Occurring in many geographical, geological and climatic environments, landslides represent a major geological hazard. In landslide prone areas, monitoring devices associated with Early Warning Systems are a cost-effective means to reduce the risk with a low environmental and economic impact, and in some cases, they can be the only solution. In this framework, particular interest has been reserved for Wireless Sensor Networks (WSNs), defined as networks of usually low-size and low-cost devices denoted as nodes, which are integrated with sensors that can gather information through wireless links.
In this thesis, data from a new prototypical ground instability monitoring instrument called Wi-GIM (Wireless sensor network for Ground Instability Monitoring) have been analysed. The system consists in a WSN made by nodes able to measure their mutual inter-distances by calculating the time of flight of an Ultra-Wide Band impulse. Therefore, no sensors are implemented in the network, as the same signals used for transmission are also used for ranging. The system has been tested in a controlled outdoor environment and applied for the monitoring of the displacements of an actual landslide, the Roncovetro mudflow in Central Italy, where a parallel monitoring with a Robotic Total Station (RTS) allowed to validate the system. The outputs are displacement time series showing the distance of each couple of nodes belonging to the same cluster. Data retrieved from the tests revealed a precision of 2–5 cm and that measurements are influenced by the temperature. Since the correlation with this parameter has proved to be linear, a simple correction is sufficient to improve the precision and remove the effect of temperature. The campaign also revealed that measurements were not affected by rain or snow, and that the system can efficiently communicate up to 150 m with a 360° angle of view without affecting precision. Other key features of the implemented system are easy and quick installation, flexibility, low cost, real-time monitoring and acquisition frequency changeability.
The comparison between Wi-GIM and RTS measurements pointed out the presence of an offset (in an order that vary from centimetric to decametric) constant for each single couple, due mainly to the presence of obstacles that can obstruct the Line Of Sight (LOS). The presence of vegetation is the main cause of the non-LOS condition between two nodes, which translates in a longer path of the signals and therefore to a less accurate distance measurements. To go further inside this issue, several tests have been carried out proving the strong influence of the vegetation over both data quantity and quality. To improve them, a MATLAB tool (R2018a, MAthWorks, Natick, MA, USA) called WiSIO (Wireless Sensor network Installation Optimizer) has been developed. The algorithm finds the best devices deployment following three criteria: (i) inter-visibility by means of a modified version of the
Hidden Point Removal operator; (ii) equal distribution; (iii) positioning in preselected priority areas. With respect to the existing viewshed analysis, the main novelty is that it works directly with 3D point clouds, without rendering them or performing any surface. This lead to skip the process of generating surface models avoiding errors and approximations, that is essential when dealing with vegetation.
A second installation of the Wi-GIM system has been therefore carried out considering the deployment suggested by WiSIO. The comparison of data acquired by the system positioned with and without the help of the proposed algorithm allowed to better comprehend the effectiveness of the tool. The presented results are very promising, showing how a simple elaboration can be essential to have more and more reliable data, improving the Wi-GIM system performances, making it even more usable in very complex environments and increasing its flexibility.
The main left limitation of the Wi-GIM system is currently the precision. Such issue is connected to the aim of using only low-cost components, and it can be prospectively overcome if the system undergoes an industrialization process. Furthermore, since the system architecture is re-adaptable, it is prone to enhancements as soon as the technology advances and new low cost hardware enters the market.