Latest SCI publications

Latest Projects

Research project (§ 26 & § 27)
Duration : 2017-02-01 - 2018-07-31

Combined analysis of Proba-V 100m and MODIS NDVI products for near-real-time detection of “disturbances” in natural vegetation and forests across Europe In the last years, a strong interest has been devoted to the development of robust methodologies for multi-temporal information extraction and analysis (Bovolo & Bruzzone, 2015). The main reasons are: i) the increased number of satellites with higher revisit period that allow the acquisition of frequent images, ii) the new policy for data distribution of archive data that makes a retrospective data analysis on large scale possible, and iii) new policies for the (free and open) distribution of satellite data
Research project (§ 26 & § 27)
Duration : 2016-05-01 - 2018-04-30

The HQ-S2 project will generate advanced pre-processing algorithms for Sentinel-2 (S-2) data and produce long time series of high quality (HQ) images corrected of cloud effects and atmospheric noise built from combined Landsat-8 and available, current, S-2 scenes. The resulting multi-temporal optical data set is an essential base product for advanced applications, which will be tested in three specific use cases in the domains of agriculture, forestry and urban areas. The project results are the first step towards a novel high quality user-oriented S-2 service potentially provided in the future via the EODC framework. As a prerequisite, the quality and accuracy of the ESA standard Sentinel-2 (S-2) L1C products in terms of geometric and radiometric calibration and correction will be investigated with a specific focus on areas with high topographic variations. Near-real-time (NRT) filtering will be applied to a image time series composed from S-2 and Landsat-8 data in order to fill observational gaps between cloud-free imagery and to remove artefacts due to (undetected) clouds and poor atmospheric conditions. HQ-S2 will take methods of data pre-processing implemented in the ESA S-2 toolbox and will further develop and extend these to meet the needs of the use cases and integrate and test these within the EODC Earth Observation data processing framework.
Research project (§ 26 & § 27)
Duration : 2016-05-01 - 2018-10-31

The objectives for natural (forest) resources management continue to broaden following changing national and international regulations. The multi-purpose objectives naturally lead to increasing information requirements, which are further amplified by increasing market pressures and more frequent hazards related to climate change. Traditional forest inventories are only partially adequate to provide the necessary detailed information for practitioners over large areas and with dense updating intervals. In this respect, the newest generation of Earth Observation (EO) satellites offers a wealth of relevant information. Compared to customary orthophotos, Sentinel-2 offers for example large area coverage at 10 m spatial resolution under constant viewing and illumination conditions at very high revisit frequency (5 days). The sensor has a powerful spectral setting specifically fine-tuned for vegetation monitoring such as red edge bands and bands in the SWIR. Importantly, all data are free. The spatial resolution of Sentinel-2 is probably adequate for most applications and in cases of higher resolution requirements, data from Pleiades satellite can be used in a complementary way. This VHR satellite offers in addition stereo capability to derive relevant 3D information and can be tasked by Austrian institutional users at reduced costs. The improved availability of EO sensors has amplified the dynamics within the research community and created a renewed interest from forest stakeholders. A number of studies have demonstrated the potential of EO for mapping forest parameters such as growing stock, tree species and habitat characteristics. However, most studies focus only on a single sensor and derive only a few forest parameters. The three overarching goals of EO4Forest are: • to assess the potential of Sentinel-2 and Pleiades data for mapping a large range of forest parameters (e.g. growing stock, tree species, deadwood, tree/crown size, gaps and patchiness, leaf area index and light conditions) and to compare the performance against those obtained using commonly available data (e.g. orthophotos), • to combine the derived forest parameters into meaningful and forest-relevant information products thereby taking into account the requirements of forest managers as well as the management of natural resources/wildlife management, • to define protocols for EO data analysis that make optimum use of classical terrestrial inventories while providing hints for improved sampling designs. EO4Forest will thereby combine different (satellite) sensors, exploit multi-sensor approaches as well as the additional information provided in dense time series of Sentinel-2 and stereo information from Pleiades. The EO data will be analyzed using a range of methods ranging from classical classification approaches and direct measurements, to object based image analysis (OBIA), the use of physically-based radiative transfer models (RTM) and modern machine learning techniques. Instead of focusing on a single forest type, the experimental set-up is chosen to include three test sites covered by largely different forests. Stakeholders from the forest management community as well as the environmental community (protected areas, wildlife) are consulted and involved to address as early as possible their respective requirements.

Supervised Theses and Dissertations