First exploratory analysis

[lorenzznerol]

Larvae example trajectories

This is 2x 10.000 example trajectories from biophysical experiments performed with Parcels.

The trajectories are a tiny subset of a much bigger collection of trajectories that have been simulated with the aim of learning about the fate of larvae of different species of fish. Note that we cannot give a lot of background on the biological study here, because we are still working on this publication.

Also note, that these trajectories should not be used for biological or physical science but merely serve as study objects for developing, testing, or benchmarking (stasticical) methods and algorithms.

Different physics

The trajectories are taken from two classes of biophysical simulations which differ in the physical processes they take into account when estimating horizontal movement of the larvae at the surface.

One experiment only took into account ocean currents from a state estimate of the true ocean circulation in the Mediterranean Sea. The other experiment took into account ocean currents and the horizontal movement due to Stokes Drift which was estimated based on a wave state estimate for the Mediterranean Sea.

Data files

There are two compressed CSV files:

trajectories_nostokes_subset_10000.csv.gz contains trajectories of particles moving only according to the ocean currents. (The file trajectories_nostokes_subset_10000_sample_10_lines.csv contains an uncompressed sample with only the first 10 lines.)
trajectories_stokes_subset_10000.csv.gz contains trajectories of particles moving according to the ocean currents and according to Stokes Drift. (The file trajectories_stokes_subset_10000_sample_10_lines.csv contains an uncompressed sample with only the first 10 lines.)

There are two netCDF files which contain the same data:

trajectories_nostokes_subset_10000.nc contains trajectories of particles moving only according to the ocean currents.
trajectories_stokes_subset_10000.nc contains trajectories of particles moving according to the ocean currents and according to Stokes Drift.

Columns and their meaning

• "obs" contains the time step since the larva started to exist. Each trajectory covers 40 days of hourly positions.
• "traj" indicates the trajectory ID.
• "MPA" indicates the marine protection area (MPA) the particle is in. 1 to 9 indicate real MPAs. 10 indicates coastal regions in general. 0 indicates open ocean. More on MPAs can be found here.
• "distance" contains the distance traveled through the water since the start of the trajectory in kilometers.
• "land" is measuring the land influence at the given position. The velocity and possibly wave data used in the simulation comes on a regular horizontal grid. Velocity on land is set to zero. At runtime of the simulation, the particle is moved according to a velocity that is linearly interpolated to the position of the particle. (This amounts to simulating trajectories with a No-slip condition.) Values land==0 indicate that no data from land points was used in the interpolation. Values land==1 indicates that all four interpolation base points were on land.
• "lat" and "lon" contain the horizontal positions in degrees Latitude and Longitude.
• "temp" contains the ambient temperature in degrees Celsius the simulated larva would have felt at the given time and position. Note that the temperature at positions under land influence (land > 0) is wrong, because there is a temperature of 0 degrees Celsius assumed for each land point.
• "time" contains time stamps.
• "z" contains the vertical positions of the simulated larva in meters counted downwards.

Source; Mediterranean Sea Trajectory Data Examples

[lorenzznerol]

[lorenzznerol]

Taken from "Create relational database as backend #13" https://github.com/siradam/DataMining_Project/issues/13

test query

SELECT obs, traj, mpa, distance, land, lat, lon, temp, "time", z
    FROM public.trajectories_nostokes_subset_10000 where traj = 0 order by obs limit 100;

The "order by" is needed, the observations are not ordered in the source.

Running:

SELECT obs, traj, mpa, distance, land, lat, lon, temp, "time", z
    FROM public.trajectories_stokes_subset_10000 where traj = 0 order by obs limit 100;

We see that the trajectories are much faster, the distance is almost double of the "nostokes" dataset.

obs 99, no Stokes drift: 99 0 1 15.530908 0 43.29053 5.2286572 13.36425 "2017-03-05 03:00:00" 1.0182366 obs 99, with Stokes drift: 99 0 1 27.268465 0.99102813 43.353794 5.2833734 0.120739646 "2017-03-05 03:00:00" 1.0182366 With Stokes drift, obs 54 has already reached the "nostokes" obs 99 distance: 54 0 1 15.538239 0.40967715 43.26185 5.3504033 7.9671803 "2017-03-03 06:00:00" 1.0182366

If it is taken away by the Stokes drift, the larvae trajectory also seems to take a slightly different way, since at the same distance, lon and lat differ quite a bit, and the land value is different.

It is unclear how interesting this comparison is, but it might lead us to add a possible connection between the two datasets not on the location and time, but on the distance.