The paper proposed a GIS-based study, which allow analysis of the impact of wildfires using multispectral satellite images. A wildfire that occurred between 30 April and 3 May 2025 north of the village Birky, Kamin-Kashirsky district Volyn region (northwestern Ukraine), formed the basis of the study. Using channels in the near-infrared (NIR) and short-wave infrared (SWIR) ranges was found to be the most promising approach for promptly detecting wildfires and determining their long-term consequences. To interpret the study results, consequences and intensity of the wildfire were assessed using the US Geological Survey scale and a normalized burning coefficient. Creating forest fire intensity maps is key to developing vegetation restoration plans after fires and assessing the potential future impact on burnt areas.
Wildfires are among the most dangerous natural disasters today. They are caused by both natural and anthropogenic factors. In any case, they lead to significant economic, environmental, and socio-cultural losses. In the context of modern climate change and rising air temperatures, the intensity of forest and steppe fires is increasing. Traditional monitoring methods often prove to be ineffective due to the vastness of the studied areas, the time lag between the start of a fire and its detection, and limited access to certain territories. In this context, remote sensing methods, particularly satellite image analysis, offer new opportunities for timely detection, impact assessment, and fire forecasting. Satellite data makes it possible to cover large areas, quickly (often in real time) identify thermal anomalies, analyse potential burned areas, and track wildfire spread dynamics. The widespread use of geographic information systems (GIS) and technologies opens up new possibilities for processing, analyzing, and visualizing a wide range of wildfire data. This integration of diverse information provides valuable support for decision-making. All of the above highlights the importance of studying wildfires using GIS and remote sensing, especially given the limited number of such studies in Ukraine.
The objective of this work is to analyze the potential of remote sensing data and various web services for wildfire analysis using GIS technologies, with a focus on the Volyn region of Ukraine.
The main tasks of the study are: to find the needed satellite images using web services such as Copernicus Browser and Earth Explorer; to analyze the image bands from Sentinel-2 (European Space Agency) and Landsat 8 (NASA) satellites for fire detection and analysis; to calculate the
Normalized Burn Ratio (NBR) and delta Normalized Burn Ratio (dNBR) for calculating the fire impact; to identify the advantages and limitations of Sentinel-2 and Landsat 8 imagery in analyzing wildfire consequences, using the example of the wildfire near the village of Birky, Kamin-Kashyrskyi district, Volyn region, Ukraine.
In modern scientific literature, there are numerous publications dedicated to this issue. For
instance, F. Sivrikaya et al., based on their analysis of Mediterranean forest areas in Turkey (in the
Yeşilova Forestry Enterprise, Kahramanmaraş), described the potential of GIS for analyzing and
assessing forest fire risk while accounting for factors such as tree species composition, forest cover
percentage, stand age, slope steepness, aspect, and distance from settlements and roads. They also
analyzed the visibility of fire watchtowers used for forest monitoring in the study area [
R. Jaiswal et al., using ArcGIS software and forest data from the Gorna Subwatershed (Madhya
Pradesh, India), developed a GIS-based model to assess wildfire risk across the study region. A
color composite image from the Indian Remote Sensing Satellite (IRS) 1D LISS III was used for
vegetation mapping. Based on the dataset, four wildfire-risk zones were identified. The
recommended GIS model appeared to be very useful with the actual fire-affected areas [
E. Chuvieco and J. Salas recommended mapping fires in Central Spain using GIS and comparing
wildfire maps with topographic, meteorological, vegetation, and human activity maps. Three maps
were created-probability of ignition, fuel hazard, and human risk-which were integrated into a
comprehensive fire danger map based on Spanish Forest Service criteria. This approach improved
understanding of spatial fire distribution, which has critical results for developing regional fire
protection plans [
A. Supriadi and T. Oswari proposed a GIS-based web application for the fire department of
Depok City (Indonesia). The system aimed to speed up fire report processing, present spatial and
non-spatial data, update wildfire records, and assist in locating new fire stations, hydrants,
wildfireprone zones, and ignition hotspots [
H. Adab et al. studied fires in northeastern Iran using extreme temperature data from the
MODIS satellite. As a result, they calculated several indexes-Structural Wildfire Index, Wildfire
Risk Index, and Hybrid Wildfire Index-for monitoring and minimizing wildfire occurrence and
related damage. Key wildfire-contributing factors included proximity to settlements and roads,
slope steepness and aspect, elevation, and vegetation moisture. All of the listed previously factors
were incorporated into a GIS model [
In Ukraine, some attention has also been given to studying wildfire using GIS and remote
sensing. V. Zatserkovnyi and others examined the use of satellite imagery for forest wildfire
monitoring through remote sensing and highlighted the advantages of this data for detecting
wildfire and assessing their impact on ecosystems. A range of morphometric analyses and index
calculations were conducted, with image classification performed using the Maximum Likelihood
method [
Researchers from Lviv Polytechnic National University studied wildfires in the Chornobyl
Exclusion Zone in 2020 using Sentinel-2 satellite data [
O. Borysenko and V. Meshkova addressed fire and pest outbreak prediction in pine forests using
GIS technologies in their monograph [
A study by O. Bandurko and O. Svynchuk focused on identifying wildfires using low-resolution
satellite images and a “fire pixel” detection algorithm for TERRA MODIS and NOAA AVHRR data
[
As seen from Ukrainian studies, most research focuses on wildfires in the Chornobyl Exclusion Zone, while other regions remain largely unexplored. Therefore, the study of wildfires in Ukraine using GIS tools is still underdeveloped and requires further investigation.
The source base of this study consists of freely available satellite images from the European Space
Agency (ESA) and NASA. We analyzed multispectral satellite images to detect and study wildfires.
This can be done through online platforms designed for working with satellite data, such as
Copernicus Browser (a free ESA resource) [
Copernicus Browser is a free online application that provides easy access to satellite imagery from Copernicus missions and combined datasets. Because of how easy it is to use and its intuitive interface allows any user to explore the planet easily using high-resolution satellite data. The platform provides ready-to-use imagery, preset visualizations, and thematic data collections. Copernicus Browser allows visualization of Sentinel-2 imagery in several modes, such as: True Color (bands B4, B3, B2); False Color (bands B8, B4, B3); False Color (Urban) (bands B12, B11, B4); Highlight Optimized Natural Color.
Various spectral indices, including: Normalized Difference Vegetation Index (NDVI) = (B8 – B4) / (B8 + B4); Moisture Index = (B8A – B11) / (B8A + B11); Normalized Difference Water Index (NDWI) = (B3 – B8) / (B3 + B8); Normalized Difference Snow Index (NDSI) = (B3 – B11) / (B3 + B11); Scene Classification Map (Sentinel-2 data processed with the ESA algorithm).
Each band represents a specific range of the electromagnetic spectrum, and the satellite sensor
captures Earth’s surface in these spectral ranges. Users can also define custom band combinations
or scripts, create time-lapse animations (e.g., for vegetation dynamics, land use, or urban growth),
perform basic measurements and calculations, and export results in multiple formats [
The USGS Earth Explorer data portal provides access to geospatial datasets from the U.S. Geological Survey. Users can search by location or coordinates to obtain Landsat satellite imagery, radar data, UAS data, digital elevation models, aerial photos, and other geospatial datasets. Earth Explorer supports visualization of Landsat 8–9 imagery in multiple modes, including: Reflective Color (bands 6, 5, 4); Thermal Browse (band 10); Quality Browse; Natural Color (bands 4, 3, 2); Color Infrared (CIR) (bands 5, 4, 3); False Color (Urban) (bands 7, 6, 5); False Color (Vegetation Analysis) (bands 6, 5, 4); Near Infrared (NIR) (band 5); Normalized Burn Ratio (NBR); Normalized Difference Moisture Index (NDMI); Normalized Difference SnowIndex (NDSI); Normalized Difference Vegetation Index (NDVI); Soil Adjusted Vegetation Index (SAVI); Thermal Band Average (bands 10, 11); Thermal (band 11).
Comparing Copernicus Browser and Earth Explorer, their basic visualization tools are similar. However, Copernicus Browser provides greater flexibility for preliminary data analysis directly within the platform.
To assess wildfire impact and burn severity, the Normalized Burn Ratio (NBR) is used, calculated as:
NBR= NIR−SWIR ,
NIR + SWIR
(1)
where: NBR – Normalized Burn Ratio; NIR – Near-infrared reflectance (Sentinel-2: Band 8;
Landsat 8–9: Band 5); SWIR – Shortwave infrared reflectance (Sentinel-2: Band 12; Landsat 8–9:
Band 7) [
To evaluate burn severity, the difference between pre-wildfire and post-wildfire NBR, called delta NBR (dNBR), is calculated: dNBR= prefire NBR− postfire NBR , (2) where: dNBR – delta Normalized Burn Ratio; prefire NBR – NBR before the fire; postfire NBR – NBR after the fire.
Higher dNBR values indicate more severe burn damage, while negative values may suggest
vegetation regrowth after the wildfire [
Burn severity classification was created using the scale proposed by the United States Geological Survey (USGS) (Table 1).
Our area of interest for this study is a wildfire in northwestern Ukraine [
The previous location of the wildfire (approximate boundary determination) and its duration
were established using the FIRMS (Fire Information for Resource Management System) web service
(Fig. 1) [
To detect and analyze wildfires in the Copernicus Browser service, certain bands or
combinations of bands are used. For example, Sentinel-2 satellite imagery and the band
combination B12, B8, B2 [
In Earth Explorer, fires are best seen using Thermal Browse (band 10), Thermal Band Average (bands 10, 11), and Thermal (band 11).
The next step is to download the satellite images. For dNBR calculations, Sentinel-2 L2A images (via Copernicus Browser) and Landsat 8-9 OLI/TIRS C2 L1 images (via Earth Explorer) were downloaded. Images were selected for both pre- and post-wildfire periods. Due to several factors-capture frequency, time of day, weather (especially cloudiness), and partial image commercialization-this task was not straightforward. The pre-wildfire images were of good quality and available for April 27, 2025 (Sentinel-2), and April 21, 2025 (Landsat 9). However, because of cloudy conditions, the first high-quality post-wildfire image was available only on June 4, 2025 (Sentinel-2) and June 16, 2025 (Landsat 8–9). This delay means that vegetation in the area (mostly wet floodplain meadows of the Pripyat River) could have already recovered. Therefore, for further analysis, we used partially cloudy images from May 7, 2025, available for both Sentinel-2 and Landsat 9 (Figs. 3, 4).
The Sentinel-2 image is of higher quality than the Landsat 9 one. This is due to Sentinel-2’s better spatial resolution (20 m per pixel vs. 30 m per pixel) and newer technology (launched 2015–2017, compared to 2013 for Landsat 9). Around 10–15% of the area was covered by clouds, reducing the precision of further analysis.
Next, we determined the exact wildfire boundaries. The images were processed in QGIS 3.16.16-Hannover. A vector layer was created, and based on different image types/band combinations (True color, False color, SWIR for Sentinel-2; Reflective Color, False Color, NBR for Landsat 8–9), the wildfire perimeter was delineated. Due to limited image availability and cloud cover, images from several dates were used.
To refine the boundary, the “Fire Boundary Script” [
Burned area sizes by damage class (Table 1) were computed in QGIS 3.16.16-Hannover using the “Raster layer unique values report” plugin.
Finally, visualization of results was done using the USGS (United States Geological Survey) classification scale for interpreting burn severity (Table 1). The threshold for wildfire-affected vegetation was set at dNBR = 100. Based on this threshold, the wildfire perimeter was delineated (Figs. 6–7).
To present the researched results, we visualized the data in QGIS 3.16.16-Hannover. For interpretation of burn severity, we applied the United States Geological Survey (USGS) classification scale (Table 1) with subsequent division into classes. The lower threshold for vegetation affected by fire was set at dNBR = 100. Based on this threshold, the wildfire boundaries were delineated.
As a result, we obtained the following outputs shown in Figures 6 and 7.
Using the “Raster layer unique values report” plugin in QGIS 3.16.16-Hannover, we calculated the burned area sizes according to the USGS classification scale (Table 2 and 3). Severity level / dNBR %
Low severity (100–269) Moderate-low severity (270–439) Moderate-high severity (440–659)
High severity (660–745)
TOTAL where * – calculated by authors. where * – calculated by authors.
8,71 12,11 3,87 0,03 34,11 2 17,29 16,81 0,01 – – 34,11 25,5 35,5 11,3 0,1 100 % 50,7 49,3 0 – – 100 3,99 1,88 0,74 0,02 34,11 2 33,74 0,37 – – – 34,11 11,7 5,5 2,2 0,1 100 % 98,9 1,1 – – – 100 The data obtained show that the wildfire near Birky village, Kamin-Kashyrskyi District, Volyn Region, Ukraine, which lasted from April 30 to May 3, 2025, covered an area of 34.11 km². Despite its massive area, the wildfire’s impact was not severe due to the dominance of meadow floodplain vegetation, which is resilient and regenerates quickly.
A major limitation of studies of this kind is access to suitable and relevant satellite imagery. Because the region is frequently cloudy, cloud-free images are often unavailable. As a result, data from multiple satellites must be combined, which complicates interpretation.
Visual analysis of the results shows the presence of “unburned” patches within the wildfire zone
on May 7, 2025 (Figs. 6–7). These areas correspond to cloud shadows (Figs. 3–4). Additionally,
lakes, open water, and unburned fragments within the burned area distort the overall dNBR
distribution. In such cases, dNBR values are inaccurate but do not affect general conclusions. This
observation aligns with Bandurka and Svynchuk (2022), who emphasized excluding cloud-covered
and water-covered fragments during image interpretation [
The highest wildfire intensity occurred in the northern sector near Hirky village, the southern part near Birky, and the western part near Vetly. The northern area, covered with forest, showed the strongest and longest-lasting burn effects. On Sentinel-2 imagery from June 4, 2025, and Landsat 8 imagery from June 16, 2025, fire traces (dNBR > 100) remained visible only in this northern zone (Figs. 6–7).
The wildfire was naturally limited by water bodies – the Prypiat River and nearby drainage channels. Since the study area consists mostly of floodplain meadows and pastures, vegetation damage was minor. Firefighters contained the spread near the villages of Hirky, Birky, Vetly, and Liubotyn.
The dNBR values vary depending on vegetation type, landscape, and local conditions. For more
accurate interpretation, field validation is recommended where possible. The dNBR classification
ranges are not rigid; threshold shifts of about ±100 points are acceptable. Outliers below -550 or
above +1350 may occur in unburned areas due to recognition errors, clouds, or other factors.
Settlements also distort results and should be excluded from analysis [
Comparison of results from different satellites shows that Sentinel-2 L2A provides more accurate data than Landsat 8–9 OLI/TIRS C2 L1. This is due to its higher spatial resolution (20 m vs. 30 m per pixel) and newer, more sensitive sensor technology. However, since the wildfire had limited ecological impact and its effects largely disappeared within a month, Landsat 8-9 data are sufficiently reliable for most analytical tasks.
Regarding result precision, satellite-based remote sensing cannot be considered high-accuracy.
According to Babushka et al. (2021), the deviation in dNBR-based burned area estimation ranges
from 6.7% to 11.5% depending on methodology [
Our results indicate that within the affected area, low-severity burns dominate: 49.3% according to Landsat 8-9 data; 61.0% (low and moderate-low severity combined) according to Sentinel-2 data. These findings are consistent with Babushka et al. (2021). Overall, the obtained results are sufficiently accurate and suitable for this type of environmental analysis.
Thus, remote sensing data processed and analyzed using geographic information systems and technologies are of high importance and great potential for monitoring and preventing adverse natural events such as wildfires. Several online platforms provide near-real-time wildfire data (e.g., FIRMS), which help detect fire occurrences. However, much broader analytical opportunities arise when working directly with satellite imagery.
The most promising approach involves analyzing bands from the near-infrared (NIR) and shortwave infrared (SWIR) ranges, which enable both the detection of active wildfire and the assessment of their long-term environmental effects. Among the numerous available platforms, the Copernicus Browser (European Space Agency) and Earth Explorer (USGS, NASA) offer the most valuable open-access datasets. These were the sources from which we obtained Sentinel-2 L2A and Landsat 8–9 OLI/TIRS C2 L1 imagery for our analysis.
The use of GIS software provides powerful tools for both analysis and visualization. In this study, we used QGIS 3.16.16-Hannover**. The calculation of the **dNBR (Differenced Normalized Burn Ratio)** enabled the evaluation of the long-term environmental impact of the wildfire.
Satellite-derived burn severity maps have strong practical applications. They can support the development of emergency rehabilitation plans and guide vegetation recovery efforts. Furthermore, such data can be used not only to assess burn intensity but also to predict potential secondary effects on burned areas, such as flooding, landslides, or soil erosion.
One of the main challenges in studies of this type is obtaining suitable and cloud-free imagery. Cloud shadows remain a significant obstacle for analysis, especially in regions with frequent cloud cover. In our case, only one fully usable and one conditionally usable image were available within a month after the wildfire, both for Sentinel-2 and Landsat 8–9. This highlights the need for improved methods to work with partially usable imagery and to integrate data from multiple satellite systems within a single analysis.
Further research on wildfires across Ukraine using remote sensing data is urgently needed. Such studies should aim to identify regional patterns, refine (calibrate) the USGS burn severity scale for different landscapes and vegetation types, and evaluate the accuracy of burn severity analysis across various satellite platforms. The authors have not employed any Generative AI tools.