Article
Meteorology & Atmospheric Sciences
Ji-Won Lee, Ki-Hong Min, Kyo-Sun Sunny Lim
Summary: Radar data assimilation into numerical prediction models using 3DVAR and hybrid methods significantly improves precipitation forecasts, with the hybrid method showing better accuracy. The key factor influencing precipitation formation is the change in water vapor amount.
ATMOSPHERIC RESEARCH
(2022)
Article
Meteorology & Atmospheric Sciences
Joshua Mccurry, Jonathan Poterjoy, Kent Knopfmeier, Louis Wicker
Summary: To address the uncertainty of moist convection, a novel Bayesian data assimilation method based on particle filtering is evaluated. Assimilating with particle filtering produces posterior variables that are more consistent with model climatology compared to ensemble Kalman filtering, reducing data assimilation bias. These differences have significant impacts on the dynamic evolution of convective systems and forecast verification scores, with implications for the selection of physical parameterization schemes and parameter estimation in data assimilation frameworks.
MONTHLY WEATHER REVIEW
(2023)
Article
Meteorology & Atmospheric Sciences
Eun-Gyeong Yang, Hyun Mee Kim
Summary: This study investigated the performance of three data assimilation methods based on the WRF model over East Asia. The hybrid E3DVAR method outperformed 3DVAR and EnKF for both January and July seasons. Adjusting background error covariance can improve forecast accuracy, and each method has different strengths in different seasons.
ATMOSPHERIC RESEARCH
(2021)
Article
Geosciences, Multidisciplinary
Tijana Janjic, Yuefei Zeng
Summary: The study introduces a new concept of combining weak constraints on mass conservation and non-negativity in convective-scale data assimilation. Results show that both weak constraints successfully improve the mass conservation property in analyses and reduce the biased increase in integrated mass-flux divergence and vorticity. The combination of both constraints achieves the least biased increase and the best forecasts.
GEOPHYSICAL RESEARCH LETTERS
(2021)
Article
Nuclear Science & Technology
Stefano Riva, Carolina Introini, Stefano Lorenzi, Antonio Cammi
Summary: Hybrid Data Assimilation (HDA) methods integrate Model Order Reduction (MOR) techniques into a Data Assimilation (DA) framework to combine mathematical models and experimental data. This paper focuses on the numerical formulation of HDA techniques and the impact of noisy data using the Generalised Empirical Interpolation Method (GEIM) and the Parameterised-Background Data-Weak (PBDW) formulation. Results show the effectiveness of both algorithms in reconstructing the system state and assessing the effect of noise on the available data.
ANNALS OF NUCLEAR ENERGY
(2023)
Article
Nuclear Science & Technology
Stefano Riva, Carolina Introini, Stefano Lorenzi, Antonio Cammi
Summary: Hybrid Data Assimilation (HDA) methods combine mathematical models and experimental observations by integrating model order reduction techniques into a data assimilation framework, reducing solution time while maintaining accuracy. These methods estimate the state of a system using measurements from various fields, but when direct measurements are not possible, efforts can be made to extract information from indirect measurements.
ANNALS OF NUCLEAR ENERGY
(2023)
Article
Mathematics, Interdisciplinary Applications
Dongrui Shao, Junyu Chu, Luonan Chen, Huanfei Ma
Summary: Data assimilation is crucial for both data driven and model driven research. The Kalman filter, a widely used data assimilation framework, has traditionally relied on theoretical models. However, recent efforts have aimed to develop model-free Kalman filters that solely rely on data. In this study, we propose a hybrid model framework that combines delay embedding theory and machine learning to bridge the gap between exact model-based and totally model-free methods. This hybrid approach is more flexible in application and has been validated using benchmark systems and real-world problems.
CHAOS SOLITONS & FRACTALS
(2023)
Article
Environmental Sciences
Peng Liu, Yi Yang, Anwei Lai, Yunheng Wang, Alexandre O. Fierro, Jidong Gao, Chenghai Wang
Summary: The study utilized a dual-resolution, hybrid 3DEnVAR method to assimilate radar data and pseudo-water vapor observations, improving short-term severe weather forecasts with the WRF model. Different coefficients were tested, with a static coefficient of 0.4 and an ensemble coefficient of 0.6 found to yield the best forecast skill in this event.
Article
Meteorology & Atmospheric Sciences
Jeffrey S. Whitaker, Anna Shlyaeva, Stephen G. Penny
Summary: This study compares two methods for incorporating a time-invariant, high-rank covariance estimate in an ensemble-based data assimilation system: the hybrid-covariance approach and the hybrid-gain approach. The results show that the simpler and less expensive hybrid-gain approach can achieve similar performance if the incremental normal-mode balance constraint applied to the ensemble-part of the hybrid-covariance update is turned off.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2022)
Article
Meteorology & Atmospheric Sciences
Yuxuan Feng, Tijana Janji, Yuefei Zeng, Axel Seifert, Jinzhong Min
Summary: Model uncertainty in cloud microphysics is a significant source of model error for convective clouds and precipitation. By introducing samples for model microphysical uncertainty into convective-scale ensemble data assimilation and combining them with large-scale additive noise, the study found that simulations with a two-moment scheme trigger more convection and produce stronger signals in the melting layer. The combination approach significantly improves short-term ensemble forecasts of radar reflectivity and hourly precipitation.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2021)
Article
Meteorology & Atmospheric Sciences
Mayeul Destouches, Thibaut Montmerle, Yann Michel, Jean-Francois Caron
Summary: This article presents the addition of hydrometeor fields as control variables to the cloud-resolving model AROME-France. Even without directly assimilating hydrometeor observations, analysis increments of hydrometeors can be produced via covariance calculations. Forecast-analysis experiments were performed over a three-month period, comparing different configurations. The experiments with hydrometeor variables show a positive impact on cloud cover and precipitation forecasts, reducing the spin-up period.
QUARTERLY JOURNAL OF THE ROYAL METEOROLOGICAL SOCIETY
(2023)
Article
Meteorology & Atmospheric Sciences
Shibo Gao, Haiqiu Yu, Chuanyou Ren, Limin Liu, Jinzhong Min
Summary: The E3DA system, utilizing three-dimensional ensemble-variational data assimilation, significantly improved quantitative convective weather forecasting skills and spatial distribution compared to 3DVar. E3DA reduced the root-mean-square error of radial velocity, enhanced the forecast accuracy of wind, temperature, and water vapor mixing ratio, and outperformed 3DVar in adjusting vertical velocity, temperature, and humidity in response to added reflectivity.
ADVANCES IN ATMOSPHERIC SCIENCES
(2021)
Article
Nuclear Science & Technology
Carolina Introini, Stefano Riva, Stefano Lorenzi, Simone Cavalleri, Antonio Cammi
Summary: This study proposes a novel approach to estimate the full state of a system by using a two-step method: determining system parameters based on measurements and estimating the full state using reduced order modelling techniques. This method can be applied to engineering systems where not all variables of interest are measurable.
ANNALS OF NUCLEAR ENERGY
(2023)
Article
Environmental Sciences
Chi Vuong Nguyen, Lionel Soulhac
Summary: This study compares three data assimilation methods for evaluating air quality at the local urban scale and finds that all three methods can improve air quality estimates, with similar performances. It demonstrates that data assimilation is a promising tool for enhancing air quality simulations at the urban scale.
ATMOSPHERIC ENVIRONMENT
(2021)
Article
Meteorology & Atmospheric Sciences
Xian Xiao, Juanzhen Sun, Xiushu Qie, Zhuming Ying, Lei Ji, Mingxuan Chen, Lina Zhang
Summary: The study proposes and implements a proof-of-concept method for assimilating total lightning observations in the 4DVAR framework. By assimilating both radar and lightning data simultaneously, improved dynamical consistency, enhanced updraft and latent heat, and improved moisture distributions are achieved. The combined data assimilation scheme proves to be robust to variations in vertical velocity profiles, radii of horizontal interpolation, binning time intervals, and relationships used to estimate maximum vertical velocity from lightning flash rates.
MONTHLY WEATHER REVIEW
(2021)
Article
Meteorology & Atmospheric Sciences
Ieva Dauzickaite, Amos S. Lawless, Jennifer A. Scott, Peter Jan van Leeuwen
Summary: This paper discusses the importance of parallel computing in data assimilation and proposes a new strategy using randomized singular value decomposition to accelerate calculations and maintain parallelism in the time domain.
QUARTERLY JOURNAL OF THE ROYAL METEOROLOGICAL SOCIETY
(2021)
Article
Meteorology & Atmospheric Sciences
S. Pathiraja, P. J. van Leeuwen
Summary: Model uncertainty quantification is crucial for effective data assimilation. This study presents a methodology for estimating the statistics of sub-grid scale processes using partial observations of the coarse scale process. The method involves minimizing the conditional sum of squared deviations to estimate model error realizations and obtaining a conditional probability distribution of additive model errors for non-Gaussian error structures.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2022)
Article
Meteorology & Atmospheric Sciences
Dandan Tao, Peter Jan van Leeuwen, Michael Bell, Yue Ying
Summary: Through the study of Hurricane Patricia, it was found that assimilating observational data significantly improves the prediction of rapid intensification. Analysis of observation impacts showed that deep-layer dropsonde observations have the greatest effect on the circulation of the entire troposphere. Verification of ensemble forecasts revealed that errors in the early vortex structure can lead to inaccuracies in later stage predictions.
JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES
(2022)
Article
Oceanography
Kyle Shackelford, Charlotte A. DeMott, Peter Jan van Leeuwen, Elizabeth Thompson, Samson Hagos
Summary: The study explores the impact of precipitation on the ocean by investigating the formation and characteristics of rain layers (RLs) in the equatorial Indian Ocean. The findings suggest that RLs cool the ocean surface and enhance sea surface temperature (SST) gradients, which can promote atmospheric convection. This highlights the importance of further research on RL feedbacks to the atmosphere.
JOURNAL OF GEOPHYSICAL RESEARCH-OCEANS
(2022)
Article
Environmental Sciences
Concetta Di Mauro, Renaud Hostache, Patrick Matgen, Ramona Pelich, Marco Chini, Peter Jan van Leeuwen, Nancy Nichols, Gunter Bloeschl
Summary: The study proposes an innovative approach based on a tempered particle filter (TPF) to assimilate probabilistic flood maps derived from synthetic aperture radar data into a flood forecasting model, showing significant improvement in model forecasting accuracy with longer-lasting benefits.
WATER RESOURCES RESEARCH
(2022)
Article
Geosciences, Multidisciplinary
C. Kevin Yang, J. Christine Chiu, Alexander Marshak, Graham Feingold, Tamas Varnai, Guoyong Wen, Takanobu Yamaguchi, Peter Jan van Leeuwen
Summary: A lack of satellite-based aerosol retrievals near low-topped clouds is addressed by developing a Convolutional Neural Network to retrieve aerosol optical depth (AOD) in cloud-free regions. The method takes into account cloud radiative effects on reflectance observations and achieves a horizontal resolution of 100-500 m. The retrieval uncertainty is 0.01 + 5%AOD, with a mean bias of approximately -2%. In satellite observations, aerosol hygroscopic growth near cloud edges leads to a 100% increase in AOD, resulting in a 55% overall increase in clear-sky aerosol direct radiative effect.
GEOPHYSICAL RESEARCH LETTERS
(2022)
Article
Mathematics, Interdisciplinary Applications
Oana Lang, Peter Jan van Leeuwen, Dan Crisan, Roland Potthast
Summary: In this work, a tempering-based adaptive particle filter is used to infer from a partially observed stochastic rotating shallow water (SRSW) model. The methodology validates the applicability of tempering and sample regeneration to high-dimensional models in geophysical fluid dynamics problems. The efficiency of the particle filter is studied in the SRSW model configuration simulating the atmospheric Jetstream.
FRONTIERS IN APPLIED MATHEMATICS AND STATISTICS
(2022)
Article
Meteorology & Atmospheric Sciences
Daniel Ayers, Jack Lau, Javier Amezcua, Alberto Carrassi, Varun Ojha
Summary: Prediction errors grow faster in some situations than in others in chaotic dynamical systems. Real-time knowledge about error growth can enable strategies to adjust modelling and forecasting infrastructure to increase accuracy and reduce computation time. In this feasibility study, supervised machine learning is used to estimate local Lyapunov exponents in place of the classical method. The machine learning algorithms accurately predict stable and unstable local Lyapunov exponents, but only somewhat accurately predict neutral exponents.
QUARTERLY JOURNAL OF THE ROYAL METEOROLOGICAL SOCIETY
(2023)
Article
Geography, Physical
Nicholas Williams, Nicholas Byrne, Daniel Feltham, Peter Jan Van Leeuwen, Ross Bannister, David Schroeder, Andrew Ridout, Lars Nerger
Summary: In the past decade, groundbreaking satellite observations have provided a better understanding of the Arctic sea ice system. This study presents the first results of a new sea ice data assimilation system that combines various observations to estimate sea ice thickness and improve modeled ice distribution. Assimilating sub-grid-scale thickness distribution has a significant impact on the performance of the sea ice model.
Article
Meteorology & Atmospheric Sciences
Javier Amezcua, Haonan Ren, Peter Jan Van Leeuwen
Summary: Numerical weather prediction systems have model errors due to missing and simplified physical processes, as well as limited model resolution. Estimating model error parameters is a challenging problem, even in low-dimensional systems. By focusing on linear and nonlinear low-dimensional systems and using state augmentation in ensemble Kalman smoothers, we can find practical solutions to this problem.
TELLUS SERIES A-DYNAMIC METEOROLOGY AND OCEANOGRAPHY
(2023)
Article
Geosciences, Multidisciplinary
Joshua Chun Kwang Lee, Javier Amezcua, Ross Noel Bannister
Summary: Hybrid ensemble-variational data assimilation methods are widely used in the mid-latitudinal context, but their benefits in the tropical context have been less explored. This study introduces and improves the hybrid ensemble-variational DA method in a tropical configuration of a simplified fluid dynamics model. The algorithm includes localization and weighting parameters, and an ensemble system is designed to generate ensemble perturbations. Sensitivity tests using observing system simulation experiments show that the hybrid method performs well with certain weighting configurations.
GEOSCIENTIFIC MODEL DEVELOPMENT
(2022)
Article
Geosciences, Multidisciplinary
Sagar K. Tamang, Ardeshir Ebtehaj, Peter Jan van Leeuwen, Gilad Lerman, Efi Foufoula-Georgiou
Summary: This paper presents the results of ensemble Riemannian data assimilation for relatively high-dimensional nonlinear dynamical systems, focusing on the Lorenz-96 model and the quasi-geostrophic model. The method infers the analysis state from a joint distribution, which effectively handles systematic biases. Comparisons with classic implementations of particle filter and stochastic ensemble Kalman filter show that this method can improve the predictability of dynamical systems with the same ensemble size.
NONLINEAR PROCESSES IN GEOPHYSICS
(2022)
Article
Physics, Multidisciplinary
Nachiketa Chakraborty, Peter Jan van Leeuwen
Summary: Measuring time lags between light curves at different wavelengths is crucial for studying multiwavelength variability in astronomy. Using a discrete mutual information function (DMIF) can accurately identify lag components and reveal lag features in nonlinear relationships.
PHYSICAL REVIEW RESEARCH
(2022)
Article
Geosciences, Multidisciplinary
Sagar K. Tamang, Ardeshir Ebtehaj, Peter J. van Leeuwen, Dongmian Zou, Gilad Lerman
Summary: In this paper, a novel ensemble data assimilation paradigm is introduced on a Riemannian manifold equipped with the Wasserstein metric, which can capture differences in probability distribution shapes and penalize non-Gaussian geophysical biases in state space. The new approach is applied to dissipative and chaotic evolutionary dynamics, highlighting potential advantages and limitations compared to classic ensemble data assimilation methods under systematic errors.
NONLINEAR PROCESSES IN GEOPHYSICS
(2021)