Harrie-Jan Hendricks-Franssen, Forschungszentrum Jülich, Institute of Bio- and Geosciences, Germany
Insa Neuweiler, Leibniz Universität Hannover, Institute of Fluid Mechanics, Germany
Matteo Camporese, University of Padova, Italy
Model predictions of water flow and related transport are essential for many applications, ranging from weather forecast over flood and drought prediction to managing of groundwater resources. The optimal merging of measurement data and terrestrial system model simulations via data assimilation is essential to improve model predictions in all terrestrial compartments. This concerns improvement of short term predictions by updating initial states. For long term predictions, estimating model parameters is also of high importance. Data assimilation can also play an important role in improving models by detecting model structural errors on the basis of systematic increments.
There are many challenges and open questions concerning data assimilation in order to improve model predictions. Although the challenges are different for different compartments, there are many challenges that apply to all of them. Methodological developments are for example needed to better handle non-Gaussian distributed model states and parameters or to obtain more robust results if only small ensembles can be calculated. The fast increasing amount of available data and new data types offers the opportunity to reduce prediction uncertainty. For this purpose, better multivariate data assimilation strategies need to be developed and tested, new observation operators are needed and the filtering and compressing of large amounts of measurement data becomes essential.
While in the past data assimilation focused on models for single compartments (e.g., the land surface), increasingly data assimilation is applied in combination with coupled models that comprise multiple compartments (e.g, land surface and subsurface). Different types of measurements can be jointly assimilated to update predictions with coupled models, but many questions remain about the best data assimilation strategy, as processes in coupled models act on very different spatial and temporal scales. Also, the processes at the interfaces are in particular difficult to capture in models. The question how to deal with possible model errors needs to be addressed. Strategies for efficient computing and parallelization are especially for those models of high importance.
This session invites contributions on sequential data assimilation for improved terrestrial system model predictions. The contributions can for example be concerned with new methodological developments for data assimilation, assimilation of new data types and the development of measurement operators, and the simultaneous assimilation of various different data types measured at different spatial scales. Uncertainty characterization and ensemble generation are critical elements in data assimilation studies and improvement in these aspects is of special interest for this session. Also of special interest are the development and evaluation of data assimilation strategies for coupled models, including studies which include the atmospheric compartment (e.g., atmosphere-land surface models) or coupled modelling of water and biogeochemical cycles. Finally, contributions showing applications in real-world case studies with verification of predictions, or case studies that highlight the challenges and the demand of future research directions are of high interest as well. We hope to receive contributions from subsurface hydrology as well as surface hydrology, land surface modeling, ecohydrology and atmospheric sciences.