforestationWaterCropsTotalOmission (%)
fire scars836040008400.476
exposed soil084000008400.000
native vegetation008400008400.000
reforestation7403763008409.167
water000084008400.000
crops231040060584027.976
Total11418408517638406055040
Inclusion (%)26.7310.0001.2930.0000.0000.000
Global Accuracy = 93.7302%; Kappa = 0.925
Table 6. Confusion matrix of the classification by neural networks of all Sentinel-2 bands and the NDVI, SAVI, NBR, and CSI indexes in the 2019 post-fire period.
Table 6. Confusion matrix of the classification by neural networks of all Sentinel-2 bands and the NDVI, SAVI, NBR, and CSI indexes in the 2019 post-fire period.
ClassFire ScarsExposed SoilNative VegetationReforestationWaterCropsTotalOmission (%)
fire scars836000408400.476
exposed soil0824000168401.905
native vegetation408342008400.714
reforestation820267275084013.452
water000084008400.000
crops010808072484013.810
Total9229328607378497405040
Inclusion (%)9.32811.5883.0231.3571.0602.162
Global Accuracy = 94.940%; Kappa = 0.939
Table 7. Confusion matrix of the classification by neural networks of all Sentinel-2 bands and the NDVI, SAVI, NBR, and CSI indexes in the post-fire 2020 period.
Table 7. Confusion matrix of the classification by neural networks of all Sentinel-2 bands and the NDVI, SAVI, NBR, and CSI indexes in the post-fire 2020 period.
ClassFire ScarsExposed SoilNative VegetationReforestationWaterCropsTotalOmission (%)
fire scars840000008400.000
exposed soil084000008400.000
native vegetation008391008400.119
reforestation000840008400.000
water000084008400.000
crops007008338400.833
Total8408408468418408335040
Inclusion (%)0.0000.0000.8270.1190.0000.000
Global Accuracy = 99.841%; Kappa = 0.998
Table 8. Confusion matrix of the classification by neural networks of all Sentinel-2 bands and the NDVI, SAVI, NBR, and CSI indexes in the post-fire period of 2021.
Table 8. Confusion matrix of the classification by neural networks of all Sentinel-2 bands and the NDVI, SAVI, NBR, and CSI indexes in the post-fire period of 2021.
ClassFire ScarsExposed SoilNative VegetationReforestationWaterCropsTotalOmission (%)
fire scars840000008400.000
exposed soil084000008400.000
native vegetation007091310084015.595
reforestation000840008400.000
water000084008400.000
crops000008408400.000
Total8408407099718408405040
Inclusion (%)0.0000.0000.00013.4910.0000.000
Global Accuracy = 97.400%; Kappa = 0.968
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Castro Abreu, P.I.d.S.; Martins, G.D.; Pereira, G.H.d.A.; Gallis, R.B.d.A.; Brito, J.L.S.; de Abreu Júnior, C.A.M.; Xavier, L.C.M.; Bravo, J.V.M. A Method for Mapping and Associating Burned Areas with Agricultural Practices Within the Brazilian Cerrado. Fire 2025, 8, 320. https://doi.org/10.3390/fire8080320

AMA Style

Castro Abreu PIdS, Martins GD, Pereira GHdA, Gallis RBdA, Brito JLS, de Abreu Júnior CAM, Xavier LCM, Bravo JVM. A Method for Mapping and Associating Burned Areas with Agricultural Practices Within the Brazilian Cerrado. Fire. 2025; 8(8):320. https://doi.org/10.3390/fire8080320

Chicago/Turabian Style

Castro Abreu, Pâmela Inês de Souza, George Deroco Martins, Gabriel Henrique de Almeida Pereira, Rodrigo Bezerra de Araujo Gallis, Jorge Luis Silva Brito, Carlos Alberto Matias de Abreu Júnior, Laura Cristina Moura Xavier, and João Vitor Meza Bravo. 2025. "A Method for Mapping and Associating Burned Areas with Agricultural Practices Within the Brazilian Cerrado" Fire 8, no. 8: 320. https://doi.org/10.3390/fire8080320

APA Style

Castro Abreu, P. I. d. S., Martins, G. D., Pereira, G. H. d. A., Gallis, R. B. d. A., Brito, J. L. S., de Abreu Júnior, C. A. M., Xavier, L. C. M., & Bravo, J. V. M. (2025). A Method for Mapping and Associating Burned Areas with Agricultural Practices Within the Brazilian Cerrado. Fire, 8(8), 320. https://doi.org/10.3390/fire8080320

Article Metrics

Back to TopTop