Janik Deutscher

Forest Assessment Using High Resolution SAR Data in X-Band

Novel radar satellite missions also include sensors operating in X-band at very high resolution. The presented study reports methodologies, algorithms and results on forest assessment utilizing such X-band satellite images, namely from TerraSAR-X and COSMO-SkyMed sensors. The proposed procedures cover advanced stereo-radargrammetric and interferometric data processing, as well as image segmentation and image classification. A core methodology is the multi-image matching concept for digital surface modeling based on geometrically constrained matching. Validation of generated surface models is made through comparison with LiDAR data, resulting in a standard deviation height error of less than 2 meters over forest. Image classification of forest regions is then based on X-band backscatter information, a canopy height model and interferometric coherence information yielding a classification accuracy above 90%. Such information is then directly used to extract forest border lines. High resolution X-band sensors deliver imagery that can be used for automatic forest assessment on a large scale.

Mapping Tropical Rainforest Canopy Disturbances in 3D by COSMO-SkyMed Spotlight InSAR-Stereo Data to Detect Areas of Forest Degradation

Assessment of forest degradation has been emphasized as an important issue for emission calculations, but remote sensing based detecting of forest degradation is still in an early phase of development. The use of optical imagery for degradation assessment in the tropics is limited due to frequent cloud cover. Recent studies based on radar data often focus on classification approaches of 2D backscatter. In this study, we describe a method to detect areas affected by forest degradation from digital surface models derived from COSMO-SkyMed X-band Spotlight InSAR-Stereo Data. Two test sites with recent logging activities were chosen in Cameroon and in the Republic of Congo. Using the full resolution COSMO-SkyMed digital surface model and a 90-m resolution Shuttle Radar Topography Mission model or a mean filtered digital surface model we calculate difference models to detect canopy disturbances. The extracted disturbance gaps are aggregated to potential degradation areas and then evaluated with respect to reference areas extracted from RapidEye and Quickbird optical imagery. Results show overall accuracies above 75% for assessing degradation areas with the presented methods.

Methods for Mapping Forest Disturbance and Degradation from Optical Earth Observation Data: a Review

Purpose of ReviewThis paper presents a review of the current state of the art in remote sensing-based monitoring of forest disturbances and forest degradation from optical Earth Observation data. Part one comprises an overview and tabular description of currently available optical remote sensing sensors, which can be used for forest disturbance and degradation mapping. Part two reviews the two main categories of existing mapping approaches: first, classical image-to-image change detection and second, time series analysis.Recent FindingsWith the launch of the Sentinel-2a satellite and available Landsat imagery, time series analysis has become the most promising but also most demanding category of degradation mapping approaches. Four time series classification methods are distinguished. The methods are explained and their benefits and drawbacks are discussed. A separate chapter presents a number of recent forest degradation mapping studies for two different ecosystems: temperate forests with a geographical focus on Europe and tropical forests with a geographical focus on Africa.SummaryThe review revealed that a wide variety of methods for the detection of forest degradation is already available. Today, the main challenge is to transfer these approaches to high-resolution time series data from multiple sensors. Future research should also focus on the classification of disturbance types and the development of robust up-scalable methods to enable near real-time disturbance mapping in support of operational reactive measures.

Updating lidar-derived crown cover density products with sentinel-2

Crown cover density (CCD) is one important forest attribute used in forest management. With remote sensing, crown cover density maps can be derived from the spectral information of optical satellite imagery or from a normalized digital surface model (nDSM). LiDAR data based applications provide the most accurate results, but LiDAR campaigns are expensive and available data are often outdated. We propose a new method to update LiDAR-derived CCD products and to map forest change. The method is based on Sentinel-2 imagery and an outdated LiDAR nDSM used to train a kNN classifier. CCD estimations are derived for two tests sites in Austria. Results are compared with the LiDAR CCD values in unchanged forest and with the latest tree cover density product of the European Copernicus High Resolution Layers Forest. Results demonstrate the operability of the workflow. User accuracies for forest change detection are very high with 87.3% and 94.8%.

Humid tropical forest monitoring with multi-temporal L-, C- and X-band SAR data

Humid tropical forest monitoring with EO is limited by frequent cloud cover and rapid forest regrowth. Both can be overcome by using temporally dense SAR image stacks. We present a method that uses the coefficient of variation of multi-temporal SAR data stacks to map tropical forest disturbances. The SAR data pre-processing and the forest change detection workflows are described and illustrated. The method is tested at a humid tropical forest site in the Republic of Congo. At this test site we use data from three different SAR sensors: ALOS PALSAR, Sentinel-1 and TerraSAR-X. The forest disturbance maps are validated by visual interpretation and compared to the Landsat based Humid Tropical Forest Disturbance Alerts available from Global Forest Watch. Change mapping accuracies for plots larger than 0.5 ha are very high: 76% for ALOS PALSAR, 96% for TerraSAR-X and 98% for Sentinel-1. For Sentinel-1, producer accuracies were derived for different forest disturbance types. The overall accuracy is 81.8%, with highest values for deforestation in oil palm plantations and burnt areas. The results are similar to the accuracy of the Humid Tropical Forest Disturbance Alert layer, which detects 85.6% of all reference areas. We also show that fusion of the disturbance maps on a result level is possible. The presented method could be adapted to near real-time processing and to a combined processing with optical EO data.

Support of Wind Resource Modeling Using Earth Observation

This contribution outlines the potential of remote sensing data to support wind resource modelling especially through improved input parameterization regarding the state and characterization of the land surface. Wind speed and wind flow is strongly influenced by land surface properties. Three different remote sensing based parameters can help to characterize wind resources: a) land cover and land use; b) digital elevation models (DEM); c) phenological information. Earth observation data are used already in wind resource models to some extent. However, the new advances and especially the possibilities which open up through the Copernicus Sentinel satellites are not considered yet. Opportunities include seasonal mapping of land cover which will allow a precise quantification of vegetation cover which has a direct influence on heat fluxes. The use of newest DEMs like Tandem-X with a 12 m resolution allows detecting also small landscape feature like rows of hedges and trees. Further, elevation models derived by either photogrammetric approaches or airborne laser scanning can further refine the information. By using EO-based information on the surface, e.g. roughness, and in-situ wind measurements, realistic wind fields for sufficiently large areas can be derived by considering also shadowing effects and wind shear.

The assessment of forest parameters by combined LiDAR and satellite data over Alpine regions – EUFODOS Implementation in Austria / Hodnotenie parametrov lesa kombináciou LIDAR-u a satelitných údajov v alpských regiónoch – implementácia systému EUFODOS v Rakúsku

Abstract Regional authorities require detailed and georeferenced information on the status of forests to ensure a sustainable forest management. One of the objectives in the FP7 project EUFODOS was the development of an operational service based on airborne laser scanning and satellite data in order to derive forest parameters relevant for the management of protective forests in the Alps. The estimated parameters are forest type, stem number, height of upper layer, mean height and timber volume. RapidEye imagery was used to derive coniferous and broadleaf forest classes using a logistic regression-based method. After the generation of a normalised Digital Surface Model and a forest mask, the forest area was segmented into homogeneous polygons, tree tops were detected, and various forest parameters are calculated. The accuracy of such an assessment was comparable with some previous studies, and the R-square between the estimated and measured values was 0.69 for tree top detection, 0.82 for upper height and 0.84 for mean height. For the calculation of timber volume, the R² for modelling is 0.82, for validation with an independent set of field plots, the R² is 0.71. The results have been successfully integrated into the regional forestry GIS and are used in forest management. Abstrakt Regionálne plánovanie zabezpečujúce trvale udržateľný manažment lesa vyžaduje detailné a georeferencované informácie o stave lesov. Jedným z cieľov projektu EUFODOS (projekt 7. RP EÚ) bolo vyvinúť operatívnu službu využívajúcu údaje leteckého laserového skenovania v kombinácii so satelitnými údajmi, pomocou ktorých sú odvodené informácie potrebné pre obhospodarovanie ochranných lesov v Alpách. Zisťované parametre sú lesný typ, počet stromov, výška hornej korunovej vrstvy, priemerná výška a kmeňová zásoba. Použilo sa snímkovanie systémom RapidEye pre odvodenie tried ihličnanov a listnáčov s použitím logistického regresného modelu. Po vygenerovaní normalizovaného digitálneho modelu povrchu a masky lesa sa plocha lesa segmentovala do homogénnych polygónov, identifikovali sa vrcholce stromov a vypočítali sa požadované porastové charakteristiky. Presnosť uvedených odhadov bola porovnateľná s predošlými štúdiami - R2 medzi odhadovanými a meranými hodnotami pozícií vrcholcov stromov bol 0,69, pre hornú výšku 0,82 a pre priemernú výšku porastu 0,84. Pri výpočte objemu dreva bol R2 príslušného modelu 0,82. Pri validácii s nezávislým súborom plôch bola dosiahnutá hodnota R2 0,71. Prezentované výsledky sa úspešne integrovali do regionálnych lesníckych GIS sú využívané pri manažmente lesa.