Marius Rüetschi

Monitoring forest areas with satellite data has become a vital tool to derive information on disturbances in European forests at large scales. An extensive validation of generated maps is essential to evaluate their potential and limitations in detecting various disturbance patterns. Here, we present the validation results of forest disturbance maps generated for four study areas in Germany using Sentinel-2 data from 2018 to 2022. We apply a time series filtering method to map annual forest disturbances larger than 0.1 ha based on spectral clustering and annual change magnitude. The presented method is part of a research study to design a precursor for a national German forest disturbance monitoring system. In this context, annual forest change areas are used to estimate affected timber volume and related economic losses. To better understand the thematic accuracies and the reliability of the area estimates, we performed an independent and extensive validation of the annual product using 20 validation sets embedded in our four study areas and comprising a total of 11 019 sample points. The collected reference datasets are based on an expert interpretation of high-resolution aerial and satellite imagery, including information on the dominant tree species, disturbance cause, and disturbance severity level. Our forest disturbance map achieves an overall accuracy of 99.1 ± 0.1% in separating disturbed from undisturbed forest. This is mainly indicative of the accuracy for undisturbed forest, as that class covers 97.2% of the total forest area. For the disturbed forest class, the user’s accuracy is 84.4 ± 2.0% and producer’s accuracy is 85.1 ± 3.4% for 2018 to 2022. The similar user’s and producer’s accuracies indicate that the total disturbance area is estimated accurately. However, for 2022, we observe an overestimation of the total disturbance extent, which we attribute to the high drought stress in that year leading to false detections, especially around forest edges. The accuracy varies widely among validation sets and seems related to the disturbance cause, the disturbance severity, and the disturbance patch size. User’s accuracies range from 31.0 ± 8.4% to 98.8 ± 1.3%, while producer’s accuracies range from 60.5 ± 37.3% to 100.0 ± 0.0% across the validation sets. These variations highlight that the accuracy of a single local validation set is not representative of a region with a large diversity of disturbance patterns, such as Germany. This emphasizes the need to assess the accuracies of large-scale disturbance products in as many different study areas as possible, to cover different patch sizes, disturbance severities, and disturbance causes.

Insect and disease outbreaks in forests are biotic disturbances that can profoundly alter ecosystem dynamics. In many parts of the world, these disturbance regimes are intensifying as the climate changes and shifts the distribution of species and biomes. As a result, key forest ecosystem services, such as carbon sequestration, regulation of water flows, wood production, protection of soils, and the conservation of biodiversity, could be increasingly compromised. Despite the relevance of these detrimental effects, there are currently no spatially detailed databases that record insect and disease disturbances on forests at the pan-European scale. Here, we present the new Database of European Forest Insect and Disease Disturbances (DEFID2). It comprises over 650,000 harmonized georeferenced records, mapped as polygons or points, of insects and disease disturbances that occurred between 1963 and 2021 in European forests. The records currently span eight different countries and were acquired through diverse methods (e.g., ground surveys, remote sensing techniques). The records in DEFID2 are described by a set of qualitative attributes, including severity and patterns of damage symptoms, agents, host tree species, climate-driven trigger factors, silvicultural practices, and eventual sanitary interventions. They are further complemented with a satellite-based quantitative characterization of the affected forest areas based on Landsat Normalized Burn Ratio time series, and damage metrics derived from them using the LandTrendr spectral-temporal segmentation algorithm (including onset, duration, magnitude, and rate of the disturbance), and possible interactions with windthrow and wildfire events. The DEFID2 database is a novel resource for many large-scale applications dealing with biotic disturbances. It offers a unique contribution to design networks of experiments, improve our understanding of ecological processes underlying biotic forest disturbances, monitor their dynamics, and enhance their representation in land-climate models. Further data sharing is encouraged to extend and improve the DEFID2 database continuously. The database is freely available at https://jeodpp.jrc.ec.europa.eu/ftp/jrc-opendata/FOREST/DISTURBANCES/DEFID2/.

Monitoring and understanding forest dynamics is essential for environmental conservation and management. This is why the Swiss National Forest Inventory (NFI) provides countrywide vegetation height maps at a spatial resolution of 0.5 m. Its long update time of 6 years, however, limits the temporal analysis of forest dynamics. This can be improved by using spaceborne remote sensing and deep learning to generate large-scale vegetation height maps in a cost-effective way. In this paper, we present an in-depth analysis of these methods for operational application in Switzerland. We generate annual, countrywide vegetation height maps at a 10-m ground sampling distance for the years 2017–2020 based on Sentinel-2 satellite imagery. In comparison to previous works, we conduct a large-scale and detailed stratified analysis against a precise Airborne Laser Scanning reference dataset. This stratified analysis reveals a close relationship between the model accuracy and the topology, especially slope and aspect. We assess the potential of deep learning-derived height maps for change detection and find that these maps can indicate changes as small as 250 m². Larger-scale changes caused by a winter storm are detected with an F1-score of 0.77. Our results demonstrate that vegetation height maps computed from satellite imagery with deep learning are a valuable, complementary, cost-effective source of evidence to increase the temporal resolution for national forest assessments.

Microwave scattering from forests generates pixel geolocation shifts in Synthetic aperture radar (SAR) data that require an adequate representation within digital elevation models (DEM) for preprocessing. We analyze the impact of DEM properties on the radiometry and geolocation of radiometric terrain corrected (RTC) Copernicus Sentinel-1 imagery of forests to improve consistency in backscatter intensities for time series analyses. To account for the penetration depth of the C-Band sensor, we approximate the structure of stands in a temperate deciduous forest using height percentiles from Aerial Laser Scanning (ALS) point clouds in the Hainich National Park, Germany. Comparing the RTC results obtained using DEMs of SRTM, Copernicus, and ALS DEMs, the latter reduces topographically induced errors, resulting in visibly smaller effects from topography and spatially shifted information. Based on the P50 ALS vegetation elevation, results show homogeneous intensities within the same orbit and reduce variance from 2.4 dB² to 1.2 dB² in the difference of mid-range data from ascending and descending azimuth directions. Over forest, we observe lower intensities on sensor-facing and increased intensities on away-facing slopes and correlations with the illuminated pixel area (IPA) and local incidence angle. We reduce this bias with linear regressions of intensity on IPA. ALS DEMs in RTC and the proposed regression correction increase the consistency of images across orbits, measured by the inter-orbit range, throughout the selected year at our study site. We suggest the proposed method applies to other areas, requiring further testing under different forest types and topography.

Bei grossflächiger Kartierung von Bodenbedeckungen basierend auf Fernerkundungsdaten entstehen oft systematische Fehler. Um solchen entgegenzuwirken, können Active Learning-Verfahren eingesetzt werden, welche eine effiziente Verbesserung der Kartengenauigkeit ermöglichen. Für die Erstellung der fernerkundungsbasierten und schweizweiten Gebüschwaldkarte im Rahmen des Schweizerischen Landesforstinventars wurde ein solches Verfahren angewandt. Anhand dieses Beispiels soll aufgezeigt werden, wie mittels Active Learning der mittlere absolute Fehler einer grossflächigen Karte mit vertretbarem Aufwand von 6.04% auf 2.68% mehr als halbiert werden konnte.

The benefits of composite products are well known to users of data from optical sensors: cloud-cleared composite reflectance or index products are commonly used as an analysis-ready data (ARD) layer. No analogous composite products are currently in widespread use that is based on spaceborne radar satellite backscatter signals. Here, we present a methodology to produce wide-area ARD composite backscatter images. They build on the existing heritage of geometrically and radiometrically terrain corrected level 1 products. By combining backscatter measurements of a single region seen from multiple satellite tracks (incl. ascending and descending), they are able to provide wide-area coverage with low latency. The analysis-ready composite backscatter maps provide flattened backscatter estimates that are geometrically and radiometrically corrected for slope effects. A mask layer annotating the local quality of the composite resolution is introduced. Multiple tracks are combined by weighting each observation by its local resolution, generating seamless wide-area backscatter maps suitable for applications ranging from wet snow monitoring to land cover classification or short-term change detection.

The Environmental Data Portal EnviDat aims to fuse data publication repository functionalities with next-generation web-based environmental geospatial information systems (web-EGIS) and Earth Observation (EO) data cube functionalities. User requirements related to mapping and visualization represent a major challenge for current environmental data portals. The new Cloud Optimized Raster Encoding (CORE) format enables an efficient storage and management of gridded data by applying video encoding algorithms. Inspired by the cloud optimized GeoTIFF (COG) format, the design of CORE is based on the same principles that enable efficient workflows on the cloud, addressing web-EGIS visualization challenges for large environmental time series in geosciences. CORE is a web-native streamable format that can compactly contain raster imagery as a data hypercube. It enables simultaneous exchange, preservation, and fast visualization of time series raster data in environmental repositories. The CORE format specifications are open source and can be used by other platforms to manage and visualize large environmental time series.

Although trees growing outside forests are less perceived in relation to those inside forests, they are also valuable and serve various functions. The biophysical thresholds of Trees Outside Forests (TOF) were recently defined and standardised by the Global Forest Resources Assessments of the United Nations Food and Agricultural Organization (UNFAO-FRA). With these definitions, standardised mapping of TOF resources became feasible at the national and international level. Here, we aimed to fill the spatial information gap of TOF resources at the national scale with an automated mapping approach based on the UNFAO-FRA definition. The approach was carried out in ArcGIS using adapted UNFAO-FRA biophysical thresholds with routinely acquired countrywide remote sensing data for the whole of Switzerland: a Vegetation Height Model (VHM), a Topographic Landscape Model (TLM Regio) land cover map, and a Forest Mask. Results were validated using stereo-image interpretation data of the Swiss National Forest Inventory (NFI), which verified 95%, 55% and 75% of overall, producer’s and user’s accuracy, respectively. Of the five forest production regions in Switzerland, the highest accuracy was achieved in the Central Plateau and the lowest in the Alps, with accuracy generally decreasing with increasing elevation. Omission and commission errors were highly correlated with the vertical and horizontal accuracy of the VHM, and the applied biophysical thresholds caused both types of error. The final TOF map produced with our approach is at the countrywide scale, is superior to existing TOF information, meets UNFAO-FRA management and reporting needs, and enables the derivation of TOF’s biomass, carbon sequestration potential and species distribution for the whole country.

In der Schweiz gibt es zahlreiche unterschiedliche Lebensräume. Die WSL hat eine digitale Verbreitungskarte erstellt.

En Suisse, il existe d'innombrables milieux naturels différents, dont l'Institut fédéral de recherches sur la forêt, la neige et le paysage (WSL) a dressé une carte synoptique numérique.

The continual increase of shrub forest in the Swiss Alps over the past few decades impacts biodiversity, forest succession and the protective function of forests. Therefore, up-to-date and area-wide information on its distribution is of great interest. To detect the shrub forest areas for the whole of Switzerland (41,285 km²), we developed an approach that uses Random Forest (RF), bias correction techniques and data from multiple remote sensing sources. Manual aerial orthoimage interpretation of shrub forest areas was conducted in a non-probabilistic way to derive initial training data. The multi-sensor and open access predictor data included digital terrain and vegetation height models obtained from Airborne Laser Scanning (ALS) and stereo-imagery, as well as Synthetic Aperture Radar (SAR) backscatter from Sentinel-1 and multispectral imagery from Sentinel-2. To mitigate the expected bias due to the training data sampling strategy, two techniques using RF probability estimates were tested to improve mapping accuracy. 1) an iterative and semi-automated active learning technique was used to generate further training data and 2) threshold-moving related object growing was applied. Both techniques facilitated the production of a shrub forest map for the whole of Switzerland at a spatial resolution of 10 m. An accuracy assessment was performed using independent data covering 7640 regularly distributed National Forest Inventory (NFI) plots. We observed the influence of the bias correction techniques and found higher accuracies after each performed iteration. The Mean Absolute Error (MAE) for the predicted shrub forest proportion was reduced from 6.04% to 2.68% while achieving a Mean Bias Error (MBE) of close to 0. The present study underscores the potential of combining multi-sensor data with bias correction techniques to provide cost-effective and accurate countrywide detection of shrub forest. Moreover, the map complements currently available NFI plot sample point data.

Künstliche Intelligenz, Deep learning, hochaufgelöste Satellitenbilder, automatisierte Auswertungen, landesweite Datensätze etc. sind alles Begriffe, mit denen viele im Bereich Wald- und Forstwissenschaften schon einmal konfrontiert worden sind. Was steckt aber nun genau dahinter bzw. wird bereits angewandt? Im folgenden Artikel wird anhand des neuen «Waldmischungsgrades LFI» aufgezeigt, was fernerkundungs-technisch alles möglich ist und wo eben auch die Grenzen liegen. Dabei wird auch deutlich, dass ein Austausch zwischen der Praxis und der Forschung wichtiger denn je ist.

Countrywide winter and summer Sentinel-1 (S1) backscatter data, cloud-free summer Sentinel-2 (S2) images, an Airborne Laser Scanning (ALS)-based Digital Terrain Model (DTM) and a forest mask were used to model and subsequently map Dominant Leaf Type (DLT) with the thematic classes broadleaved and coniferous trees for the whole of Switzerland. A novel workflow was developed that is robust, cost-efficient and highly automated using reference data from aerial image interpretation. Two machine learning approaches based on Random Forest (RF) and deep learning (UNET) for the whole country with three sets of predictor variables were applied. 24 subareas based on aspect and slope categories were applied to explore effects of the complex mountainous topography on model performances. The reference data split into training, validation and test data sets was spatially stratified using a 25 km regular grid. Model accuracies of both RF and UNET were generally highest with Kappa (K) around 0.95 when predictors were included from both S1/S2 and the topographic variables aspect, elevation and slope from the DTM. While only slightly lower accuracies were obtained when using S2 and DTM data, lowest accuracies were obtained when only predictors from S1 and DTM were included, with RF performing worse than UNET. While on countrywide level RF and UNET performed overall similarly, substantial differences in model performances, i.e. higher variances and lower accuracies, were found in subareas with northwest to northeast orientations. The combined use of S1/S2 and DTM predictors mitigated these problems related to topography and shadows and was therefore superior to the single use of S1 and DTM or S2 and DTM data. The comparison with independent National Forest Inventory (NFI) plot data demonstrated precisions of K around 0.6 in the predictions of DLT and indicated a trend of increasing deviations in mixed forests. A comparison with the Copernicus High Resolution Layer (HRL) DLT 2018 revealed overall higher map accuracies with the exception of pure broadleaved forest. Although, spatial patterns of DTL were overall similar, UNET performed better than RF in areas with a distinct DLT on forest stand level, with the largest differences occurring when only S1 and DTM data was used. In contrast, predictions obtained from RF were more accurate in mixed stands. This study goes beyond the case study level and meets the requirements of countrywide data sets, in particular regarding repeatability, updating, costs and characteristics of training data sets. The 10 m countrywide DLT maps add complementary and spatially explicit information to the existing NFI estimates and are thus highly relevant for forestry practice and other related fields.

Strong winds may uproot and break trees and represent a major natural disturbance for European forests. Wind disturbances have intensified over the last decades globally and are expected to further rise in view of the effects of climate change. Despite the importance of such natural disturbances, there are currently no spatially explicit databases of wind-related impact at a pan-European scale. Here, we present a new database of wind disturbances in European forests (FORWIND). FORWIND is comprised of more than 80 000 spatially delineated areas in Europe that were disturbed by wind in the period 2000–2018 and describes them in a harmonized and consistent geographical vector format. The database includes all major windstorms that occurred over the observational period (e.g. Gudrun, Kyrill, Klaus, Xynthia and Vaia) and represents approximately 30 % of the reported damaging wind events in Europe. Correlation analyses between the areas in FORWIND and land cover changes retrieved from the Landsat-based Global Forest Change dataset and the MODIS Global Disturbance Index corroborate the robustness of FORWIND. Spearman rank coefficients range between 0.27 and 0.48 (p value < 0.05). When recorded forest areas are rescaled based on their damage degree, correlation increases to 0.54. Wind-damaged growing stock volumes reported in national inventories (FORESTORM dataset) are generally higher than analogous metrics provided by FORWIND in combination with satellite-based biomass and country-scale statistics of growing stock volume. The potential of FORWIND is explored for a range of challenging topics and scientific fields, including scaling relations of wind damage, forest vulnerability modelling, remote sensing monitoring of forest disturbance, representation of uprooting and breakage of trees in large-scale land surface models, and hydrogeological risks following wind damage. Overall, FORWIND represents an essential and open-access spatial source that can be used to improve the understanding, detection and prediction of wind disturbances and the consequent impacts on forest ecosystems and the land–atmosphere system. Data sharing is encouraged in order to continuously update and improve FORWIND. The dataset is available at https://doi.org/10.6084/m9.figshare.9555008 (Forzieri et al., 2019).

Informationen über die Ausdehnung und die Entwicklung von Gebüschwäldern sind für verschiedene forstliche und ökologische Fragestellungen von Bedeutung, jedoch sind aktuelle und flächendeckende Datensätze kaum verfügbar. Im Rahmen dieser Arbeit wurde die Kartierung des Gebüschwaldes mit Dominanz von Alpenerle, Legföhre und Hasel für den Kanton Graubünden auf Basis von landesweit verfügbaren Fernerkundungsdaten untersucht. Es wurden die Satellitendaten von Sentinel-1 und Sentinel-2 sowie ein Vegetationshöhenmodell und ein Geländemodell verwendet. Für die überwachte Klassifikation mit Random Forest, einem auf Entscheidungsbäumen basierenden Verfahren des maschinellen Lernens, wurden vom Kanton bereitgestellte und durch Luftbildinterpretation ergänzte Trainingsflächen verwendet. Die unabhängige Validierung der Ergebnisse wurde mit Daten des Landesforstinventars (LFI) durchgeführt. Alpenerlen- und Legföhrenwälder liessen sich mit einer hohen Genauigkeit von 92.1% bzw. 86.7% klassifizieren, für Haselwälder betrug die interne Modellgüte hingegen nur 66.7%. Die resultierende Flächenausdehnung des Gebüschwaldes war vergleichbar mit den Angaben des LFI. Beim direkten Vergleich mit den LFI-Luftbildinterpretationspunkten wurden grössere Abweichungen festgestellt. Der Hauptgrund dafür ist der unterschiedliche räumliche Detaillierungsgrad. LFI-Flächen mit einem hohen Strauchanteil wurden jedoch zuverlässig als Gebüschwald klassifiziert. Die hier vorgestellte Methode unterstreicht das Potenzial von schweizweit verfügbaren Fernerkundungsdaten für eine weitgehend objektive, kosteneffiziente und grossflächige Kartierung von Gebüschwäldern in einer für die Praxis genügenden Genauigkeit.

Die satellitengestützte Fernerkundung hat das Potenzial, nach einem Sturmereignis rasch und kostengünstig eine grobe Übersicht über die entstandenen Windwürfe zu ermöglichen. Ein neuartiges Verfahren wurde entwickelt, das sich der frei verfügbaren Radardaten des Satellitenpaars Sentinel1A/B bedient. Im vorliegenden Artikel wird über die Erfahrungen der grossflächigen Anwendung im Zuge des Sturms Burglind berichtet.

Storm events are capable of causing windthrow to large forest areas. A rapid detection of the spatial distribution of the windthrown areas is crucial for forest managers to help them direct their limited resources. Since synthetic aperture radar (SAR) data is acquired largely independent of daylight or weather conditions, SAR sensors can produce temporally consistent and reliable data with a high revisit rate. In the present study, a straightforward approach was developed that uses Sentinel-1 (S-1) C-band VV and VH polarisation data for a rapid windthrow detection in mixed temperate forests for two study areas in Switzerland and northern Germany. First, several S-1 acquisitions of approximately 10 before and 30 days after the storm event were radiometrically terrain corrected. Second, based on these S-1 acquisitions, a SAR composite image of before and after the storm was generated. Subsequently, after analysing the differences in backscatter between before and after the storm within windthrown and intact forest areas, a change detection method was developed to suggest potential locations of windthrown areas of a minimum extent of 0.5 ha - as is required by the forest management. The detection is based on two user-defined parameters. While the results from the independent study area in Germany indicated that the method is very promising for detecting areal windthrow with a producer’s accuracy of 0.88, its performance was less satisfactory at detecting scattered windthrown trees. Moreover, the rate of false positives was low, with a user’s accuracy of 0.85 for (combined) areal and scattered windthrown areas. These results underscore that C-band backscatter data have great potential to rapidly detect the locations of windthrow in mixed temperate forests within a short time (approx. two weeks) after a storm event. Furthermore, the two adjustable parameters allow a flexible application of the method tailored to the user’s needs.