Next Article in Journal
Cognitive FDA-MIMO Radar Network’s Transmit Element Selection Algorithm for Target Tracking in a Complex Interference Scenario
Previous Article in Journal
High-Precision Sub-Wavelength Motion Compensation Technique for 3D Down-Looking Imaging Sonar Based on an Acoustic Calibration System
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 17
Issue 1
10.3390/rs17010051
remotesensing-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

The Detection of Small-Scale Open-Burning Agriculture Fires Through Remote Sensing

by
Eduardo R. Oliveira
1,2,*,
Bárbara T. Silva
2,
Diogo Lopes
2,
Sofia Corticeiro
2,
Fátima L. Alves
2,
Leonardo Disperati
1 and
Carla Gama
2
1
Department of Physical Sciences, Earth and Environment, University of Siena, 53100 Siena, Italy
2
Department of Environment and Planning, CESAM-Centre for Environmental and Marine Studies, University of Aveiro, 3810-193 Aveiro, Portugal
*
Author to whom correspondence should be addressed.
Remote Sens. 2025, 17(1), 51; https://doi.org/10.3390/rs17010051
Submission received: 4 November 2024 / Revised: 13 December 2024 / Accepted: 23 December 2024 / Published: 27 December 2024
Browse Figures
Versions Notes

Abstract

:
The open burning of agricultural residues is a widespread practice with significant environmental implications. This study explores the potential of satellite remote sensing to detect and analyze small-scale agricultural fires in Portugal, focusing on their spatial and temporal characteristics. Using active fire detection products from various satellite platforms, including VIIRS, MODIS, SLSTR, and SEVIRI, we conducted a detailed analysis across two local case studies and a national-scale assessment. This study evaluates both active fire detections and post-fire burned area estimations, using high-resolution satellite imagery to overcome the limitations associated with the small size and low intensity of these fires. The results indicate that while active fire detections are feasible for larger-scale burning, challenges remain for smaller fires due to resolution constraints. A systematic comparison with an agricultural burning request database further highlights the need for the enhancement of temporal and spatial precision in data to improve detection reliability. Despite these limitations, this work underscores the importance of remote sensing tools in monitoring agricultural burning practices and enhancing environmental management efforts.

1. Introduction

The open burning of agricultural residues is a common practice worldwide, whether for eliminating agricultural residues or to promote land clearing [1]. In Portugal, this practice is particularly prevalent during the wet season, which usually runs from October to April. Typically, these fires involve the pile burning of agricultural waste and shrubbery (e.g., pruning and cuttings), or the deliberate burning of vegetation to promote the creation of pastures for livestock, control invasive species, and reduce wildfire hazard [2,3,4,5].
As single events, these burning practices are characterized by low intensities and small extents. However, the generalized practice of the open burning of agricultural residues can have several negative environmental impacts. Burning biomass releases greenhouse gasses [6], contributing to climate change, and particulate matter and gaseous atmospheric pollutants [7,8,9], impacting air quality and posing health risks to nearby communities. The cumulative effects on ecosystems are also significant, as frequent burning disrupts soil structure, depletes essential nutrients, and can hinder natural vegetation growth [10]. Over time, these disturbances may reduce soil fertility, alter biodiversity, and weaken ecosystem resilience, potentially transforming land into degraded areas susceptible to erosion [11] and invasive species [12]. Furthermore, in certain conditions, open burning may inadvertently escalate into uncontrolled wildfires, intensifying the environmental impact and heightening risks to public safety.
While open-burning is typically an unregulated practice, often carried out by non-experts, prescribed fire differs greatly in many aspects, particularly in terms of safety, purpose, and environmental impacts [13]. Prescribed fires are carefully controlled burns conducted by trained professionals to achieve specific management objectives like reducing wildfire fuel loads, enhancing wildlife habitats, and supporting native vegetation [14,15]. Prescribed fires are subject to legal regulations and follow a detailed plan that includes site mapping, weather monitoring, and safety protocols, ensuring that they benefit ecosystems while minimizing risks and impacts [14].
Satellite remote sensing (RS) is widely used to study fire phenomena. This approach typically includes active fire detection through the detection of thermal anomalies. The detection of fire anomalies is usually carried through high-temporal-resolution satellites, equipped with optical multispectral sensors, which often operate in the long-wavelength infrared range (e.g., MODIS, VIIRS, Sentinel-3) [16,17,18,19,20,21]. While higher-spatial-resolution satellites (e.g., Landsat series, Sentinel-2) have been explored for active fire detection, their relatively long revisit times (spanning several days) are insufficient to reliably capture such ephemeral occurrences [18,20,22,23,24]. Instead, higher-spatial-resolution satellites are better suited for post-fire analysis, such as the determination of burned extents or severity. These applications are capable of depicting the comparatively more persistent effects of fires and are not constrained by high-temporal-resolution requirements [25,26,27,28].
Considering the environmental impacts of the open burning of agricultural residues and the relatively unknown locations and timings of these events, remote sensing holds significant potential for providing valuable insights into these practices. However, the relatively low intensity, small spatial extent, and short duration of these fires—combined with the fact that they often occur during cloudy conditions or in the presence of haze from fire smoke itself, severely hindering the operability of multispectral optical sensors—pose significant challenges for satellite-based RS analysis [20,21].
This paper aims to determine whether the small-scale open burning of agricultural residues can be detected using RS. Moreover, it has the objective of identifying the best methods and products for systematic analysis and validation. To achieve this, a series of case studies will be analyzed, involving the testing of various satellite products and RS techniques. Additionally, the paper includes a comparison of RS products and methods against a comprehensive dataset of open-burning requests, serving as a ‘best possible’ validation reference. The findings of this study can contribute to improving environmental monitoring programs and informing agricultural policies aimed at reducing the impacts of open burning. By providing reliable tools for detecting and validating such activities, this research may also support efforts to mitigate air pollution and enhance sustainable land management practices.

2. Material and Methods

In order to verify the potential of the analysis of the small-scale open burning of agricultural residues through RS, this study focuses on several cases studies with varying spatial extents: two local-scale analyses, which were validated on the field at the ground level, and a systematic national-scale analysis.

2.1. Case Studies

The two local-scale analyses were conducted in the northern Portuguese mainland (Figure 1) and corresponded to known events involving extensive controlled biomass burning (Figure 1-Terras de Bouro and Póvoa do Varzim). One of these events consisted of a prescribed fire that despite the differences from agricultural open burnings—particularly regarding potential impacts on soil and vegetation—was included to assess the capability of RS to monitor small-scale fires. The national-level analysis consisted of a systematic analysis of mainland Portugal between 2019 and 2023 (Figure 1-Portugal Mainland).

2.1.1. Terras De Bouro

The first local case study corresponds to several agriculture extensive burnings that took place nearby the Peneda-Gerês National Park in Terras de Bouro (Figure 1-Terras de Bouro). These occurrences are a common practice in this region, particularly for creating pastures [29]. Some of these events were observed through on-site observations conducted by the authors of this study on 16th February 2023 and likely led to significant local air quality degradation. While rural areas like this lack direct air quality monitoring data, the existing literature on small-scale experimental field fires provides valuable insights into the potential air quality impacts of such activities, helping to substantiate this assessment [30,31]. For example, [30] measured dangerously high average particulate matter concentrations (>1000 µg·m−3) near the source during a shrubland burn (<1 h). Similarly, [31] demonstrated that firefighters are exposed to air pollution levels exceeding the Occupational Exposure Standard limits, emphasizing the urgent need for the adoption of preventive measures to mitigate these exposure risks. The cumulative evidence from these studies suggests that even controlled burns can lead to significant air quality degradation, reinforcing concerns about the local impact of small-scale open-burning agriculture practices. Although prescribed burns in Portugal require permits, the regulations focus primarily on ambient conditions such as wind speed, relative humidity, and temperature to minimize wildfire risk, rather than specifically addressing air quality impacts, as is common practice in the United States of America (USA) [32].
According to the official yearly database of burned areas by the Institute for Nature Conservation and Forests (ICNF) [33], this region also experienced several other events from 31 January to 24 February 2023. Among those, 2 occurrences corresponded to intentional fires for hunting and wildlife management, 2 involved the negligent pile burning of forest or agricultural residues, 3 involved the negligent burning of forest or agricultural residues, 24 corresponded to negligent burning for pasture management, and 5 were arson fires (see Section 3.1).

2.1.2. Póvoa De Varzim

The second local case study corresponds to two consecutive controlled prescribed fires that occurred in a recent clearcut of a Eucalyptus plantation, which took place in Balazar, Póvoa de Varzim, from 2 to 4 March 2023 (Figure 1-Póvoa de Varzim). This action was undertaken by the local Forest Association Portucalea in collaboration with forestry sappers from Vila do Conde and the volunteer firefighters of Póvoa de Varzim. This was first performed on March 2nd as a very controlled and low-intensity burning along a thin strip of about 30 × 100 m (Figure 2), and on 4 March, another larger section within the same clearcutting plot was burned (see Section 3.2). The fire was conducted downslope, against the wind, to ensure a very low fire danger class. As with all prescribed fires, this case study is not accounted for by the ICNF yearly burned areas database.

2.1.3. Portugal Mainland

Besides the local case studies, this study also includes a national-level analysis of agriculture burning requests for mainland Portugal (Figure 1–Portugal Mainland). For this analysis, we considered a national-level database of agricultural burning requests (ABR) for mainland Portugal, provided by ICNF. This database results from the current legislation (DL 82/2021, from 13 October), which states that agricultural residue burning is allowed between November 1st and May 31st on days when the fire danger is not classified as being among the highest risk levels, with the only requirement of notifying ICNF via an app or directly through local municipalities. The dataset includes 4,099,487 requests, covering the period from 4 December 2019 to 30 November 2023, and contains information such as location (parish, municipality, district), geographical coordinates and, in some cases, additional data like the type of burning, the motive, the burned area, and the authorization status.

2.2. Fire Detection Methods

2.2.1. Active Fire Detections

The case studies were analyzed in terms of active fire detections, initially by comparing the results with the observations of Fire Radiative Power products (FRPs) obtained from different satellite platforms. The FRP represents the rate of outgoing thermal radiative energy resulting from a burning fire, which is usually expressed in megawatts (MW) [34]. Table 1 summarizes the FRP-related products from the Visible Infrared Imaging Radiometer Suite (VIIRS), the Moderate-Resolution Imaging Spectroradiometer (MODIS), the Sea and Land Surface Temperature Radiometer (SLSTR), and the Spinning-Enhanced Visible Infra-Red Imager (SEVIRI), which were considered in this study.
Despite the high temporal resolution of the considered FRPs, one of the key limitations in terms of detecting small fires is their limited spatial resolution. For this reason, in our study, we also considered the Landsat 8 and 9 platforms, which together have a revisit time of 8 days. In both cases, these satellites are currently active and equipped with Thermal Infrared Sensors (respectively, TIRS and TIRS-2), which offer 100 m spatial resolutions. This represents a significant improvement with respect to the FRPs shown in Table 1. However, each of these satellites is also equipped with an Operational Land Imager (respectively, OLI and OLI-2), which simultaneously images several multispectral bands, operating from the visible to shortwave infrared range, at a 30 m resolution [38,39]. The Landsat Active Fire and Thermal Anomalies product, produced by the joint NASA/USGS OLI sensors using Landsat 8 and 9, is based on the algorithm developed by [18] and is currently available for the continental USA, southern Canada, and northern Mexico [40]. Since our case studies (Section 2) fall outside these regions, we directly utilized the [18] algorithm to detect active fires in these locations at a 30 m resolution. This algorithm has been found to effectively detect fires in both daytime and nighttime conditions, including small-scale events such as biomass burning, gas flares, and active volcanoes. Despite recognizing minimal spectral and radiometric differences between the OLI and OLI-2 sensors, we used this algorithm with the same numerical conditional values for both Landsat 8 and 9 images.

2.2.2. Estimation of Burned Extents

Unlike active fire detections, which require a high temporal resolution due to the need to capture the fire in progress, burned area estimation using satellite remote sensing is less demanding in this regard. For active fire detection, the fire must be burning at the exact time of satellite image acquisition, while cloud-cover free conditions are necessary for accurate detection [41]. In contrast, post-burn analysis can take advantage of the lasting effects of fire—such as charred vegetation or burn scars—that persist for extended periods after the fire has been extinguished [27]. This persistence reduces the urgency of capturing the event at a specific time, thus allowing for the use of satellites with higher-spatial-resolution but lower temporal frequency, as they are more likely to acquire clear imagery of the burned area after the fire event.
Post-fire analysis using remote sensing is a common practice in disaster management and environmental monitoring. The Copernicus Emergency Management Service often provides detailed mapping and assessments of disasters, including wildfires, on a global scale. However, for smaller-scale agricultural fires, which often do not trigger emergency protocols, alternative solutions are necessary.
The MINDED-FBA tool [42] is a remote sensing based tool used to determine the extent of flooding and burning. The tool incorporates optical multispectral (OMS) and synthetic aperture radar (SAR) modules, which allow us to perform automatic multi-index differencing threshold selection procedures aimed at image classification. The OMS module is particularly effective for determining the extent of an area burned area, which is useful for validating active fire detections and evaluating extensive biomass burning areas [27,43].
The implementation of the MINDED-FBA tool requires the input of both pre-event (t0) and post-event (t1) scenes, which are used to calculate a set of pre-defined indices and corresponding single-index differences. To estimate burned areas, the tool pre-selects four multispectral indices (i.e., NBRs, NBRl, NBR2 and NDVI). The tool performs the automatic thresholding of these single-index differencing scenes, assuming the further away from the theoretical zero, the higher the magnitude of change. This procedure is then followed by density slicing to classify pixels as ‘no change’ (Nc), ‘low-magnitude change’ (LMc), or ‘high-magnitude change’ (HMc). These classifications are then combined into an overall map with a pixel-by-pixel majority analysis. If no majority is found among the four single-index difference classifications, that pixel is assigned with the ‘mixed’ class [27,42,43].
The MINDED_FBA tool was utilized for both the Terras de Bouro and Póvoa de Varzim case studies, using the OMS module with Level-2 surface reflectance products derived from Landsat and Sentinel-2 data [38,39,44] (Table 2). All the available pre-processing procedures were selected, including cloud/cloud shadow masking and permanent water body masking (considering thematic data from [45] as well as topographic correction (using ALOWS WORLD 3D-30 m, [46]). Since the processing included burned area estimation, the resulting classifications (i.e., Nc, LMc, HMc and Mixed) should correspond to burn-related changes.
The ICNF releases the yearly burned area dataset. This consists of a multi-source survey, combining fieldwork and remote sensing interpretation. This dataset includes information about the starting and ending date, as well as the location, type (e.g., negligent or intentional), and cause of fire (e.g., accidental, arson, pile, or extensive residue burnings). This official dataset represents the best ground-truth alternative for validating burned area extent estimations in Portugal [42].

2.2.3. Active Fire Detections of Agricultural Residue Open-Burning

Besides the satellite active fire detections and post-burn analysis, this study considered the ABR, which was made available by ICNF for this study. This dataset represents the best available register of agricultural residue burning occurrences in Portugal. However, it should be highlighted that not all registered burns are executed, and that neither are all burn activities included in the ABR database. In many cases, these practices are not undertaken (at least on the requested date) due to unfavorable local weather conditions, which have a direct influence on the moisture level of the biomass residues, which must be dry enough to allow burning. On the other hand, not all agricultural burning requests are processed and integrated into the ICNF ABR dataset, as some municipalities are still not integrated into this system. Finally, despite the legal requirements for registering such practices, unauthorized burns remain widespread in Portugal.
To evaluate the capability of detecting the open burnings of agricultural residues through satellite RS, we used geospatial analysis tools (ArcGIS Pro, v3.2.2) to intersect the ABR database with the different RS fire detection products.
Beyond the direct intersection between both datasets, we implemented supplementary procedures to address potential spatial and temporal inconsistencies between them. In respect to the ABR database, we tested the application of distance buffers of 10 m and 100 m to account for Global Positioning System (GPS) inaccuracies, as well as temporal buffers from the reported request dates—including ±1 day and ±1 week. As for the FRPs, spatial buffers were implemented in respect to each pixel centroid, considering the corresponding sensor spatial resolution (Table 1). The analysis of the intersections according to different buffering considerations provides insight into the detection capabilities of different remote sensing products, the optimal time windows for detection, and the characteristics of various types of agricultural residue burning.

3. Results

3.1. Terras De Bouro

In Table 3, we can find the available products used to detect active fires in Terras de Bouro (as well as for the Póvoa de Varzim case study-Section 3.2). Regarding the FRPs, we listed the initial and end time of acquisitions from VIIRS, MODIS and SLSTR, including those scenes resulting in positive and negative detections. As for SEVIRI, we only listed the acquisition time for positive detections since this platform is capable of producing FRPs every 15 min.
For the Terras do Bouro case study, we can verify that most FRP detections during 16 February 2023, correspond to the SEVIRI, with a total of 21 scenes with positive detections. In particular, these are concentrated within the 9:30–16:15 period. As for the remaining sensors, the VIIRS instruments included most detections, including one from S-NPP during the early nighttime period, which was detected by neither VIIRS/NOAA-20 nor MODIS/Aqua at around the same period. As for daytime detections, all satellites registered positive detections from 10:50 to 14:30. Using the Daytime Active Fire Detection Algorithm, we could also find positive detections for February 16th using the Landsat 9 OLI-2 SR data.
Figure 3a shows that all FRP acquisitions from February 16th were detected northern of the Terras de Bouro village, corresponding to 14 detections for SEVIRI, 10 for VIIRS, 4 for MODIS, and 3 for SLSRT. The most southern detection corresponds to the single nighttime detection provided by VIIRS/S-NPP (02:12–02-18, previously mentioned from the analysis of Table 3), which recorded a relatively low-magnitude FRP (4.9 MW) (Figure 3b). The improved spatial resolution of the Landsat 9 OLI-2 seems to indicate that 6 different active fires took place for around 11 h, demonstrating good spatial correlation with the remaining lower-resolution FRPs (Figure 3c). The highest-magnitude FRP detections were obtained from SEVIRI/MSG (74.4 MW), SLSTR/Sentinel-3B (72.0 MW), MODIS/Terra (56.1 MW), and VIIRS/NOAA-20 (35.8 MW). For such higher-magnitude FRP detections, the pixel centroids from VIIRS/NOAA-20 and MODIS/Aqua are located within less than 1 km in respect to the SLSTR/Sentinel-3B pixel centroid (Figure 3b).
In Figure 3a, we also included every positive detection, in terms of FRP for VIIRS, MODIS, and SEVIRI, during the antecedent and following days with respect to 16 February. For these periods (as well as for the subsequent national-level analysis in Section 3.3), we did not include the SLSTR/Sentinel-3 data, since we could not find any readily available catalog in archive mode (e.g., NETCDF). Within both periods, most detections were registered by SEVIRI (respectively 107 and 100), followed by VIIRS (51 and 74) and MODIS (5 and 9).
Regarding the creation of the MINDED-FBA map (Figure 3f), which was produced during the whole period from 31 January to 24 February, we found several areas classified as showing low-magnitude changes. Such areas seem to have a good spatial correlation with both FRP and DAFDA, as well as with the ICNF dataset showing the yearly burned area. The few pixels classified as ‘mixed’ resemble “salt-and-pepper” noise [27,43].
The analysis of the ABR dataset map (Figure 3d) from 31 January to 24 February indicates that there were a total of 389 requests. Among these, there were only 3 records corresponding to extensive burning requests (all from 7 February), while the remainder corresponded to burning requests for agricultural residues (31/01–15/02: 311; 16/02: 23; and 17/02–24/02: 52). When comparing the ABR database (Figure 3d) with the remaining datasets, there were few apparent spatial correlations with either active fire detections (Figure 3a,b) or post-burn extents (Figure 3e,f). The few exceptions seemed to correspond to a couple of extensive burning requests from 7 February 2023, which seemed to be detected by VIIRS. However, such detections occurred on 6 February 2023 (from 2:01 to 2:49), corresponding to –1 day difference. However, when applying a distance buffer equivalent to each satellite instrument resolution (Table 1), we found 156 intersections with SEVIRI, corresponding to 11 burning requests and 2 extensive burning requests. However, if we extended the temporal window to minus or plus 1 day, we found 358 intersections with SEVIRI, 2 with VIIRS, and 1 with MODIS. These corresponded to 29 agriculture burning requests and 2 extensive burnings. Likewise, if we consider a temporal window of minus or plus 1 week, we obtain 1565 intersections with SEVIRI, 6 with MODIS, and 4 with VIIRS.

3.2. Póvoa De Varzim

Regarding the Póvoa de Varzim case study, the analysis shown in Table 3 included one positive detection from VIIRS/NOAA for 2 March, and two positives for March 4th employing both VIIRS options (S-NPP and NOAA). All options were assessed during the early afternoon period.
The analysis of Figure 4a,b allows us to verify that only one detection occurred on 2 March, which registered the lowest magnitude of FRP (1.9 MW). For 4 March, VIIRS/S-NPP included another detection, which resulted in the overall highest FRP magnitude (16.9 MW), while NOAA-20 detected the other two points (with 4.1 and 4.2 MW values). In the Sentinel-2B scene from 3 March (i.e., corresponding to the day in between the two controlled fire events) (Figure 4c), it is possible to confirm the outline of the first fire from 2 March (Figure 4a), whose pixel centroid is located circa 100 m (i.e., below the VIIRS spatial resolution size) from the fire perimeter of that day (Figure 4e). As for 4 March, the other 3 points are also located within a certain distance under the VIIRS spatial resolution limit.
Regarding the MINDED-FBA results (Figure 4d), we can verify the detection of both high-magnitude and low-magnitude changes, showing good spatial correlation with the overall fire perimeter. However, besides the mixed pixels, which once again resemble ‘salt-and-pepper’ noise, it is also possible to find other patches of low-magnitude change pixels distributed around the Eucalyptus clear-cut plot. When comparing Sentinel-2 scenes from 28 February (Figure 4a) and 15 March (Figure 4d) (which corresponds to the t0 and t1 scenes used to implement the MINDED-FBA tool), it is possible to verify that such low-magnitude change pixels seem to be related with land-use changes, particularly from plowing operations external to the study area.
Regarding the ABR dataset, we found a single register for March 2nd, corresponding to a burning request, which was located on the access pathway adjacent to the plantation plot at approximately 100 m distance from the overall burned area (Figure 4c).

3.3. Portugal Mainland

In addition to the two local case studies, we conducted a systematic comparison for mainland Portugal between remote sensing detections of active fires and ABR from 2019 to 2023.
Regarding the active fire detections, we considered those FRPs which were readily available in archive catalogs within the considered period. These included VIIRS (S-NPP and NOAA-20), MODIS (Terra and Aqua), and SEVIRI (Meteosat Second Generation). After acquiring all available FRPs, we applied spatial buffers to account for the resolution of each product and performed direct spatial intersections with the ABR coordinates (Table 4). This intersection resulted in 3410 matches, including 3044 from SEVIRI, 312 from MODIS, and 54 from VIIRS. In addition to the assessing total number of intersections, we also analyzed the number of unique burning requests that resulted in at least one detection using the FRPs (Table 5). This analysis considered both types of agricultural burning requests, resulting in 1239 standard burning requests and 132 extensive burning requests.
Table 4 also includes the total number of intersections when additional spatial buffers (10 m and 100 m) and temporal buffers (±1 day and ±1 week) are applied. The results indicate that the temporal buffers substantially increase the number of intersections compared to the spatial buffers. While the total number of intersections between the FRPs and the ABR dataset is 3410, adding spatial buffers of 10 m and 100 m results in overall increases of 1% and 13%, respectively, compared to the original dataset. Instead, applying a temporal buffer of ±1 day increases intersections by 171%, rising to 174% and 205% when combined with the spatial buffers of 10 m and 100 m, respectively. The effect of a temporal buffer of ±1 week is even more pronounced, increasing the intersections by 1302%, and, respectively, 1314% and 1438% when combined with both 10 m and 100 m spatial buffers. As for the individual sensors, this increase is even greater for SEVIRI, followed by VIIRS and MODIS.
As for the unique burning request analysis (Table 5), we observe a similar effect from the different buffering parameters. Introducing 10 m and 100 m spatial buffers results in increases of 1% and 14%, respectively, while applying ±1-day and ±1-week temporal buffers alone leads to increases of 143% and 972%. When combining a 10 m spatial buffer with ±1-day and ±1-week buffers, the increases reach 147% and 982%, respectively, whereas a 100 m buffer combined with ±1-day and ±1-week buffers increases reach to 173% and 1077%. The impact of both spatial and temporal buffers is more pronounced for standard burnings than for extensive burnings, with maximum increases of 1182% and 92%, respectively, under the ±1-week and 100 m buffer configurations. Comparing Table 4 with Table 5 shows that the total number of detections is substantially higher than the unique number of requests.
Figure 5 presents a comparison between the timeseries of active fire detections and the ABR database, highlighting their overall intersections (without any spatial or temporal buffers). From the analysis of Figure 5a, we observe a significant number of fire detections during the summer months of 2023, 2020, and 2022 (mainly in July and August), as well as a secondary period of high frequency after January 2023. In contrast, in Figure 5b, the ABR data appear more evenly distributed, with peak activity occurring from late autumn to early spring (October to May). The overall intersections between both datasets (Figure 5c) show significantly lower counts but tend to follow the same seasonal pattern as the ABR database, with peaks occurring from late autumn to spring. However, certain months with high ABR counts (e.g., November–December 2022) do not show a corresponding increase in the intersection series.
Figure 6 shows the distribution of the different FRP instruments and times of day when intersections with the ABR occur (again without any buffers). Among VIIRS and MODIS, most detections occur from morning to early afternoon, corresponding to the acquisition times of most instruments. Another period of detection is observed during early dawn (02:00 to 03:30), particularly when using NOAA-20 and S-NPP. For the remainder of the day, SEVIRI is the only instrument capable of detecting active fires, with the highest detection period generally occurring between 16:00 and 23:30. In terms of FRP magnitude, SEVIRI records the highest values, followed by MODIS/Aqua and VIIRS (S-NPP and NOAA-20).

4. Discussion

The results of this study demonstrate the capability of various remote sensing products to detect small-scale agricultural fires and prescribed fire interventions.
In the local-scale case study of Terras de Bouro, which involved multiple extensive agricultural burnings, we found a strong spatial correlation between FRPs from all instruments considered (VIIRS, MODIS, SLSTR, and SEVIRI) and the Daytime Active Fire Detection Algorithm applied to a Landsat 9 OLI-2 scene (at 30 m spatial resolution). Despite using the [18] algorithm on a Landsat 9 OLI-2 scene, when using the same numerical conditional values originally indicated for Landsat 8 OLI, we were still successful in detecting a potential daytime active fire. Additionally, we observed a good correlation with post-fire analysis; this included the MINDED-FBA results derived from surface reflectance products of Landsat 8 and 9 scenes (also at 30 m resolution). The extent of the burned areas was further confirmed by the official yearly burned area dataset from ICNF, which attributed most of these fires to negligent extensive burning performed for pasture management. However, when compared to the agriculture burning request dataset, the correspondence was more limited and was concentrated primarily in extensive burning requests.
For the Póvoa de Varzim case study, which focused on a smaller area with a low-intensity controlled prescribed fire for fuel management, carried out over two days, we found that only VIIRS (S-NPP and NOAA-20) detected active fires. This suggests that, despite the small scale of the event—especially on 2 March, when the burned area was approximately 3400 m2—VIIRS S-NPP successfully detected the fire, even though the event’s size was several times smaller than the instrument’s pixel resolution (375 m, corresponding to an area of over 140,000 m2). Other FRPs from MODIS, SLSTR, and SEVIRI failed to detect fires on either day, which was likely due to the insufficient spatial resolution and sensitivity for such small-scale, low-intensity events.
Although the ICNF yearly burned area dataset was unavailable for this case, fire-related changes were identified using the MINDED-FBA tool and Sentinel-2 imagery (with 20 m spatial resolution). Despite detecting several high-magnitude changes within the known burned area, these classification results are likely not indicative of high burn severity. Indeed, the small extent of the study area may exacerbate the weight of fire-related changes relative to the overall image statistics (hence affecting the results of automatic thresholding). Additionally, the MINDED-FBA outputs also included low-magnitude changes that are not related to fires but are instead caused by land-use changes related to plowing. This indicates that, despite the pre-processing steps, the tool is still sensitive to producing false positives in such cases, which is to be expected anyway, as this is a non-supervised classification method. Nevertheless, the MINDED-FBA tool seems to effectively resolve much of the salt-and-pepper noise, as these pixels are classified as ‘mixed.’
Regarding the agriculture burning request database drawn from ICNF, the Póvoa de Varzim case study only includes one record corresponding to this event. Such findings suggest that, as may be the case with many similar instances, these practices do not always occur at the exact requested locations and dates, whether due to unfavorable conditions or user-induced inconsistencies. Therefore, these findings justify the introduction of additional considerations into the subsequent systematic analysis of mainland Portugal.
This national-scale analysis of Portugal’s mainland was conducted for the period from 2019 to 2023. When comparing the timeseries of FRP detections (Figure 5a) with the ABR database (Figure 6a), we verify that the majority of active fire detections corresponded to wildfires occurred during July and August (70,0%), clearly outside the period for agriculture burning requests. Associated to the significant size of both remote sensing and reference datasets, as well as their significant spatial-temporal overlap, we found that the data structure was unsuitable for efforts to perform an accuracy assessment following the procedure based on the confusion matrix approach (e.g., [49]). Nevertheless, in Figure 5c and Table 4, we identified a significant number of correspondences when intersecting the FRP and ABR databases. These had an equivalent meaning to true positive detections. To account for potential spatial and temporal inaccuracies in the ABR dataset, we applied spatial buffers of 10 m and 100 m to the ABR coordinates, along with temporal buffers of ±1 day and ±1 week relative to the ABR entry dates. These assumptions resulted in a substantial increase in the number of intersections with the FRPs, with the temporal buffer yielding a much larger increase compared to the two spatial buffer scenarios. Nevertheless, the maximum total number of potentially detected individual requests—16,141 (with a 100 m buffer and ±1 week)—remained significantly lower (less than 1%) than the total number of requests recorded in the ABR dataset for the 2019–2023 period (4,099,487). This discrepancy was likely due to the following conditions:
  • Not every request resulting in an actual burning;
  • Inconsistencies in the reported location coordinates and/or dates (extending beyond the considered spatial and temporal buffers);
  • Non-favorable cloudy conditions or haze from fire smoke during the satellite image acquisition, which may prevent accurate or any FRP measurements;
  • Insufficient spatial and/or temporal resolutions of the acquiring sensors;
  • A combination of one or more of the above-mentioned conditions.
By considering the best available reference datasets for both agriculture burning requests and official yearly burned areas, both provided by INCF, the results of this study highlight the benefits and limitations of using remote sensing products to detect small agriculture fires, while acknowledging inherent uncertainties. These uncertainties arise not only from sensor limitations (e.g., spatial resolution and temporal coverage) but also from the assumed ground truth data, as the burning requests database may not fully reflect actual ground conditions. For instance, requests to burn are made several days in advance, but meteorological conditions or other factors may lead to changes in plans, resulting in no actual burning.
For the Portuguese mainland, the current setup of satellites capable of producing FRP with sub 1 km spatial resolutions is mostly operational during the daytime (Figure 6a). The remaining periods are in some cases only covered by SEVIRI, which includes the worst spatial resolution (i.e., 3 km). The observed gradual decrease in FRP detections from nighttime to dawn likely reflects the cooling of burnings from the preceding daytime period. Despite their spatial resolution limitations, we found the lowest resolution sensors—SEVIRI and MODIS—resulted in the highest FRP magnitudes [MW] (i.e., up to ~500 MW). These higher magnitudes likely result from the increased temporal resolution of these platforms (i.e., 15 min for SEVIRI and 1 day for MODIS), increasing the likelihood of capturing larger agriculture fires at their peak intensity. Nevertheless, such magnitudes are still significantly lower than other confirmed summertime wildfires which, during this period, achieve FRP magnitudes over 1200 MW.
For the systematic analysis of small agricultural fires, such as the recurrent practices of burning in Portugal, FRPs seem to be particularly effective for detecting extensive burnings, such as those used for pasture creation. However, for smaller and low-severity events, such as those observed in the Póvoa de Varzim case study, higher-resolution satellite data can progressively detect smaller fires. In particular, the VIIRS satellites offer the best compromise between temporal and spatial resolutions. Nevertheless, the Póvoa de Varzim prescribed fire is still considerably larger than many agriculture small-scale burning events, such as those involving hte pile burning of biomass residues, which are typically only a few meters wide.
The use of active fire detection algorithms integrating higher-spatial-resolution imagery could, in theory, enable the detection of small-scale piles [18,24]. However, the combined revisit times of Landsat 8 and 9 (8 days) and Sentinel-2A and Sentinel-2B (5 days) are still insufficient to consistently detect the majority of these events due to their reduced spatial extents and short durations [22]. Other approaches for detecting burned areas using higher-spatial-resolution imagery, like MINDED-FBA, are more consistent, particularly for extensive open-fire agriculture burnings. These practices tend to occur during the period of vegetation dormancy and are often followed by the slow natural regeneration of pastures. Consequently, the effects of such fires can often be identified weeks or even months after the event. In contrast, the burned areas from smaller events like pile burning are less likely to be detected, and may only be captured using active fire detections. For this reason, we consider active fire detection algorithms based on higher-spatial-resolution data to be relevant, since they may be the only possibility of detecting smaller agriculture burnings. Despite their reduced temporal resolution, even a few detections may still represent an advance over current air quality models. In the future, it would be valuable to explore how FRP data could be used in conjunction with the ABR database to create prediction models for small-scale fire events. This would require systematic ground-level truth checking, improved temporal and spatial accuracy of the ABR database, and more detailed descriptions of each record.

5. Conclusions

This study demonstrates the capabilities of various remote sensing products and approaches for detecting small-scale fires. The analysis of two local case studies shows that fire radiative power data from VIIRS, MODIS, SLSTR, and SEVIRI has a strong spatial correlation with higher-resolution data, including Landsat 8, Landsat 9, and Sentinel-2. While these products are most effective in detecting extensive burning, such as that used for land clearing and pasture creation, they struggle to detect small-scale agriculture fires, such as those resulting from the pile burning of agriculture residues. These findings confirm that active fire detections become progressively more dependent on the spatial resolution of the sensor. Among the tested products, VIIRS-equipped satellites offer the best compromise between temporal and spatial resolutions.
At a national level, the intersection of fire radiative power products with an agricultural burning request database reveals limited overlap, even when applying spatial and temporal buffers. While these buffers significantly increased intersections, they still represented less than 1% of the total dataset, as not all requests resulted in burnings and many unregistered negligent burnings took place.
Despite these limitations, even partial detections have the potential to be valuable for multiple purposes, such as regulatory monitoring, wildfire risk management, and refining atmospheric emission estimates. Beyond Portugal, these findings hold broader implications for agricultural and environmental management policies. The improved detection and monitoring of small-scale fires using remote sensing products can aid policymakers in developing strategies to regulate agricultural burning, mitigate fire-related emissions, and reduce wildfire risks in other regions with similar challenges. Furthermore, integrating these technologies into environmental monitoring programs can enhance efforts to track fire-induced land cover changes and assess their long-term ecological impacts on a global scale.
Future efforts could explore the integration of additional and diverse satellite data (e.g., thermal infrared, high spatial resolution, and SAR) for systematic active fire detections. Such products could be combined with agriculture burning request databases to develop predictive models for small-scale fire events. This would require systematic ground-level validation, improvements in the precision of the request database, and the use of more detailed metadata for each entry.

Author Contributions

Conceptualization, E.R.O. and C.G.; methodology, E.R.O., B.T.S. and C.G.; software, E.R.O. and B.T.S.; validation, E.R.O. and B.T.S.; formal analysis, E.R.O. and B.T.S.; investigation, E.R.O., B.T.S., D.L., S.C. and C.G.; resources, F.L.A., L.D. and C.G.; data curation, E.R.O., B.T.S. and C.G.; writing—original draft preparation, E.R.O., B.T.S. and C.G.; writing—review and editing, E.R.O., B.T.S., D.L., S.C., F.L.A., L.D. and C.G.; visualization, E.R.O., B.T.S., C.G.; supervision, E.R.O., D.L., F.L.A., L.D. and C.G.; project administration, D.L. and C.G.; funding acquisition, D.L., S.C. and C.G. All authors have read and agreed to the published version of the manuscript.

Funding

The authors acknowledge the financial support from the Fundação para a Ciência e a Tecnologia (FCT) through national funds within the projects PRUNING (DOI: 10.54499/2022.01045.PTDC) and FirEProd (DOI: 10.54499/PCIF/MOS/0071/2019), as well as the research contract granted to Carla Gama (DOI: 10.54499/2021.00732.CEECIND/CP1659/CT0006), and the CESAM Associated Laboratory (UIDP/50017/2020 + UIDB/50017/2020 + LA/P/0094/2020). The authors also acknowledge the support provided by the ERA Chair BESIDE project, funded by the European Union’s Horizon 2020 program under grant agreement No. 951389 (DOI: 10.3030/951389). Thanks are also due to the Agreement between DSFTA-UNISI and IGG-CNR (2018/9150000123).

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors on request.

Acknowledgments

The authors wish to thank the Instituto da Conservação da Natureza e das Florestas (ICNF) for providing access to the national database of agricultural burning requests for mainland Portugal. The authors also thank RAIZ-Instituto da Floresta e Papel for the support in the establishment and trials monitoring, and Portucalea Forest Association for the support mainly related to providing the study area, for conducting the prescribed fire operation, and for their assistance in establishing, measuring, and maintaining the field experiment.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Webb, J.; Hutchings, N.; Amon, B.; Nielsen, O.-K.; Phillips, R.; Dämmgen, U. Field Burning of Agricultural Residues. EMEP/EEA Emiss. Invent. Guideb. 2013, 2013, 1–14. [Google Scholar]
  2. Souza, M.E.B.; Pacheco, A.P.; Teixeira, J.G. Systematising Experts’ Understanding of Traditional Burning in Portugal: A Mental Model Approach. Int. J. Wildl. Fire 2023, 32, 1558–1575. [Google Scholar] [CrossRef]
  3. Santín, C.; Doerr, S.H. Fire Effects on Soils: The Human Dimension. Philos. Trans. R. Soc. B Biol. Sci. 2016, 371, 28–34. [Google Scholar] [CrossRef]
  4. Espinosa, J.; Palheiro, P.; Loureiro, C.; Ascoli, D.; Esposito, A.; Fernandes, P.M. Fire-Severity Mitigation by Prescribed Burning Assessed from Fire-Treatment Encounters in Maritime Pine Stands. Can. J. For. Res. 2019, 49, 205–211. [Google Scholar] [CrossRef]
  5. Antunes, C.C.; Viegas, D.X.; Mendes, J.M.; Em, M.; Sociais, D.; Naturais, R.; Catedrático, P.; Auxiliar, P. Avaliação Do Risco de Incêndio Florestal No Concelho de Arganil. Silva Lusit. 2011, 19, 165–179. [Google Scholar] [CrossRef]
  6. Xu, Y.; Huang, Z.; Ye, J.; Zheng, J. Hourly Emissions of Air Pollutants and Greenhouse Gases from Open Biomass Burning in China during 2016–2020. Sci. Data 2023, 10, 629. [Google Scholar] [CrossRef] [PubMed]
  7. López-Caravaca, A.; Vicente, E.D.; Figueiredo, D.; Evtyugina, M.; Nicolás, J.F.; Yubero, E.; Galindo, N.; Ryšavý, J.; Alves, C.A. Gaseous and Aerosol Emissions from Open Burning of Tree Pruning and Hedge Trimming Residues: Detailed Composition and Toxicity. Atmos. Environ. 2024, 338, 120849. [Google Scholar] [CrossRef]
  8. Alves, C.A.; Vicente, E.D.; Evtyugina, M.; Vicente, A.; Pio, C.; Amado, M.F.; Mahía, P.L. Gaseous and Speciated Particulate Emissions from the Open Burning of Wastes from Tree Pruning. Atmos. Res. 2019, 226, 110–121. [Google Scholar] [CrossRef]
  9. Gonçalves, C.; Evtyugina, M.; Alves, C.; Monteiro, C.; Pio, C.; Tomé, M. Organic Particulate Emissions from Field Burning of Garden and Agriculture Residues. Atmos. Res. 2011, 101, 666–680. [Google Scholar] [CrossRef]
  10. Pearson, P.; Bodin, S.; Gittelson, A.; Kinney, S.; McCarty, J.L.; Stevenson, G.; Albertengo, J. Fire in the Fields: Moving Beyond the Damage of Open Agricultural Burning on Communities, Soil, and the Cryosphere Fire in the Fields: Moving Beyond the Damage of Open Agricultural Burning on Communities, Soil, and the Cryosphere. A CCAC Project Summary Report: Impacts and Reduction of Open Burning in the Andes, Himalayas—and Globally; Co-Authors for ICCI: Fire in the Fields: Moving Beyond the Damage of Open Agricultural Burning on Communities, Soil, and the Cryosphere; Climate and Clean Air Coalition to Reduce Short-Lived Climate Pollutants (CCAC): Paris, France, 2015. [Google Scholar] [CrossRef]
  11. Parente, J.; Girona-García, A.; Lopes, A.R.; Keizer, J.J.; Vieira, D.C.S. Prediction, Validation, and Uncertainties of a Nation-Wide Post-Fire Soil Erosion Risk Assessment in Portugal. Sci. Rep. 2022, 12, 2945. [Google Scholar] [CrossRef]
  12. Nunes, L.J.R.; Raposo, M.A.M.; Meireles, C.I.R.; Gomes, C.J.P.; Ribeiro, N.M.C.A. The Impact of Rural Fires on the Development of Invasive Species: Analysis of a Case Study with Acacia Dealbata Link. in Casal Do Rei (Seia, Portugal). Environment 2021, 8, 44. [Google Scholar] [CrossRef]
  13. Han, X.; Frey, G.E.; Sun, C. Regulation and Practice of Forest-Management Fires on Private Lands in the Southeast United States: Legal Open Burns versus Certified Prescribed Burns. J. For. 2020, 118, 385–402. [Google Scholar] [CrossRef]
  14. Fernandes, P.M.; Botelho, H.S. A Review of Prescribed Burning Effectiveness in Fire Hazard Reduction. Int. J. Wildl. Fire 2003, 12, 117–128. [Google Scholar] [CrossRef]
  15. Hmielowski, T.L.; Carter, S.K.; Spaul, H.; Helmers, D.P.; Radeloff, V.C.; Zedler, P.H. Prioritizing Land Management Efforts at a Landscape Scale: A Case Study Using Prescribed Fire in Wisconsin. Ecol. Appl. 2015, 26, 1018–1029. [Google Scholar] [CrossRef] [PubMed]
  16. Matin, M.A.; Chitale, V.S.; Murthy, M.S.R.; Uddin, K.; Bajracharya, B.; Pradhan, S. Understanding Forest Fire Patterns and Risk in Nepal Using Remote Sensing, Geographic Information System and Historical Fire Data. Int. J. Wildl. Fire 2017, 26, 276. [Google Scholar] [CrossRef]
  17. Schroeder, W.; Oliva, P.; Giglio, L.; Csiszar, I.A. The New VIIRS 375m Active Fire Detection Data Product: Algorithm Description and Initial Assessment. Remote Sens. Environ. 2014, 143, 85–96. [Google Scholar] [CrossRef]
  18. Schroeder, W.; Oliva, P.; Giglio, L.; Quayle, B.; Lorenz, E.; Morelli, F. Active Fire Detection Using Landsat-8/OLI Data. Remote Sens. Environ. 2016, 185, 210–220. [Google Scholar] [CrossRef]
  19. Vilar, L.; Camia, A.; San-Miguel-Ayanz, J. A Comparison of Remote Sensing Products and Forest Fire Statistics for Improving Fire Information in Mediterranean Europe. Eur. J. Remote Sens. 2015, 48, 345–364. [Google Scholar] [CrossRef]
  20. Hall, J.V.; Loboda, T.V.; Giglio, L.; McCarty, G.W. A MODIS-Based Burned Area Assessment for Russian Croplands: Mapping Requirements and Challenges. Remote Sens. Environ. 2016, 184, 506–521. [Google Scholar] [CrossRef]
  21. Liu, T.; Mickley, L.J.; Singh, S.; Jain, M.; DeFries, R.S.; Marlier, M.E. Crop Residue Burning Practices across North India Inferred from Household Survey Data: Bridging Gaps in Satellite Observations. Atmos. Environ. X 2020, 8, 100091. [Google Scholar] [CrossRef]
  22. Murphy, S.W.; de Souza Filho, C.R.; Wright, R.; Sabatino, G.; Correa Pabon, R. HOTMAP: Global Hot Target Detection at Moderate Spatial Resolution. Remote Sens. Environ. 2016, 177, 78–88. [Google Scholar] [CrossRef]
  23. Kumar, S.S.; Roy, D.P. Global Operational Land Imager Landsat-8 Reflectance-Based Active Fire Detection Algorithm. Int. J. Digit. Earth 2018, 11, 154–178. [Google Scholar] [CrossRef]
  24. Hu, X.; Ban, Y.; Nascetti, A. Sentinel-2 MSI Data for Active Fire Detection in Major Fire-Prone Biomes: A Multi-Criteria Approach. Int. J. Appl. Earth Obs. Geoinf. 2021, 101, 102347. [Google Scholar] [CrossRef]
  25. Kavzoglu, T.; Erdemir, M.Y.; Tonbul, H. Evaluating Performances of Spectral Indices for Burned Area Mapping Using Object-Based Image Analysis. In Proceedings of the 11th International Symposium on Spatial Accuracy 2016, East Lansing, MI, USA, 8–11 July 2016; pp. 1–7. [Google Scholar]
  26. Chuvieco, E.; Martín, M.P.; Palacios, A. Assessment of Different Spectral Indices in the Red-near-Infrared Spectral Domain for Burned Land Discrimination. Int. J. Remote Sens. 2002, 23, 5103–5110. [Google Scholar] [CrossRef]
  27. Oliveira, E.R.; Disperati, L.; Alves, F.L. A New Method (MINDED-BA) for Automatic Detection of Burned Areas Using Remote Sensing. Remote Sens. 2021, 13, 5164. [Google Scholar] [CrossRef]
  28. Jahanianfard, D.; Parente, J.; Gonzalez-Pelayo, O.; Benali, A. Multidecadal Satellite-Derived Portuguese Burn Severity Atlas (1984–2022). Earth Syst. Sci. Data 2024, 2024, 1–33. [Google Scholar]
  29. de Oliveira, E.; Colaço, M.C.; Fernandes, P.M.; Sequeira, A.C. Remains of Traditional Fire Use in Portugal: A Historical Analysis. Trees For. People 2023, 14, 100458. [Google Scholar] [CrossRef]
  30. Amorim, J.H.; Valente, J.; Cascão, P.; Ribeiro, L.M.; Viegas, D.X.; Ottmar, R.; Miranda, A.I. Near-Source Grid-Based Measurement of CO and PM2.5 Concentration during a Full-Scale Fire Experiment in Southern European Shrubland. Atmos. Environ. 2016, 145, 19–28. [Google Scholar] [CrossRef]
  31. Miranda, A.I.; Martins, V.; Cascão, P.; Amorim, J.H.; Valente, J.; Tavares, R.; Borrego, C.; Tchepel, O.; Ferreira, A.J.; Cordeiro, C.R.; et al. Monitoring of Firefighters Exposure to Smoke during Fire Experiments in Portugal. Environ. Int. 2010, 36, 736–745. [Google Scholar] [CrossRef]
  32. Jaffe, D.A.; O’Neill, S.M.; Larkin, N.K.; Holder, A.L.; Peterson, D.L.; Halofsky, J.E.; Rappold, A.G. Wildfire and Prescribed Burning Impacts on Air Quality in the United States. J. Air Waste Manag. Assoc. 2020, 70, 583–615. [Google Scholar] [CrossRef] [PubMed]
  33. ICNF Cartografia Nacional de Áreas Ardidas Dos Anos 1975 à Data Atual Em Conformidade Com o Disposto No n.° 5 Do Art.° 2.° Do Decreto-Lei no 124/2006, de 28 de Junho, Na Redação Dada Pelo Decreto-Lei n.° 17/2009, de 14 de Janeiro. Available online: https://sigservices.icnf.pt/server/rest/services/BDG/areas_ardidas/MapServer (accessed on 1 October 2024).
  34. Wooster, M.J.; Roberts, G.J.; Giglio, L.; Roy, D.; Freeborn, P.; Boschetti, L.; Justice, C.; Ichoku, C.; Schroeder, W.; Davies, D.; et al. Satellite Remote Sensing of Active Fires: History and Current Status, Applications and Future Requirements. Remote Sens. Environ. 2021, 267, 112694. [Google Scholar] [CrossRef]
  35. NASA. NASA Active Fire Data. Available online: https://www.earthdata.nasa.gov/learn/find-data/near-real-time/firms/active-fire-data (accessed on 1 October 2024).
  36. ESA. SLSTR Products. Available online: https://sentiwiki.copernicus.eu/web/slstr-products#S3-SLSTR-Products-L2-FRP-Products (accessed on 1 October 2024).
  37. Eumetsat LSA SAF. Available online: https://datalsasaf.lsasvcs.ipma.pt/PRODUCTS/MSG/FRP-PIXEL/HDF5/ (accessed on 1 October 2024).
  38. USGS. Landsat 8 Data Users Handbook; Department of the Interior U.S. Geological Survey: Sioux Falls, SD, USA, 2019; Volume 8.
  39. USGS. Landsat 9 Data Users Handbook Landsat 9 Data Users Handbook Version 1.0; Department of the Interior U.S. Geological Survey: Sioux Falls, SD, USA, 2022.
  40. NASA Layer Information: LANDSAT OLI (8 & 9) Fire and Thermal Anomalies (Day|Night, 30 Meters). Available online: https://firms.modaps.eosdis.nasa.gov/descriptions/FIRMS_Landsat_Firehotspots.html (accessed on 1 October 2024).
  41. 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]
  42. Oliveira, E.R.; Disperati, L.; Alves, F.L. MINDED-FBA: An Automatic Remote Sensing Tool for the Estimation of Flooded and Burned Areas. Remote Sens. 2023, 15, 724. [Google Scholar] [CrossRef]
  43. Oliveira, E.R.; Disperati, L.; Cenci, L.; Pereira, L.G.; Alves, F.L. Multi-Index Image Differencing Method (MINDED) for Flood Extent Estimations. Remote Sens. 2019, 11, 1305. [Google Scholar] [CrossRef]
  44. ESA. Sentinel-2 User Handbook; ESA: Paris, France, 2015; Volume 2. [Google Scholar]
  45. APA. Surface Water Bodies Rivers of Mainland Portugal: SNIAmb Spatial Data Set. Available online: https://data.europa.eu/data/datasets/massas-de-agua-superficiais-rios-de-portugal-continental-conjunto-de-dados-geografico-sniamb?locale=en (accessed on 1 October 2024).
  46. JAXA ALOS Global Digital Surface Model “ALOS World 3D–30 m (AW3D30)”. Available online: https://www.eorc.jaxa.jp/ALOS/en/dataset/aw3d30/aw3d30_e.htm (accessed on 18 January 2023).
  47. USGS. Earth Explorer. Available online: https://earthexplorer.usgs.gov/ (accessed on 12 November 2020).
  48. Copernicus Copernicus Browser. Available online: https://browser.dataspace.copernicus.eu/ (accessed on 1 October 2024).
  49. Congalton, R.G. A Review of Assessing the Accuracy of Classifications of Remotely Sensed Data. Remote Sens. Environ. 1991, 37, 35–46. [Google Scholar] [CrossRef]
Remotesensing 17 00051 g001
Figure 1. The location of the three case study areas: 1—Terras de Bouro; 2—Póvoa de Varzim; 3—Portugal Mainland (Coordinate Reference System: ETRS89/TM06).
Figure 1. The location of the three case study areas: 1—Terras de Bouro; 2—Póvoa de Varzim; 3—Portugal Mainland (Coordinate Reference System: ETRS89/TM06).
Remotesensing 17 00051 g001
Remotesensing 17 00051 g002
Figure 2. Prescribed fire of 2 March 2023, nearby Balazar, Póvoa de Varzim (photo by the authors).
Figure 2. Prescribed fire of 2 March 2023, nearby Balazar, Póvoa de Varzim (photo by the authors).
Remotesensing 17 00051 g002
Remotesensing 17 00051 g003
Figure 3. Terras de Bouro: (a) the date of active thermal anomalies detected with the Fire Radiative Power obtained from VIIRS (S-NPP and NOAA-20), MODIS (Terra and Aqua), and SLRS (Sentinel-3B) [35,37,47,48]; (b) the magnitude of active thermal anomalies detected with the Fire Radiative Power (FRP [MW]) obtained from VIIRS (S-NPP and NOAA-20), MODIS (Terra and Aqua)], and SLRS (Sentinel-3B) [35,37,47,48]; (c) detection results of the Daytime Active Fire Algorithm from Lansat-9 OLI-2 data [18,47]; (d) agriculture burning request (ABR) [47]; (e) burned area-related changes obtained from the MINDED-FBA tool [42,47]; (f) the official yearly database of burned areas, with an indication of occurrence date [33,47] (Coordinate Reference System: ETRS89/TM06).
Figure 3. Terras de Bouro: (a) the date of active thermal anomalies detected with the Fire Radiative Power obtained from VIIRS (S-NPP and NOAA-20), MODIS (Terra and Aqua), and SLRS (Sentinel-3B) [35,37,47,48]; (b) the magnitude of active thermal anomalies detected with the Fire Radiative Power (FRP [MW]) obtained from VIIRS (S-NPP and NOAA-20), MODIS (Terra and Aqua)], and SLRS (Sentinel-3B) [35,37,47,48]; (c) detection results of the Daytime Active Fire Algorithm from Lansat-9 OLI-2 data [18,47]; (d) agriculture burning request (ABR) [47]; (e) burned area-related changes obtained from the MINDED-FBA tool [42,47]; (f) the official yearly database of burned areas, with an indication of occurrence date [33,47] (Coordinate Reference System: ETRS89/TM06).
Remotesensing 17 00051 g003
Remotesensing 17 00051 g004
Figure 4. Póvoa de Varzim: (a) date of active thermal anomalies detected with the Fire Radiative Power obtained from VIIRS (S-NPP and NOAA-20) [35,48]; (b) magnitude of active thermal anomalies detected with the Fire Radiative Power (FRP [MW]) obtained from VIIRS (S-NPP and NOAA-20) [35,48]; (c) agriculture burning request (ABR) [48]; (d) burned area-related changes obtained from the MINDED-FBA tool [42,48]; (e) Location of Eucalyptus clear-cut plantation and prescribed fire perimeters [48] (Coordinate Reference System: ETRS89/TM06).
Figure 4. Póvoa de Varzim: (a) date of active thermal anomalies detected with the Fire Radiative Power obtained from VIIRS (S-NPP and NOAA-20) [35,48]; (b) magnitude of active thermal anomalies detected with the Fire Radiative Power (FRP [MW]) obtained from VIIRS (S-NPP and NOAA-20) [35,48]; (c) agriculture burning request (ABR) [48]; (d) burned area-related changes obtained from the MINDED-FBA tool [42,48]; (e) Location of Eucalyptus clear-cut plantation and prescribed fire perimeters [48] (Coordinate Reference System: ETRS89/TM06).
Remotesensing 17 00051 g004
Remotesensing 17 00051 g005
Figure 5. (a) Timeseries of detections of active fires from FRPs (including VIIRS, MODIS and SEVIRI); (b) timeseries of agriculture burning requests (ICNF); (c) intersections between active fire detections and agriculture burning requests (ABRs), without buffers (Bin size = 1 week).
Figure 5. (a) Timeseries of detections of active fires from FRPs (including VIIRS, MODIS and SEVIRI); (b) timeseries of agriculture burning requests (ICNF); (c) intersections between active fire detections and agriculture burning requests (ABRs), without buffers (Bin size = 1 week).
Remotesensing 17 00051 g005
Remotesensing 17 00051 g006
Figure 6. Frequency of intersections between FRP (by satellite) and ABR (without buffers) according to (a) instrument and time of the day and (b) magnitude of FRP power [MW].
Figure 6. Frequency of intersections between FRP (by satellite) and ABR (without buffers) according to (a) instrument and time of the day and (b) magnitude of FRP power [MW].
Remotesensing 17 00051 g006
Table 1. Summary of FRPs considered in this study.
Table 1. Summary of FRPs considered in this study.
InstrumentSatelliteSpatial ResolutionTemporal ResolutionFRP
VIIRSS-NPP375 m1 dayVNP14IMGTDL_NRT [35]
VJ114IMGTDL_NRT [35]
NOAA-20
MODISTerra1000 m1 dayMOD14ML [35]
MYD14ML [35]
Aqua
SLSTRSentinel-3A
Sentinel-3B		1000 m2–3 dayS3A_SL_2_FRP [36]
S3B_SL_2_FRP [36]
SEVIRIMSG 13100 m15 minLSA SAF SEVIRI FRP-PIXEL [37]
1 Meteosat Second Generation (MSG).
Table 2. Pre- (t0) and post-event (t1) scenes processed with the OMS module of the MINDED-FBA tool [47,48].
Table 2. Pre- (t0) and post-event (t1) scenes processed with the OMS module of the MINDED-FBA tool [47,48].
Case StudyPre-Event ScenePost-Event Scene
Terras de BouroLandsat 9 OLI-2
31 January 2023		Landsat 8 OLI
24 February 2023
Póvoa de VarzimSentinel-2B MSI
28 February 2023		Sentinel-2A MSI
15 March 2023
Table 3. Acquisition times of FRPs and surface reflectance products used to perform active fire detections for the Terras do Bouro and Póvoa do Varzim case studies, including positive (†) and negative detections [35,37,47,48].
Table 3. Acquisition times of FRPs and surface reflectance products used to perform active fire detections for the Terras do Bouro and Póvoa do Varzim case studies, including positive (†) and negative detections [35,37,47,48].
Case StudyDateFRPSR
VIIRSMODISSLSTRSEVIRILandsat 9
S-NPPNOAA-20TerraAquaSentinel 3ASentinel 3BMSG *OLI-2
Terras de Bouro16 February 202302:12–02:18 †
13:36–13:42 †		01:18–01:24
03:00–03:06
12:42–12:48 †
14:24–14:30 †		10:50–10:55 †
21:55–22:00		02:10–02:15
13:15–13:20 †		 11:02–11:07 †
22:22–22:27		00:15; 09:30;
09:45; 10:00;
10:15; 11:00;
11:30; 12:00;
13:30; 13:45;
14:15; 14:30;
15:00; 15:30;
15:45; 16:15;
19:00; 19:15;
20:00; 21:00;
23:30		11:13–11:14 †
Póvoa de Varzim2 March 202302:48–02:54
12:30–12:36
14:12–14:18		01:54–02:00
03:36–03:42
13:18–13:24 †		10:35–10:40
21:40–21:45		02:10–02:15
02:15–02:20
13:20–13:25		10:39–10:42
21:59–22:02		10:48–10:51
22:08–22:11		--
4 March 202302:12–02:18
13:36–13:42 †		01:18–01:24
03:00–03:06
12:42–12:48 †
14:24–14:30		10:15–10:20
11:55–12:00
21:20–21:25
21:25–21:30
23:00–23:05		02:00–02:05
03:35–03:40
13:05–13:10
14:45–14:50		10:39–10:42
21:59–22:02		10:48–10:51
22:08–22:11
* The MSG/SEVIRI is capable of acquiring and producing FRPs every 15 min; therefore, only positive acquisitions are listed.
Table 4. Intersection between different FRPs (including VIIRS, MODIS, and SEVIRI) with agriculture burning requests, considering different additional spatial buffers (10 m and 100 m) and different temporal buffers (0 d, ±1 day, and ±1 week).
Table 4. Intersection between different FRPs (including VIIRS, MODIS, and SEVIRI) with agriculture burning requests, considering different additional spatial buffers (10 m and 100 m) and different temporal buffers (0 d, ±1 day, and ±1 week).
Without Spatial BufferBuffer 10 mBuffer 100 m
0 d±1 d±1 w0 d±1 d±1 w0 d±1 d±1 w
InstrumentVIIRS541043445610835267125419
MODIS312494108733152411474727691703
SEVIRI3044864846,3843071871046,7203324950850,323
Total3410924647,8153458934248,219386310,40252,445
Table 5. Unique request types detected with FRPs (including VIIRS, MODIS, and SEVIRI), considering different additional spatial buffers (10 m and 100 m) and different temporal buffers (0 d, ±1 day, and ±1 week).
Table 5. Unique request types detected with FRPs (including VIIRS, MODIS, and SEVIRI), considering different additional spatial buffers (10 m and 100 m) and different temporal buffers (0 d, ±1 day, and ±1 week).
Without Spatial BufferBuffer 10 mBuffer 100 m
0 d±1 d±1 w0 d±1 d±1 w0 d±1 d±1 w
Individual request typeBurning1239317314,4751255321414,6041406356215,887
Extensive 132160222136166229154186254
Total1371333314,6971391338014,8331560374816,141
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

Oliveira, E.R.; Silva, B.T.; Lopes, D.; Corticeiro, S.; Alves, F.L.; Disperati, L.; Gama, C. The Detection of Small-Scale Open-Burning Agriculture Fires Through Remote Sensing. Remote Sens. 2025, 17, 51. https://doi.org/10.3390/rs17010051

AMA Style

Oliveira ER, Silva BT, Lopes D, Corticeiro S, Alves FL, Disperati L, Gama C. The Detection of Small-Scale Open-Burning Agriculture Fires Through Remote Sensing. Remote Sensing. 2025; 17(1):51. https://doi.org/10.3390/rs17010051

Chicago/Turabian Style

Oliveira, Eduardo R., Bárbara T. Silva, Diogo Lopes, Sofia Corticeiro, Fátima L. Alves, Leonardo Disperati, and Carla Gama. 2025. "The Detection of Small-Scale Open-Burning Agriculture Fires Through Remote Sensing" Remote Sensing 17, no. 1: 51. https://doi.org/10.3390/rs17010051

APA Style

Oliveira, E. R., Silva, B. T., Lopes, D., Corticeiro, S., Alves, F. L., Disperati, L., & Gama, C. (2025). The Detection of Small-Scale Open-Burning Agriculture Fires Through Remote Sensing. Remote Sensing, 17(1), 51. https://doi.org/10.3390/rs17010051

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop