Ensemble Dressing of North American Land Data Assimilation version 2 (EDN2)

Most datasets of surface meteorology are deterministic, yet many applications using these datasets require or can benefit from uncertainty estimates in meteorological fields. Motivated by this gap, we applied a locally-weighted spatial regression technique with the widely-used North American Land Data Assimilation version 2 (NLDAS-2) dataset values to generate ensemble estimates for daily precipitation, daily mean temperature, and diurnal temperature range. The approach is a form of ensemble dressing. This uncertainty dataset and methods from this work are made publicly available to support research such a data assimilation or model uncertainty studies.

The dataset includes a 100-member ensemble for daily precipitation, temperature and diurnal temperature range at 1/8th degree for the NLDAS-2 domain (25 to 53 North, 125 to 67 West), for the time period 1979-2019. It also includes the spatial regression coefficients and other inputs needed to run the Gridded Meteorological Ensemble Tool (GMET) used to generate the ensembles. A limited number of summary statistical analyses of the dataset are also included.