Storage and Reset: How Indian Ocean Seasons Remember and Erase Zonal Aerosol Gradients—A 24-Year MODIS MAIAC Perspective (2001–2024)

Aerosol loading contrasts sharply between the Arabian Sea and the Bay of Bengal. Still, earlier work treated this east–west difference as a fixed pattern or a simple trend, leaving two questions open: how long a gradient formed in one season persists, and whether that persistence is linked to the long-term trend. The present study addresses this using 24 years (2001–2024) of 1 km MODIS MAIAC (MCD19A2) Aerosol Optical Depth (AOD) datasets for Kavaratti (10.56°N, Arabian Sea) and Port Blair (11.62°N, Bay of Bengal). The sites act as a methodological control: they are separated by ~20° in longitude but only ~1° in latitude, so their AOD difference reflects zonal transport and source contrasts rather than solar geometry or monsoon regime, and both are small remote island observatories where the signal is transported rather than locally emitted. Seasonal zonal gradients (longitude-normalised AOD differences) were tested for trend using Mann–Kendall and Sen's slope, and for persistence using autocorrelation and e-folding timescales.

The gradient is negative in every season. Two independent diagnostics agree on a storage–reset structure: March, April, May (MAM) and September, October, November (SON) hold a positive autocorrelation at lags of one and two years with finite e-folding times (~0.9 and ~0.8 years), whereas December, January, February (DJF) and June, July, August (JJA) decorrelate at short lags and give negative or undefined e-folding times—consistent with winter ventilation and monsoon wet scavenging. Critically, DJF has the weakest memory, yet is the only season with a significant long-term weakening, so persistence and trend are decoupled: memory governs seasonal predictability, while trend reflects external forcing. These mechanisms are inferred from persistence structure and regional meteorology rather than co-analysed rainfall or boundary-layer data.

Persistence thus gives climate models a benchmark beyond mean AOD and trend, which a model can match for the wrong reasons; reproducing storage–reset timing tests scavenging and ventilation parameterisations directly.