Bryan Zhang, Dhruv Rao, Adam Yang, Lawrence Zhang, Rodz Andrie Amor
Weather forecasting is a vitally important tool for tasks ranging from planning day to day activities to disaster response planning. However, modeling weather has proven to be challenging task due to its chaotic and unpredictable nature. Each variable, from temperature to precipitation to wind, all influence the path the environment will take. As a result, all models tend to rapidly lose accuracy as the temporal range of their forecasts increase. Classical forecasting methods use a myriad of physics-based, numerical, and stochastic techniques to predict the change in weather variables over time. However, such forecasts often require a very large amount of data and are extremely computationally expensive. Furthermore, as climate and global weather patterns change, classical models are substantially more difficult and time-consuming to update for changing environments. Fortunately, with recent advances in deep learning and publicly available high quality weather datasets, deploying learning methods for estimating these complex systems has become feasible. The current state-of-the-art deep learning models have comparable accuracy to the industry standard numerical models and are becoming more ubiquitous in practice due to their adaptability. Our group seeks to improve upon existing deep learning based forecasting methods by increasing spatial resolutions of global weather predictions. Specifically, we are interested in performing super resolution (SR) on GraphCast temperature predictions by increasing the global precision from 1 degree of accuracy to 0.5 degrees, which is approximately 111km and 55km respectively.
{"title":"Super Resolution On Global Weather Forecasts","authors":"Bryan Zhang, Dhruv Rao, Adam Yang, Lawrence Zhang, Rodz Andrie Amor","doi":"arxiv-2409.11502","DOIUrl":"https://doi.org/arxiv-2409.11502","url":null,"abstract":"Weather forecasting is a vitally important tool for tasks ranging from\u0000planning day to day activities to disaster response planning. However, modeling\u0000weather has proven to be challenging task due to its chaotic and unpredictable\u0000nature. Each variable, from temperature to precipitation to wind, all influence\u0000the path the environment will take. As a result, all models tend to rapidly\u0000lose accuracy as the temporal range of their forecasts increase. Classical\u0000forecasting methods use a myriad of physics-based, numerical, and stochastic\u0000techniques to predict the change in weather variables over time. However, such\u0000forecasts often require a very large amount of data and are extremely\u0000computationally expensive. Furthermore, as climate and global weather patterns\u0000change, classical models are substantially more difficult and time-consuming to\u0000update for changing environments. Fortunately, with recent advances in deep\u0000learning and publicly available high quality weather datasets, deploying\u0000learning methods for estimating these complex systems has become feasible. The\u0000current state-of-the-art deep learning models have comparable accuracy to the\u0000industry standard numerical models and are becoming more ubiquitous in practice\u0000due to their adaptability. Our group seeks to improve upon existing deep\u0000learning based forecasting methods by increasing spatial resolutions of global\u0000weather predictions. Specifically, we are interested in performing super\u0000resolution (SR) on GraphCast temperature predictions by increasing the global\u0000precision from 1 degree of accuracy to 0.5 degrees, which is approximately\u0000111km and 55km respectively.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"41 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142267517","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
AI data-driven models (Graphcast, Pangu Weather, Fourcastnet, and SFNO) are explored for storyline-based climate attribution due to their short inference times, which can accelerate the number of events studied, and provide real time attributions when public attention is heightened. The analysis is framed on the extreme atmospheric river episode of February 2017 that contributed to the Oroville dam spillway incident in Northern California. Past and future simulations are generated by perturbing the initial conditions with the pre-industrial and the late-21st century temperature climate change signals, respectively. The simulations are compared to results from a dynamical model which represents plausible pseudo-realities under both climate environments. Overall, the AI models show promising results, projecting a 5-6 % increase in the integrated water vapor over the Oroville dam in the present day compared to the pre-industrial, in agreement with the dynamical model. Different geopotential-moisture-temperature dependencies are unveiled for each of the AI-models tested, providing valuable information for understanding the physicality of the attribution response. However, the AI models tend to simulate weaker attribution values than the pseudo-reality imagined by the dynamical model, suggesting some reduced extrapolation skill, especially for the late-21st century regime. Large ensembles generated with an AI model (>500 members) produced statistically significant present-day to pre-industrial attribution results, unlike the >20-member ensemble from the dynamical model. This analysis highlights the potential of AI models to conduct attribution analysis, while emphasizing future lines of work on explainable artificial intelligence to gain confidence in these tools, which can enable reliable attribution studies in real-time.
{"title":"Harnessing AI data-driven global weather models for climate attribution: An analysis of the 2017 Oroville Dam extreme atmospheric river","authors":"Jorge Baño-Medina, Agniv Sengupta, Allison Michaelis, Luca Delle Monache, Julie Kalansky, Duncan Watson-Parris","doi":"arxiv-2409.11605","DOIUrl":"https://doi.org/arxiv-2409.11605","url":null,"abstract":"AI data-driven models (Graphcast, Pangu Weather, Fourcastnet, and SFNO) are\u0000explored for storyline-based climate attribution due to their short inference\u0000times, which can accelerate the number of events studied, and provide real time\u0000attributions when public attention is heightened. The analysis is framed on the\u0000extreme atmospheric river episode of February 2017 that contributed to the\u0000Oroville dam spillway incident in Northern California. Past and future\u0000simulations are generated by perturbing the initial conditions with the\u0000pre-industrial and the late-21st century temperature climate change signals,\u0000respectively. The simulations are compared to results from a dynamical model\u0000which represents plausible pseudo-realities under both climate environments.\u0000Overall, the AI models show promising results, projecting a 5-6 % increase in\u0000the integrated water vapor over the Oroville dam in the present day compared to\u0000the pre-industrial, in agreement with the dynamical model. Different\u0000geopotential-moisture-temperature dependencies are unveiled for each of the\u0000AI-models tested, providing valuable information for understanding the\u0000physicality of the attribution response. However, the AI models tend to\u0000simulate weaker attribution values than the pseudo-reality imagined by the\u0000dynamical model, suggesting some reduced extrapolation skill, especially for\u0000the late-21st century regime. Large ensembles generated with an AI model (>500\u0000members) produced statistically significant present-day to pre-industrial\u0000attribution results, unlike the >20-member ensemble from the dynamical model.\u0000This analysis highlights the potential of AI models to conduct attribution\u0000analysis, while emphasizing future lines of work on explainable artificial\u0000intelligence to gain confidence in these tools, which can enable reliable\u0000attribution studies in real-time.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"52 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142267515","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Irene Ortiz, Ermioni Dimitropoulou, Pierre de Buyl, Nicolas Clerbaux, Javier García-Heras, Amin Jafarimoghaddam, Hugues Brenot, Jeroen van Gent, Klaus Sievers, Evelyn Otero, Parthiban Loganathan, Manuel Soler
Aviation's non-CO$_2$ effects, especially the impact of aviation-induced contrails, drive atmospheric changes and can influence climate dynamics. Although contrails are believed to contribute to global warming through their net warming effect, uncertainties persist due to the challenges in accurately measuring their radiative impacts. This study aims to address this knowledge gap by investigating the relationship between aviation-induced contrails, as observed in Meteosat Second Generation (MSG) satellite imagery, and their impact on radiative forcing (RF) over a two-week study. Results show that while daytime contrails generally have a cooling effect, the higher number of nighttime contrails results in a net warming effect over the entire day. Net RF values for detected contrails range approximately from -8 TW to 2.5 TW during the day and from 0 to 6 TW at night. Our findings also show a 41.03% increase in contrail coverage from January 24-30, 2023, to the same week in 2024, accompanied by a 128.7% rise in contrail radiative forcing (CRF), indicating greater warming from the added contrails. These findings highlight the necessity of considering temporal factors, such as the timing and duration of contrail formation, when assessing their overall warming impact. They also indicate a potential increase in contrail-induced warming from 2023 to 2024, attributable to the rise in contrail coverage. Further investigation into these trends is crucial for the development of effective mitigation strategies.
{"title":"Satellite-Based Quantification of Contrail Radiative Forcing over Europe: A Two-Week Analysis of Aviation-Induced Climate Effects","authors":"Irene Ortiz, Ermioni Dimitropoulou, Pierre de Buyl, Nicolas Clerbaux, Javier García-Heras, Amin Jafarimoghaddam, Hugues Brenot, Jeroen van Gent, Klaus Sievers, Evelyn Otero, Parthiban Loganathan, Manuel Soler","doi":"arxiv-2409.10166","DOIUrl":"https://doi.org/arxiv-2409.10166","url":null,"abstract":"Aviation's non-CO$_2$ effects, especially the impact of aviation-induced\u0000contrails, drive atmospheric changes and can influence climate dynamics.\u0000Although contrails are believed to contribute to global warming through their\u0000net warming effect, uncertainties persist due to the challenges in accurately\u0000measuring their radiative impacts. This study aims to address this knowledge\u0000gap by investigating the relationship between aviation-induced contrails, as\u0000observed in Meteosat Second Generation (MSG) satellite imagery, and their\u0000impact on radiative forcing (RF) over a two-week study. Results show that while\u0000daytime contrails generally have a cooling effect, the higher number of\u0000nighttime contrails results in a net warming effect over the entire day. Net RF\u0000values for detected contrails range approximately from -8 TW to 2.5 TW during\u0000the day and from 0 to 6 TW at night. Our findings also show a 41.03% increase\u0000in contrail coverage from January 24-30, 2023, to the same week in 2024,\u0000accompanied by a 128.7% rise in contrail radiative forcing (CRF), indicating\u0000greater warming from the added contrails. These findings highlight the\u0000necessity of considering temporal factors, such as the timing and duration of\u0000contrail formation, when assessing their overall warming impact. They also\u0000indicate a potential increase in contrail-induced warming from 2023 to 2024,\u0000attributable to the rise in contrail coverage. Further investigation into these\u0000trends is crucial for the development of effective mitigation strategies.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"57 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142267521","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Precipitation nowcasting is crucial for mitigating the impacts of severe weather events and supporting daily activities. Conventional models predominantly relying on radar data have limited performance in predicting cases with complex temporal features such as convection initiation, highlighting the need to integrate data from other sources for more comprehensive nowcasting. Unlike physics-based models, machine learning (ML)-based models offer promising solutions for efficiently integrating large volumes of diverse data. We present EF4INCA, a spatiotemporal Transformer model for precipitation nowcasting that integrates satellite- and ground-based observations with numerical weather prediction outputs. EF4INCA provides high-resolution forecasts over Austria, accurately predicting the location and shape of precipitation fields with a spatial resolution of 1 kilometre and a temporal resolution of 5 minutes, up to 90 minutes ahead. Our evaluation shows that EF4INCA outperforms conventional nowcasting models, including the operational model of Austria, particularly in scenarios with complex temporal features such as convective initiation and rapid weather changes. EF4INCA maintains higher accuracy in location forecasting but generates smoother fields at later prediction times compared to traditional models. Interpretation of our model showed that precipitation products and SEVIRI infrared channels CH7 and CH9 are the most important data streams. These results underscore the importance of combining data from different domains, including physics-based model products, with ML approaches. Our study highlights the robustness of EF4INCA and its potential for improved precipitation nowcasting. We provide access to our code repository, model weights, and the dataset curated for benchmarking, facilitating further development and application.
{"title":"Integrated nowcasting of convective precipitation with Transformer-based models using multi-source data","authors":"Çağlar Küçük, Aitor Atencia, Markus Dabernig","doi":"arxiv-2409.10367","DOIUrl":"https://doi.org/arxiv-2409.10367","url":null,"abstract":"Precipitation nowcasting is crucial for mitigating the impacts of severe\u0000weather events and supporting daily activities. Conventional models\u0000predominantly relying on radar data have limited performance in predicting\u0000cases with complex temporal features such as convection initiation,\u0000highlighting the need to integrate data from other sources for more\u0000comprehensive nowcasting. Unlike physics-based models, machine learning\u0000(ML)-based models offer promising solutions for efficiently integrating large\u0000volumes of diverse data. We present EF4INCA, a spatiotemporal Transformer model\u0000for precipitation nowcasting that integrates satellite- and ground-based\u0000observations with numerical weather prediction outputs. EF4INCA provides\u0000high-resolution forecasts over Austria, accurately predicting the location and\u0000shape of precipitation fields with a spatial resolution of 1 kilometre and a\u0000temporal resolution of 5 minutes, up to 90 minutes ahead. Our evaluation shows\u0000that EF4INCA outperforms conventional nowcasting models, including the\u0000operational model of Austria, particularly in scenarios with complex temporal\u0000features such as convective initiation and rapid weather changes. EF4INCA\u0000maintains higher accuracy in location forecasting but generates smoother fields\u0000at later prediction times compared to traditional models. Interpretation of our\u0000model showed that precipitation products and SEVIRI infrared channels CH7 and\u0000CH9 are the most important data streams. These results underscore the\u0000importance of combining data from different domains, including physics-based\u0000model products, with ML approaches. Our study highlights the robustness of\u0000EF4INCA and its potential for improved precipitation nowcasting. We provide\u0000access to our code repository, model weights, and the dataset curated for\u0000benchmarking, facilitating further development and application.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"198 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142267520","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Kirsten J. Mayer, Katherine Dagon, Maria J. Molina
Previous research has demonstrated that specific states of the climate system can lead to enhanced subseasonal predictability (i.e., state-dependent predictability). However, biases in Earth system models can affect the representation of these states and their subsequent evolution. Here, we present a machine learning framework to identify state-dependent biases in Earth system models. In particular, we investigate the utility of transfer learning with explainable neural networks to identify tropical state-dependent biases in historical simulations of the Energy Exascale Earth System Model version 2 (E3SMv2) relevant for midlatitude subseasonal predictability. Using a perfect model framework, we find transfer learning may require substantially more data than provided by present-day reanalysis datasets to update neural network weights, imparting a cautionary tale for future transfer learning approaches focused on subseasonal modes of variability.
{"title":"Can Transfer Learning be Used to Identify Tropical State-Dependent Bias Relevant to Midlatitude Subseasonal Predictability?","authors":"Kirsten J. Mayer, Katherine Dagon, Maria J. Molina","doi":"arxiv-2409.10755","DOIUrl":"https://doi.org/arxiv-2409.10755","url":null,"abstract":"Previous research has demonstrated that specific states of the climate system\u0000can lead to enhanced subseasonal predictability (i.e., state-dependent\u0000predictability). However, biases in Earth system models can affect the\u0000representation of these states and their subsequent evolution. Here, we present\u0000a machine learning framework to identify state-dependent biases in Earth system\u0000models. In particular, we investigate the utility of transfer learning with\u0000explainable neural networks to identify tropical state-dependent biases in\u0000historical simulations of the Energy Exascale Earth System Model version 2\u0000(E3SMv2) relevant for midlatitude subseasonal predictability. Using a perfect\u0000model framework, we find transfer learning may require substantially more data\u0000than provided by present-day reanalysis datasets to update neural network\u0000weights, imparting a cautionary tale for future transfer learning approaches\u0000focused on subseasonal modes of variability.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"189 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142267518","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Windstorms significantly impact the UK, causing extensive damage to property, disrupting society, and potentially resulting in loss of life. Accurate modelling and understanding of such events are essential for effective risk assessment and mitigation. However, the rarity of extreme windstorms results in limited observational data, which poses significant challenges for comprehensive analysis and insurance modelling. This dissertation explores the application of generative models to produce realistic synthetic wind field data, aiming to enhance the robustness of current CAT models used in the insurance industry. The study utilises hourly reanalysis data from the ERA5 dataset, which covers the period from 1940 to 2022. Three models, including standard GANs, WGAN-GP, and U-net diffusion models, were employed to generate high-quality wind maps of the UK. These models are then evaluated using multiple metrics, including SSIM, KL divergence, and EMD, with some assessments performed in a reduced dimensionality space using PCA. The results reveal that while all models are effective in capturing the general spatial characteristics, each model exhibits distinct strengths and weaknesses. The standard GAN introduced more noise compared to the other models. The WGAN-GP model demonstrated superior performance, particularly in replicating statistical distributions. The U-net diffusion model produced the most visually coherent outputs but struggled slightly in replicating peak intensities and their statistical variability. This research underscores the potential of generative models in supplementing limited reanalysis datasets with synthetic data, providing valuable tools for risk assessment and catastrophe modelling. However, it is important to select appropriate evaluation metrics that assess different aspects of the generated outputs. Future work could refine these models and incorporate more ...
{"title":"Using Generative Models to Produce Realistic Populations of the United Kingdom Windstorms","authors":"Etron Yee Chun Tsoi","doi":"arxiv-2409.10696","DOIUrl":"https://doi.org/arxiv-2409.10696","url":null,"abstract":"Windstorms significantly impact the UK, causing extensive damage to property,\u0000disrupting society, and potentially resulting in loss of life. Accurate\u0000modelling and understanding of such events are essential for effective risk\u0000assessment and mitigation. However, the rarity of extreme windstorms results in\u0000limited observational data, which poses significant challenges for\u0000comprehensive analysis and insurance modelling. This dissertation explores the\u0000application of generative models to produce realistic synthetic wind field\u0000data, aiming to enhance the robustness of current CAT models used in the\u0000insurance industry. The study utilises hourly reanalysis data from the ERA5\u0000dataset, which covers the period from 1940 to 2022. Three models, including\u0000standard GANs, WGAN-GP, and U-net diffusion models, were employed to generate\u0000high-quality wind maps of the UK. These models are then evaluated using\u0000multiple metrics, including SSIM, KL divergence, and EMD, with some assessments\u0000performed in a reduced dimensionality space using PCA. The results reveal that\u0000while all models are effective in capturing the general spatial\u0000characteristics, each model exhibits distinct strengths and weaknesses. The\u0000standard GAN introduced more noise compared to the other models. The WGAN-GP\u0000model demonstrated superior performance, particularly in replicating\u0000statistical distributions. The U-net diffusion model produced the most visually\u0000coherent outputs but struggled slightly in replicating peak intensities and\u0000their statistical variability. This research underscores the potential of\u0000generative models in supplementing limited reanalysis datasets with synthetic\u0000data, providing valuable tools for risk assessment and catastrophe modelling.\u0000However, it is important to select appropriate evaluation metrics that assess\u0000different aspects of the generated outputs. Future work could refine these\u0000models and incorporate more ...","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"4 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142267519","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Assaf Shmuel, Teddy Lazebnik, Oren Glickman, Eyal Heifetz, Colin Price
Wildfires pose a significant natural disaster risk to populations and contribute to accelerated climate change. As wildfires are also affected by climate change, extreme wildfires are becoming increasingly frequent. Although they occur less frequently globally than those sparked by human activities, lightning-ignited wildfires play a substantial role in carbon emissions and account for the majority of burned areas in certain regions. While existing computational models, especially those based on machine learning, aim to predict lightning-ignited wildfires, they are typically tailored to specific regions with unique characteristics, limiting their global applicability. In this study, we present machine learning models designed to characterize and predict lightning-ignited wildfires on a global scale. Our approach involves classifying lightning-ignited versus anthropogenic wildfires, and estimating with high accuracy the probability of lightning to ignite a fire based on a wide spectrum of factors such as meteorological conditions and vegetation. Utilizing these models, we analyze seasonal and spatial trends in lightning-ignited wildfires shedding light on the impact of climate change on this phenomenon. We analyze the influence of various features on the models using eXplainable Artificial Intelligence (XAI) frameworks. Our findings highlight significant global differences between anthropogenic and lightning-ignited wildfires. Moreover, we demonstrate that, even over a short time span of less than a decade, climate changes have steadily increased the global risk of lightning-ignited wildfires. This distinction underscores the imperative need for dedicated predictive models and fire weather indices tailored specifically to each type of wildfire.
{"title":"Global Lightning-Ignited Wildfires Prediction and Climate Change Projections based on Explainable Machine Learning Models","authors":"Assaf Shmuel, Teddy Lazebnik, Oren Glickman, Eyal Heifetz, Colin Price","doi":"arxiv-2409.10046","DOIUrl":"https://doi.org/arxiv-2409.10046","url":null,"abstract":"Wildfires pose a significant natural disaster risk to populations and\u0000contribute to accelerated climate change. As wildfires are also affected by\u0000climate change, extreme wildfires are becoming increasingly frequent. Although\u0000they occur less frequently globally than those sparked by human activities,\u0000lightning-ignited wildfires play a substantial role in carbon emissions and\u0000account for the majority of burned areas in certain regions. While existing\u0000computational models, especially those based on machine learning, aim to\u0000predict lightning-ignited wildfires, they are typically tailored to specific\u0000regions with unique characteristics, limiting their global applicability. In\u0000this study, we present machine learning models designed to characterize and\u0000predict lightning-ignited wildfires on a global scale. Our approach involves\u0000classifying lightning-ignited versus anthropogenic wildfires, and estimating\u0000with high accuracy the probability of lightning to ignite a fire based on a\u0000wide spectrum of factors such as meteorological conditions and vegetation.\u0000Utilizing these models, we analyze seasonal and spatial trends in\u0000lightning-ignited wildfires shedding light on the impact of climate change on\u0000this phenomenon. We analyze the influence of various features on the models\u0000using eXplainable Artificial Intelligence (XAI) frameworks. Our findings\u0000highlight significant global differences between anthropogenic and\u0000lightning-ignited wildfires. Moreover, we demonstrate that, even over a short\u0000time span of less than a decade, climate changes have steadily increased the\u0000global risk of lightning-ignited wildfires. This distinction underscores the\u0000imperative need for dedicated predictive models and fire weather indices\u0000tailored specifically to each type of wildfire.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"52 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142267559","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper presents a new precipitation dataset that is daily, has a spatial resolution of one degree on a quasi-global scale, and spans more than 42 years, using machine learning techniques. The ultimate goal of this dataset is to provide a homogeneous daily precipitation dataset for several decades without gaps, which is suitable for climate analysis. As a first step, 42 years of daily precipitation data was generated using machine learning techniques. The machine learning methods are supervised learning, and the reference data are estimated precipitation datasets from 2001 to 2020. The three machine learning methods are random forest, gradient-boosted decision trees, and convolutional neural networks. The input data are satellite observations and atmospheric circulations from reanalysis, which are somewhat modified based on knowledge of the climatological background. Using the trained statistical models, we predict back to 1979, when daily precipitation data was almost unavailable globally. The detailed procedures are described in this paper. The produced data have been partially evaluated. However, additional evaluations from different perspectives are needed. The advantages and disadvantages of this precipitation dataset are also discussed. Currently, this GPC/m precipitation dataset version is GPC/m-v1-2024.
{"title":"GPC/m: Global Precipitation Climatology by Machine Learning; Quasi-global, Daily, and One Degree Spatial Resolution","authors":"Hiroshi G. Takahashi","doi":"arxiv-2409.09639","DOIUrl":"https://doi.org/arxiv-2409.09639","url":null,"abstract":"This paper presents a new precipitation dataset that is daily, has a spatial\u0000resolution of one degree on a quasi-global scale, and spans more than 42 years,\u0000using machine learning techniques. The ultimate goal of this dataset is to\u0000provide a homogeneous daily precipitation dataset for several decades without\u0000gaps, which is suitable for climate analysis. As a first step, 42 years of\u0000daily precipitation data was generated using machine learning techniques. The\u0000machine learning methods are supervised learning, and the reference data are\u0000estimated precipitation datasets from 2001 to 2020. The three machine learning\u0000methods are random forest, gradient-boosted decision trees, and convolutional\u0000neural networks. The input data are satellite observations and atmospheric\u0000circulations from reanalysis, which are somewhat modified based on knowledge of\u0000the climatological background. Using the trained statistical models, we predict\u0000back to 1979, when daily precipitation data was almost unavailable globally.\u0000The detailed procedures are described in this paper. The produced data have\u0000been partially evaluated. However, additional evaluations from different\u0000perspectives are needed. The advantages and disadvantages of this precipitation\u0000dataset are also discussed. Currently, this GPC/m precipitation dataset version\u0000is GPC/m-v1-2024.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"21 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142267522","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In recent years, AI-based weather forecasting models have matched or even outperformed numerical weather prediction systems. However, most of these models have been trained and evaluated on reanalysis datasets like ERA5. These datasets, being products of numerical models, often diverge substantially from actual observations in some crucial variables like near-surface temperature, wind, precipitation and clouds - parameters that hold significant public interest. To address this divergence, we introduce WeatherReal, a novel benchmark dataset for weather forecasting, derived from global near-surface in-situ observations. WeatherReal also features a publicly accessible quality control and evaluation framework. This paper details the sources and processing methodologies underlying the dataset, and further illustrates the advantage of in-situ observations in capturing hyper-local and extreme weather through comparative analyses and case studies. Using WeatherReal, we evaluated several data-driven models and compared them with leading numerical models. Our work aims to advance the AI-based weather forecasting research towards a more application-focused and operation-ready approach.
{"title":"WeatherReal: A Benchmark Based on In-Situ Observations for Evaluating Weather Models","authors":"Weixin Jin, Jonathan Weyn, Pengcheng Zhao, Siqi Xiang, Jiang Bian, Zuliang Fang, Haiyu Dong, Hongyu Sun, Kit Thambiratnam, Qi Zhang","doi":"arxiv-2409.09371","DOIUrl":"https://doi.org/arxiv-2409.09371","url":null,"abstract":"In recent years, AI-based weather forecasting models have matched or even\u0000outperformed numerical weather prediction systems. However, most of these\u0000models have been trained and evaluated on reanalysis datasets like ERA5. These\u0000datasets, being products of numerical models, often diverge substantially from\u0000actual observations in some crucial variables like near-surface temperature,\u0000wind, precipitation and clouds - parameters that hold significant public\u0000interest. To address this divergence, we introduce WeatherReal, a novel\u0000benchmark dataset for weather forecasting, derived from global near-surface\u0000in-situ observations. WeatherReal also features a publicly accessible quality\u0000control and evaluation framework. This paper details the sources and processing\u0000methodologies underlying the dataset, and further illustrates the advantage of\u0000in-situ observations in capturing hyper-local and extreme weather through\u0000comparative analyses and case studies. Using WeatherReal, we evaluated several\u0000data-driven models and compared them with leading numerical models. Our work\u0000aims to advance the AI-based weather forecasting research towards a more\u0000application-focused and operation-ready approach.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"77 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142267523","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This study explores the application of diffusion models in the field of typhoons, predicting multiple ERA5 meteorological variables simultaneously from Digital Typhoon satellite images. The focus of this study is taken to be Taiwan, an area very vulnerable to typhoons. By comparing the performance of Conditional Denoising Diffusion Probability Model (CDDPM) with Convolutional Neural Networks (CNN) and Squeeze-and-Excitation Networks (SENet), results suggest that the CDDPM performs best in generating accurate and realistic meteorological data. Specifically, CDDPM achieved a PSNR of 32.807, which is approximately 7.9% higher than CNN and 5.5% higher than SENet. Furthermore, CDDPM recorded an RMSE of 0.032, showing a 11.1% improvement over CNN and 8.6% improvement over SENet. A key application of this research can be for imputation purposes in missing meteorological datasets and generate additional high-quality meteorological data using satellite images. It is hoped that the results of this analysis will enable more robust and detailed forecasting, reducing the impact of severe weather events on vulnerable regions. Code accessible at https://github.com/TammyLing/Typhoon-forecasting.
{"title":"Estimating atmospheric variables from Digital Typhoon Satellite Images via Conditional Denoising Diffusion Models","authors":"Zhangyue Ling, Pritthijit Nath, César Quilodrán-Casas","doi":"arxiv-2409.07961","DOIUrl":"https://doi.org/arxiv-2409.07961","url":null,"abstract":"This study explores the application of diffusion models in the field of\u0000typhoons, predicting multiple ERA5 meteorological variables simultaneously from\u0000Digital Typhoon satellite images. The focus of this study is taken to be\u0000Taiwan, an area very vulnerable to typhoons. By comparing the performance of\u0000Conditional Denoising Diffusion Probability Model (CDDPM) with Convolutional\u0000Neural Networks (CNN) and Squeeze-and-Excitation Networks (SENet), results\u0000suggest that the CDDPM performs best in generating accurate and realistic\u0000meteorological data. Specifically, CDDPM achieved a PSNR of 32.807, which is\u0000approximately 7.9% higher than CNN and 5.5% higher than SENet. Furthermore,\u0000CDDPM recorded an RMSE of 0.032, showing a 11.1% improvement over CNN and 8.6%\u0000improvement over SENet. A key application of this research can be for\u0000imputation purposes in missing meteorological datasets and generate additional\u0000high-quality meteorological data using satellite images. It is hoped that the\u0000results of this analysis will enable more robust and detailed forecasting,\u0000reducing the impact of severe weather events on vulnerable regions. Code\u0000accessible at https://github.com/TammyLing/Typhoon-forecasting.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"59 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142215328","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：