Carlos Carrillo
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Carlos Carrillo has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- Ana Cortés
Universitat Autónoma de Barcelona
8 shared publications - Tomàs Margalef
Universitat Autónoma de Barcelona
7 shared publications - Irene González
Universitat Autónoma de Barcelona
5 shared publications Ana Cortés
Computer Architecture and Operating Systems Department, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Spain
3 shared publicationsAntonio Espinosa
3 shared publicationsTomàs Margalef
Computer Architecture and Operating Systems Department, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Spain
3 shared publications
Latest Publications
- Open Access
- 10 October 2026
- Published by Elsevier in SSRN Electronic Journal
GPU-Accelerated Elliptical Wave Propagation for High-Resolution Wildfire Simulation in IoT-Enabled Decision-Support Systems
The Internet of Things (IoT) is enabling advanced environmental monitoring and disaster-management systems in which heterogeneous sensing data can support time-critical decision making. In wildfire management, however, transforming environmental observations into high-resolution fire-spread predictions remains computationally demanding, limiting the integration of simulation models into operational IoT-enabled workflows. This limitation is particularly relevant for Elliptical Wave Propagation (EWP) simulators, where perimeter propagation and reconstruction become increasingly expensive as the fire-front resolution increases. This work addresses this bottleneck through a GPU-based redesign of the two main stages of EWP fire-front evolution. Perimeter Propagation is reformulated as a fine-grained data-parallel process that maps independent fire-front points onto CUDA threads. Perimeter Reconstruction combines parallel segment preprocessing with a bounded candidate-intersection search. For the evaluated wildfire scenario, a candidate radius of ten segments reproduced the reconstruction obtained with larger neighborhoods across all considered resolutions. For a fixed neighborhood size, this strategy reduces intersection-search complexity from quadratic to linear in the number of perimeter segments. The proposed approach was integrated into FARSITE and evaluated it on the Arkadia wildfire case with perimeter resolutions from 100m to 5m. At the finest resolution, more than 19.5 million perimeter points were processed, reducing total simulation time from 57.1 to 2.8 minutes and achieving an end-to-end speedup above 20. These results demonstrate the potential of GPU-accelerated EWP simulation as a computational component of time-sensitive IoT-enabled wildfire decision-support systems.
- 10 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessGPU-Accelerated Elliptical Wave Propagation for High-Resolution Wildfire Simulation in IoT-Enabled Decision-Support Systems
The Internet of Things (IoT) is enabling advanced environmental monitoring and disaster-management systems in which heterogeneous sensing data can support time-critical decision making. In wildfire management, however, transforming environmental observations into high-resolution fire-spread predictions remains computationally demanding, limiting the integration of simulation models into operational IoT-enabled workflows. This limitation is particularly relevant for Elliptical Wave Propagation (EWP) simulators, where perimeter propagation and reconstruction become increasingly expensive as the fire-front resolution increases. This work addresses this bottleneck through a GPU-based redesign of the two main stages of EWP fire-front evolution. Perimeter Propagation is reformulated as a fine-grained data-parallel process that maps independent fire-front points onto CUDA threads. Perimeter Reconstruction combines parallel segment preprocessing with a bounded candidate-intersection search. For the evaluated wildfire scenario, a candidate radius of ten segments reproduced the reconstruction obtained with larger neighborhoods across all considered resolutions. For a fixed neighborhood size, this strategy reduces intersection-search complexity from quadratic to linear in the number of perimeter segments. The proposed approach was integrated into FARSITE and evaluated it on the Arkadia wildfire case with perimeter resolutions from 100m to 5m. At the finest resolution, more than 19.5 million perimeter points were processed, reducing total simulation time from 57.1 to 2.8 minutes and achieving an end-to-end speedup above 20. These results demonstrate the potential of GPU-accelerated EWP simulation as a computational component of time-sensitive IoT-enabled wildfire decision-support systems.
- Open Access
- 19 April 2026
- Published by Elsevier in SSRN Electronic Journal
Boundary α-Shape: A novel way to generate complex polygons from planar point sets
The Boundary α-Shape method is a 2D shape‑reconstruction proposal designed to infer coherent object boundaries from an unstructured set of points. Although it belongs to the Delaunay triangulation family of algorithms, it specifically addresses several limitations commonly found in traditional α‑shape and triangulation based approaches. In particular, the method is engineered to correctly handle empty regions within the point distribution, allowing it to recover shapes with holes or internal cavities. It also supports the representation of multiple disconnected objects without requiring manual intervention.A key contribution of the Boundary α-Shape approach is its robustness against noise and outliers, reducing the need for user‑defined parameters that often make classical methods sensitive or difficult to tune. The algorithm aims to automatically adapt to non‑uniform point densities by adjusting its reconstruction criteria in a data‑driven manner. Overall, it provides a more flexible and reliable reconstruction pipeline that preserves detailed boundary structures while minimizing parameter dependency and improving performance on real‑world, imperfect datasets.
- 19 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessBoundary α-Shape: A novel way to generate complex polygons from planar point sets
The Boundary α-Shape method is a 2D shape‑reconstruction proposal designed to infer coherent object boundaries from an unstructured set of points. Although it belongs to the Delaunay triangulation family of algorithms, it specifically addresses several limitations commonly found in traditional α‑shape and triangulation based approaches. In particular, the method is engineered to correctly handle empty regions within the point distribution, allowing it to recover shapes with holes or internal cavities. It also supports the representation of multiple disconnected objects without requiring manual intervention.A key contribution of the Boundary α-Shape approach is its robustness against noise and outliers, reducing the need for user‑defined parameters that often make classical methods sensitive or difficult to tune. The algorithm aims to automatically adapt to non‑uniform point densities by adjusting its reconstruction criteria in a data‑driven manner. Overall, it provides a more flexible and reliable reconstruction pipeline that preserves detailed boundary structures while minimizing parameter dependency and improving performance on real‑world, imperfect datasets.
- Open Access
- 2 April 2026
- Published by Elsevier in Environmental Modelling & Software
- Vol. 201
Optimizing canopy cover evaluation: A machine learning approach using LiDAR data
- 2 April 2026
- Published by Elsevier in Environmental Modelling & Software
- Vol. 201
Open AccessOptimizing canopy cover evaluation: A machine learning approach using LiDAR data
- 6 July 2025
- Published by Springer Nature
- p. 63-71
Remote Sensing AI for Crop Planting in Wildfire Fuel Mapping
- 6 July 2025
- Published by Springer Nature
- p. 63-71
Remote Sensing AI for Crop Planting in Wildfire Fuel Mapping
- Open Access
- 1 July 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-16
Zone adaptive fuel mapping for high resolution wildfire spread forecasting
Extreme wildfire events (EWE), although a rare natural hazard, account for a substantial portion of global wildfire damage, requiring proactive anticipation and mitigation due to their increasing occurrence. Wildfire spread simulators are crucial for reducing damage, but they rely heavily on accurate fuel maps, which are often outdated, have low resolution, or are unavailable in many regions. While land cover maps are more up-to-date, high-resolution globally, and widely available, there is no universally accepted method to convert land cover maps into fuel maps. In this work, an automatic methodology for generating high-resolution fuel maps from land cover maps called zone-adaptive fuel mapping (ZAFM) is proposed. ZAFM is a consistent local approach that makes use of public resources to create a fuel map. The proposed methodology has been tested using, as a study case, an EWE that occurred in the north-east of Spain during the summer of 2022. To assess the accuracy of the proposed fuel mapping method, we compared the fire spread forecast using the ZAFM fuel map with fire evolutions based on different fuel maps derived from the land cover map of the study area. The accuracy assessment, based on the F2-score metric, reveals that ZAFM achieves the highest F2-score of approximately 0.90, while the F2-scores for the other fuel maps range from 0.78 to 0.89, with no individual simulation reaching 0.90. ZAFM was also evaluated against other publicly available fuel maps covering Catalonia, and once again achieved higher F2-scores in the case study simulations. These results highlight the superior predictive performance of ZAFM and underscore the importance of using up-to-date, high-resolution data to improve wildfire spread forecasts. Furthermore, since ZAFM relies on open-access data maps, it can be applied worldwide with any available high-resolution land cover map.
- 1 July 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-16
Open AccessZone adaptive fuel mapping for high resolution wildfire spread forecasting
Extreme wildfire events (EWE), although a rare natural hazard, account for a substantial portion of global wildfire damage, requiring proactive anticipation and mitigation due to their increasing occurrence. Wildfire spread simulators are crucial for reducing damage, but they rely heavily on accurate fuel maps, which are often outdated, have low resolution, or are unavailable in many regions. While land cover maps are more up-to-date, high-resolution globally, and widely available, there is no universally accepted method to convert land cover maps into fuel maps. In this work, an automatic methodology for generating high-resolution fuel maps from land cover maps called zone-adaptive fuel mapping (ZAFM) is proposed. ZAFM is a consistent local approach that makes use of public resources to create a fuel map. The proposed methodology has been tested using, as a study case, an EWE that occurred in the north-east of Spain during the summer of 2022. To assess the accuracy of the proposed fuel mapping method, we compared the fire spread forecast using the ZAFM fuel map with fire evolutions based on different fuel maps derived from the land cover map of the study area. The accuracy assessment, based on the F2-score metric, reveals that ZAFM achieves the highest F2-score of approximately 0.90, while the F2-scores for the other fuel maps range from 0.78 to 0.89, with no individual simulation reaching 0.90. ZAFM was also evaluated against other publicly available fuel maps covering Catalonia, and once again achieved higher F2-scores in the case study simulations. These results highlight the superior predictive performance of ZAFM and underscore the importance of using up-to-date, high-resolution data to improve wildfire spread forecasts. Furthermore, since ZAFM relies on open-access data maps, it can be applied worldwide with any available high-resolution land cover map.
- Open Access
- 13 June 2025
- Published by Elsevier in Journal of Computational Science
- Vol. 90
Realistic wildfire growth simulations applying a differentiable parametric representation of the fire front based on Composite Bézier curves
- 13 June 2025
- Published by Elsevier in Journal of Computational Science
- Vol. 90
Open AccessRealistic wildfire growth simulations applying a differentiable parametric representation of the fire front based on Composite Bézier curves
- Open Access
- 2 May 2025
- Published by Elsevier in Journal of Computational Science
- Vol. 88
Edge computing driven forest fire spread simulation: An energy-aware study
- 2 May 2025
- Published by Elsevier in Journal of Computational Science
- Vol. 88
Open AccessEdge computing driven forest fire spread simulation: An energy-aware study
- 28 June 2024
- Published by Springer Nature
- p. 151-165
Enhancing the Realism of Wildfire Simulation Using Composite Bézier Curves
- 28 June 2024
- Published by Springer Nature
- p. 151-165
Enhancing the Realism of Wildfire Simulation Using Composite Bézier Curves
- Open Access
- 27 April 2024
- Published by Elsevier in Environmental Modelling & Software
- Vol. 177
Cloud-based urgent computing for forest fire spread prediction
- 27 April 2024
- Published by Elsevier in Environmental Modelling & Software
- Vol. 177
Open AccessCloud-based urgent computing for forest fire spread prediction
- Open Access
- 7 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 187-191
A new paradigm for forest fire spread prediction: Faster decisions at high resolution
- 7 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 187-191
Open AccessA new paradigm for forest fire spread prediction: Faster decisions at high resolution