The Danube Delta represents a dynamic and vulnerable ecosystem, where monitoring changing processes requires modern analysis tools. This material presents the paradigm shift in monitoring the Delta, illustrating the potential of Earth Observation (EO) data in comparison to classic topographic maps. The study analyzes the vegetation fires of February–March 2025, examined through Sentinel-2, Sentinel-3, MODIS/VIIRS data, burnt area products (EFIS), and Landsat 9 images. The integration of these sources allowed for a detailed observation of the spatial and temporal dynamics of the fires in the first part of 2025, as well as the multi-annual dynamics, starting from the winter 2016–2017 season. Furthermore, the assessment of Sentinel-2 metadata from the 2015–2025 period shows the constant frequency of images with low cloud cover, allowing their applicability in periodic monitoring with a very high temporal frequency. All data were processed in an open-source environment and are freely available, the presented methodology being replicable in other environments. The results underline the importance of EO as operational support in the sustainable management and monitoring of the Danube Delta.

Copernicus, Danube Delta, earth observation, FIIRMS, fire, Landsat, satellite, Sentinel-2

Wetlands play an essential role in maintaining global ecological balance and health. They are important for biodiversity, water filtration, and play a significant role in the hydrological regime. The Danube Delta, as part of the UNESCO World Heritage and a Ramsar site, represents an area of major interest for research. The natural dynamics, as well as anthropic interventions in this area, have attracted researchers from various fields such as geography, biology, chemistry, anthropology, etc. In the current context of the dynamics of natural phenomena, anthropic augmented, constant and detailed monitoring of this area is considered a priority.

Although field research and monitoring in this area began to be methodically carried out as early as the 19^th century, the capacity to monitor all occurring changes in detail was limited. Difficult access to disparate map collections, as well as the low frequency of topographic surveys, represented the main impediment to detailed monitoring of the Danube Delta. In this paradigm, field observations offered a punctual and relatively fragmented image of the natural and human-induced changes that occurred in the deltaic space. Current studies have focused on how the monitoring of the Delta was carried out (Constantinescu et al. 2010; Tănăsescu and Constantinescu 2020; Constantinescu et al. 2024) in the 19^th century and the beginning of the 20^th century. However, this information was mostly inaccessible to researchers even in the early 2000s.

Starting from the early years of the 20^th century, a series of topographic surveys covered the delta area. These works, carried out at different scales by various institutions from Romania, offered an important starting point and a solid basis for research related to the evolution of the Danube Delta. However, it is noteworthy to mention that only 9 detailed maps were produced for the study area (between 1909 and 2019) – approximately one map every 13 years, with no cartographic product being published after 2019 (Table 1).

Among the presented materials, the most recent is the 1:50,000 topographic map published by ANCPI in 2019. Prior to this, the previous map product was 1:150,000 map of the Danube Delta published in 1983. Numerous changes and human interventions took place in the region during this 36-year interval.

With the launch of the Landsat program in 1972, the beginning of satellite observations opened new perspectives in the field of environmental monitoring.

Although Landsat images from the first decades were sporadic and had a low spatial resolution (30–80 m) (Goward et al. 2017), they represented a genuine improvement in analysis methods. In the case of the Danube Delta, data from the Landsat archive today allow the reconstruction of significant morphological and hydrological transformations, such as coastline evolution or river bank changes of the Sfântu Gheorghe branch. However, the lack of continuity observations from the pre-2000 period, as well as the difficulty in accessing and processing images in the years prior to the full opening of the archive in 2008 (Goward et al. 2017), limited the practical impact of this data.

Between 1972 and 2015, at the level of the study area, 917 available Landsat (EROS 2020) scenes were identified, and 675 of these were acquired before 2008. After this year, due to the parallel operation of Landsat 5, Landsat 7, and Landsat 8 satellites, a much larger number of images were acquired, reaching 51 scenes per year for 2014 and 2015 (Fig. 1). With the increase in observation frequency and the improved accessibility of satellite data, the capacity to conduct research in Danube Delta significantly improved recently. These factors have led to a major change in how research on the Danube Delta can be conducted.

Figure 1. 

Provides a visual overview of the satellite coverage of the Danube Delta using Landsat imagery, structured by year and by platform/sensor. It also includes two charts showing the number of scenes before and after the data was made publicly available, as well as an annual cumulative evolution graph of the total number of scenes.

The paradigm shift occurred with the emergence and consolidation of the European Union’s Copernicus program, especially through the launch of the Sentinel-2 missions (2015–present), which provide free optical data with a spatial resolution of up to 10 meters and an acquisition frequency of 5 days. Currently, the Sentinel-2 constellation numbers three satellites, the last two having been launched in 2017 and 2024. This constellation of three satellites offers unparalleled satellite image coverage of both land and water surfaces.

The accessibility, consistency, and quality of this data today allow for the almost continuous monitoring of phenomena such as: seasonal variations in water surface extent, vegetation health, the occurrence of vegetation fires, water quality parameters (Constantin et al. 2017; Constantin et al. 2024) or changes in the waterline or submerged sandbar (Tătui and Constantin 2020) in the maritime zone of the delta. Furthermore, the existence of dedicated platforms like the Copernicus Data Space Ecosystem (CDSE) democratizes access to satellite data, allowing specialists, as well as local institutions, NGOs, or even the general public, to quickly visualize and interpret the information. Besides the satellite component, the Copernicus Program includes 6 different types of services that offer pan-European data, freely and openly for 5 of them, including at the level of the Danube Delta, for various environmental parameters.

In parallel, data from other programs, such as National Aeronautics and Space Administration (NASA) Fire Information for Resource Management System (FIRMS), which integrates detections from Moderate Resolution Imaging Spectroradiometer (MODIS) and Visible Infrared Imaging Radiometer Suite (VIIRS) sensors, or the damage assessment services within Copernicus Emergency Management Service (CEMS), complete the picture regarding events with ecological impact. These services provide information on vegetation fires with daily updates, enabling researchers to monitor these events in near real-time.

In the case of the Danube Delta, numerous opportunities based on EO data are feasible, and within this material, a case study regarding vegetation fires from the period January – March 2025 was analyzed. The integration of these multiple EO data sources allows for the rapid identification of fire outbreaks and extent, and a clear understanding of their seasonality and persistence.

Besides recent satellite data sources, the importance of historical maps, available through open data initiatives such as geo-spatial.org, remains relevant in the long-term analysis of the deltaic landscape transformations. The comparison between historical maps and current satellite images offers valuable perspectives on channel dynamics, morphological diversity, and anthropic interventions over the last century.

The paper aims to highlight the added value that Earth Observation data brings to the understanding and monitoring of the Danube Delta, using a selection of examples from the recent period (2015–2025), rather than providing an exhaustive analysis, the objective of this work highlights the tremendous potential of such a wealth of data sources. By integrating Sentinel-2 images, Landsat data, FIRMS detections, and EFFIS products, a coherent and replicable monitoring approach is outlined, adaptable also to other wetland regions or protected areas in Romania and around the world.

The analysis presented in this paper aims to demonstrate the utility of Earth Observation (EO) data for monitoring vegetation fires in the Danube Delta. The proposed methodology involves integration of open access satellite data sources, and thematic information from dedicated platforms, combined through processing, and visualization in open-source environments (Fig. 2).

Figure 2. 

Illustrates the workflow and data processing steps leading to the final results. The cells on the left list the data sources and dataset names, the center presents the main software tools used, and the right highlights the type of obtained results.

Satellite data were preliminarily assessed using the Copernicus Data Space Ecosystem (CDSE), the structure that improves access and exploitation to EU’s Copernicus satellite data and contributing missions. Sentinel-2 L1C products were selected for the period 2015–2025, corresponding to tiles 35TPL, 35TQL, 35TQK, and 35TPK, which fully cover the Danube Delta. A filter was applied to select only scenes with cloud coverage less than 10% and 50% and metadata was used to assess the distribution of available images. The thresholds of 10% and 50% were chosen to illustrate the large volume of usable data, even though scenes exceeding 50% cloud cover may still contain partially usable information.

Images with less than 10% cloud cover are considered nearly ideal and can be used for a wide range of analyses. For Landsat images and their frequency analysis, the Earth Explorer platform was used, and metadata research was conducted for the two tiles covering the Danube Delta: Path 181, Rows 28 & 29, without considering cloud-cover filters.

For the visual detection of fires, standard Sentinel-2 natural color (bands 4,3,2) and shortwave infrared (SWIR) composite (bands 12, 8A, 4) were used. These specific band combinations highlight recent burns and the extent of affected areas. The Sentinel-2 images were acquired on February 20 and 25, 2025. Additionally, a natural color composite of a Landsat 9 image from February 20, 2025, was used to perform a visual analysis of fire events.

To contextualize the vegetation burning phenomena, data from NASA FIRMS services were integrated, providing daily updates with active fire detections obtained through MODIS (1 km resolution) and VIIRS (375 m resolution) sensors. This data was downloaded in shapefile format (.shp), covering detections for the winter-spring 2025 season. Additionally, data provided by the Sentinel-3 (Wooster et al. 2012) World Fire Atlas (S3-WFA) were also used, for detections starting from the winter season 2016–2017 (Sentinel-3A was launched on February 16, 2016). This information is available in vector format, point type. In addition, products from EFFIS (European Forest Fire Information System) (San-Miguel-Ayanz et al. 2012) were integrated, particularly the Burnt Area Rapid Damage Assessment data. These are automatically generated within the service based on MODIS and Sentinel-2 observations and provide delimitations in the form of polygon vectors of burnt areas, with a weekly update frequency and a minimum area affected of 30 ha.

An additional source was represented by the collection of historical maps available through standard web services WMS / WMTS from geo-spatial.org.

Within the study, the authors also used the visualization and analysis capabilities available through online platforms such as CDSE, Earth Explorer and FIRMS. Thus, the composites related to Sentinel-2 and Landsat 9 images were processed online. The obtained results were used without local post-processing, inherent to classic approaches.

After downloading the data, in various forms of processing, they were analyzed using open-source solutions such as QGIS or SAGA GIS (Conrad 2015).

Vegetation fires in the Danube Delta have become, in recent years, a rather frequent and more visible phenomenon during the cold season, winter – spring. In 2025, since mid-February and the first decade of March, the area between Letea and the Sfântu Gheorghe branch was characterized by intense burning activity. In this context, data obtained from Earth Observation sources offer a comprehensive and detailed picture of how these events can be monitored in near real-time, at spatial and temporal resolutions that were unavailable before the satellite era.

The preliminary analysis was performed based on Sentinel-2 data. Two key moments, February 20 and 25, 2025, provide a clear sequence of the fire’s evolution. In natural color composites (bands 4,3,2), the smoke plumes from February 25 are clearly visible, in contrast to the image from February 20, where only a few incipient signs of burning can be observed. (Fig. 3 below).

Figure 3. 

Shows the progression of wildfires in the Danube Delta between 20 and 25 February 2025. On the left are natural color images using bands 4, 3, 2; on the right is a SWIR composite using bands 12, 8A, 4. Both images were acquired by the Sentinel-2 satellite.

This evolution is even more evident in the SWIR composites (bands 12, 8A, 4), where pixels corresponding to the active burning areas are marked by intense red or orange tones, while the burned areas (scars) appear in brown tones, in contrast with the unaffected vegetation, rendered in shades of dark green. The compact shape of the fire, its south-eastern orientation, and the visible expansion in just five days indicate rapid burning, driven by wind.

In parallel, a complementary analysis was carried out using data from other components of the Copernicus program, as well as from global EO sources. Sentinel-3 data, available through the World Fire Atlas, indicated the presence of several burning events in the targeted area (Fig. 4A), with a peak of activity in the second half of February. These detections, although spatially coarser than those from Sentinel-2, provide support for the rapid assessment of the phenomenon at a regional scale, with daily updates. The S3-WFA data, available since the 2016–2017 winter season, offer an important starting point for identifying the general trend of vegetation fires. For a clearer understanding of their temporal dynamics, a seasonal analysis was performed for the available period (Fig. 4C). Most fires occur in winter and spring, with an annual minimum identified in summer. Although a positive trend is noted regarding the number of detected events, there are also years when events are minimal (2021, 2023 and 2024). In 2025, the highest number of observations since 2016 is noted: over 40 events. Continuous monitoring is crucial throughout the year, as the phenomenon may not remain constant over time.

Additionally, data provided by EFFIS (Fig. 4B) also highlights the burnt areas. There is an expansion of the affected areas by over 10000 hectares in the Danube Delta (Fig. 4D) in the first three months of 2025. Given the used methodology (a cumulative detection between MODIS and Sentinel-2), the events may appear extensive in both space and time. However, most individual occurrences are limited, lasting no more than 5 days and covering areas under 1000 hectares.

Figure 4. 

Presents a composite of results and information structured as follows: A. Fires extracted from the Sentinel-3 World Fire Atlas, overlaid on a Sentinel-2 image from 22 March 2025; B. Burned areas provided by EFFIS for the January–March 2025 period, overlaid on the same S-2 image from 22 March; C. Number of fires per season, starting from 2017, based on Sentinel-3 World Fire Atlas data; D. Area affected by wildfires and the number of days on which fires were observed – a correlation based on EFFIS data.

For the consolidation of results, the integration of data originating from the FIRMS (NASA) and the Landsat archive was also carried out. On February 20, the Landsat 9 image (Fig. 5A) captures an active event, and the superimposed MODIS and VIIRS detections confirm the position and orientation of the fire. These detections, marked by orange points, correspond to pixels with high thermal signatures and correlate with Sentinel-2 observations from the same day. On March 22, the Sentinel-2 image over the area, in parallel with MODIS and VIIRS data from the first three months of 2025, reveals a persistence of the phenomenon, with active burning visible in multiple hotspots (Fig. 5B).

Figure 5. 

Presents a composite of results and information structured as follows: A. Points indicating locations where FIRMS highly confidently detects fires on 20 February 2025, overlaid on a Landsat image from the same date where the fire is visibly confirmed; B. Fires detected by FIRMS for the January–March 2025 period, overlaid on a Sentinel-2 image from 22 March 2025, clearly showing the correlation between datasets; C. Total cumulative fire radiative power (FRP) normalized by the number of detections, as provided by FIRMS between January and March 2025; D. Cumulative number of fire detections grouped in 10-day intervals, as reported by FIRMS from January to March 2025.

The temporal analysis of FIRMS detections shows a significant increase in activity between February 21 and 28, 2025, with over 800 points detected within a 10-day interval (Fig. 5D). This peak is supported by the high values of the Fire Radiative Power (FRP – “the rate of emitted radiative energy by the fire at the time of the observation, measured in Wats (W)”, CEOS Working Group on Calibration and Validation) indicator.

Overall, the integration of Sentinel-2 images with derived products from Sentinel-3, FIRMS data, and EFFIS provides a replicable mythology for monitoring vegetation fires. EO data not only confirm the presence and extent of the fire but also allow a more eloquent understanding of the temporal and spatial distribution of the analyzed phenomena. The event from February–March 2025 serves as a representative example demonstrating the applicability of Earth Observation data for specific research or monitoring interests.

At the same time, the analysis of Sentinel-2 metadata reveals that, despite a hypothetical revisiting time, cloud cover remains one of the most limitative factors. This aspect underscores the added value of sensors with high temporal resolution (e.g. Sentinel-3, MODIS, VIIRS) and derived products (e.g. EFFIS) for bypassing these limitations.

The metadata corresponding to the four Sentinel-2 tiles (35TPL, 35TQL, 35TQK, 35TPK), for all three active platforms, for the period 2015–2025 show an average of over 1400 scenes per tile (Fig. 6). Tile 35TPK, records the highest number of images (1964), and filtering scenes with <10% clouds we obtain between 400 and 700 clear images per tile. This frequency allows, on average, very precise observations approximately every 10 days – for cloud cover <10%, for 35TPL tile, a remarkable density for monitoring dynamic processes in the Danube Delta.

As mentioned in the the beginning of this material, temporal coverage of topographic maps is scarce, with approximately one at every 13 years., we can now clearly state that with higher temporal frequency of Earth Observation data, we are witnessing a paradigm shift in the research and monitoring methods of the Danube Delta and beyond.

Figure 6. 

Displays a Sentinel-2 composite of the Danube Delta, alongside the spatial division into the four Sentinel-2 tiles that overlap the area. It also shows the number of available images for each tile, as well as several metadata parameters specific to the Sentinel-2 platform.

The satellite data used in this study — originating from the Copernicus and Landsat programs, together with global products such as S3-WFA, FIRMS, and EFFIS — have highlighted a series of capabilities for monitoring, analysis, and understanding natural and anthropic phenomena in the Danube Delta, with an emphasis on the dynamics of vegetation fires. Wetlands can be affected by vegetation fires, resulting in various environmental impacts, including changes in water quality.

During the winter and early spring of 2025, a series of extensive fires were observed in the Danube Delta. The combined use of Sentinel-2 products (RGB and SWIR) allowed for detection and analysis of their extent over short time intervals. Also, the impact on the landscape was assessed. Complementary data from other EO programs provided an objective estimation of the affected areas, the number of events, and their multi-annual trend, and offered an overview of the spatial distribution of vegetation fires.

One of the main contributions of this paper is demonstrating the utility of systematic monitoring based entirely on open data.

The evaluation of Sentinel-2 scene availability highlighted a high density of useful images during the period 2016–2025, with hundreds of valid scenes even after applying strict cloud filters (<10%). This conclusion validates the use of Sentinel-2 as an operational tool for monitoring the Danube Delta.

Although historical maps are a very important tool for understanding the deltaic space from a perspective where anthropic interventions were limited and targeted towards punctual elements, currently, without the use of Earth Observation data, monitoring the changes and dynamics present in the Danube Delta in real-time and with modern research means is not possible.

Data were provided by the European Forest Fire Information System – EFFIS (https://forest-fire.emergency.copernicus.eu) of the European Commission Joint Research Centre. We acknowledge the use of data and/or imagery from NASA’s Fire Information for Resource Management System (FIRMS) (https://www.earthdata.nasa.gov/data/tools/firms), part of NASA’s Earth Science Data and Information System (ESDIS).