Doctoral School

Reading date: 02/07/2026

• ACEVEDO ROCHA, ALFONSO DARÍO: Modelo de gestión para la restauración del suelo dedicados a la actividad ganadera en el departamento de Córdoba, Colombia. Un aporte de la ganadería regenerativa al desarrollo sostenible del territorio- estudios de casos comparativos
  
  Author: ACEVEDO ROCHA, ALFONSO DARÍO
  Programme: DOCTORAL DEGREE IN SUSTAINABILITY
  Department: University Research Institute for Sustainability Science and Technology (IS.UPC)
  Mode: Normal
  Deposit date: 04/06/2026
  Reading date: pending
  Reading time: pending
  Reading place: pending
  Thesis director: MORATO FARRERAS, JORDI
  Thesis abstract: The impact of livestock farming on soil degradation has been widely discussed, with cattle production being attributed a significant impact both globally and locally. This research, entitled "Management model for the restoration of land used for livestock farming in the department of Córdoba, Colombia. A contribution of regenerative livestock farming to the sustainable development of the territory - Comparative case studies," analyzes the evolution of regenerative livestock farming, especially in Latin America, and adapts a methodology to evaluate the efficiency of this model on the soil using biological and spatial indicators, comparing it with a conventional model on farms in Córdoba. It also evaluates the contribution of these models to the Sustainable Development Goals (SDGs) and proposes a conceptual model for the sustainable management and restoration of livestock soils.An analysis was carried out on the progress of regenerative livestock farming worldwide and in Latin America, finding a significant increase in its adoption since 2018, especially in 2019, when the preservation of biodiversity, water, and soil health began to be valued.Macro indicators of soil condition were defined and 12 farms were selected to analyze conventional and regenerative production models. Edaphic macro and mesofauna were evaluated, satellite images were analyzed, and information on vegetation cover was collected, using biological and spatial indicators to measure the restorative capacity of the models. A SWOT (Strengths, Weaknesses, Opportunities, and Threats) analysis was also conducted, concluding that the regenerative model is more efficient in grazing practices, profitability, and biodiversity, allowing for passive restoration of the ecosystem.A methodology was developed to analyze the efficiency of regenerative livestock farming, establishing methods for collecting biotic and spatial data, which facilitates continuous soil diagnosis and monitoring. The study area, the duration of the analysis, the phases of the process, the indicators, and the multitemporal analysis of the plant component were defined using remote sensing and GIS technologies, providing the necessary data for a detailed analysis. It was concluded that regenerative livestock farming is an economical and sustainable low-cost model that improves productivity without agrochemicals and prevents erosion, promoting soil restoration through sustainable practices.This model also contributes to the fulfillment of the SDGs, especially in food security, climate action, and ecosystem restoration, promoting greater biological biodiversity and vegetation cover in the study areas. The results show a positive impact on soil health, vegetation development, and macrofauna, with notable root growth and no degraded soils. Regenerative livestock farming works in harmony with nature, improving the living conditions and health of livestock.Regenerative livestock farming is presented as a key tool for addressing the environmental, social, and productive challenges of livestock farming, restoring soils and ecosystems, and reversing the damage caused by overgrazing and the use of agrochemicals in conventional models. It restores soil fertility, promotes a natural cycle of regeneration, and facilitates the recovery of native species of flora and fauna, increasing the ecosystem's resistance to external disturbances.In terms of species abundance, a total of 763 individuals were quantified between both systems, distributed across 19 taxonomic orders, with 630 individuals in the regenerative model and 133 in the conventional model. The Shannon-Wiener Index (H') under the regenerative model was 2.06, reflecting high biological diversity and efficiency in restoration and long-term sustainability.
  

• RODRIGUEZ AGUIRRE, STEPHANIA MABEL: Assessment of Sediment Production, Connectivity, and Delivery in Mountainous Catchments under Environmental Changes. The Role of Shallow Landslides in the Upper Llobregat Basin (Pre-Pyrenees, Catalonia)
  
  Author: RODRIGUEZ AGUIRRE, STEPHANIA MABEL
  Programme: DOCTORAL DEGREE IN GEOTECHNICAL ENGINEERING
  Department: Department of Civil and Environmental Engineering (DECA)
  Mode: Normal
  Deposit date: 01/06/2026
  Reading date: 02/07/2026
  Reading time: 10:00
  Reading place: ETSECCPB.UPC, Campus NordBuilding D2. Classroom: 216C/Jordi Girona, 1-308034 Barcelona
  Thesis director: HÜRLIMANN ZIEGLER, MARCEL | DE MEDINA IGLESIAS, VICENTE CÉSAR
  Thesis abstract: Soil erosion in mountainous landscapes is strongly influenced by shallow landslides, particularly during extreme precipitation events. Despite their geomorphic significance, quantitative soil erosion studies often prioritize fine sediment dynamics and agricultural land-use calibrations, thereby underrepresenting the complexity of landslide processes driven by intrinsic slope parameters and external triggering forces. This study examines sediment production and connectivity from shallow landslides under climate change and land-cover transitions. The primary objective aims to quantify sediment mobilization from slope instabilities to fluvial networks, focusing on the Saldes River Basin (102 km²) for model calibration and applying the results to the Upper Llobregat River Basin (504 km²).The quantification approach used a distributed, event-based method with the FSLAM landslide stability model. Projected climate scenarios predict increased extreme precipitation events, resulting in heightened slope mass wasting. Future simulations indicate landslide mobilization rates increase by 10%, peaking at 16% during 20-year rainfall events. These findings underscore the significance of hillslope processes in sediment budget assessments and the role of extreme rainfall in sediment production.To address sediment connectivity, this study introduces the Random Connect (RC) model. The model uses a random-walk framework and applies the Index of Connectivity (IC) to estimate sediment volume fluxes from source areas to channel networks. The input data includes a digital elevation model, IC maps, and sediment source maps derived from shallow landslides. The model outputs consist of raster maps that describe sediment flux. The receive maps show accumulated sediment volumes along transfer paths, and the release maps indicate the source cells contributing sediment to the target river.The RC model application in the Saldes Basin highlighted how land cover affects sediment transport. Field surveys identified sediment hotspots and validated model outputs, incorporating forest cover as a factor in connectivity analyses. This enhanced IC methodology delineated transfer pathways, improving overland flow representation in RC simulations. Validation showed an average annual sediment budget of 1,545 t·km⁻²·year⁻¹, with the model estimating a sediment delivery ratio (SDR) of 0.34. Reservoir SDR values ranged from 0.12 to 0.24, illustrating the impact of connectivity thresholds on reservoir siltation. A detailed LULC analysis in the ULRB shows that steep, sparsely vegetated areas have increased by 43%, leading to more bare soil and sediment sources. Conversely, forested slopes and vegetated areas act as buffers, reducing sediment delivery. However, climate projections suggest intensified rainfall and prolonged droughts may reduce vegetation resilience, reactivating sediment pathways. Results show that sediment production increases by up to 22% during extreme rainfall in future scenarios, while forest expansion can reduce sediment yield by up to 14%. Combined climate and land cover scenarios indicate non-linear responses, with far-future events generating up to 18% more sediment than baseline conditions. The findings enhance understanding of sediment dynamics in reservoir-contributing basins, where traditional monitoring struggles to disentangle geomorphic processes. The study underscores the dual influence of forest recovery and extreme rainfall on sediment production and connectivity. The modelling framework provides a robust tool for sediment budget evaluation, aiding watershed management in erosion-prone mountainous areas.
  

• SUVORKIN, VLADIMIR: Robust positioning using hybridization of GNSS with other measurements of opportunity
  
  Author: SUVORKIN, VLADIMIR
  Programme: DOCTORAL DEGREE IN AEROSPACE SCIENCE AND TECHNOLOGY
  Department: Department of Physics (FIS)
  Mode: Normal
  Deposit date: 21/04/2026
  Reading date: 02/07/2026
  Reading time: 13:00
  Reading place: Sala de Juntes, Edifici B3, Campus Nord
  Thesis director: ROVIRA GARCIA, ADRIÀ | GONZÁLEZ CASADO, GUILLERMO
  Thesis abstract: Achieving robust positioning across ground and UAV platforms remains challenging under multipath, partial satellite visibility, and rapidly changing measurement quality, especially in urban and embedded scenarios. At the same time, modern smartphones and embedded receivers increasingly provide multi-constellation, dual-frequency observations, carrier-phase measurements, and IMU streams that can be exploited for aided positioning.The present thesis addresses these conditions through a robust GNSS/IMU integration framework, evaluated across heterogeneous sensor grades, from navigation-grade platforms to consumer smartphones. The framework is designed not only for offline post-processing, but also for real-time and embedded operation, following a deterministic execution structure suitable for on-board use.The principal conclusions are:First, a unified processing framework has been developed for GNSS/IMU integration using raw, undifferenced, uncombined GNSS observables (code, carrier phase, Doppler) and inertial measurements.Second, a Square-Root Information Filter (SRIF) architecture has been adopted as the estimator core, enabling a numerically robust implementation and a shared software structure across GNSS-only processing, loosely coupled (LC) fusion, and tightly coupled (TC) fusion modes.Third, Allan deviation analysis has been used systematically to identify inertial noise parameters and to configure the process-noise model of the navigation filter across different IMU classes, improving consistency of tuning when detailed manufacturer specifications are incomplete.Fourth, experimental validation on multiple independent datasets, spanning different GNSS conditions and equipment grades, shows that LC fusion provides the most consistent practical gains, especially in continuity and robustness during short GNSS degradations, and often improves typical solution behaviour when inertial data quality is adequate.Fifth, in dense urban conditions with strong multipath and masking, positioning performance remains fundamentally constrained by GNSS measurement quality and correction level; inertial aiding mitigates short-term disruptions but does not eliminate the GNSS-side error ceiling.Finally, the thesis demonstrates a transferable and numerically stable GNSS/IMU integration framework, a reproducible Allan-based methodology for configuring heterogeneous IMUs in the filter, and a realistic path toward robust positioning under practical field conditions.
  

Reading date: 03/07/2026

• GIL RAMS, DÍDAC: Splitting of separatrices in generalized standard maps
  
  Author: GIL RAMS, DÍDAC
  Programme: DOCTORAL DEGREE IN APPLIED MATHEMATICS
  Department: School of Mathematics and Statistics (FME)
  Mode: Normal
  Deposit date: 15/05/2026
  Reading date: 03/07/2026
  Reading time: 16:30
  Reading place: Sala d'Actes de l'FME, Edifici U, Campus Sud Enllaç videoconferència: https://meet.google.com/ghn-dzuh-gte
  Thesis director: MARTIN DE LA TORRE, PAU | BALDOMA BARRACA, INMACULADA CONCEPCION
  Thesis abstract: We consider generalized standard maps, that is, families of area-preserving maps on the plane F:(x,y) = (x+y+f(x;h),y+f(x;h)) with h the perturbative parameter and f fulfilling quite general assumptions. This generalization includes the Chirikov standard map, the Hénon map or the perturbed McMillan map, among others.We study transverse intersections between the invariant manifolds (stable and unstable) associated to a hyperbolic fixed point of F. These intersections give rise to homoclinic orbits related to the hyperbolic fixed point. The existence of these kind of orbits is one of the most celebrated methods to prove the existence of chaotic dynamics in a system. Indeed, the Birkhoff-Smale Homoclinic Theorem ensures that, if there exist transverse intersections between the invariant manifolds of the same invariant object, the system is locally conjugate to a Smale horseshoe with infinite symbols.The classical Melnikov theory is a first order perturbative theory that is often used to measure the intersection angle between the invariant manifolds. However, when the Melnikov function is exponentially small in the perturbative parameter, the first order analysis fails. Despite its exponentially small character, the Melnikov function still provides the correct size of the splitting in some systems. This is not our case, which is singular. To deal with it, we use a complex time matching technic involving the inner equation related to F.In the first part of our work, by considering an small enough perturbative parameter h, we obtain an asymptotic formula for the Lazutkin's invariant, a quantity analogous to the intersection angle between the invariant manifolds, related to the primary homoclinic points of a wide class of maps F. An exponentially small upper bound was obtained by Fontich. Later, we provide an asymptotic formula for a controversial example exponentially bigger than the naïve guess.The leading term of the obtained asymptotic formulas depends on a constant, often called Stokes constant, that comes from the study of the inner equation. The second part of our work contains a general algorithm, based on the study of the inner equation related to F, to compute an interval containing such constants by means of a computer assisted proof in CAPD. Finally, we apply this algorithm to prove that the Stokes constants related to maps F of the form f(x;h)=e f_0(x), with e = 4 sinh^2(h/2) and f_0 a polynomial of degree 2<=d_0<= 970 or trigonometric polynomial of any degree, are different from zero (including the Standard and Hénon maps). We also provide an example of a generalized standard map such that its Stokes constant is zero.
  

Reading date: 06/07/2026

• DE LIMAS SANTANA, ALEXANDRE: Beyond 512-bit vectors: Optimized and performance portable AI operators on vector-length-agnostic architectures
  
  Author: DE LIMAS SANTANA, ALEXANDRE
  Programme: DOCTORAL DEGREE IN COMPUTER ARCHITECTURE
  Department: Department of Computer Architecture (DAC)
  Mode: Normal
  Deposit date: 11/05/2026
  Reading date: 06/07/2026
  Reading time: 11:00
  Reading place: Sala de Juntes, UPC planta 1, B6
  Thesis director: CASAS GUIX, MARC | ARMEJACH SANOSA, ADRIÀ
  Thesis abstract: This thesis addresses the challenges of producing high-performance, portable code for arithmetic-intensive operations in Deep Neural Network (DNN) workloads, such as convolutions and matrix multiplications.It targets emerging vector and matrix architectures exposing software interfaces to program accelerators via Instruction Set Architecture (ISA) extensions.It proposes techniques to handle three main challenges related to accelerating DNN workloads on contemporary data-parallel CPU architectures: i) extrapolating software optimization techniques designed primarily for high-end general-purpose vector processors to also accomodate edge devices and long vector accelerators hardware, ii) providing performance portability for open standard vector ISA ecosystems, like RISC-V, characterized by their unprecedent micro-architectural discrepancy among implementations, and iii) designing matrix multiplication ISA extensions that uphold the core principles of emerging vector architecture concerning implementation flexibility, such as vector length agnosticism.This thesis presents empirical evidence that existing software optimization techniques for generating high-performance implementations of DNN operators for modern vector processors are biased toward high-end general-purpose systems with 512-bit vectors (e.g., Intel Cascade Lake, Fugaku AF64X).The techniques presented in the thesis support the idea of hardware/software co-design and the need to consider microarchitectural features, such as vector length, non-conventional memory subsystems, and the presence/lack of out-of-order pipelines, when generating code for vector and matrix processors.Specifically, this thesis employs runtime specialization to adapt established DNN algorithms to a broader range of processor designs, equipping applications with flexible code generators that dynamically optimize the algorithms for the specific combination of the platform's microarchitectural features and the operation's hyperparameters. The thesis makes three key contributions.First, it provides the first performance analysis of convolution workloads on a 16,384-bit vector processor, identifies cache conflict misses in existing techniques, and proposes software corrections that yield up to 1.83x speedups over prior approaches.Second, it introduces a dynamic matrix multiplication and convolution code generator that adapts algorithmic optimization techniques, such as register unrolling, to the vector width of the target platform.A performance analysis of three commercially available RISC-V systems running computer vision workloads shows geometric speedups ranging from 1.43x to 3.58x compared to state-of-the-art libraries.Third, it proposes a Matrix Tile Extension (MTE) to supplement vector-length-agnostic ISAs with geometry-agnostic matrix multiplication and memory operations.Two microarchitectures supporting this extension are described: a lean extension of an 8192-bit vector processor and a systolic-array-based design, both of which utilize vector registers for matrix storage and demonstrate 1.35x geometric mean speedups on GEMM and convolution workloads.