Assimilation of Multiple Data Types to a Regional Ionosphere Model With a 3D‐Var Algorithm (IDA4D)

Abstract

For the purpose of building a regional (bound 20–60°N in latitude and 110–160°E in longitude) ionospheric nowcast model, we investigated the performance of IDA4D (Ionospheric Data Assimilation Four‐Dimension) technique considering International Reference Ionosphere model as the background. The data utilized in assimilation were slant total electron content (STEC) from 27 ground GPS (Global Positioning System) receiver stations and NmF2 (ionospheric F2 peak density) from five ionosondes and COSMIC (Constellation Observing System for Meteorology, Ionosphere, and Climate) Data Analysis Archive Center. The period analyzed covered both geomagnetic quiet and disturbed days (15–18 March 2015). Assimilations were run under the following data combinations (cases): (1) GPS‐STEC's only; (2) GPS‐STEC's and NmF2's from five ionosondes; (3) only NmF2's from five ionosondes; and (4) GPS‐STEC's and NmF2's from both five ionosondes and COSMIC. Results showed that under case 1 the root‐mean‐square error (RMSE) in STEC reduced by 44% over the background International Reference Ionosphere values and on averaged over all ionosonde stations in the analysis RMSE values of foF2 (F2 layer critical frequency) reduced by 21%. Furthermore, foF2 RMSE values under Case 2 were 36% smaller than those under Case 1. Under Case 4, IDA4D performance improved even further in areas not covered by GPS and ionosonde measurements. Therefore, IDA4D is a potential candidate for regional ionosphere modeling that exhibits improved performance with assimilation of different data types.

Modeling total electron content derived from radio occultation measurements by COSMIC satellites over the African region

“…These grid resolutions resulted in 24 and 16 latitudinal and longitudinal bins, respectively. Several studies (e.g., Krankowski et al, 2011 andMengist et al, 2019) that have used COS-MIC data commonly consider measurements with horizontal smear > 1500 km prone to errors, and they reject such measurements. We established that after applying this restriction, there were ∼ 40 RO measurements per day during the year 2013 over our study area (not shown here).…”

“…As one of the IGS analysis centers, Center for Orbit Determination in Europe (CODE) provides Global Ionosphere Maps (GIMs) containing vertical TEC data daily using the GNSS data collected from over 200 tracking stations of IGS and other institutions. Several studies have used GIMs from CODE and other IGS analysis centers such as the Jet Propulsion Laboratory (JPL) to construct TEC models (Jakowski et al, 2011a;Mukhtarov et al, 2013;Ercha et al, 2012;Sun et al, 2017). Jakowski et al (2011a) proposed the Global Neustrelitz TEC Model (NTCM-GL) that describes the average TEC under quiet geomagnetic conditions.…”

“…The NTCM-GL was developed using GIMs during 1998-2007 provided by CODE. A global background TEC model was also built using CODE GIMs by Mukhtarov et al (2013). The model describes the climatological behavior of the ionosphere.…”

3‐D Regional Ionosphere Imaging and SED Reconstruction With a New TEC‐Based Ionospheric Data Assimilation System (TIDAS)

“…In particular, 3DVAR does not involve the time evolution of the state within a given time window (e.g., 15 min) and assumes all measurements have the same time‐stamp weighting. This is very similar to the second task (i.e., Update) of Kalman filters and thus is normally used in data assimilation based on empirical ionospheric models where the forward prediction is not directly available (although sometimes this can be implemented by using a Gauss‐Markov filter method; e.g., Aa et al., 2016; Bust et al., 2004, 2007; Mengist et al., 2019). In contrast, 4DVAR takes into account the time‐dependent weighting of different terms in the cost function within a given window, which makes the estimation of the error covariance an even more challenging and time‐costing issue (Ssessanga et al., 2019; C. Wang et al., 2004).…”

“…These include: Appropriate estimation of the background error covariance: A major difficulty in developing an accurate and reliable assimilation system lies in a proper description of the background error covariance, which is a key factor that determines the weighting contribution and the spread of information. Many data assimilation efforts, especially variation‐based algorithms, often use time‐invariant static background error covariance, based on certain assumptions of unbiased Gaussian distribution of ionospheric correlation length as deduced from climatological statistics (e.g., Aa et al., 2016; Bust et al., 2004, 2007; Mengist et al., 2019; Ssessanga et al., 2019; C. Wang et al., 2004; Yue et al., 2007). In reality, however, the ionospheric spatial correlation length and associated background error covariance may not be optimally represented under these assumptions, especially when the ionosphere exhibits considerable disturbances during space weather events.…”