Copyright 2025, Drift+Noise GmbH

PRIIMA is free software: you can redistribute it and/or modify it under the
terms of the GNU General Public License as published by the Free Software
Foundation, either version 3 of the License, or (at your option) any later
version.
PRIIMA is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with
PRIIMA. If not, see https://www.gnu.org/licenses/gpl-3.0.html.

PRIIMA

PRIIMA stands for PRedicted Ice IMAges and combines a static SAR or otherwise remotely sensed image of sea ice with a modelled sea ice drift forecast to produce a series of manipulated images visualizing potential future positions of sea-ice. Merging all the produced forecast satellite images into an animated gif or a video file creates animations of sea-ice movement. You can think of it as something similar to the moving rain radar images in a modern weather app, showing you a predicted rain pattern of the next hours. An example PRIIMA animation is shown on the PRIIMA project overview. The development of PRIIMA was funded by the ESA kickstart program.

Possible model inputs are either sea-ice drift forecasts from the pan-Arctic NeXtSIM system available via the Copernicus Marine Service or, the globally available wind forecast from the ICON weather model provided by the German Weather Service DWD. While NeXtSIM forecasts the ice drift over the next nine days, ICON only forecasts wind for about 3 days into the future. NeXtSIM ice drift can be used as-is, while the wind values are scaled to guesstimate the expected ice drift from them by applying the Nansen rule (2.5% of wind speed and wind direction +/- 25° on Northern/Southern hemisphere).

Usage

Pull the repository. Submodules need to be installed either by appending --recurse-submodules to the git clone command or by running

git submodule update --init --recursive

Then build a priima docker image by executing

docker compose build

Necessary open data is then downloaded in the process of building the image:

• land polygons from OpenStreetMap
• ICON model grid transformation tables from DWD

Preparations

Get SAR imagery in GRD format from your source of trust (for example the Copernicus Dataspace holding Sentinel-1 imagery or the EODMS Portal) and (optional but recommended) transform the images into a suitable projection (polar stereographic) by running a geometric correction on them. This has the effect that the images will be correctly converted from radar geometry (GRD) into a projected map format using a DEM. GDAL can also do the projection but will not take into account SAR geometry artifacts and, more severely, GDAL is flawed when it comes to the poles or to the antimeridian (180° E/W). The priima program operates with polar stereographic projections and will use GDAL to reproject incoming images. Lastly, it is important that the used image file contains the acquisition date in the format _<YYYYMMDD>T<hhmmss>.

Forecast ice drift or wind data is also required. It needs to cover the timespan of the image forecast, starting with the timestamp of the image and going into the future for as many hours as needed. All forecast files need to be present in a single directory. For NEXTSIM, files of the product cmems_mod_arc_phy_anfc_nextsim_hm are required. ICON wind forecast files can be found here.

Running a Forecast

A priima docker container can then be run like so:

docker compose run --rm --volume /input/model/data/:/model_data \
--volume /out/data/:/out_dir --volume /image/data:/image_data priima \
--image /image_data/s1_9572_arctic_20241115T060000.tif \
--forecast-duration 12 --data-source NEXTSIM

You'll see that there are three mounted volumes:

• the input model data directory needs to be mapped to /model_data. Data for the time in question needs to be already present and is not downloaded automatically.
• the output directory needs to be mapped to /out_dir. An appropriate output subdirectory will be created there.
• the directory containing the input image (specified with --image) needs to be mounted in the container

Followed by the docker image name priima (or a custom name you may have given during build).

Furthermore, the following parameters are passed:

• --image the path of the input image, starting from the mount point of the volume
• --forecast-duration the forecast duration in hours
• --data-source either NEXTSIM or ICON, the model with which to forecast the ice drift

There are some other parameters possible for fine tuning results:

• --num-gcps the number of ground control points which can be 4, 9, 16, 25 or 36. The default is 4.
• --output-resolution the image resolution of the output image, default is 200.
• --gtif-out to output images in GeoTiff format instead of cloud-optimized GeoTiffs (default)
• --video will output a .avi video by combining all .tif images found in the output directory

Further Documentation

• Algorithm explanation and basic principles
• overview images of the algorithm are found in docs/img