Project start: 2014

Duration: 3 + 3 years (phase I and phase II)

Funding:   DFG, part of the DFG Research Unit 2131 "Data Assimilation for Improved Characterization of Fluxes across Compartmental interfaces”

Project lead: Harald Kunstmann

Involved scientists: Barbara Haese

Project partners:
        Prof. Dr. Dr. András Bárdossy, University of Stuttgart
        Dr. Sebastian Hörning, University of Queensland, Australia




The main focus of our subproject P3 is on the simulation of the correct spatio-temporal distribution of precipitation.

In phase I of P3, we developed a methodology that combines various precipitation observations into one precipitation field. The observation data used for this purpose come from precipitation stations, radio links in the mobile network and weather radars. The big advantage of our method is that you can simulate with it an ensemble of fields of any size, where each precipitation field reflects the observations.

Example of an ensemble of 50 stochastically simulated precipitation fields; (a) reconstruction of a single precipitation field, (b) uncertainty of the ensemble calculated as difference of the 90th and 10th percentile, (c) example of the variation of precipitation and uncertainty along one commercial microwave link path (CML), (d) variation of the precipitation of the single reconstruction along the CML example compared to the observation.


In the current phase II of the project, we address the question of how these observation-based precipitation fields can improve and influence the prediction of precipitation itself and its dependent variables (eg soil moisture or surface runoff). For this purpose, we are testing various approaches to supply these precipitation fields to a data saturation system for an atmospheric land surface soil model. In the first approach, the model precipitate is simply replaced by the ensemble of observation-based precipitation fields (insertion). Two other approaches use the precipitation fields directly to assimilate the atmosphere. In the first approach, a weakly coupled model and in the second approach a complete model is used. The goal is to develop an efficient method that minimizes errors in predictions.

The scheme shows the various planned data assimilation experiments. Our model platform consists of an atmospheric model, a land surface model and a sub-surface model.


Additional informationen:

FOR2131-web page