Sediment runoff from disturbed coastal catchments is a major threat to marine ecosystems. Understanding where sediments are produced and where they are delivered enables managers to design more effective strategies for improving water quality. In this context, targeted restoration of degraded terrestrial areas provides opportunities to reduce land-based runoff from coastal areas and consequently foster coral reef conservation. To do this strategically, a systematic approach is needed to identify watersheds where restoration actions will provide the highest conservation benefits for coral reefs. This dataset is associated with the article "Global forest restoration opportunities to foster coral reef conservation", where we developed a systematic approach for identifying global forest restoration opportunities that would also result in large decreases in the flux of sediments to coral reefs. The dataset presented here consists of a global sediment export layer produced using the Sediment Delivery Ratio model of InVEST.

The sediment export per pixel is calculated using the soil erosion per pixel, as well the amount of sediment eroded from each pixel that actually reaches a stream or similar water course (Hamel et al. 2015). To model soil erosion, we used a high-resolution global potential soil erosion model developed by Borrelli et al. (2017) based on the revised universal soil loss equation (RUSLE; Renard et al. 1997), estimating soil erosion in each watershed. Sediment export was quantified using the InVEST sediment delivery model (Hamel et al. 2015), which implements a soil loss algorithm linked to the sediment connectivity algorithm proposed by Borselli et al. (2008). More details about the methodological approach can be found in Suárez-Castro et al. (2021).

Following Borrelli et al, we have capped the sediment export values to a maximum of 325 tonnes per hectare per year for pixels classified as barren, 250 tonnes per hectare per year for pixels classified as agricultural land and 50 tonnes per hectare per year for pixels classified as forest.  

Although substantial variations in model performance exist across regions, there is a high correlation between observations and predictions for 58 out of 65 basins used for the model calibration (We only calibrated our model for coastal basins having an effect on coral reefs). In addition, our model showed good agreement with the spatial patterns of the soil erosion model presented by Borrelli et al. (2017) (Spearman correlation coefficient of 0.91, Suarez-Castro et al. 2021) for all the studied regions. Even though the number of basins available for model validation was low, the good fit with Borrelli’s erosion model provides confidence that our model was able to correctly identify high and low sediment export areas at global scales.

Our aim was to generate estimates that allow us to optimally capture relative differences and opportunities among watersheds, rather than describing fine-scale within watersheds processes. This modelling approach is thus suitable for our study as it allowed us to capture major differences between sub-watersheds that can help to identify hotspots of higher sediment export where priority actions should focus. However, pixel level estimates should be taken with caution, and calibration with localised data would be ideal if this dataset is used in local scale studies.