Impact of dynamical representational errors on an Indian Ocean ensemble data assimilation system

Abstract

This study investigates the impact of dynamical representational error (RE) on the analysis of an ocean ensemble Kalman filter‐based data assimilation system, LETKF‐ROMS (Local Ensemble Transform Kalman Filter – Regional Ocean Modelling System) configured for the Indian Ocean and assimilating in situ temperature and salinity observations from Argo. Three different approaches to account for the RE are studied and inter‐compared: (a) static RE (varies in horizontal and vertical direction), (b) dynamic RE (varies in space and time) estimated from concurrent observations, and (c) dynamic RE estimated using concurrent high‐resolution model outputs. RE estimated from the model outputs exhibits rich spatial and temporal variability with an estimated temporal mean RE for temperature below 0.5 and 0.2 °C in the surface and deep layers, respectively, and reaching up to 1 °C in the thermocline layers. The region encompassing the Great Whirl displays a large seasonal variability reaching up to 0.8 °C, and the South Equatorial Current (SEC) a large interannual variability reaching up to 0.4 °C. Neglecting such spatio‐temporal variations of RE and assimilating with a static RE limited the benefits of assimilation by entertaining over‐fitting issues that caused degradations in the Bay of Bengal, the western parts of the Arabian Sea, and the equatorial Indian Ocean. Assimilating with the observations‐based dynamic RE improved the results in these regions, but the best performances were obtained with the configuration using the model‐based dynamic RE, which yielded further improvements, e.g. reduction of sea surface height root‐mean‐square errors reaches 30% with respect to the observations‐based dynamic RE. The latter also better handled the rich spatial variability regions and areas not well sampled by the observations. Improved estimates of the spatial and temporal variations of RE helped to better exploit the assimilated observations and provided enhanced analyses less prone to assimilation shocks.

Impact of Atmospheric and Model Physics Perturbations on a High‐Resolution Ensemble Data Assimilation System of the Red Sea

“…The improved ensemble and balanced ocean forecasting system using ensemble atmospheric forcing and perturbed model physics presented in this study is an important step toward the development of the first ocean forecasting and reanalysis system for the Red Sea. Further enhancement of the performances of the present system by using flow‐and‐depth‐dependent observation error variances (e.g., Sanikommu et al, 2019) will be the focus of our next effort.…”

“…Observation error variance is an important element of any data assimilation system (e.g., Hoteit et al, 2010; Sanikommu et al, 2019) and should account for errors due to: deficiencies in measurement devices, unresolved processes, unresolved subgridscale dynamics, and numerical errors in interpolation. Temporally static and spatially homogeneous observational error variance values of (0.04 m) 2 , (0.5°C) 2 , and (0.3 psu) 2 are used for the satellite along‐track SLA, the in situ T and S, respectively.…”

Impact of Atmospheric and Model Physics Perturbations On a High-Resolution Ensemble Data Assimilation System of the Red Sea

“…Gittings et al, 2018;Zhan et al, 2018 Observation error variances, which comprises errors due to instruments and unresolved scales and processes and interpolation, is an important element of the data assimilation system (e.g. Hoteit et al, 2010;Sivareddy et al, 2019). Temporally static and spatially homogeneous observational error variance values of (0.04 m) 2 , (0.5°C) 2 and (0.3psu) 2 are used for the satellite along-track SLA, the in situ T and S, respectively.…”

“…Temporally static and spatially homogeneous observational error variance values of (0.04 m) 2 , (0.5°C) 2 and (0.3psu) 2 are used for the satellite along-track SLA, the in situ T and S, respectively. These error variances for T and S, which are chosen in accordance with the suggested ranges of in situ observational errors by earlier assimilation studies (e.g., Richman et al, 2005;Forget and Wunsch, 2007;Oke and Sakov, 2008;Karspeck, 2016), are intended to account the expected dominant errors from unresolved scales and processes (Sivareddy et al 2019). The SLA observational error of (0.04 m) 2 , which is slightly larger than the suggested altimeter accuracy (AVISO 2015;Hoteit et al, 2002), is based on the sensitivity of our assimilation system to various choices of error variances, (0.04 m) 2 , (0.07 m) 2 , and (0.1 m) 2 (results not shown).…”

“…Many ocean studies later followed, implementing ensemble ocean data assimilation systems with perturbed atmospheric forcing and showing important improvements in the forecasts (e.g. Lisaeter et al, 2003;Evenson, 2004;Wan et al, 2008;Shu et al, 2011;Sakov et al, 2012;Karspect et al, 2013;Penny et al, 2015;Sivareddy et al, 2017Sivareddy et al, , 2019. Other studies assessed the background/model error characteristics based on free model ensemble runs perturbing different parameters (Brankart et al, 2015), boundary conditions (Sandery et al, 2014), bathymetry (e.g.…”