Gliderflight is a python module to calibrate a model that predicts the glider flight through water. The model results can be used to estimate the speed through water, a parameter which is required to compute turbulent dissipation rates from temperature microstructure or shear probe data, collected with a turbulence profiler mounted on top of an ocean glider.
The dissipation rate of turbulent kinetic energy is a parameter that plays a key role in many physical and biogeo chemical processes in oceans and coastal seas. However, direct oceanic measurements of turbulence are relatively scarce, as most observations stem from free-falling profilers, operated from seagoing vessels.
An emerging alternative to ship-based profiling is the use of ocean gliders with mounted turbulence profilers. A required parameter in the processing of microstructure shear and temperature measurements is the speed of flow past the sensors. This speed can be measured directly with additional sensor, such as an electromagnetic current meter or mounted acoustic Doppler current profiler, but often gliders are not equipped with additional velocity sensors. Alternatively, a glider flight model can be used to estimate the speed through water. Such a model is described in the paper A dynamic flight model for Slocum gliders and implications for turbulence microstructure measurements [merckelbach2019]. This Python model implements the steady-state and dynamic glider flight models, described therein.
Documentation of this software package can be found at http://dockserver0.hzg.de/software/gliderflight/index.html
The steady-state model implemented, considers a horizontal and vertical force balance. Vertical forces are a balance between buoyancy, gravity and the vertical components of the lift and drag forces. The horizontal force balance consists of the horizontal components of the lift and drag forces only. These two equations can be solved for the angle of attack and the speed through water, determining the flight at any instance of time.
Input to the model comes from parameters measured by the glider, such as the measured pitch angle (m_pitch), buoyancy change (m_ballast_pumped or m_de_oil_vol) and the in-situ density. Furthermore, the model requires the specification of a number of coefficients:
- mg: mass of the glider (kg)
- Vg: volume of the glider (m^{3})
- Cd0: parasite drag coefficient
- epsilon: compressibility of the hull (Pa^{-1})
- ah: lift angle coefficient due to the hull (rad^{-1})
- Cd1: induced drag coefficient (rad^{-2})
Using the depth-rate from the pressure sensor as only model constraint, the mass (or glider volume) and the parasite drag coefficient can be determined. To determine the lift angle coefficient requires an additional constraint that contains a horizontal velocity component. Details of this procedure are given in XXXXXX.
In addition to a steady-state model, this code also implements a dynamic model, that is, including the inertial terms. Since this model needs to be integrated, for which the Runge-Kutta method is used, it is more computational expensive. The dynamic model produces more accurate results when forcing conditions change rapidly, such as when crossing a sharp pycnocline or during the transition from dive to climb. Apart from the mathematical model underlying, the interfaces to both models are the same.
To calibrate a model, either steady-state or dynamic, we may wish not to include all the data in the evaluation of the cost-function. To that end, data can be masked. The Calibrate class provides boolean operators to do this:
- OR()
- AND()
- NAND()
By default a mask set to False for all data. To mask data for which a condition evaluates to True, the OR() method should be used. For example,
gm = SteadyStateCalibrate(rho0=1024) gm.set_input_data(datadict) condition = depth<10 gm.OR(condition)
which would exclude all data points for which the depth is less than 10 m from the evaluation of the cost-function.
A truth table:
| mask | conditon | OR | AND | NAND |
| 0 | 0 | 0 | 0 | 1 |
| 1 | 0 | 1 | 0 | 1 |
| 1 | 1 | 1 | 1 | 0 |
| 0 | 1 | 1 | 0 | 1 |
An example to calibrate a model:
# create a dictionary with the data
data = dict(time=t, pressure=P, pitch=pitch, buoyancy_change=deltaV)
gm = SteadyStateCalibrate()
# we have to define mass and volume at the minimum
gm.define(mg=70, Vg=70)
gm.set_input_data(data)
# mask all data below 10 m
gm.OR(pressure*10<10)
# mask all data exceeding 60 m
gm.OR(pressure*10>60)
result = gm.calibrate("mg", "Cd0")
print("Calibrated parameters:")
for k,v in result.items():
print("{}: {}".format(k,v)
# Instead of printing the parameters from the results, we could also
# get them from the corresponding attributes: print("Cd0:", gm.Cd0).
print("Cd0:", gm.Cd0)
# We also don't need to run the model again either. The model output
# is also accessible from attributes:
#
# gm.t # time
# gm.U # incident velocity
# gm.alpha # angle of attack
# gm.ug # horizontal speed
# gm.wg # vertical speed
# gm.w # vertical water velocity
# if we want to run a model with a given set of parameters
fm = DynamicGLiderModel(dt=1, rho0=1024, k1=0.02, k2=0.92)
# copy the settings from the steady state model
fm.copy_settings(gm)
solution = fm.solve(data)
# solution is now a named tuple, according to the definition:
# Modelresult = namedtuple("Modelresult", "t u w U alpha pitch ww")
When you publish results that were obtained with this software, please use the following citation:
- Copyright (c) 2018, 2019 Helmholtz Zentrum Geesthacht, Germany
- Lucas Merckelbach, [email protected]
Software is licensed under the MIT licence.
| [merckelbach2019] | 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 |