Predicting meiofauna abundance to define preservation and impact zones in a deep‐sea mining context using random forest modelling

Pedro Martínez Arbizu,Thomas Kuhn,Katja Uhlenkott,Melinda Coleman,Annemiek Vink

doi:10.1111/1365-2664.13621

Abstract

Abstract There is a strong economic interest in commercial deep‐sea mining of polymetallic nodules and therefore a need to define suitable preservation zones in the abyssal plain of the Clarion Clipperton Fracture Zone (CCZ). However, besides ship‐based multibeam data, only sparse continuous environmental information is available over large geographic scales. We test the potential of modelling meiofauna abundance and diversity on high taxonomic level on large geographic scale using a random forest approach. Ship‐based multibeam bathymetry and backscatter signal are the only sources for 11 predictor variables, as well as the modelled abundance of polymetallic nodules on the seafloor. Continuous meiofauna predictions have been combined with all available environmental variables and classified into classes representing abyssal habitats using k‐means clustering. Results show that ship‐based, multibeam‐derived predictors can be used to calculate predictive models for meiofauna distribution on a large geographic scale. Predicted distribution varies between the different meiofauna response variables. To evaluate predictions, random forest regressions were additionally computed with 1,000 replicates, integrating varying numbers of sampling positions and parallel samples per site. Higher numbers of parallel samples are especially useful to smoothen the influence of the remarkable variability of meiofauna distribution on a small scale. However, a high number of sampling positions is even more important, integrating a greater amount of natural variability of environmental conditions into the model. Synthesis and applications. Polymetallic nodule exploration contractors are required to define potential mining and preservation zones within their licence area. The biodiversity and the environment of preservation zones should be representative of the sites that will be impacted by mining. Our predicted distributions of meiofauna and the derived habitat maps are an essential first step to enable the identification of areas with similar ecological conditions. In this way, it is possible to define preservation zones not only based on expert opinion and environmental proxies but also integrating evidence from the distribution of benthic communities.

Highlights

The deep sea represents the largest ecosystem on Earth, it remains largely unexplored (Ramirez-Llodra et al, 2010)
In the Clarion Clipperton Fracture Zone (CCZ) in the north-eastern equatorial Pacific Ocean, such resources are widespread in the form of polymetallic nodules (Wedding et al, 2015), which are a potential source of copper, cobalt, nickel and manganese (Wegorzewski & Kuhn, 2014)
The polymetallic nodule belt in the CCZ is in the focus of exploration activities and plans for monitoring and conservation of areas threatened by potential future mining activities are of great importance (Kaiser et al, 2017)

Summary

Introduction

The deep sea represents the largest ecosystem on Earth, it remains largely unexplored (Ramirez-Llodra et al, 2010). In the Clarion Clipperton Fracture Zone (CCZ) in the north-eastern equatorial Pacific Ocean, such resources are widespread in the form of polymetallic nodules (Wedding et al, 2015), which are a potential source of copper, cobalt, nickel and manganese (Wegorzewski & Kuhn, 2014). Mining such resources will inherently have a severe impact on environment and fauna, the scale of which is difficult to ascertain despite the considerable number of disturbance experiments that have been carried out so far (Jones et al, 2017). It will be crucial to establish additional preservation zones on smaller spatial scales within the contractor areas in order to protect the environment from serious harm, promote recovery and provide baseline conditions against which the impacts of mining in an adjacent, ecologically similar area can be assessed (Vanreusel, Hilario, Ribeiro, Menot, & Martínez Arbizu, 2016)

Methods

Findings

Discussion

Conclusion