Data assimilation is a technique in which data collected from field observations or experiments are used to improve model performance. In almost all cases, models that are developed to predict how a system will change over time and space incorporate some level of uncertainty or approximation, or are sensitive to initial conditions (i.e., exhibit chaotic behavior). As a consequence of model imperfections or inaccuracies in initial conditions, almost all model predictions will diverge from the true system over time. A classic example of such systems arise in weather prediction where both model approximations and sensitivity to initial conditions cause weather forecasts to become inaccurate after a few days. However, by continuously updating the model based on current and past measurements, the best possible predictions can be obtained. Data assimilation can also be used to obtain better estimates of model parameters under current operating conditions, which is also known as parameter estimation. Highlighted below are some examples of data assimilation applied to biogeochemical modeling.

Food Web Model Microbial Food Web
Data assimilation used to improve microbial food web model using microcosm experiments.
Metabolism Model
Estuarine Metabolism
In this example data assimilation is used to estimate estuarine gross and net production as well as community respiration by combining dissolved oxygen measurementes with our 1D AD model.