Using a suite of mesoscale models, we show the forecast skills for 2-meter temperature over the continental United States. In this study, we use models and/or gridded fields such as the mesoETA, WRF-ARW, NDFD, and their ensemble mean. Our final resolution for the multimodel Superensemble is at 32 km at 6-hourly temporal intervals. The forecast length is 60 hours.

This study utilizes an optimization for the training length (number of days) to arrive at the best results. The period of the study includes several past seasons. The main metric for the forecast evaluation is the mean absolute error. Secondary measures include percent improvement as well as consistency of least error from case to case. The results show a major improvement for the deployment of the multimodel superensemble, which is shown to outperform the member models and their ensemble mean.