Seagrass bed - Seagrass (e.g. Enhalus, Thalassodendron) in tropical waters - ecosystem component from the WIO Symphony tool

A number of seagrass data sources in the WIO region and globally were combined, including WCMC, Allen Coral Atlas and Living Ocean Foundation maps, and a new predictive model was constructed for the Saya del Mahla banks where sparse observations but no map existed. The maps were ranked by quality, and adjusted and combined based on ranking, information from overlapping higher resolution maps, and in some instances the Gebco depth model to remove unlikely locations. The combined 1-km resolution seagrass map was divided into tropical and temperate maps using a Sea Surface Temperature Copernicus data composite with a continuous gradient between 23-25.5 degC. The major steps included: (1) Mapping the Allen Coral Atlas seagrass polygons presence to a 250-m grid, then calculating proportion of 250-m presence pixels within each 1-km pixel. In addition, total map footprint proportion were mapped to 1-km pixels. (2) Same process as (1) for three additional high-resolution maps (Seagrasses in estuaries in ZAF from Nelson Mandela University, Seagrasses in Chagos and Seychelles from Khaled bin Sultan Living Ocean Foundation) (3) Global Seagrass map from the WCMC portal was mapped same as (1), but due to variable spatial resolution and accuracy it was adjusted with a coefficient calculated by comparing the mean cover in the Allen Atlas 1-km map with the WCMC 1-km map. Further, unlikely/deep areas were masked using Gebco bathymetry. (4) Seagrass observations (points) from the WCMC portal were buffered with 750-m radius and mapped as presence in 250-m pixels -> proportion of 1-km pixels (5) A BRT machine learning model was developed for Saya Del Mahla Bank, modelling seagrass coverage based on % cover data published by Ramah S. et al. The model uncertainty is high, but predicts seagrasses in regions of the banks known to have seagrass (personal conversation Dr. Rawat and others) which are currently not available in any open maps. (6) A Global Maxent model from the WCMC portal was mapped same as (1), then adjusted to the Allen data and bathymetry (same as 3). (7) All maps were mosaiced together in order of their ranking, using map footprint to mask out lower ranked maps, then choosing the max value in the remaining overlap. The mean value was used for the overlapping area in Saya del Mahla between the WCMC polygon and modelled map. Known limitations: The combination of the many maps represents an effort to put the best available information together, still the maps are likely to over- or underpredict seagrasses in most areas. Especially the seagrasses in deeper and in more turbid waters are poorly mapped. In Chagos, seagrasses have been reported to 70 m depth but only the shallow ones are represented in the map. The Saya Del Mahla banks have notations to be the largest seagrass beds on the planet but with the poor data available we do not know. The model in the WIO Symphony seagrass map needs much more observational data to be reliable. In all, the full extent of the seagrass ecosystems remains unknown, and planners should note that large important seagrass areas may not yet be in the map.