Lack of data, appropriate information and capacities put a serious constraint on effective monitoring and management of wetlands in Africa. Still, the availability of the growing volume of environmental data from the Sentinels, combined with data from long-term Earth Observation archives (e.g. ENVISAT and Landsat) represents a unique opportunity for the operational usage of EO for wetland inventorying, assessment and monitoring on global scale. However, it also poses major challenges in terms of data access and exploitation to achieve its full potential. To meet these challenges the European Space Agency launched the GlobWetland Africa project as a major initiative to facilitate the exploitation of satellite observations for the conservation, wise-use and effective management of wetlands in Africa. GlobWetland Africa is building on the legacy of its two GlobWetland precursors (i..e GlobWetland-I [2003-2008] and GlobWetland-II [2009-2014]) by further developing tools and products to effectively use the enhanced capacities of the Sentinel missions, to address wetland conservation and management from a local, national and regional river basin perspective with a continental focus, being centered in Africa. One of the key deliverables of the GlobWetland Africa project is an open source software toolbox for mapping, modelling and monitoring the wetlands across Africa and capacity building in Earth observation data analysis. The GlobWetland Africa toolbox unifies proven and stable open source software into a single graphical user interface enabling users to access and exploit the increasing capabilities of new and freely available satellite observations primarily from the Sentinel missions of the European Copernicus initiative. The toolbox has full end-to-end image processing capabilities for wetland delineation, wetland habitat mapping, monitoring of inundation regimes and water quality as well as for mangrove mapping and river basin hydrology assessments. This presentation will introduce the GlobWetland Africa toolbox and through dedicated application cases, demonstrate the versatility and multipurpose capacity of the toolbox and the Sentinel missions to support wetland assessment, inventory and monitoring.

With the growing demand for water and land the management and planning of wetlands is becoming more and more challenging around the globe. In this situation best practice should be made of available tools that support wetland management, planning, and reporting.
The SWOS project (Satellite-based Wetland Observation Service – a Horizon-2020 project funded by the European Union, SWOS project coordinator Jena-Optronik GmbH) provides monitoring tools and information on wetland ecosystems derived from Earth Observation data.

The presentation at the ESA Living Planet Symposium will introduces tools valuable for wetland management, Ramsar and SDG reporting and ‘smart planning’ specifically in the context of sustainable urban development for the provision of drinking water and sanitation, urban health and human wellbeing, but also in the context of flooding and climate change adaptation. Examples of different test sites the application of tools developed in the SWOS project will be demonstrated. Furthermore, wetland management and policy challenges will be presented that are also representative for numerous other countries and regions around the globe.

Moreover, the software toolbox GEOclassifier will be demonstrated together with mapping and indicator examples. The toolbox has been developed in the frame of SWOS and other previous projects at Jena-Optronik GmbH. The free available desktop version provides a suite of powerful tools for viewing, processing, and analyzing remote sensing images, also offering information about the status and trends of wetlands and their ecological character.
Regarding new demands on data acquisition and processing a cloud version on one of the DIAS platforms was developed and will be presented live during the session.

The presentation will show how these tools can support national reporting obligations and implementation of international and European agreements and directives, provide information such as indicators to inform wetland management decisions, assess wetland related ecosystem functions and services, and monitor degradation and fragmentation of individual wetlands.

Wetlands are amongst the planet‘s most productive ecosystems providing a wealth of ecosystem services, e.g., nutrition, flood control and protection, or support of biodiversity, to name a few. Nevertheless, wetlands are exposed to multiple threats due to climate change, agricultural pressure, hydrological modifications, fragmentations, etc. Thus, a consistent mapping and monitoring of global wetland ecosystems is very important to track changes and trends aiming at supporting wetland conservation and sustainable management as well as for the EO-based support in monitoring the Sustainable Development Goals indicator 6.6.1 on water-related ecosystems. Although EO data are ideal for large-scale inventorying of wetlands, the tremendous diversity of them makes remote detection particular challenging. This diversity and resulting challenge has been tackled by many researchers applying different sensors (optical and radar) and mapping techniques to delineate wetland from non-wetland areas. A global and homogenous inventory, however, is still not available and scope of ongoing research.

Throughout the past years, we have developed an innovative methodology to derive information about wetlands and demonstrated the capability of the approach through the following projects:
• The Copernicus Pan European Land Monitoring Service – Lot4: Water and Wetness Product
• GlobWetland Africa
• Global Partnership for Sustainable Development Data funded EO based wetland inventory of Uganda

The creation of the wetland information for the above-mentioned activities is based on large amounts of observations over continuous time periods. Thus, multi-sensor EO satellite data of high-resolution radar (Sentinel-1) as well as optical (Landsat-5, -7, -8, and Sentinel-2) data has been considered and supported by historical input from the Envisat ASAR mission to fully cover the dynamics of open surface water and wet surfaces.

Using a hybrid sensor approach, i.e., combining optical and radar observations, provides a more robust wetland delineation compared to traditional (either optical or radar) approaches with optical imagery being more sensitive to the vegetation cover and radar imagery to soil moisture content. Additionally, the higher frequency of observations stemming from the combined data streams contributes to a better characterization of seasonal dynamics which is important so that seasonal and temporary changes do not lead to false conclusions of the overall long-term trend in wetland extent. Within the domain of optical remote sensing, the identification of wetlands is based on the enhancement of the spectral signature using bio-physical indices sensitive to water and wetness including automated dynamic image thresholding methodologies and subsequent derivation of a water and wetness probability index (Ludwig et al., 2017). The radar-based algorithm (Naemi et al., 2009; Wagner et al., 1999) builds on geophysical parameters, surface soil moisture dynamics and water bodies, derived from historical Envisat ASAR and Sentinel-1 backscatter time series to identify permanent/temporary wet and flooded areas. In addition, it is possible to identify flooded vegetation according to the double-bounce scattering principle in densely vegetated wetlands. The non-flood prone areas are masked using the Height Above Nearest Drainage (HAND; Rennó et al., 2008) index.

After the separate processing of the optical and radar imagery, the data is fused into a combined water and wetness product (Water and Wetness Presence Index or WWPI) making use of the advantages of both sensor systems. The fused products give information about the probability that a particular location contains water or wet soils, for which a rule-based classification can be applied to finally derive the wetland extent.

The methodology does not detect wetlands in the ecological sense, but rather identifies the physical properties of water and wet soils. However, when mapping wetlands areas, one general issue is that no single, scientific definition of wetlands does exist (cf. Tiner, 2016). The methodology applied on Pan-European scale provides a consistent and objective map of water and wet surfaces which builds the foundation for a final wetland classification according to specific user definitions and needs. This means that additional information (e.g., information about land use) might be incorporated to serve these requirements. Nonetheless, tracking changes of the WWPI over longer time periods will allow to detect changes in water and wet surface areas in a subjective manner without the difficulties of varying nomenclatures through the various definitions.

References:

Ludwig, C., Walli, A., Schleicher, C., Weichselbaum, J., and Riffler, M. (2017), A highly automated algorithm for wetland detection using multi-temporal optical satellite data, Manuscript submitted for publication.

Naeimi, V., Scipal, K., Bartalis, Z., Hasenauer, S., Wagner, W (2009). An Improved soil moisture retrieval algorithm for ERS and METOP scatterometer observations. IEEE Trans. Geosci. Remote Sens., 47, 1999–2013.

Wagner, W., Lemoine, G., Rott, H. (1999). A Method for Estimating Soil Moisture from ERS Scatterometer and Soil Data. Remote Sensing of Environment, 70, 191-207.

Rennó, C. D., Nobre, A. D., Cuartas, L. A., Soares, J. V., Hodnett, M. G., Tomasella, J., & Waterloo, M. J. (2008). HAND, a new terrain descriptor using SRTM-DEM: Mapping terra-firme rainforest environments in Amazonia. Remote Sensing of Environment, 112(9), 3469-3481.

Tiner, R. W. (2016). Wetland indicators: a guide to wetland formation, identification, delineation, classification, and mapping. Crc Press.

The Global Mangrove Watch is an international collaborative project established within the framework of Japan Aerospace Exploration Agency (JAXA) Kyoto & Carbon Initiative and Wetlands International’s Mangrove Capital Africa programme. It was set up to provide geospatial information about mangrove extent and changes – both losses and gains – at national to global scales in support to the Ramsar Wetlands Convention, NGO’s and the public.
A time-series of maps of the global mangrove extent has been generated based on 25 meter resolution global satellite mosaic data from JAXA’s L-band SAR satellites (JERS-1, ALOS and ALOS-2), combined (for the baseline year 2010) with Landsat-5 and -7 data. As of November 2018, maps for seven annual epochs have been produced: 1996, 2007, 2008, 2009, 2010 (baseline), 2015 and 2016. The 2017 extent map is foreseen to be completed in late 2018.
The classification approach comprised the following stages: (a) generation of a coastal water mask from L-band HH and HV backscatter; (b) generation of a mangrove “habitat” layer that identified areas that were actually or potentially able to support mangroves, using Digital Elevation Model (SRTM), Water Occurrence (JRC) and historical mangrove extent maps (Spalding et al; Giri et al.); (c) generation of a baseline classification for 2010 using a combination of ALOS PALSAR and Landsat composites to improve the distinction of the landward border between mangroves and other terrestrial land covers; (d) quality assurance assessment of the 2010 baseline product, and (e) classification of changes (gains and losses) for the years 1996, 2007, 2008, 2009, 2015 and 2016, relative to the 2010 baseline; f) quality assurance assessment of the change products.
The Global Mangrove Watch maps provide an effective means for periodic mapping and monitoring over national to regional and global scales, in a uniform manner as the same type of data and classification algorithms are used over all areas and over all temporal epochs. The data are relevant to support countries reporting on indicators for the Sustainable Development Goals (SDGs) – in particular to SDG Indicator 6.6.1, which refers to Change in the extent of water-related ecosystems over time – and to reporting on Nationally Determined Contributions under the Paris Agreement and the REDD+ scheme under the UN Framework Convention on Climate Change (UNFCCC).
All satellite data and classification software used within the GMW are free and open source. The Global Mangrove Watch maps are available for public download free of charge (www.globalmangrovewatch.org) and provided in standard Geographical Information Systems (GIS) data format.

Wetlands cover approximately 18% of the province of Alberta, Canada, comprising important ecosystems with diverse functions. The Alberta Wetland Policy plays an important part in both recognizing the value and retention of wetlands in the landscape. Wetland management requires wetland identification, delineation, and assessment. The goal of the Alberta Wetland Classification System (AWCS 2015) is to facilitate consistent characterization of wetlands across a range of settings.

The McClelland Lake Wetland Complex is a conglomeration of wetland types dominated by peatland in the boreal forest of northeastern Alberta, covering approximately 200 sq. km. The major habitat types include non-patterned fen, swamp, and bog habitat that feed into an elongated patterned fen at the south end of McClelland Lake. Existing wetland datasets have been derived by manual photo-interpretation or with in-situ visits and rely on specialized experts to validate these delineations. These are not always repeatable for monitoring types of analysis. This research uses the AWCS as the target system and investigates a multi-step integration and analysis of multi-source and multi-temporal Earth observation data using object-based and Random Forest (RF) methods to identify wetland classes and vegetated forms as a basis for on-going monitoring. The study attempts to explicitly address assignment of wetlands classes over broad scale objects in a manner compatible with the approach of manual human interpretation that has formed the basis of wetlands datasets used for regulatory purposes.

High-resolution multispectral WorldView-3 satellite imagery, airborne LiDAR, and multi-date Sentinel-1 and Sentinel-2 data formed the basis for object delineation and RF modelling. Stand-level objects were created using multi-data and multiresolution segmentation. The stand-level objects were further sub-divided to individual tree crowns to capture proportions of vegetated form for each larger stand-level object. Each tree crown object was classified as coniferous or deciduous using RF to facilitate the AWCS class and form assignment (e.g., coniferous swamp must contain at least 25% tree cover, with at least 75% of trees being coniferous). Outputs from this classification were integrated into the final wetland form assignments for each larger object. Training data of the RF classification relied on field surveys, site photos, aerial photography, and field-based soil samples and derived soil map. Candidate predictive wetland class and form variables were drawn from a broad set of multispectral, radar, and LiDAR height and intensity metrics.

RF approach was very effective at crown type discrimination. Out-of-Bag (OOB) error estimates from the model were 2% and 3% for coniferous and deciduous crown respectively. Tamarack (Larix laricina), a coniferous species that senesces in fall turning colour and dropping needles was often confused with deciduous forms with OOB error of 24%.

Overall accuracy of the RF wetland classification based on the OOB error estimates was 94%. Following the AWCS, identifying and delineating fen wetlands performed best with OOB error of approximately 2%. Swamp wetlands performed relatively well, with an OOB error of approximately 20% and less than 10% error of commission. Classification of bogs yielded the lowest accuracy results with an OOB error of about 62% since the field identification of bogs in the area is challenging and training data were limited. All non-wetland classes yielded high levels of accuracy, with OOB error estimates from the model of 1.8% error for grassland and shrubland, and 1.7% for forest.

The variables of importance resulting from the RF wetland modeling, highlights the contribution of vegetation height, percent tree cover, and topography in improving RF prediction accuracy. Terrain derivatives such as Topographic Wetness Index (TWI), Terrain Classification Index, and Valley Depth were also found among the highest-ranking variables. Likewise, LiDAR metrics such as standard deviation of intensity above 3 m, mean intensity, and 50th percentile of intensity were strong predictors.

The integrated approach developed in this work demonstrates how RF can be used to identify significant variables from a comprehensive suite of candidates, including for an object-based approach. Through assignment of vegetated form to larger wetland class areas according to proportionality of coverage, an AWCS classification was achieved that more closely matches what a human interpreter would produce and would be further familiar to regulators.

Wetlands are considered key elements of future food production in Africa since they provide good conditions to grow crops. This is mainly due to their capability to store water. The Kilombero floodplain in Tanzania is one of the largest wetlands in Africa, and it is one of the southern agricultural growth corridor (SAGCOT) key development clusters in Tanzania. However, the area is listed as a Ramsar site because of its unique value as habitat of many species such as the puku antelope (Kobus vardoni). The area is invaluable as it provides numerous ecosystem services. Over the past decades, the area experienced immigration and economic growth causing extensive land cover and land use changes. In this study, we classified Landsat image composites of four decadal periods (1970s, 1994, 2004, 2014) to identify and quantify major areas of past land conversion. The composites were generated using multi-temporal metrics of the spectral bands and spectral indices. The annual flood extent during the rainy season reaches up to 3000 km² inside the floodplain. To date, about 1100 km² of this area are already used for rice crop production. An area of only 1900 km² remains as grassland area with predominantly natural vegetation in the center of the floodplain. About 4000 km² of natural vegetation, predominantly grassland, were converted to cropland since the 1970s. Most of the changes occurred in the period of 2004-2014. The cropland area spreads along the margins of the floodplain, leaving only little space for further changes outside the regularly flooded area. Consequently, crop production increasingly extends inside the flooded areas where the farmers take the risk of high water tables during the rainy season drowning their crops. Whereas the ecological function of the wetland as a corridor for migrating species became largely lost, its role as food production environment grew rapidly over the last years. Suitable food production areas are exploited almost to their limits. As demand for land for food production will further increase, future crop production will likely expand on less suitable places. With the availability of the Copernicus Sentinel-1 and Sentinel-2 data, it is feasible to generate similar composites on an annual basis and at higher spatial resolution. This allows us to produce timely land use information updates that are highly appreciated by decision-makers and stakeholders for better land management.
The work presented here was conducted within the research projects GlobE (funded by the German Federal Ministry for Economic Cooperation and Development and the German Federal Ministry of Education and Research) and SWOS (H2020 project funded by EU).