Power has become a central theme in sustainability transitions research (STR) since the field’s political turn in the 2010s. Yet this literature remains difficult to navigate because it draws on diverse conceptualisations whose differences are not always made explicit. Readers may therefore struggle to identify how power is understood and how it is traced empirically. This article addresses this problem through a scoping review of 57 peer-reviewed journal articles that explicitly engage with power, combining a systematic Scopus search, backward and forward snowballing, and thematic analysis. The review identifies three broad analytical views of power centred on agency, structure, and their interaction. It also develops two practical tools. The first is a glossary that introduces recurring power-related concepts and points readers towards further reading. The second is a reading template that helps identify how studies define, locate, and operationalise power through their conceptual views, units of analysis, methods, adjacent traditions, and empirical signals. Together, these contributions provide an accessible entry point into power research in STR, while supporting experienced researchers in reflecting on their approaches and their dialogue with other conceptualisations in the field.
Last-mile delivery by Uncrewed Aerial Vehicles (UAVs) has gained increasing attention as a promising solution for urban parcel transport, offering a sustainable and cost-effective alternative to traditional ground-based logistics. Such operations, however, introduce significant risks, requiring effective mitigation strategies and efficient path-planning tools. Risk-aware path planning in complex urban environments remains computationally demanding, as it involves a highly combinatorial sequential decision-making problem while requiring efficient online replanning to ensure reactivity. Efficient planners often rely on graph-based search methods, in which the heuristic used to guide the search plays a crucial role. However, in risk-aware path planning, heuristic estimation is particularly challenging, as risk evaluation forms part of the cost estimation and differs significantly from classical line-of-sight or Euclidean approaches. To address this limitation, this study presents a novel Deep Learning framework for estimating accurate and admissible heuristic functions in risk-aware UAV path planning. To this end, a comprehensive database of cost-to-go maps was generated using realistic 3D urban scenarios and a graph-based optimisation framework. This dataset was used to train a Vision Transformer (ViT) network capable of providing accurate cost-to-go estimates for various objective-function configurations. The proposed architecture incorporates an admissibility correction mechanism that regulates the trade-off between heuristic accuracy and admissibility, promoting either solution optimality or search efficiency. The method was benchmarked against conventional heuristics in terms of computation time and cost-to-go. Results show speed-ups of up to 20 times compared to traditional approaches while maintaining near-optimal solutions, with marginal increases in average path cost of the order of 0.1%.
Invasive alien species (IAS) are linked to the extinction of numerous organisms, being a problem for which urgent measures must be taken. Their management requires robust decision-support tools capable of integrating ecological, economic and social impacts. Cost-benefit analysis (CBA) has been highlighted by bodies such as the European Union as an important tool for making sound decisions in IAS management. This study focuses on the restoration of plots invaded by Acacia dealbata in the SAC Sil Canyon (NW Spain), using a novel integrated assessment framework that links remote sensing and contingent valuation (CV) within an extended environmental CBA. Sentinel-2 data and a Random Forest algorithm were used to map Acacia plots with an accuracy of 98%, given that no data on the occupancy of the species were available for this location. Knowing the Acacia distribution made it possible to characterize the slope and location of the plots in order to estimate the restoration costs. The intervention yields benefits in terms of timber from native species and non-market benefits, which were monetized by CV, obtaining a median willingness to pay of 6.31 & euro;/year over 50 years. All costs and revenues were incorporated into the CBA to assess the feasibility and impact of the project. After a sensitivity analysis using a Monte Carlo simulation, the restoration was found to produce a net present value (NPV) of 2,793,476.09 & euro; over a 120-year horizon, with the non-market goods contributing most to this amount. The results demonstrate that integrating remote sensing and stated preference methods within an extended CBA provides a sound basis for evaluating IAS management alternatives, supporting transparent and evidence-based environmental decision-making in protected areas.
Ulvan, a sulfated polysaccharide derived from green macroalgae, exhibits promising biological and functional properties. However, its high molar mass limits its applicability. This study develops a kinetic model describing ulvan depolymerization using sequential ultrasonication (US) and hydrogen peroxide-ascorbic acid (H2O2/AH2) redox treatment and evaluates the influence of key operational variables on apparent weight-average molar mass (Mw) reduction. US depolymerization followed a one-phase exponential decay model, rapidly decreasing Mw, achieving a 50% reduction within 12.9 min (60% amplitude, 0.64 W/mL). Secondly, a quadratic regression model accurately described H2O2/AH2 driven depolymerization, identifying temperature, reaction time, and redox system-to-ulvan ratio as the main factors involved. The combined US-redox process reduced from 1058 to 31.7 kDa within 60 min, whereas acid hydrolysis control (0.1 M HCl) achieved only a reduction from 1022 to 429 kDa under the same treatment time. Optimal depolymerization conditions predicted an H2O2/AH2 ratio of 3.5, Redox system/Ulvan ratio of 4, temperature of 60 degrees C, and reaction time of 85.5 min, yielding a 15 kDa product, which falls within the 95% prediction interval of the model. This product was further concentrated via TFF membranes and spray-dried, with HPSEC analysis confirming no significant change in molar mass distribution and Mw after drying. FTIR and 13C-1H NMR analysis revealed that the depolymerized and spray-dried ulvan retained functional groups and structure, and that hydrolysis only occurred at the anomeric centers without significant side reactions. Overall, the sequential US-H2O2/AH2 approach offers a scalable and wellcontrolled strategy for producing low-molar-mass ulvan for specific applications.
Truss bridges are key transportation infrastructures, recognized for their ability to support heavy loads across long spans. Despite being designed with built-in redundancy, they are exposed to several forms of deterioration that compromise structural integrity, often initiating local failures and progressive collapse. Progressive collapse remains a critical failure mode in such bridges, requiring advanced monitoring and damage identification methods. This study presents an optimized Machine Learning (ML) framework for early-stage localization of initial component failure. The framework involved a validated Finite Element (FE) model of a scaled truss bridge, experimentally verified under controlled damage scenarios. The task is a multi-class classification covering all component failures under unseen loading conditions. Input features include natural frequencies and nodal deflections, with the dataset augmented using 5% Gaussian noise to simulate real-world measurement uncertainty. Four supervised ML models are tuned and evaluated using conventional performance metrics together with epistemic uncertainty. Initial results demonstrate consistent performance across three models (Deep MLP, 1D-CNN, and SVM) with an accuracy of ≈70% and strong confidence, capturing damage-sensitive behavior rather than load-specific patterns. The extended analyses reveal stable behavior among the three models under varying noise levels, class configurations, and feature sets. The observed misclassification trends reflect the bridge’s structural symmetry, while prediction confidence aligns with identified sensitivity patterns. The findings reveal that predictive performance is primarily influenced by the quality and representation of physics-informed features rather than ML’s architecture. This provides practical insights into sensor integration, feature selection, and uncertainty-aware decisions, supporting more reliable and computationally efficient monitoring solutions.
