New EMODnet biology data product: Gridded maps of Calanus finmarchicus and Calanus helgolandicus
The goal of this work was to create gridded maps displaying the abundance of two taxa, Calanus finmarchicus and Calanus helgolandicus, using the variational method DIVAnd. Such gridded maps can have different applications, for instance data visualization or reference for model validation. Calanus finmarchicus is more commonly found in the subarctic waters of the North Atlantic, with depths ranging from surface to about 4000 m. The species is sensitive to environmental parameters such as temperature, salinity, oxygen and nitrates concentration. Calanus helgolandicus is found in the Atlantic Ocean, from the North Sea south to the western coast of Africa. It is sensitive to temperature with a suitable range of 9°C-20°C.
The Continuous Plankton Recorder (CPR) is an instrument, towed by volunteer merchant ships, designed to capture plankton samples. CPR datasets are unique in the sense that data have been acquired in a consistent way (same method) for more than 70 years. In the instrument, plankton collected by continuously moving bands of filter silk is filtered and then the silk, stored in a cassette, is analysed in a laboratory. Each sample represents approximately 10 nautical miles (< 20 km) of towing. Spatial scales below that value cannot be reproduced in the gridded fields. The instrument towed at a depth of about 5 -10 meters at a speed up to 25 knots (46 km/h). More info can be found in the CPR survey webpage: https://www.cprsurvey.org/services/the-continuous-plankton-recorder/.
DIVAnd (Data-Interpolating Variational Analysis in n dimensions) is a software tool designed to perform an n-dimensional variational analysis (or gridding) of arbitrarily located observations. The code, written in the Julia language (https://julialang.org/), is available from GitHub (https://github.com/gher-uliege/DIVAnd.jl).
The main differences of DIVAnd with respect to other gridding methods is that it takes into account physical constraints such as the topography or the advection by currents. The Julia language ensures that computation using large amounts of data points can be performed in reasonable times. The method is described in Barth et al. (2014).
The gridded field is obtained by minimizing a cost function that takes into account:
- The regularity of the field: the gridded field is expected to be relatively smooth, i.e., it won’t exhibit strong gradients or variations (depending on the scales of interest).
- The closeness to the data: in a pure interpolation the field has to pass through all the observations; in the variational method, the gridded field has to be close to the observations.
Other constraints can be added to the cost function, the gridded field is the one that minimizes it. In this version of DIVAnd, the analysis is univariate, meaning that one variable is processed at the time (in this case, the abundance of one species).
You can view the layers here: https://emodnet.ec.europa.eu/geoviewer/?layers=12984:1:1,12979:1:0,12991:1:0,12990:1:0&basemap=ebwbl&active=12984&bounds=-74.42190789489621,16.821087382915223,54.59268906239001,69.23326739681275&filters=