Article
Engineering, Civil
Yiliang Du, Quan J. Wang, Wenyan Wu, Qichun Yang
Summary: Short-term precipitation forecasts often require statistical calibration to improve accuracy and reliability. This study finds that using a regionally optimized two parameter value approach can achieve good forecast calibration performance while maintaining spatial consistency.
JOURNAL OF HYDROLOGY
(2022)
Article
Geosciences, Multidisciplinary
T. Honda, A. Amemiya, S. Otsuka, J. Taylor, Y. Maejima, S. Nishizawa, T. Yamaura, K. Sueki, H. Tomita, T. Miyoshi
Summary: This study investigates the benefits of a 30-second-updating NWP system in predicting convective precipitation events, and shows that continuously assimilating PAWR observations can improve forecast accuracy by consistently modifying moisture and dynamical fields.
GEOPHYSICAL RESEARCH LETTERS
(2022)
Article
Engineering, Civil
Pengcheng Zhao, Quan J. Wang, Wenyan Wu, Qichun Yang
Summary: This paper proposes a two-step calibration approach that combines the strengths of joint probability models and the useful information included in the ensemble spread. In the first step, the ensemble mean is calibrated using a seasonally coherent calibration model. In the second step, the ensemble forecasts are re-calibrated to incorporate the ensemble spread information. The results show that forecasts calibrated using the two-step calibration approach have better skills.
JOURNAL OF HYDROLOGY
(2022)
Article
Geosciences, Multidisciplinary
Fabricio Polifke da Silva, Alfredo Silveira da Silva, Maria Gertrudes Alvarez Justi da Silva, Gisele Dornelles Pires
Summary: This study focused on identifying extreme precipitation events in a city in Brazil and evaluating precipitation forecasts using the WRF model. Findings showed that increasing horizontal resolution improved model performance and better predicted extreme precipitation events.
Article
Green & Sustainable Science & Technology
Dazhi Yang, Guoming Yang, Bai Liu
Summary: This study focuses on optimally combining quantiles of ensemble solar forecasts that have been post-processed. Numerical weather prediction (NWP) is used to provide dynamical ensemble irradiance forecasts for solar energy grid integration. However, these ensemble members often lack dispersion, leading to the use of statistical calibration techniques such as quantile regression (QR) and ensemble model output statistics (EMOS). Due to the numerous variants of QR and EMOS, it is unclear which variant performs best in different situations, motivating the combination of quantile forecasts. A framework for combining solar forecasts in the form of quantiles is proposed and demonstrated using a constrained quantile regression averaging scheme. Results show that combining quantiles is an effective strategy in improving the calibrated forecasts.
Article
Meteorology & Atmospheric Sciences
Enric Casellas, Joan Bech, Roger Veciana, Nicolau Pineda, Josep Ramon Miro, Jordi More, Tomeu Rigo, Abdel Sairouni
Summary: Heavy snowfall events can cause transport disruption and negative socioeconomic impact in regions with infrequent snowfall, especially if there are rapid transitions between different precipitation phases. Previous studies have used precipitation-phase nowcasting techniques, and this research developed and evaluated a nowcasting scheme considering the precipitation phase.
QUARTERLY JOURNAL OF THE ROYAL METEOROLOGICAL SOCIETY
(2021)
Article
Engineering, Civil
Velpuri Manikanta, Jew Das, K. Nikhil Teja, N. V. Umamahesh
Summary: This study examines the capability of ensemble precipitation forecasts obtained from two NWP models, and post-processes the raw ensemble members using two methods. The findings suggest that QRF post-processed forecasts are superior to other forecasts in terms of all verification measures at shorter lead times. However, the skill of both raw and post-processed forecasts declines at higher lead times.
JOURNAL OF HYDROLOGY
(2023)
Article
Computer Science, Information Systems
Cenker Sengoz, Sheela Ramanna, Scott Kehler, Rushil Goomer, Paul Pries
Summary: In this study, a multimodel (ensemble) forecasting approach was adopted to deploy an optimal machine learning-based weather model for real-time precipitation forecasting. Machine learning techniques were used to combine precipitation data from multiple NWP models and improve upon their individual results. The best results were achieved by neural network variants, which showed significant improvements in various error metrics compared to the baseline model.
Article
Computer Science, Hardware & Architecture
Chang-Hoo Jeong, Mun Yong Yi
Summary: This study proposes a deep learning method to improve the performance of numerical weather prediction models. By learning the transformation relationship between the output of the numerical model and the observed data using a generative adversarial network, the method corrects the forecasts of the numerical model. Experiments demonstrate that the correction of forecast data using this method improves prediction performance, especially for heavy rainfall events.
JOURNAL OF SUPERCOMPUTING
(2023)
Article
Environmental Sciences
Mekonnen Gebremichael, Haowen Yue, Vahid Nourani
Summary: This study evaluates the performance of the Global Forecast System (GFS) for precipitation forecasts in the Volta river basin in West Africa. The results show that the accuracy of the GFS forecasts varies depending on the climate zone, with underestimation bias in dry areas and overestimation bias in wet areas. However, aggregating the forecasts over longer timescales improves the skill of the GFS.
Article
Engineering, Civil
Aline S. Falck, Javier Tomasella, Fabio L. R. Diniz, Viviana Maggioni
Summary: The study demonstrates the potential of a stochastic error model to generate precipitation ensemble fields from a regional numerical weather forecasting model, reducing both systematic and random errors. Compared to the more sophisticated ensemble techniques used by the ECMWF model, SREM2D is proven to be an efficient technique with low computational cost.
JOURNAL OF HYDROLOGY
(2021)
Article
Meteorology & Atmospheric Sciences
Akshay Singhal, Muhammed Jaseem, Sanjeev K. Jha
Summary: Understanding the spatial variability of extreme precipitation events is a challenging task, especially with the complications brought by climate change. In this study, the researchers evaluated the performance of four gridded Quantitative Precipitation Forecasts (QPFs) in detecting the spatial connections among extreme precipitation events in the Ganga River basin of India. The study found that the European Centre for Medium-range Weather Forecasts (ECMWF) model performed well and could be a suitable substitute for the national agency model.
ATMOSPHERIC RESEARCH
(2022)
Article
Multidisciplinary Sciences
Tanja C. Portele, Christof Lorenz, Berhon Dibrani, Patrick Laux, Jan Bliefernicht, Harald Kunstmann
Summary: Increased frequencies of droughts require proactive preparedness, especially in semi-arid regions. Seasonal forecasting systems such as SEAS5 can provide valuable economic savings in water management decision making by predicting such hydrometeorological extremes several months ahead. Consideration of seasonal forecasts is advantageous and necessary in hydrological decision making to avoid significant economic losses.
SCIENTIFIC REPORTS
(2021)
Article
Geosciences, Multidisciplinary
Oliver Branch, Thomas Schwitalla, Marouane Temimi, Ricardo Fonseca, Narendra Nelli, Michael Weston, Josipa Milovac, Volker Wulfmeyer
Summary: The study shows that WRF performs well in predicting temperature and humidity parameters during the daytime in the UAE, with some nocturnal cold bias. There are varying degrees of biases for temperature (T-2m) and dew point (TD2m) in different regions, while wind speed (UV 10m) performance still needs improvement.
GEOSCIENTIFIC MODEL DEVELOPMENT
(2021)
Article
Meteorology & Atmospheric Sciences
Pengcheng Zhao, Quan J. Wang, Wenyan Wu, Qichun Yang
Summary: Deterministic numerical weather prediction models and ensemble numerical weather prediction models are both used worldwide to assist weather forecasting. Ensemble forecasts are found to outperform deterministic forecasts in terms of correlation, accuracy, and reliability when comparing their performance in forecasting daily precipitation in Australia over a 3-year period, despite their coarser resolution. Post-processing greatly improves the forecasts from both models, narrowing the performance gap between them.
QUARTERLY JOURNAL OF THE ROYAL METEOROLOGICAL SOCIETY
(2021)
Article
Medicine, General & Internal
Chang Sun, Serge Richard, Takemasa Miyoshi, Naohiro Tsuzu
Summary: This paper presents an agent-based model and a particle filter approach for studying the spread of COVID-19. The research focuses mainly on Tokyo but also briefly surveys other prefectures in Japan. A novel method for evaluating the effective reproduction number and other unknown parameters is introduced. The stability of the computations and the testing and discussion of uncertain quantities are also examined.
JOURNAL OF CLINICAL MEDICINE
(2022)
Article
Mathematics, Applied
Q. Sun, T. Miyoshi, S. Richard
Summary: We introduce an extended SEIR infectious disease model with data assimilation to study the spread of COVID-19, taking into account undetected asymptomatic and pre-symptomatic cases. An ensemble Kalman filter is implemented to assimilate reliable observations and estimate the effective reproduction number and unobservable subpopulations. The analysis is conducted for three main prefectures of Japan and the entire country, revealing more stable estimated reproduction numbers compared to a different method. Sensitivity tests also suggest the decreasing efficiency of states of emergency.
JOURNAL OF COMPUTATIONAL AND APPLIED MATHEMATICS
(2023)
Article
Meteorology & Atmospheric Sciences
Jianyu Liang, Koji Terasaki, Takemasa Miyoshi
Summary: This study applies machine learning as a substitute for physically based radiative transfer models in satellite data assimilation, known as ML-OO, and constructs a reference system using the nonhydrostatic icosahedral atmospheric model and local ensemble transform Kalman filter. The experiment shows that the assimilation using ML-OO is slightly inferior to RTTOV-OO but outperforms assimilation based on traditional observations alone.
JOURNAL OF THE METEOROLOGICAL SOCIETY OF JAPAN
(2023)
Article
Meteorology & Atmospheric Sciences
Tatiana Nomokonova, Philipp J. Griewank, Ulrich Loehnert, Takemasa Miyoshi, Tobias Necker, Martin Weissmann
Summary: This study focuses on the potential improvement of short-term forecasts of low-level wind using a network of Doppler lidars. A new methodology based on ensemble sensitivity analysis (ESA) is developed to assess the impact. The study demonstrates that a network of 20-30 Doppler lidars leads to a considerable variance reduction and can significantly reduce the 1-3 hr forecast error by a factor of 1.6-3.3 compared to surface-wind observations only. These findings provide the basis for designing an operational network of Doppler lidars for the renewable energy sector.
QUARTERLY JOURNAL OF THE ROYAL METEOROLOGICAL SOCIETY
(2023)
Article
Meteorology & Atmospheric Sciences
James Taylor, Takumi Honda, Arata Amemiya, Shigenori Otsuka, Yasumitsu Maejima, Takemasa Miyoshia
Summary: A sensitivity analysis was conducted for a numerical weather prediction system that uses observations from a new-generation weather radar to update a 500m mesh with a 30-second refresh rate. The aim was to determine the optimal scale for short-range forecasting of convective systems and understand the model's behavior to rapid updates. The results showed that while the model performed well within a 30-minute lead time, it consistently overestimated rainfall and did not outperform simpler nowcast models. Using a larger localization scale generated more intense convection in the analyses and hindered forecast accuracy.
WEATHER AND FORECASTING
(2023)
Article
Meteorology & Atmospheric Sciences
A. Amemiya, M. Shlok, T. Miyoshi
Summary: This study demonstrates the application of machine learning, particularly recurrent neural networks, in model bias correction for weather prediction. Idealized experiments using different architectures of neural networks and simple linear regression are performed to compare their effectiveness. The results show that neural networks generally outperform linear regression, and recurrent neural networks perform the best in detecting and reducing systematic biases in weather models.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2023)
Article
Water Resources
Mao Ouyang, Shunji Kotsuki, Yuka Ito, Tomochika Tokunaga
Summary: This study implemented a hydraulic model to investigate flood events in Mobara city, Japan. The model accurately reproduced the floods in the study area, and it was found that the extent of flooding had increased by a factor of 5.5 over a 50-year period. Therefore, it is important to consider changes in land use/land covers, topographies, and input river water levels in urban flood mapping.
JOURNAL OF HYDROLOGY-REGIONAL STUDIES
(2022)
Article
Oceanography
Shun Ohishi, Takemasa Miyoshi, Misako Kachi
Summary: This paper presents an eddy-resolving local ensemble transform Kalman filter (LETKF)-based research analysis system, LORA-WNP and LORA-MC, for the western North Pacific (WNP) and Maritime Continent (MC) regions, respectively. Validation comparisons with JCOPE2M reanalysis and AVISO observational datasets show that LORA-WNP has better agreement in surface horizontal velocity with independent drifter buoy observations in the mid-latitude region, especially along the Kuroshio Extension (KE), while LORA-MC has closer agreement in surface velocity with independent drifter buoys in the equatorial coastal region. The results also demonstrate that LORA-WNP and LORA-MC have sufficient accuracy for geoscience research applications as well as for fisheries, marine transport, and environment consultants.
Article
Meteorology & Atmospheric Sciences
Akira Yamazaki, Koji Terasaki, Takemasa Miyoshi, Shunsuke Noguchi
Summary: This study evaluates the contribution of assimilating AMSU-A satellite-based radiance measurements to a global data assimilation system. The observations were found to have the strongest impact in the upper troposphere, particularly in the austral midlatitudes where westerly jets exist. The accumulated observation impact was tied to dynamic processes in the upper-tropospheric and general stratospheric circulation.
WEATHER AND FORECASTING
(2023)
Article
Multidisciplinary Sciences
Iyan E. Mulia, Naonori Ueda, Takemasa Miyoshi, Takumu Iwamoto, Mohammad Heidarzadeh
Summary: In this study, a deep learning method based on generative adversarial networks (GAN) was used to translate parametric model outputs into realistic atmospheric forcing fields resembling numerical weather prediction (NWP) model results. Lead-lag parameters were also introduced to incorporate a forecasting feature in the model. The results showed that the storm surge model accuracy with forcings generated by GAN was comparable to that of the NWP model and outperformed the parametric model. This novel GAN model offers an alternative for rapid storm forecasting and has the potential to improve forecasts further by combining diverse data sources.
SCIENTIFIC REPORTS
(2023)
Article
Meteorology & Atmospheric Sciences
Shunji Kotsuki, Koji Terasaki, Masaki Satoh, Takemasa Miyoshi
Summary: This study improves precipitation forecasts using GPM DPR data through model parameter estimation. The NICAM-LETKF and GPM DPR observations are used to estimate a model cloud physics parameter related to snowfall terminal velocity. By using a two-dimensional histogram-based parameter estimation method, the gap between simulated and observed data is effectively reduced, leading to improved 6 hr precipitation forecasts.
JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES
(2023)
Article
Geosciences, Multidisciplinary
Mao Ouyang, Keita Tokuda, Shunji Kotsuki
Summary: Controlling weather is a challenging task due to the chaotic nature of the atmosphere. This study uses a control simulation experiment on the Lorenz-63 model to demonstrate that variables can be controlled by adding perturbations with a constant magnitude. By investigating the impact of controls on system instability, the researchers propose an adaptive method to update the magnitude of perturbations, leading to a reduction in control times and magnitudes. The results suggest that understanding the effects of control on instability is beneficial for designing feasible methods to control the complex atmosphere.
NONLINEAR PROCESSES IN GEOPHYSICS
(2023)
Article
Geosciences, Multidisciplinary
Qiwen Sun, Takemasa Miyoshi, Serge Richard
Summary: The control simulation experiment (CSE) is a newly developed approach for studying the controllability of dynamical systems, based on the well-known observing system simulation experiment (OSSE) in meteorology. The CSE aims to reduce extreme weather events in the Lorenz-96 model by exploiting the system's sensitivity to initial conditions and applying minimal perturbations. The study discusses the impact of various parameters of the CSE on the reduction of extreme events over a 100-year simulation.
NONLINEAR PROCESSES IN GEOPHYSICS
(2023)
Article
Geosciences, Multidisciplinary
Tobias Necker, David Hinger, Philipp Johannes Griewank, Takemasa Miyoshi, Martin Weissmann
Summary: This study investigates vertical localization based on a unique 1000-member ensemble simulation and derives an empirical optimal localization that minimizes sampling error. The results suggest that vertical localization should consider variables, vertical levels, and correlation types.
NONLINEAR PROCESSES IN GEOPHYSICS
(2023)