As the EU’s old-growth forests continue to disappear and productive forests are increasingly endangered by climate-related disasters, acquiring comprehensive EU-level forest data has become essential to preventing and combating illegal logging and the negative impacts of more frequent and intense natural forest disturbances, such as wildfires, windstorms, and pest outbreaks. After all, we cannot protect and sustainably manage what we do not know or understand well. In this context, the European Commission's proposal in late 2023 to establish a forest monitoring framework has reopened the debate on the need for enhanced knowledge about Europe’s forests.

In recent years, Earth Observation (EO) technologies have continued to evolve and provide new opportunities for acquiring data on the status and trends of forests. Studies have shown that remote sensing can generate timely, reliable and comprehensive information at a continental scale in a standardised manner. Comparable data is needed not only for scientific assessments but also plays a role in tracking progress towards European policy targets and objectives relating to forests, such as those formulated in the EU Biodiversity Strategy or the EU Regulation for the Land Use, Land Use Change and Forestry sector, as well as international commitments such as the Paris Climate Change Agreement and the Kunming-Montreal Agreement on biodiversity.

Currently, important forest-related statistics are captured by Member States as part of their National Forest Inventories (NFIs), which are based on field data for most countries. NFIs provide highly reliable, detailed data on parameters such as forest area, distribution, tree species composition, tree growth rates, etc, but indicator definitions and data collection procedures vary across Europe. Many NFI indicators were developed to reflect the particular forest types, climates and topographies of the different European countries, thus resulting in statistics whose relevance and dissemination are limited to specific regions. Additionally, the frequency of NFIs varies between countries. For instance, Spain adopts a 10-year interval between NFI assessments, while Finland updates its NFI on a 5-year cycle, with data collection occurring continuously through measurements distributed over this period. Such differences might create incomparable and patchy data at the EU level, preventing accurate assessments and comparisons.

Aerial image illustrating the vegetation index of a boggy forest, captured through remote sensing (Credit: Adobe Stock)

Remote sensing, on the other hand, enables measurements covering extensive areas that are easily repeatable at a higher frequency and require low manpower or added costs, thus supplementing existing NFI data. Significant advancements in EO-based forest monitoring have been made possible in recent years thanks to the launch of new satellite missions. The Sentinel satellite fleet, launched by the European Space Agency from 2014 onwards, the Landsat 9 satellite, launched by NASA in 2021, and sensors and micro/satellites operated by private companies provide reliable data that can help track forest dynamics in the current climate change context. Combined, these sensors offer a wealth of high temporal and spatial resolution imagery for monitoring any location on the Earth’s surface at a near-daily interval.

Besides technological advancements, open data policies among private companies and government institutions have also contributed to making a rich amount of information available for research and operational uses, particularly as all Landsat imagery was made freely accessible over the Internet by the US Geological Survey in 2008, and with the Copernicus Earth Observation programme of the European Union further sharply increasing free and open data availability.

Yet, the generation of large amounts of geospatial data – in different formats and from various sources such as governments, companies and researchers – came with its own challenges, often exceeding the storage and processing capacities of individual computers. In practice, these barriers have limited users' possibilities to take advantage of available data for research and other operational purposes. Cloud computing platforms and sophisticated algorithms have been essential to increasing data processing and management capabilities, allowing the analysis and visualisation of data that enable high-impact applications, for instance, to monitor deforestation, drought and disaster, food security, water management, climate, and environmental protection, among others. As these technologies develop further, analyses that previously involved only satellite scenes or limited study areas have now been expanded to include entire regions, nations, or even global applications.

Sentinel-2A in the vacuum chamber during testing at IABG in Munich, Germany (Credit: IABG, ©ESA)

Types and uses of remote sensing technologies

Essentially, remote sensing instruments come in two modalities: passive and active sensors. Passive sensors record energy such as sunlight reflected by the Earth’s surface. They include multispectral and hyperspectral technologies, present in satellite and airborne sensors that detect light in the visible, near-infrared, and short-wave infrared regions of the light spectrum. This kind of optical data is highly valuable for a variety of applications in the forest sector, such as natural resources management and monitoring, atmosphere characterisation and monitoring, land cover classification, environmental risk management, and biodiversity monitoring. However, there are issues associated with the use of passive optical imagery. First, the presence of clouds and topographic landscape features like mountains may create shadows by obstructing solar radiation. Second, a “saturation effect” might occur when the top layer of vegetation is dense, making it hard for some sensors to detect changes in wood volume or biomass in lower canopy bands.

Unlike passive sensors, active sensors emit their own energy, such as electromagnetic waves, toward their target area. Active sensors such as LiDAR, which uses light detection through laser pulses, and RADAR, which emit radio waves, boast 3D and multi-dimensional remote sensing capabilities. Such sensors can provide high-resolution, accurate geospatial data regardless of lighting conditions, and, in the case of RADAR, are less dependent on weather.

Laser scanners, for example, can be classified into different types based on the platforms used, including Airborne Laser Scanner (ALS), Terrestrial Laser Scanning (TLS), Mobile Laser Scanners (MLS), and Personal Laser Scanning (PLS). ALS, which captures data using sensors attached to aircraft such as helicopters or airplanes, is the most extensively studied and utilised 3D remote sensing technology for forestry applications and is being adopted to generate forest attribute maps in combination with NFI information in Northern European countries like Norway and Sweden. Due to its three-dimensional capability, it captures fine-scale terrain and vegetation features in higher resolution than many other remote sensing techniques. The technology has proven effective in mapping forest attributes including habitat traits relevant to biodiversity assessments such as tree height and canopy structure. For instance, tree height offers information on the vertical structure of a forest, which influences species composition and diversity. To complement this, spaceborne laser scanning data, such as that collected by the Global Ecosystem Dynamics Investigation (GEDI) gives the chance to better understand such attributes at the global level. The integration of GEDI data with Landsat time series optical data is expected to enable a historical analysis of forest height and dynamics spanning multiple decades.

Copernicus Sentinel-2B satellite views of wildfires blazing in Siberia, with SWIR (enhanced infrared) views of the event. Copyright: Contains modified Copernicus Sentinel data (2018)/ processed by ESA

Wildfire prevention and management

Earth Observation monitoring systems can help identify risk exposure - the degree to which forests are exposed to certain hazards considering elements such as biomass, vegetation cover density, forest types, etc. For instance, in the case of wildfire prevention and management, elevated biomass might create a higher fuel load for forest fires in fire-prone regions. Through remote sensing, biomass and forests’ growing stock can be estimated, thus allowing the calculation of risk exposure to fire disturbances.

Fuel types (tree species) and their conditions can be identified using optical datasets from Sentinel-2, while ALS data provides insights into forest structure, such as gap sizes. This information supports monitoring before, during, and after fires, aiding prevention and management measures. Additionally, Synthetic Aperture Radar and thermal data can support the active-phase management of wildfires. However, current Earth Observation sensors may not be ideal for detecting smaller forest fires in case of lower spatial resolution or long revisit time, which may result in missing the fire’s active phase.

At the European level, the European Forest Fire Information System, managed by the Joint Research Center (JRC), tracks and reports forest fires in Europe using a variety of satellite, meteorological, land cover, vegetation and historical fire data, detailing incidents, their severity, and affected areas.

Drought monitoring

Droughts can be monitored using optical imagery to derive vegetation indices, which serve as proxies for assessing vegetation health. The European Drought Observatory, operated by the JRC, offers extensive open datasets on droughts, including graphs and time-series data at the European level.

Storm damage monitoring

Storm-affected areas are widely monitored through Earth Observation techniques. High-frequency sensors like Sentinel-1 (microwave) and Sentinel-2 (optical) are crucial for timely windthrow detection. However, there is a lack of comprehensive spatial databases on wind disturbances across Europe. The FORWIND database, with over 80,000 areas affected by wind disturbances between 2000 and 2018, is a valuable resource for understanding and predicting wind impacts on forests.

Illegal logging

Along with climate-driven forest disturbance agents, illegal logging is a significant issue and human-caused disturbance in European countries such as Greece, Romania, Latvia, Cyprus, Ukraine, Kosovo and Russia. Detecting individual tree felling is challenging and currently best achieved using commercial Very High-Resolution datasets. Spectral mixture analysis can detect illegal logging by identifying increased shadow fractions in the forest canopy. Change detection methods using microwave and optical data can also be employed. Still, two key challenges are the need for timely information to enable immediate action and distinguishing between legal and illegal logging. The latter requires information on forest concession and protection status, which often comes from non-Earth Observation sources.

Forest die-off due to bark beetle outbreak in the Šumava National Park in the Czech Republic (Credit: Adobe Stock)

Pest and disease monitoring

Optical-based Earth Observation techniques using vegetation indices can offer valuable insights into pest and biotic stressors. Vegetation indices are mathematical formulas used to analyse of the health and density of vegetation based on how they reflect different types of light. While Sentinel-2-based methods can detect issues at the stand level, for instance, using vegetation indices to distinguish infestation classes, the more difficult task of identifying infested trees during the initial attack phase requires examining individual trees, which current sensors cannot adequately do.

Although at the local and national levels, various efforts have been made to monitor pests and biotic stressors, currently, there is no comprehensive database on insect and disease risks across Europe. Challenges relating to data harmonisation, accessibility and the need to involve multiple stakeholders limit the provision of information to the European level. To address them, the JRC is developing a Database of European Forest Insect & Disease Disturbances to create a detailed record of insect and disease damage in European forests since 1981.

What’s next for remote sensing in forest monitoring?

Information regarding the status and condition of forests in the European Union is highly fragmented. Historically, each Member State developed its own forest monitoring and assessment approaches using different combinations of field data, remote sensing and Earth Observation technologies. Since Earth Observation data is acquired in a standardised way across Europe, it can help reduce interoperability issues and inconsistencies in forest information across the continent.

Yet, there are also limitations when monitoring disturbances with remote sensing applications. Some of these are the difficulty in attributing disturbances to specific disturbance agents based on remote sensing data, and differentiating between natural and human-induced disturbances, which are still very complex tasks. Disturbance patterns can also vary according to location and forest structure characteristics, and the intensity of the disturbance affects its detection and interpretation accuracy. Clearcutting and severe tree mortality caused by pests and large fires can be more accurately identified through remote sensing than small disturbances such as early-stage pest infestations, as the latter creates isolated pockets of tree mortality or tree defoliation, resulting in subtler changes per pixel, for instance, in the case of optical remote sensors.

Moreover, integrating remote sensing data with ground-based data presents challenges such as data compatibility and resolution differences. Plot data, often used for validation, can be inadequate due to its limited spatial coverage and variability in data collection methods across regions. This inadequacy underscores the necessity for improved integration techniques and complementary data sources to enhance the reliability of Earth Observation observations. Developing standardised protocols and leveraging machine learning techniques could help overcome these barriers.

References