Oct 7
Before the Flames Arrive: Satellite Technologies That Provide Fire Warning
- Gülüzar Yiğit
- Oct 7
- 15 min read
We always talk about forest fires afterward: how many hectares were burned, how many days did it last, how many people intervened? But the real question often lingers: Could we have foreseen this? In this article, we look not at the past, but at the possibilities. Thanks to two recent studies based on satellite data, we'll examine what nature tells us about where and when fires might break out. So, it's a little science, a little data, and a little effort to understand nature. It's not too late, but we may need to raise our sights a little higher.
Contents
Can Fires Really Be Predicted with Satellite Technologies?
Every summer, we burn more and more. As temperatures rise, high winds make land more vulnerable to fire. As a result, we delay intervention, and nature loses. But this is no longer a surprise. According to Ertuğrul and Mertol's (2019) climate projections, Turkey will be "prepared" for longer fire seasons and more risky areas by 2050. Of course, this preparation is for nature, not for us.
There are ample reasons to believe that wildfires are no coincidence: rising temperatures, human intervention, and decreasing rainfall. In drought-prone areas, fire seasons are being brought forward, and the duration of fires is lengthening. In other words, nature is saying, "I could catch fire soon," but most of the time, we don't hear it.
High temperatures, low humidity, harsh topography, and winds, combined with the combination of these factors, can quickly cause fires to spread out of control, particularly in the Mediterranean region. Add to this the human impacts of roadsides, fields, and residential areas, and the situation escalates even further.
But when you look at this picture from above, things change. With remote sensing and satellite technologies, it's now possible to get information about risky areas before a fire breaks out. In other words, satellites don't just record what's already happened; with some data, they can also provide insights into what's to come.
Indeed, a study conducted in Alberta, Canada, found that forest fires peak during the short window of time when the snow has melted but the vegetation hasn't yet sprouted. The soil is bare, humidity is low, and the environment is ripe. Before nature is even awake, the fires are already well underway.
Fire News Comes from Satellite
Every time Sentinel-2 visits a fire zone, it downloads a kind of "update" on the forest's health. Is that plant still green, or has it begun to dry out and turn yellow? It's all recorded in the spectral bands. NDVI (Normalized Vegetation Index) deciphers this language: If it's healthy, the index is high; if it's drying, the index is low.
But it's not just about the shade of the foliage. If the surface temperature (LST) has risen and soil moisture has decreased, the forested area has quietly entered red alert. Add to that the proximity of a roadside or residential area—bingo: all the conditions for a fire are on the table. Such temperature data, such as in this study, are taken from MODIS's MOD11A1 product and calculated by NASA using a split-window algorithm. In other words, the temperature data is not field measurements, but derived from satellite thermal bands and used as a ready-made product.
These parameters provide signals before a fire breaks out, not after. Is there a sudden drop in NDVI? Has LST increased? Has the soil dried out? In the Alberta example, the NDVI time series was decomposed using the STL (Seasonal-Trend Decomposition Using Loess) method to reveal seasonal trends. Additionally, the phenocam data was smoothed with a double logistic curve. This eliminated short-term changes and revealed more reliable patterns. These are all moments when nature silently cries out, "I'm here."
Satellites can hear this cry. Because remote sensing is not just an observer documenting what's happening; it's an early warning system that can anticipate what's about to happen. If accurate data is interpreted correctly, maps can reveal the future.
How to Make a Fire Map: An Example from Golestan National Park
Creating a fire risk map is like asking nature, "How easily do you burn right now?" A team in Golestan National Park sought to answer this question and compiled 10 different parameters in 2024:
They built a model that listens to both nature's inner voice and human footprints, including NDVI, surface temperature, precipitation, wind, land cover, roads, and settlements. The park is located in northern Iran, on the border of the Hirkani forests and semi-arid steppes. In other words, it's a land on the brink of both greenery and fire.
The data were analyzed using AHP and fuzzy logic (Fuzzy Logic) methods. AHP matrices were constructed based on expert opinions, and then the consistency ratio (CR) was calculated. The CR value reported in the study was 0.08, below the acceptable limit of 0.1. This demonstrates that the weighting method was methodically consistent. The result? Approximately 24% of the park is classified as "high" or "very high" fire risk. And most of these areas, as you might expect, are concentrated in the areas closest to human settlements and roads.
The Golestan National Park fire risk map is a multi-criteria fire risk map created by evaluating various indicators, including NDVI, temperature, precipitation, and settlement. Extremely high risk is particularly concentrated along the southern and western borders; these areas also overlap with past fire hotspots.
The strength of this approach is its ability to spatially assess numerous parameters. But let's not forget: The model's accuracy will depend on the resolution and timeliness of the data used. MODIS NDVI data has a resolution of 250 m. In the Alberta study, this resolution was deemed sufficient for analyzing large-area fire trends, but it was supplemented by Phenocam data for micro-scale predictions. If the satellite is late, the fire will arrive early. Therefore, while MODIS is deemed sufficient for wide-area analyses, higher-resolution data or ground observations should be preferred for situations requiring ground-level precision.
It's clear that the actual fires and the model's predicted high-risk areas match up almost perfectly. The data in Figure 3 demonstrates that satellites provide not only predictions but also accurate forecasts. The majority of the fires occurred in areas previously classified as "very high risk."
Canada's 10-Day Fire Window
Alberta, Canada. The year is 2024. The MODIS satellite is looking down from the sky, phenocams from the ground. They're recording plant greenup times between 2000 and 2022. The goal is simple: When does nature wake up, when do the fires wake up?
The picture that emerges is quite striking: Fires peak on average 10 days before vegetation greens up. In other words, while the NDVI hasn't yet budged and the soil is still bare and moisture-free, the flames are waiting for an opportunity.
The developed model can predict greening dates with 76% accuracy. However, the study doesn't include many factors that shape fires, such as topography, wind, and socioeconomic factors. Therefore, the model doesn't predict where fires will break out; it simply suggests when to be vigilant.
Can We Be Prepared for Fire?
The Golestan study asks, “Where is the likelihood of fire most likely?” while the Alberta study asks, “When should we be on alert?” Combined, we have not just a map, but a timeline.
What is the use of this?
Local governments can identify risk areas and critical periods in advance.
Response teams can plan resources more accurately.
Civilians can be warned early, without saying "let the summer come and we'll see."
So, nature sends a signal, technology picks it up. The rest is up to us: to answer the question, “When will we take action?”
Fire Warning Satellite Technologies and Daily Life Suggestions
Parameters such as NDVI, LST, and soil moisture used in pre-fire risk assessment are currently provided periodically by numerous satellites. Integrating this data into practical systems beyond academic publications is crucial for strengthening early warning mechanisms. Neither study utilized GNSS-based atmospheric corrections or precise location data. However, integrating GNSS-supported meteorological observations into the model could improve the spatial accuracy of forecasts, particularly in regions where topography affects fire risk.
First, local governments can establish automated analysis systems that interpret satellite data. These systems can be designed to alert relevant authorities when levels fall below or exceed certain thresholds. The data used can be sourced from open-access platforms such as ESA's Sentinel-2, NASA's MODIS, or VIIRS.
Fire risk maps should be detailed not only at the forestry administration level, but also at the district and neighborhood level. These maps can be used by decision-makers and field teams in pre-intervention resource planning. Local support can be obtained with high-resolution imagery or unmanned aerial vehicles (UAVs) to increase spatial resolution.
Early Fire System Advantages
Early warning systems can be enhanced by integrating with meteorological and topographic data. For example, combining daily data such as temperature, wind speed, and direction with NDVI and LST analyses improves the accuracy of models predicting short-term fire potential.
Disseminating data is also crucial. Risk maps should be shared not only with administrative agencies but also with stakeholders such as irrigation associations, agricultural cooperatives, and rural municipalities. These institutions are often the first to observe or become aware of fire outbreaks in areas where they occur.
Finally, transforming models developed by research centers and universities into decision support systems with open-source interfaces contributes to the wider use of information. Such models should be updated and optimized based on feedback from the field, both for accuracy and usability.
Monitoring Post-Fire Recovery with Satellites
Fires wipe out a large portion of wildlife, drastically reducing populations. But in the wake of this destruction, nature makes room for new players. Some creatures see this environment as an opportunity. There are no longer any rivals to defend their territories or predators to hunt them. This is where resilient microorganisms and plants come into play.
The first arrivals, carried by wind-borne spores and seeds, colonize these burned areas. This process is known in the scientific community as ecological succession . First come herbaceous plants, then shrubs, and then, over the years, tree seedlings. Over time, small animals like mice, insects, and birds return to the area.
So, as you can see: We can't defeat nature. From microorganisms to their top predators, no land, even burned, remains empty. Nature continues to heal, and the ecological order is being reestablished.
How to Determine Post-Fire Recovery?
Thanks to remote sensing technologies, satellites not only "see" the Earth; they also use wavelengths of light to measure the condition of soil, water, and plants. So, even if nature doesn't yet appear to have recovered to the naked eye, we can tell it's beginning to recover.
How Can Vegetation Recovery Be Monitored with Satellites?
Pre- and post-fire imagery, such as optical satellite imagery like Sentinel-2 and Landsat-8, are the most widely used resources worldwide for monitoring post-fire landscape changes. Their high spatial and spectral resolution allows us to precisely track burn scars, severity levels, and regrowth over time.
Satellites and remote sensing technologies are used to monitor vegetation recovery after wildfires using specific metrics and applications. NBR (Normalized Burn Ratio) is used to determine the burned area. Then, dNBR (difference NBR) reveals the burn intensity. Indices such as NDVI can then be used to monitor the rate of vegetation return, the extent of recovery, and the level of vigour. In other words, all these analyses can be performed immediately after a fire, even before foliage is visible.
Even though the trees aren't growing, we hear them saying, "I'm coming." These invisible signals of recovery are filtered by satellites and transformed into statistics, maps, and reports. Thanks to remote sensing, we can feel the pulse of nature even before it fully recovers.
An Example of Post-Fire Recovery: The Yellowstone Fire
The dNBR map, created with Landsat imagery after the 1988 Yellowstone fire, visualized fire severity in five different classes: no burning, low burning, moderate burning, high burning, and regrowth class.
These classifications quantify where the land has sustained the most damage. Such analyses aren't limited to just a few days after a fire; we engineers track how nature repairs itself, and how quickly which areas recover, using satellite data collected over years.
In the summer of 1988, lightning strikes, drought, and human-caused factors devastated more than one-third of Yellowstone National Park (approximately 793,000 acres) by a massive wildfire. Landsat satellites tracked the fire's effects not just that year but also step by step for 32 years.
In the 1987 images, dark green areas represented healthy pine forests, while light green and yellow areas represented meadows and plains.
In images taken in 1988, these areas are replaced by scorched, blackened ground. The reddish tones clearly highlight the fire-affected areas. In some areas, active fires are even visible as bright pink.
The 1989 image reveals the full impact of the fire. The images show that the fire did not create a homogeneous devastation, but burned at different intensities in different areas. This is one of the most valuable insights remote sensing offers: the ability to measure differences in intensity and recovery within a visually consistent framework.
As time progresses, particularly in images taken in years 5, 10, and 17, the burn scars appear to fade and new vegetation emerges. Initially, herbaceous plants and wildflowers, followed by small pine seedlings, begin to grow. Because the new trees are young, they photosynthesize more rapidly and produce spectrally different reflectances. This manifests as detectable recovery signals in the NDVI and SWIR bands.
But this recovery is not easy: due to the high-altitude plateaus, short growing seasons, and harsh winters, recovery takes decades. Indeed, even in 2019, satellites can still see traces of the fires.
The three false-color Landsat images below show the effects of the devastating 1988 wildfire and the natural recovery process over the years. Each image is created by combining Landsat's SWIR (shortwave infrared), NIR (near infrared), and visible bands. This combination of bands allows us to see changes in vegetation much more clearly.
From left to right:
1993 : 5 years after the fire. Reddish tones are widespread, representing areas where vegetation has not yet recovered and the soil surface remains exposed.
1998 : 10 years after the fire. Green tones are beginning to return in some areas, indicating the re-introduction of grasses and small shrubs to the area.
2005 : 17 years after the fire. The area is now largely green, indicating that the forest ecosystem has largely recovered and that slow-growing conifers are beginning to return.
Real-Time Fire Monitoring: Free Data Sources for Türkiye
Large fires can periodically break out in dozens of forest and shrubland areas in Turkey. These fires threaten human life and ecosystems, and real-time data is crucial for rapid response and situation assessment. Fortunately, organizations like NASA and ESA make satellite data publicly available, enabling active fire monitoring on a global scale.
For example, NASA's Worldview site performed emergency mapping of the fires in Izmir in July 2025, using a FIRMS layer of red dots on a VIIRS satellite image. These free tools provide vital information for surveyors and related professionals. Below, we'll primarily cover the NASA FIRMS and Sentinel Hub EO Browser platforms, as well as local volunteer mapping and technical tips.
NASA FIRMS – Active Fire Data (MODIS & VIIRS)
NASA's FIRMS (Fire Information for Resource Management System) service provides active fire and thermal anomalies from the MODIS sensors on the Aqua/Terra satellites and the VIIRS sensors on the Suomi NPP/NOAA satellites. The data is typically updated with a 3-hour delay (NRT). FIRMS allows you to track the location, intensity (perceived brightness), and timing of fires worldwide. For example, you can receive email alerts to notify you of new fires in a specific region, or you can download the data as SHP/KML and use it in your own GIS projects.
Steps for Viewing and Downloading Data with FIRMS
Fire Visualization on the FIRMS Map: You can directly view current fires in Turkey or worldwide by following this map link via NASA's FIRMS (Fire Information for Resource Management System) platform. Active fire hotspots are shown on the map based on thermal anomalies from the MODIS and VIIRS satellites. While there is a lag of approximately three hours, the data is quite current and is updated frequently.
FIRMS Data Download: If you want to download and analyze fire data, you can use the Download Active Fire Data page of the FIRMS platform. On this page:
By selecting the last 24 hours, the last 7 days, or a specific date range
In formats such as SHP, KML, CSV or TXT
You can download data from MODIS or VIIRS. This data can then be imported into desktop software like QGIS or ArcGIS.
Tip: You can perform spatiotemporal analyses over Turkey by adding downloaded SHP or KML data to QGIS. Point symbols are a particularly useful visualization method for fire intensity.
With these steps, you can quickly view fires and download historical data for analysis via FIRMS. The FIRMS API can also be used with Python/R for large-scale investigations or automation.
High Resolution Fire Monitoring with Sentinel Hub EO Browser
The ESA-supported Sentinel Hub EO Browser allows for real-time viewing and analysis of various satellite images, particularly Sentinel-2. Sentinel-2 data is updated at 10-meter resolution (every few days), making it particularly useful for pre- and post-fire investigations. In the EO Browser, you can easily use band combinations such as True Color and False Color (SWIR/NIR/Red) , or predefined analyses such as NDVI/NBR .
EO Browser Usage Steps:
Selecting the Relevant Location: Open EO Browser . Select a region in Turkey on the map from the top left (e.g., mark the province or forest area where the fire broke out).
Selecting the Dataset: From the right-hand menu, go to the Data section. Select the Sentinel-2 L2A (atmosphere-corrected) data collection. In the Date filter, select the days following the fire (updates will occur every 2–5 days until new imagery becomes available).
Visualization: From the Layers / Bands section, you can select analysis modes such as NDVI or NBR . With NDVI ((B8-B4)/(B8+B4)), vibrant green vegetation is highlighted; vegetation in burned areas has a low value for this index. NBR is calculated using Band 8 (NIR) and Band 12 (SWIR) (NBR=(B8−B12)/(B8+B12)), and burned areas are highlighted with low NBR values. If pre-built indices are unavailable, you can perform similar calculations by adding the B8 and B12 bands to the formula in the Custom section.
Time Series and Comparison: Create a time series or side-by-side comparison between two dates. For example, view pre- and post-fire imagery sequentially to observe changes in vegetation in affected areas. You can also create animations across multiple dates with time sliders.
EO Browser, with its user-friendly interface, allows you to quickly perform satellite analyses of fire zones. Furthermore, with the false-color SWIR band combination (e.g., B12, B8A, B4), healthy vegetation appears green, burnt soil appears brown, and bare soil appears brown. You too can easily perform your fire zone analyses with this tool.
Local Initiatives and Open Maps
Volunteers sometimes create provisional fire maps on platforms like OpenStreetMap (OSM) or MapHub. For example, during fire seasons, communities may incorporate satellite data or news reports they receive into OSM maps. Because these maps are temporary and voluntary , they should be used with caution. Unofficial data sources may be outdated or processed incorrectly. Therefore, always compare your data with official announcements from official institutions such as the General Directorate of Forestry and AFAD (Disaster and Emergency Management Presidency) .
Local OSM maps may have fire boundaries or hotspots drawn on them. You can find publicly available maps shared through Google My Maps.
Some NGOs or university groups are producing interactive maps augmented with additional data layers (e.g., MODIS/VIIRS), but these are generally provisional.
For official and approved information, the most reliable way would be to follow the geographic data of the Ministry of Agriculture and Forestry, the fire instant monitoring service (if available) or the 112 Emergency call center.
These initiatives provide useful insights, but reliability is important: Be wary of maps whose source you do not know.
Technical Tips and Suggestions
Bands Used: The infrared (NIR) and short-wave infrared (SWIR) bands are invaluable in fire monitoring. For example, in Sentinel-2 imagery, Bands 8 (NIR) and 12 (SWIR) are used to distinguish between burning and unburned areas. The NBR index is calculated using these bands.
NDVI Formülü Indices (NDVI/NBR): The NDVI formula (B8-B4)/(B8+B4) measures the presence of green vegetation. A high NDVI value indicates dense vegetation, while a low value indicates open land/burned areas. NBR highlights burned areas with a low value. Post-fire dNBR (Delta NBR) can be calculated to determine the severity of the burned area.
NBR Formülü GIS Tools: You can import satellite layers into your maps by establishing WMS/WCS connections with software like QGIS. For example, you can add FIRMS WMS using the steps described above and view real-time fire hotspots in QGIS. You can also download Sentinel-2 imagery from the ESA Copernicus Hub and process it with the Raster Calculator in QGIS to create NDVI/NBR maps.
Automation & API: Advanced users can set up automation using Python or Google Earth Engine (GEE). You can receive instant alerts for specific regions by activating NASA FIRMS's email alert service. Sentinel data can be accessed via scripts using ESA's APIs or the Sentinel Hub API. This saves time on big data analysis and reporting.
These technical approaches help you effectively use fire data in engineering projects. Rapid access to accurate data is crucial during critical disasters like fires; the tools and methods in this article aim to empower everyone, from surveyors to enthusiasts, to participate in this process with technical expertise and awareness.
Resources
Bitek, Deniz & Ocakbaşı, Emrullah & Erenoğlu, Cüneyt. (2022). Forest Fire Analysis with Sentinel-2 Satellites: The 2018 Greece Mati Example. Academic Perspective Procedia. 5. 210-219. 10.33793/acperpro.05.02.5941.
Polat, N., & Kaya, Y. (2021). Damage Assessment in Forest Fires Using Multispectral Satellite Imagery. Bartın Faculty of Forestry Journal, 23(1), 172-181. https://doi.org/10.24011/barofd.837507
Ertuğrul, Mertol. (2019). Future Forest Fire Danger Projections Using Global Circulation Models (GCM) in Turkey. Fresenius Environmental Bulletin. 28. 3261-3269.
Gunay Ercan, Delfin & UYGUNOL, Mehmet. (2023). Detection of Burned Areas by Remote Sensing and Comparison with Land Cover Classes, Muğla/Marmaris Example. 10.15659/uzalcbs2022.12754.
Zahra Parvar, Sepideh Saeidi, Seyedhamed Mirkarimi,Integrating meteorological and geospatial data for forest fire risk assessment,Journal of Environmental Management,Volume 358,2024,120925,ISSN 0301-4797, https://doi.org/10.1016/j.jenvman.2024.120925.
Pickell , PD, Coops, NC, Ferster, CJ et al. An early warning system to forecast the close of the spring burning window from satellite-observed greenness. SciRep 7 , 14190 (2017). https://doi.org/10.1038/s41598-017-14730-0
Çiçekli, SY (2024). Investigation of Burnt Forest Areas with Remote Sensing Techniques: The Kozan Forest Fire Example. Çukurova University, Journal of Engineering Faculty, 39(2), 503-514.
Christine Conte. How Ecosystems Recover After a Fire? https://earth.org/how-ecosystems-recover-after-a-fire/
Franks, S., Masek, J. G. & Turner, M. G. (2013). Monitoring forest regrowth following large scale fire using satellite data – A case study of Yellowstone National Park, USA. European Journal of Remote Sensing , 46:1, 551-569, DOI: 10.5721/ EuJRS20134632. https://doi.org/10.5721/EuJRS20134632
NASA Scientific Visualization Studio. (2012, July 23). Yellowstone Burn Recovery . NASA. https://svs.gsfc.nasa.gov/11029
NASA Earth Observatory. (n.d.). World of Change: Burn Recovery in Yellowstone . NASA. https://earthobservatory.nasa.gov/world-of-change/Yellowstone/show-all
Comments