1. Introduction
Landslides are extremely important geological/geomorphologic processes. Although they are natural evolutionary processes of relief, they often claim many victims and/or cause significant material losses, as well as help unleash other events, such as erosion processes. Urban consolidation in mid-size and large towns and cities and the consequent disorganised land occupation in the past few decades are the prime factors responsible for catastrophes in various regions of Brazil.
In land planning and organisation, one of most important stages involves zoning in accordance with different risk levels. To formulate zoning rules, it is necessary to distinguish between the concepts of susceptibility and risk. The first refers to the probability that a possible fact will affect a zone with certain intensity, irrespective of harm to the population. The concept of risk includes the possible existence of damages to the population, facilities, infrastructures or activities. Therefore, when zoning a piece of land, not only the susceptibility of the zones to natural phenomena should be considered but also the existence of residents, infrastructure, facilities and other factors in the area that is vulnerable and may be affected. Several studies have been made during these past decades to identify areas susceptible to mass movements, due to the large number of geological-geotechnical accidents that have occurred causing damage to the population.
When the study involves vast areas, it is preferable for the analysis to be done with the help of a geographic information system (GIS). A GIS is a tool that can organise georeferenced databases, address a large volume of data and reduce inaccuracy in comparison with the work done by hand.
The characterisation of the physical environment of a study area can provide key information for rational planning of the use and conservation of land and water. In this context, the use of a GIS permits the integration of the data more accurately and faster than the traditional analytical methods. A GIS also enables better use of existing data and from it can provide further information, thereby permitting more efficient organisation of actions. GISs are, in the environmental area, especially in developing or Third World countries, valuable tools for control and more rational use of scarce financial resources.
Developing computer techniques has endowed cartographic information processing with incredible speed and precision. The use of the tool in a GIS environment helps integrate information from various sources and on different topics. This is why it is extremely important for land planning and specifically risk management.
2. Methodology
Several proposals for risk assessment methods have already been made, but most of them basically consist of the production and use of preliminary maps, such as susceptibility and/or hazard charts. The model proposed here involves a procedure to assess a landslide risk based on a three-tier mapping structure: thematic, susceptibility or hazard and risk maps.
2.1. Thematic Maps
Maps or charts are cartographic documents used for various purposes. Thematic maps present information relating to one or more aspects of the biotic, human and physical environment.
Map production using geographic information system technology is an incredible advance in the area, by linking geographic data to alphanumerical data, the latter of which are normally represented in the form of tables, producing thematic maps that combine information with major benefits compared to the traditional systems.
One of the major problems when preparing charts and maps concerns the cartographic principles and content of each document. One of the aspects of the environment that is recorded in maps and charts relates to the physical environmental components, namely rocks, unconsolidated materials, water, relief, climate conditions and related aspects, such as vegetation, and so on.
Therefore, thematic maps are cartographic documents concerning the physical environment and are produced using physical data of the region and its current status that are the basis for undertaking a cartographic analysis. The proposed thematic maps in this model are those that record the spatial variability of physical aspects contributing to landslide occurrences.
There are a number of methods to prepare natural state maps. They consist of office work such as photo-interpretation and geoprocessing, as well as field work. The data come from earlier studies and aerial photos.
The topographical map is the basic document that must be available as a source of data. The restriction when making these charts is precisely the scale, range of the relief and equidistance of the contours where they appear. All data from the existing cartographic basis (topographical map) is put together to form a database. The information must be georeferenced to obtain the following maps:
• Gradient map;
• Vegetation map;
• Natural drainage map;
• Map of geological-geotechnical domains;
• Map of construction vulnerability;
• Regional geographic map.
The maps characterise a distinct theme that refers to the physical environment. Each theme consists of classes (attributes) that are associated with a deduced probability corresponding to specialist opinions regarding existing risk factors. Considering the absence of statistical data for the relationship between existing risk factors and landslide occurrences, probabilities are adopted, deduced by specialists, which express the confidence with which each attribute contributes, to a greater or lesser degree, to the likelihood of landslide occurrence. The existing risk factors correspond to the set of environmental, geologic and geometric conditions in which landslides will occur. The model in question does not include factors directly responsible for causing landslides (rainfall, erosion, temperature variation, etc.) due to the scarcity of data.
Probability by judgement is adopted to quantify the attributes in each thematic class, which is a way of formally capturing specialist opinions in figures and then combining these opinions in models. The uncertainty captured in this way has a numerical value that depends on the specialist’s personal skill in judging uncertainties developed from past experience.
Current experience suggests that at least in the early stages of specialist deduction, verbal descriptions are more intuitive than numbers. Thus, such descriptions are included as components within event or fault trees. Accordingly, approximate transformations between verbal descriptions and approximations by quantifying probabilities by judgement can be fixed for component events.
Table 1 gives a list of verbal descriptions with their values adapted from studies by [1,2]. These values were attributed in each thematic class to express confidence, by judgement of each situation contributing to landslide occurrence.