The Latest from the Data Assimilation Research Testbed: New Algorithms for Non-Gaussian Distributions and Sampling Errors; New Memory Management for Larger Models and Faster Execution; New Model and Observation Interfaces for Land, Upper Atmosphere, and Ocean Biogeochemistry.

The Data Assimilation Research Testbed (DART) is a community facility for ensembledata assimilation developed and maintained by the National Center for AtmosphericResearch (NCAR). This poster highlights new capabilities that have been added toDART in the last year that are useful to researchers doing data assimilation for Earthsystem applications:

Completion of a decade of CAM6+DART reanalysis, which provides ensembleatmospheric forcing for surface models, and more.

New or improved model interfaces include: TIEGCM, MITgcm_ocean NBLING(biogeochemistry), a low order tracer advection model (student contribution), andCAM Spectral Element.

New observation interfaces include: snow water equivalent, soil moisture, with SolarInduced Fluorescence (SIF) and snow cover in development.

Efficiency gains and handling higher resolution models by compaction of sparsestate vectors in ocean and land models, and improved caching (studentcontributions).

Quantile conserving and quantile regressing filters support a wide range of non-Gaussian and nonlinear ensemble updates.

A randomized dormant ensemble filter deals with sampling errors.

More flexibility in defining observation quantities, localization, and model interfacebuilding.

Upgraded web site documentation search and a CLM5+DART tutorial.

Improved soil moisture estimates over China using CLM+DART and soil moistureobservations (ESA-CCI).