2nd On-line SeaDataCloud User Workshop
Hi-Tech to High Knowledge

Data interpolation: what can we do {for, with} you?

C. Troupin, A. Barth, J.-M. Beckers & the VRE team

What is DIVA


๐Ÿ”ง Software tool to interpolate in situ observations

What do we mean by

Getting gridded fields from sparse, in situ data โคต๏ธ

(โ‰ƒ getting information where we don't have measurements)

Why interpolation
in oceanography ๐ŸŒŠ
is challenging

  1. A large (huge) amount of data points available
  2. Regions without any observation
  3. Physical obstacles: the topography
  4. Multi-dimensional: longitude, latitude, depth and time

How does DIVA work

1. Observations influence their neighbourhood
over a certain distance

2. Observations are assigned a certain weigh,
translating the confidence of the measurement

3. The tool minimises a cost function taking into account:

  • The distance between observation and analysis
  • The regularity (or smoothness) of the interpolated field


DIVA (1991-2016) aka DIVA classic

  • Interpolation in 2 dimensions, based on a finite-element solver
  • Coded in Fortran 77 and 95
  • Not developed anymore, except bug fixes


DIVAnd (2014-), alias the new DIVA

  • Interpolation in n dimensions (n=2, 3, 4, ...)
  • Coded in Julia
  • โ‰  not a new release of DIVA


Why Julia

Fast, high-level, dynamic language

Started in 2012 | Now at version 1.5.1

Open source | โ†—๏ธ growing user community

Why Julia


How to run & use

Install Julia and download the code


DIVA (2D) within Ocean Data View


DIVA-on-Web (2D)


Jupyter notebooks as user guidelines


Jupyter notebooks inside
the Virtual Research Environment

Further deployments

Docker container

Used in the Virtual Research Environment
Singularity container

Under development in the PHIDIAS project (HPC)

Who's using it

SeaDataCloud regional leaders

Regional climatologies for temperature and salinity


EMODnet Chemistry regional leaders

Gridded fields of nutrient concentrations


EMODnet Biology

Creation of specific abundance products


Latest developments

High-frequency radar interpolation

Adding physical constrains to improve the reconstruction

Heatmap based on presence/absence data

Pluto notebooks instead of Jupyer

lightweight: written in pure Julia

simple: no hidden workspace state, cool user interface

reactive: automatically updated when a cell changes

Concluding remarks

Why may I not be able to use it

Hofstadter's Law:

It always takes longer than you expect,
even when you take into account Hofstadter's Law."

Does it work only with oceanographic data

Does it work only with oceanographic data

Fire heatmap, based on remote sensing data (MODIS and VIIRS)

Your state after this presentation

Bad work, I don't want to use this tool

Not bad, but I'll never use it

Cool, I want to try it with a dataset of mine

Quick recap

  1. DIVAnd is a software tool specifically designed for
    the spatial interpolation of oceanographic data
  2. The code is written in the Julia language and
    optimised to process large amounts of data
  3. The tool is currently used in different EMODnet lots
    and deployed in the frame of other Europea initiaves
    such as PHIDIAS, BlueCloud, EOSC-Hub