Numerical Prediction of the Migrating Diurnal Tide Total Variability in the Mesosphere and Lower Thermosphere

We present a forecast model for the total variability of DW1 and study its prediction accuracy against the actual variability from the extended Canadian Middle Atmospheric Model (eCMAM) simulations and Sounding of the Atmosphere using Broadband Emission Radiometry (SABER) observations in the mesosphere and lower thermosphere region. To account for the long-term variability (periods >30 days), we fit the data with a multi-linear regression model that contains the solar cycle, El Nino Southern Oscillation, quasi-biennial oscillation, and the seasonal harmonics at 12, 6, 4, and 3 months. The fitting coefficients/amplitudes from each deterministic variability are examined between the eCMAM and SABER. To predict the short-term tidal variability, we adopted an auto-regression (AR) model from Vitharana et al. (2019, ). The forecast model is a combination of the multi-linear regression model and the AR model. The forecast model can predict the total tidal variability of DW1 with high accuracy. The prediction accuracy (correlation coefficient) ranges between 0.81 and 0.98 (average 0.93) for eCMAM and 0.76 and 0.92 (average 0.82) for SABER. The prediction accuracy for the short-term tidal variability is very high for both eCMAM and SABER and shows little variation with correlation coefficients at similar to 0.95. The prediction accuracy for the total tidal variability follows that for the long-term tidal variability. The discrepancy between the forecast model and the actual total tidal variability is mainly because the forecast model cannot capture all the long-term tidal variability (periods >30 days). Besides the well-known periods for long-term variability, the other periods change on a year-to-year basis.