Meteorol Atmos Phys 93, 129–145 (2006) DOI 10.1007/s00703-005-0152-9 Centre for Atmospheric Sciences, Indian Institute of Technology Delhi, New Delhi, India 4D-variational assimilation over the Bay of Bengal using the Adjoint of a Variable Resolution Global Ocean Model S. K. Deb, H. C. Upadhyaya, and O. P. Sharma With 10 Figures Received February 3, 2005; accepted April 19, 2005 Published online: May 8, 2006 # Springer-Verlag 2006 Summary The adjoint technique has been widely used over the last two decades in applications involving the dynamics of the Atmosphere and Ocean. This present study applies this technique in the assimilation of oceanic parameters such as temperature and net-surface heat flux, in a variable resolution Ocean General Circulation Model (OGCM ). The National Center for Environmental Prediction (NCEP ) 1999 July–August global data (sea-surface temperature and derived surface heat flux) and vertical temperature profiles collected from ORV Sagar Kanya and INS Sagardhwani during the Bay of Bengal Monsoon Experiment (BOBMEX 1999) are used for assimilation over the Bay of Bengal. The net-surface heat flux and temperature obtained after assim- ilation show good agreement with their respective seasonal patterns over the whole Indian Ocean, though there are some differences in the numerical values of heat fluxes. The vertical profiles of temperature after assimilation over the Bay of Bengal show strong resemblance with the cor- responding ship observations, emphasizing the importance of the adjoint technique in ocean data assimilation. 1. Introduction A numerical model along with a good data assimilation technique has proved to be a power- ful method in estimating unobserved data from available observations, generating data for data- void regions of the globe with physical consisten- cy, particularly over the ocean. The development of remote sensing techniques along with their application in the Oceanographic field has revo- lutionized data assimilation. Currently, Ocean General Circulation Models (OGCM s) are not quite capable of providing a complete diagnosis and understanding of the complicated ocean sys- tem response to observed atmospheric forcing. Although OGCM s are equipped with complex dynamics and run with eddy resolving resolution and realistic geometry, their capability of simu- lating the three dimensional ocean circulation structure is limited not only by poorly known model input parameters (such as surface momen- tum, net heat flux, eddy mixing coefficients), but also the lack of direct measurements. In addition to this insufficient knowledge of air sea interac- tion process is another limiting factor. It is also well verified that traditional bulk aerodynam- ics formulas are no longer a reliable tool for calculating the flux distribution. Among many available data assimilation techniques, the ad- joint method is perhaps the most promising and practical approach in combining complex Ocean General Circulation models with large data sets. The adjoint technique is a well-known technique used in problems related to the dynamics of atmosphere and ocean. It is based on the varia- tional optimal control technique (Sasaki, 1970; Talagrand and Courtier, 1987; Talagrand, 1991; Yu and O’Brien, 1991; Chao and Chang, 1992;