Skip to main content

Bringing Earth Observation Services for Monitoring Dynamic Forest Disturbances to the Users EOMonDis

Periodic Reporting for period 2 - EOMonDis (Bringing Earth Observation Services for Monitoring Dynamic Forest Disturbances to the Users EOMonDis)

Reporting period: 2017-02-01 to 2019-07-31

Natural Forests provide valuable ecosystem services and at the same time these ecosystems are affected by direct and indirect human impacts. Especially in tropical countries with a high percentage of remaining natural forests, human induced land use change is one of the main drivers for increasing deforestation rates. During the last decade remote sensing became an integral part of national and international forest policy programmes to address the issue of increasing deforestation rates. One of these is the United Nations Framework Convention on Climate Change (UNFCCC) hosted Reducing Emission from Deforestation and forest Degradation programme (REDD). This program requires the establishment of robust Monitoring, Reporting and Verification (MRV) systems to account for emissions from deforestation and forest degradation. The role of Earth Observation (EO) and in-situ measurements has been underscored as fundamental tools for the evaluation of activity data and emission factors that are essential inputs for the estimation of Green House Gases (GHG). Furthermore, consumers are increasingly aware of the damage caused by deforestation in tropical countries. This has triggered the retail industry and large commodity companies to take action to ensure that their supply chain is seen as responsible and sustainable in terms of the use and preservation of forest resources. Robust forest monitoring systems are therefore a basic necessity to enable companies in verifying their Zero Deforestation commitments.
The EOMonDis project has five overall objectives to support the REDD program, Zero Deforestation initiatives and sustainable forest management in humid and dry tropical countries. These are:
1.) Conduct a User requirements and market assessment for REDD+ and Low Emission Development (LED) programmes; this involves the identification of the needs of stakeholders in terms of the REDD+ policy drivers, working practices and decision making cycles as well as the funding available from the markets.
2.) To develop improved methods based on the User, ZD and the REDD+ policy requirements that will benefit from the improved quality, coverage, revisit times of Copernicus satellite data and will make use of the combination of different data sources from the Sentinel-1, Sentinel-2 satellite missions as well as other companion data.
3.) To assess the utility of the innovative methods, by testing and comparing them on demonstration sites that are representative of both the humid and dry tropical forests.
4.) To develop a service model that ensures a sustainable project outcome for at least 3 years after the project has been completed.
5.) To disseminate the results of the project to a wide audience that includes the user and donor communities via promotion and marketing activities to ensure that the overall impact of the project is optimised.
The first objective of conducting a user requirements and market assessment was carried out during the first months of the project and a deliverable with compiled information was submitted in time. During this time the prototyping sites and demonstration sites were negotiated and set in consultation with the Users. A questionnaire with technical system requirements for the Service Platform was sent to the Users and the major challenges to access funding sources for the United Nations Framework Convention on Climate Change (UNFCCC) policy on reducing emissions from deforestation and degradation (REDD+) were also identified. The fragmented nature of the funding sources and the lack of dedicated budgets for the Earth Observation (EO) applications for Forest Monitoring within these programs were identified as the major obstacles for a sustainable market for European EO industry. Monitoring the evolution of the REDD+ policy and Zero Deforestation (ZD) programmes will be followed throughout the project lifetime. The second objective relates to the development of improved algorithms and models for the mapping of forest disturbances, above-ground biomass and land use land cover. For both, the optical and radar data domain, automated processing chains that are necessary to harmonise multi-temporal data stacks and prepare data for the subsequent analyses were developed. The integration of two different sensors, namely Landsat 8 and Sentinel-2, into the time series data stack was a main task, but first results for forest cover mapping were generated as well. Besides the data harmonisation of Sentinel-1 SAR data, a new approach to detect forest disturbances was investigated and will be further developed towards an implementation in an operational forest disturbance monitoring system. Another main objective of EOMonDis is to disseminate the results of the project to a wide audience. A project website was built and does contain a specific section for news updates. Additionally, a flyer and printout were designed, that is both downloadable from the website and can be given to Users and donors at conferences or meetings. The project team has contributed to eight conferences during the first reporting period and submitted four full papers for publication, of which three are conference proceedings. The other two objectives, which are assessing the utility of the innovative methods and develop a Business Model, will be addressed at a later stage during project lifetime.
During the first 12 month of EOMonDis project the consortia members could achieve progress beyond the state of the art in both data domains. New processing chains and algorithms were developed for optical imagery and radar data. In the optical data processing, there were issues of geometric and radiometric inconsistencies between Landsat 8 and Sentinel-2 satellite data. In order to use both data types within one consistent time series, these misalignments were corrected by a developed processing routine. The correction needs to be automatic in order to process the large amount of satellite data in a timely manner. Time series analysis methods based on optical data so far used only one sensor and are mostly focussed on tropical evergreen humid forest. In the radar data domain a specific pre-processing chain has been developed in order to process the intensive data amount that is provided by the Sentinel-1 satellite constellation. The processing chain automatically downloads, calibrates, orthorectifies and filters Sentinel-1 data. The multi-image filter used is particularly adapted to forest disturbances mapping and takes advantage of the whole SAR time series dataset. It is worth noting that pre-processed S1 images are directly superposed to Sentinel-2 tiles, which is particularly suitable when combining S1 and S2 data for forest mapping. The processing chain, which allows the fast pre-processing of a large amount of Sentinel-1data in an efficient way by using new and free tools to create “ready-to-use” time series of high temporal and spatial resolution, is unique.