HR: 0800h
AN: SA51B-0256    [Abstracts]
TI: Validation of the Objective Analysis Algorithm IDA3D
AU: * Bust, G S
EM: gbust@arlut.utexas.edu
AU: Garner, T W
EM: garner@arlut.utexas.edu
AU: Taylor, B
EM: taylor@arlut.utexas.edu
AB: We have developed a set of metrics and skill scores to analyze the performance of the objective analysis algorithm Ionospheric Data Assimilation Three-Dimensional (IDA3D). The objective of the skill score is to quantify the improvement in model predictions after either assimilation of a new data source, or improvement in the model itself. The metrics chosen to be evaluated by the skill score include electron density in the E-regin, F-region, and topside of the ionosphere, total electron content (TEC), F-region peak height and F-region layer thickness. The metrics can be sub-binned into different geographic regions, solar and magnetic activity levels and time of day. For a first principles data assimilation model that is capable of forecasting, the metric can also be a function of prediction time beyond the analysis time. The skill score we have developed begins with a pure climate prediction of the metric values. Then we run the data assimilative model and produce assimilative prediction of the metric values. The skill score is then given as \begin{equation} \label{eqn:skill} S = 1.0 - \frac{<|\vec D - \vec T|>}{<|\vec M - \vec T|>}. \end{equation} Where $\vec D$ is the array of metric values ouput from the data assimilation model, $\vec T$ is the independent measurements of the metric that serve as ``truth'', and $\vec M$ is the array of climate predictions of the metric values. The skill is a function of time in the sense that every time the data assimilative model is improved we generate new skill scores. The values of the skill score range from $1.0$ for when the data assimilative model is perfect (that is agrees with the measurements), to $0.0$ when the data assimilative model is no better than the climate predictions, to negative infinity as the data assimilative model does much worse than climate. We will present preliminary results for several days of analysis with incoherent scatter observatins of electron density serving as truth.
DE: 6982 Tomography and imaging
DE: 2447 Modeling and forecasting
DE: 2494 Instruments and techniques
SC: SPA-Aeronomy [SA]
MN: 2004 AGU Fall Meeting