LAND: SAR Monitoring of the Brazilian Amazon and Central African Rainforests - a PhD project to support EU FP7 ReCover

The goal of this project is to improve tropical forest monitoring to reduce deforestation and forest degradation (REDD) by developing operational methods to use cloud penetrating synthetic aperture radar (SAR) satellite data to map forest and forest change. This project assists two projects that Norut is involved in: the EU FP7 ReCover project (Science based remote sensing services to support REDD and sustainable forest management in tropical region), where Norut is service provider for the Democratic Republic of Congo (DRC) and Colombia, and a Tanzanian-Norwegian MRV collaboration project (Enhancing the measuring, reporting and verification (MRV) of forests in Tanzania through the application of advanced remote sensing). In both projects, Norut is responsible for the use of SAR data to support the countries REDD MRV activities. A literature research on bio- and geophysical as well as technical aspects of tropical forest monitoring was carried out. Historical C- and L-band SAR data from ERS-2, Envisat A(dvanced)SAR and ALOS Palsar (Phased Array type L-band Synthetic Aperture Radar) as well as background information of the regions, in form of existing thematic maps and other ancillary information (i.e. DEMs) were collected. Processing of optical data sets for multi-sensor approaches and VHR data for training and validation are taken care of by ReCover and MRV project partners. Accurate SAR pre-processing, i.e. geocoding and radiometric calibration, is a pre-requisite. Noruts in-house developed SAR pre-processing system, GSAR, has been further developed and adapted to all SAR satellites during this project by including slope correction of the SAR backscatter coefficient. A fully automatic pre-processing system has been developed and delivered to Kongsberg Satellite Station (KSAT) for the Tanzanian MRV project. Similar processing systems are to be implemented in the future in DRC. The ReCover project has been successfully terminated in Dec 2013 and Norut has fulfilled the Service Level Agreement (SLA) that had been signed with the users Observatoire Satellital des Forêts d Afrique Central (OSFAC) in DRC and Instituto de Hidrología, Meteorología y Estudios Ambientales (IDEAM) in Colombia. For the DRC case, Norut has developed methods and delivered the following products based on C- and L-band SAR data over a 68000 km2 test site: SAR backscatter mosaics, forest/non-forest (FNF) maps, biomass maps, forest change and forest degradation maps for the years 2000, 2005 and 2010 [Haarpaintner et al, 2012]. Feedback from the user, the collection of ground reference data during fieldwork in March 2013 [Haarpaintner et al, 2014] and making use of seasonal variation and time series, led to an improvement and a final service product delivery in September 2013 [Haarpaintner et al, 2015a]. Forest change was detected using a multi-sensor approach based on both optical and SAR based FNF maps. Instead of forest degradation, a degraded forest map was produced by mapping forests that were converted to areas of subsistence agriculture around settlements. For the Colombia case, Norut has provided image mosaics based on C-band SAR from the Envisat ASAR WS (wides-swath) and APS (alternate polarization) modes for the periods 2006/2007 and 2010/2011, respectively, the corresponding FNF maps, a forest change map between those periods and a test product for a deforestation alert system based on ASAR APS data. Specific statistical analysis studies of time series of SAR signatures have been performed for several dominant land cover classes to captured the temporal evolution and seasonal effects [Einzmann et al, 2012; Kohling & Haarpaintner, 2013]. Cooperation with OSFAC will continue inside the ESA-funded SAR for REDD project until 2017 in order to provide SAR processing capabilities to OSFAC. SAR mosaics, FNF maps, annual forest change maps have also been produced for the two Tanzanian study sites: Amani, a protected forest reserve, and Liwale that has a broad range of land use and land cover related activities (e.g. deforestation). Methods and products from 2007 to 2010 are described in [Haarpaintner et al, 2015b] and have been delivered to the Sakoine University of Agriculture in Tanzania. A field data set from a national forest inventory project served for a statistical analysis of the evolution of class specific SAR-signatures to analyse seasonal patterns. Aerial LiDAR scanner data collected by partners have been used to build regression lines and produce vegetation height estimation from SAR data for the Liwale region. The methods are highly automized and are implemented in Noruts in-house processing system GSAR for operational forest monitoring. Norut has also organized and contributed to several capacity building workshops inside ReCover as well as organized by the Global Forest Observation Initiativ. Finally, the project contributed to a recent publication in Nature Climate Change (Reiche et al., 2016).

Tropical rainforest plays a key role in our climate and needs to be protected. Norway contributes financially to several REDD efforts, and effective monitoring of deforestation is an important task. However, persistent cloud cover in the tropics often pre vent reliable monitoring by optical sensors. The main objective of this project is to improve current monitoring by developing new methods using cloud-penetrating SAR satellite imagery to detect deforestation and forest degradation. Norut is involved in s everal tropical forest monitoring projects (EU FP7 Recover project and a project inside the GEO Forest Carbon Tracking Task) and other international REDD related proposals. We wish to strengthen our capacity by giving a young researcher the opportunity fo r a PhD project that will profit of the available datasets and a vast international network of highly experienced research groups, including the Brazilian Space Institute (INPE) and the Observatoire Satellital des Forêts d'Afrique Centrale (OSFAC). Severa l different satellite SAR systems (especially C-and L-band) are in orbit. We will evaluate these and future sensors by their operational availability and their performance for rainforest monitoring. We will collect historical optical and SAR satellite dat a, analyzed satellite products, wall-to-wall mosaics and ground observation to study more rigorously the SAR signatures of different states of forest in order to establish statistical relationships and look-up tables. These will then serve to develop new detection methods using change detection, feature detection and classification techniques that will be included in Norut's SAR monitoring system. The idea is to complement INPE's current monitoring systems for the Amazon and to contribute in the establis hment of monitoring systems in the Congo Basin through OSFAC. Capacity building and technical transfer is also included in the projects and several workshops are planned in the involved developing countries.