Original paper

Strategies for soil initialization of regional decadal climate predictions

Kothe, Steffen; Tödter, Julian; Ahrens, Bodo

Meteorologische Zeitschrift (2015)

74 references

published online: Jul 6, 2016
manuscript accepted: Dec 23, 2015
manuscript revision received: Oct 23, 2015
manuscript revision requested: Sep 16, 2015
manuscript received: Aug 3, 2015

DOI: 10.1127/metz/2016/0729

BibTeX file

O

Open Access (paper can be downloaded for free)

Download paper for free

Abstract

The deep soil shows a long-term climate memory. Thus, the initialization method of the soil state in climate simulations potentially has an impact on climate predictions. This study focuses on regional decadal climate predictions with the model COSMO-CLM for Africa and Europe, driven by the global climate model MPI-ESM. The impacts of five soil initialization methods of different complexity are compared and assessed against 2 m-temperature and precipitation observations. Even though the results are heterogeneous in space and time with high uncertainties, some basic conclusions can be drawn. The simplest approach, i.e. interpolating the soil initial fields of the driving global climate model, is worst. The interpolation of soil data from a re-analysis product (here ERA-Interim) or extracting the initial state from a long-term spin-up simulation with COSMO-CLM driven by ERA-Interim give better results. Another approach extracts the initial state from a long-term spin-up simulation with COSMO-CLM's offline soil model TERRA-ML driven by gridded and improved observational data (here the WATCH data). The additional assimilation of satellite-based surface soil moisture data into this TERRA-ML simulation further improves the climate prediction in some regions. In conclusion, decadal climate prediction systems with sophisticated soil initialization schemes have the potential to make use of the soil's long-term memory. Most promising, but also most costly, is deep soil initialization by means of data assimilation. Remaining challenges are the persisting inconsistencies between driving data, assimilated observations, and soil model.

Keywords

COSMO-CLMregional climate modeldecadal predictionsoil initializationdata assimilationsatellite observations