Identifying Atmospheric Model Trends and Tendencies using Observations and Analyses

Many atmospheric and oceanic modeling centers perform validation and verification to determine the accuracy of their models and to understand inherent model biases that occur on a daily basis. The US Navy has done similar work in order to provide verification support to the weather forecasters and the warfighter. However, these verification metrics are typically produced for the short term, allowing forecasters to interpret model biases over the recent past, usually the last 30 days or fewer. While extremely valuable, knowledge of model performance over the past month does not address seasonal tendencies of the model to over or under predict occurrences, nor does it address the potential dependence of model bias on climate variations and change. Additionally, the applications of verification are typically focused on real-time forecasting as opposed to retrospective case studies.

Work is currently underway to produce longer-term climatological verification support, referred to as Trends and Tendencies (TnT). The TnT project involves ingesting model, analysis, and observational data in both the recent and distant past, analyzing the data, and producing simple yet informative graphics and metrics for interpretation by weather forecasters, helping them to correctly account for past model biases observed on various time scales. These products can be produced for a plethora of stations throughout the globe for any month, season, or year, as long as archived data is accessible. Through the use of several Python scripts, a few prototypes for TnT have been developed, allowing us to analyze trends for various stations and regions. The TnT metrics will have the potential to be used both operationally for on-the-fly trend analysis and forecasting as well as for retrospective case studies focused on specific regions, seasons, and models. Results from two TnT prototypes are presented along with the methods used to obtain them.