| description abstract | he Lagrangian evolution of cloud cover and cloud-controlling variables is well approximated using red noise processes with different autocorrelation time scales for each variable. Trajectories within the subtropical marine boundary layer are generated using winds from ECMWF Re-Analysis data for low cloud decks in four eastern subtropical ocean basins. Cloud cover, liquid water path, and boundary layer depth are sampled at 12-h intervals using A-Train satellites, and droplet concentration is sampled every 24 h. Lower-tropospheric stability and vertical velocity are sampled concurrently using reanalysis data. Samples are converted to seasonal and diurnal anomalies. Data are spatially averaged over a range of length scales. The e-folding decay times τ for autocorrelation are calculated for each variable based on lag times of 12, 24, 36, and 48 h. Using lag 24 h and an averaging radius of 100 km, τ ≈ 15?17 h for liquid water path and vertical velocity, τ ≈ 19 h for cloud cover, τ ≈ 24?25 h for boundary layer depth and droplet concentration, and τ ≈ 53 h for lower-tropospheric stability.Time scales vary somewhat between regions and are shortest in the eastern Indian Ocean. Decay time τ increases with averaging scale and the autocorrelation e-folding length of a variable at a fixed time. Diurnal analysis shows cloud cover anomalies have a stronger memory during morning breakup, while other variables show stronger memory as clouds reform in the evening. Lagrangian cloud anomalies are less persistent than anomalies at a fixed location. For the latter, estimated τ values can vary significantly at different lag times, so a red noise assumption is inappropriate. | |
