The impact of modelling method selection on predicted extent and distribution of deep-sea benthic assemblages

Abstract

Predictive modelling of deep-sea species and assemblages with multibeam acoustic datasets as input variables is now a key tool in the provision of maps upon which spatial planning and management of the marine environment can be based. However, with a multitude of methods available, advice is needed on the best methods for the task at hand. In this study, we predictively modelled the distribution and extent of three vulnerable marine ecosystems (VMEs) at the assemblage level ('Lophelia pertusa reef frameworks'; 'Stylasterids and lobose sponges'; and 'Xenophyophore fields') on the eastern flank of Rockall Bank, using three modelling methods: MaxEnt; RandomForests classification with multiple assemblages (gRF); and RandomForests classification with the presence/absence of a single VME (saRF). Performance metrics indicated that MaxEnt performed the best, but all models were considered valid. All three methods broadly agreed with regard to broad patterns in distribution. However, predicted extent presented a variation of up to 35 % between the different methods, and clear differences in predicted distribution were observed. We conclude that the choice of method is likely to influence the results of predicted maps, potentially impacting political decisions about deep-sea VME conservation.

Predicting the Distribution of Indicator Taxa of Vulnerable Marine Ecosystems in the Arctic and Sub-arctic Waters of the Nordic Seas

“…Estimates from both sources have their own uncertainty, but modeling approaches like MaxEnt do not incorporate this uncertainty in the predictions. Finally, the selection of the modeling approach itself can introduce biases in the prediction (Piechaud et al, 2015). In our study we utilized a single modeling framework, although some authors have suggested the use of ensemble models, averaging predictions from different models (Georgian et al, 2019).…”

Ecological Role of Submarine Canyons and Need for Canyon Conservation: A Review

“…Habitat suitability modeling (also called species distribution modeling and predictive habitat mapping) offers a means to produce maps of the distribution of specific species and/or communities, within and between canyon systems. Its potential use in deep-sea conservation and management has been demonstrated in other deep-sea habitats (Piechaud et al, 2015) and across deep-sea regions (Ross and Howell, 2013). However, production of such maps again requires a robust biological classification system (where communities are considered) as well as a firm understanding of the drivers of community and species distributions.…”

On the Use of Maps and Models in Conservation and Resource Management (Warning: Results May Vary)

“…Boosted regression trees Supervised Costa et al, 2014;Hewitt et al, 2015 Classification rule with unbiased interaction selection and estimation Supervised Ierodiaconou et al, 2011 Discriminant function analysis Supervised Degraer et al, 2008 Ecological niche factor analysis Supervised Tong et al, 2012;Sánchez-Carnero et al, 2016 Fuzzy k-means Unsupervised Falace et al, 2015 Generalized additive model Supervised Schmiing et al, 2013;Touria et al, 2015 Generalized Quick, unbiased, efficient tree Supervised Ierodiaconou et al, 2011;Hasan et al, 2012 Random forest Both Hasan et al, 2012;Diesing et al, 2014;Piechaud et al, 2015 Support vector machine Supervised Hasan et al, 2012 Frontiers in Marine Science | www.frontiersin.orgFIGURE 2 | Example of how different methods can produce different outcomes. The input data were bathymetric data, backscatter data, and topographic data (i.e., slope, easterness, northerness, and relative deviation from mean value) (see Lecours et al, 2016b).…”