Estimating the state of the atmosphere from observations is essential for monitoring climate and also for the initialization of weather, sub-seasonal and seasonal weather forecasts. A popular way of making such estimates is using a method called the Ensemble Kalman Filter.

A known weakness of the Ensemble Kalman filter approach is that its ability to provide state estimates that closely match densely distributed observations is very limited.

This paper describes a computationally inexpensive innovative variation on the technique that greatly ameliorates this difficulty.