Challenges in Data Assimilation

In my research what I'm trying to do is to prove that data assimilation would be an effective way to gain better results in ocean numerical modelling specifically in the sea ice. However today I read an article about the errors arise from data assimilation. It was an eye opener.

There are three types of errors


1. Temporal error:

This is the kind of error that arise from mismatch of time and space.  For an example the value of temperature that is being measured might be a point data while the data used by the computation model depends upon the time step, usually in my case it's one day. This also applies for the position of data obtained. Matching the grid points and actual position of data obtaining points is quite challenging.

2. Instrumental error:

This is the error rising from the error in instruments. The observational data that we use in assimilation might be wrong and the error can get accumulated with cycles.

3. Assimilation Residuals:

The above errors might be easy to avoid with proper handling of data measurement but residuals are inherent to the concept of data assimilation. Data assimilation tends to nudge the results towards observation results. This might lead to avoid the extreme weather predictions.

Other than these errors, it must also be considered that obtaining observational data is extremely difficult in oceanography.