Title:On the dynamics and evolution of forecast uncertainty and forecast sensitivity

Author: Alexander Goldstein (University of Wisconsin - Madison)

Ensemble variance is a useful tool in the diagnosis of the potential regions of uncertainty within a numerical weather prediction (NWP) model forecast. Based upon the examination of several cases, it appears that areas of high uncertainty within the forecast state are often co-located with areas of error in the deterministic forecast. Ensemble variance can propagate from upstream as well as be generated directly within a verifying region. This prompts the question, what relationship, if any, exists between forecast uncertainty and sensitivity to initial conditions? The relationship between adjoint sensitivity and ensemble NWP variance is explored with respect to two different scenarios: one in which ensemble variance propagates into a region prior to the development of synoptic-scale features, e.g., mid-latitude and/or tropical cyclones, the other in which variance is associated with the development of the feature itself. It is suggested that high forecast uncertainty often originates within areas of sparsely sampled data, e.g., within the high latitudes. If a flow pattern is present such that it intersects an area of high extant uncertainty, it is shown that this uncertainty propagates and evolves with the flow in a manner similar to that of a convective (as opposed to an absolute) instability. Sensitivity of specific metrics of the forecast (e.g., cyclone intensity) are computed for cases of both high and low predictability, and cases where forecast uncertainty has clear origins upstream versus those generated within a synoptic-scale weather system.

GMAO Head: Steven Pawson Global Modeling and Assimilation Office NASA Goddard Space Flight Center

Curator: Nikki Privé Last Updated: Feb 9 2015