The goal of this project is the consistent aggregation of experimental data, simulation, and additional information about the soil water movement for two experimental test sites: ASSESS and Grenzhof.


Monitoring soil-water dynamics with Ground Penetrating Radar

The measurement of dynamic processes like infiltration or drainage is a key to quantification of the soil as soil-water dynamics is highly dependent on properties and architecture of the soil. Especially infiltration in sandy soils includes a number of challenging phenomena that range from the strong non-linearity through multiscale heterogeneity to non-equilibrium phenomena like fingering.

Most experiments take place at the ASSESS-Site which consists of three different sandy materials. The measurement method includes radar measurement in frequencies between 200 MHz and 600 MHz with an automated GPR-scanner as well as TDR measurements at different positions and depths of the soil material.

Contact: Lisa Hantschel

Workshop Contribution to "Terrestrial systems - frontiers of our understanding", September 21 - 24, 2015, Freudenstadt, Germany:













Managing Uncertainties

The representation of soil water movement faces uncertainties in all components: states (water content), dynamics (propagating the water content), subscale physics (material properties and their corresponding parameters and architecture) and forcing (boundary conditions and initial state).
As the state can be measured (see above) and calculated with dynamics, subscale physics and forcing, the redundant information has the potential to reduce uncertainties in all components.

We employ and enhance a data assimilation method, the Ensemble Kalman Filter (EnKF), to estimate states, parameters, architecture, and boundary condition based on water content measurements - currently from the Grenzhof test site.

Contact: Hannes Bauser

Conference contribution to AGU Fall Meeting, December 14 - 17, 2015, San Francisco, USA:

Improving the Method

In soil hydrology we have highly nonlinear processes, sharp fronts, and discontinuities across layer boundaries. This leads to non-Gaussian probability distributions, which is at odds with the assumption of Gaussian probability density functions (pdf) of the EnKF.

Particle filters do not rely on a Gaussian assumption and are capable to handle arbitrary pdfs. In contrast the EnKF does not suffer from the 'curse of dimensionality'. We try to improve our data assimilation methods by using a combination of a particle filter and an EnKF combining the strength of both methods.

Contact: Daniel Berg

Conference contribution to AGU Fall Meeting, December 14 - 17, 2015, San Francisco, USA:


Deutsche Forschungsgemeinschaft (DFG) funds this research through project RO 1080/12-1.