Day 4

Detailed paper information

Back to list

Paper title Geomatics and EO for landslides: from mapping and monitoring to teaching
Authors
  1. Vasil Yordanov Politecnico di Milano Speaker
  2. Lorenzo Amici Politecnico di Milano
  3. Quang Xuan Truong Information Technology Faculty, Hanoi University of Natural Resources and Environment
  4. Maria Antonia Brovelli Politecnico di Milano
Form of presentation Poster
Topics
  • D1. Managing Risks
    • D1.01 Satellite EO for Geohazard Risks
Abstract text Landslides are one of the most dangerous and disastrous geological hazards worldwide, posing threats to human life, infrastructures and to the natural environment. In this domain it was initiated a joint project between Politecnico di Milano, Italy and Hanoi University of Natural Resources and Environment, Vietnam. The project is funded on the Italian side by the Ministry of Foreign Affairs and International Cooperation (MAE) and on the Vietnamese side by …... Its main focus is on the problem related to the landslides phenomenon which is relevant in both countries. The goal is to join efforts and experience in the field of geodata science, focusing on the most innovative approaches and designing and implementing sustainable new observation processing strategies. These include studying and applying new techniques for landslide susceptibility mapping through machine learning algorithms; and landslide displacement monitoring through earth observation satellite and UAV data, and citizen science applications for thematic data collection. Moreover, the project has an important consequence also in building new capacities, which will be transferred into the universities’ teachings and professional refresher training, with a direct impact on students and an indirect influence on technology transfer outside the academic environment.
Currently the project has been ongoing for almost a year and has already achieved the target milestones where the main results reached to the current date can be presented into four main tracks: (1) susceptibility mapping, (2) citizen science, (3) landslide monitoring, and (4) capacity building.

1. Susceptibility mapping.
Landslide susceptibility mapping is a topic of crucial importance in risk mitigation. A machine learning approach based on the Random Forests algorithm is adopted to produce landslide susceptibility maps over two areas in Northern Lombardy, Italy (Val Tartano and Upper Valtellina). Random Forests algorithm has been employed, because it has already proven its good performance in the field of landslide susceptibility analysis. As per standard procedure in susceptibility mapping, a landslide inventory (records of past events) usually is used to feed a model with information about the presence of an event; however, in many cases the information of absence is often neglected and usually it is represented by simply including areas for which are missing landslide records. As it can be considered an important aspect, it was introduced an innovative factor, namely the No Landslide Zone (NLZ) which was defined by geological criteria. The main aim of its introduction is to determine areas with a very low possibility of landslides. For that purpose, it was defined a threshold combining slope angle and Intact Uniaxial Compressive Strength of the terrain lithology:

(slope < 5°) OR [(5° < slope < 15° OR slope > 70°) AND (IUCS > 100MPa)]

Upon verification of its consistency the NLZs depicted an error in the margin of 1.7% for Upper Valtellina and 0.5% for Val Tartano. By these means, the model was provided with information about landslide absence in addition to that of past landslide events. The resulted susceptibility maps (i.g., Figure 1) were subsequently validated with state-of-the-art metrics, depicting very satisfactory results when NLZ was included.

2. Citizen science.
Landslide inventory is always a key factor in the hazard studies and as such it is crucial to be as a complete and up-to-date as possible. Most of the times they are lacking some past events, or simply the provided attributes are incomplete. In order to allow faster and more complete landslide data collection, it was developed an open-source thematic mobile application based on citizen science approach. The app allows any user with a mobile device to map and add information about past landslides, by sharing the location of it and compiling a standard geological questionary. Naturally, potential citizens that can contribute may have various levels of knowledge about the landslide phenomena, which was taken into account in the app by choosing between questionnaires related to non-experience users or to professional one. For accessing the collected data were developed two means. The first one is in the form of a plugin for QGIS which allows the user to directly download locally the collected records, including the landslides’ locations and related information. The second distribution mean is through a web application which allows simple data exploration in a map or tabular views (Figure 2). In addition, the webapp can visualize some statistics for the observations using the collected fields or to create a dashboard for a specific landslide.

3. Landslide monitoring.
Whilst susceptibility studies can be of great aid in preventing threats posed by future events, active landslides need to be monitored to reduce the risk of damages and casualties. With this aim, this work proposes a way to compute landslide displacements through time, by exploiting the great availability of high-quality multispectral satellite images. The developed procedure produces maps of displacement magnitude and direction by means of local cross-correlation of Sentinel-2 images (Figure 3). The Ruinon landslide, an active landslide in Upper Valtellina, was analyzed during two different time windows (yearly analysis between 2015 and 2020, monthly analysis in July, August and September 2019). The main preprocessing steps are starting from creating a suitable multi-temporal stack according to the AOI and cloud cover; image co-registration to ensure that the images become spatially aligned so that any feature in one image overlaps as well as possible its footprint in all other images in the stack; histogram matching to transform one image so that the cumulative distribution function (CDF) of values in each band matches the CDF of bands in another image. The main processing is based on the Maximum Cross-Correlation procedure implemented on master-slave couples of images. The approach needs an optimal moving window to test whether a location (pixel) from master is at the corresponding location (pixel) in the slave image, or it is displaced in the boundaries of the search window. The outputs are shifts (in pixels) in X and Y directions which are actually the distances required to register the window of the slave with the one of the master. The spatial resolution of Sentinel-2 images can be considered a bit lower for the landslide’s size under considerations. However, the implemented approach depicted the most major displacements during the landslide’s most active periods. To compare and evaluate the performance of the cross-correlation approach were used products from photogrammetric point cloud comparisons (provided by the local environmental agency ARPA Lombardia) created from UAV observations in periods close to the considered ones for satellite monitoring.

4. Capacity building.
In order to transfer the knowledge and experience, from the project activities, to students it was organized a joined course activities between Italian and Vietnamese partner universities, which are offered to 50 students from both countries. The activities comprehended two preparatory webinars that presented the problem of landslides in Vietnam and Italy. In addition, practical sessions are offered to all students involved to ensure a homogeneous basic preparation adequate to face the proposed project. The project focuses on the creation of landslide susceptibility maps and their presentation in a webGIS. Where the purpose of the project proposed is to analyze case studies, both in Italy and Vietnam, based on the new observation processing GIS strategies designed and implemented in the framework of the Bilateral Scientific Research project. The students are tutored together by Italian and Vietnamese tutors. Where it is expected that the outcomes from students’ work to be presented during a workshop organized by the project partners.