Storms can cause significant damage to forest areas, affecting biodiversity and infrastructure and leading to economic loss. Thus, rapid detection and mapping of windthrows are crucially important for forest management. Recent advances in computer vision have led to highly-accurate image classification algorithms such as Convolutional Neural Network (CNN) architectures. In this study, we tested and implemented an algorithm based on CNNs in an ArcGIS environment for automatic detection and mapping of damaged areas. The algorithm was trained and tested on a forest area in Bavaria, Germany. . It is a based on a modified U-Net architecture that was optimized for the pixelwise classification of multispectral aerial remote sensing data. The neural network was trained on labeled damaged areas from after-storm aerial orthophotos of a ca. 109 k m 2 forest area with RGB and NIR bands and 0.2-m spatial resolution. Around 10 7 pixels of labeled data were used in the process. Once the network is trained, predictions on further datasets can be computed within seconds, depending on the size of the input raster and the computational power used. The overall accuracy on our test dataset was 92 % . During visual validation, labeling errors were found in the reference data that somewhat biased the results because the algorithm in some instance performed better than the human labeling procedure, while missing areas affected by shadows. Our results are very good in terms of precision, and the methods introduced in this paper have several additional advantages compared to traditional methods: CNNs automatically detect high- and low-level features in the data, leading to high classification accuracies, while only one after-storm image is needed in comparison to two images for approaches based on change detection. Furthermore, flight parameters do not affect the results in the same way as for approaches that require DSMs and DTMs as the classification is only based on the image data themselves, and errors occurring in the computation of DSMs and DTMs do not affect the results with respect to the z component. The integration into the ArcGIS Platform allows a streamlined workflow for forest management, as the results can be accessed by mobile devices in the field to allow for high-accuracy ground-truthing and additional mapping that can be synchronized back into the database. Our results and the provided automatic workflow highlight the potential of deep learning on high-resolution imagery and GIS for fast and efficient post-disaster damage assessment as a first step of disaster management.

<p>Photosynthesis is a major driver of terrestrial ecosystem dynamics. Unfortunately, gross primary productivity (GPP), or the rate at which solar energy is captured and stored into sugar molecules during photosynthesis, cannot be directly measured from remote sensing (RS) signals. Several RS signals related to vegetation pigments and to canopy structure can, however, serve as proxies for GPP. They can further be combined with different types and degrees of modelling to derive spatio-temporal estimations of GPP. Different strategies exist to do so, which often vary with respect to how much they depend on an in-situ reference for GPP, the gold standard being those derived from eddy covariance (EC) measurements at flux tower sites.</p> <p>Here we investigate several such strategies with a specific goal: to explore the potential contribution of Sentinel satellites to improve GPP estimation. The Sentinel fleet is maintained by the European Union’s Copernicus programme, thereby guaranteeing a certain longevity and enabling the establishment of operational services that do not depend on single satellite missions. The main RS signals we consider are: the OLCI global vegetation index (OGVI) and OLCI terrestrial chlorophyll index (OTCI) from the Sentinel-3 OLCI instrument; daytime and night-time land surface temperature (LST) from Sentinel-3 SLSTR; and sun-induced chlorophyll fluorescence (SIF) from TROPOMI on-board of Sentinel-5-P. We further use time series of Sentinel-2 data to quantify the spatial homogeneity within the observational footprints of these coarser spatial resolution products in order to ensure a proper comparison to flux-tower data. The whole exercise is part of the Sen4GPP project funded by the European Space Agency (ESA).</p> <p>The three strategies we explore to derive GPP are: (1) empirical SIF-based estimation of GPP, including a version involving spatial downscaling to reach a finer resolution of SIF; (2) deterministic modelling based on a quantum yield light use efficiency (LUE) model calibrated on EC flux towers; and (3) purely data-driven machine learning (ML) based on EC measurements at flux towers using dedicated 10-fold cross-validation using the FLUXCOM-X framework. The cross-comparison is done for independent flux tower sites over Europe based on the Warm Winter 2020 database, covering the recent past (2018-2020) when TROPOMI SIF observations are available.</p> <p>The results indicate that the ML approach clearly outperforms the process-based LUE approach, which itself performs better than SIF. However, this order also reflects a decreasing reliance in flux tower data and possibly increasing capacity to extrapolate to situations not present in the learning dataset. The results further indicate that the ML approach using Sentinel data can perform better than a baseline using MODIS data alone, probably due to the inclusion of SIF information. Results also illustrate how ensuring the spatial consistency between grid and tower does improve performance, strengthening the rational for spatially downscaling coarse RS signals such as SIF. Overall, these encouraging results bode well for the potential use of Sentinel data to improve our current capacity to monitor biogeochemical process at global scale.</p>

<p class="part" data-startline="9" data-endline="9">While eddy covariance (EC) is a standard for measuring total ecosystem evaporation (evapotranspiration, ET), upscaling from the tower to the regional and global scales is still marred with uncertainties. Here, we explore this scale translation via the FLUXCOM-X framework, which links data from EC measurements and remote sensing to machine learning techniques to produce models for generating globally gridded products. In particular, we explore potential sources of uncertainty inherent to this pipeline and how these influence the resulting gridded products: training data quality including site-selection and coverage of in-situ data, and modelling distinct water flux components (i.e., transpiration (T) and abiotic evaporation) individually compared to total evapotranspiration.</p> <p class="part" data-startline="11" data-endline="11">Overall, changes in the FLUXCOM-X framework compared to previous versions (Jung et al. 2019) results in tangible improvements to the spatial and temporal patterns of the global evapotranspiration products. Furthermore, the predictions of a corresponding transpiration product provide an empirical estimate of plant controlled water fluxes. The resulting global T/ET ratios are consistent with current estimates from isotopic analyses, but with the advantage of high spatio-temporal coverage. Lessons learned from this analysis provide a more targeted line of inquiry into potential avenues for further improvements in global evapotranspiration modeling.</p> <p class="part" data-startline="14" data-endline="14">Jung, M., Koirala, S., Weber, U. et al. The FLUXCOM ensemble of global land-atmosphere energy fluxes. Sci Data 6, 74 (2019). https://doi.org/10.1038/s41597-019-0076-8</p>

<p>Due to climate change the number of storms and, thus, forest damage has increased over recent years. The state of the art of damage detection is manual digitization based on aerial images and requires a great amount of work and time. There have been numerous attempts to automatize this process in the past such as change detection based on SAR and optical data or the comparison of Digital Surface Models (DSMs) to detect changes in the mean forest height. By using Convolutional Neural Networks (CNNs) in conjunction with GIS we aim at completely streamlining the detection and mapping process.</p><p>We developed and tested different CNNs for rapid windthrow detection based on Planet data that is rapidly available after a storm event, and on airborne data to increase accuracy after this first assessment. The study area is in Bavaria (ca. 165 square km) and data was provided by the agency for forestry (LWF). A U-Net architecture was compared to other approaches using transfer learning (e.g. VGG32) to find the most performant architecture for the task on both datasets.  U-Net was originally developed for medical image segmentation and has proven to be very powerful for other classification tasks.</p><p>Preliminary results highlight the potential of Deep Learning algorithms to detect damaged areas with accuracies of over 91% on airborne data and 92% on Planet data. The proposed workflow with complete integration into ArcGIS is well-suited for rapid first assessments after a storm event that allows for better planning of the flight campaign, and first management tasks followed by detailed mapping in a second stage.</p>