Ensemble data assimilation for soil-vegetation-atmosphere systems

There are many open research questions in the field of data assimilation for land surface models. We know that all models are imperfect and are probably biased. We also know that our knowledge of the initial conditions and forcing for those models is imperfect. The observations of interest may not be represented explicitly in the model and both the models and the observations have uncertainties and differences in representativeness. The observations may be biased. An effective data assimilation system must address all of these while producing a model state that contains the information that may be derived from those observations.

A successful ensemble data assimilation system must capture and reflect the uncertainty in our knowledge of the system being modeled as well as the variability of the system itself. Land surface processes are challenging in this regard given the tremendous heterogeneity of the land surface and the range of scales of interest; from individual plants to watersheds to continental-scale responses. Furthermore, the equations governing ecological processes are not nearly as well-defined as those for atmospheric modeling, for example. Our goal is to produce an ensemble of land surface states that results in a skillful short-term forecast. The accuracy of this forecast is our measure of the success of our ensemble system. A good forecast is believed to depend on a good initial state and accurate model dynamics and so is a challenging measure of success.

The Data Assimilation Research Testbed (DART) is a community facility for ensemble data assimilation developed and maintained at the National Center for Atmospheric Research (NCAR). DART is a software environment that makes it easy to explore a variety of data assimilation methods and observations with different numerical models and is designed to facilitate the combination of assimilation algorithms, models, and real (as well as synthetic) observations to allow increased understanding of all three. Land surface models supported by DART are the Community Land Model (CLM) and the uncoupled mode of the Noah Land Surface Model (Noah LSM). This talk will present an overview and results of several assimilation experiments and summarize the challenges and future direction of land data assimilation research. 1) CLM is used to assimilate MODIS snow cover fraction observations with the goal of improving daily estimates of snow water equivalent. 2) Noah LSM is used at a single site to assimilate hourly soil moisture estimates from a neutron probe and verified against (withheld) in-situ soil moisture estimates. 3) CLM is also used at a single site to assimilate flux tower observations and is compared to open-loop simulations. 4) The ability to assimilate observations into CLM in a fully coupled earth system model framework is demonstrated. 5) Recent efforts into supporting AMSR-E brightness temperatures and GRACE water storage observations will be discussed as time permits.