5. THE SOFTWARE TOOL “REviSED”
The interface tool to link remote sensing data and oil
drift model is REviSED (REmote SEnsing interface for
Drift modelling). This core element of the processing
chain was developed at BfG within the project using
Interactive Data Language (IDL). It generates all
necessary drift model input files using either aircraft or
satellite data. In the following section, the program
workflow is described in detail:
When started, REviSED searches for input files and
accesses the data. If a polygon file exists, the tool will
create the model input files for Level-II processing.
Otherwise, only the basis file with centre coordinates
for Level-1 processing will be generated.
For Level-II processing, spill information like date and
time of observation are taken directly from the remote
sensing data (no specification file required). Further
parameters are either read from the specification file -
as far as provided - or from the default file.
The oil amount is an important input parameter for
predicting the oil distribution. It can not be directly
determined from radar data. If the oil amount is
estimated during an aircraft mission, the assumed value
can be written to the specification file by the user.
Otherwise, either the default value (e.g. 10 tons) may be
used or the volume can be estimated using an areavolume-relationship
based on Fay’s equations ([15],
[12]) by setting the keyword FAY as parameter in the
basis file for the oil amount. Fay [15] assumes, that
when the spreading process stops the polluted area
reaches a maximum extent A max :
5 3/4
¡ - 1<>- i - (1)
Using this relationship for calculating the oil volume
(V) from the area of a pollution (taken as A niax ), implies
that the spreading process lias already reached its end
and thus the area of the spill lias reached its maximum
at the time a radar image is taken. Though this will not
be tme in many cases, this approach provides a
conservative estimate of the volume, that means the real
volume can only be the same or higher than the estimate
(assumed undisturbed spreading and if the correct oil
type is applied). The explanation is, that if the spreading
process continues the oil thickness decreases and if it
stops, it reaches a minimum. Thus, if there are two
(theoretical) slicks with the same surface area, there will
be more oil per area if the spreading is ongoing, than if
it lias stopped.
For Level-II processing, the approach that is used to
generate the particle cloud is to equally distribute a
number of maximum 1000 particles over the area
enclosed by a polygon (Fig. 7). For this, the existing
IDL’s polyfill v algorithm to raster the vector data is
enhanced to come as close as possible to a number of
1000 particles and to include all edges. Nevertheless,
there are - depending on the polygon - sometimes a few
more particles generated. In this case, the excess
particles are deleted by random selection (Fig. 8, orange
symbol in zoom). After the particle cloud is generated,
the distribution’s centre of mass is calculated and builds
the first point of the drift path. As a last step, REviSED
creates all drift model input files.
Figure 7. Example of drift model run: Level-II
processing. 1000 particles are distributed over a
shapefde using REviSED algorithm (zoom). The new
center coordinate is also calculated (red star) and the
dri ft path is shown.
6. CASE STUDY
In the selected case study, there is one pair of
observations: one observation by satellite and one -
with a certain time delay - by aircraft. The latter
confirmed the possible pollution detected with help of
the satellite service as a real mineral oil spill.
Starting with the satellite sliapefile, REviSED and the
oil drift model were nm assuming the standard
parameter for oil age of 3 hours. All figures concerning
oil spills and drift model outputs shown in this paper
refer to this same oil slick. E.g. Fig. 3. shows the SAR
and SLAR images and Fig. 8 shows the generated
particle cloud and the computed drift path.
Within the project, such pairs of observations are used
to validate the model results. Fig. 8 shows the location
of the oil spill at time to when detected by the CSNservice
and the aircraft observation at time ti three hours
later.
