TSU is creating a neural network for forecasting the spring flood

The Faculty of Geology and Geography and the Faculty of Innovative Technologies are testing a neural network created to assess the risk to human settlements during the spring flood. A computer program analyzes data accumulated over more than 20 years, identifies trends, and correlates them with operational hydrometeorological data. After the results are processed, the neural network provides information on the dynamics of the water level in the risk zone and predicts which areas may be flooded in the next 48 hours.

- The neural network is now monitoring the situation that is developing around two settlements on the banks of the Ob River, Molchanovo and Nikolskoye villages,- says Vadim Khromykh, one of the authors of the project and Associate Professor at TSU. - If earlier the forecast was built fairly simply, based on the water level in the river at gauging stations and the distance to the village, now the neural network analyzes many factors, including meteorological data, snow reserves, data from gauging stations, and the influence of large tributaries of the Ob that contribute to flooding. The computer program makes a much more accurate calculation of current water levels at different points in the area and gives a short-term forecast for the next two to three days.

The control of the flood situation in other localities of the region is managed by the Ministry of Emergency Situations in the Tomsk Region using the Geoportal, a software package created by TSU in the interests of the regional administration and emergency response services. This year, detailed models of more than 30 villages located on the bank of the Ob River and in the flooding zone have been added to the geographic information system (I-GIS).

Geoportal in real time in the format of 3-D-models gives information about the current and critical water level near each settlement, the total population in it, the number of people registered in houses in the flooded area, and in a separate line the number of children.
- Now we are completing an address database, which includes data about each house, about the tenants registered in it, and in the future, maybe even their phone numbers,- says Vadim Khromykh. - This base will be added to the I-GIS system. If an evacuation is required, the Emergencies Ministry will be able to send messages to the cell phones of the residents of a community threatened with flooding. Early warning will help avoid situations in which the villagers have found themselves, for example, in 2010, when during a severe flood people did not have time to collect things and secure livestock.

As Vadim Khromykh notes, the evacuation process itself is often fraught with panic, because people do not know which way is currently open and safe. The TSU scientists will deal with the solution to this problem with colleagues from Tokyo Metropolitan University. The researchers plan to develop an automated warning system (real-time model): an interactive online map, which at each moment of time highlights in different colors the flooded area, the area free of water, and the paths through which people can leave the risk zone.
- Our colleagues from Tokyo have created such a system for some settlements in their country because Japan is characterized by mudflows, earthquakes, tsunamis, and the associated flooding of large areas, - says Vadim Khromykh. - We are going to try a new approach using the example of one object, the village of Chernaya Rechka, which is located near the Tom' River and often finds itself in the flood zone.

TSU geographers will implement this idea together with Daichi Nakayama, a scientist at the Tokyo Metropolitan University (TMU), who will soon arrive at the University on a working visit. Nakayama is one of the leading experts in GIS technologies and the author of the Japanese version of the decision tree, a computer model that predicts landslides.

Provided by Tomsk State University