A Strategy for Assimilating Data from Microsats

The Earth observing system is rapidly changing from a few, long-lived, well-calibrated, but expensive legacy satellites towards many smaller, more flexible, and less expensive microsats. A constellation of microsats can provide a high volume of high temporal resolution data, fill data gaps, and mitigate risk via data redundancy; however, microsats exhibit higher uncertainty, larger variability in sensor performance, shorter lifetimes, and unknown biases. We must therefore develop new methods to adapt to microsats and rapidly integrate them into our NWP models.

More specifically, we need to develop a capability to optimally sample high-quality, high-impact data, and adaptively weight observations by varying their uncertainty specifications as a function of space and time. We plan to 1) replace our current globally uniform satellite radiance sampling strategy with dynamic sampling concentrated in regions of interest, and 2) replace our static observation uncertainty specifications with adaptive ones. For our first task, essentially intelligent data selection, we will use a large archive of forecast sensitivity to observation impact (FSOI) data and machine learning methods to predict which observations are likely to have the greatest positive impact on our forecasts, and preferentially select those observations. For our second task, we will start by extending the Desroziers 2005 technique to estimate observation error variances and covariances as a function of time and region in order to find the observation uncertainty values that optimize information content. Machine learning methods may prove useful here as well, using a host of metadata as predictors of observation error variance, to augment the Desroziers technique.