Glidersim is an attempt to simulate the behaviour of (typically Slocum) ocean gliders. The key reason for developing this software is that for some glider operations, notably in German waters, it is compulsary to provide trajectory forecasts of each deployed glider with a time horizon of 12 hours. Based on these forecasts, regions of bounding boxes are defined, spanning the forecasted tracks. This information is sent to coastal ship traffic control centres, who issue warnings to ships informing them regarding the ongoing scientific experiments.
In order to run these simulations, the software implements
- a simulator of the glider's software, so that given the same mission file active on a glider, the behaviour of the simulator is similar to that of the real glider;
- a simulator of the glider's actuators (buoyancy pump, pitch battery motor, and fin);
- a simulator of the glider's dynamics;
- a digital environment for the glider to fly in.
The first two simulators are primarily coded in this python package. The glider dynamics are based on the python gliderflight package, which is based on the published paper A dynamic flight model for Slocum gliders and implications for turbulence microstructure measurements [1].
The digital environment plays a major role in how the glider actually moves, and provides the input to the glider flight model. Each environment class implements a method that return for a given time and (latitude, longitude, depth) coordinate the in-situ density, water depth, surface elevation and a three-dimensional current vector.
For the source of digital environments a few options are available. For the purpose described above, environmental data are obtained from the most recent data files sent back by the glider, supplemented by a bathymetry data file of the region. The currents are represented by an estimate (by the glider) of depth-averged currents, averaged in time over a subsurface period. In particular for tidal regions this is not so accurate. An improved method to obtain currents from glider observations uses a Kalman filter [2]. Other sources of input can be model output. The python code provides ways for the GETM model, which uses curvilinear sigma grids and netCDF files for output.
The source code of glidersim can be downloaded from github. Most dependencies can be taken care of by running 'pip install -r requirements.txt', but some python packages are not available via pypi and need to be downloaded from github and installed manually. This applies to the packages latlon and GliderNetCDF. The following bash script can be used to install all required pakcages.
#!/bin/bash
# get the source of additional packages from github
git clone https://github.com/smerckel/latlon.git
git clone https://github.com/smerckel/GliderNetCDF.git
pip install ./latlon ./GliderNetCDF
# Get and install glidersim.
git clone https://github.com/smerckel/glidersim.git
pip install glidersim
An example of how this software can be used is given by the script
gs_example.py, which is to be run from the subdirectory
examples. This script simulates two different missions carried out
with the glider * Comet* in the North Sea in August 2019. The
corresponding data files can be found in the subdirectory data, see
the tree below.
glidersim
├── data
│ ├── cac <- cache files required to read glider binary data files
│ └── comet
│ ├── from-glider <- glider binary data files
│ ├── nsb3 <- directory with mission files for nsb3 mission
│ │ ├── mafiles <- all required ma files, as they were on the glider
│ │ └── missions <- the nsb3.mi mission file
│ └── spiral <- directory with mission files for nsb3 mission
│ ├── mafiles <- all required ma files, as they were on the glider
│ ├── missions <- the spiral.mi mission file
│ └── trim <- glider binary data used to calibrate the glider flight model.
├── examples
└── glidersim
Below is an example script, adapted from gs_example.py shipped with this software.
import glidersim
import glidersim.configuration
import glidersim.environments
# Define a glider model. In this case a Slocum shallow 100 m.
glider_model = glidersim.glidermodels.Shallow100mGliderModel()
# Set the glider flight parameters.
glider_model.initialise_gliderflightmodel(Cd0=0.20, mg=73.3, Vg=71.542e-3, T1=2052, T2=-35.5, T3=0.36)
# The bathymetry is read from a netcdf file. Set the names of the fields that nead to be read.
glidersim.environments.GliderData.NC_ELEVATION_NAME='bathymetry'
glidersim.environments.GliderData.NC_ELEVATION_FACTOR=1
glidersim.environments.GliderData.NC_LAT_NAME='latc'
glidersim.environments.GliderData.NC_LON_NAME='lonc'
# Tell dbdreader where to get the cache files from
glidersim.environments.GliderData.DBDREADER_CACHEDIR = '../data/cac'
# Use current estimates from the BSH model, accessible via an online API:
environment_model = glidersim.environments.DriftModel("comet", download_time=24,
gliders_directory='../data',
bathymetry_filename='../data/bathymetry.nc')
# Create a configuarion dictionary
conf = glidersim.configuration.Config('nsb3.mi', # the mission name to run
description="test", # descriptive text used in the output file
datestr='20190821', # start date of simulation
timestr='13:54', # and time
lat_ini=5418.9674, # starting latitude
lon_ini=724.5902, # starting longitude
mission_directory='../data/comet/nsb3', # where the missions and mafiles directories are found
output='comet-nsb3-nsb3.nc', # name of output file (pickled files (.pck) can also be used
sensor_settings= dict(c_wpt_lat=5418.000,# set some glider parameters
c_wpt_lon= 725.800,
m_water_vx=0.365,
m_water_vy=-0.099),
special_settings={'glider.gps.acquiretime':100., # how long the GPS should take to get a reading
'mission_initialisation_time':400}, # how much time the glider needs to initialise.
mission_start="pickup") # if not set, a new mission is assumed, otherwise it is a continuation of
# a previous dive.
# Create a GliderMission object, specifying the glider hardware and environment.
GM=glidersim.glidersim.GliderMission(conf,verbose=False,
glider_model=glider_model,
environment_model = environment_model)
# load the mission and show the contents.
GM.loadmission(verbose=True)
# Run the simulation with 0.5 seconds time step. The glider CPU cycle
# is set to 4 seconds. We simulate only 7 hours, and show some
# diagnostic output.
GM.run(dt=0.5,CPUcycle=4,maxSimulationTime=7/24, end_on_surfacing=False, verbose=True)
# Save the results in a file for later analysis
GM.save()
The output is written to a netCDF file (default) or to pickled files. The netCDF file contains a long list of variables. Most of them start either with c_ m_ u_ or x_ and are equivalent in meaning with respect to the glider parameters. Some other parameters are added for diagnosis.
For example, the glider depth is represented by the variable
m_depth, and can be plotted as function of time (in seconds since
1970) using the variable m_present_time, which is in the netCFD file
equal to the time dimension. The script examples/validate_model.py
compares the output generated with the glidersim model with the actual
glider data files for the two simulated missions.
The glidersim software is not a one-to-one implementation of the actual glider software. The actual glider source code is not only rather complex, but also closed source. The glidersim code implements only the essential bits of the glider software, required for simulating normal operation. The implementation is on the basis of how I think the glider software is programmed, or how I think the glider behaves (from experience). This software comes as is, see also the GPLv3 license.
Lucas Merckelbach
[1] Merckelbach, L., A. Berger, G. Krahmann, M. Dengler, and J. Carpenter, 2019: A dynamic flight model for Slocum gliders and implications for turbulence microstructure measurements. J. Atmos. Oceanic Technol. 36(2), 281-296, doi:10.1175/JTECH-D-18-0168.1
[2] Merckelbach, L.: 2016 Depth-averaged instantaneous currents in a tidally dominated shelf sea from glider observations Biogeosciences 13(24):6637-6649 DOI: 10.5194/bg-13-6637-2016