Centre forMedium-Range Weather Forecasts Research Articles

Ensemble Predictability of Week 3/4 Precipitation and Temperature over the United States via Cluster Analysis of the Large-Scale Circulation

Abstract Forecasting the week 3/4 period presents many challenges, resulting in a need for improvements to forecast skill. At this distance from initial conditions, numerical models struggle to present skillful forecasts of temperature, precipitation, and associated extremes. One approach to address this is to utilize more predictable large-scale circulation regimes to make forecasts of temperature and precipitation anomalies, using the association between the regimes and surface weather obtained from reanalysis products. This study explores the utility of k-means cluster analysis on geopotential heights and their ability to make skillful regime predictions in the week 3/4 period. Using 14-day running means of European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5) 500-hPa geopotential heights for the wintertime December–February (DJF) period, circulation regimes are identified using k-means clustering. Each period is assigned a cluster number, allowing the compositing of any reanalysis or observation variable to form cluster maps. Maps of 500-hPa height, 2-m temperature, precipitation, and storm-track anomalies are some of the variables composited. The utility of these relationships in a dynamical forecast setting is tested via Global Ensemble Forecast System v12 (GEFSv12) hindcasts and real-time ensemble suite forecasts. Week 3/4 deterministic and probabilistic experimental forecasts are then derived from cluster assignments using several methods. We find, via a conditional skill analysis, forecasts strongly correlated with a cluster exhibit greater skill for both dynamical model and cluster-derived forecasts. Our preliminary results represent a step forward to aid forecasters make more skillful assessments of the circulation regime and its associated surface weather for this challenging forecast time scale. Significance Statement Our paper links the local statistics of 14-day precipitation and temperature within the United States to very broad preferred patterns of winds and pressure over the wide Pacific–North American region. Such patterns, known as “circulation regimes,” provide the background that heavily influences local surface conditions. We find that forecasts for which midatmospheric anomalies are closely aligned with one or more circulation regimes are better at predicting the U.S. weather probabilities than other week 3–4 forecasts. A tool based on this finding identifies forecasts for which confidence is higher. Future work will be designed to further exploit the links between circulation regimes and weather to improve the accuracy of week 3–4 forecasts.

Combining low- and high-frequency microwave radiometer measurements from the MOSAiC expedition for enhanced water vapour products

Abstract. In the central Arctic, high-quality water vapour observations are sparse due to the low density of meteorological stations and uncertainties in satellite remote sensing. Different reanalyses also disagree on the amount of water vapour in the central Arctic. The Multidisciplinary drifting Observatory for the Study of the Arctic Climate (MOSAiC) expedition provides comprehensive observations that are suitable for evaluating satellite products and reanalyses. Radiosonde observations provide high-quality water vapour estimates with a high vertical but a low temporal resolution. Observations from the microwave radiometers (MWRs) on board the research vessel Polarstern complement these observations through high temporal resolution. In this study, we demonstrate the high accuracy of the combination of the two MWRs HATPRO (Humidity and Temperature Profiler) and MiRAC-P (Microwave Radiometer for Arctic Clouds – Passive). For this purpose, we developed new retrievals of integrated water vapour (IWV) and profiles of specific humidity and temperature using a neural network approach, including observations from both HATPRO and MiRAC-P to utilize their different water vapour sensitivity. The retrievals were trained with the European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis version 5 (ERA5) and synthetic MWR observations simulated with the Passive and Active Microwave radiative TRAnsfer tool (PAMTRA). We applied the retrievals to synthetic and real observations and evaluated them with ERA5 and radiosondes launched during MOSAiC, respectively. To assess the benefit of the combination of HATPRO and MiRAC-P compared to single MWR retrievals, we compared the errors with respect to MOSAiC radiosondes and computed the vertical information content of the specific humidity profiles. The root mean squared error (RMSE) of IWV was reduced by up to 15 %. Specific humidity biases and RMSE were reduced by up to 75 % and 50 %, respectively. The vertical information content of specific humidity could be increased from 1.7 to 2.4 degrees of freedom. We also computed relative humidity from the retrieved temperature and specific humidity profiles and found that RMSE was reduced from 45 % to 15 %. Finally, we show a case study demonstrating the enhanced humidity profiling capabilities compared to the standard HATPRO-based retrievals. The vertical resolution of the retrieved specific humidity profiles is still low compared to radiosondes, but the case study revealed the potential to resolve major humidity inversions. To what degree the MWR combination detects humidity inversions, also compared to satellites and reanalyses, will be part of future work.

CY GNSS significant wave height inversion model based on multivariate machine learning

The Cyclone Global Navigation Satellite System (CYGNSS) provides high-quality Global Navigation Satellite System Reflectometry (GNSS-R) data, which can be reliably used for the inversion of Significant Wave Height (SWH). Due to the high dynamics of CYGNSS, the received signal is easily affected by environmental factors, and the complexity of the sea conditions makes it difficult for simple models to accurately invert SWH. In order to solve the above problems, this paper proposes a multivariate SWH inversion model based on machine learning. According to the formation mechanism of waves and the correlation analysis between CYGNSS parameters and SWH, relevant parameters are selected, and three training schemes of 5 parameters, 9 parameters and 17 parameters are designed. Subsequently, the inversion model was trained and validated using Random Forest (RF) and Convolutional Neural Network (CNN), and the SWH inversion results were compared with the reference values of the European Centre for Medium-range Weather Forecasts (ECMWF). The best inversion model was the 17-parameter CNN inversion model with an RMSE of = 0.1840 m.

Estimation of nighttime PM2.5 concentrations over Seoul using Suomi NPP/VIIRS Day/Night Band

With rapid economic development and urban growth, Seoul experiences severe air pollution due to fine particulate matter with an aerodynamic diameter of ≤2.5 μm (PM2.5), which is detrimental to human health. Although recent studies have extensively focused on estimating daytime PM2.5 concentrations using various types of satellite data, there remains a significant lack of research on nighttime PM2.5 estimations. This study estimated nighttime PM2.5 in Seoul from December 2018 to November 2019 using multiple linear regression (MLR) and random forest (RF) models. These models, which incorporated data on radiance, moon illumination fraction, and terrain height from the Visible Infrared Imaging Radiometer Suite (VIIRS) Day/Night Band (DNB) on board the Suomi National Polar-orbiting Partnership satellite covering all moon phases, also utilized meteorological data from the ERA5 reanalysis by the European Centre for Medium-Range Weather Forecasts (ECMWF). To address multicollinearity, seasonal models were developed using forward stepwise regression and variance inflation factor analysis. DNB radiance analysis indicates that the high intensity of artificial light sources in Seoul significantly reduces the impact of moonlight, leading to notable changes in the DNB radiation associated with PM2.5 concentrations. Consequently, this study estimated nighttime PM2.5 over Seoul across all moon phases. These estimates were then validated through 10-fold cross-validation. The RF model exhibited superior accuracy, with a coefficient of determination (R2) of 0.65–0.90, compared to MLR, with R2 of 0.15–0.50, reflecting seasonal fluctuations in the model performance. The developed models can be applied to estimate reliable nighttime PM2.5 concentrations in megacities with strong artificial light sources, utilizing a comprehensive dataset from satellite observations for all moon phases. Additionally, our findings can serve as scientific data for establishing environmental policies by providing valuable insights into understanding air pollution primarily caused by PM2.5.

A Quick Look at the Atmospheric Circulation Leading to Extreme Weather Phenomena on a Continental Scale

The study delves into the primary large-scale atmospheric features contributing to extreme weather events across Europe during early September 2023. The period was examined using a dataset composed by the European Centre for Medium-Range Weather Forecasts (ECMWF) analysis and satellite imagery. In early September 2023, an omega blocking pattern led to the development of a low-pressure system over the Iberian Peninsula producing heavy precipitation and flooding over Spain and acting as a mechanism for a mineral dust outbreak. A second low-pressure system developed over Greece. Extreme precipitation was recorded across Greece, Turkey, and Bulgaria as the system gradually shifted southward over the Mediterranean. The system earned the name “Storm Daniel” as it acquired subtropical characteristics. It caused floods over Libya and its associated circulation favoured the transport of mineral dust over Northern Egypt as it moved eastward. Meanwhile, the high-pressure blocking system associated with the omega pattern induced heatwave temperatures in countries further north. This period was compared with the large-scale circulation observed in mid-September 2020, when severe weather also affected the Mediterranean region. However, the weather systems were not directly connected by the large-scale circulation, as shown in September 2023. Although mesoscale conditions are relevant to formation and intensification of some atmospheric phenomena, the establishment of an omega blocking pattern in early September 2023 showed how large-scale atmospheric dynamics can produce abnormal weather conditions on a continental scale over several days.

A segmented grid model for vertical adjustment of precipitable water vapor in China

Precipitable Water Vapor (PWV) is vital in climate research, monitoring in troposphere stability and weather disasters. This study investigates the detailed vertical variation of PWV in China, utilizing data from the 5th generation European Centre for Medium-Range Weather Forecasts (ECMWF). A PWV segmented vertical adjustment grid model (CPWV-H) was developed based on a sliding window algorithm, with a spatial resolution of 0.5°×0.5° and a temporal resolution of one day. The performance of the CPWV-H model was evaluated using multi-source PWV data and compared with the conventional empirical model (EPWV-H) and the recently released C-PWVC1 model. Results show that the CPWV-H model has superior accuracy and stability. Using ERA5 PWV profiles as a reference, the CPWV-H model has a mean Bias of -0.45 mm and a mean RMSE of 1.24 mm, improving by 21.52% over the EPWV-H model and 31.11% over the C-PWVC1 model. With radiosonde profiles as a reference, it has a mean Bias of -0.10 mm and a mean RMSE of 2.86 mm, improving by 7.14% and 13.86% over the EPWV-H and C-PWVC1 models, respectively. Consequently, the CPWV-H model exhibits enhanced stability and vertical adjustment accuracy in China, offering significant utility for the fusion of multi-source water vapor data, comparative analysis, and climate research.

Radar-Derived Convective Features and Atmospheric Environments of Hail-Producing Storms over the Contiguous United States

Abstract Hail smaller than 0.75 in. is known to cause economic impacts yet remains understudied due to report biases towards recording larger hail sizes (≤1 in.). In this study, we assembled ground hail reports during 2017–22 from the National Centers for Environmental Information (NCEI) Storm Data, Community Collaborative Rain, Hail and Snow Network (CoCoRaHS), and Meteorological Phenomena Identification Near the Ground (mPING) databases. Then, these reports are collocated with the attributes of radar-derived convective features from the Multi-Radar Multi-Sensor (MRMS) system and fifth major global reanalysis produced by European Centre for Medium-Range Weather Forecasts (ERA5) atmospheric vertical profiles to construct a dataset describing properties of a full spectrum of hailstorms. The characteristics of radar reflectivity and atmospheric profiles are examined for hail of different sizes reported within selected regions over the contiguous United States (CONUS). In addition to the seasonal and diurnal variations, the morphology of convective features shows apparent regional differences from west to east in CONUS. The maximum expected size of hail (MESH) performance against reported hail sizes shows underestimation of hail with significant sizes and overestimation of small hail sizes. ERA5 vertical atmospheric profiles are explored to form relationships between storm environment and hail sizes. In addition to the relationships between wind shear and hail sizes, the roles of low-level relative humidity and freezing level height in regard to hail melting are discussed.

Station keeping control method based on deep reinforcement learning for stratospheric aerostat in dynamic wind field

Aiming at the problem of long-endurance station keeping of stratospheric aerostats under energy constraints in dynamic wind environment, a stratospheric aerostat station keeping controller was designed using the D3QN (Dueling Double Deep Q Network) algorithm of deep reinforcement learning in this article. The performance of the station keeping controllers are evaluated by the average station keeping radius and station keeping effective time ratio. Meanwhile, the effects of different reward functions on the performance of station keeping controller are studied. The simulation results of typical scenarios with European Centre for Medium-Range Weather Forecasts (ECMWF) data show that: under the task constraints of three days of station keeping time and 50 km of station keeping radius, the average station keeping radius of the stratospheric aerostat without power propulsion is 28.25 km, the station keeping time ratio is 90.56%, and the State of Charge (SOC) of the battery is more than 50%. Under the condition of propulsion, the control effect of the average station keeping radius of stratospheric aerostat is significantly improved. The control effect of adding dynamic propulsion in the north-south direction is better than that in the east-west direction, which is close to the dual-channel control effect. At the same time, this study verifies that the station keeping controller based on deep reinforcement learning has strong robustness, and demonstrates that the station keeping controller can be designed with different reward functions for the requirements of different station keeping task.

Random Forests’ Postprocessing Capability of Enhancing Predictive Skill on Subseasonal Time Scales—A Flow-Dependent View on Central European Winter Weather

Abstract Weather predictions 2–4 weeks in advance, called the subseasonal time scale, are highly relevant for socioeconomic decision-makers. Unfortunately, the skill of numerical weather prediction models at this time scale is generally low. Here, we use probabilistic random forest (RF)-based machine learning models to postprocess the subseasonal to seasonal (S2S) reforecasts of the European Centre for Medium-Range Weather Forecasts (ECMWF). We show that these models are able to improve the forecasts slightly in a 20-winter mean at lead times of 14, 21, and 28 days for wintertime central European mean 2-m temperatures compared to the lead-time-dependent mean bias-corrected ECMWF’s S2S reforecasts and RF-based models using only reanalysis data as input. Predictions of the occurrence of cold wave days are improved at lead times of 21 and 28 days. Thereby, forecasts of continuous temperatures show a better skill than forecasts of binary occurrences of cold wave days. Furthermore, we analyze if the skill depends on the large-scale flow configuration of the atmosphere at initialization, as represented by weather regimes (WRs). We find that the WR at the start of the forecast influences the skill and its evolution across lead times. These results can be used to assess the conditional improvement of forecasts initialized during one WR in comparison to forecasts initialized during another WR. Significance Statement Forecasts of winter temperatures and cold waves 2–4 weeks in advance done by numerical weather prediction (NWP) models are often unsatisfactory due to the chaotic characteristics of the atmosphere and limited predictive skill at this time range. Here, we use statistical methods, belonging to the so-called machine learning (ML) models, to improve forecast quality by postprocessing predictions of a state-of-the-art NWP model. We compare the forecasts of the NWP and ML models considering different weather regimes (WRs), which represent the large-scale atmospheric circulation such as the typical westerly winds in Europe. We find that the ML models generally yield better temperature forecasts for 14, 21, and 28 days in advance and better forecasts of cold wave days 21 and 28 days in advance. The quality of forecasts depends on the WR present at the forecast start. This information can be used to assess the conditional improvement of forecasts.

An Empirical Model for Skillful Rogue Wave Forecasts

Abstract One of the leading goals of rogue wave research is to develop a robust rogue wave warning system to mitigate the danger they pose. One such system has been developed by the European Centre for Medium-Range Weather Forecasts (ECMWF), called the freak wave warning system (FWWS), based on nonlinear wave effects. The FWWS predicts maximum expected wave envelope height as a risk parameter for forecasts. Recently, a data-driven alternative has been proposed by Häfner et al., which was distilled from a neural network using wave buoy observations. However, it has yet to be evaluated by a spectral wave model for application to operational wave forecasting. The data-driven, learned model emphasizes bandwidth-controlled linear superposition as the predominant mechanism in crest-to-trough rogue wave generation, while nonlinear effects are a secondary term. The present work evaluates the performance of the empirical model using output from an ECMWF global wave hindcast. We find that the prediction models based on bandwidth effects have the highest log likelihood scores, with the empirical model outperforming all other tested models. In contrast, the expected maximum envelope wave height from the FWWS does not predict the occurrence of rogue waves. These results indicate that the empirical model with wave model input is a skillful predictor and should be considered for operational implementation to improve rogue wave forecasting. Significance Statement Rogue waves are unexpectedly large and unpredictable waves. Encounters with rogue waves can result in damage to marine vessels and offshore infrastructures. Research is devoted to developing systems to predict the risk of rogue wave events. A novel predictive model has been shown to perform well in predicting rogue wave occurrences based on wave buoy observations. This model is a symbolic expression distilled from an artificial neural network that incorporates known rogue wave dynamics and that can be evaluated alongside traditional risk estimates with minor adjustments to operational forecasting systems. This study evaluates the newly proposed empirical model in a forecasting setting. Our work demonstrates the efficacy of the empirical model for operational forecasting purposes.

Typhoon Trajectory Prediction by Three CNN+ Deep-Learning Approaches

The accuracy in predicting the typhoon track can be key to minimizing their frequent disastrous effects. This article aims to study the accuracy of typhoon trajectory prediction obtained by combining several algorithms based on current deep-learning techniques. The combination of a Convolutional Neural Network with Long Short-Term Memory (CNN+LSTM), Patch Time-Series Transformer (CNN+PatchTST) and Transformer (CNN+Transformer) were the models chosen for this work. These algorithms were tested on the best typhoon track data from the China Meteorological Administration (CMA), ERA5 data from the European Centre for Medium-Range Weather Forecasts (ECMWF), and structured meteorological data from the Zhuhai Meteorological Bureau (ZMB) as an extension of existing studies that were based only on public data sources. The experimental results were obtained by testing two complete years of data (2021 and 2022), as an alternative to the frequent selection of a small number of typhoons in several years. Using the R-squared metric, results were obtained as significant as CNN+LSTM (0.991), CNN+PatchTST (0.989) and CNN+Transformer (0.969). CNN+LSTM without ZMB data can only obtain 0.987, i.e., 0.004 less than 0.991. Overall, our findings indicate that appropriately augmenting data near land and ocean boundaries around the coast improves typhoon track prediction.

Warming events and their causes at the Bering Sea section B in summer of 1999–2019

Based on hydrological data for the Bering Sea from 1999 to 2019 during the Chinese National Arctic Research Expeditions (CHINAREs), along with oceanographic and meteorological data from the National Centers for Environmental Information (NCEI) and the European Centre for Medium-Range Weather Forecasts (ECMWF), we investigated trends in summer at the Bering Sea section B over the past 20 years and their underlying causes. The results revealed a pronounced rise in temperature at the Bering Sea section B in 2003, 2014, 2016, 2018, and 2019, suggesting a shift towards a new phase of warmth starting in 2014. In warm years (2003, 2014, 2016, 2018, and 2019), the mean maximum temperature in the upper layer in the basin north of 57°N and on the continental shelf were roughly 2.7 °C, 2.6 °C higher than that in cold years (1999, 2008, 2010, and 2012). Additionally, the mean minimum temperature of the cold water in the middle layer in the basin was roughly 0.7 °C higher in the warm years compared to the cold years. This difference between the warm and cold years was evident in the heat content: the average heat content in the upper to middle layers in the basin north of 57°N and on the continental shelf were about 1.2 and 0.9 GJ/m2 higher, respectively, in the warm years than the cold years. However, there was a modest warming trend in the basin south of 57°N, and the temperature in the mid-layer on the continental shelf showed year-to-year stability, in contrast to the significant warming observed elsewhere. Further study suggested that temperature fluctuation in the basin was closely tied to seasonal 2-m air temperature. In addition, the southern basin, divided at 57°N, experienced larger positive or negative anomalous wind stress curl compared to the northern basin. This discrepancy resulted in differing causes of temperature variation between the two regions. Temperature change in the upper layer on the continental shelf was attributed to an increase in cumulative net heat flux resulting from reduced sea ice, while the temperature in the middle layer on the continental shelf was limited by the enhanced density gradient between the upper and middle layer caused by melt ice, preventing significant warming. Overall, the thermal state for the Bering Sea section B was influenced by a combination of factors, including net heat flux, sea ice, and atmospheric conditions, contributing to the year-to-year variation characterized by alternating cold and warm regimes.

PreciDBPN: A customized deep learning approach for hourly precipitation downscaling in eastern China

Long-term series of high-resolution gridded precipitation datasets are essential for hydrological and meteorological research. Producing high-resolution precipitation data from regional models demands substantial computational resources and labor. Global Reanalyses offer long-term coverage and effectively capture annual and seasonal precipitation patterns. However, they have inadequate resolution and frequently have difficulties depicting extreme conditions. This study proposes an efficient and accurate approach for generating long-term series of high spatial and temporal resolution precipitation. It is achieved by leveraging deep learning techniques to integrate the fifth generation of the European Centre for Medium-Range Weather Forecasts (ECMWF) global climate reanalysis (ERA5, 0.25°, hourly) data with high-resolution precipitation fusion datasets. Considering the heavy-tailed distribution of precipitation, we developed the PreciDBPN model structure, which combines a classification network with a super-resolution network and incorporates physically relevant indices into the model's input. We trained and evaluated the PreciDBPN and baseline models in eastern China using the China Meteorological Administration Land Data Assimilation System (CLDAS) precipitation dataset (0.0625°, hourly, 2017–2022). When compared to baseline methods and ERA5, our model excels in multiple metrics and provides a more precise representation of relative rainfall frequency. Independent verification was performed using station observations during the period of 2010–2015 when CLDAS data were unavailable. During this verification, the PreciDBPN demonstrated exceptional performance and greater robustness compared to the baseline models. Because our method can efficiently downscale precipitation and bias-correct reanalysis data using minimal computational resources, it can be used to generate high-resolution precipitation datasets (0.0625°, hourly) from 1979 to 2022 while correcting for heavy precipitation underestimations in reanalysis data.

Sensitivity of horizontal resolution and land surface model in operational WRF forecast for Online Nuclear Emergency Response System (ONERS)

Accurate Meteorological forecasts are crucial for the assessment of plume dispersion and dose prediction in nuclear power plant (NPP) sites. In this work the forecast sensitivity of the Weather Research and Forecasting (WRF) model is tested by running a series of forecast simulations for horizontal resolution, and land surface models (LSM) in the context of Online Nuclear Emergency Response System (ONERS) for Indian NPP sites. 72 h forecast simulations are made for three seasons viz. summer, southeast and northeast monsoon using the Global Forecast data. Three simulation experiments, namely 2 km-NOAH, 3 km-NOAH and 3 km-NOAHMP are conducted using two different nested domain configurations (18–6–2 km and 9–3 km) and two LSM schemes (NOAH and NOAH-MP) and tested at four different NPP sites. Forecast comparison of surface winds, relative humidity, temperature, heat fluxes and planetary boundary layer heights with data from meteorological tower, radiosonde and the Modern-Era Retrospective analysis for Research and Applications 2 (MERRA-2) shows 3 km-NOAH is equally capable in predicting surface parameters as well as vertical profiles compared to 2 km-NOAH with marginal differences. 3 km-NOAHMP shows less mean bias and better correlation for boundary layer height and heat fluxes. Comparison of spatial flow-field with 5th generation European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis (ERA5) data shows synoptic scale seasonal winds, sea level pressure systems and temperature hot-spots are better captured by 3 km-NOAHMP compared to 6 km coarse domain in the 18–6–2 km configuration. The daily accumulated rainfall by all simulations is overestimated compared to ERA5 data. The predictions by 3 km-NOAHMP better agree with Integrated Multi-satellite Retrievals for Global Precipitation Measurement (GPM-IMERGE) data whereas 2 km-NOAH predicts delayed rainfall occurrence. Dispersion simulations of hypothetical plume release from a coastal NPP site with all three forecasts properly show the influence of local scale diurnal land-sea breeze and seasonal winds on the plume movement. Therefore the 9–3 km domain with NOAHMP LSM is found to be a suitable choice for operational weather forecast in ONERS for Indian NPP sites.

Assessment of Wind Energy Potential in Kolaka Using Era5-land Reanalysis Data

This research aims to evaluate the potential for wind energy in Kolaka Regency using ERA5-Land reanalysis data from the output of the European Center for Medium-Range Weather Forecasts (ECMWF) model, which has a spatial resolution of 9 km with hourly temporal resolution for the period of 2001 to 2020. This research also compared the quality of ERA5-Land reanalysis wind speed data on data from nearby meteorological stations using statistical assessment methods. The analysis results reveal that the highest average wind speed is discovered at a height of 80 to 150 meters with a range of 4.99 to 5.93 m/s. The spatial distribution map of wind power density indicates that the energy potential at a height of 80 to 150 meters can reach 60 to 170 W/m2. The largest wind energy potential in Kolaka Regency can be generated in the DJF period (December-January-February) with the highest average monthly wind power density in January and the JJA period (June-July-August) with the highest average in July. In general, Kolaka Regency can be classified as an area with first-class wind energy potential, so it tends to be more suitable for the development of small-scale power plants. The suitable locations for developing small-scale wind power plants in Kolaka Regency are the southern Pomalaa coast and the Tanggetada coast because both locations have better wind energy sources than other areas.

Abnormal high water temperature prediction in nearshore waters around the Korean Peninsula using ECMWF ERA5 data and a deep learning model

The abnormally high-water temperature (AHWT) phenomena have caused the mass stranding of farmed fish in the Korean coastal waters, leading to a substantial monetary loss in recent decades. It is most important to predict the HWT occurrence and take responsive measures before the HWT arrival to prevent such loss, we proposed a methodology to predict HWT occurrences using a deep-learning technology. Firstly, we trained a long short-term memory (LSTM) deep-learning model using the sea surface temperature data from the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 product to estimate future water temperature in advance. Secondly, we used the estimated water temperature data to predict HWT occurrences from 1 day to 7 days later. We computed root mean square error (RMSE), mean absolute percentage error (MAPE) metrics, and F1-scores to evaluate the accuracy of the proposed LSTM model. In the cases of 1-day and 7-day water temperature predictions, RMSE and MAPE values between the estimated data and the sea-truth data were 0.293 degrees Celsius with 1.313 % and 0.854 degrees Celsius with 4.175 %, respectively. The F1-scores of the classification algorithm of 1- and 7-day HWT predictions were 0.96 and 0.74, respectively. This study contributes to developing measures to reduce the monetary loss of HWT damage on fish farms.

Windformer: A novel 4D high-resolution system for multi-step wind speed vector forecasting based on temporal shifted window multi-head self-attention

Accurate wind speed forecasting (WSF) plays a pivotal role in anticipating the power output of wind farms. Nevertheless, the stochastic and variable nature of wind speed presents a significant challenge in achieving accurate WSF. Hence, this study proposes a novel 4D high-resolution system, Windformer, for wind speed vector forecasting (WSVF). Windformer combines the ability of convolutional neural networks (CNNs) for feature extraction and transformers based on attention mechanisms for information fusion. In the Windformer, the input and output layers are mainly composed of 3D CNNs. The input layer is employed for feature extraction and information compression, while the output layer is tasked with recovering the WSV field. The key components of the model consist of encoders and decoders built upon temporal shifted window multi-head self-attention. This architecture is capable of effectively integrating spatio-temporal information. Trained on 39 years of regional reanalysis data, Windformer obtains the most accurate forecast results compared to 5 WSF baseline models. Most importantly, within the next 6 hours, Windformer demonstrates higher accuracy compared to the High-Resolution Deterministic Forecast (HRES) from the European Centre for Medium-Range Weather Forecasts (ECMWF).

A Study on the Impact of Vertical Grid Parameter Perturbations in the Regional Ocean Modeling System

In this study, the Regional Ocean Modeling System (ROMS) is employed to construct a three-dimensional barotropic ocean model with a monodirectional upper boundary and homogeneous and steady wind covering the entire computation area. Eight perturbation experiments are designed to determine the vertical grid distribution difference with high resolution at the surface and bottom. Two types are considered in the model, including removing the Coriolis force (type 1) and employing a different Coriolis force (type 2). According to the experiments, the velocity of the current in type 1 yields uncertainty, and wind energy could penetrate the upper ocean and reach the abyss. The surface velocity in type 2 is fundamentally compatible with the empirical relationship constructed by Ekman, and the curved lines of the vertical distribution of horizontal currents nearly match. For type 1, the velocity is very strong from the sea surface to the bottom. When comparing type 1 and type 2 cases, the Coriolis force obstructs the wind energy transfer into the deep ocean. In addition, the European Centre for Medium-Range Weather Forecasts (ECMWF)’s global surface wind distribution indicates that the realistic ocean upper wind boundary is similar to the numerical experiment in the Pacific and Atlantic oceans, where the wind direction is along the latitude line at the equator. In order to make the experimental situation as close as possible to the real ocean, validation experiments are conducted in this study to consider the uncertainty in the current profile at the equator. The simulation results of type 1 differ significantly from the data obtained from the real ocean. This uncertainty may transfer the signal to higher latitudes, causing incorrect simulation results, especially in the critical region. Overall, this research not only makes discoveries in physical ocean theory but also guides predictive and forecasting techniques for ocean modeling.

An Automated Machine Learning Approach to the Retrieval of Daily Soil Moisture in South Korea Using Satellite Images, Meteorological Data, and Digital Elevation Model

Soil moisture is a critical parameter that significantly impacts the global energy balance, including the hydrologic cycle, land–atmosphere interactions, soil evaporation, and plant growth. Currently, soil moisture is typically measured by installing sensors in the ground or through satellite remote sensing, with data retrieval facilitated by reanalysis models such as the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis 5 (ERA5) and the Global Land Data Assimilation System (GLDAS). However, the suitability of these methods for capturing local-scale variabilities is insufficiently validated, particularly in regions like South Korea, where land surfaces are highly complex and heterogeneous. In contrast, artificial intelligence (AI) approaches have shown promising potential for soil moisture retrieval at the local scale but have rarely demonstrated substantial products for spatially continuous grids. This paper presents the retrieval of daily soil moisture (SM) over a 500 m grid for croplands in South Korea using random forest (RF) and automated machine learning (AutoML) models, leveraging satellite images and meteorological data. In a blind test conducted for the years 2013–2019, the AutoML-based SM model demonstrated optimal performance, achieving a root mean square error of 2.713% and a correlation coefficient of 0.940. Furthermore, the performance of the AutoML model remained consistent across all the years and months, as well as under extreme weather conditions, indicating its reliability and stability. Comparing the soil moisture data derived from our AutoML model with the reanalysis data from sources such as the European Space Agency Climate Change Initiative (ESA CCI), GLDAS, the Local Data Assimilation and Prediction System (LDAPS), and ERA5 for the South Korea region reveals that our AutoML model provides a much better representation. These experiments confirm the feasibility of AutoML-based SM retrieval, particularly for local agrometeorological applications in regions with heterogeneous land surfaces like South Korea.

Microphysical Characteristics of Monsoon Precipitation over Yangtze-and-Huai River Basin and South China: A Comparative Study from GPM DPR Observation

It is rare to conduct a comparative analysis of precipitation characteristics across regions based on long-term homogeneous active satellite observations. By collocating the Global Precipitation Measurement Dual-frequency Precipitation Radar (GPM DPR) observations with European Centre for Medium-Range Weather Forecasts 5th Reanalysis (ERA5) data, this study comparatively examines the microphysics of monsoon precipitation in the rainy season over the Yangtze-and-Huai River Basin (YHRB) and South China (SC) from 2014 to 2023. The comparative analysis is made in terms of precipitation types and intensities, precipitation efficiency index (PEI), and ice phase layer (IPL) width. The results show that the mean near-surface precipitation rate and PEI are generally higher over SC (2.87 mm/h, 3.43 h−1) than over YHRB (2.27 mm/h, 3.22 h−1) due to the more frequent occurrence of convective precipitation. The DSD characteristics of heavy precipitation in the wet season for both regions are similar to those of deep ocean convection, which is associated with a greater amount of water vapor. However, over SC, there are larger but fewer raindrops in the near-surface precipitation. Moreover, moderate PEI precipitation is the main contributor to heavy precipitation (>8 mm/h). Stratiform precipitation over YHRB is frequent enough to contribute more than convective precipitation to heavy precipitation (8–20 mm/h). The combined effect of stronger convective available potential energy and low-level vertical wind favors intense convection over SC, resulting in a larger storm top height (STH) than that over YHRB. Consequently, it is conducive to enhancing the microphysical processes of the ice and melt phases within the precipitation. The vertical wind can also influence the liquid phase processes below the melting layer. Collectively, these dynamic microphysical processes are important in shaping the efficiency and intensity of precipitation.