Shallow and deep deformation in northern Apennines region using seismological data

For its particular position and the complex geological history, the Northern Apennines has been considered as a natural laboratory to apply several kinds of investigations. By the way, it is complicated to joint all the knowledge about the Northern Apennines in a unique picture that explains the structural and geological emplacement that produced it. The main goal of this thesis is to put together all information on the deformation - in the crust and at depth - of this region and to describe a geodynamical model that takes account of it. To do so, we have analyzed the pattern of deformation in the crust and in the mantle. In both cases the deformation has been studied using always information recovered from earthquakes, although using different techniques. In particular the shallower deformation has been studied using seismic moment tensors information. For our purpose we used the methods described in Arvidsson and Ekstrom (1998) that allowing the use in the inversion of surface waves [and not only of the body waves as the Centroid Moment Tensor (Dziewonski et al., 1981) one] allow to determine seismic source parameters for earthquakes with magnitude as small as 4.0. We applied this tool in the Northern Apennines and through this activity we have built up the Italian CMT dataset (Pondrelli et al., 2006) and the pattern of seismic deformation using the Kostrov (1974) method on a regular grid of 0.25 degree cells. We obtained a map of lateral variations of the pattern of seismic deformation on different layers of depth, taking into account the fact that shallow earthquakes (within 15 km of depth) in the region occur everywhere while most of events with a deeper hypocenter (15-40 km) occur only in the outer part of the belt, on the Adriatic side. For the analysis of the deep deformation, i.e. that occurred in the mantle, we used the anisotropy information characterizing the structure below the Northern Apennines. The anisotropy is an earth properties that in the crust is due to the presence of aligned fluid filled cracks or alternating isotropic layers with different elastic properties while in the mantle the most important cause of seismic anisotropy is the lattice preferred orientation (LPO) of the mantle minerals as the olivine. This last is a highly anisotropic mineral and tends to align its fast crystallographic axes (a-axis) parallel to the astenospheric flow as a response to finite strain induced by geodynamic processes. The seismic anisotropy pattern of a region is measured utilizing the shear wave splitting phenomenon (that is the seismological analogue to optical birefringence). Here, to do so, we apply on teleseismic earthquakes recorded on stations located in the study region, the Sileny and Plomerova (1996) approach. The results are analyzed on the basis of their lateral and vertical variations to better define the earth structure beneath Northern Apennines. We find different anisotropic domains, a Tuscany and an Adria one, with a pattern of seismic anisotropy which laterally varies in a similar way respect to the seismic deformation. Moreover, beneath the Adriatic region the distribution of the splitting parameters is so complex to request an appropriate analysis. Therefore we applied on our data the code of Menke and Levin (2003) which allows to look for different models of structures with multilayer anisotropy. We obtained that the structure beneath the Po Plain is probably even more complicated than expected. On the basis of the results obtained for this thesis, added with those from previous works, we suggest that slab roll-back, which created the Apennines and opened the Tyrrhenian Sea, evolved in the north boundary of Northern Apennines in a different way from its southern part. In particular, the trench retreat developed primarily south of our study region, with an eastward roll-back. In the northern portion of the orogen, after a first stage during which the retreat was perpendicular to the trench, it became oblique with respect to the structure.

Estimating crustal deformation from geodetic data

By the end of the 19th century, geodesy has contributed greatly to the knowledge of regional tectonics and fault movement through its ability to measure, at sub-centimetre precision, the relative positions of points on the Earth’s surface. Nowadays the systematic analysis of geodetic measurements in active deformation regions represents therefore one of the most important tool in the study of crustal deformation over different temporal scales [e.g., Dixon, 1991]. This dissertation focuses on motion that can be observed geodetically with classical terrestrial position measurements, particularly triangulation and leveling observations. The work is divided into two sections: an overview of the principal methods for estimating longterm accumulation of elastic strain from terrestrial observations, and an overview of the principal methods for rigorously inverting surface coseismic deformation fields for source geometry with tests on synthetic deformation data sets and applications in two different tectonically active regions of the Italian peninsula. For the long-term accumulation of elastic strain analysis, triangulation data were available from a geodetic network across the Messina Straits area (southern Italy) for the period 1971 – 2004. From resulting angle changes, the shear strain rates as well as the orientation of the principal axes of the strain rate tensor were estimated. The computed average annual shear strain rates for the time period between 1971 and 2004 are γ˙1 = 113.89 ± 54.96 nanostrain/yr and γ˙2 = -23.38 ± 48.71 nanostrain/yr, with the orientation of the most extensional strain (θ) at N140.80° ± 19.55°E. These results suggests that the first-order strain field of the area is dominated by extension in the direction perpendicular to the trend of the Straits, sustaining the hypothesis that the Messina Straits could represents an area of active concentrated deformation. The orientation of θ agree well with GPS deformation estimates, calculated over shorter time interval, and is consistent with previous preliminary GPS estimates [D’Agostino and Selvaggi, 2004; Serpelloni et al., 2005] and is also similar to the direction of the 1908 (MW 7.1) earthquake slip vector [e.g., Boschi et al., 1989; Valensise and Pantosti, 1992; Pino et al., 2000; Amoruso et al., 2002]. Thus, the measured strain rate can be attributed to an active extension across the Messina Straits, corresponding to a relative extension rate ranges between < 1mm/yr and up to ~ 2 mm/yr, within the portion of the Straits covered by the triangulation network. These results are consistent with the hypothesis that the Messina Straits is an important active geological boundary between the Sicilian and the Calabrian domains and support previous preliminary GPS-based estimates of strain rates across the Straits, which show that the active deformation is distributed along a greater area. Finally, the preliminary dislocation modelling has shown that, although the current geodetic measurements do not resolve the geometry of the dislocation models, they solve well the rate of interseismic strain accumulation across the Messina Straits and give useful information about the locking the depth of the shear zone. Geodetic data, triangulation and leveling measurements of the 1976 Friuli (NE Italy) earthquake, were available for the inversion of coseismic source parameters. From observed angle and elevation changes, the source parameters of the seismic sequence were estimated in a join inversion using an algorithm called “simulated annealing”. The computed optimal uniform–slip elastic dislocation model consists of a 30° north-dipping shallow (depth 1.30 ± 0.75 km) fault plane with azimuth of 273° and accommodating reverse dextral slip of about 1.8 m. The hypocentral location and inferred fault plane of the main event are then consistent with the activation of Periadriatic overthrusts or other related thrust faults as the Gemona- Kobarid thrust. Then, the geodetic data set exclude the source solution of Aoudia et al. [2000], Peruzza et al. [2002] and Poli et al. [2002] that considers the Susans-Tricesimo thrust as the May 6 event. The best-fit source model is then more consistent with the solution of Pondrelli et al. [2001], which proposed the activation of other thrusts located more to the North of the Susans-Tricesimo thrust, probably on Periadriatic related thrust faults. The main characteristics of the leveling and triangulation data are then fit by the optimal single fault model, that is, these results are consistent with a first-order rupture process characterized by a progressive rupture of a single fault system. A single uniform-slip fault model seems to not reproduce some minor complexities of the observations, and some residual signals that are not modelled by the optimal single-fault plane solution, were observed. In fact, the single fault plane model does not reproduce some minor features of the leveling deformation field along the route 36 south of the main uplift peak, that is, a second fault seems to be necessary to reproduce these residual signals. By assuming movements along some mapped thrust located southward of the inferred optimal single-plane solution, the residual signal has been successfully modelled. In summary, the inversion results presented in this Thesis, are consistent with the activation of some Periadriatic related thrust for the main events of the sequence, and with a minor importance of the southward thrust systems of the middle Tagliamento plain.

Multiresolution spherical wavelet analysis in global seismic tomography

Every seismic event produces seismic waves which travel throughout the Earth. Seismology is the science of interpreting measurements to derive information about the structure of the Earth. Seismic tomography is the most powerful tool for determination of 3D structure of deep Earth's interiors. Tomographic models obtained at the global and regional scales are an underlying tool for determination of geodynamical state of the Earth, showing evident correlation with other geophysical and geological characteristics. The global tomographic images of the Earth can be written as a linear combinations of basis functions from a specifically chosen set, defining the model parameterization. A number of different parameterizations are commonly seen in literature: seismic velocities in the Earth have been expressed, for example, as combinations of spherical harmonics or by means of the simpler characteristic functions of discrete cells. With this work we are interested to focus our attention on this aspect, evaluating a new type of parameterization, performed by means of wavelet functions. It is known from the classical Fourier theory that a signal can be expressed as the sum of a, possibly infinite, series of sines and cosines. This sum is often referred as a Fourier expansion. The big disadvantage of a Fourier expansion is that it has only frequency resolution and no time resolution. The Wavelet Analysis (or Wavelet Transform) is probably the most recent solution to overcome the shortcomings of Fourier analysis. The fundamental idea behind this innovative analysis is to study signal according to scale. Wavelets, in fact, are mathematical functions that cut up data into different frequency components, and then study each component with resolution matched to its scale, so they are especially useful in the analysis of non stationary process that contains multi-scale features, discontinuities and sharp strike. Wavelets are essentially used in two ways when they are applied in geophysical process or signals studies: 1) as a basis for representation or characterization of process; 2) as an integration kernel for analysis to extract information about the process. These two types of applications of wavelets in geophysical field, are object of study of this work. At the beginning we use the wavelets as basis to represent and resolve the Tomographic Inverse Problem. After a briefly introduction to seismic tomography theory, we assess the power of wavelet analysis in the representation of two different type of synthetic models; then we apply it to real data, obtaining surface wave phase velocity maps and evaluating its abilities by means of comparison with an other type of parametrization (i.e., block parametrization). For the second type of wavelet application we analyze the ability of Continuous Wavelet Transform in the spectral analysis, starting again with some synthetic tests to evaluate its sensibility and capability and then apply the same analysis to real data to obtain Local Correlation Maps between different model at same depth or between different profiles of the same model.

The rupture process of recent tsunamigenic earthquakes by geophysical data inversion

Subduction zones are the favorite places to generate tsunamigenic earthquakes, where friction between oceanic and continental plates causes the occurrence of a strong seismicity. The topics and the methodologies discussed in this thesis are focussed to the understanding of the rupture process of the seismic sources of great earthquakes that generate tsunamis. The tsunamigenesis is controlled by several kinematical characteristic of the parent earthquake, as the focal mechanism, the depth of the rupture, the slip distribution along the fault area and by the mechanical properties of the source zone. Each of these factors plays a fundamental role in the tsunami generation. Therefore, inferring the source parameters of tsunamigenic earthquakes is crucial to understand the generation of the consequent tsunami and so to mitigate the risk along the coasts. The typical way to proceed when we want to gather information regarding the source process is to have recourse to the inversion of geophysical data that are available. Tsunami data, moreover, are useful to constrain the portion of the fault area that extends offshore, generally close to the trench that, on the contrary, other kinds of data are not able to constrain. In this thesis I have discussed the rupture process of some recent tsunamigenic events, as inferred by means of an inverse method. I have presented the 2003 Tokachi-Oki (Japan) earthquake (Mw 8.1). In this study the slip distribution on the fault has been inferred by inverting tsunami waveform, GPS, and bottom-pressure data. The joint inversion of tsunami and geodetic data has revealed a much better constrain for the slip distribution on the fault rather than the separate inversions of single datasets. Then we have studied the earthquake occurred on 2007 in southern Sumatra (Mw 8.4). By inverting several tsunami waveforms, both in the near and in the far field, we have determined the slip distribution and the mean rupture velocity along the causative fault. Since the largest patch of slip was concentrated on the deepest part of the fault, this is the likely reason for the small tsunami waves that followed the earthquake, pointing out how much the depth of the rupture plays a crucial role in controlling the tsunamigenesis. Finally, we have presented a new rupture model for the great 2004 Sumatra earthquake (Mw 9.2). We have performed the joint inversion of tsunami waveform, GPS and satellite altimetry data, to infer the slip distribution, the slip direction, and the rupture velocity on the fault. Furthermore, in this work we have presented a novel method to estimate, in a self-consistent way, the average rigidity of the source zone. The estimation of the source zone rigidity is important since it may play a significant role in the tsunami generation and, particularly for slow earthquakes, a low rigidity value is sometimes necessary to explain how a relatively low seismic moment earthquake may generate significant tsunamis; this latter point may be relevant for explaining the mechanics of the tsunami earthquakes, one of the open issues in present day seismology. The investigation of these tsunamigenic earthquakes has underlined the importance to use a joint inversion of different geophysical data to determine the rupture characteristics. The results shown here have important implications for the implementation of new tsunami warning systems – particularly in the near-field – the improvement of the current ones, and furthermore for the planning of the inundation maps for tsunami-hazard assessment along the coastal area.

Development of Clinical Decision Support Systems based on Mathematical Models of Physiological Systems

In the last years of research, I focused my studies on different physiological problems. Together with my supervisors, I developed/improved different mathematical models in order to create valid tools useful for a better understanding of important clinical issues. The aim of all this work is to develop tools for learning and understanding cardiac and cerebrovascular physiology as well as pathology, generating research questions and developing clinical decision support systems useful for intensive care unit patients. I. ICP-model Designed for Medical Education We developed a comprehensive cerebral blood flow and intracranial pressure model to simulate and study the complex interactions in cerebrovascular dynamics caused by multiple simultaneous alterations, including normal and abnormal functional states of auto-regulation of the brain. Individual published equations (derived from prior animal and human studies) were implemented into a comprehensive simulation program. Included in the normal physiological modelling was: intracranial pressure, cerebral blood flow, blood pressure, and carbon dioxide (CO2) partial pressure. We also added external and pathological perturbations, such as head up position and intracranial haemorrhage. The model performed clinically realistically given inputs of published traumatized patients, and cases encountered by clinicians. The pulsatile nature of the output graphics was easy for clinicians to interpret. The manoeuvres simulated include changes of basic physiological inputs (e.g. blood pressure, central venous pressure, CO2 tension, head up position, and respiratory effects on vascular pressures) as well as pathological inputs (e.g. acute intracranial bleeding, and obstruction of cerebrospinal outflow). Based on the results, we believe the model would be useful to teach complex relationships of brain haemodynamics and study clinical research questions such as the optimal head-up position, the effects of intracranial haemorrhage on cerebral haemodynamics, as well as the best CO2 concentration to reach the optimal compromise between intracranial pressure and perfusion. We believe this model would be useful for both beginners and advanced learners. It could be used by practicing clinicians to model individual patients (entering the effects of needed clinical manipulations, and then running the model to test for optimal combinations of therapeutic manoeuvres). II. A Heterogeneous Cerebrovascular Mathematical Model Cerebrovascular pathologies are extremely complex, due to the multitude of factors acting simultaneously on cerebral haemodynamics. In this work, the mathematical model of cerebral haemodynamics and intracranial pressure dynamics, described in the point I, is extended to account for heterogeneity in cerebral blood flow. The model includes the Circle of Willis, six regional districts independently regulated by autoregulation and CO2 reactivity, distal cortical anastomoses, venous circulation, the cerebrospinal fluid circulation, and the intracranial pressure-volume relationship. Results agree with data in the literature and highlight the existence of a monotonic relationship between transient hyperemic response and the autoregulation gain. During unilateral internal carotid artery stenosis, local blood flow regulation is progressively lost in the ipsilateral territory with the presence of a steal phenomenon, while the anterior communicating artery plays the major role to redistribute the available blood flow. Conversely, distal collateral circulation plays a major role during unilateral occlusion of the middle cerebral artery. In conclusion, the model is able to reproduce several different pathological conditions characterized by heterogeneity in cerebrovascular haemodynamics and can not only explain generalized results in terms of physiological mechanisms involved, but also, by individualizing parameters, may represent a valuable tool to help with difficult clinical decisions. III. Effect of Cushing Response on Systemic Arterial Pressure. During cerebral hypoxic conditions, the sympathetic system causes an increase in arterial pressure (Cushing response), creating a link between the cerebral and the systemic circulation. This work investigates the complex relationships among cerebrovascular dynamics, intracranial pressure, Cushing response, and short-term systemic regulation, during plateau waves, by means of an original mathematical model. The model incorporates the pulsating heart, the pulmonary circulation and the systemic circulation, with an accurate description of the cerebral circulation and the intracranial pressure dynamics (same model as in the first paragraph). Various regulatory mechanisms are included: cerebral autoregulation, local blood flow control by oxygen (O2) and/or CO2 changes, sympathetic and vagal regulation of cardiovascular parameters by several reflex mechanisms (chemoreceptors, lung-stretch receptors, baroreceptors). The Cushing response has been described assuming a dramatic increase in sympathetic activity to vessels during a fall in brain O2 delivery. With this assumption, the model is able to simulate the cardiovascular effects experimentally observed when intracranial pressure is artificially elevated and maintained at constant level (arterial pressure increase and bradicardia). According to the model, these effects arise from the interaction between the Cushing response and the baroreflex response (secondary to arterial pressure increase). Then, patients with severe head injury have been simulated by reducing intracranial compliance and cerebrospinal fluid reabsorption. With these changes, oscillations with plateau waves developed. In these conditions, model results indicate that the Cushing response may have both positive effects, reducing the duration of the plateau phase via an increase in cerebral perfusion pressure, and negative effects, increasing the intracranial pressure plateau level, with a risk of greater compression of the cerebral vessels. This model may be of value to assist clinicians in finding the balance between clinical benefits of the Cushing response and its shortcomings. IV. Comprehensive Cardiopulmonary Simulation Model for the Analysis of Hypercapnic Respiratory Failure We developed a new comprehensive cardiopulmonary model that takes into account the mutual interactions between the cardiovascular and the respiratory systems along with their short-term regulatory mechanisms. The model includes the heart, systemic and pulmonary circulations, lung mechanics, gas exchange and transport equations, and cardio-ventilatory control. Results show good agreement with published patient data in case of normoxic and hyperoxic hypercapnia simulations. In particular, simulations predict a moderate increase in mean systemic arterial pressure and heart rate, with almost no change in cardiac output, paralleled by a relevant increase in minute ventilation, tidal volume and respiratory rate. The model can represent a valid tool for clinical practice and medical research, providing an alternative way to experience-based clinical decisions. In conclusion, models are not only capable of summarizing current knowledge, but also identifying missing knowledge. In the former case they can serve as training aids for teaching the operation of complex systems, especially if the model can be used to demonstrate the outcome of experiments. In the latter case they generate experiments to be performed to gather the missing data.

Kinematic description of the rupture from strong motion data: strategies for a robust inversion

We present a non linear technique to invert strong motion records with the aim of obtaining the final slip and rupture velocity distributions on the fault plane. In this thesis, the ground motion simulation is obtained evaluating the representation integral in the frequency. The Green’s tractions are computed using the discrete wave-number integration technique that provides the full wave-field in a 1D layered propagation medium. The representation integral is computed through a finite elements technique, based on a Delaunay’s triangulation on the fault plane. The rupture velocity is defined on a coarser regular grid and rupture times are computed by integration of the eikonal equation. For the inversion, the slip distribution is parameterized by 2D overlapping Gaussian functions, which can easily relate the spectrum of the possible solutions with the minimum resolvable wavelength, related to source-station distribution and data processing. The inverse problem is solved by a two-step procedure aimed at separating the computation of the rupture velocity from the evaluation of the slip distribution, the latter being a linear problem, when the rupture velocity is fixed. The non-linear step is solved by optimization of an L2 misfit function between synthetic and real seismograms, and solution is searched by the use of the Neighbourhood Algorithm. The conjugate gradient method is used to solve the linear step instead. The developed methodology has been applied to the M7.2, Iwate Nairiku Miyagi, Japan, earthquake. The estimated magnitude seismic moment is 2.6326 dyne∙cm that corresponds to a moment magnitude MW 6.9 while the mean the rupture velocity is 2.0 km/s. A large slip patch extends from the hypocenter to the southern shallow part of the fault plane. A second relatively large slip patch is found in the northern shallow part. Finally, we gave a quantitative estimation of errors associates with the parameters.

Inferences on earthquake kinematic properties from data inversion: two different approaches

During my PhD, starting from the original formulations proposed by Bertrand et al., 2000 and Emolo & Zollo 2005, I developed inversion methods and applied then at different earthquakes. In particular large efforts have been devoted to the study of the model resolution and to the estimation of the model parameter errors. To study the source kinematic characteristics of the Christchurch earthquake we performed a joint inversion of strong-motion, GPS and InSAR data using a non-linear inversion method. Considering the complexity highlighted by superficial deformation data, we adopted a fault model consisting of two partially overlapping segments, with dimensions 15x11 and 7x7 km2, having different faulting styles. This two-fault model allows to better reconstruct the complex shape of the superficial deformation data. The total seismic moment resulting from the joint inversion is 3.0x1025 dyne.cm (Mw = 6.2) with an average rupture velocity of 2.0 km/s. Errors associated with the kinematic model have been estimated of around 20-30 %. The 2009 Aquila sequence was characterized by an intense aftershocks sequence that lasted several months. In this study we applied an inversion method that assumes as data the apparent Source Time Functions (aSTFs), to a Mw 4.0 aftershock of the Aquila sequence. The estimation of aSTFs was obtained using the deconvolution method proposed by Vallée et al., 2004. The inversion results show a heterogeneous slip distribution, characterized by two main slip patches located NW of the hypocenter, and a variable rupture velocity distribution (mean value of 2.5 km/s), showing a rupture front acceleration in between the two high slip zones. Errors of about 20% characterize the final estimated parameters.

Early Warning For Large Earthquakes: Observations, Models and Real-Time Data Analysis

This thesis is a collection of works focused on the topic of Earthquake Early Warning, with a special attention to large magnitude events. The topic is addressed from different points of view and the structure of the thesis reflects the variety of the aspects which have been analyzed. The first part is dedicated to the giant, 2011 Tohoku-Oki earthquake. The main features of the rupture process are first discussed. The earthquake is then used as a case study to test the feasibility Early Warning methodologies for very large events. Limitations of the standard approaches for large events arise in this chapter. The difficulties are related to the real-time magnitude estimate from the first few seconds of recorded signal. An evolutionary strategy for the real-time magnitude estimate is proposed and applied to the single Tohoku-Oki earthquake. In the second part of the thesis a larger number of earthquakes is analyzed, including small, moderate and large events. Starting from the measurement of two Early Warning parameters, the behavior of small and large earthquakes in the initial portion of recorded signals is investigated. The aim is to understand whether small and large earthquakes can be distinguished from the initial stage of their rupture process. A physical model and a plausible interpretation to justify the observations are proposed. The third part of the thesis is focused on practical, real-time approaches for the rapid identification of the potentially damaged zone during a seismic event. Two different approaches for the rapid prediction of the damage area are proposed and tested. The first one is a threshold-based method which uses traditional seismic data. Then an innovative approach using continuous, GPS data is explored. Both strategies improve the prediction of large scale effects of strong earthquakes.

Towards the 3D attenuation imaging of active volcanoes: methods and tests on real and simulated data

The purpose of my PhD thesis has been to face the issue of retrieving a three dimensional attenuation model in volcanic areas. To this purpose, I first elaborated a robust strategy for the analysis of seismic data. This was done by performing several synthetic tests to assess the applicability of spectral ratio method to our purposes. The results of the tests allowed us to conclude that: 1) spectral ratio method gives reliable differential attenuation (dt*) measurements in smooth velocity models; 2) short signal time window has to be chosen to perform spectral analysis; 3) the frequency range over which to compute spectral ratios greatly affects dt* measurements. Furthermore, a refined approach for the application of spectral ratio method has been developed and tested. Through this procedure, the effects caused by heterogeneities of propagation medium on the seismic signals may be removed. The tested data analysis technique was applied to the real active seismic SERAPIS database. It provided a dataset of dt* measurements which was used to obtain a three dimensional attenuation model of the shallowest part of Campi Flegrei caldera. Then, a linearized, iterative, damped attenuation tomography technique has been tested and applied to the selected dataset. The tomography, with a resolution of 0.5 km in the horizontal directions and 0.25 km in the vertical direction, allowed to image important features in the off-shore part of Campi Flegrei caldera. High QP bodies are immersed in a high attenuation body (Qp=30). The latter is well correlated with low Vp and high Vp/Vs values and it is interpreted as a saturated marine and volcanic sediments layer. High Qp anomalies, instead, are interpreted as the effects either of cooled lava bodies or of a CO2 reservoir. A pseudo-circular high Qp anomaly was detected and interpreted as the buried rim of NYT caldera.

Noises: a Nuisance and a Resource. Development of New Decomposition Methods of Noisy Data

Noise is constant presence in measurements. Its origin is related to the microscopic properties of matter. Since the seminal work of Brown in 1828, the study of stochastic processes has gained an increasing interest with the development of new mathematical and analytical tools. In the last decades, the central role that noise plays in chemical and physiological processes has become recognized. The dual role of noise as nuisance/resource pushes towards the development of new decomposition techniques that divide a signal into its deterministic and stochastic components. In this thesis I show how methods based on Singular Spectrum Analysis have the right properties to fulfil the previously mentioned requirement. During my work I applied SSA to different signals of interest in chemistry: I developed a novel iterative procedure for the denoising of powder X-ray diffractograms; I “denoised” bi-dimensional images from experiments of electrochemiluminescence imaging of micro-beads obtaining new insight on ECL mechanism. I also used Principal Component Analysis to investigate the relationship between brain electrophysiological signals and voice emission.

Data analysis with merge trees

Today’s data are increasingly complex and classical statistical techniques need growingly more refined mathematical tools to be able to model and investigate them. Paradigmatic situations are represented by data which need to be considered up to some kind of trans- formation and all those circumstances in which the analyst finds himself in the need of defining a general concept of shape. Topological Data Analysis (TDA) is a field which is fundamentally contributing to such challenges by extracting topological information from data with a plethora of interpretable and computationally accessible pipelines. We con- tribute to this field by developing a series of novel tools, techniques and applications to work with a particular topological summary called merge tree. To analyze sets of merge trees we introduce a novel metric structure along with an algorithm to compute it, define a framework to compare different functions defined on merge trees and investigate the metric space obtained with the aforementioned metric. Different geometric and topolog- ical properties of the space of merge trees are established, with the aim of obtaining a deeper understanding of such trees. To showcase the effectiveness of the proposed metric, we develop an application in the field of Functional Data Analysis, working with functions up to homeomorphic reparametrization, and in the field of radiomics, where each patient is represented via a clustering dendrogram.

Data driven regularization models of non-linear ill-posed inverse problems in imaging

Imaging technologies are widely used in application fields such as natural sciences, engineering, medicine, and life sciences. A broad class of imaging problems reduces to solve ill-posed inverse problems (IPs). Traditional strategies to solve these ill-posed IPs rely on variational regularization methods, which are based on minimization of suitable energies, and make use of knowledge about the image formation model (forward operator) and prior knowledge on the solution, but lack in incorporating knowledge directly from data. On the other hand, the more recent learned approaches can easily learn the intricate statistics of images depending on a large set of data, but do not have a systematic method for incorporating prior knowledge about the image formation model. The main purpose of this thesis is to discuss data-driven image reconstruction methods which combine the benefits of these two different reconstruction strategies for the solution of highly nonlinear ill-posed inverse problems. Mathematical formulation and numerical approaches for image IPs, including linear as well as strongly nonlinear problems are described. More specifically we address the Electrical impedance Tomography (EIT) reconstruction problem by unrolling the regularized Gauss-Newton method and integrating the regularization learned by a data-adaptive neural network. Furthermore we investigate the solution of non-linear ill-posed IPs introducing a deep-PnP framework that integrates the graph convolutional denoiser into the proximal Gauss-Newton method with a practical application to the EIT, a recently introduced promising imaging technique. Efficient algorithms are then applied to the solution of the limited electrods problem in EIT, combining compressive sensing techniques and deep learning strategies. Finally, a transformer-based neural network architecture is adapted to restore the noisy solution of the Computed Tomography problem recovered using the filtered back-projection method.

Data surveillance for infectious diseases: models, complex networks and machine learning

In this thesis, we investigate the role of applied physics in epidemiological surveillance through the application of mathematical models, network science and machine learning. The spread of a communicable disease depends on many biological, social, and health factors. The large masses of data available make it possible, on the one hand, to monitor the evolution and spread of pathogenic organisms; on the other hand, to study the behavior of people, their opinions and habits. Presented here are three lines of research in which an attempt was made to solve real epidemiological problems through data analysis and the use of statistical and mathematical models. In Chapter 1, we applied language-inspired Deep Learning models to transform influenza protein sequences into vectors encoding their information content. We then attempted to reconstruct the antigenic properties of different viral strains using regression models and to identify the mutations responsible for vaccine escape. In Chapter 2, we constructed a compartmental model to describe the spread of a bacterium within a hospital ward. The model was informed and validated on time series of clinical measurements, and a sensitivity analysis was used to assess the impact of different control measures. Finally (Chapter 3) we reconstructed the network of retweets among COVID-19 themed Twitter users in the early months of the SARS-CoV-2 pandemic. By means of community detection algorithms and centrality measures, we characterized users’ attention shifts in the network, showing that scientific communities, initially the most retweeted, lost influence over time to national political communities. In the Conclusion, we highlighted the importance of the work done in light of the main contemporary challenges for epidemiological surveillance. In particular, we present reflections on the importance of nowcasting and forecasting, the relationship between data and scientific research, and the need to unite the different scales of epidemiological surveillance.

Meaningful insights: explainability techniques for black-box models on tabular data

Artificial Intelligence (AI) and Machine Learning (ML) are novel data analysis techniques providing very accurate prediction results. They are widely adopted in a variety of industries to improve efficiency and decision-making, but they are also being used to develop intelligent systems. Their success grounds upon complex mathematical models, whose decisions and rationale are usually difficult to comprehend for human users to the point of being dubbed as black-boxes. This is particularly relevant in sensitive and highly regulated domains. To mitigate and possibly solve this issue, the Explainable AI (XAI) field became prominent in recent years. XAI consists of models and techniques to enable understanding of the intricated patterns discovered by black-box models. In this thesis, we consider model-agnostic XAI techniques, which can be applied to Tabular data, with a particular focus on the Credit Scoring domain. Special attention is dedicated to the LIME framework, for which we propose several modifications to the vanilla algorithm, in particular: a pair of complementary Stability Indices that accurately measure LIME stability, and the OptiLIME policy which helps the practitioner finding the proper balance among explanations' stability and reliability. We subsequently put forward GLEAMS a model-agnostic surrogate interpretable model which requires to be trained only once, while providing both Local and Global explanations of the black-box model. GLEAMS produces feature attributions and what-if scenarios, from both dataset and model perspective. Eventually, we argue that synthetic data are an emerging trend in AI, being more and more used to train complex models instead of original data. To be able to explain the outcomes of such models, we must guarantee that synthetic data are reliable enough to be able to translate their explanations to real-world individuals. To this end we propose DAISYnt, a suite of tests to measure synthetic tabular data quality and privacy.