Articles de revista
http://hdl.handle.net/2117/3924
2024-03-29T07:30:55Z
-
ENN: a neural network with DCT adaptive activation functions
http://hdl.handle.net/2117/405515
ENN: a neural network with DCT adaptive activation functions
Martínez Gost, Marc; Pérez Neira, Ana Isabel; Lagunas Hernandez, Miguel A.
The expressiveness of neural networks highly depends on the nature of the activation function, although these are usually assumed predefined and fixed during the training stage. Under a signal processing perspective, in this paper we present Expressive Neural Network (ENN), a novel model in which the non-linear activation functions are modeled using the Discrete Cosine Transform (DCT) and adapted using backpropagation during training. This parametrization keeps the number of trainable parameters low, is appropriate for gradient-based schemes, and adapts to different learning tasks. This is the first non-linear model for activation functions that relies on a signal processing perspective, providing high flexibility and expressiveness to the network. We contribute with insights in the explainability of the network at convergence by recovering the concept of bump, this is, the response of each activation function in the output space. Finally, through exhaustive experiments we show that the model can adapt to classification and regression tasks. The performance of ENN outperforms state of the art benchmarks, providing above a 40% gap in accuracy in some scenarios.
2024-03-28T12:13:49Z
Martínez Gost, Marc
Pérez Neira, Ana Isabel
Lagunas Hernandez, Miguel A.
The expressiveness of neural networks highly depends on the nature of the activation function, although these are usually assumed predefined and fixed during the training stage. Under a signal processing perspective, in this paper we present Expressive Neural Network (ENN), a novel model in which the non-linear activation functions are modeled using the Discrete Cosine Transform (DCT) and adapted using backpropagation during training. This parametrization keeps the number of trainable parameters low, is appropriate for gradient-based schemes, and adapts to different learning tasks. This is the first non-linear model for activation functions that relies on a signal processing perspective, providing high flexibility and expressiveness to the network. We contribute with insights in the explainability of the network at convergence by recovering the concept of bump, this is, the response of each activation function in the output space. Finally, through exhaustive experiments we show that the model can adapt to classification and regression tasks. The performance of ENN outperforms state of the art benchmarks, providing above a 40% gap in accuracy in some scenarios.
-
Sidelobe suppression for multicarrier signals via structured spectral precoding
http://hdl.handle.net/2117/405505
Sidelobe suppression for multicarrier signals via structured spectral precoding
Hussain, Khawar; López Valcarce, Roberto; Rey Micolau, Francesc; Sala Álvarez, José; Villares Piera, Nemesio Javier
Reducing the large sidelobes of multicarrier signals is crucial to prevent adjacent channel interference. Spectral precoding is an effective approach toward this goal, at the expense of throughput loss due to precoder redundancy; thus, it is of interest to explore alternative precoder designs with improved performance at lower redundancies. We present a novel precoder which minimizes radiated power within a user-selectable frequency region. The structure of the precoding matrix is chosen to allow efficient mitigation of in-band distortion at the receiver by means of iterative and successive interference cancellation, while completely avoiding distortion to protected and pilot subcarriers. By exploiting the low-rank properties of constituent blocks, computational complexity can be significantly reduced with little impact on sidelobe reduction. Simulation results show the benefits of the proposed design, which is particularly effective in redundancy-limited settings targeting high spectral efficiency.
2024-03-28T11:13:20Z
Hussain, Khawar
López Valcarce, Roberto
Rey Micolau, Francesc
Sala Álvarez, José
Villares Piera, Nemesio Javier
Reducing the large sidelobes of multicarrier signals is crucial to prevent adjacent channel interference. Spectral precoding is an effective approach toward this goal, at the expense of throughput loss due to precoder redundancy; thus, it is of interest to explore alternative precoder designs with improved performance at lower redundancies. We present a novel precoder which minimizes radiated power within a user-selectable frequency region. The structure of the precoding matrix is chosen to allow efficient mitigation of in-band distortion at the receiver by means of iterative and successive interference cancellation, while completely avoiding distortion to protected and pilot subcarriers. By exploiting the low-rank properties of constituent blocks, computational complexity can be significantly reduced with little impact on sidelobe reduction. Simulation results show the benefits of the proposed design, which is particularly effective in redundancy-limited settings targeting high spectral efficiency.
-
Evaluating scalability, resiliency, and load balancing in software-defined networking
http://hdl.handle.net/2117/405073
Evaluating scalability, resiliency, and load balancing in software-defined networking
Barbecho Bautista, Pablo; Comellas Colomé, Jaume; Urquiza Aguiar, Luis Felipe
With emerging technologies like cloud computing and big data, managing traditional networks has become more demanding. Software-defined networking (SDN) promises faster implementation, flexibility, and simplified network management. However, due to SDN’s centralized nature, it encounters limitations. SDN controllers should have enough processing power to deal with a high amount of flow. In addition, a single point of failure may affect the network’s resiliency. For these issues, multi-instance implementation enables distributed control. However, this solution implies an intrinsic controller-to-controller synchronization channel. In this article, we propose different failure scenarios in both the data and control planes to provide network administrators with a clear view of the constraints of network reliability, load balancing, and scalability in SDN environments. The simulation results show that, regarding resiliency, SDN networks require half the time compared to traditional networks in order to recover from a link failure. Regarding load-balancing capabilities, load balancing is not guaranteed with the reactive forwarding approach (on-demand flow installation). Lastly, the SDN multi-instance solution impacts the network performance by between 1% and 21% compared to the single-instance case.
This article belongs to the Proceedings of XXXI Conference on Electrical and Electronic Engineering, Quito, Ecuador, 29 November–1 December 2023
2024-03-21T11:27:03Z
Barbecho Bautista, Pablo
Comellas Colomé, Jaume
Urquiza Aguiar, Luis Felipe
With emerging technologies like cloud computing and big data, managing traditional networks has become more demanding. Software-defined networking (SDN) promises faster implementation, flexibility, and simplified network management. However, due to SDN’s centralized nature, it encounters limitations. SDN controllers should have enough processing power to deal with a high amount of flow. In addition, a single point of failure may affect the network’s resiliency. For these issues, multi-instance implementation enables distributed control. However, this solution implies an intrinsic controller-to-controller synchronization channel. In this article, we propose different failure scenarios in both the data and control planes to provide network administrators with a clear view of the constraints of network reliability, load balancing, and scalability in SDN environments. The simulation results show that, regarding resiliency, SDN networks require half the time compared to traditional networks in order to recover from a link failure. Regarding load-balancing capabilities, load balancing is not guaranteed with the reactive forwarding approach (on-demand flow installation). Lastly, the SDN multi-instance solution impacts the network performance by between 1% and 21% compared to the single-instance case.
-
Modified oriented dihedral model for scattering characteristic description with PolSAR data
http://hdl.handle.net/2117/403917
Modified oriented dihedral model for scattering characteristic description with PolSAR data
Chen, Yifan; Zhang, Lamei; Mallorquí Franquet, Jordi Joan; Zou, Bin
Dihedral is a common structure in polarimetric SAR images and can be found on many man-made targets. Many researchers have proposed different dihedral models, but the accuracy of these models is limited. In this case, the feature extraction methods based on these models are also not effective enough, which affects the subsequent applications such as target detection. Therefore, it is necessary to propose a new and accurate scattering model, which can be applied to dihedral with different orientation angles for feature extraction and target detection. In this paper, a general scattering model called modified oriented dihedral scattering model (MODM) is proposed based on physical optics (PO) and geometric optics (GO) of high-frequency approximation techniques. By analyzing the propagation and reflection of electromagnetic wave, MODM can accurately describe the scattering characteristic of dihedral for all observation conditions. In order to apply the model to real PolSAR images, MODM is introduced into a new feature extraction method, which is called five-scattering component polarimetric decomposition method (MODM-5SD). Feature extraction and target detection experiments of buildings with various oriented dihedral structures are performed using different data sets, which show that dihedral scattering components from oriented dihedral structures can be more effectively extracted by MODM-5SD. In addition, more buildings with oriented dihedral structures can be detected with the features from MODM-5SD. The experimental results show that MODM can more accurately describe the scattering characteristic of dihedral, which can be further applied for scattering characterization and feature extraction of targets with typical dihedral structures.
2024-03-07T11:16:45Z
Chen, Yifan
Zhang, Lamei
Mallorquí Franquet, Jordi Joan
Zou, Bin
Dihedral is a common structure in polarimetric SAR images and can be found on many man-made targets. Many researchers have proposed different dihedral models, but the accuracy of these models is limited. In this case, the feature extraction methods based on these models are also not effective enough, which affects the subsequent applications such as target detection. Therefore, it is necessary to propose a new and accurate scattering model, which can be applied to dihedral with different orientation angles for feature extraction and target detection. In this paper, a general scattering model called modified oriented dihedral scattering model (MODM) is proposed based on physical optics (PO) and geometric optics (GO) of high-frequency approximation techniques. By analyzing the propagation and reflection of electromagnetic wave, MODM can accurately describe the scattering characteristic of dihedral for all observation conditions. In order to apply the model to real PolSAR images, MODM is introduced into a new feature extraction method, which is called five-scattering component polarimetric decomposition method (MODM-5SD). Feature extraction and target detection experiments of buildings with various oriented dihedral structures are performed using different data sets, which show that dihedral scattering components from oriented dihedral structures can be more effectively extracted by MODM-5SD. In addition, more buildings with oriented dihedral structures can be detected with the features from MODM-5SD. The experimental results show that MODM can more accurately describe the scattering characteristic of dihedral, which can be further applied for scattering characterization and feature extraction of targets with typical dihedral structures.
-
On the design of a network digital twin for the radio access network in 5G and beyond
http://hdl.handle.net/2117/403890
On the design of a network digital twin for the radio access network in 5G and beyond
Vilà Muñoz, Irene; Sallent Roig, Oriol; Pérez Romero, Jordi
A Network Digital Twin (NDT) is a high-fidelity digital mirror of a real network. Given the increasing complexity of 5G and beyond networks, the use of an NDT becomes useful as a platform for testing configurations and algorithms prior to their application in the real network, as well as for predicting the performance of such algorithms under different conditions. While an NDT can be defined for the different subsystems of the network, this paper proposes an NDT architecture focusing on the Radio Access Network (RAN), describing the components to represent and model the operation of the different RAN elements, and to perform emulations. Different application use cases are identified, and among them, the paper puts the focus on the training of Reinforcement Learning (RL) solutions for the RAN. For this use case, the paper introduces a framework aligned with O-RAN specifications and discusses the functionalities needed to integrate the NDT. This use case is illustrated with the description of a RAN NDT implementation used for training an RL-based capacity-sharing solution for network slicing. Presented results demonstrate that the implemented RAN NDT is a suitable platform to successfully train the RL solution, achieving service-level agreement satisfaction values above 85%.
2024-03-06T15:31:03Z
Vilà Muñoz, Irene
Sallent Roig, Oriol
Pérez Romero, Jordi
A Network Digital Twin (NDT) is a high-fidelity digital mirror of a real network. Given the increasing complexity of 5G and beyond networks, the use of an NDT becomes useful as a platform for testing configurations and algorithms prior to their application in the real network, as well as for predicting the performance of such algorithms under different conditions. While an NDT can be defined for the different subsystems of the network, this paper proposes an NDT architecture focusing on the Radio Access Network (RAN), describing the components to represent and model the operation of the different RAN elements, and to perform emulations. Different application use cases are identified, and among them, the paper puts the focus on the training of Reinforcement Learning (RL) solutions for the RAN. For this use case, the paper introduces a framework aligned with O-RAN specifications and discusses the functionalities needed to integrate the NDT. This use case is illustrated with the description of a RAN NDT implementation used for training an RL-based capacity-sharing solution for network slicing. Presented results demonstrate that the implemented RAN NDT is a suitable platform to successfully train the RL solution, achieving service-level agreement satisfaction values above 85%.
-
An assessment of SAOCOM L-Band PolInSAR capabilities for canopy height estimation: A case study over managed forests in Argentina
http://hdl.handle.net/2117/403450
An assessment of SAOCOM L-Band PolInSAR capabilities for canopy height estimation: A case study over managed forests in Argentina
Seppi, Santiago Ariel; López Martínez, Carlos; Joseau, Marisa Jacqueline
This work presents the first results of canopy height mapping with L -Band SAOCOM data by means of Polarimetric SAR Interferometry (PolInSAR). For this study case, a colocated temporal series of SAOCOM fully polarimetric images covering the year 2021 was acquired, with a temporal baseline of 8 to 16 days between each acquisition. The study area corresponds to managed forests in Corrientes, Argentina, one of the main forest production regions in the country. Field measurements provided by forest owners were available in order to validate the obtained results, along with canopy height measurements from the Global Ecosystem Dynamics Investigation mission. A multi-baseline PolInSAR approach was adopted, following the methodology proposed by previous works. The results show that this multi-baseline selection approach is more appropriate than a single-baseline one, given the high variability that SAOCOM presents in the spatial baseline between acquisitions, and the lack of orbital control over this parameter. The height maps obtained over a validation site yielded an average R2 of 0.72 and a mean RMSE of 2.35 m., in agreement with the figures obtained in similar studies. This study presents, for the first time, canopy height maps of 8 and 16-day temporal baselines, L -band, orbital interferograms. The conclusions yielded by this research represent a qualitative and quantitative evaluation of SAOCOM PolInSAR capabilities for forest height mapping. It is also relevant to assess the suitability of the RVoG model to characterize the structure of these forests with L-band data retrieved by the Argentinean constellation.
2024-02-29T12:31:19Z
Seppi, Santiago Ariel
López Martínez, Carlos
Joseau, Marisa Jacqueline
This work presents the first results of canopy height mapping with L -Band SAOCOM data by means of Polarimetric SAR Interferometry (PolInSAR). For this study case, a colocated temporal series of SAOCOM fully polarimetric images covering the year 2021 was acquired, with a temporal baseline of 8 to 16 days between each acquisition. The study area corresponds to managed forests in Corrientes, Argentina, one of the main forest production regions in the country. Field measurements provided by forest owners were available in order to validate the obtained results, along with canopy height measurements from the Global Ecosystem Dynamics Investigation mission. A multi-baseline PolInSAR approach was adopted, following the methodology proposed by previous works. The results show that this multi-baseline selection approach is more appropriate than a single-baseline one, given the high variability that SAOCOM presents in the spatial baseline between acquisitions, and the lack of orbital control over this parameter. The height maps obtained over a validation site yielded an average R2 of 0.72 and a mean RMSE of 2.35 m., in agreement with the figures obtained in similar studies. This study presents, for the first time, canopy height maps of 8 and 16-day temporal baselines, L -band, orbital interferograms. The conclusions yielded by this research represent a qualitative and quantitative evaluation of SAOCOM PolInSAR capabilities for forest height mapping. It is also relevant to assess the suitability of the RVoG model to characterize the structure of these forests with L-band data retrieved by the Argentinean constellation.
-
Wavefront-modified vector beams for THz cornea spectroscopy
http://hdl.handle.net/2117/403444
Wavefront-modified vector beams for THz cornea spectroscopy
Lamberg, Joel; Zarrinkhat, Faezeh; Tamminen, Aleksi; Baggio, Mariangela; Ala-Laurinaho, Juha; Rius Casals, Juan Manuel; Romeu Robert, Jordi; Khaled, Elsayed E. M.; Taylor, Zachary
Terahertz spectroscopy is a promising method to diagnose ocular diseases, where the cornea is typically imaged by Gaussian beams. However, the beam’s mismatch with the cornea’s spherical surface produces a 5-10 % error in analysis. We investigate cornea spectroscopy with wavefront-modified vector beams, reducing the original analysis error to less than 0.5 %. Vector beams are synthesized by our developed 3D Angular Spectrum Method expanded to vector spherical harmonic presentation, allowing wavefront modification and scattering analysis from 100-layer cornea models. We show that wavefront-modified spherical vector beams possess increased accuracy and non-sensitive focusing on cornea spectroscopy compared to the Gaussian beams. Additionally, we investigate wavefront-modified cylindrical vector beams, which show frequency-dependent scattering power arising from s- and p-polarizations. As a result, these beams are unsuitable for cornea spectroscopy, although they have potential for optical force applications. Wavefront-modified vector beams can be applied to spherical target spectroscopy and optical force applications, such as medicine, medical imaging, and optical tweezers.
2024-02-29T11:44:12Z
Lamberg, Joel
Zarrinkhat, Faezeh
Tamminen, Aleksi
Baggio, Mariangela
Ala-Laurinaho, Juha
Rius Casals, Juan Manuel
Romeu Robert, Jordi
Khaled, Elsayed E. M.
Taylor, Zachary
Terahertz spectroscopy is a promising method to diagnose ocular diseases, where the cornea is typically imaged by Gaussian beams. However, the beam’s mismatch with the cornea’s spherical surface produces a 5-10 % error in analysis. We investigate cornea spectroscopy with wavefront-modified vector beams, reducing the original analysis error to less than 0.5 %. Vector beams are synthesized by our developed 3D Angular Spectrum Method expanded to vector spherical harmonic presentation, allowing wavefront modification and scattering analysis from 100-layer cornea models. We show that wavefront-modified spherical vector beams possess increased accuracy and non-sensitive focusing on cornea spectroscopy compared to the Gaussian beams. Additionally, we investigate wavefront-modified cylindrical vector beams, which show frequency-dependent scattering power arising from s- and p-polarizations. As a result, these beams are unsuitable for cornea spectroscopy, although they have potential for optical force applications. Wavefront-modified vector beams can be applied to spherical target spectroscopy and optical force applications, such as medicine, medical imaging, and optical tweezers.
-
Curved boundary integral method for electromagnetic fields
http://hdl.handle.net/2117/403441
Curved boundary integral method for electromagnetic fields
Lamberg, Joel; Zarrinkhat, Faezeh; Tamminen, Aleksi; Ala-Laurinaho, Juha; Rius Casals, Juan Manuel; Romeu Robert, Jordi; Khaled, Elsayed; Taylor, Zachary
The angular spectrum method is a rigorous method to synthesize near and far-field electromagnetic beams from planar field distributions. However, this limitation of planar surfaces has restricted its applicability to beams with simple focal planes. We propose a curved boundary integral method (CBIM) to synthesize electromagnetic beams from arbitrary surfaces to address this limitation and expand the method’s scope to synthesize beams from and between shaped objects. This study presents a detailed theoretical framework behind the CBIM and validates its effectiveness and accuracy with a comprehensive set of simulations. Additionally, we present mathematical proof to support our proposal. The proposed method satisfies Maxwell’s equations and significantly benefits optical systems and inverse beam design. It allows for analyzing electromagnetic forward/backward propagation between optical elements using a single method. It is also valuable for optical force beam design and analysis.
© 2023 Optica Publishing Group under the terms of the Open Access Publishing Agreement. Users may use, reuse, and build upon the article, or use the article for text or data mining, so long as such uses are for noncommercial purposes and appropriate attribution is maintained. All other rights are reserved.
2024-02-29T11:24:51Z
Lamberg, Joel
Zarrinkhat, Faezeh
Tamminen, Aleksi
Ala-Laurinaho, Juha
Rius Casals, Juan Manuel
Romeu Robert, Jordi
Khaled, Elsayed
Taylor, Zachary
The angular spectrum method is a rigorous method to synthesize near and far-field electromagnetic beams from planar field distributions. However, this limitation of planar surfaces has restricted its applicability to beams with simple focal planes. We propose a curved boundary integral method (CBIM) to synthesize electromagnetic beams from arbitrary surfaces to address this limitation and expand the method’s scope to synthesize beams from and between shaped objects. This study presents a detailed theoretical framework behind the CBIM and validates its effectiveness and accuracy with a comprehensive set of simulations. Additionally, we present mathematical proof to support our proposal. The proposed method satisfies Maxwell’s equations and significantly benefits optical systems and inverse beam design. It allows for analyzing electromagnetic forward/backward propagation between optical elements using a single method. It is also valuable for optical force beam design and analysis.
-
Quality control of Doppler spectra from a vertically pointing, S-band profiling radar
http://hdl.handle.net/2117/403437
Quality control of Doppler spectra from a vertically pointing, S-band profiling radar
Belak, Susan L.; Tanamachi, Robin; Asel, Matthew L.; Dennany, Grant; Gnanasambandam, Abhiram; Frasier, Steve; Rocadenbosch Burillo, Francisco
This study describes a novel combination of methods to remove spurious spectral peaks, or “spurs,” from Doppler spectra produced by a vertically pointing, S-band radar. The University of Massachusetts S-band frequency-modulated, continuous-wave radar (UMass FMCW) was deployed to monitor the growth of the CBL over northern Alabama during the VORTEX–Southeast field campaign in 2016. The Doppler spectra contained spurs caused by high-voltage switching power supplies in the traveling wave tube amplifier. In the original data processing scheme for this radar, a median filtering method was used to eliminate most of the spurs, but the largest ones persisted, which significantly degraded the quality of derived radar moments (e.g., reflectivity, Doppler velocity, and spectrum width) and hindered further analysis of these data (e.g., hydrometeor classification and boundary layer height tracking). Our technique for removing the spurs consists of three steps: (i) a Laplacian filter identifies and masks peaks in the spectra that are characteristic of the spurs in shape and amplitude, (ii) an in-painting method then fills in the masked area based on surrounding data, and (iii) the moments data (e.g., reflectivity, Doppler velocity, and spectrum width) are then recomputed using a coherent power technique. This combination of techniques was more effective than the median filter at removing the largest spurs from the Doppler spectra, and preserved more of the underlying Doppler spectral structure of the scatterers. Performance of both the median-filter and the in-painting methods are assessed through statistical analysis of the spectral power differences. Downstream products, such as boundary layer height detection, are more easily derived from the recomputed moments.
2024-02-29T10:50:02Z
Belak, Susan L.
Tanamachi, Robin
Asel, Matthew L.
Dennany, Grant
Gnanasambandam, Abhiram
Frasier, Steve
Rocadenbosch Burillo, Francisco
This study describes a novel combination of methods to remove spurious spectral peaks, or “spurs,” from Doppler spectra produced by a vertically pointing, S-band radar. The University of Massachusetts S-band frequency-modulated, continuous-wave radar (UMass FMCW) was deployed to monitor the growth of the CBL over northern Alabama during the VORTEX–Southeast field campaign in 2016. The Doppler spectra contained spurs caused by high-voltage switching power supplies in the traveling wave tube amplifier. In the original data processing scheme for this radar, a median filtering method was used to eliminate most of the spurs, but the largest ones persisted, which significantly degraded the quality of derived radar moments (e.g., reflectivity, Doppler velocity, and spectrum width) and hindered further analysis of these data (e.g., hydrometeor classification and boundary layer height tracking). Our technique for removing the spurs consists of three steps: (i) a Laplacian filter identifies and masks peaks in the spectra that are characteristic of the spurs in shape and amplitude, (ii) an in-painting method then fills in the masked area based on surrounding data, and (iii) the moments data (e.g., reflectivity, Doppler velocity, and spectrum width) are then recomputed using a coherent power technique. This combination of techniques was more effective than the median filter at removing the largest spurs from the Doppler spectra, and preserved more of the underlying Doppler spectral structure of the scatterers. Performance of both the median-filter and the in-painting methods are assessed through statistical analysis of the spectral power differences. Downstream products, such as boundary layer height detection, are more easily derived from the recomputed moments.
-
Land, jet stream, and other atmospheric effects on burned area estimation during the South Asian heatwave of 2022
http://hdl.handle.net/2117/403429
Land, jet stream, and other atmospheric effects on burned area estimation during the South Asian heatwave of 2022
Irawan, Amir Mustofa; Vall-Llossera Ferran, Mercedes Magdalena; López Martínez, Carlos; Camps Carmona, Adriano José; Chaparro Danon, David; Portal González, Gerard; Pablos Hernández, Miriam; Alonso González, Alberto
Understanding the key variables that characterise fire propagation is important for a better estimation of fire events and their impacts. This study uses machine learning combined with satellite remote sensing and atmospheric modelled data to enhance estimations of burned areas. It focuses on the intense early summer weather patterns in South Asia during April and May 2022 and explores the relationship between environmental factors and fire spread. The study employs various algorithms, including random forest, extra trees, extreme gradient boosting (XGBoost), gradient boosting regressor, support vector regressor and neural networks. XGBoost proves to be the most accurate approach. An isolation forest algorithm is used to adjust for outliers in burned area estimations. The comprehensive analysis conducted includes the identification of key variables and sensitivity tests incorporating changes of up to 25 % in natural environmental conditions to assess the model’s consistency. The results indicate that integrating vegetation, atmospheric, and human-related variables with the XGBoost algorithm, and incorporating outlier adjustments leads to the most effective performance (R2 ≥ 0.7), with jet stream variables enhancing the accuracy by approximately 11.5 %. The study highlights the notable impact on fire propagation of increases in the value of 300-hPa meridional circulation index flow (MCI300) and a high 500-hPa geopotential height anomaly (ΔZ500), indicating the development of strong atmospheric blocking (upper tropospheric ridge). As compared to other factors, e.g. land surface temperature, vapour pressure deficit, soil moisture and vegetation optical depth, the impact of changes in jet stream metrics (MCI300 and ΔZ500) was more pronounced, indicating greater sensitivity. These insights emphasise the complexity of fire spread, and the importance of using atmospheric factors to estimate burned areas, particularly during severe heatwaves.
2024-02-29T07:59:40Z
Irawan, Amir Mustofa
Vall-Llossera Ferran, Mercedes Magdalena
López Martínez, Carlos
Camps Carmona, Adriano José
Chaparro Danon, David
Portal González, Gerard
Pablos Hernández, Miriam
Alonso González, Alberto
Understanding the key variables that characterise fire propagation is important for a better estimation of fire events and their impacts. This study uses machine learning combined with satellite remote sensing and atmospheric modelled data to enhance estimations of burned areas. It focuses on the intense early summer weather patterns in South Asia during April and May 2022 and explores the relationship between environmental factors and fire spread. The study employs various algorithms, including random forest, extra trees, extreme gradient boosting (XGBoost), gradient boosting regressor, support vector regressor and neural networks. XGBoost proves to be the most accurate approach. An isolation forest algorithm is used to adjust for outliers in burned area estimations. The comprehensive analysis conducted includes the identification of key variables and sensitivity tests incorporating changes of up to 25 % in natural environmental conditions to assess the model’s consistency. The results indicate that integrating vegetation, atmospheric, and human-related variables with the XGBoost algorithm, and incorporating outlier adjustments leads to the most effective performance (R2 ≥ 0.7), with jet stream variables enhancing the accuracy by approximately 11.5 %. The study highlights the notable impact on fire propagation of increases in the value of 300-hPa meridional circulation index flow (MCI300) and a high 500-hPa geopotential height anomaly (ΔZ500), indicating the development of strong atmospheric blocking (upper tropospheric ridge). As compared to other factors, e.g. land surface temperature, vapour pressure deficit, soil moisture and vegetation optical depth, the impact of changes in jet stream metrics (MCI300 and ΔZ500) was more pronounced, indicating greater sensitivity. These insights emphasise the complexity of fire spread, and the importance of using atmospheric factors to estimate burned areas, particularly during severe heatwaves.