5. Data assimilation development and evaluation
Errors in data assimilation estimate can originate from various sources. This sub-theme discusses the impacts of uncertainties in model parameters, biases in forcing data, and errors in assimilated observations. In addition, a series of data assimilation algorithms (EnKF, PF, and 4DVAR, etc.) have been widely investigated, and their efficiency and applicability are presented. This sub-theme also discusses the status and plans for community-based open-source tools for hydrologic and land DA to support such investigations.
|Omar Ghattas, UT-Austin
|Leila Farhadi, The George Washington University: “Development of Data Assimilation Techniques for Hydrological Applications”
|Hydrologic data assimilation aims to utilize the knowledge of hydrological process as embodied in a hydrologic model and the information gained from observations. Model predictions and observations are both imperfect and contain different kinds of information. Nonetheless, when used together, they provide an accuracy level that cannot be obtained when used separately. This process is extremely valuable for providing initial conditions for hydrological system prediction and/or correcting hydrological system prediction, and improving parameterization of hydrological system.
In this presentation recent developments of data assimilation in hydrology are briefly reviewed and development of two new data assimilation algorithms for two different case studies is introduced.
In the first case, an F/T assimilation algorithm is developed for the NASA Goddard Earth Observing System, version 5 (GEOS-5) modeling and assimilation framework. The algorithm includes a newly developed observation operator that diagnoses the landscape F/T state in the GEOS-5 Catchment land surface model. The F/T analysis is a rule-based approach that adjusts Catchment model state variables in response to binary F/T observations, while also considering forecast and observation errors. To evaluate the accuracy of the model, a regional observing system simulation experiment is conducted using synthetically generated F/T observations. For F/T classification errors below 20%, the assimilation of F/T observations reduced the root-mean-square errors (RMSE) of surface temperature and soil temperature when compared to model estimates. This F/T assimilation scheme is being developed to exploit planned operational F/T products from the NASA Soil Moisture Active Passive (SMAP) mission.
In the second case, a data assimilation approach is developed for estimating key parameters governing moisture and heat diffusion equation and the closure function which links these equations. Parameters of the system are estimated by developing objective functions that link atmospheric forcing, surface states, and unknown parameters. A single objective function is expressed that measures moisture and temperature dependent errors solely in terms of observed forcings and surface states. This objective function is minimized with respect to the parameters to identify evaporation and drainage models and estimate water and energy balance flux components. Uncertainty of parameter estimates is obtained from the inverse of Hessian of the objective function, which is an approximation of the error covariance matrix. Uncertainty analysis and analysis of the covariance approximation, guides the formulation of a well-posed estimation problem. Accuracy of this method is examined through its application over different field sites. The applicability of this approach to diverse climates and land surface conditions with different spatial scales, using remotely sensed measurements is also presented.
|In this talk I will summarize recent advances in Particle Markov chain Monte Carlo (PMCMC) simulation for environmental (hydrologic) data assimilation. PMCMC uses a ensemble of particles (trajectories) to approximate the state and output forecast distribution and avoids sampling collapse by periodic MCMC resampling steps. This approach allows for joint parameter and state estimation but care should be exercised how to estimate the model parameters. I will illustrate some preliminary results and discuss the advantages / limitations of the proposed methodology for operational real-time forecasting.
|Kun Yang, ITP/Chinese Academy of Sciences, Beijing, China: “Microwave Data Assimilation Practice in the Tibetan Plateau and Its Implication for Land Hydrological Modeling”
|Sujay V Kumar, NASA GSFC/SAIC, USA: “Multivariate Assimilation of Satellite-derived Remote Sensing Datasets in the North American Land Data Assimilation System (NLDAS)”
|Feng Huang, China Agricultural University, China: “Investigating Crop Water Productivity by Using Remotely Sensed Approach for Summer Maize at Hebei Plain in North China”