Sun, R., A. C. Subramanian, M. R.
Mazloff, S. Sanikommu, G. Gopalakrishnan, B. D. Cornuelle, T.-Y. Hsu, A. J. Miller, P. Ag and
I. Hoteit, 2026:
Exploring air–sea interactions in a regional weakly
coupled data assimilation system: An observing system
experiment for the 2023 Indian Ocean Monsoon and
Cyclone Biparjoy.
Journal of Advances in Modeling Earth Systems, 18, e2026MS005804.
Abstract.
Prior to the onset of the South Asian monsoon, the Arabian Sea experiences a warming phase
We examined air–sea interactions in a weakly coupled ocean–atmosphere data assimilation (DA)
system by performing an observing system experiment (OSE). The OSE is performed for the Arabian Sea
monsoon onset event in 2023, which also coincides with the passage of the tropical cyclone Biparjoy. Results
show that assimilating both oceanic and atmospheric data produces a better forecast of the ocean–atmosphere
state (fit to the assimilated data) than either experiment assimilating only oceanic or atmospheric data. This
indicates that the coupled system provides more consistent initial conditions, yielding more balanced oceanic
and atmospheric states when assimilating all the data. However, we found that assimilating only atmospheric
data improves both the ocean and atmosphere state forecasts, whereas assimilating only oceanic data improves
the ocean state forecasts but degrades the atmospheric state forecasts, when compared with the free run (coupled
model simulation without assimilating any data). We found that this counterintuitive model performance is due
to the systematic misrepresentation of air–sea exchanges. First, errors in the cyclone path increase when
atmospheric data are not assimilated. Then, assimilation of sea surface temperature (SST) leads to a warmer SST
below the misplaced cyclone (with cyclone path offset of about 200 km relative to IBTrACS data), resulting in
increased cyclone intensity. This misrepresentation of air–sea exchanges not only occurs during the cyclone
passage but also persists throughout the early monsoon season. Since errors in one component can affect the
other and result in positive feedback, leading to error growth, this study highlights an interesting effect on
weakly coupled ocean–atmosphere DA systems when atmospheric data are unavailable or limited in space and
time, particularly during high‐impact atmospheric events such as tropical cyclones.
Reprint (pdf)