Development of a European Land Data Assimilation
System to Predict floods and droughts (ELDAS)
Coordinated by B.J.J.M. van den Hurk
KNMI, PO Box 201, 3730 AE De Bilt, The Netherlands; hurkvd@knmi.nl
Introduction
On 1 December 2001 the EU-funded project Development of a European Land Data
Assimilation System to Predict floods and droughts (ELDAS) was officially started.
ELDAS aims at the design and implementation of a system for soil moisture data
assimilation in a Numerical Weather Prediction (NWP) environment, and to assess its
added value for improving hydrology-related forecasts such as floods and seasonal
drought events. This document provides a description of structure, goals and work
program of ELDAS. Since the project is still in its initialization stage, a review of the
current plans by the Global Land Atmosphere System Study (GLASS) consortium is
desirable.
Motivation and background
ELDAS is strongly inspired by a number of recent activities in the area of land data
assimilation and related modeling and observational programs:
 The US Land Data Assimilation System (LDAS) and its Global LDAS
(GLDAS) counterpart not only gave inspiration for the ELDAS acronym.
ELDAS was originally drafted as a European contribution to this (G)LDAS
program, which was considered desirable in order to
o Develop European expertise in land data assimilation
o Provide European forcing and validation observations to the GLDAS
archive, considered to be useful if (G/E)LDAS products are considered
to form a long term data base for climate monitoring purposes
o Extend GLDAS methodologies by expanding on the number of groups,
models and assimilation procedures involved in the land data
assimilation work
 A number of European NWP centers have developed (semi-)operational soil
assimilation systems, and a combination of existing experience was warranted
 A number of near-future satellite observation programs (METEOSAT Second
Generation, SMOS) make preparation for the data that will become available
necessary
 Long term meteorological (re-)analysis archives such as ERA15 have been
used extensively for analyzing the performance of model components or
trends in past climate patterns. An improved system for soil moisture
assimilation may be a valuable means for extending this work towards the
analysis of the quality of land surface process simulations or the role of land
surface in the regional or global hydrological cycle.
The European Union program for support of Research and Technological
Development (RTD) was considered as the optimal platform for applying for financial
support. EU applications are usually only successful when a strong social-economic
component is present. The combination of the developments in the scientific fields
and EU requirements have resulted in an ELDAS-project that has the following key
elements:
 It combines European expertise on (operational) soil moisture data
assimilation to build a prototype of a common data assimilation infrastructure
 It anticipates on using new observations from present and future satellite
platforms (METEOSAT/MSG, SMOS)
 The infrastructure is implemented in a number of NWP-centers, to make
operational implementation after the end of the ELDAS project
straightforward
 ELDAS combines the development of the data assimilation structure with a
wide range of validation and case studies, to demonstrate the potential benefits
of the system for routine NWP forecasting, prediction of floods or strong
drought events, and for the detection of systematic errors in current land
surface parameterization schemes used for NWP and climate modeling.
The focus on assessing the added value of ELDAS products for operational tasks has
lead to the definition of a so-called Combined RTD/Demonstration project, in which a
number of case studies are being considered as demonstration tasks (for which less
EU support is made available).
Main deliverables
ELDAS is not designed to run as a near-realtime operational scheme on a European or
global scale within the timeframe of the project. Instead, it will result in a common
data assimilation infrastructure at a number of NWP-centers, and a number of
validation and case studies, addressing the value of the resulting soil fields. For these
case studies, a number of dedicated forcing data sets will be assembled, which will
help to define the optimal use of available observations once the systems will run
operationally. One central dataset will cover the whole European area for at least one
annual cycle at a resolution of 0.2. Soil moisture fields derived from this data set will
be analyzed in an extensive NWP experiment. Apart from the data sets and designed
system implementations, the main deliverables will consist of assessments of the
added value of ELDAS soil moisture fields for operational forecasting and climate
research purposes.
Structure of the project
As usual, the work in the project is subdivided in a number of workpackages. These
are lumped into 5 categories (see Figure 1):
 WP1000: organization and coordination
 WP2000: preparation of databases
 WP3000: Set-up of infrastructure and generation of soil moisture fields
 WP4000: Validation studies
 WP5000: Case studies (Demonstration component).
A brief description of each group of workpackages is given below.
Figure 1: Organizational layout of ELDAS
Infrastructure
Soil moisture fields will be generated by making use of an independent NWP forecast
suite in which first guess values of precipitation and surface radiation are replaced by
observed/analyzed fields (see below). In addition, corrections to the simulated soil
moisture content will be applied by transferring forecast errors in near surface
quantities (temperature and relative humidity) and surface temperature heating rates
into soil moisture increments. The whole system will be operated in a fully coupled
mode, that is, a land surface scheme coupled to an atmospheric model, to account for
the impact of surface-atmosphere feedback on the soil moisture increments (see
Figure 2).
The data assimilation procedure will consist of a simplified column-by-column
variational optimization of a cost function, measuring the distance of the first guess to
both the observations and to the background field. The background error is adjusted
dynamically using a simplified Kalman filter (Hess, 2001). The minimization makes
use of the assumption that close to the optimum, the gradient of the cost function with
respect to the control variables may be assumed to be linear. This implies that these
gradients may be obtained from a finite difference approach, making only one
additional model integration for each control variable necessary.
Soil moisture fields are produced for a suite of land surface schemes, to acknowledge
the strong dependence of the assimilated fields on the model structure. This will be
achieved by distributing the soil moisture production system over three different
institutes: ECMWF, the German Weather Service (DWD) and Meteo France. By this
strategy soil moisture fields for the ECMWF land surface scheme TESSEL, the
surface scheme in the DWD Lokal Modell, and the ISBA surface scheme will be
obtained. The involvement of these schemes varies per validation/case study
workpackage. The handling of forcing observations, definition of the background
error and data assimilation procedure, and the specification of target domains and
time frames have been harmonized for studies in which results for these schemes will
be intercompared.
Figure 2: Schematic layout of the ELDAS data assimilation system
Observations
A number of demonstration databases will be constructed in the context of ELDAS.
The first one will cover the whole of Europe for a 15-month period between 1
October 1999 and 31 December 2000. When MSG will be flying, a 2nd database is
constructed near real-time during the project, covering the growing season in 2003.
2100: Precipitation
The basic precipitation product is an objective analysis at 1/6º resolution, based on an
optimum interpolation of bias corrected gauge measurements from 10000 stations
spread over Europe. Sub-daily fields will be derived from this for various areas in
Europe where a transnational qualitative radar database is available that can be used
to disaggregate the daily sums to 3-hrly intervals. For two case studies (5100 and
5400) quantitative radar derived precipitation products will be combined with the
gauge analysis.
2200: Surface radiation
Surface radiation will be derived at 3-hrly intervals from operational short-term
forecasts from the ECMWF model, corrected for observed cloud properties retrieved
from METEOSAT/MSG satellite data. A one-year demonstration database over
Europe at approximately 20km resolution will be constructed in this workpackage. In
addition, a direct surface shortwave radiation estimate from METEOSAT/MSG data
is obtained using a simplified radiation transfer scheme (McNider et al, 1995).
2300: Surface heating rates
The rate at which surface temperature increases after sunrise is partially dependent on
the moisture condition of the surface. Based on this principle, soil moisture
corrections will be obtained from surface heating rates retrieved from
METEOSAT/MSG data archives. Particular attention is paid to the atmospheric
correction of the surface temperature data, cloud screening and the definition of
observation errors. A one-year archive of heating rates for the European continent will
be constructed.
2400: Atmospheric forcing fields
The soil assimilation will make use of analyzed fields of near surface temperature and
relative humidity (see below). These fields will be taken from the operational surface
analysis carried out at ECMWF. In addition, upper air analyzed fields from ECMWF
will be used as lateral boundary condition for the number of limited area model
experiments explored during ELDAS.
Validation studies
Four workpackages are devoted to validation of the products.
4100: GSWP validation
The emphasis in this workpackage will be on validation of the 2000-dataset using
methodologies developed in the context of the Global Soil Wetness Project.
Validation data will be gathered from river discharge records, flux tower and
scintillometer stations scattered across the domain, in situ soil moisture records and
synoptic observations.
4200: SSM/I or AMSR validation
Using radiative transfer models, the top of atmosphere microwave radiation will be
generated from the surface state in the data assimilation modules. These will be
compared to the record of SSM/I measurements for the 2000 data base, and possibly
with AMSR data for the 2003 database. SSM/I is not an optimal sensor for soil
moisture detection, but careful site selection and signal filtering has given rise to
valuable comparison material from recent studies.
4300: GLDAS
In this workpackage, results from the N/GLDAS system will be compared to the
outcome from the ELDAS system. N/GLDAS forcings will be fed to the ELDAS data
assimilation system, and the resulting hydrological budgets from a number of major
American river basins will be evaluated.
4400: French river basins
Starting from the Rhone basin (as evaluated in the so-called RhoneAGG experiment),
CNRM/Meteo France has operationalized the production of surface forcings for a
dense 8km grid over France. River discharge observations from major river basins are
available for evaluation of the hydrological balance calculated by the surface schemes
part of the ELDAS system. A control simulation of the surface hydrology by the
surface schemes without data assimilation allows to evaluate the added value of the
data assimilation component of the land surface simulations.
Case studies
In a number of case studies, semi-operational practice cases are used to evaluate how
ELDAS will possible improve routine results.
5100: UK Floodings
The flooding events in the autumn of 2000 in the UK were partially affected by a
combination of high initial soil water content and excessive precipitation. The
workpackage will focus on how ELDAS may enable a proper estimate of the initial
soil moisture content prior to the precipitation events, aiming at an improved timing
of the actual floodings.
5200: Rhine floodings
In this workpackage two parallel NWP cycles will be operated, one with and one
without soil moisture adjustments. Again the skill of the runoff generated by the NWP
cycles will be compared to the rover discharge of the Rhine river for a number of
events that still have to be selected.
5300: European NWP cycles
Similarly, a number of parallel NWP cycles will be carried out overlying the entire
European continent, and covering the full 15-month period in 1999-2000. The focus
of this experiment will be the skill of ELDAS to produce an adequate seasonal cycle
in the land surface moisture conditions, aiming at better predictions or diagnosis of
period with relative drought. ELDAS will be compared to current practice of soil
moisture assimilation based on synoptic data solely.
5400: Climate hydrology in the Baltic area
This workpackage aims at assessing the way soil moisture data assimilation can be
used to diagnose systematic errors in the participating surface schemes. Systematic
errors in for instance the runoff fraction will show up as a persistent need for soil
moisture corrections. Initially, the workpackage is formulated to focus on the Baltic
area, but the analysis will likely be carried out for a larger area in the ELDAS domain.
Project partners, meetings and logistics
ELDAS is coordinated by KNMI, and a total of 10 institutes are participating. In
addition, subcontracts with 2 additional organizations have been included. A full
description of the participating institutes can be found at the ELDAS web-page,
http://www.knmi.nl/samenw/eldas, which also lists the contact persons.
The project will organize a yearly progress meeting, associated with a special
workshop in a number of cases. Workshops devoted to microwave data assimilation
and general land data assimilation aspects are foreseen to be organized partly in the
context of ELDAS.
More information
More information on ELDAS can be obtained from the web-site of ELDAS
(http://www.knmi.nl/samenw/eldas), or via the coordinator Bart van den Hurk,
hurkvd@knmi.nl.