FYRE Climate: a high-resolution reanalysis of daily precipitation and temperature in France from 1871 to 2012

Surface observations are usually too few and far between to properly assess multidecadal variations at the local scale and characterize historical local extreme events at the same time. A data assimilation scheme has been recently presented to assimilate daily observations of temperature and precipitation into downscaled reconstructions from a global extended reanalysis through an Ensemble Kalman fitting approach and to derive high-resolution fields. Recent studies also showed that assimilating observations at high temporal resolution does not guarantee correct multidecadal variations. The current paper thus proposes (1) to apply the data assimilation scheme over France and over the 1871–2012 period based on the SCOPE Climate reconstructions background dataset and all available daily historical surface observations of temperature and precipitation, (2) to develop an assimilation scheme at the yearly timescale and to apply it over the same period and lastly, (3) to derive the FYRE Climate reanalysis, a 25-member ensemble hybrid dataset resulting from the daily and yearly assimilation schemes, spanning the whole 1871–2012 period at a daily and 8 km resolution over France. Assimilating daily observations only allows reconstructing accurately daily characteristics, but fails in reproducing robust multidecadal variations when compared to independent datasets. Combining the daily and yearly assimilation schemes, FYRE Climate clearly performs better than the SCOPE Climate background in terms of bias, error, and correlation, but also better than the Safran reference surface reanalysis over France available from 1958 onward only. FYRE Climate also succeeds in reconstructing both local extreme events and multidecadal variability. It is freely available at https://doi.org/10.5281/zenodo.4005573 (precipitation, Devers et al., 2020b) and https://doi.org/10.5281/zenodo.4006472 (temperature, Devers et al., 2020c).

The improvements to the regional South China Sea Operational Oceanography Forecasting System (SCSOFSv1)

South China Sea Operational Oceanography Forecasting System (SCSOFS) had been constructed and operated in National Marine Environmental Forecasting Center of China to provide daily updated hydrodynamic forecasting in SCS for the future 5 days since 2013. This paper presents recent comprehensive updates of the configurations of the physical model and data assimilation scheme in order to improve SCSOFS forecasting skills. It highlights three of the most sensitive updates, including sea surface atmospheric forcing method, tracers advection discrete scheme, and modification of data assimilation scheme. Inter-comparison and accuracy assessment among five versions during the whole upgrading processes are performed by employing OceanPredict Inter-comparison and Validation Task Team Class4 metrics. The results indicate that remarkable improvements have been achieved in SCSOFSv2 with respect to the original version known as SCSOFSv1. Domain averaged monthly mean root mean square errors decrease from 1.21 °C to 0.52 °C for sea surface temperature, from 21.6 cm to 8.5 cm for sea level anomaly, respectively.

Assessing the Impact of the Assimilation of SWOT Observations in a Global High-Resolution Analysis and Forecasting System Part 1: Methods

The future Surface Water Ocean Topography (SWOT) mission due to be launched in 2022 will extend the capability of existing nadir altimeters to enable two-dimensional mapping at a much higher effective resolution. A significant challenge will be to assimilate this kind of data in high-resolution models. In this context, Observing System Simulation Experiments (OSSEs) have been performed to assess the impact of SWOT on the Mercator Ocean and Copernicus Marine Environment Monitoring Service (CMEMS) global, high-resolution analysis and forecasting system. This paper focusses on the design of these OSSEs, in terms of simulated observations and assimilation systems (ocean model and data assimilation schemes). The main results are discussed in a companion paper. Two main updates of the current Mercator Ocean data assimilation scheme have been made to improve the assimilation of information from SWOT data. The first one is related to a different parametrisation of the model error covariance, and the second to the use of a four-dimensional (4D) version of the data assimilation scheme. These improvements are described in detail and their contribution is quantified. The Nature Run (NR) used to represent the “truth ocean” is validated by comparing it with altimeter observations, and is then used to simulate pseudo-observations required for the OSSEs. Finally, the design of the OSSEs is evaluated by ensuring that the differences between the assimilation system and the NR are statistically consistent with the misfits between real ocean observations and real-time operational systems.

The improvements to the regional South China Sea Operational Oceanography Forecasting System

South China Sea Operational Oceanography Forecasting System (SCSOFS) had been built up and operated in National Marine Environmental Forecasting Center of China to provide daily updated hydrodynamic forecasting in SCS for the future 5 days since 2013. This paper presents comprehensive updates had been conducted to the configurations of the physical model and data assimilation scheme in order to improve SCSOFS forecasting skills in recent years. It highlights three of the most sensitive updates, sea surface atmospheric forcing method, tracers advection discrete scheme, and modification of data assimilation scheme. Scientific inter-comparison and accuracy assessment among five versions during the whole upgrading processes are performed by employing Global Ocean Data Assimilation Experiment OceanView Inter-comparison and Validation Task Team Class4 metrics. The results indicate that remarkable improvements have been achieved in SCSOFSv2 with respect to the original version SCSOFSv1. Domain averaged monthly mean root mean square errors decrease from 1.21 °C to 0.52 °C for sea surface temperature, from 21.6 cm to 8.5 cm for sea level anomaly, respectively.

Development of an OMI AI data assimilation scheme for aerosol modeling over bright surfaces – a step toward direct radiance assimilation in the UV spectrum

Using the Vector LInearized Discrete Ordinate Radiative Transfer (VLIDORT) code as the main driver for forward model simulations, a first-of-its-kind data assimilation scheme has been developed for assimilating Ozone Monitoring Instrument (OMI) aerosol index (AI) measurements into the Naval Aerosol Analysis and Predictive System (NAAPS). This study suggests both RMSE and absolute errors can be significantly reduced in NAAPS analyses with the use of OMI AI data assimilation, when compared to values from NAAPS natural runs. Improvements in model simulations demonstrate the utility of OMI AI data assimilation for improving the accuracy of aerosol model analysis over cloudy regions and bright surfaces. However, the OMI AI data assimilation alone does not out-perform aerosol data assimilation that uses passive-based aerosol optical depth (AOD) products over cloud free skies and dark surfaces. Further, as AI assimilation requires the deployment of a fully-multiple-scatter-aware radiative transfer model in the forward simulations, computational burden is an issue. Nevertheless, the newly-developed modeling system contains the necessary ingredients for assimilation of radiances in the ultra-violet (UV) spectrum, and our study shows the potential of direct radiance assimilation at both UV and visible spectrums, possibly coupled with AOD assimilation, for aerosol applications in the future. Additional data streams can be added, including data from TROPOspheric Monitoring Instrument (TROPOMI), Ozone Mapping and Profiler Suite (OMPS) and eventually with the Plankton, Aerosol, Cloud and ocean Ecosystem (PACE) mission.

Learning missing part of physics-based models within a variational data assimilation scheme

<p>The analogy between data assimilation and machine learning has already been shown and is still being investigated to address the problem of improving physics-based models. Even though both techniques learn from data, machine learning focuses on inferring model parameters while data assimilation concentrates on hidden system state estimation with the help of a dynamical model. <br> <br>Also, neural networks and more precisely ResNet-like architectures can be seen as dynamical systems and numerical schemes, respectively. They are now considered state of the art in a vast amount of tasks involving spatio-temporal forecasting. But to train such networks, one needs dense and representative data which is rarely the case in earth sciences. At the same time, data assimilation offers a proper Bayesian framework allowing to learn from partial, noisy and indirect observations. Thus, each of this field can profit from the other by providing either a learnable class of dynamical models or dense data sets.</p><p>In this work, we benefit from powerful and flexible tools provided by the deep learning community based on automatic differentiation that are clearly suitable for variational data assimilation, avoiding explicit adjoint modelling. We use a hybrid model divided into 2 terms. The first term is a numerical scheme that comes from the discretisation of physics-based equations, the second is a convolutional neural network that represents the unresolved part of the dynamics. From the Data Assimilation point of view, our network can be seen as a particular parametrisation of the model error. We then jointly learn this parameterisation and estimate hidden system states within a variational data assimilation scheme. Indirectly, the issue of incorporating physical knowledge into machine learning models is also addressed. </p><p>We show that the hybrid model improves forecast skill compared to traditional data assimilation techniques. The generalisation of the method on different models and data will also be discussed.</p>