# Adjoint assimilation

From AMS Glossary

## adjoint assimilation

A form of variational data assimilation in which adjoint equations are used to obtain gradients of a scalar measure of a forecast (

*J*) with respect to model control variables.The complete assimilation generally involves an iterative procedure to reduce the cost-function in which a minimization algorithm is used to adjust initial conditions based on a sensitivity gradient (

**∇**_{x}*J*) provided by the adjoint model.*See*adjoint sensitivity, variational objective analysis.