"Preludium 13" for innovative landslide identification

– I got into it while getting my Master’s degree. At first, I handled landslide scarps identification, then moved on to integration of different data type in order to predict landslides, next – after receiving Innovative Doctorate – I focused on integrating the data from airborne laser scanning and interferometric synthetic aperture radar in order to identify and monitor landslides, and now I’m moving on to identyfying landslides using the object method – explains Kamila Pawłuszek, a doctoral student at the Faculty of Environmental Engineering and Geodesy, who received a grant in the ”Preludium 13” project for research called ”Automation of landslide identification using the data from airborne laser scanning, object classification and machine learning”.

– The object method, which offers an approach close to human perception has recently revolutionized the field of remote sensing, it allows to transform what humans see into a library of mathematical, logical, geometrical rules that a computer understands. Unlike the conventional methods of classification based on pixels, OBIA is based on segments that create homogenic and conceptually logical segments or objects. Such objects are much better for illustrating real world elements than geometrical configurations of single pixels are – explains Pawłuszek.

– In case of landslides it’s going to work like this – we’ll take into consideration some distinctive features of landslides and dependencies between their particular elements. Distinctive topographic traces typical for landslides are reflected in point cloud of laser scanning. I’m going to enter the landslide descriptors library, created in cooperation with geologists, into the computer and, based on that group of rules and their hierarchy, the software will identify on the numeric area model of laser scanning the areas where landslides occur – adds the student. This method was put into use along with satellite optical images but the basic limitation for its use in Polish conditions is the lack of vegetation disappearance within landslides. In Poland it’s rare for the vegetation within a landslide to disappear completely.

Most of the landslides in Poland occur in the Carpathian Mountains due to their flysch structure – on average there’s one landslide per every square kilometre. In 2010, after heavy rainfall, the landslides posing a threat to buildings occured in 107 municipalities in the Carpathian area. Almost 2300 buildings were damaged, 560 of them destroyed completely. That’s why landslides are identified, monitored and predicted so that no investments are realized in dangerous places or so that people can get prepared to deal with the danger.

As part of the project, Kamila Pawłuszek has planned the purchase of proper software, two consultation-oriented trips to Italy and two visits in the Polish Geological Institute. – It’s one thing to do remote work, with abstract data, it’s another thing to do fieldwork, that’s why I’ll go to the institute to see geologists work, to observe how they identify landslides in the field, and to try to translate that identification into mathematical, geometrical, and logical sets of rules – she explains. When it comes to Italy, I went to prof. Tarolli as part of the Innovative Doctorate. During the project, I’m going to consult the results of my research twice so that I’m sure I’m heading in the right direction. And I need to admit that receiving  both the Innovative Doctorate and the ”Preludium” grant gave me a confidence boost and motivated me to put a lot of the ideas buzzing around in my head into action.

mj