{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,10]],"date-time":"2026-06-10T14:29:25Z","timestamp":1781101765189,"version":"3.54.1"},"reference-count":44,"publisher":"MDPI AG","issue":"2","license":[{"start":{"date-parts":[[2022,2,12]],"date-time":"2022-02-12T00:00:00Z","timestamp":1644624000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Entropy"],"abstract":"<jats:p>The initial field has a crucial influence on numerical weather prediction (NWP). Data assimilation (DA) is a reliable method to obtain the initial field of the forecast model. At the same time, data are the carriers of information. Observational data are a concrete representation of information. DA is also the process of sorting observation data, during which entropy gradually decreases. Four-dimensional variational assimilation (4D-Var) is the most popular approach. However, due to the complexity of the physical model, the tangent linear and adjoint models, and other processes, the realization of a 4D-Var system is complicated, and the computational efficiency is expensive. Machine learning (ML) is a method of gaining simulation results by training a large amount of data. It achieves remarkable success in various applications, and operational NWP and DA are no exception. In this work, we synthesize insights and techniques from previous studies to design a pure data-driven 4D-Var implementation framework named ML-4DVAR based on the bilinear neural network (BNN). The framework replaces the traditional physical model with the BNN model for prediction. Moreover, it directly makes use of the ML model obtained from the simulation data to implement the primary process of 4D-Var, including the realization of the short-term forecast process and the tangent linear and adjoint models. We test a strong-constraint 4D-Var system with the Lorenz-96 model, and we compared the traditional 4D-Var system with ML-4DVAR. The experimental results demonstrate that the ML-4DVAR framework can achieve better assimilation results and significantly improve computational efficiency.<\/jats:p>","DOI":"10.3390\/e24020264","type":"journal-article","created":{"date-parts":[[2022,2,13]],"date-time":"2022-02-13T21:08:43Z","timestamp":1644786523000},"page":"264","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":26,"title":["A Framework for Four-Dimensional Variational Data Assimilation Based on Machine Learning"],"prefix":"10.3390","volume":"24","author":[{"given":"Renze","family":"Dong","sequence":"first","affiliation":[{"name":"College of Meteorology and Oceanography, National University of Defense Technology, Changsha 410000, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hongze","family":"Leng","sequence":"additional","affiliation":[{"name":"College of Meteorology and Oceanography, National University of Defense Technology, Changsha 410000, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8412-2472","authenticated-orcid":false,"given":"Juan","family":"Zhao","sequence":"additional","affiliation":[{"name":"College of Meteorology and Oceanography, National University of Defense Technology, Changsha 410000, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Junqiang","family":"Song","sequence":"additional","affiliation":[{"name":"College of Meteorology and Oceanography, National University of Defense Technology, Changsha 410000, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shutian","family":"Liang","sequence":"additional","affiliation":[{"name":"School of Resources and Environmental Engineering, Hefei University of Technology, Hefei 230000, China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1968","published-online":{"date-parts":[[2022,2,12]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"47","DOI":"10.1038\/nature14956","article-title":"The quiet revolution of numerical weather prediction","volume":"525","author":"Bauer","year":"2015","journal-title":"Nature"},{"key":"ref_2","unstructured":"Haltiner, G.J., and Williams, R.T. (1980). Numerical Prediction and Dynamic Meteorology, Wiley. Technical Report."},{"key":"ref_3","first-page":"1","article-title":"Das problem der wettervorhersage, betrachtet vom standpunkte der mechanik und der physik, translation by y. mintz: The problem of weather forecasting as a problem in mechanics and physics. los angeles, 1954","volume":"21","author":"Bjerknes","year":"1904","journal-title":"Meteor. Zeit"},{"key":"ref_4","doi-asserted-by":"crossref","unstructured":"Kalnay, E. (2003). Atmospheric Modeling, Data Assimilation and Predictability, Cambridge University Press.","DOI":"10.1017\/CBO9780511802270"},{"key":"ref_5","doi-asserted-by":"crossref","first-page":"1143","DOI":"10.1002\/qj.49712656415","article-title":"The ECMWF operational implementation of four-dimensional variational assimilation. I: Experimental results with simplified physics","volume":"126","author":"Rabier","year":"2000","journal-title":"Q. J. R. Meteorol. Soc."},{"key":"ref_6","doi-asserted-by":"crossref","first-page":"2089","DOI":"10.1175\/1520-0493(2001)129<2089:IOTDFA>2.0.CO;2","article-title":"Impact of the digital filter as a weak constraint in the preoperational 4DVAR assimilation system of M\u00e9t\u00e9o-France","volume":"129","author":"Gauthier","year":"2001","journal-title":"Mon. Weather Rev."},{"key":"ref_7","first-page":"347","article-title":"The Met Office global four-dimensional variational data assimilation scheme","volume":"133","author":"Rawlins","year":"2007","journal-title":"Q. J. R. Meteorol. Soc. A J. Atmos. Sci. Appl. Meteorol. Phys. Oceanogr."},{"key":"ref_8","doi-asserted-by":"crossref","first-page":"2339","DOI":"10.1175\/MWR3394.1","article-title":"Extension of 3DVAR to 4DVAR: Implementation of 4DVAR at the Meteorological Service of Canada","volume":"135","author":"Gauthier","year":"2007","journal-title":"Mon. Weather Rev."},{"key":"ref_9","first-page":"3465","article-title":"A pre-operational variational data assimilation system for a non-hydrostatic model at the Japan Meteorological Agency: Formulation and preliminary results","volume":"131","author":"Honda","year":"2005","journal-title":"Q. J. R. Meteorol. Soc. A J. Atmos. Sci. Appl. Meteorol. Phys. Oceanogr."},{"key":"ref_10","doi-asserted-by":"crossref","first-page":"675","DOI":"10.1007\/s13351-020-9847-6","article-title":"Research and Operational Development of Numerical Weather Prediction in China","volume":"34","author":"Shen","year":"2020","journal-title":"J. Meteorol. Res."},{"key":"ref_11","first-page":"1367","article-title":"A strategy for operational implementation of 4D-Var, using an incremental approach","volume":"120","author":"Courtier","year":"1994","journal-title":"Q. J. R. Meteorol. Soc."},{"key":"ref_12","first-page":"6","article-title":"Data Assimilation: A Key Technology for NWP\u2014Technical Review of Data Assimilation in ECMWF","volume":"3","author":"Jiandong","year":"2013","journal-title":"Adv. Meteorol. Sci. Technol."},{"key":"ref_13","doi-asserted-by":"crossref","first-page":"607","DOI":"10.1002\/qj.2982","article-title":"A review of operational methods of variational and ensemble-variational data assimilation","volume":"143","author":"Bannister","year":"2017","journal-title":"Q. J. R. Meteorol. Soc."},{"key":"ref_14","doi-asserted-by":"crossref","first-page":"4489","DOI":"10.1175\/MWR-D-15-0440.1","article-title":"Review of the ensemble Kalman filter for atmospheric data assimilation","volume":"144","author":"Houtekamer","year":"2016","journal-title":"Mon. Weather Rev."},{"key":"ref_15","doi-asserted-by":"crossref","first-page":"20200089","DOI":"10.1098\/rsta.2020.0089","article-title":"Learning earth system models from observations: Machine learning or data assimilation?","volume":"379","author":"Geer","year":"2021","journal-title":"Philos. Trans. R. Soc. A"},{"key":"ref_16","unstructured":"Dueben, P., Modigliani, U., Geer, A., Siemen, S., Pappenberger, F., Bauer, P., Brown, A., Palkovi\u010d, M., Raoult, B., and Wedi, N. (2021). Technical Memo, European Centre for Medium-Range Weather Forecasts."},{"key":"ref_17","doi-asserted-by":"crossref","first-page":"20200097","DOI":"10.1098\/rsta.2020.0097","article-title":"Can deep learning beat numerical weather prediction?","volume":"379","author":"Schultz","year":"2021","journal-title":"Philos. Trans. R. Soc. A"},{"key":"ref_18","doi-asserted-by":"crossref","first-page":"195","DOI":"10.1038\/s41586-019-0912-1","article-title":"Deep learning and process understanding for data-driven Earth system science","volume":"566","author":"Reichstein","year":"2019","journal-title":"Nature"},{"key":"ref_19","doi-asserted-by":"crossref","unstructured":"Camps-Valls, G., Reichstein, M., Xiaoxiang, Z., and Tuia, D. (2021). Deep Learning for Earth Sciences, Wiley.","DOI":"10.1002\/9781119646181"},{"key":"ref_20","doi-asserted-by":"crossref","first-page":"e2020MS002109","DOI":"10.1029\/2020MS002109","article-title":"Improving data-driven global weather prediction using deep convolutional neural networks on a cubed sphere","volume":"12","author":"Weyn","year":"2020","journal-title":"J. Adv. Model. Earth Syst."},{"key":"ref_21","doi-asserted-by":"crossref","first-page":"6289","DOI":"10.1029\/2018GL078510","article-title":"Prognostic validation of a neural network unified physics parameterization","volume":"45","author":"Brenowitz","year":"2018","journal-title":"Geophys. Res. Lett."},{"key":"ref_22","doi-asserted-by":"crossref","first-page":"9684","DOI":"10.1073\/pnas.1810286115","article-title":"Deep learning to represent subgrid processes in climate models","volume":"115","author":"Rasp","year":"2018","journal-title":"Proc. Natl. Acad. Sci. USA"},{"key":"ref_23","doi-asserted-by":"crossref","first-page":"e2020GL091363","DOI":"10.1029\/2020GL091363","article-title":"Use of neural networks for stable, accurate and physically consistent parameterization of subgrid atmospheric processes with good performance at reduced precision","volume":"48","author":"Yuval","year":"2021","journal-title":"Geophys. Res. Lett."},{"key":"ref_24","doi-asserted-by":"crossref","first-page":"e2021MS002609","DOI":"10.1029\/2021MS002609","article-title":"Improved weather forecasting using neural network emulation for radiation parameterization","volume":"13","author":"Song","year":"2021","journal-title":"J. Adv. Model. Earth Syst."},{"key":"ref_25","unstructured":"Krasnopolsky, V. (2020). Using machine learning for model physics: An overview. arXiv."},{"key":"ref_26","doi-asserted-by":"crossref","first-page":"e2021MS002477","DOI":"10.1029\/2021MS002477","article-title":"Machine learning emulation of gravity wave drag in numerical weather forecasting","volume":"13","author":"Chantry","year":"2021","journal-title":"J. Adv. Model. Earth Syst."},{"key":"ref_27","doi-asserted-by":"crossref","first-page":"e2020MS002232","DOI":"10.1029\/2020MS002232","article-title":"Machine learning for model error inference and correction","volume":"12","author":"Bonavita","year":"2020","journal-title":"J. Adv. Model. Earth Syst."},{"key":"ref_28","doi-asserted-by":"crossref","first-page":"e2021MS002521","DOI":"10.1029\/2021MS002521","article-title":"Building Tangent-Linear and Adjoint Models for Data Assimilation With Neural Networks","volume":"13","author":"Hatfield","year":"2021","journal-title":"J. Adv. Model. Earth Syst."},{"key":"ref_29","doi-asserted-by":"crossref","first-page":"e2021MS002554","DOI":"10.1029\/2021MS002554","article-title":"Deep Emulators for Differentiation, Forecasting and Parametrization in Earth Science Simulators","volume":"13","author":"Nonnenmacher","year":"2021","journal-title":"J. Adv. Model. Earth Syst."},{"key":"ref_30","doi-asserted-by":"crossref","first-page":"1949","DOI":"10.1134\/S000511791911002X","article-title":"Complete statistical theory of learning","volume":"80","author":"Vapnik","year":"2019","journal-title":"Autom. Remote Control"},{"key":"ref_31","unstructured":"Nielsen, M. (2022, January 17). A Visual Proof That Neural Nets Can Compute Any Function. Available online: http:\/\/neuralnetworksanddeeplearning.com\/chap4.html."},{"key":"ref_32","doi-asserted-by":"crossref","first-page":"436","DOI":"10.1038\/nature14539","article-title":"Deep learning","volume":"521","author":"LeCun","year":"2015","journal-title":"Nature"},{"key":"ref_33","doi-asserted-by":"crossref","unstructured":"Fablet, R., Ouala, S., and Herzet, C. (2018, January 3\u20137). Bilinear residual neural network for the identification and forecasting of geophysical dynamics. Proceedings of the 2018 IEEE 26th European Signal Processing Conference (Eusipco), Rome, Italy.","DOI":"10.23919\/EUSIPCO.2018.8553492"},{"key":"ref_34","doi-asserted-by":"crossref","first-page":"5742","DOI":"10.1029\/2018GL078202","article-title":"Could machine learning break the convection parameterization deadlock?","volume":"45","author":"Gentine","year":"2018","journal-title":"Geophys. Res. Lett."},{"key":"ref_35","doi-asserted-by":"crossref","unstructured":"Lewis, J.M., Lakshmivarahan, S., and Dhall, S. (2006). Dynamic Data Assimilation: A Least Squares Approach, Cambridge University Press.","DOI":"10.1017\/CBO9780511526480"},{"key":"ref_36","doi-asserted-by":"crossref","first-page":"399","DOI":"10.1175\/1520-0469(1998)055<0399:OSFSWO>2.0.CO;2","article-title":"Optimal sites for supplementary weather observations: Simulation with a small model","volume":"55","author":"Lorenz","year":"1998","journal-title":"J. Atmos. Sci."},{"key":"ref_37","unstructured":"Lorenz, E.N. (1995, January 4\u20138). Predictability: A problem partly solved. Proceedings of the Seminar on Predictability, Reading, UK."},{"key":"ref_38","doi-asserted-by":"crossref","first-page":"2884","DOI":"10.1175\/1520-0493(2001)129<2884:AEAKFF>2.0.CO;2","article-title":"An ensemble adjustment Kalman filter for data assimilation","volume":"129","author":"Anderson","year":"2001","journal-title":"Mon. Weather Rev."},{"key":"ref_39","doi-asserted-by":"crossref","first-page":"4977","DOI":"10.1175\/MWR-D-17-0094.1","article-title":"Can we optimize the assimilation order in the serial ensemble Kalman filter? A study with the Lorenz-96 model","volume":"145","author":"Kotsuki","year":"2017","journal-title":"Mon. Weather Rev."},{"key":"ref_40","doi-asserted-by":"crossref","unstructured":"Nishizawa, S. (2021). 4D-Var data assimilation using an adjoint model of a neural network surrogate model. Earth Space Sci. Open Arch. ESSOAr.","DOI":"10.1002\/essoar.10507570.1"},{"key":"ref_41","doi-asserted-by":"crossref","first-page":"2463","DOI":"10.1175\/JHM-D-14-0230.1","article-title":"Evaluation of the Global Land Data Assimilation System (GLDAS) air temperature data products","volume":"16","author":"Ji","year":"2015","journal-title":"J. Hydrometeorol."},{"key":"ref_42","doi-asserted-by":"crossref","first-page":"679","DOI":"10.1016\/j.ijforecast.2006.03.001","article-title":"Another look at measures of forecast accuracy","volume":"22","author":"Hyndman","year":"2006","journal-title":"Int. J. Forecast."},{"key":"ref_43","doi-asserted-by":"crossref","unstructured":"Peng, K., Cao, X., Liu, B., Guo, Y., Xiao, C., and Tian, W. (2021). Polar Vortex Multi-Day Intensity Prediction Relying on New Deep Learning Model: A Combined Convolution Neural Network with Long Short-Term Memory Based on Gaussian Smoothing Method. Entropy, 23.","DOI":"10.3390\/e23101314"},{"key":"ref_44","doi-asserted-by":"crossref","first-page":"282","DOI":"10.1016\/0022-1694(70)90255-6","article-title":"River flow forecasting through conceptual models part I\u2014A discussion of principles","volume":"10","author":"Nash","year":"1970","journal-title":"J. Hydrol."}],"container-title":["Entropy"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/1099-4300\/24\/2\/264\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,10]],"date-time":"2025-10-10T22:18:13Z","timestamp":1760134693000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/1099-4300\/24\/2\/264"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2022,2,12]]},"references-count":44,"journal-issue":{"issue":"2","published-online":{"date-parts":[[2022,2]]}},"alternative-id":["e24020264"],"URL":"https:\/\/doi.org\/10.3390\/e24020264","relation":{},"ISSN":["1099-4300"],"issn-type":[{"value":"1099-4300","type":"electronic"}],"subject":[],"published":{"date-parts":[[2022,2,12]]}}}