Forest ecosystems around the globe are facing increasing natural and human disturbances. Increasing disturbances can challenge forest resilience, that is the capacity of forests to sustain their functions and services in the face of disturbance. Quantifying resilience across large spatial extents yet remains challenging, as it requires the assessment of both forest disturbances and the ability of forests to recover from disturbance. Moderate-resolution remote sensing systems such as Landsat or Sentinel-2 might offer ways for overcoming those challenges, but studies at the European scale are lacking. We fill this gap by analyzing the resilience of Europe’s forests by means of Landsat-based disturbance and recovery indicators. Specifically, we used a comprehensive set of manually interpreted reference plots and random forest regression to model annual canopy cover from smoothed annual Landsat time series across more than 30 million disturbance patches mapped from Landsat time series across Europe and over the time period 1986-2016. From the annual time series of canopy cover, we estimated the time it takes disturbed areas to recover to pre-disturbance canopy cover levels using space-for-time substitution (defined as recovery intervals). We following quantified forest resilience as the ratio between disturbance intervals (i.e., the average time between two disturbance events) and recovery intervals, with critical resilience defined as areas where canopy disturbances occurred faster than canopy recovery. We found that for the majority of forests in Europe, forests cover returns to pre-disturbance values within 30 years post disturbance. The resilience of Europe’s forests to recent disturbance is thus high, with recovery being >10 times faster than disturbance for approx. 70 % of Europe’s forests. However, 12 % of Europe’s forests had low or critical resilience, with disturbances occurring as fast or faster than forest canopy can recover. We conclude that Europe’s forests are widely resilient to past disturbance regimes, yet changing climate, disturbance and management regimes could erode resilience. We further conclude that Landsat and similar moderate-resolution sensors are key to monitoring European forest dynamics and resilience.

Severe and sustained droughts experienced recently in western Europe resulted in bark beetle outbreaks at unprecedented levels, resulting in high spruce tree mortality. In France, the north eastern part was strongly affected by tree mortality, endangering forest health and resulting in a strong economic impact for the timber industry. Operational monitoring tools allowing large-scale detection of bark beetle outbreaks are urgently needed to better understand beetle outbreak dynamics, quantify surfaces and volumes impacted, and help in decision making of stakeholders in the forestry sector, if possible with early warning systems. Satellite remote sensing shows strong potential to contribute to such an operational monitoring system.
Remotely-sensed detection of bark beetle infestation relies on the detection of symptoms expressed by trees when attacked. Infested trees go through three stages, the early stage is called the green-attack stage, mainly characterized from the ground by visual identification of the boring holes in the bark and the resulting sawdust, with little to no change in color of the foliage in the visible domain. It is followed by a red-attack stage with foliage turning red because of changes in foliage pigment content, and finally a grey-attack stage when foliage is falling. Even though the green attack stage’s characteristics make it particularly difficult to detect using remote sensing, multiple publications identified the potential of near infrared and shortwave infrared information to discriminate green-attack from healthy trees, due to the subtle differences in terms of water content of the foliage. However, given the relatively long period with favorable conditions for bark beetle attacks, the efficiency of a monitoring system allowing early detection requires frequent observations.
Sentinel-2 satellites acquire high spatial resolution multispectral images within a five days revisit period. The decametric spatial resolution is appropriate for fine scale monitoring within forest plots, with potential capacity to identify symptoms occurring on small patches of individual trees. Here, we took advantage of the level-2A Sentinel-2 time series produced by the Theia data and service center and developed a method based on anomaly detection over the seasonal signal obtained from a spectral index specifically designed to inform about foliage water content. The method allows for pixel-wise analysis, as an harmonic model is fitted for each pixel on the signal from the first two seasons of satellite acquisitions, supposedly corresponding to healthy status. New acquisitions showing a deviation from the healthy seasonal model can then be identified automatically as anomalies using a threshold. The method produces raster and vector outputs including the date of acquisition of the Sentinel-2 image resulting in three successive anomalies.
We tested our method specifically on spruce forests identified in the north eastern part of France and reported by the national forest database, and used ground observations collected by forest officers and expert observers during three years, provided as geolocated polygons including the date of observation along with the forest stand health status (healthy, green-attack, red-attack, grey-attack). This validation showed a strong degree of agreement between ground observations and anomalies, with a false-positive rate of 2% for healthy trees and a detection rate of 85% for stands including all stages of attack. In addition, the method showed promising capacity to identify bark beetle attacks from the early green-attack stage, with a detection rate of 68%. Anomalies corresponding to trees identified as red-attack stage on the ground were identified on average four months and up to ten months prior observation, while anomalies corresponding to trees identified as grey-attack stage on the ground were identified on average fourteen months and up to twenty months prior observation.
The method was applied to process all Sentinel-2 images over more than 120 000 km2 and 21 Sentinel-2 tiles, in order to produce maps of bark beetle outbreaks from 2018 to present days, and provide these maps to the National Forest Office and governmental forest services for dissemination among forest stakeholders. The assessment of these region-scale products is currently ongoing.
In order to ensure scaling up, continuous production, and dissemination to remote sensing and forestry communities, the method was implemented in a Python package, named “Fordead”, that will be soon released with an open-source license. It gives a fully automated processing workflow, and includes a collection of processing tools that make use of Sentinel-2 data and time-series analyses easier, from time series processing to visualization. Research perspectives are now moving towards the use of this method to identify and assess the dieback detection caused by other factors and affecting other types of forests.

The Forest Flux Innovation Action project of the EU Horizon 2020 programme, Grant Agreement #821860, developed a seamless service chain for the estimation of forest structural and primary production variables. The services were provided for pilot users. The inputs for the computation of the primary production variables were the structural variable predictions as well as daily temperature and precipitation data. The production naturally aims at as small uncertainty as possible. An uncertainty in the structural variable estimation consequently affects the uncertainty level of the primary production output.

The main EO data source of Forest Flux was Sentinel-2 imagery. On one of the nine pilot sites, in Eastern-Central Finland, also airborne laser scanning (ALS) data with 0.5 points/㎡ density were available. We studied how much the Sentinel-2 based structural variable estimates potentially would improve by introducing the ALS data together with the satellite imagery. Sentinel-2 image from 14.6.2019 and ALS data acquired in 2019 were used for the study. The applied estimation method was the in-house Probability method and the estimation was computed using the Forestry Thematic Exploitation Platform (F-TEP) for which the ALS data processing tools were developed.

Openly available field sample plots by the Finnish Forest Centre from 2019 were used as reference data. The sample plots were randomly divided to training and test sets. 601 plots were used for model training and 248 plots were left for the uncertainty assessment. The methods to compute the uncertainty characteristics were the relative root mean square error (RMSE) and bias. Seven Sentinel-2 bands were utilized for the predictions after an initial testing. From the ALS point clouds, six metrics were computed.

Three ALS features produced similar results to using seven Sentinel-2 bands for stem basal area and stem volume, better for tree mean height and tree mean diameter and worse for tree species proportions which could be expected. By using all six ALS features, the results were better than with seven Sentinel-2 bands, except for tree species. Inclusion of the seven Sentinel-2 and six ALS features led to a similar result to using three ALS features only together with the Sentinel-2.

ALS features improved estimation particularly of the high growing stock volume forest. In addition, the overall averaging of the estimation was reduced. The relative root means square error for tree mean height decreased from 24% to 10%, when ALS features were added. For tree mean diameter the decrease was from 27% to 17% and for stem volume from 44% to 31%. It was concluded that combination of ALS and Sentinel-2 bands improve the results compared to using these data sets alone.

Central Europe has recently experienced several extremely hot and dry summers accompanied by a substantially increased risk of forest fires compared to previous years. Forest fires have not played a significant role in the forest history of Central Europe. However, heat waves and forest fires are likely to become more frequent in the future, highlighting the need for more research on forest fires. To this end, satellite-based information on forest fire history can help to inform fire research and for developing operational risk assessments.
The objective of this study was to analyze the fire history of a fire-prone region in Germany by developing annual burned area maps using Landsat and Sentinel-2 time series. The federal state of Brandenburg is one of the most densely forested regions in Germany, dominated by Scots pine. In this study, we used all Landsat and Sentinel-2 images with a cloud cover percentage less than 70, acquired between 1984 and 2020. We used the Framework for Operational Radiometric Correction for Environmental monitoring (FORCE) to prepare the image time series, which includes atmospheric correction, geometric correction, BRDF-correction, cloud masking, and data harmonization, (Frantz, 2019). We then calculated the normalized burn ratio (NBR) for the harmonized time series resulting in a long intra-annual time series for the last forty years for every pixel. Fires cause abrupt changes in NBR. To detect and characterize these changes, we applied the breakpoint detection method employed by the Breaks For Additive Season and Trend (BFAST) algorithm (Verbesselt et al., 2010, Zeileis, 2005) to the NBR time series. The breakpoint detection algorithm results in a segmentation of the time series, in which OLS segments including linear trend terms and harmonic season terms are separated by breakpoints. We then extract for each breakpoint several metrics including the magnitude of change, the pre-disturbance value, and the rate of change before and after the event (Oeser et al., 2020). These breakpoint variables are then used as predictor variables in a random forest classification model to separate burned areas from logging, insect disturbances, and wind breakage. To build a reference database for model training and validation we used location data from the state forest administration in combination with onscreen digitized polygons. At the Living Planet Symposium 2022, we will present the results of the burned area mapping. Our analysis shows that dense time series are needed to accurately capture forest fires in Central Europe. Forest fires in Central Europe are often ground fires and salvaged relatively soon, which complicates fire detection. Our study will contribute to a better understanding on how Copernicus Sentinel-2 can contribute to forest fires research in Central Europe.


References:
Frantz, D., 2019. FORCE—Landsat + Sentinel-2 Analysis Ready Data and Beyond. Remote Sensing, 11
Oeser, J., Pflugmacher, D., Senf, C., Heurich, M., & Hostert, P., 2017. Using Intra-Annual Landsat Time Series for Attributing Forest Disturbance Agents in Central Europe. Forests, 8, 251
Verbesselt, J.; Hyndman, R.; Newnham, G.; Culvenor, D. Detecting trend and seasonal changes in satellite image time series. Remote Sens. Environ. 2010, 114, 106–115.
Zeileis, A. A unified approach to structural change tests based on ML Scores, F statistics, and OLS residuals. Econom. Rev. 2005, 24, 445–466.

Combining GEDI and Sentinel data for structural forest parameter estimation
Authors: Manuela Hirschmugl, Florian Lippl, Hannah Scheicher
Background
Forests have a major impact on the carbon cycle (Mitchard 2018). The majority of the stored carbon dioxide in the biosphere emitted by fossil fuels and industry is absorbed by forests (Pugh et al. 2019). However, the magnitude of its contribution and its distribution as carbon sink is not yet fully understood and remains highly uncertain (Pan et al. 2011). Due to human induced climate change, biodiversity is rapidly declining and habitat is being destructed (Turner et al. 2003; Jetz et al. 2007). In order to mitigate and understand the effects on the ecosystem, continuous spatial measurement frameworks for land cover and vegetation are needed (Bergen et al. 2009). Forest variables such as canopy height, canopy vertical height profiles and biomass have to be analyzed. Pre-launch calibration and validation studies employing simulated GEDI waveforms processed from Airborne LiDAR Instruments (ALS) show promising results and suggest real GEDI data as well suited for capturing vegetation patterns and biomass products and hence being used as a reference data (Rishmawi et al. 2021; Qi et al. 2019; Schneider et al. 2020; Duncanson et al. 2020). Since the release of version 1 GEDI data, various studies have been published assessing the accuracy of GEDI data by evaluating ground elevation and canopy height estimates against airborne laser scanning height data (Adam et al. 2020; Spracklen and Spracklen 2021; Lang et al. 2021; Potapov et al. 2021). These studies are in good agreement to each other and highlight the applicability of GEDI data to forest structure investigations. Furthermore, the ability of the spaceborne laser to analyze complex forest structures with dense and multilayered canopies enables not only AGB estimations but gives also new valuable insights into biodiversity (Guerra-Hernández and Pascual 2021; Spracklen and Spracklen 2021). This could help to a better understanding of the carbon cycle and ecological forecasting (Schneider et al. 2020).
In our project “GEDI-Sens”, we investigate the relations and combination options between forest parameters provided by GEDI and data from the Copernicus Sentinel-1 (S-1) and Sentinel-2 (S-2) satellites. Previous works show varying levels of agreement between GEDI and S-1 (Verhelst et al., 2021) or S-2 data (Lang et al., 2019; Pereira-Pires et al., 2021). The works had different foci, mainly targeting canopy height and/or above ground biomass (AGB). Some authors also integrated both S-1 and S-2 data to improve the relationship (Chen et al., 2021; Debastiani et al., 2019).
In the first step of the project, we investigated the quality of the GEDI data compared to ALS data for a mountainous forest area in the National Park Kalkalpen, Austria. We found the accuracy of the DTM height from GEDI to decreases with increasing slope inclination from an RMSE of 2.71 m for slopes < 10° up to 10.6 m for slopes > 50°. The mean RMSE is 7.6 m. This error is also visible in the evaluation of the canopy heights. The RH100 from GEDI compared to the maximum height of the ALS data shows an RMSE of 7.92 m and a low R² of only 0.38, even if the winter data is excluded (mainly deciduous forests). When excluding all changed areas in the forest cover such as storm damages between the ALS acquisition (2018) and the GEDI data (2019-2020), the RMSE only slightly improves to 7.91 m. In the next step, we used the correct ALS-based terrain height instead of the GEDI inherent terrain height to calculate a “corrected” vegetation height. The resulting R² improved slightly to 0.39, but with an RMSE of 8.01 m. These results suggest that the usability of GEDI for canopy height measurements in mountainous areas is limited. A similar analysis will be done in our second test site in the tropical forests of Uganda, where flat to hilly terrain prevails.
The vertical structure of the vegetation however should be independent of height errors and thus we expect better correlation. This remains to be analysed in the next step. Given a positive result, we will use time series data from both S-1 and S-2, their reflectance/backscatter as well as indices and textural features. We expect the results compared to both ALS derived vertical structure as well as compared to field plots by the time of the symposium. We are also investigating the use of the joint NASA-ESA Multi-Mission Algorithm and Analysis Platform (MAAP) platform for this purpose.
This study is supported by the Austrian Research Agency FFG under the Austrian Space Application Programme (ASAP) No. 38308664.

(a) (b)
Fig. 1: Relation of ALS-based vegetation height with (a) GEDI RH100 and (b) with the canopy height deducted from GEDI top-of-canopy height and ALS-based terrain height.
References:
Adam, M., M. Urbazaev, C. Dubois, and C. Schmullius (2020). “Accuracy assessment of GEDI terrain elevation and canopy height estimates in European temperate forests: Influence of environmental and acquisition parameters”. Remote Sensing 12.23, p. 3948.
Bergen, K. M., S. J. Goetz, R. O. Dubayah, G. M. Henebry, C. T. Hunsaker, M. L. Imhoff, R. F. Nelson, G. G. Parker, and V. C. Radeloff (2009). “Remote sensing of vegetation 3-D structure for biodiversity and habitat: Review and implications for lidar and radar spaceborne missions”. Journal of Geophysical Research: Biogeosciences 114.G2. doi: https://doi.org/10.1029/2008JG000883. url: https://agupubs.onlinelibrary.wiley.com/doi/abs/10.1029/2008JG000883.
Chen, L., Ren, C., Bai, Z., Wang, Z., Liu, M., Man, W., Liu, J., 2021. Improved estimation of forest stand volume by the integration of GEDI LiDAR data and multisensor imagery in the Changbai Mountains Mixed forest Ecoregion (CMMFE), northeast China. Int. J. Appl. Earth Obs. Geoinformation 100. https://doi.org/10.1016/j.jag.2021.102326
Debastiani, A.B., Sanquetta, C.R., Dalla Corte, A.P., Pinto, N.S., Rex, F.E., 2019. Evaluating SAR-optical sensor fusion for aboveground biomass estimation in a Brazilian tropical forest. Ann. For. Res. 62, 109–122.
Duncanson, L., A. Neuenschwander, S. Hancock, N. Thomas, T. Fatoyinbo, M. Simard, C. A. Silva, J. Armston, S. B. Luthcke, M. Hofton, et al. (2020). “Biomass estimation from simulated GEDI, ICESat-2 and NISAR across environmental gradients in Sonoma County, California”. Remote Sensing of Environment 242, p. 111779.
Jetz, W., D. S. Wilcove, and A. P. Dobson (2007). “Projected impacts of climate and land-use change on the global diversity of birds”. PLoS biology 5.6, e157. Lang, N., N. Kalischek, J. Armston, K. Schindler, R. Dubayah, and J. D. Wegner (2021). “Global canopy height estimation with GEDI LIDAR waveforms and Bayesian deep learning”. arXiv preprint arXiv:2103.03975.
Lang, N., Schindler, K., Wegner, J.D., 2019. Country-wide high-resolution vegetation height mapping with Sentinel-2. Remote Sens. Environ. 233, 111347. https://doi.org/10.1016/j.rse.2019.111347
Mitchard, E. T. (2018). “The tropical forest carbon cycle and climate change”. Nature 559.7715, pp. 527–534.
Pan, Y., R. A. Birdsey, J. Fang, R. Houghton, P. E. Kauppi, W. A. Kurz, O. L. Phillips, A. Shvidenko, S. L. Lewis, J. G. Canadell, et al. (2011). “A large and persistent carbon sink in the world’s forests”. Science 333.6045, pp. 988–993.
Pereira-Pires, J.E., Mora, A., Aubard, V., Silva, J.M.N., Fonseca, J.M., 2021. Assessment of Sentinel-2 spectral features to estimate forest height with the new GEDI data. Dr. Conf. Comput. Electr. Ind. Syst. 626, 123–131. https://doi.org/10.1007/978-3-030-78288-7_12
Potapov, P., X. Li, A. Hernandez-Serna, A. Tyukavina, M. C. Hansen, A. Kommareddy, A. Pickens, S. Turubanova, H. Tang, C. E. Silva, J. Armston, R. Dubayah, J. B. Blair, and M. Hofton (2021). “Mapping global forest canopy height through integration of GEDI and Landsat data”. Remote Sensing of Environment 253, p. 112165. issn: 0034-4257. doi: https://doi.org/10.1016/j.rse.2020.112165. url: https://www.sciencedirect.com/science/article/pii/S0034425720305381. 3 Bibliography
Pugh, T. A. M., M. Lindeskog, B. Smith, B. Poulter, A. Arneth, V. Haverd, and L. Calle (2019). “Role of forest regrowth in global carbon sink dynamics”. Proceedings of the National Academy of Sciences 116.10, pp. 4382–4387. issn: 0027-8424. doi: 10.1073/pnas.1810512116. url: https://www.pnas.org/content/116/10/4382.
Qi, W., S. Saarela, J. Armston, G. Ståhl, and R. Dubayah (Oct. 2019). “Forest biomass estimation over three distinct forest types using TanDEM-X InSAR data and simulated GEDI lidar data”. Remote Sensing of Environment 232, p. 111283. doi: 10.1016/j.rse.2019.111283.
Rishmawi, K., C. Huang, and X. Zhan (2021). “Monitoring Key Forest Structure Attributes across the Conterminous United States by Integrating GEDI LiDAR Measurements and VIIRS Data”. Remote Sensing 13.3. issn: 2072-4292. url: https://www.mdpi.com/2072-4292/13/3/442.
Schneider, F. D., A. Ferraz, S. Hancock, L. I. Duncanson, R. O. Dubayah, R. P. Pavlick, and D. S. Schimel (2020). “Towards mapping the diversity of canopy structure from space with GEDI”. Environmental Research Letters 15.11, p. 115006.
Spracklen, B. and D. V. Spracklen (2021). “Determination of Structural Characteristics of Old-Growth Forest in Ukraine Using Spaceborne LiDAR”. Remote Sensing 13.7, p. 1233.
Turner, W., S. Spector, N. Gardiner, M. Fladeland, E. Sterling, and M. Steininger (2003). “Remote sensing for biodiversity science and conservation”. Trends in ecology & evolution 18.6, pp. 306–314.
Verhelst, K., Gou, Y., Herold, M., Reiche, J., 2021. Improving Forest Baseline Maps in Tropical Wetlands using GEDI-based forest height information and Sentinel-1. Forests 12. https://doi.org/10.3390/f12101374

Vast areas of Central and Northern Europe experienced a pronounced drought in 2018. Germany, among other countries, was heavily affected. In some parts of the country, exceptionally dry conditions continued into spring 2021. The effects of the 2018 drought had a strong impact on Central European forests, particularly in the Czech Republic and Germany. Extensive droughts cause severe stress to trees, which is amplified by the specific situation in Germany, where forests are often located in hilly regions or on poor soils, and many trees are planted at the margins of their climatic niche. Once stressed by drought, trees are generally more susceptible to insect damage. While deciduous trees often have the potential to recover from insect infestations, the situation is different for coniferous trees. The European spruce bark beetle (Ips typographus [L.]) is one of the most damaging pest insects of spruce forests in Europe: successful infestation is typically fatal to trees. During the 2018-2020 drought, bark beetle management in Germany had a strong focus on the prevention of outbreak expansion by massive salvage and sanitation logging in outbreak areas and their surroundings. Actual numbers of the associated forest loss are provided based on statistical sampling and are not spatially explicit. Besides, the temporal development can only be traced at annual intervals.
Remote sensing has proven to be valuable in detecting forest changes, particularly stand replacing changes. However, annual change maps typically use annual best-pixel composites or temporal metrics. These can lead to some ambiguity in correctly assigning a change to a particular year, as images taken under optimal conditions are typically weighted more heavily than winter acquisitions. Hence, changes happening late in a year are likely attributed to the next year. Common silvicultural practice in Germany avoids large-scale clear-cuts. This has changed in response to the recent drought. Clear-cuts are common practice to implement salvage logging. To our knowledge there is currently no comprehensive, spatially-explicit assessment of clear-cuts and tree loss in Germany.
We demonstrate an efficient method to map clear-cuts in temperate Central European forests with high spatial (10 m) and temporal (monthly) resolution. We present a first spatially-explicit assessment of the tree-loss areas in response to the 2018-2020 drought in Germany. To achieve this goal, we used time series of Sentinel-2 and Landsat 8 data and a spectral index largely insensitive to illumination conditions, the disturbance index (DI, Healey et al., 2005). The dense time series was aggregated to monthly composites, thereby removing outliers. From the monthly time series (January 2018-April 2021), we computed anomalies with respect to a reference period (2017) and applied simple thresholding to separate clear-cuts and dead trees from healthy and stressed forest stands. We identified changes (i.e. tree loss) persisting over the monitoring period, determined tree loss dates at per-pixel scale and aggregated the results to different administrative levels.
Our results reveal that about 588,489 ha of forest were lost in Germany between January 2018 and April 2021, corresponding to more than 5 % of the total forest area. This figure contains also dead trees that were not yet logged, but mainly refers to cleared forests. In 2018, the tree loss area was still rather low as it took some time for the trees to die in response to the heavy 2018 drought. Most of the cleared areas of 2018 are likely the result of the removal of windthrown trees in the aftermath of 2017 summer storms (e.g. near Passau in Bavaria, South-East Germany) and 2018 winter storms such as “Friederike” (e.g. in Northern and Eastern Germany). Drought induced mortality in beech and spruce trees started in 2018, and was accelerated by bark beetle infestation in spruce trees, which started in 2018 and continued in several outbreak phases until 2021. Salvage logging as radical management strategy started already in 2018 in some federal states such as Saxony-Anhalt, but accelerated through 2019 and 2020, particularly in Hesse and North Rhine-Westphalia. Consequently, the spatial pattern of tree loss changed from larger areas in Eastern and South-Eastern Germany in 2018 to dominant changes in Central and Western Germany in 2019, 2020 and 2021. Considering all forest types, tree loss was evident throughout Germany, even though Northern and Southern Germany were less affected than Central Germany. Central Western and Eastern Germany were most heavily affected with regard to forest loss in coniferous forests. In a belt ranging from the western to the eastern borders of the country, a large share of the coniferous forests was cleared, in some areas more than three quarters. At district level (Landkreis), the pattern becomes clearer than on federal state level. The district of Soest in North Rhine-Westphalia, for example, lost two thirds of its coniferous forests.
While existing annual crown condition assessments are a valuable source to identify general (long-term) forest health developments, spatially-explicit mapping of tree loss is still missing in Germany. We aim to support forest management and scientific understanding with this first assessment of tree loss after the 2018-2020 drought years.