RICERCA ITALIANA     

Forecast and control systems for landslides: local sensor distributed networks integration, monitoring techniques and hydro-geological models.

Università degli Studi de L'Aquila

Abstract

The research project aims at combining complementary expertises from three Center of Excellence, namely CETEMPS, DEWS and TELEGEOMATICA, in order to solve problems of major social relevance, such as monitoring and control of unstable slopes. Our research proposal is innovative, as we intend to combine locally deployed sensors and remote sensing for acquisition of a multitude of physical parameters that are directly or indirectly related to those phenomena that may determine landslides events in unstable slopes. If compared to currently employed techniques, the method we intend to develop and test in our research is characterized by the use of advanced technologies such as remote sensing e dense wireless sensor networks: thus, a large set of measurements can be performed and collected, and then used to feed advanced models and algorithms, with the ultimate goal of performing more reliable and economic forecast tools.

The project consists of five Work Package (WP). WP0 is concerned with identification and classification of requirements of the whole monitoring system. WP1, WP2 and WP3 are concerned with specific activities developed and coordinated by DEWS, TELEGEOMATICA and CETEMPS, respectively. WP4 is entirely devoted to the integration of the various activities, and will foster closer cooperation among partners as long as the activities proceed. Integration will be finally focused on developing a test bed, that should demonstrate how to predict a landslide moving from tracking of spatial and temporal evolution of climate and ground status.

The design of the monitoring system based on WSN will be carried out following the design methodology recently introduced that is known as Platform Based Design (PBD) [43], [51]. According to this paradigm, the optimal design solution is reached with successive, iterative refinement steps. PBD is a "meet-in-the-middle" design methodology. It consists of two parts: a top-down and a bottom-up phase. In the top-down process, the monitoring system requirements are identified together with the phenomena to be monitored. In addition, communication protocols and data processing algorithms are defined to fulfil the requirements. In the bottom-up process, starting from a possible implementation of the monitoring platform, consisting of one or more WSN, Public Networks (wired or wireless [44], [45]), Satellites, WEB [46]-[49], ecc.), the platform itself is characterized so that it can be seen as a set of services offered to the functionality that has to be realized. The implementation process consists of mapping the functionality onto the platform in an "optimal" way, where optimal is intended in the sense of satisfying constraints and minimizing resource consumption. The mapping phase is actually the most critical one as it coordinates and conjoins the top-down and the bottom-up processes. The end result is a good solution obtained quickly and that re-uses components that have been developed independently of the application.
This approach motivates splitting the activities of each WP (WP1, WP2, WP3) in methodology oriented tasks and application oriented tasks. Methodological tasks are concerned with definition of phenomena of interest, monitoring strategies, communication protocols for networked embedded systems (e.g. power aware routing, distributed source coding, energy scavenging, and other techniques oriented to minimize energy spending). The output of these tasks will consist of an abstract description of system components, which is independent from actual implementation details. Application oriented tasks will pursue instead configuration of the monitoring platform to meet higher level (application) requirements. <<<

Principal Investigator

Maria Domenica DI BENEDETTO Università degli Studi de L'AQUILA

Research Objectives

Our project will pursue objectives in both methodological and applicative domains. In fact, we aim at developing methods for monitoring and control of unstable slopes, using measurement data collected by locally deployed wireless sensor network and remote sensing systems. The following main physical parameters will be taken into account for in situ measurements: temperature and humidity of the terrain at various levels, temperature and humidity of the lower atmosphere, acceleration. From the remote sensing side, we will be able to collect data on localization and intensity of rainfalls and snowfalls, which are intended as external solicitations to the dynamical model that describes a landslide.
From a detailed system model and a possibly rich sets of measurement reports, we try to estimate the risk of an incoming landslide on the monitored area. In this context, definition and validation of models that represent the status of a slope, their dependence on i) severe atmospheric events that may happen in that area, ii) humidity, temperature and other climate parameters, iii) humidity and temperature of the terrain), is a major efforts for performance and effectiveness of the entire system. Therefore, methodological tasks include (high level) mathematical models for landslide prediction, which combines the abovementioned set of measurements, as well as models for describing the behaviour of the energy critical wireless (sensor) networking context. While simulation tools will be largely adopted for performance assessment of methodologies at this stage, a test bed will be developed on the application side for validation of models and algorithms over an actual platform and with the aim of testing "real" environmental conditions (e.g. areas subject to heavy rainfall)

According to the scheme depicted in the figure, the test bed will consist of a wireless sensor network (WSN), a gateway station that provides an interface to the Internet through e.g. GSM or GPRS, a WEB-based system for data presentation and network management, an integrated remote sensing system that provides rainfall and snowfall maps by combining measurements from ground radars, satellites and sensors in situ, and a processing system that at last provides risk level estimates.

As far as the WSN component is concerned, we will adopt the MOTES platform that has been already used in other research activities at DEWS. This platform is based on an open source SW platform, with the (light) TinyOS operating system specifically conceived for networked embedded system, and the TinyDB application layer. This latter SW stack will be adequately modified to implement algorithms and models developed in the methodological tasks and will provide the important feature of allowing a "technology unaware" application developer to access the WSN with a high level set of primitives (i.e. the WSN will be handled at this level as a database). A further component is represented by the access WEB page that will be implemented with Java technology, since Java supports modularity and portability over various systems according to the paradigm Write Once, Run Anywhere. At the same time, the WEB server will implement
data storage. The Gateway will be provided with secure and reliable access to the WSN even in critical conditions and will be also provided with a Web camera for eventual real time acquisition of images from the site.

To summarize, the monitoring system we intend to propose and validate, is a complex system that combines advanced modelling techniques and recent technology advances in wireless networking and remote sensing to provide continuous monitoring, risk estimation and alerting the public in correspondence of critical condition detection. This is an innovative approach with respect to currently used systems that are based exclusively on direct and expensive techniques for measurement and for reporting of physical parameters that are directly related to landslides.
In view of the potential impact of the proposed approach, significant efforts will be devoted to validation in both simulation and test bed domains. An important aspect of this validation with respect to currently used systems, is related to the use of instrumentation and topographical techniques already available at the Center TELEGEOMATICA for classical monitoring. For example, we can mention the Mobile Mapping System, MMS, provided with GPS, EGNOS, high quality INS, that has been already used for monitoring of slopes along a valley. MMS is provided with digital cameras, that are able to collect pictures of a road and slopes behind it. The INS system, integrated with GPS, provides parameters for external steering of photo rendering. This system is also equipped with transversal laser scanning that may detect large deviations in a slope shape. The entire system is mounted on a van. <<<

Timescale

24 months

National and international background

The combination of geographic, geological and climatic factors makes Italy one of the European countries more prone to landslides. Italy shares with USA, India and Japan the record of the greatest economical loss due to landslides. The total amount of "landslide costs" (direct and indirect) for each of these countries ranges between 1 and 5 billions of dollars per year.
Heavy and/or persistent precipitations and sudden snow melt represent some relevant phenomena at the origin of landslides [37]. By experience, can be said that very often the typical condition that lead to the landslides are characterized by a period of persistent rain (15-20 days), within the seasonal averages, on which is superimposed an exceptional event of short duration (2, 3 days).

However, flow and outflow water modeling in areas with a complex orography represents an ambitious frontier for research activities in geophysics and in particular in hydrology. As opposed to the concentrated parameters hydrologic approach, based on simple space-independent models, scientific community is focused on the development of spatially semi-distributed models or fully distributed models. If on one hand this trend allowed a more realistic and geographically dependent soil state, on the other hand the high resolution characterization of geo-morphological and hydro-geological parameters represents a critical aspect from the soil parameter setting and in-situ measurements point of view. Today, at the Center of Excellence CETEMPS, where these studies are in progress, rain precipitation measurements are gathered as provided by Regional Agency for Agriculture Services (ARRSSA) and by Hydrographic Institute of Abruzzo Region. In areas non covered by sensors, rain precipitations are estimated through radar and satellite monitoring. The simple sliding scheme, along with the hourly reconstruction of the precipitation fields, can be used to evaluate a flooding alarm index that is useful for identifying segments of the water-drainage net exposed in the event of severe precipitations.
The most recent technologies allow to perform prompt data acquisitions. In the past, motorized "total stations" have been used to repeatedly measure distances and azitmuthal angles from fixed points to sample points (reflecting prisms) in the landslides. Data were sent by cable to a control centre.
At the end of 1980 some catastrophic events lead to the institution of a landslide monitoring center in Sondrio, equipped with various sensors (extensometer, inclinometer and geophones) placed on unstable sides.

The GPS system was particularly suitable for the remote monitoring, via the raw phase data transmission to a processing center, equipped with a refrence GPS receiver. Alternatively, raw data from the central reference station can be sent to the single receivers on the landslide, which elaborate second derivative equations and send back the results to te center. RTK processing is also possible for fast movements of lanslides.

In the future, GALILEO and a renewed GLONASS will allow monitoring in deep valleys and urban areas where the availability of the satellite signals is poor.
In the past also photogrammetry was used, with some difficulties due to slope views and distance. At present photogrammetry is substituted by terrestrial laser scanner in particular for fast landslides movements. Interferential satellite and terrestrial SAR approaches have been also proposed.
The last methods require precise tracking of satellites on the monitoring area. Satellites are also used to recognize severe precipitations which could lead to the landslide phenomena.

After all, the availability of an environmental monitoring platform, such as the one we intend to develop, allowing climatic and micro-climatic measurements on the ground, can be used to calibrate the described satellite based monitoring systems and, in general, can represent a fundamental component of a modern landslides monitoring system.

Starting from these considerations, the framework in which this research unit intends to operate moves from the development of the monitoring system which will be realized through the use of ad-hoc wireless sensor networks (WSN). This solution allows to realize, through a single radio communication network, a very through monitoring of parameters of interest for the observation and control of side stability, is easily scalable in side to the size of the monitoring area, is easily and promptly deployable, has long operative lifetime (low energy consumption and, consequently, limited human intervention on the system) and, most of all, is characterized by low costs with respect to available monitoring solutions. At the same time, meteorological data acquisition (humidity, temperature, etc.) can support definition activities of forecasting models.
More precisely, the system will be developed following the design methodology known as Platform Based Design (PBD) [43], [51]. According to this paradigm, the optimal design solution is reached through successive, iterative refinement steps. From one side, the monitoring system requirements as well as the models of phenomena to be monitored are identified. In addition, communication protocols and data processing algorithms are defined in order to fulfill the above requirements. From the other side, acting orthogonally with respect to the previous step, starting from a possible implementation of the monitoring platform, constituted by one or more WSN, Public Networks (wired or wireless [44], [45]), Satellites, WEB [46]-[49], etc., the platform itself is customized so that it can provide to algorithms and protocols previously defined the required services. The two identified developing phases are carried on in parallel through successive refinements, coming to the final definition of the integrated monitoring system.

In defining the state of the art for the overall activity of the project, it is important to consider on one side, the modelling of the phenomena to be monitored and the corresponding control mechanisms, which belong to the "top-down" part of the PBD methodology; on the other, the development of the implementation platform with particular attention to ad-hoc networks.
As for the models, we propose to use techniques that are characteristics of the area of the so-called "hybrid" systems, i.e., models that combine different semantic models as discrete, continuous dynamics and distributed systems. The use of such models has required the development of new theories and new design methods. In these past few years, particular attention has been devoted to the study of these systems (see [52, 53]). However, in our opinion, much still needs to be done in this field, both on the theoretical aspects and to make the results robust and applicable.
As for the ad-hoc networks, we will pay particular attention to recent results and perspectives for future research (e.g. [8]). WSN normally contains a large amount of nodes, with a simplified architecture, equipped by one or more sensors, a supply unit, a data processing unit and a communication unit. Nodes could operate in various environmental conditions and, in general, some of them could be moving or in unpredictable positions.
Therefore, algorithms and protocols will have to be adaptive [2],[8].
The need for WSN to operate in zones that are often remote or not easily accessible deals with the need to power the nodes not by a distribution network, but with batteries. Moreover, we want to guarantee to the WSN a maximal duration (without the need for a human intervention). Consequently, a true technological challenge arises: power consumption reduction, at every developing level, from electronic circuitry, which must foresee the possibility of harvesting the energy from the environment (energy scavenging) [31]-[36] to communication techniques implemented in the entire protocol stack.

The energy needed for transmitting a packet from node A toward node B is proportional to b*d^a, where d is the distance among nodes, a is the path loss, a coefficient between 2 and 4 (4 if the communication happens near to the soil) and b a coefficient representing the overhead for each transmitted bit. It can be demonstrated that generally it is more energetic convenient to use a multihop network, where packets are sent throughout smaller intermediate hops, rather than a direct communication end-to-end among nodes. One can imagine that the maximum of efficiency might be obtained in a network where communications happen over the greater number of hops. Actually, it must be considered that each node spends some energy also in receiving and transmitting each bit. Thus, if we want to reduce the energy consumption, we have also to reduce the number of transmitted bits: compression techniques are of paramount importance from this point of view. The sensor networks are formed by an high density of nodes, therefore it often might show a spatial-temporal correlation among measured data. Therefore, it is needed to remove that redundancy, so that an efficient joint description of sensor data can be obtained. An intense coordination activity among nodes thus seems to be needed. Actually, this coordination can be kept at a reasonable level using distributed source coding techniques [1]-[10]. By means of such techniques, it is possible to compress node data partially using (or not) information coming from other nodes.
In a typical configuration of sensor network, nodes must perform information passing towards a gateway, acting as interface to external world. Moving from a node to another one, information is routed based on some rules. Therefore, it is possible to foresee data aggregation, using in network functions (integrated in routing algorithms [11-21]) like suppression (duplicates elimination), minimum, maximum or average computations or other more sophisticated algorithms. Provided that data processing is less energy consuming than communication, aggregation could enable relevant energy savings.
In producing efficient protocol solutions for these sensor networks, it must accounted for constraints. In particular, for those applications where node mobility and reliable links are required, routing protocols based on distance vector algorithms, such as DSDV [26], AODV [27], DSR [28], have to be considered, while in applications where network life is of paramount importance, algorithms like directed-diffusion [25], zone routing protocol [24], energy aware routing [22], LEACH [23], or others [29], [30], must be considered.
Besides protocols design, network implementation aspects are to be accounted for. As an example, nodes packaging solutions for hostile environments are to be considered.
As far as the experimentation is concerned, a sufficiently flexible platform will be used. A valid choice is represented by the Motes family [39]. This platform is based on programmable devices, equipped with an open source operating system (TinyOS [40]) and with a radio interface of the last generation, and also with an interesting application which allows to view the underlying WSN as a database [41], [42]. <<<