Independent model validation is critical feedback to understand where our forecast models are performing well and where they are unreliable. Scientifically, validating solar wind models is a necessary…

Independent model validation is critical feedback to understand where our forecast models are performing well and where they are unreliable. Scientifically, validating solar wind models is a necessary goal to improve the accuracy and understand where physics of the system is correctly captured. Heliospheric models are regularly updated with new physics incorporated, or improved definitions of empirical relationships, but with these improvements the question still remains, how accurately does a model reproduce observed solar wind conditions? This question is multifaceted and needs to be addressed piecewise. When is the model accurate? To understand when a model is accurate, there fundamentally needs to be time resolution in any of the statistics used, i.e., a bulk measurement such as a mean squared difference is not enough. In addition, the statistics representing accuracy needs to be able to cope with state changes in the system. The solar wind will persist in a consistent state with stable conditions for a length of time and then rapidly change to a new state (e.g., the transition between fast solar wind stream to slow). The accuracy statistics need to capture this fundamentally non-gaussian, state-changing behavior. How do models compare against each other? To fully understand the accuracy of a model, IT is necessary to compare multiple models against each other. These models might represent two different physical approaches to modeling the corona (e.g., a potential field source surface [pfss] or magnetohydrodynamic [MHD] approach), or might represent two different versions of the same physical model (e.g., wang-sheeley-arge [WSA] model versions 1.4 and 2.2). As new physics is added into a model or a physical theory is reinterpreted in a model update, robust statistics of how this update is propagated into the solar wind is needed. We believe traditional kirk ET AL. Evaluating and validating heliospheric models against data NNH17ZDA001N-HSWO2R 2 techniques of statistical error analysis are not enough to fully capture the subtle differences between models. In this proposal, we seek to develop a regular set of statistical measurements to evaluate solar wind models that are: resolved in time, not bound to a gaussian distribution assumption, and directly comparable between models. These statistics will empower forecasters and future model developers to fully understand when and how accurately any model reproduces the solar wind.