Performance Analysis of a Strong Constraint 4DVar and 4DEnVar on Regional Ionosphere Imaging

Data assimilation (DA) techniques have recently gained traction in the ionospheric community, particularly at regional operational centers where more precise data are becoming prevalent. At center stage is the argument over which technique or scheme merits realization. At 4DSpace, we have in-house developed and assessed the performance of two regional flavors of short-term forecast strong constraint four-dimensional (4D, space and time) variational (SC4DVar) DA schemes; the orthodox incremental (SC4DVar-Inc) and ensemble-based (SC4DEnVar) approach. SC4DVar-Inc is bottled-necked by expensive Tangent Linear Models (TLMs) and model Ad-joints (MAs), while SC4DEnVar design mitigates these limitations. Both schemes initialize from the same background (IRI-2016), and electron densities forward propagated (30-min) by a Gauss Markov filter- the densities take on a log-normal distribution to assert the mandatory ionosphere density positive definiteness. Preliminary assimilation is performed only with ubiquitous Global Navigation Satellite System observables from ground-based receivers, with a focus on moderately stable mid-latitudes, specifically the Japanese archipelago and neighboring areas. Using a simulation analysis, we find that under model space localization, 30 member Ensembles are sufficient for regional SC4DEnVar. Verification of reconstructions is with independent observations from ground-based ionosonde and satellite radio occultations: the performance of both schemes is fairly adequate during the quiet period when the background has a better estimation of the hmF2. SC4DVar-Inc is slightly better over areas densely populated with measurements, but SC4DEnVar estimates the overall 3D ionosphere picture better, particularly in remote areas and during severe conditions. These results warrant SC4DEnVar as a better candidate for precise short-time regional forecasts. We have developed and assessed the performance of two flavors of short-term forecast four-dimensional (space and time) variational schemes. The first scheme relies on the quality of Tangent Linear Models and model Ad-joints in its design. The second scheme adopts an ensemble-based approach. Both schemes initialize from the same background and electron densities propagated by a Gauss Markov filter. Our verification analysis indicates that the ensemble approach approximates peak density-height variations better and offers improved estimates of the overall 3D ionosphere picture, particularly in remote areas and during severe conditions. Therefore, we recommend employing the ensemble-based approach for precise short-time regional forecasts. Specifying regional ionosphere with strong constraint 4DVar and 4DEnVar while ingesting ground GNSS data4DEnVar is computationally efficient at handling linearization and provides better 3D analyses in severe conditions and remote areas4DEnVar is recommended for regional ionosphere short-time forecast