home

Offline SVAT simulations
Last update: 24 March 2004

Contents

Purpose

Multiyear offline simulations with the land surface schemes used in the data assimilation allow an evaluation of the climatological soil water evolution in the area. This climatology is necessary to express the added value of the data assimilation in terms of ability to detect atypical events. Moreover, a systematic difference between the offline climatology and the assimilated time series is a means to detect systematic errors in the model, assuming the forcings to be correct.
Offline simulations have been carried out with the TESSEL scheme. So far, a trial simulation is carried out using a forcing derived from ERA40 by Ulf Hansson (SMHI) for the year 1995. Soon, an ELDAS-based forcing for 2000 will be available.
Here a brief description of the interface between the offline model and the forcing is given, followed by a comparison between the offline TESSEL simulation and ERA40 soil water content in Europe.

The forcing interface

Ulf Hansson has prepared a netCDF-based database of ERA40 values of
  • precipitation
  • 2m temperature
  • 10 m total wind speed
  • lowest level (approx. 10m) specific humidity
  • surface pressure
  • downward longwave and shortwave radiation
This forcing was reformatted to comply with the data format that was used in the RhoneAGG experiment, following ALMA conventions on variable labeling and gridding. Precipitation was distributed over snowfall and rainfall using the 2m temperature as indicator. No attempt was made to interpolate the specific humidity to 2m height. The header of the forcing netCDF file is given here. In this file, a 1 x 1 degree grid is specified between -14 and +38E, and +35 and +72N, at a time resolution of 6 hrs.
Similarly, netCDF files containing the land surface characteristics was created. The header is found here, and the contents of the file are compatible with the data needs of TESSEL. The parameters were all taken from the ECOCLIMAP database file for ELDAS, by picking the closest grid point value.
Initial prognostic variables (soil temperature and water content, snow cover, snow albedo and density, interception reservoir) were given default uniform values for each grid box. The contents of the file containing the intial prognostics is found here.

Example with TESSEL

The standard version of TESSEL was run for 1995. A one-year repetition was executed to estimate an initial state for 1995. The offline code was developed for the various PILPS-like experiments, and updates of the physical parameterization implemented after the RhoneAgg experiment were not considered. In this code, wind speed is interpolated to 2m height using the local roughness length and assuming neutral stratification. The specific humidity provided by Ulf (lowest model level) was considered to be identical at 2m height and not changed. The code interpolates the forcing (proved at 6hrly timesteps) to the model time step (0.5 hour) linearly. Shortwave radiation is not corrected for sunrise/sunset timing in the morning/evening intervals.
This figure  (click to enlarge) shows a time series of total soil water content, averaged for all land grid points in the simulations domain. Shown is a comparison between the offline TESSEL scheme and soil water retrieved from ERA40. ERA40 data were not spatially interpolated, but aggregated to the offline grid by picking the closest land grid point value. The total soil water content is expressed relative to its maximum value in spring, which is slightly higher in ERA40 (0.915m) than in the offline simualtion (0.88m). Both timeseries are simulated with the same TESSEL model formulation, and roughly driven by the same forcings (apart from interpolation issues). The main difference between the two simulations is the initialization (one year repetition run in the offline model comapred to the continuous ERA40 cycle) and the soil adjustments made in ERA40. These keep the soil in a wetter state throughout the summer.

Spatial information per month can be regarded from the animated cycle of plots, accessible via this link (unzip and run on linux with xanim). The overall higher soil water content in ERA40 is clearly seen from this animation, as well as the increase of the difference between the offline simulation and ERA40 from May onwards. It is also evident that the largest differences occur in the south of Europe, where the offline scheme has a more pronounced drying cycle.

Another movie shows the evolution of the soil water profile through the year for both data sets. It is clear that the offset between ERA40 and TESSEL is constantly present in the bottom soil layer, which probably remains at a higher equilibrium water content by absorbing part of the ERA40 soil water increments. The largest deviations between the two data sets occurs in the top layers. A clear contribution of the ERA40 increments to the evolution of the soil water profile can not be easily detected from this animation.

More work will be carried out by running a multiyear suite as well as preparing and running an ELDAS forcing covering 2000 at the original ELDAS 0.2 degree resolution. For comments: contact Bart vd Hurk.