Study Group 2nd BEES Group – ESGI 134. Problem from Koblenz-Touristik
This year our BEEs project (see the previous one) was continued with a very nice study group in Koblenz. Twelve participants from St. Petersburg took part in the event. Here we would like to present couple of problems considered there. The first one was proposed by Koblenz-Touristik and called “The flood prediction problem”.
The famous Deutsche Eck (German corner) with monument to German Emperor Wilhelm I is located in Koblenz where the Mosel river joins the Rhine. Because of the whims of the weather this location often suffers from extreme rising of water level. Such situations produce nice photos when only the emperor remains dry.
On the opposite side of the Mosel river the city of Koblenz runs a camping-group with certain special installation with bathrooms and toilets.
In case of high water levels and the danger of flooding these installations need to be removed. So, the question was whether one could predict a critical time when these installations have to be removed using only the past data on the water levels?
The initial data for this task consisted of the detailed water level measurements for several years, heights map of the region and so-called critical water level value (when the installations are to be flooded). In the same way it is known that the removal procedure usually takes at least 8 hours. That means that the crane has to be called in 8 hours before the flood.
The proposed solution (T. Pogarskaia, M. Churilova, S. Lupuleac, O. Minevich, M. Petukhova, N. Zaitseva, J. Shinder) was based on the analysis of the available measurements and the main observation was that the rising of the water level usually had a linear character. The team proposed to approximate the water level in the future by a linearization procedure. Based on already known values of the water level different approaches of derivative estimation were tested and, moreover, for accurate estimations, the smoothing methods were used to get rid of the noise in original data. The approach was tested on the available data sets for some years and the results gave quite fairly accurate estimates of the future rising water level.
The team created an application that calls the alarm clock at the right time before the flood based on incoming data on the current water level. This application allows to specify the critical level value and the prediction time interval (safe time), thus, it is rather customizable.
Postscriptum: The last flood happened in Koblenz three weeks ago (see photo from Jan. 7, 2018). Unfortunately our recently developed software was not used…