Thermospheric Weather as Observed by Ground‐Based FPIs and Modeled by GITM

• Published in: Journal of geophysical research. Space physics Vol. 124; no. 2; pp. 1307 - 1316
• Main Authors: Harding, Brian J., Ridley, Aaron J., Makela, Jonathan J.
• Format: Journal Article
• Language: English
• Published: Washington Blackwell Publishing Ltd 01-02-2019
• Subjects: Atmosphere; Atmospheric models; Auroral zone; Auroral zones; Autocorrelation function; Autocorrelation functions; Climate models; Correlation; Correlation analysis; Data correlation; first‐principles model; Fluctuations; FPI; Ground-based observation; Interferometers; Ionosphere; Ionospheric models; Lower atmosphere; Thermosphere; Thermospheric winds; Time lag; Upper atmosphere; variability; Weather; Weather forecasting; Wind
• ISSN: 2169-9380; 2169-9402
• DOI: 10.1029/2018JA026032

The first long‐term comparison of day‐to‐day variability (i.e., weather) in the thermospheric winds between a first‐principles model and data is presented. The definition of weather adopted here is the difference between daily observations and long‐term averages at the same UT. A year‐long run of the Global Ionosphere Thermosphere Model is evaluated against a nighttime neutral wind data set compiled from six Fabry‐Perot interferometers at middle and low latitudes. First, the temporal persistence of quiet‐time fluctuations above the background climate is evaluated, and the decorrelation time (the time lag at which the autocorrelation function drops to e−1) is found to be in good agreement between the data (1.8 hr) and the model (1.9 hr). Next, comparisons between sites are made to determine the decorrelation distance (the distance at which the cross‐correlation drops to e−1). Larger Fabry‐Perot interferometer networks are needed to conclusively determine the decorrelation distance, but the current data set suggests that it is ∼1,000 km. In the model the decorrelation distance is much larger, indicating that the model results contain too little spatial structure. The measured decorrelation time and distance are useful to tune assimilative models and are notably shorter than the scales expected if tidal forcing were responsible for the variability, suggesting that some other source is dominating the weather. Finally, the model‐data correlation is poor (−0.07 < ρ < 0.36), and the magnitude of the weather is underestimated in the model by 65%.
Plain Language Summary
Much like in the lower atmosphere, weather in the upper atmosphere is harder to predict than climate. Physics‐based models are becoming sophisticated enough that they can in principle predict the weather, and we present the first long‐term evaluation of how well a particular model, Global Ionosphere Thermosphere Model, performs. To evaluate the model, we compare it with a year of data from six ground‐based sites that measure the thermospheric wind. First, we calculate statistics of the weather, such as the decorrelation time, which characterizes how long weather fluctuations persist (1.8 hr in the data and 1.9 hr in the model). We also characterize the spatial decorrelation by comparing weather at different sites. The model predicts that the weather is much more widespread than the data indicates; sites that are 790 km apart have a measured correlation of 0.4, while the modeled correlation is 0.8. In terms of being able to actually predict a weather fluctuation on a particular day, the model performs poorly, with a correlation that is near zero at the low latitude sites, but reaches an average of 0.19 at the midlatitude sites, which are closer to the source that most likely dominates the weather: heating in the auroral zone.
Key Points
A long‐term data‐model comparison of day‐to‐day thermospheric variability finds that GITM represents the weather poorly (−0.07 < ρ < 0.36)
The average measured decorrelation time of 1.8 hr agrees with the modeled time of 1.9 hr
The weather in GITM contains too little spatial structure, when compared with the measured ∼1,000‐km decorrelation distance