Prediction Of Extreme Events Research Articles

Dynamic pathway linking Pakistan flooding to East Asian heatwaves.

In July to August 2022, Pakistan suffered historic flooding while record-breaking heatwaves swept southern China, causing severe socioeconomic impacts. Similar extreme events have frequently coincided between two regions during the past 44 years, but the underlying mechanisms remain unclear. Using observations and a suite of model experiments, here, we show that the upper-tropospheric divergent wind induced by convective heating over Pakistan excites a barotropic anomalous anticyclone over eastern China, which further leads to persistent heatwaves. Atmospheric model ensemble simulation indicates that this dynamic pathway linking Pakistan flooding and East Asian heatwaves is intrinsic to the climate system, largely independent of global sea surface temperature forcing. This dynamic connection is most active during July to August when convective variability is large over Pakistan and the associated divergent flow excites barotropic Rossby waves that propagate eastward along the upper troposphere westerly waveguide. This robust waveguide and the time delay offer hopes for improved subseasonal prediction of extreme events in East Asia.

Evaluation of COSMO-CLM Model Parameter Sensitivity in the Study of Extreme Events across the Eastern Region of India

The present study aims to identify the parameters from the Consortium for Small-scale Modelling in CLimate Mode (COSMO-CLM) regional climate model that strongly controls the prediction of extreme events over West Bengal and the adjoining areas observed between 2013 to 2018. Metrics, namely Performance Score (PS) screen out the most persuasive parameter on model output. Additionally, the Performance Index (PI) measure the reliability of the model and Skill Score (SS) establishes the model performance against the reference simulation leading to the optimization of the model for a given variable. In this study, parameter screening for four output variables such as 2m-temperature, surface latent heat flux, precipitation and cloud cover of COSMO-CLM is accomplished. For heat wave simulations, 2m-temperature and surface latent heat flux are explored whereas cloud cover and precipitation are examined for extreme rainfall events. A total of 25 adjustable parameters representing the following parameterization schemes: turbulence, land surface process, microphysics, convection, radiation and soil. Out of the six parameterization schemes, the scaling factor of the laminar boundary layer for heat (rlam_heat) and the ratio of laminar scaling factors for heat over sea and land (rat_sea) from the land surface process is sensitive to SLH, TP. The exponent to get the effective surface area (e_surf) from the land surface has a large impact on 2m-temperature. A few parameters from microphysics (cloud ice threshold for auto conversion), convection (mean entrainment rate for shallow convection) and radiation (parameter for computing the amount of cloud cover in saturated conditions) play a significant role in producing TP, and TCC fields. It is evident from the results that the parameter sensitivities on model performance depend on the choice of the meteorological field. Furthermore, in almost all input model parameters, the model performance reveals the opposite character in different domains for a given meteorological field.

A Review of Solar and Wind Energy Resource Projection Based on the Earth System Model

Many countries around the world are rapidly advancing sustainable development (SD) through the exploitation of clean energy sources such as solar and wind energy, which are becoming the core of the sustainable energy transition. In recent years, the continuous advancement of Earth system models (ESMs) has facilitated numerous studies utilizing them to predict long-term and large-scale meteorological elements, consequently enabling forecasts of wind and solar energy. These forecasts provide critical guidance for formulating national renewable energy policies. Nevertheless, the current literature on ESMs predicting wind and solar energy lacks sufficient integration. Hence, to comprehend the focal points and future research prospects, we conducted this systematic review, employing four academic search tools to comprehensively analyze the relevant literature from the past five years. We summarized the general analytical process and compared the content and conclusions of the literature. The study reveals that future photovoltaic (PV) potential for electricity generation may increase in certain regions but decrease in others, while the global potential for concentrated solar power (CSP) may diminish, influenced by diverse factors and displaying significant regional disparities. In addition, wind resource trends vary in different regions, and forecasts exhibit considerable uncertainty. Therefore, many studies have corrected wind speeds prior to predicting wind energy. Subsequent research endeavors should concentrate on optimizing ESMs, investigating the impacts of technological innovation, and enhancing the prediction and analysis of extreme weather events.

The GFDL Variable‐Resolution Global Chemistry‐Climate Model for Research at the Nexus of US Climate and Air Quality Extremes

AbstractWe present a variable‐resolution global chemistry‐climate model (AM4VR) developed at NOAA's Geophysical Fluid Dynamics Laboratory (GFDL) for research at the nexus of US climate and air quality extremes. AM4VR has a horizontal resolution of 13 km over the US, allowing it to resolve urban‐to‐rural chemical regimes, mesoscale convective systems, and land‐surface heterogeneity. With the resolution gradually reducing to 100 km over the Indian Ocean, we achieve multi‐decadal simulations driven by observed sea surface temperatures at 50% of the computational cost for a 25‐km uniform‐resolution grid. In contrast with GFDL's AM4.1 contributing to the sixth Coupled Model Intercomparison Project at 100 km resolution, AM4VR features much improved US climate mean patterns and variability. In particular, AM4VR shows improved representation of: precipitation seasonal‐to‐diurnal cycles and extremes, notably reducing the central US dry‐and‐warm bias; western US snowpack and summer drought, with implications for wildfires; and the North American monsoon, affecting dust storms. AM4VR exhibits excellent representation of winter precipitation, summer drought, and air pollution meteorology in California with complex terrain, enabling skillful prediction of both extreme summer ozone pollution and winter haze events in the Central Valley. AM4VR also provides vast improvements in the process‐level representations of biogenic volatile organic compound emissions, interactive dust emissions from land, and removal of air pollutants by terrestrial ecosystems. We highlight the value of increased model resolution in representing climate–air quality interactions through land‐biosphere feedbacks. AM4VR offers a novel opportunity to study global dimensions to US air quality, especially the role of Earth system feedbacks in a changing climate.

A Study of the Application of AI & ML to Climate Variation, with Particular Attention to Legal & Ethical Concerns

INTRODUCTION: This research investigates the utilization of artificial intelligence and machine learning in comprehending various climatic variations, emphasizing the associated use of legal and ethical considerations. This escalating impact of climatic change necessitates innovative approaches and the potential of AI/ML to offer tools for analysis and prediction. OBJECTIVES: The primary objective here, was to assess the effectiveness of AI/ML in the deciphering of varying climatic patterns and projecting the future trends. Concurrently, this study aims for the identification and analysis of legal and ethical challenges that may arise from the integration of these technologies in climatic research and policy. METHODS: Here, the literature review forms the basis for understanding various AI/ML applications related to climate science. This study employs various case analyses to examine the existing models to gauge the accuracy and efficiency of predictions. Legal frameworks and ethical principles need to be scrutinized through the qualitative analysis of relevant policies and guidelines. RESULTS: This extensive research reveals the various significant contributions of AI/ML in the enhancement of climatic modeling precision and the prediction of extreme events. However legal and ethical considerations such as data privacy, accountability, and transparency also emerged as crucial challenges which required careful attention. CONCLUSION: While AI/ML exhibited great potential in the advancement of climate research, a balanced approach is imperative to navigate the associated legal and ethical concerns. Striking this equilibrium will be pivotal for ensuring responsible and effective deployment of these technologies in the pursuit of best understanding and mitigating varying climatic variations.

Reservoir computing-based advance warning of extreme events

Physics-based computing exploits nonlinear or disorder-induced complexity, for example, to realize energy-efficient and high-throughput computing tasks. A particularly difficult but useful task is the prediction of extreme events that can occur in a wide range of complex systems. We prepare an experiment based on a microcavity semiconductor laser that produces statistically rare extreme events resulting from the interplay of deterministic nonlinear dynamics and spontaneous emission noise. We then evaluate the performance of three reservoir computing training approaches in predicting the occurrence of extreme events. We show that Dual Training Reservoir Computing (which in turn can be implemented with fast semiconductor laser dynamics) can provide meaningful early warnings up to 15 times the typical linear correlation time of the dynamics.

Modulation of Western South Atlantic Marine Heatwaves by Meridional Ocean Heat Transport

AbstractMarine heatwaves and cold spells are extreme surface temperature events that have been associated with adverse societal and ecosystem impacts in several regions around the globe. Predicting these events presents a challenge because of their generally short‐lived nature and dependence on air‐sea interactions, both locally and remotely. Here we analyze oceanic propagating features that promote the occurrence of marine heatwaves and cold spells in the western subtropical South Atlantic. The main interannual feature detected from satellite sea level data since 1993 shows a westward propagating zonal pattern with a periodicity of 3–5 years. The pattern has a significant in‐phase correlation with sea surface temperature (SST) anomalies in the western South Atlantic, explaining 82% of the daily extreme warm (90th percentile) and cold (10th percentile) SST anomalies and consequently modulating interannual variations in the intensity and duration of marine heatwave and cold spell events. It is found that meridional oceanic advection plays an important role in the regional heat budget associated with the westward‐propagating mode, modulating the meridional exchange of tropical (warm) and extratropical (cold) waters in the western subtropical South Atlantic region and thereby setting a baseline for temperature extremes on interannual timescales. This propagating mode is well correlated (r > 0.6) with the strength of the meridional overturning circulation at 25°S and 30°S with a lag of approximately 3–9 months. The lagged response provides a potential source of predictability of extreme events in the western South Atlantic.

Improved prediction of extreme ENSO events using an artificial neural network with weighted loss functions

The El Niño–Southern Oscillation (ENSO) causes a wide array of abnormal climates and extreme events, including severe droughts and floods, which have a major impact on humanity. With the development of artificial neural network techniques, various attempts are being made to predict ENSO more accurately. However, there are still limitations in accurately predicting ENSO beyond 6 months, especially for abnormal years with less frequent but greater impact, such as strong El Niño or La Niña, mainly due to insufficient and imbalanced training data. Here, we propose a new weighted loss function to improve ENSO prediction for abnormal years, in which the original (vanilla) loss function is multiplied by the weight function that relatively reduces the weight of high-frequency normal events. The new method applied to recurrent neural networks shows significant improvement in ENSO predictions for all lead times from 1 month to 12 months compared to using the vanilla loss function; in particular, the longer the prediction lead time, the greater the prediction improvement. This method can be applied to a variety of other extreme weather and climate events of low frequency but high impact.

Intensity duration and frequency of Heat wave in different phases of MJO over India

The Madden-Julian Oscillation (MJO) is an important tropical atmospheric phenomenon that plays a significant role in modulating weather patterns across India. Heat Waves (HW) are extreme weather events that have severe impacts on human health, agriculture, and various socio-economic sectors. This research paper aims to investigate the dynamical relationship between MJO phases, as measured by the Real-time Multivariate MJO (RMM) index, and the occurrence of HW in India, as defined by the Indian Meteorological Department. This paper demonstrated the relationship among three aspects, i.e., intensity, duration, frequency of HW during the different MJO phases. The differential dynamics associated for about two-fold rise in the HW occurrence in the South East, Eastern and North West India in the MJO active phase. Inactive phase witnessed enhancement (suppression) of HW during wet (dry) over the regions. The novel finding of this paper is that, the weaker the MJO, the higher probability of HW occurrence in respective regions depending on the position of the MJO convection center. The maximum duration is observed to be significantly decreased by about 2 days in the active MJO phase over NW and north central parts of the country. It is also inferred from this study that there is a significantly decrease of 0.5 °C in HW intensity during the active wet phase of MJO over the NW India. To explain the linkage of HW with MJO, geopotential height (GPH), wind circulation, velocity potential (VP), sea surface temperature (SST) and outgoing longwave radiation (OLR) composite anomalies have been computed and used in the analysis. It is observed that all these parameters have a direct impact on the evolution, strength and spread of the HW in India. This dynamical study mechanism can be prescribed in the numerical weather prediction models for the improved prediction of such extreme weather events during summer for proactive disaster management.

Beyond Accuracy: Building Trustworthy Extreme Events Predictions Through Explainable Machine Learning

Extreme events, despite their rarity, pose a significant threat due to their immense impact. While machine learning has emerged as a game-changer for predicting these events, the crucial challenge lies in trusting these predictions. Existing studies primarily focus on improving accuracy, neglecting the crucial aspect of model explainability. This gap hinders the integration of these solutions into decision-making processes. Addressing this critical issue, this paper investigates the explainability of extreme event forecasting using a hybrid forecasting and classification approach. By focusing on two economic indicators, Business Confidence Index (BCI) and Consumer Confidence Index (CCI), the study aims to understand why and when extreme event predictions can be trusted, especially in the context of imbalanced classes (normal vs. extreme events). Machine learning models are comparatively analysed, exploring their explainability through dedicated tools. Additionally, various class balancing methods are assessed for their effectiveness. This combined approach delves into the factors influencing extreme event prediction accuracy, offering valuable insights for building trustworthy forecasting models.

Prediction of extreme rainfall events in 21st century - The results based on Bayesian Markov Chain Monte Carlo

The escalation of climate change caused more extreme rainfall, increasing flood damage in Japan since around 2010. Therefore, it is crucial to examine climate change's impact on extreme rainfall events over the past 100 years and assess whether the existing return level standards effectively address the requirements for mitigating urban flood risks in the present scenario. The annual maximum hourly rainfall data from six major cities in Japan are divided into two distinct periods: 1921–1970 and 1971–2020. These datasets were then subjected to fitting with the Generalized Extreme Value distribution using two methods: Maximum Likelihood Estimation (MLE) and Bayesian Markov Chain Monte Carlo (MCMC). The results demonstrate that the MCMC method provides more precise and accurate estimates, resulting in a narrower range of uncertainty for the high period of interest. Furthermore, the return levels for the 100-year return period in Nagoya and Fukuoka, based on the data for 1971–2020, are 24.4 mm/h and 13.4 mm/h greater than the corresponding standard values, respectively. This suggests significant changes in extreme rainfall patterns in certain cities in Japan and highlights the need for Nagoya and Fukuoka to update their criteria for return levels based on the latest rainfall data using the MCMC method.

A multiscale approach to statistical downscaling of daily precipitation: Israel as a test case

AbstractRainfall in the Eastern Mediterranean is strongly modulated by complex topography and localized mesoscale processes. General circulation models (GCMs) struggle to capture daily precipitation variability in the region, both in time and in space. Rain in the Eastern Mediterranean occurs within a hierarchy of scales, as synoptic scale structures often drive local rainfall patterns. Daily rain prediction in the region can therefore benefit from analog downscaling—a nonlinear regression of a high‐resolution predictand from past synoptic‐scale predictors. We present a multiscaled downscaling algorithm of daily rain over Israel. The underlying goal is to create a mechanism‐based tool that will improve the analysis and prediction of precipitation on short time scales in models that cannot produce the field explicitly. We train the algorithm using coarse grid ERA5 reanalysis data and measurements from 21 rain gauges. The routine uses a k‐nearest neighbours algorithm to find the most similar past instances (i.e., analogs) for every predicted day. Analog selection is performed in two steps, based on scale (synoptic and local), as to not overshadow correlative but local predictors. The algorithm also includes several unique aspects tailored to Mediterranean climate: subdaily predictors of cyclone life cycles; representation of upper level cyclonic drivers; and the inclusion of rainfall potential using the Modified K‐Index (MKI). The proposed algorithm has better accuracy (66% correct predictions) compared to non‐downscaled reanalysis and climatological predictions. It better captures the spatial rainfall variance, mitigates the “drizzle bias,” and improves skill in extreme event prediction. However, it underestimates very rainy events and has trouble fully representing the spatial variance in the region. Nonetheless, our algorithm represents the potential for computationally inexpensive downscaling of daily precipitation in the Mediterranean with various possible applications, for example, characterization of droughts and storms, linking hydrological and synoptic scale processes and introducing uncertainty estimates using large ensembles.

Prediction of extreme rain in Kototabang using Himawari-8 satellite based on differences in cloud brightness temperature

The accurate prediction of extreme rain events is essential for the management and mitigation of hydrometeorological disasters. The Himawari-8 satellite provides cloud observations that can accurately forecast short-term extreme rain events using the Brightness Temperature (BT) data and Brightness Temperature Difference (BTD) method. The statistical evaluation of this method was conducted using optical rain gauge data taken from Kototabang, West Sumatra, Indonesia (100.32° E, 0.20° S) to determine the best threshold for detecting extreme rain. We used three bands, bands 11 (B11), 13 (B13) and 15 (B15) and tested three combinations of BTDs from these bands, namely BTD1 (B11-B13), BTD2 (B13-B15), and BTD3 (BTD1-BTD2). This study emphasizes the importance of parameter selection in extreme rain identification and forecasting using cloud BT and BTD methods. Effective parameter optimization is essential for adapting these approaches to different rainfall intensities, therefore selecting appropriate thresholds is necessary. The research particularly highlights the impact of temperature and rainfall intensity on BT accuracy, with BT excelling at 250 K for light rain but showing reduced accuracy as rainfall intensity increases. BTD1 demonstrates improved accuracy with higher rainfall intensity, especially at a 3 K threshold, allowing for predictions of extreme rain events with a 10–20 min lead time. However, the limitation of this threshold is shown by consistent critical success index (CSI). Therefore, BTD1 with threshold 0 K give better performances with good accuracy, CSI and false alarm ratio (FAR) for various rain intensities. BTD2 shows improved accuracy at lower thresholds with reduced rainfall intensity and at the 3 K threshold, offering potential for extreme rainfall anticipation, but with declining CSI as rainfall intensity rises. The 0 K threshold, despite high probability of detection (POD), yields increased FAR in moderate to severe rainfall scenarios. BTD3 generally exhibits increased accuracy and CSI with rising rainfall intensity, except at the −3 K threshold, with the 0 K threshold standing out as the optimal choice, providing a 10–20 min lead time for intense rain predictions. This study shows that the selection of the appropriate BT and BTD techniques, and parameter values should align with specific rainfall levels and forecasting goals.

Improving the prediction of extreme wind speed events with generative data augmentation techniques

Extreme Wind Speed events (EWS) are responsible for the worst damages caused by wind in wind farms. An accurate estimation of the frequency and intensity of EWS is essential to avoid wind turbine damage and to minimize cut-out events in these facilities. In this paper we discuss how generative Data Augmentation (DA) techniques improve the performance of Machine Learning (ML) and Deep Learning (DL) algorithms in EWS prediction problems. These problems are usually tackled as classification tasks, which are highly unbalanced due to the small number of EWS events in wind farms. Different versions of Variational AutoEncoders (VAE) are proposed and analysed in this work (VAEs, Conditional VAEs (CVAEs) and Class-Informed VAEs (CI-VAE)) as generative DA techniques to balance the data in EWS problems, leading to better performance of the prediction systems. The proposed generative DA techniques have been compared against traditional DA algorithms in a real problem of EWS prediction in Spain, considering ERA5 reanalysis data as predictive variables. The results showed that the CI-VAE with a Convolutional Neural Network approach obtained the best results, with values of Precision 0.62, Recall 0.74 and F1 score 0.67, improving up to 4% the results of the method without data augmentation techniques.

Real-time GNSS tropospheric parameter prediction of extreme rainfall events in China based on WRF multi-source data assimilation

In recent years, extreme rainfall events have frequently occurred frequently, and heavy rainfall can cause drastic changes in the troposphere. Therefore, achieving to achieve real-time high-precision numerical prediction of key tropospheric parameters during heavy rainfall has become a major problem in global navigation satellite system (GNSS) meteorology. In this paper, two extreme rainfall events in southern China (Guangdong region) and northern China (Shandong region) in 2022 are used as case studies. Twenty-four-hour real-time numerical forecasts of key tropospheric parameters (atmospheric weighted mean temperature (Tm), precipitable water vapor (PWV), and GNSS zenith tropospheric delay (ZTD)) are obtained using three models, namely, the HGPT2, GPT3, and WRF models. Two optimization models, i.e., WRFDA (am) and WRFDA (pre), are then constructed by assimilating two types of data (global upper air and surface weather observations and daily advanced microwave sounding unit A (AMSU-A) brightness temperature) based on the WRF model. The experimental results for heavy rainfall show that (1) the WRF model predicts the key tropospheric parameters with better accuracy than the HGPT2 and GPT3 models, and the WRFDA (pre) model predicts PWV and ZTD with the highest accuracy; (2) the WRFDA (pre) model achieves a higher accuracy than the WRF model in predicting PWV and ZTD, where the PWV prediction accuracy is improved relative to the WRF model (in the south: MAE: 32.7 %; RMSE: 33.9 %; MAPE: 36.8 %; in the north: MAE: 27.3 %; RMSE: 24.2 %; MAPE: 28.0 %); this model achieves an MAE of 2.17 cm and an RMSE of 2.70 cm in 24-h ZTD prediction in the south, while the MAE reaches 2.48 cm, and the RMSE is 3.18 cm in the north; (3) the models provide a higher forecast accuracy in the southern region than in the northern region for heavy rainfall. The WRFDA (pre) model provides a favourable ZTD accuracy at GNSS stations near the ocean, while the WRFDA (am) model provides a satisfactory ZTD accuracy at inland GNSS stations, and the WRFDA (am) model provides the highest ZTD prediction accuracy at GNSS stations above 100 m.

Targeted Observations on Arctic Sea Ice Concentration for Improving Extended‐Range Prediction of Ural Blocking

AbstractThe predictability of certain extreme weather events can exceed the traditional 2 weeks by considering the boundary conditions. Targeted observations in sensitive areas (SA) on Arctic sea ice concentration (SIC) can improve the extended‐range (4 pentads) forecast skills of long‐lasting and strong Ural blocking (UB). The SA are determined based on the SIC optimally growing boundary errors, obtained by the conditional nonlinear optimal perturbation method. The SA are mainly located in the Barents Sea, Greenland Sea, and Okhotsk Sea. The results of observing system simulation experiments for 8 UB cases indicate that the targeted observations can remarkably improve the prediction skills of UB in the fourth pentad. Targeted observations have a positive effect on 75% of 160 experiment members, reduce 35% forecast errors of the fourth pentad mean blocking index, and perform even better when the original forecast errors are greater. Further diagnosis shows that targeted observations contribute to more accurate SIC boundary conditions in the Barents Sea, Greenland Sea, and Okhotsk Sea and reduce temperature errors in the lower and middle troposphere. It further results in well‐described westerly winds in the Ural region and its adjacent regions, corresponding to the more skillful predictions of blocking circulations. The above results supply a theoretical base for the design of Arctic SIC observations and more skillful extended‐range predictions for mid‐latitude extreme weather.

Regimes of Soil Moisture–Wet-Bulb Temperature Coupling with Relevance to Moist Heat Stress

Abstract Human heat stress depends jointly on atmospheric temperature and humidity. Wetter soils reduce temperature but also raise humidity, making the collective impact on heat stress unclear. To better understand these interactions, we use ERA5 to examine the coupling between daily average soil moisture and wet-bulb temperature (Tw) and its seasonal and diurnal cycle at global scale. We identify a global soil moisture–Tw coupling pattern with both widespread negative and positive correlations in contrast to the well-established cooling effect of wet soil on dry-bulb temperature. Regions showing positive correlations closely resemble previously identified land–atmosphere coupling hotspots where soil moisture effectively controls surface energy partition. Soil moisture–Tw coupling varies seasonally closely tied to monsoon development, and the positive coupling is slightly stronger and more widespread during nighttime. Local-scale analysis demonstrates a nonlinear structure of soil moisture–Tw coupling with stronger coupling under relatively dry soils. Hot days with high Tw values show wetter-than-normal soil, anomalous high latent and low sensible heat flux from a cooler surface, and a shallower boundary layer. This supports the hypothesis that wetter soil increases Tw by concentrating surface moist enthalpy flux within a shallower boundary layer and reducing free-troposphere-air entrainment. We identify areas of particular interest for future studies on the physical mechanisms of soil moisture–heat stress coupling. Our findings suggest that increasing soil moisture might amplify heat stress over large portions of the world including several densely populated areas. These results also raise questions about the effectiveness of evaporative cooling strategies in ameliorating urban heat stress. Significance Statement The purpose of this study is to provide a global picture of the relationship between soil moisture anomalies and a heat stress metric that includes the joint effects of temperature and humidity. This is important because a better understanding of this relationship will help improve the prediction of extreme heat stress events and inform strategies for ameliorating heat stress. We find a widespread positive correlation between soil moisture and heat stress, in contrast to studies relying on temperature alone. This raises the possibility that, over much of the world, and in the most populous regions, strategies like irrigation or “greening” that can reduce temperature might be ineffective or even harmful in reducing heat stress with humidity incorporated.

Attractor Inspired Deep Learning for Modelling Chaotic Systems

Predicting and understanding the behavior of dynamic systems have driven advancements in various approaches, including physics-based models and data-driven techniques like deep neural networks. Chaotic systems, with their stochastic nature and unpredictable behavior, pose challenges for accurate modeling and forecasting, especially during extreme events. In this paper, we propose a novel deep learning framework called Attractor-Inspired Deep Learning (AiDL), which seamlessly integrates actual statistics and mathematical models of system kinetics. AiDL combines the strengths of physics-informed machine learning and data-driven methods, offering a promising solution for modeling nonlinear systems. By leveraging the intricate dynamics of attractors, AiDL bridges the gap between physics-based models and deep neural networks. We demonstrate the effectiveness of AiDL using real-world data from various domains, including catastrophic weather mechanics, El Niño cycles, and disease transmission. Our empirical results showcase AiDL’s ability to substantially enhance the modeling of extreme events. The proposed AiDL paradigm holds promise for advancing research in Time Series Prediction of Extreme Events and has applications in real-world chaotic system transformations.

West-WRF 34-Year Reforecast: Description and Validation

Abstract This study presents a high-resolution regional reforecast based on the Weather Research and Forecasting (WRF) Model, tailored for the prediction of extreme hydrometeorological events over the western United States (West-WRF) spanning 34 cool seasons (1 December–31 March) from 1986 to 2019. The West-WRF reforecast has a 9-km domain covering western North America and the eastern Pacific Ocean and a 3-km domain covering much of California. The West-WRF reforecast is generated by dynamically downscaling the control member of the Global Ensemble Forecasting System (GEFS) v10 reforecast. Verification of near-surface temperature, wind, and humidity highlight the added value in the reforecast relative to GEFS. Analysis of geopotential height indicates that West-WRF reduces the bias throughout much of the troposphere during early lead times. The West-WRF reforecast also shows clear improvement in atmospheric river characteristics (intensity and landfall) over GEFS. Analysis of mean areal precipitation (MAP) shows that at the basin scale, the reforecast can improve MAP relative to GEFS and reveals a consistent low bias in the reforecast for a coastal watershed (Russian) and a high bias observed in a Northern Sierra watershed (Yuba). The reforecast has a dry bias in seasonal precipitation in the northern Central Valley and Coast Mountain ranges, and a wet bias in the Northern Sierra Nevada, consistent with other operational high-resolution (<25 km) regional models. The applications of this high-resolution multiyear reforecast include process-based studies, assessment of model performance, and machine learning applications.

Prediction of dragon king extreme events using machine learning approaches and its characterizations

In this study, we employ a machine learning approach to infer the complex dynamics of dragon king extreme events. Specifically, we utilize two distinct machine learning techniques: Echo State Network and Gated Recurrent Unit. To do so, we consider three distinct systems for predicting dragon kings behavior: a pair of electronic circuits, coupled logistic maps, and Hindmarsh-Rose neurons. We discover that a few actual time series data points, accompanied by their corresponding system parameters, are adequate to capture dragon kings nature. Initially, we demonstrate that systems under consideration possess characteristics of extreme events, with signal amplitudes greater than the critical amplitude threshold. The presence of dragon kings within these observed extreme events is discerned by the emergence of hump-like behavior in the tail distribution of the probability density function and the statistical measures. Finally, we calculate the root mean square error to determine the accuracy of the predicted dynamics.