There is a great need to accurately predict short-term precipitation, which has socioeconomic effects such as agriculture and disaster prevention. Recently, the forecasting models have employed multi-source data as the multi-modality input, thus improving the prediction accuracy. However, the prevailing methods usually suffer from the desynchronization of multi-source variables, the insufficient capability of capturing spatio-temporal dependency, and unsatisfactory performance in predicting extreme precipitation events. To fix these problems, we propose a short-term precipitation forecasting model based on spatio-temporal alignment attention, with SATA as the temporal alignment module and STAU as the spatio-temporal feature extractor to filter high-pass features from precipitation signals and capture multi-term temporal dependencies. Based on satellite and ERA5 data from the southwestern region of China, our model achieves improvements of 12.61% in terms of RMSE, in comparison with the state-of-the-art methods.
短期降水对农业和防灾等社会经济影响巨大，因此亟需准确预测短期降水。近年来，预报模式采用多源数据作为多模态输入，从而提高了预报精度。然而，现有方法通常存在多源变量不同步、捕捉时空依赖性的能力不足以及预测极端降水事件的性能不理想等问题。为了解决这些问题，我们提出了一种基于时空配准注意力的短期降水预报模型，以 SATA 作为时空配准模块，以 STAU 作为时空特征提取器，从降水信号中过滤高通特征并捕捉多期时空依赖性。基于中国西南地区的卫星和ERA5数据，我们的模型在均方根误差（RMSE）方面与最先进的方法相比提高了12.61%。
{"title":"STAA: Spatio-Temporal Alignment Attention for Short-Term Precipitation Forecasting","authors":"Min Chen, Hao Yang, Shaohan Li, Xiaolin Qin","doi":"arxiv-2409.06732","DOIUrl":"https://doi.org/arxiv-2409.06732","url":null,"abstract":"There is a great need to accurately predict short-term precipitation, which\u0000has socioeconomic effects such as agriculture and disaster prevention.\u0000Recently, the forecasting models have employed multi-source data as the\u0000multi-modality input, thus improving the prediction accuracy. However, the\u0000prevailing methods usually suffer from the desynchronization of multi-source\u0000variables, the insufficient capability of capturing spatio-temporal dependency,\u0000and unsatisfactory performance in predicting extreme precipitation events. To\u0000fix these problems, we propose a short-term precipitation forecasting model\u0000based on spatio-temporal alignment attention, with SATA as the temporal\u0000alignment module and STAU as the spatio-temporal feature extractor to filter\u0000high-pass features from precipitation signals and capture multi-term temporal\u0000dependencies. Based on satellite and ERA5 data from the southwestern region of\u0000China, our model achieves improvements of 12.61% in terms of RMSE, in\u0000comparison with the state-of-the-art methods.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"2 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142215406","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}
Because of the rudimentary reporting methods and general lack of documentation, the creation of a severe weather database within the Philippines has been difficult yet relevant target for climatology purposes and historical interest. Previous online severe weather documentation i.e. of tornadoes, waterspouts, and hail events, has also often been few, inconsistent, or is now defunct. Many individual countries or continents maintain severe weather information through either government-sponsored or independent organizations. In this case, Project SWAP is intended to be a collaborative exercise, with clear data attribution and open avenues for augmentation, and the creation of a common data model to store the severe weather event information will assist in maintaining and updating the database in the Philippines. For this work, we document the methods necessary for creating the SWAP database, provide broader climatological analysis of spatio-temporal patterns in severe weather occurrence within the Philippine context, and outline potential use cases for the data. We also highlight its key limitations, and emphasize the need for further standardization of such documentation.
{"title":"Project Severe Weather Archive of the Philippines (SWAP). Part 1: Establishing a Baseline Climatology for Severe Weather across the Philippine Archipelago","authors":"Generich H. Capuli","doi":"arxiv-2409.03211","DOIUrl":"https://doi.org/arxiv-2409.03211","url":null,"abstract":"Because of the rudimentary reporting methods and general lack of\u0000documentation, the creation of a severe weather database within the Philippines\u0000has been difficult yet relevant target for climatology purposes and historical\u0000interest. Previous online severe weather documentation i.e. of tornadoes,\u0000waterspouts, and hail events, has also often been few, inconsistent, or is now\u0000defunct. Many individual countries or continents maintain severe weather\u0000information through either government-sponsored or independent organizations.\u0000In this case, Project SWAP is intended to be a collaborative exercise, with\u0000clear data attribution and open avenues for augmentation, and the creation of a\u0000common data model to store the severe weather event information will assist in\u0000maintaining and updating the database in the Philippines. For this work, we\u0000document the methods necessary for creating the SWAP database, provide broader\u0000climatological analysis of spatio-temporal patterns in severe weather\u0000occurrence within the Philippine context, and outline potential use cases for\u0000the data. We also highlight its key limitations, and emphasize the need for\u0000further standardization of such documentation.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"42 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142215404","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}
Thomas Nils Nipen, Håvard Homleid Haugen, Magnus Sikora Ingstad, Even Marius Nordhagen, Aram Farhad Shafiq Salihi, Paulina Tedesco, Ivar Ambjørn Seierstad, Jørn Kristiansen, Simon Lang, Mihai Alexe, Jesper Dramsch, Baudouin Raoult, Gert Mertes, Matthew Chantry
A data-driven model (DDM) suitable for regional weather forecasting applications is presented. The model extends the Artificial Intelligence Forecasting System by introducing a stretched-grid architecture that dedicates higher resolution over a regional area of interest and maintains a lower resolution elsewhere on the globe. The model is based on graph neural networks, which naturally affords arbitrary multi-resolution grid configurations. The model is applied to short-range weather prediction for the Nordics, producing forecasts at 2.5 km spatial and 6 h temporal resolution. The model is pre-trained on 43 years of global ERA5 data at 31 km resolution and is further refined using 3.3 years of 2.5 km resolution operational analyses from the MetCoOp Ensemble Prediction System (MEPS). The performance of the model is evaluated using surface observations from measurement stations across Norway and is compared to short-range weather forecasts from MEPS. The DDM outperforms both the control run and the ensemble mean of MEPS for 2 m temperature. The model also produces competitive precipitation and wind speed forecasts, but is shown to underestimate extreme events.
{"title":"Regional data-driven weather modeling with a global stretched-grid","authors":"Thomas Nils Nipen, Håvard Homleid Haugen, Magnus Sikora Ingstad, Even Marius Nordhagen, Aram Farhad Shafiq Salihi, Paulina Tedesco, Ivar Ambjørn Seierstad, Jørn Kristiansen, Simon Lang, Mihai Alexe, Jesper Dramsch, Baudouin Raoult, Gert Mertes, Matthew Chantry","doi":"arxiv-2409.02891","DOIUrl":"https://doi.org/arxiv-2409.02891","url":null,"abstract":"A data-driven model (DDM) suitable for regional weather forecasting\u0000applications is presented. The model extends the Artificial Intelligence\u0000Forecasting System by introducing a stretched-grid architecture that dedicates\u0000higher resolution over a regional area of interest and maintains a lower\u0000resolution elsewhere on the globe. The model is based on graph neural networks,\u0000which naturally affords arbitrary multi-resolution grid configurations. The model is applied to short-range weather prediction for the Nordics,\u0000producing forecasts at 2.5 km spatial and 6 h temporal resolution. The model is\u0000pre-trained on 43 years of global ERA5 data at 31 km resolution and is further\u0000refined using 3.3 years of 2.5 km resolution operational analyses from the\u0000MetCoOp Ensemble Prediction System (MEPS). The performance of the model is\u0000evaluated using surface observations from measurement stations across Norway\u0000and is compared to short-range weather forecasts from MEPS. The DDM outperforms\u0000both the control run and the ensemble mean of MEPS for 2 m temperature. The\u0000model also produces competitive precipitation and wind speed forecasts, but is\u0000shown to underestimate extreme events.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"32 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142215407","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}
Shengchen Zhu, Yiming Chen, Peiying Yu, Xiang Qu, Yuxiao Zhou, Yiming Ma, Zhizhan Zhao, Yukai Liu, Hao Mi, Bin Wang
Accurate weather forecasting is essential for understanding and mitigating weather-related impacts. In this paper, we present PuYun, an autoregressive cascade model that leverages large kernel attention convolutional networks. The model's design inherently supports extended weather prediction horizons while broadening the effective receptive field. The integration of large kernel attention mechanisms within the convolutional layers enhances the model's capacity to capture fine-grained spatial details, thereby improving its predictive accuracy for meteorological phenomena. We introduce PuYun, comprising PuYun-Short for 0-5 day forecasts and PuYun-Medium for 5-10 day predictions. This approach enhances the accuracy of 10-day weather forecasting. Through evaluation, we demonstrate that PuYun-Short alone surpasses the performance of both GraphCast and FuXi-Short in generating accurate 10-day forecasts. Specifically, on the 10th day, PuYun-Short reduces the RMSE for Z500 to 720 $m^2/s^2$, compared to 732 $m^2/s^2$ for GraphCast and 740 $m^2/s^2$ for FuXi-Short. Additionally, the RMSE for T2M is reduced to 2.60 K, compared to 2.63 K for GraphCast and 2.65 K for FuXi-Short. Furthermore, when employing a cascaded approach by integrating PuYun-Short and PuYun-Medium, our method achieves superior results compared to the combined performance of FuXi-Short and FuXi-Medium. On the 10th day, the RMSE for Z500 is further reduced to 638 $m^2/s^2$, compared to 641 $m^2/s^2$ for FuXi. These findings underscore the effectiveness of our model ensemble in advancing medium-range weather prediction. Our training code and model will be open-sourced.
{"title":"PuYun: Medium-Range Global Weather Forecasting Using Large Kernel Attention Convolutional Networks","authors":"Shengchen Zhu, Yiming Chen, Peiying Yu, Xiang Qu, Yuxiao Zhou, Yiming Ma, Zhizhan Zhao, Yukai Liu, Hao Mi, Bin Wang","doi":"arxiv-2409.02123","DOIUrl":"https://doi.org/arxiv-2409.02123","url":null,"abstract":"Accurate weather forecasting is essential for understanding and mitigating\u0000weather-related impacts. In this paper, we present PuYun, an autoregressive\u0000cascade model that leverages large kernel attention convolutional networks. The\u0000model's design inherently supports extended weather prediction horizons while\u0000broadening the effective receptive field. The integration of large kernel\u0000attention mechanisms within the convolutional layers enhances the model's\u0000capacity to capture fine-grained spatial details, thereby improving its\u0000predictive accuracy for meteorological phenomena. We introduce PuYun, comprising PuYun-Short for 0-5 day forecasts and\u0000PuYun-Medium for 5-10 day predictions. This approach enhances the accuracy of\u000010-day weather forecasting. Through evaluation, we demonstrate that PuYun-Short\u0000alone surpasses the performance of both GraphCast and FuXi-Short in generating\u0000accurate 10-day forecasts. Specifically, on the 10th day, PuYun-Short reduces\u0000the RMSE for Z500 to 720 $m^2/s^2$, compared to 732 $m^2/s^2$ for GraphCast and\u0000740 $m^2/s^2$ for FuXi-Short. Additionally, the RMSE for T2M is reduced to 2.60\u0000K, compared to 2.63 K for GraphCast and 2.65 K for FuXi-Short. Furthermore,\u0000when employing a cascaded approach by integrating PuYun-Short and PuYun-Medium,\u0000our method achieves superior results compared to the combined performance of\u0000FuXi-Short and FuXi-Medium. On the 10th day, the RMSE for Z500 is further\u0000reduced to 638 $m^2/s^2$, compared to 641 $m^2/s^2$ for FuXi. These findings\u0000underscore the effectiveness of our model ensemble in advancing medium-range\u0000weather prediction. Our training code and model will be open-sourced.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"18 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142215409","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}
Climate models lack the necessary resolution for urban climate studies, requiring computationally intensive processes to estimate high resolution air temperatures. In contrast, Data-driven approaches offer faster and more accurate air temperature downscaling. This study presents a data-driven framework for downscaling air temperature using publicly available outputs from urban climate models, specifically datasets generated by UrbClim. The proposed framework utilized morphological features extracted from LiDAR data. To extract urban morphological features, first a three-dimensional building model was created using LiDAR data and deep learning models. Then, these features were integrated with meteorological parameters such as wind, humidity, etc., to downscale air temperature using machine learning algorithms. The results demonstrated that the developed framework effectively extracted urban morphological features from LiDAR data. Deep learning algorithms played a crucial role in generating three-dimensional models for extracting the aforementioned features. Also, the evaluation of air temperature downscaling results using various machine learning models indicated that the LightGBM model had the best performance with an RMSE of 0.352{deg}K and MAE of 0.215{deg}K. Furthermore, the examination of final air temperature maps derived from downscaling showed that the developed framework successfully estimated air temperatures at higher resolutions, enabling the identification of local air temperature patterns at street level. The corresponding source codes are available on GitHub: https://github.com/FatemehCh97/Air-Temperature-Downscaling.
{"title":"Machine Learning Framework for High-Resolution Air Temperature Downscaling Using LiDAR-Derived Urban Morphological Features","authors":"Fatemeh Chajaei, Hossein Bagheri","doi":"arxiv-2409.02120","DOIUrl":"https://doi.org/arxiv-2409.02120","url":null,"abstract":"Climate models lack the necessary resolution for urban climate studies,\u0000requiring computationally intensive processes to estimate high resolution air\u0000temperatures. In contrast, Data-driven approaches offer faster and more\u0000accurate air temperature downscaling. This study presents a data-driven\u0000framework for downscaling air temperature using publicly available outputs from\u0000urban climate models, specifically datasets generated by UrbClim. The proposed\u0000framework utilized morphological features extracted from LiDAR data. To extract\u0000urban morphological features, first a three-dimensional building model was\u0000created using LiDAR data and deep learning models. Then, these features were\u0000integrated with meteorological parameters such as wind, humidity, etc., to\u0000downscale air temperature using machine learning algorithms. The results\u0000demonstrated that the developed framework effectively extracted urban\u0000morphological features from LiDAR data. Deep learning algorithms played a\u0000crucial role in generating three-dimensional models for extracting the\u0000aforementioned features. Also, the evaluation of air temperature downscaling\u0000results using various machine learning models indicated that the LightGBM model\u0000had the best performance with an RMSE of 0.352{deg}K and MAE of 0.215{deg}K.\u0000Furthermore, the examination of final air temperature maps derived from\u0000downscaling showed that the developed framework successfully estimated air\u0000temperatures at higher resolutions, enabling the identification of local air\u0000temperature patterns at street level. The corresponding source codes are\u0000available on GitHub:\u0000https://github.com/FatemehCh97/Air-Temperature-Downscaling.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"53 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142215408","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}
The subpolar gyre is at risk of crossing a tipping point which would result in the collapse of convection in the Labrador Sea. It is important to understand the mechanisms at play and how they are represented in climate models. In this study we use causal inference to verify whether the proposed mechanism for bistability of the subpolar gyre is represented in CMIP6 models. In many models an increase of sea surface salinity leads to a deepening of the mixed layer resulting in a cooling of the water at intermediate depth, in line with theory. The feedback from the subsurface temperature through density to the strength of the gyre circulation is more ambiguous, with fewer models indicating a significant link. Those that do show a significant link do not agree on its sign. One model (CESM2) contains all interactions, with both a negative and delayed positive feedback loop.
{"title":"Subpolar Gyre Variability in CMIP6 Models: Is there a Mechanism for Bistability?","authors":"Swinda K. J. Falkena, Anna S. von der Heydt","doi":"arxiv-2408.16541","DOIUrl":"https://doi.org/arxiv-2408.16541","url":null,"abstract":"The subpolar gyre is at risk of crossing a tipping point which would result\u0000in the collapse of convection in the Labrador Sea. It is important to\u0000understand the mechanisms at play and how they are represented in climate\u0000models. In this study we use causal inference to verify whether the proposed\u0000mechanism for bistability of the subpolar gyre is represented in CMIP6 models.\u0000In many models an increase of sea surface salinity leads to a deepening of the\u0000mixed layer resulting in a cooling of the water at intermediate depth, in line\u0000with theory. The feedback from the subsurface temperature through density to\u0000the strength of the gyre circulation is more ambiguous, with fewer models\u0000indicating a significant link. Those that do show a significant link do not\u0000agree on its sign. One model (CESM2) contains all interactions, with both a\u0000negative and delayed positive feedback loop.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"104 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142215411","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}
Bernat Jiménez-Esteve, David Barriopedro, Juan Emmanuel Johnson, Ricardo Garcia-Herrera
Anthropogenic climate change (ACC) is altering the frequency and intensity of extreme weather events. Attributing individual extreme events (EEs) to ACC is becoming crucial to assess the risks of climate change. Traditional attribution methods often suffer from a selection bias, are computationally demanding, and provide answers after the EE occurs. This study presents a ground-breaking hybrid attribution method by combining physics-based ACC estimates from global climate models with deep-learning weather forecasts. This hybrid approach circumvents the framing choices and accelerates the attribution process, paving the way for operational anticipated global forecast-based attribution. We apply this methodology to three distinct high-impact weather EEs. Despite some limitations in predictability, the method uncovers ACC fingerprints in the forecasted fields of EEs. Specifically, forecasts successfully anticipate that ACC exacerbated the 2018 Iberian heatwave, deepened hurricane Florence, and intensified the wind and precipitable water of the explosive cyclone Ciar'an.
人为气候变化（ACC）正在改变极端天气事件的频率和强度。将个别极端事件（EEs）归因于 ACC 正成为评估气候变化风险的关键。传统的归因方法往往存在选择偏差，计算量大，而且是在 EE 发生后才提供答案。本研究提出了一种开创性的混合归因方法，将全球气候模型中基于物理学的 ACC 估值与深度学习天气预报相结合。这种混合方法避免了框架选择，加快了归因过程，为基于全球预测的业务预期归因铺平了道路。我们将这一方法应用于三种不同的高影响天气 EE。尽管在可预测性方面存在一些限制，但该方法在预测的 EEs 领域中发现了 ACC 指纹。具体来说，预测成功地预测到气候变化加剧了 2018 年伊比利亚热浪，加深了佛罗伦萨飓风，并增强了爆炸性气旋 Ciar'an 的风力和可降水量。
{"title":"AI-driven weather forecasts enable anticipated attribution of extreme events to human-made climate change","authors":"Bernat Jiménez-Esteve, David Barriopedro, Juan Emmanuel Johnson, Ricardo Garcia-Herrera","doi":"arxiv-2408.16433","DOIUrl":"https://doi.org/arxiv-2408.16433","url":null,"abstract":"Anthropogenic climate change (ACC) is altering the frequency and intensity of\u0000extreme weather events. Attributing individual extreme events (EEs) to ACC is\u0000becoming crucial to assess the risks of climate change. Traditional attribution\u0000methods often suffer from a selection bias, are computationally demanding, and\u0000provide answers after the EE occurs. This study presents a ground-breaking\u0000hybrid attribution method by combining physics-based ACC estimates from global\u0000climate models with deep-learning weather forecasts. This hybrid approach\u0000circumvents the framing choices and accelerates the attribution process, paving\u0000the way for operational anticipated global forecast-based attribution. We apply\u0000this methodology to three distinct high-impact weather EEs. Despite some\u0000limitations in predictability, the method uncovers ACC fingerprints in the\u0000forecasted fields of EEs. Specifically, forecasts successfully anticipate that\u0000ACC exacerbated the 2018 Iberian heatwave, deepened hurricane Florence, and\u0000intensified the wind and precipitable water of the explosive cyclone Ciar'an.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"9 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142215413","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}
Athul Rasheeda Satheesh, Peter Knippertz, Andreas H. Fink
Numerical weather prediction (NWP) models often underperform compared to simpler climatology-based precipitation forecasts in northern tropical Africa, even after statistical postprocessing. AI-based forecasting models show promise but have avoided precipitation due to its complexity. Synoptic-scale forcings like African easterly waves and other tropical waves (TWs) are important for predictability in tropical Africa, yet their value for predicting daily rainfall remains unexplored. This study uses two machine-learning models--gamma regression and a convolutional neural network (CNN)--trained on TW predictors from satellite-based GPM IMERG data to predict daily rainfall during the July-September monsoon season. Predictor variables are derived from the local amplitude and phase information of seven TW from the target and up-and-downstream neighboring grids at 1-degree spatial resolution. The ML models are combined with Easy Uncertainty Quantification (EasyUQ) to generate calibrated probabilistic forecasts and are compared with three benchmarks: Extended Probabilistic Climatology (EPC15), ECMWF operational ensemble forecast (ENS), and a probabilistic forecast from the ENS control member using EasyUQ (CTRL EasyUQ). The study finds that downstream predictor variables offer the highest predictability, with downstream tropical depression (TD)-type wave-based predictors being most important. Other waves like mixed-Rossby gravity (MRG), Kelvin, and inertio-gravity waves also contribute significantly but show regional preferences. ENS forecasts exhibit poor skill due to miscalibration. CTRL EasyUQ shows improvement over ENS and marginal enhancement over EPC15. Both gamma regression and CNN forecasts significantly outperform benchmarks in tropical Africa. This study highlights the potential of ML models trained on TW-based predictors to improve daily precipitation forecasts in tropical Africa.
{"title":"Machine learning models for daily rainfall forecasting in Northern Tropical Africa using tropical wave predictors","authors":"Athul Rasheeda Satheesh, Peter Knippertz, Andreas H. Fink","doi":"arxiv-2408.16349","DOIUrl":"https://doi.org/arxiv-2408.16349","url":null,"abstract":"Numerical weather prediction (NWP) models often underperform compared to\u0000simpler climatology-based precipitation forecasts in northern tropical Africa,\u0000even after statistical postprocessing. AI-based forecasting models show promise\u0000but have avoided precipitation due to its complexity. Synoptic-scale forcings\u0000like African easterly waves and other tropical waves (TWs) are important for\u0000predictability in tropical Africa, yet their value for predicting daily\u0000rainfall remains unexplored. This study uses two machine-learning models--gamma\u0000regression and a convolutional neural network (CNN)--trained on TW predictors\u0000from satellite-based GPM IMERG data to predict daily rainfall during the\u0000July-September monsoon season. Predictor variables are derived from the local\u0000amplitude and phase information of seven TW from the target and\u0000up-and-downstream neighboring grids at 1-degree spatial resolution. The ML\u0000models are combined with Easy Uncertainty Quantification (EasyUQ) to generate\u0000calibrated probabilistic forecasts and are compared with three benchmarks:\u0000Extended Probabilistic Climatology (EPC15), ECMWF operational ensemble forecast\u0000(ENS), and a probabilistic forecast from the ENS control member using EasyUQ\u0000(CTRL EasyUQ). The study finds that downstream predictor variables offer the\u0000highest predictability, with downstream tropical depression (TD)-type\u0000wave-based predictors being most important. Other waves like mixed-Rossby\u0000gravity (MRG), Kelvin, and inertio-gravity waves also contribute significantly\u0000but show regional preferences. ENS forecasts exhibit poor skill due to\u0000miscalibration. CTRL EasyUQ shows improvement over ENS and marginal enhancement\u0000over EPC15. Both gamma regression and CNN forecasts significantly outperform\u0000benchmarks in tropical Africa. This study highlights the potential of ML models\u0000trained on TW-based predictors to improve daily precipitation forecasts in\u0000tropical Africa.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"159 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142215412","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}
J. Hernandez-Bernal, A. Spiga, F. Forget, D. Banfield
Thermal tides are atmospheric planetary-scale waves with periods that are harmonics of the solar day. In the Martian atmosphere thermal tides are known to be especially significant compared to any other known planet. Based on the data set of pressure timeseries produced by the InSight lander, which is unprecedented in terms of accuracy and temporal coverage, we investigate thermal tides on Mars and we find harmonics even beyond the number 24, which exceeds significantly the number of harmonics previously reported by other works. We explore comparatively the characteristics and seasonal evolution of tidal harmonics and find that even and odd harmonics exhibit some clearly differentiated trends that evolve seasonally and respond to dust events. High-order tidal harmonics with small amplitudes could transiently interfere constructively to produce meteorologically relevant patterns.
{"title":"High-Order harmonics of Thermal Tides observed in the atmosphere of Mars by the Pressure Sensor on the Insight lander","authors":"J. Hernandez-Bernal, A. Spiga, F. Forget, D. Banfield","doi":"arxiv-2408.15745","DOIUrl":"https://doi.org/arxiv-2408.15745","url":null,"abstract":"Thermal tides are atmospheric planetary-scale waves with periods that are\u0000harmonics of the solar day. In the Martian atmosphere thermal tides are known\u0000to be especially significant compared to any other known planet. Based on the\u0000data set of pressure timeseries produced by the InSight lander, which is\u0000unprecedented in terms of accuracy and temporal coverage, we investigate\u0000thermal tides on Mars and we find harmonics even beyond the number 24, which\u0000exceeds significantly the number of harmonics previously reported by other\u0000works. We explore comparatively the characteristics and seasonal evolution of\u0000tidal harmonics and find that even and odd harmonics exhibit some clearly\u0000differentiated trends that evolve seasonally and respond to dust events.\u0000High-order tidal harmonics with small amplitudes could transiently interfere\u0000constructively to produce meteorologically relevant patterns.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142215416","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}
Sungduk Yu, Brian L. White, Anahita Bhiwandiwalla, Musashi Hinck, Matthew Lyle Olson, Tung Nguyen, Vasudev Lal
Detecting and attributing temperature increases due to climate change is crucial for understanding global warming and guiding adaptation strategies. The complexity of distinguishing human-induced climate signals from natural variability has challenged traditional detection and attribution (D&A) approaches, which seek to identify specific "fingerprints" in climate response variables. Deep learning offers potential for discerning these complex patterns in expansive spatial datasets. However, lack of standard protocols has hindered consistent comparisons across studies. We introduce ClimDetect, a standardized dataset of over 816k daily climate snapshots, designed to enhance model accuracy in identifying climate change signals. ClimDetect integrates various input and target variables used in past research, ensuring comparability and consistency. We also explore the application of vision transformers (ViT) to climate data, a novel and modernizing approach in this context. Our open-access data and code serve as a benchmark for advancing climate science through improved model evaluations. ClimDetect is publicly accessible via Huggingface dataet respository at: https://huggingface.co/datasets/ClimDetect/ClimDetect.
{"title":"ClimDetect: A Benchmark Dataset for Climate Change Detection and Attribution","authors":"Sungduk Yu, Brian L. White, Anahita Bhiwandiwalla, Musashi Hinck, Matthew Lyle Olson, Tung Nguyen, Vasudev Lal","doi":"arxiv-2408.15993","DOIUrl":"https://doi.org/arxiv-2408.15993","url":null,"abstract":"Detecting and attributing temperature increases due to climate change is\u0000crucial for understanding global warming and guiding adaptation strategies. The\u0000complexity of distinguishing human-induced climate signals from natural\u0000variability has challenged traditional detection and attribution (D&A)\u0000approaches, which seek to identify specific \"fingerprints\" in climate response\u0000variables. Deep learning offers potential for discerning these complex patterns\u0000in expansive spatial datasets. However, lack of standard protocols has hindered\u0000consistent comparisons across studies. We introduce ClimDetect, a standardized\u0000dataset of over 816k daily climate snapshots, designed to enhance model\u0000accuracy in identifying climate change signals. ClimDetect integrates various\u0000input and target variables used in past research, ensuring comparability and\u0000consistency. We also explore the application of vision transformers (ViT) to\u0000climate data, a novel and modernizing approach in this context. Our open-access\u0000data and code serve as a benchmark for advancing climate science through\u0000improved model evaluations. ClimDetect is publicly accessible via Huggingface\u0000dataet respository at: https://huggingface.co/datasets/ClimDetect/ClimDetect.","PeriodicalId":501166,"journal":{"name":"arXiv - PHYS - Atmospheric and Oceanic Physics","volume":"21 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142215418","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：