Article
Geosciences, Multidisciplinary
Mohamad El Gharamti, James L. McCreight, Seong Jin Noh, Timothy J. Hoar, Arezoo RafieeiNasab, Benjamin K. Johnson
Summary: Predicting major floods during extreme rainfall events remains a challenge due to rapid flow changes and model errors. This study presents a data assimilation framework to improve flood predictions, using Hurricane Florence as a case study. The framework demonstrates improved streamflow estimates and highlights the importance of adaptive inflation and along-the-stream localization in addressing model errors.
HYDROLOGY AND EARTH SYSTEM SCIENCES
(2021)
Article
Computer Science, Interdisciplinary Applications
L. Minah Yang, Ian Grooms
Summary: By incorporating generative models and patching schemes, the scalability of the constructed analog ensemble optimal interpolation method for data assimilation has been enhanced, resulting in improved assimilation performance on complex dynamical models.
JOURNAL OF COMPUTATIONAL PHYSICS
(2021)
Article
Meteorology & Atmospheric Sciences
Jonathan Poterjoy
Summary: This study explores the use of iterative and hybrid strategies for improving localized particle filters in geophysical models. The experiments show that these strategies have the largest benefits in observation-sparse regimes with non-Gaussian prior errors.
QUARTERLY JOURNAL OF THE ROYAL METEOROLOGICAL SOCIETY
(2022)
Article
Water Resources
Ignacio Martin Santos, Mathew Herrnegger, Hubert Holzmann
Summary: The authors developed a seasonal forecast system for discharge in the Upper Danube basin upstream of Vienna, but found limited skill in the results. The system is not yet sufficient to be incorporated into water resources decision support systems in the region.
JOURNAL OF HYDROLOGY-REGIONAL STUDIES
(2021)
Article
Meteorology & Atmospheric Sciences
Matthias Morzfeld, Daniel Hodyss
Summary: Covariance localization is crucial for successful ensemble data assimilation, especially in global numerical weather prediction. We review and synthesize optimal and adaptive localization methods that are based on sampling error theory and defined by optimality criteria. These methods show similarity in idealized numerical experiments and the most important attribute is damping spurious long-range correlations.
MONTHLY WEATHER REVIEW
(2023)
Article
Meteorology & Atmospheric Sciences
Andrew Walsworth, Jonathan Poterjoy, Elizabeth Satterfield
Summary: In order to obtain accurate state estimates for dynamical models, observation uncertainty needs to be specified accurately. The Desroziers method based on innovation diagnostics is commonly used to estimate observation uncertainty, but it depends greatly on the prescribed background uncertainty. For ensemble data assimilation, inflation and localization are required to address under sampling and these uncertainties come from statistics calculated from ensemble forecasts.
MONTHLY WEATHER REVIEW
(2023)
Article
Meteorology & Atmospheric Sciences
Lili Lei, Zhongrui Wang, Zhe-Min Tan
Summary: This study introduces two integrated hybrid Ensemble Kalman Filter (EnKF) methods that update both the ensemble mean and ensemble perturbations by a hybrid background error covariance in the framework of EnKF, using climatological ensemble perturbations to approximate the static background error covariance. The integrated hybrid EnKFs are superior to traditional hybrid assimilation methods, demonstrating the benefits of updating ensemble perturbations by the hybrid background error covariance.
MONTHLY WEATHER REVIEW
(2021)
Article
Computer Science, Interdisciplinary Applications
J. N. Stroh, Bradford J. Smith, Peter D. Sottile, George Hripcsak, David J. Albers
Summary: Mechanical ventilation is a necessary tool in managing ARDS, but it carries the risk of VILI. This study proposes a hypothesis-driven LVS modeling strategy that achieves robust personalization using a non-physiological model with pre-defined parameters. Model inversion through windowed data assimilation transforms observed waveforms into interpretable parameter values, characterizing the data rather than quantifying physiological processes. Accurate model-based inference on human-ventilator data shows model flexibility and utility across various breath types, including dyssynchronous LVS breaths.
JOURNAL OF BIOMEDICAL INFORMATICS
(2023)
Article
Meteorology & Atmospheric Sciences
Will Boyles, Matthias Katzfuss
Summary: The EnKF is a popular technique for data assimilation in high-dimensional nonlinear state-space models. Our proposed method involves regularizing a sparse Cholesky factor of the inverse covariance matrix and non-zero Cholesky entries, resulting in improved accuracy and robustness to parameter errors.
MONTHLY WEATHER REVIEW
(2021)
Article
Meteorology & Atmospheric Sciences
Eugene Wahl, Eduardo Zorita, Andrew Hoell
Summary: This paper introduces an offline paleo-data assimilation methodology that combines the analog assimilation method (AA) and the Kalman filter (KF) to reconstruct climate fields. The results suggest that the AA method performs well in reconstructing sea level pressure (SLP) geopotential height fields, and the addition of the KF postprocessor enhances the reconstruction skill in certain regions.
JOURNAL OF CLIMATE
(2022)
Article
Computer Science, Interdisciplinary Applications
David Albers, Melike Sirlanci, Matthew Levine, Jan Claassen, Caroline Der Nigoghossian, George Hripcsak
Summary: This paper focuses on predicting physiological processes using data assimilation (DA) with sparse, non-stationary electronic health record data in the intensive care unit. The results show that introducing constrained inference can greatly improve prediction accuracy and robustness, while minimizing data requirements.
JOURNAL OF BIOMEDICAL INFORMATICS
(2023)
Article
Computer Science, Information Systems
Paulo S. G. De Mattos Neto, Paulo Renato A. Firmino, Hugo Siqueira, Yara De Souza Tadano, Thiago Antonini Alves, Joao Fausto L. De Oliveira, Manoel Henrique Da Nobrega Marinho, Francisco Madeiro
Summary: The concentration of PM in the air poses health risks to humans and the environment, prompting the development of systems for monitoring, forecasting, and controlling emissions. The use of forecasting systems based on ANN ensembles, particularly combined with MLP, shows better performance in PM10 and PM2.5 time series forecasting, aiding in addressing air pollution issues.
Article
Environmental Sciences
Xinghong Cheng, Zilong Hao, Zengliang Zang, Zhiquan Liu, Xiangde Xu, Shuisheng Wang, Yuelin Liu, Yiwen Hu, Xiaodan Ma
Summary: A new inversion method combining sensitivity analysis and 3DVAR techniques was developed for linear and nonlinear emission source modeling, with a simulation experiment conducted for a heavy haze case in the Beijing-Tianjin-Hebei region in 2016. Results showed that using a posteriori inversed ES could significantly improve underestimations of SO2 and NO2 during the heavy haze period, with simulation errors decreasing significantly.
ATMOSPHERIC CHEMISTRY AND PHYSICS
(2021)
Article
Meteorology & Atmospheric Sciences
Bo Huang, Xuguang Wang, Daryl T. Kleist, Ting Lei
Summary: The scale-dependent localization method improves global forecasts up to 5 days, with the SDL-Cross experiment showing more accurate tropical storm-track forecasts in the short term compared to the operationally tuned level-dependent scale-invariant localization. Tighter horizontal localization generally improves global forecasts below 100 hPa in the W3 SDL experiments, but degrades forecasts above 50 hPa.
MONTHLY WEATHER REVIEW
(2021)
Article
Engineering, Civil
Amir Mazrooei, A. Sankarasubramanian, Andrew W. Wood
Summary: This study introduces a novel VAR DA method which uses a basin-wide scaling factor to update the soil moisture conditions of the VIC model, resulting in improved hydrologic predictions.
JOURNAL OF HYDROLOGY
(2021)
Article
Meteorology & Atmospheric Sciences
Luyu Sun, Stephen G. Penny, Matthew Harrison
Summary: This study investigates the impact of an augmented-state Lagrangian data assimilation method using the local ensemble Kalman transform filter on a regional ocean data assimilation system. The results show that this method can provide more accurate estimates of ocean variables and perform better in estimating surface currents under special weather conditions.
MONTHLY WEATHER REVIEW
(2022)
Article
Meteorology & Atmospheric Sciences
Chih-Chien Chang, Shu-Chih Yang, Stephen G. Penny
Summary: A newly developed regional hybrid gain data assimilation system (WRF-HGDA) using the Weather Research and Forecasting model (WRF) is presented. It combines the ensemble-based Kalman filter (WRF-LETKF) with the variational analysis system (WRF-3DVAR) by utilizing their gain matrices. The performance of WRF-HGDA is evaluated using observations from GNSS radio occultation (RO) experiments and the results show that the variational correction can improve the analysis of WRF-LETKF, with the equal-weighted WRF-HGDA performing better than its component DA systems in terms of moisture and wind fields when assimilating conventional observations. Assimilating additional RO data further enhances the performance of WRF-LETKF and WRF-HGDA.
Article
Meteorology & Atmospheric Sciences
S. G. Penny, T. A. Smith, T-C Chen, J. A. Platt, H-Y Lin, M. Goodliff, H. D. Abarbanel
Summary: This article introduces the integration of data assimilation (DA) with machine learning for entirely data-driven online state estimation. Recurrent neural networks (RNNs) are used as pretrained surrogate models to replace key components in numerical weather prediction (NWP) and can be initialized using DA methods to estimate the state of a system.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2022)
Article
Geosciences, Multidisciplinary
Gregory J. Hakim, Chris Snyder, Stephen G. Penny, Matthew Newman
Summary: The study shows that a Kalman filter with a linear emulator can efficiently assimilate strongly coupled data, improving the forecast skill of ocean analyses. The experiment results demonstrate that daily assimilation of observations using a linear inverse model reduces the analysis errors of sea-surface temperature by 20% compared to a control experiment. Additionally, it enhances the forecast skill for at least 50 days. However, the assimilation of coupled data leads to an increase in forecast errors of extratropical Northern Hemisphere 2 m air temperature.
GEOPHYSICAL RESEARCH LETTERS
(2022)
Article
Meteorology & Atmospheric Sciences
Jonathan Demaeyer, Stephen G. Penny, Stephane Vannitsem
Summary: This study presents a method utilizing eigenfunctions of the Koopman or Perron-Frobenius operators to construct reliable ensemble forecasts, demonstrating that projecting initial conditions onto a subset characterized by fast-decaying oscillations can produce highly reliable forecasts across various lead times.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2022)
Article
Fisheries
M. F. Cronin, S. Swart, C. A. Marandino, C. Anderson, P. Browne, S. Chen, W. R. Joubert, U. Schuster, R. Venkatesan, C. Addey, O. Alves, F. Ardhuin, S. Battle, M. A. Bourassa, Z. Chen, M. Chory, C. Clayson, R. B. de Souza, M. du Plessis, M. Edmondson, J. B. Edson, S. T. Gille, J. Hermes, V Hormann, S. A. Josey, M. Kurz, T. Lee, F. Maicu, E. H. Moustahfid, S-A Nicholson, E. S. Nyadjro, J. Palter, R. G. Patterson, S. G. Penny, L. P. Pezzi, N. Pinardi, J. E. J. Reeves Eyre, N. Rome, A. C. Subramanian, C. Stienbarger, T. Steinhoff, A. J. Sutton, H. Tomita, S. M. Wills, C. Wilson, L. Yu
Summary: The Observing Air-Sea Interactions Strategy (OASIS) is a program that aims to improve Earth system forecasts, CO2 uptake assessments, and provide ocean information for decision makers. It focuses on creating a global network of mobile air-sea observing platforms, a satellite network optimized for measuring air-sea fluxes, and improving the representation of air-sea coupling in Earth system models. The program consists of various activities such as network design, model improvement, partnership building, and best practices experiments.
ICES JOURNAL OF MARINE SCIENCE
(2023)
Article
Meteorology & Atmospheric Sciences
Junjie Dong, Luyu Sun, James A. Carton, Stephen G. Penny
Summary: This study extends previous work by Sun and Penny and Sun et al. to improve the analysis of the ocean by including path information from surface drifters using an augmented-state Lagrangian data assimilation. The study focuses on the Gulf of Mexico during Hurricane Isaac in 2012 and uses a regional ocean model to quantify improvements in sea surface velocity, temperature, and height analysis. By assimilating drifter positions and vertical profiles, the study shows significant improvements in analyzing the ocean state under hurricane conditions, which can also be applicable to other tropical oceans.
MONTHLY WEATHER REVIEW
(2023)
Article
Meteorology & Atmospheric Sciences
Isabel A. Houghton, Stephen G. Penny, Christie Hegermiller, Moriah Cesaretti, Camille Teicheira, Pieter B. Smit
Summary: An ensemble-based method for wave data assimilation using significant wave height observations is implemented and skillful analysis fields are generated, resulting in reduced forecast errors up to 2.5 days. The Local Ensemble Transform Kalman Filter (LETKF) method provides more physically realistic model state updates and better reflects the underlying sea state dynamics and uncertainty compared to optimal interpolation methods. LETKF shows advantages over optimal interpolation in skill assessment far from observations and specific storm events. This advancement is valuable in improving sea state predictions and enabling future coupled data assimilation and utilization of global surface observations.
Editorial Material
Multidisciplinary Sciences
Yan Li, Shan Sang, Safa Mote, Jorge Rivas, Eugenia Kalnay
Summary: With the recognition of coupled human and natural systems (CHANS), modeling CHANS with two-way feedbacks has become a crucial tool for achieving sustainability. This paper discusses the challenges in CHANS modeling and the opportunities to advance its science and applications in promoting the sustainability of CHANS.
NATIONAL SCIENCE REVIEW
(2023)
Article
Mathematics, Applied
Jason A. Platt, Stephen G. Penny, Timothy A. Smith, Tse-Chun Chen, Henry D. I. Abarbanel
Summary: Drawing on ergodic theory, this paper introduces a novel training method for machine learning based forecasting methods for chaotic dynamical systems. The method enforces dynamical invariants in the systems, such as the Lyapunov exponent spectrum and the fractal dimension, which enables longer and more stable forecasts when operating with limited data. The technique is demonstrated using reservoir computing, a specific kind of recurrent neural network, and the effectiveness is verified with typical test cases.
Review
Geosciences, Multidisciplinary
Eugenia Kalnay, Travis Sluka, Takuma Yoshida, Cheng Da, Safa Mote
Summary: Different coupled data assimilation strategies were assessed using a range of coupled models, and the analysis accuracy was compared. The strongly coupled ensemble Kalman filter (EnKF) method with a short assimilation window showed comparable accuracy to the long assimilation window 4D-Var method. The strongly coupled approach outperformed weakly coupled and uncoupled approaches for both EnKF and 4D-Var, and produced more accurate analyses compared to other coupled data assimilation approaches.
NONLINEAR PROCESSES IN GEOPHYSICS
(2023)
Article
Meteorology & Atmospheric Sciences
Timothy A. Smith, Stephen G. Penny, Jason A. Platt, Tse-Chun Chen
Summary: The computational cost of traditional numerical weather and climate models has led to the development of machine learning-based emulators. However, subsampling the training data in terms of temporal resolution can negatively affect the quality of the emulator's predictions, leading to increased bias at small spatial scales. Different machine learning architectures show different levels of sensitivity to this subsampling.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2023)
Article
Oceanography
Eric Hackert, S. Akella, L. Ren, K. Nakada, J. A. Carton, A. Molod
Summary: This study evaluates the TAO/TRITON array using data denial assimilation experiments and assesses its impact on El Nino/Southern Oscillation (ENSO) predictions. The results show that assimilating TAO/TRITON data generally improves comparisons with other in situ observations, especially for temperature. It is found that TAO/TRITON data can deepen the mixed layer depth, amplify the El Nino downwelling signal, and improve the amplitude of the ENSO signal.
JOURNAL OF GEOPHYSICAL RESEARCH-OCEANS
(2023)
Article
Geosciences, Multidisciplinary
Chu-Chun Chang, Eugenia Kalnay
Summary: This study examined the feasibility of the correlation cutoff method as an alternative spatial localization and demonstrated that it can deliver comparable analysis to the traditional localization more efficiently, especially under a more complicated model with reduced ensemble and observation sizes.
NONLINEAR PROCESSES IN GEOPHYSICS
(2022)
Article
Geosciences, Multidisciplinary
Zhiqiang Liu, Ning Zeng, Yun Liu, Eugenia Kalnay, Ghassem Asrar, Bo Wu, Qixiang Cai, Di Liu, Pengfei Han
Summary: Atmospheric inversion techniques have advanced the understanding of carbon sources and sinks, although most studies have focused on fluxes rather than CO2 concentrations. This study applies a constrained ensemble Kalman filter approach to ensure the conservation of global CO2 mass, resulting in improved accuracy in tracking CO2 concentrations and predicting seasonal fluxes.
GEOSCIENTIFIC MODEL DEVELOPMENT
(2022)