Application of the AHP method for prioritizing actions to reduce risk associated with gravitational mass movements in areas along the margins of watercourses of the city of Rio Branco, Brazil

Problems related to gravitational mass movements at the margins of watercourses are present in different parts of the world, but in the Amazon, whose hydrological regime has unique characteristics, and whose socioeconomic situation may be considered underdeveloped, such problems come into sharp focus. Rotational landslides and creeps on the ground are present with high frequency and magnitude in several municipalities in the Western Amazon. Such processes act in association and annually cause significant losses and material damage. In this context, this study proposes the structuring of a decision-making process that aims to assist in the indication of critical areas, as well as proposing possible actions to mitigate risks. The criteria and procedures adopted for risk mapping in a portion of the urban center of Rio Branco-AC are described. Risk was measured, based on the multicriteria analysis of attributes related to the physical, anthropic environment and the characteristics of the processes considered. The hierarchical analysis method was the technique adopted to structure the risk assessment model, which was later integrated into the GIS environment, to allow spatial analysis. The results were presented in the form of cartograms, descriptive tables and graphs, and aim to help control the mapped risks.


Introduction
In Brazil, in particular and in the world, in general, uncontrolled growth and high rates of urbanization contribute to increasing the vulnerability of the exposed population and the potential for damage to infrastructure and facilities, which makes the issue of risk assessment and analysis, an essential step for the adoption of policies aimed at reducing natural disasters.
As a result, several authors (Cerri et al. 2007;Faria and Augusto Filho 2013;Sobreira and Souza 2012;Nola and Zuquette 2021;Wadadar and Mukhopadhyay 2022) have been developing methods and techniques (progressive detailing, cumulative precedent experience, etc.) to map risks, especially those related to external geodynamics processes.
In Brazil and in many other countries, risk mapping is based on predominantly qualitative assessments, in which the situational risk is diagnosed by the team that performs the mapping, based on field surveys, through surface geological-geotechnical investigations (Cerri et al. 2007;Faria and Augusto Filho 2013). However, more recently, several researchers (Cascini et al. 2005;Jasiwal 2011;Vega and Hidalgo 2016;Nola and Zuquette 2021) have developed procedures to allow some quantitative risk assessment.
Observing the diversity of adverse processes (falls, landslides, creeps, inundations, flash floods, etc.) and exposed elements (populations, properties and goods, economic and social activities, etc.), it is possible for a specialist to understand that such a diagnosis needs to go through the evaluation of the phenomenon based on an appropriate interpretation of its controlling factors; there is a constant need to describe them in a systematic way, compatible with the scale to be applied.
Therefore, it is essential that there is some prior knowledge about the dangerous event (its type, mechanism, material involved, magnitude, speed, trajectory, severity, destructive power, etc.) and the vulnerability of the exposed element.
Even recognizing the inaccuracies and uncertainties inherent in qualitative assessments, Cerri et al. (2007) emphasize the importance of this activity for the effectiveness of an intervention policy.
To improve the risk mapping approaches, Cascini et al. (2005), Jasiwal (2011), Faria and Augusto Filho (2013), and Vega and Hidalgo (2016) considered pertinent to suggest and incorporate quantitative techniques that can contribute to reducing subjectivity.
Although the indeterminacy of an ideal solution may be evident, since there are no unique constraints, or even a single perception of the problem, it is essential that the diagnosis presents some level of hierarchy regarding the solutions considered and generates assistance in allowing the validation of the results. This hierarchy should be based on qualitative criteria (type of movement, characteristics of land use and occupation, construction standards, etc.) and quantitative criteria (number of properties and affected population, volume that can be moved, affected area, engineering intervention costs, etc.).
In this sense, there is the perspective that the structuring of a decision-making process, based on the use of the decision support routine called Analytical Hierarchy Process-AHP, can simplify the valuation of the conditions, reduce inconsistencies and provide assistance that allow a more objective indication of risk situations.
The results presented in this article are related to a study developed with a focus on the indication of critical areas regarding the risk of gravitational mass movements in areas adjacent to rivers and streams in the Brazilian Amazon hydrographic system, especially creeps or rotational landslides. This research has as its central premise the fact that the social and economic damages related to such processes (creeps and rotational landslides) are significant in areas adjacent to the watercourses of Amazonian cities, and that the knowledge regarding the mechanisms and principles that govern them is still incipient (Carvalho et al. 2009;Labadessa 2011;Matos and Cursinho 2012;Pacheco and Brandão 2012;Magalhães and Gomes 2013).
The perception regarding the use and occupation of land in these areas, as well as the precarious infrastructure and the lack of uniformity in the language and criteria adopted, has limited the efficiency of the strategies necessary for adequate planning and coping with risk in these spaces.
In this context, the objective of this work was the development of a procedure based on the Hierarchical Analysis Method, capable of evaluating, quantifying and hierarchizing, the conflicting relationships between the use and occupation of areas along the margins of watercourses in Amazonian cities and the magnitude of dangerous phenomena, to then indicate a set of mitigating actions.
The research was carried out in a test area of the urban core of Rio Branco (AC), located in the Northern region of Brazil. This area is known for presenting frequent creep movements and rotational landslides, in an area with materials with geological and geotechnical characteristics that shows high susceptibility to the occurrence of such gravitational mass movements (Maia 2003;Oliveira and Ferreira 2006), in addition to occupations with different patterns and stages of consolidation. The results presented in the form of cartograms, descriptive tables and tables that aim to help control the mapped risks.

The hierarchical analysis method (AHP)
The delimitation and assessment of areas subject to the occurrence of dangerous events and, therefore, areas with geological-geotechnical risk depend on knowledge and surveying of the attributes that influence the predisposition and outbreak of the considered processes (Rodrigues and Zuquete 2006;Sobreira and Souza 2012;Salehpour et al. 2021;Wadadar and Mukhopadhyay 2022). The great challenge in this type of assessment consists in relating, in a systematic way, the characteristics of the physical environment and the anthropic actions, with the signs of movement identified to, then, indicate the potentialities and limitations of the different terrains. An important aspect to be considered with regard to the characterization of the physical environment refers to the identification of the processes responsible for the observed events, which are responsible for the appearance/creation of the areas of geological-geotechnical risk.
AHP has been applied in several areas of knowledge and has proved to be extremely useful in the analysis of geological-geotechnical risk of mass movements and in other natural disasters, especially in situations involving subjective judgments, as pointed out by the studies by Rodrigues and Zuquete (2006), Intarawichian and Dasananda (2010), Paula and Cerri (2012), Kayastha et al. (2013) and Faria and Augusto Filho (2013).
AHP is a technique based on a decision tree, proposed by Saaty (1970), which aims to reproduce human reasoning in the comparison of elements, as a tool to aid decisionmaking in complex problems.
Despite the structuring of the problem in a more simplified form, the method does not rule out the need for specialists to carry out decision-making (specialist knowledge) in regard to the problem, but it has been used in several works cited in the academic literature.
The criteria are selected and arranged in a descending hierarchical structure in the form of an inverted decision tree, whose structure descends from the goal (objective) to the criteria, sub-criteria and alternatives, in successive levels. Rodrigues and Zuquete (2006) and Paula and Cerri (2012) summarize the recommended steps for applying the method: 1. Define the problem, that is, what one seeks to know (risk, susceptibility, etc.); 2. Create a hierarchical structure, by breaking down the problem into different levels, in which the top is occupied by the goal (general objective) and the last level by the more specific elements, such as attribute classes or intervention alternatives; 3. Build decision matrices for each level, make parity comparisons and define the relative importance of each element. Saaty (1990) proposes that the importance of one element in relation to another, be defined based on a scale that varies from 1 to 9, where 1 refers to the elements in the matrix that present the same importance, and 9 when one element presents a higher degree of influence over another, to explain the problem; 4. Check the consistency of the judgment in each matrix; 5. Obtain the vector of priorities for each matrix and perform the global assessment of the problem through the weighted linear combination of the various elements of each hierarchical level Saaty (1990).
Conducting a sensitivity analysis, in which the consistency of a positive reciprocal matrix requires that its maximum eigenvalue (λ max ) is very close to the dimension of the matrix (n), results in a greater consistency of the data, insofar as how much the closer λ max is to n, the more consistent the result will be (Gomes 2009). The Consistency Index (CI) indicates how far the given value is away from the expected theoretical value n, in which this difference is measured in relation to the number of degrees of freedom of the matrix (n − 1). Thus, the Consistency Index is given by Eq. (1).
If CI is small enough, the decision-maker's comparisons are likely to be consistent. The degree of inconsistency is also verified by calculating the "Consistency Ratio" (CR) which is given by Eq. (2).
where RI is the random consistency index for the value n (matrix order number), which represents the value that would be obtained in a reciprocal matrix of order n and in which no logical judgments were made.

Definition of the study area
The study was applied to a portion of the urban core of Rio Branco (AC), in a representative region of areas adjacent to rivers and streams in the Western Brazilian Amazon (1) CI = max − n n − 1 (2) CR = CI RI ( Fig. 1), where events related to creeps and rotational landslides annually cause significant losses and material damage. This area, with a surface area of approximately 1.1 km 2 , involved the city's historic center, the edge of the river Acre and five precarious settlements. It presents a sensitive diversity of densities and occupation patterns, which are inserted in the same geomorphological context (Rio Branco depression) and presents degrees of susceptibility, that vary from low, to high, to the development of creeps and rotational landslides (Nascimento 2016). This diversity of circumstances was considered adequate for the simulation of different risk scenarios.
Acre River is a 100 m width channel, comprises the largest drainage of Rio Branco municipality and presents flows that can vary from 50 to 900 m 3 /s on the same hydrological year.
Cenozoic sedimentary rocks of continental origin, composed by claystones and siltstones, massive or very laminated, from Solimões Formation, sustain local morphology (Brasil 1976).
Rio Branco depression is characterized by its suave hills with flat to slightly convex tops and "V" shaped valleys with low dissection degree. Soils are mainly plastic, with clay and silty clay texture laid over sedimentary rocks from Solimões Formation, forming an angular unconformity (Brasil 1976;Oliveira and Ferreira 2006;Borges 2007Borges , 2019. Although block fall into the river bed is frequent, creep and rotational landslides, mobilizing both soil and rock material, are the most significant mass movements processes of the region, both for its magnitude as for its coverage, as well as for its associated damage (Nascimento and Simões 2017).
To support the application of the proposed method, the authors initially sought to characterize the types of gravitational mass movements (their typology, mechanisms and materials involved) based on an integrated analysis of the components of the physical environment (relief, watercourses, scars features) previous events and geological materials involved). In this context, two ranges of land adjacent to the main watercourses in the urban area of Rio Branco, the Acre River and one of its tributaries, the São Francisco stream, were selected for this activity.
The mapping work was carried out by Nascimento (2016), in the zoning phase, which allowed inferring that creeps and rotational slides are the most significant processes, both by the frequency with which they occur and by the magnitude of the registered events.
To classify the susceptibility of the terrain, the area described in Fig. 2 was subdivided into homogeneous physiographic units, considering aspects such as slope, geological substract and unconsolidated (Soil) cover.
Field works and photointerpretation performed by Nascimento (2016) on 2012 and 2014 resulted in the identification of 73 instability features, so allowing the recognition that creep and landslides processes are intrinsically associated with the effects of rain infiltration and fluvial dynamics (erosion and water level fluctuation).
The creeps are associated with the plastic soils that cover the hills of the Solimões Formation, occur more frequently in terrain with a slope greater than 12% and are not limited to the flooded section of the channels.
Although rain intensity can be recognized as a slope instability agent, this variable was not considered in zoning, as data used did not allowed the authors to differentiate any Fig. 2 Geotechnical zoning of mass movements susceptibility. Source: Nascimento (2016) relation, both in time and in space, between this factor and the presence of the features identified in the study area.
The influence of the fluvial dynamic was considered by introducing factors related to water velocity distribution (margin longitudinal profiles) and to the compartment area subjected to seasonal flooding.
It is known that in river channels with meandering pattern, the maximum velocity and turbulence zones are located in the vicinity of concave river margins, decreasing toward lower depth and convex margins. So, the longitudinal profile was used to describe, in a qualitative way, the influence of fluvial processes (erosion and sedimentation) on the global stability of each compartment, defining the following classes: Missing, Concave, Convex, Straight and Sinuous.
The area subjected to seasonal flooding was estimated by considering elevations 132 and 134 m of topographic map, which are commonly used by Rio Branco municipal Civil Defense to identify areas susceptible to exceptional or seasonal flooding. Three classes were defined for this factor: NI (No flood zone), comprising compartments totally inserted on superior elevations to the major riverbed; PI (Partially flood zone), comprising the compartments the present a maximum of 1/3 of its area on the major riverbed; and I (flood zones) associated to the units totally inserted on the minor river bed or with at least 1/3 of its area on the major riverbed.
The zoning map is the result of the combination between the potential of homogeneous physiographic units to develop events, and the considered induction factors, by using weighted overly techniques including analytic hierarchy process (AHP) and weighted linear combination (WLC).
According to Nascimento (2016), rotational landslides are frequently observed in flooded areas and also on slope surfaces already weakened by the creep, which are often induced by the intense fluctuations of river discharges that coincide with periods of intense and prolonged rain. There are depressions between 0.4 and 2.2 m high and traction cracks that are hundreds of meters long at the top and mid-slope. The evidence of creep is widespread, with an increase in its rate of displacement during the rainy season, evidenced by the misalignment of the foundations of wooden buildings (Fig. 3). The movements take place in predominantly silty clayey soils, as described in Fig. 4, which represents section AB (indicated in Fig. 3), which was adapted from the work of Nascimento (2016).

Materials
The basic reference information obtained in this research comes from the Territorial Information System-SITgeo (PMRB 2013). The main materials obtained from SITgeo were as follows: • Digital mosaic of aerial photography from 2013, with a spatial resolution of 5 m, on a scale of 1:5000, developed by Aeroimagem S/A and obtained through a digital aerophotogrammetric camera-LEICA ADS-80; • Digital topographic map of the urban area, with contour lines equidistant at one meter on a scale of 1:5000, obtained by airborne laser profiling, in 2013; • Registry map on a scale of 1:5000, from the year 2013, containing information regarding plans for public places, lots, paved roads, land use, rivers, neighborhoods and flood zones; • Fluviometric data from 1970 to 2014 provided by the State Civil Defense Coordinator; • Geotechnical zoning database for susceptibility to gravitational mass movements developed by Nascimento (2016): shape files, topographical and geological feature records, field records, geological-geotechnical sections and 23 standard penetration test logs; and The Geographic Information System (GIS), of the ArcGIS 9.3 program, an ArcMap operations module developed by ESRI, was the tool used to manipulate all available or generated information plans. This took place by using projection "Universal Transversa de Mercator"-UTM, WGS-84, zone 19S.

Steps and methodological procedures
The following is a detailed description of the methodological procedures used in carrying out the work.

E1: Definition of the land unit
In general, the susceptibility maps provide important aid in planning the occupation of territorial space, especially with regard to the potential of a region's physical environment for the development of a given process.
However, it must be understood that these cartographic products have limitations related to the definition of actions necessary to mitigate geological risks and natural disasters. For there to be an indication of where, and what actions will be necessary to face the risk, in the short and medium term, it is necessary that the study be conducted on a higher, more detailed, scale, which in addition to the natural terrain susceptibility, also considers the degree of development of the processes and their relationship with anthropic induction factors and the vulnerability of the exposed elements.
Considering the diversity of uses, occupations, stages of consolidation, installed infrastructure, processes and their stages of development, it was decided to select a regular area, named cells, with dimensions of 50 by 50 m, to represent the smallest unit of risk mapping, which allowed the tabulation and cross-indexing of information related to the criteria defined within the AHP systematics, without the concern of overestimating or underestimating the extent of each evaluated criterion.
The study area was, therefore, subdivided into 467 cells that constituted the information plan called Regular Grid. This information plan was generated in vector format, in the Arc-Gis program, using the Vector Grid operator of the GeoWizard 11.2 extension.

E2: Risk breakdown
The decision problem (risk assessment) was broken down into four levels of hierarchy, as can be seen in Fig. 5. The first level was occupied by the components that determine the risk in the land unit, that is, the Hazard Index (HI) and the Vulnerability Index (VI); the second level was composed of the indicators of these components (Natural Susceptibility, Potential of Induction, Potential of Destruction, Degree of Resilience and Potential of Damage); the third level was made up of the attributes (natural, anthropic and process characteristics); and the fourth level, and last one, by the classes of each attribute.
The susceptibility indicator was obtained by overlapping the Zoning Information plan with the Regular Grid information plan. Although this indicator allows estimating favorable or unfavorable characteristics to the development of the considered processes, the presence of signs of movement in a given cell may indicate that it already has all the characteristics necessary for the development of a geological mass movement.
In this case, the quantification of the hazard must consider the level of development of the process, which can be estimated based on signs of movement identified through fieldwork.
The information related to the signs of movement was obtained based on overlaps between the shapefile of Regular Grid, with the registration of unstable areas carried out by Nascimento (2016); the contour lines with a frequency of 1 m (PMRB 2013). This allowed identification of rugged regions, landslide features and those areas with evidence of creep. The level of development of the mapped signs and evidences was based on the field record information collected by Nascimento (2016).
The record of these occurrences in the geographic database was carried out by inserting a column for each attribute related to the characteristics of the mapped processes, in the information plan of Regular Grid. The magnitude of the evidence and the area of influence of these signals in the cell were considered as necessary attributes to estimate the Potential of Destruction indicator.
The magnitude of the signs of movement was assessed based on criteria presented by Cerri et al. (2007), as follows: • Absent-there is no evidence of the development of instability processes; • Incipient-evidence of instability suggests that the process is in an early stage, develops slowly or is dormant. In this case, monitoring is perfectly possible; • Significant-there are indications that the process is in full development. There are diverse abatement measures, cracks in the terrain and / or in buildings, or even scars from past landslides. Monitoring of the area is still possible, but it is perfectly likely that landslides will occur during intense episodes of rain and / or sudden variations in level in the watercourses; and • Severe-the evidence or features of instability are expressive, indicating that landslides may occur at any time, and it is not possible to monitor them.
For the evaluation of the area occupied by the signs or features of instability in the cell, the following classes were defined: absent; < 25%; 25-50%; 50-75%; 75-100%. For estimating the interference of human action in the evolution of processes (Potential of Induction), the possible damages to properties and goods as a result of a destructive process (Potential of Damage) and the capacity of population to restore its condition prior to a geological accident (Degree of Resilience), seven anthropic attributes were defined. These attributes were as follows: type of occupation, technical-constructive mode of implantation of buildings, area occupied by buildings and infrastructure, occupation pattern, surface water concentration, concentration of garbage/rubble on slope or escarpment and ground cover.
The information used for tabulation of classes related to anthropic attributes was obtained through fieldwork and from orthophoto of 2013, the information plans from the SitGeo database (lots, places, buildings and road system) and Google Street View images from the year 2012. In Fig. 6, there are images that illustrate some of these observations.

E3: Valuation of risk
In this stage of application development, the comparison matrices were set up for each of the levels mentioned and represented in Fig. 5. Then, comparisons were made between the elements of the same hierarchical level, with respect to the element of the higher level, defining, within a pre-defined scale of values (Table 1), the relative importance of a given element in relation to another.
In the peer-to-peer comparison, the weights of the elements of each matrix were defined by calculating the "given vector" or "priority vector", while the verification of the coherence of the judgments was performed based on the results of the "given value" calculation.
The judgments made are essentially related to the authors' experience, aided by consultations with specialists in the region studied.
To obtain the normalized weights, the maximum given value, the consistency ratio (CR) and the consistency index (CI), the authors used the APH module, from the ASSISTAT 7.7 Fig. 6 Images that illustrate the method used for gathering information of anthropic attributes statistical program (Silva and Azevedo 2009), using as input data the comparisons made in each matrix.
The parameter adopted to assess the consistency of the judgment made was the Consistency Ratio with a value less than 0.1 (Saaty 1990).
The values obtained in the priority vector of each matrix corresponded to the normalized relative weights assigned to each of the evaluated elements (classes, attributes, indicators and components).
After the tabulation of the attribute classes was completed, the Attribute Table of the Regular Grid information plan was exported to Microsoft Excel. Then, a spreadsheet was created containing the standardized weights of all attribute classes. With the aid of the Microsoft Excel VLOOKUP function, the attribute class contained in each cell was converted to its normalized relative weight.
The hazard (HI) and vulnerability (VI) indexes were obtained through the linear combination between the weights of the classes, attributes and indicators, according to Eq. (3).
where IC is the component's value (hazard or vulnerability) in cell i; RWI is normalized relative weight of the indicator present in cell i; RWA is normalized relative weight of the attribute in cell i; RWC is normalized weight of the attribute class present in cell i; n refers to the number of criteria.
The columns of the Excel spreadsheet that contained the values of HI and VI were inserted in the table of attributes of Regular Grid and they were later converted to the matrix format, for subsequent overlaps.
This procedure was performed using the Spatial Analyst module (Feature to Raster) from the ArcGis program. Subsequently, in the ArcGis program, the HI and VI values were divided into four different ranges (Natural Breaks option), representing, in ascending order, the low, moderate, high and very high classes for each component.
The results of this classification generated the Hazard and Vulnerability cartograms, which were validated through the overlay of the orthophoto of 2013, the digital topographic  (Nascimento 2016), based on the scenarios presented in Tables 2 and 3. The values of HI and VI were divided into four classes (low, moderate, high and very high), using the Natural Breaks function, from ArcGis. The risk evaluation (RI) resulted from the combination of these classes and was obtained through Eq. (4): where RI is degree of risk in cell i; HI is weight of the class of the hazard component present in cell i; VI is the weight of the class of the vulnerability component present in cell i.
The risk in a given cell was conditioned to the mutual existence of hazard and vulnerability. If there is no propensity for damage to occur due to a particular physical event, there is no hazard and, therefore, there is no risk. In order for this relationship to be translated into numerical terms, the matrix shown in Table 4 was constructed.
In this sense, it was considered that risk situations only exist when one of these components is not incipiently present, which is why a scale of 0 to 3 was defined.
The RI equal to 0 (low risk) indicates the least critical condition, in which, by maintaining the existing conditions, the possibility of social or economic damage is minimal, as it is unlikely that the event will occur, or due to the condition of low vulnerability of the exposed element. Areas with a low degree of hazard were considered to be stable.
The RI of 1-3 (moderate risk) indicates that, if the existing conditions are maintained, it is perfectly possible that social or economic losses will occur, especially in the months of more (4) RI = HI × VI It is unlikely that a destructive event occur in a one-year period Medium Hazard indicators are of medium potential to destructive processes. Building occupation and Consolidation of occupation should be carefully evaluated. The need for specific geotechnical design and monitoring must be adopted for high vulnerability areas High Hazard indicators are of high potential, in way that the occurrence of an event sis very likely to occur in one-year period. New occupation areas, including buried infrastructure must be avoided. Monitoring must be constant and preferably performed by using instrumentation on more vulnerable units Very High There are general evidences of creep and rotational mass movements. It is not possible to predict when new movements will occur Composed by consolidated or under consolidation areas. Damages associated to the occurrence of one instability process are significant and are essentially related to the population high fragility occupying these areas intense rainfall. It is, therefore, recommended to relocate buildings located in areas with a high or very high degree of hazard and to monitor those with a moderate threat potential. The RI greater than 3 (high risk) indicates the most critical condition. The signs of instability are expressive, in number and/or magnitude. If the existing conditions are maintained, events with high destructive power are very likely to occur. Due to the conditions of vulnerability, significant social and economic damage is expected in these areas, which is why it is recommended to relocate the buildings that are present there.
Considering the situations described, it is considered extremely important that, concurrent with defining the degree of risk, the type of action and some deadline for its execution should also be defined.
The decision on the type of action and the respective deadline resulted from the crossindexing the level of risk and the degree of hazard, as follows: • Do not intervene-low risk and low hazard; • Do not occupy-low risk and medium, high or very high hazard; • Monitoring-moderate or high risk and medium hazard; • Monitoring with scheduled relocation-moderate or high risk and high hazard; • Immediate relocation-high risk and very high hazard.

Results and discussion
Considering the proposed system, the relative normalized weights for indicators (RWI), attributes (RWA) and classes (RWC) were defined using the AHP method, which were later related by means of a linear combination to support the evaluation of the risk components (HI-Hazard Index and VI-Vulnerability Index). The results are summarized in Table 5. Figure 7 illustrates the overlapping of the geotechnical zoning information plan developed by Nascimento (2016) with the Regular Grid information plan. Approximately 53% of the assessed area was classified as having high susceptibility and the remaining regions as having low susceptibility. According to Nascimento (2016), high susceptibility is related to the slope of the compartments, which varies between 12 and 20% and the fluctuations of fluvial discharges, which act directly on the unconsolidated sediments and clay from the   Solimões formation, generating several features of instability, as described. The regions with low susceptibility are above the flood zone, in compartments whose slope varies between 6 and 12%. In Figs. 8 and 9, the cartograms related to the vulnerability and hazard components are presented, respectively.
The simulation carried out indicated that 79.87% of the investigated area has a high and very high vulnerability to these mass movements (Fig. 10). An indication of very high vulnerability, in general, was associated, by the risk assessment model, to regions with a high concentration of subnormal settlements. There is, also, high vulnerability, on the other hand, of the historic and administrative center of the city, characterized by well-structured buildings and a greater concentration of installed infrastructure. Low vulnerability (17.77%), in general, was associated with areas devoid of occupations and infrastructure, normally located in the flood zones of the watercourse, while the average vulnerability was related to squares and open spaces, but endowed with some infrastructure.
Favorable conditions for the development of geological accidents (classes with high and very high HI) were observed in 38.76% of the investigated area. The Hazard cartogram indicated that cells with a higher level of hazard, concentrated in the concave section of the river, within the flood zones, since the level of high hazard was associated with the regions of widespread evidence of creeps and abatement measures. Moderate hazard, in turn, was associated with compartments with high natural susceptibility (Fig. 7), in which the signs of instability were observed incipiently or are absent. Low hazard was associated with the risk model, with compartments of low natural susceptibility, and without signs or evidence of instability. The cartogram that expresses the risk (Fig. 10) resulted from the cross-indexing of the VI (Fig. 8) and HI (Fig. 9) cartograms.
From the data presented in Table 6, it can be seen that the most critical situations (classes with moderate and high risk) were identified in 36% of the investigated area.
In the decision on the type of action to be implemented, and the level of urgency for each action resulted from the cross-indexing the risk cartography and the hazard cartography. The results are summarized in Table 7 and represented in the cartogram in Fig. 11.
Based on the superposition of instability scars registries and orthogonal photographs' mosaic, and on the actions presented in Fig. 12, the coherence between the proposed action and the identified hazard was qualitatively evaluated. Validation process was completed by performing an in situ surface survey on all moderate and high-risk cells, leading to the following observations: (a) The emergency actions related to the immediate relocation occurred in only 5% of the mapped area and correspond to the cells located in the vicinity of the largest channel bed (Acre River), occupied by buildings with low construction standards, precarious infrastructure and, therefore, very vulnerable to the landslides frequently observed in this area; (b) The presence of a small concentration of constructions in a single cell defined as presenting very high hazard was sufficient for providing a priority action indication (immediate relocation of all people within the cell), as can be observed on the yellow cell in Fig. 12; (c) Cells with No occupation indication correspond to open areas or areas with no constructions, in general, located on the inundation zone of Acre River, which is the zone in which the majority of rotational mass movement scars are concentrated;

Conclusions
This work sought to provide aid in assisting in the management of the risk associated with gravitational mass movements, composed of creep and rotational landslides, along watercourse margins and urban areas of a municipality in the Western Brazilian Amazon. It is noteworthy that these processes are of common occurrence throughout the western region of the Amazon. Due to the specificity of the theme, several criteria had to be developed or adapted for the analysis to consider certain environmental, economic and social particularities of the region.
The Hierarchical Analysis Method was the technique adopted to structure the decisionmaking process, so that the hierarchy of the problem and the prioritization of alternatives were based on the logical and quantified judgment that the method provides.
In terms of use and occupation, especially the construction pattern and the level of income of the population in the area studied, it was observed that the lower income population, characterized by subnormal settlements that occupy riverside areas, is in fact the most exposed to the processes mapped. If, on the one hand, this population has developed ways to live with creep (wooden constructions, with limited dimensions, located in an area free from infrastructure, among others), their physical and social condition makes them very vulnerable to landslides.
It was observed that the criteria and procedures described allowed the establishment of standards for the collection, organization, management and validation of the decisionmaking process, which contributes to the reduction of subjectivity in the identification of zones of different degrees of risk and in the implementation of short-term and medium term actions to reduce geological accidents.
The results suggest that the methodological proposal can be applied in other areas, for the study of the risk associated with other processes, which does not exempt it from the need for adequate knowledge about those processes, their relationship with the physical environment and the exposed elements.