Frequently Cycled Data Assimilation with Global MPAS at Convective-Allowing Resolution

In August 2018, The Weather Company (TWC) operationalized a custom version of the NCAR Model for Prediction Across Scales (MPAS) at uniform 15-km resolution, which runs out to 72 hours (and experimentally out to 144 hours) to replace a 13-km WRF-based global system. The 4x-daily forecasts are initialized with the 0.25-degree NCEP GFS analyses, NASA SPoRT 2-km SST, and 4-km NESDIS VIIRS green vegetation fraction data.

In parallel, TWC is developing an hourly-updating “Global High Resolution Atmospheric Forecasting” system, based on variable resolution (15/3-km) MPAS and the Gridpoint Statistical Interpolation data assimilation software. The forecasts target 15-hour lead times, in part to drive TWC’s short-term “Forecast On-Demand” and “Currents On-Demand” capabilities. To facilitate rapid improvement of this system, a re-forecast environment evaluates how changes to the model or data assimilation impact forecast quality. Verification is performed through a combination of the DTC Model Evaluation Tools for gridded precipitation and point comparisons via a custom database.

In this presentation, we will describe the development and implementation of the (i) data assimilation, (ii) re-forecast, and (iii) verification frameworks. We will also summarize the assimilation experiment results to date and discuss the implementation of a cycled forecast system and the acquisition of low-latency observations.