Adaptive Observations in a 4D-Var Framework Applied to the Nonlinear Burgers Equation Model

In 4D-Var data assimilation for geophysical models, the goal is to reduce the lack of fit between model and observations (strong constraint approach assuming perfect model). In the last two decades four dimensional variational technique has been extensively used in the numerical weather prediction due to the fact that time distributed observations are assimilated to obtain a better initial condition thus leading to more accurate forecasts using the above 4D-Var approach. The use of large-scale unconstrained minimization routines to minimize a cost functional measuring lack of fit between observations and model forecast requires availability of the gradient of the cost functional with respect to the control variables. Nonlinear Burgers equation model is used as numerical forecast model. First order adjoint model can be used to find the gradient of the cost functional. The use of targeted observations supplementing routine observations contributes to the reduction of the forecast analysis error and can provide improved forecast of weather events of critical societal impact, for instance, hurricanes, tornadoes, sharp fronts etc. The optimal space and time locations of the adaptive observations can be determined by using a singular vector approach. In our work we use both adjoint sensitivity and sensitivity to observation approaches to identify the optimal space and time locations for targeted observations at future time aimed at providing an improved forecast. Both approaches are compared in this work and some conclusions are outlined.