Operational Oil Spill Modelling: From Science to Engineering Applications in the Presence of Uncertainty - Mathematical Modelling and Numerical Simulation of Oil Pollution Problems - page 110

Fig. 5.4 HyosPy structure for automatic updating of forecasts and visualizations. Here each forecast
wind/tidal condition drives a single hydrodynamics and oil spill model, with each model based on
the latest available data and forecasts. Visualizations include both the latest forecasts and prior
forecasts
for HyosPy is the NOAA PyGnome model, which is a new Linux/Python version
of the GNOME operational oil spill model used by emergency response agencies
across the USA [ 4 , 33 ].
HyosPy was run as a real-time demonstration on Feb 4, 2014 using COS and
forecast data downloaded from the internet as it became available and an imaginary
oil spill near a beach in Corpus Christi Bay (Texas, USA). As shown in Figs. 5.5 and
5.6 , HyosPy produced tracks 3h apart, where each track is based on combination
of the latest available hindcast and forecast data with new instances of both the
hydrodynamic model and the oil spill model. The oil spill diffusivity coefficient in
these models is low, so the Lagrangian particle stay close together in each track,
which provides clearer visualization of how the model operates.
Key innovations of HyosPy are (i) automated re-running of the hydrodynamic
model as new data becomes available, and (ii) using successive forecasts in the visu-
alization to provide insight into how themodel predictions are changing as newdata is
added. Because HyosPy is a wrapper around models rather than a model itself, it can
be readily modified to include multiple perturbed forecasts, hydrodynamic models,
and oil spill models along each of the linear paths of Fig. 5.4 ; that is, we can implement
Figs. 5.2 and 5.3 by creating multiple model instances within the existing system. As
an operational approach, HyosPy could be set up with a single continuously-running
hydrodynamic model using the latest wind/tidal forecast data. When a spill occurs,
additional computer resources can be added to allow multiple hydrodynamics mod-
els and oil spill models to be run using perturbed forecast data. Recent tests have
shown that HyosPy can be operated as a web service with a continuously-running
hydrodynamic model [ 18 , 19 ]. For a real (or imaginary) oil spill, the user enters the
oil spill location, time, oil type, and quantity into boxes on a web site and HyosPy
produces, then automatically updates, a series of predicted spill tracks. HyosPy han-
dles the data manipulation, reformatting, and initiation calls between models without
user guidance, which makes the complex multi-model processing entirely invisible.