IGACO Data Assimilation

From Earth Science Information Partners (ESIP)

<Bach to IGACO_Framework

Data Assimilation Executive Summary, WMO GAW Report 169

Introduction

IGACO (Integrated Global Atmospheric Chemistry Observations) is a component of the IGOS (Integrated Global Observing Strategy) partnership. When implemented, its purpose is to improve the availability of comprehensive, reliable and accurate information about the changing atmosphere to those responsible for environmental policy development, weather and air quality prediction, and to those responsible for the related research. The IGACO Report defined the rationale, reviewed the past, present and planned global air chemistry observing system and recommended a strategy to integrate ground-based, aircraft and satellite observations using atmospheric models and data assimilation.

ACCENT (Atmospheric Composition Change: the European Network of Excellence) is a network of excellence for atmospheric chemistry, set up under the EU 6th Framework programme. Two ACCENT activities, T&TP (Transport and Transformation of Pollutants in the Troposphere) and AT2 (ACCENT TROPSAT-2), together with the WMO (World Meteorological organisation), sponsored the workshop. The topic, the integration of the observations using comprehensive models and data assimilation, was identified as the most pressing research issue, which needs to be resolved before IGACO can be fully operational.

The resulting workshop on Chemical Data Assimilation for the Observation of the Earth's Atmosphere, sponsored jointly by ACCENT and the WMO Global Atmosphere Watch Programme, was held at the WMO in Geneva from the 24th to the 26th April 2006. The meeting was focussed on data assimilation and observational needs for three classes of modelling applications:

  • Group 1: Numerical weather and air quality forecasting;
  • Group 2: Integration of atmospheric chemistry observations through re-analysis; and
  • Group 3: Inverse modelling (e.g. assessing chemical source/sinks, atmospheric radiative forcing).

The principal recommendations resulting from the scientific discussions are as follows.

Data quality and availability

To make accurate forecasts and provide reliable concentration fields of chemical components, data assimilation techniques requires high quality observational data with as large a temporal and spatial resolution as possible. The recommendations express concerns about obtaining such data, whether groundbased or satellite, the potential future gaps in tropospheric chemical satellite data, and the continued absence of plans for satellites which will yield chemical observations in the middle to lower troposphere on the hourly basis required for air quality. Access to observational data for assimilation and for model evaluation and validation is a continuing concern in this area, as is the quality of observational data.

Data Issues (Group 1: Forecasting)

  • Immediate attention should be given to ensure the continued provision of good quality satellite data of sufficient temporal and spatial resolution in the future.
  • Assimilation for air quality purposes requires at least hourly measurements of key chemical species. Plans should be made to provide such data as soon as possible.
  • Within Europe, there is a need to improve the dialogue between meteorological, research and environmental agencies to facilitate and improve real-time forecasting. Cooperation should aim to support or improve:
    • global chemistry transport models that provide boundary conditions for high resolution regional air quality models;
    • weather & air quality forecasts;
    • collaboration related to attaining EU air quality standards (for example for NO2, fine particles and ozone); and
    • surveillance and reassessment of international agreements (Kyoto Protocol, Montreal Protocol).
  • Since an important aspect of IGACO is to enhance forecasting, there should be chemical data transmission globally in near real time to support the initialization of forecast models on an operational basis. In detail:
    • Transmit routine composition observations (e.g. surface ozone and PM)as well as more comprehensive data taken in campaigns or certain parts of the year on the Global Telecommunications System(GTS) of WMO.
    • More advanced research grade measurements need more care with respect to intellectual property rights. Whatever the case, all data transmitted in real or near real time should comply with accepted data quality protocols.
  • Both ozone and particulate matter data should be regarded as ‘essential’ in the terms of WMO Resolution 40 (unrestricted exchange of meteorological and related data and products)

Do we have sufficient data to do a meaningful reanalysis? (Group 2: Reanalysis)

  • There is an urgent need to perform a reanalysis of total column ozone. These measurements have nearly continuous coverage since 1978.
  • Free access to and standardisation of observational data is needed.
  • There is an urgent need to improve the spatial coverage of surface-based measurements worldwide, and to improve the access to them.

Observations (Group 3: Inverse Modeling)

  • Quantitative retrievals of atmospheric species are required, so effective merging of satellite, air-borne and surface-based in situ and remote sensing observations is needed to obtain the most accurate picture of the concentration fields. (i.e. characterization)
  • High accuracy, high temporal resolution surface concentration measurements remain indispensable, especially close to source regions
  • A surface network of total column measurements is recommended to improve the accuracy of satellite retrievals and their usefulness for models. Information about the precision and accuracy of measurements should always be provided.
  • Measurement sites used for comparison and validation should be as representative of the satellite view as possible.
  • Strategies should be developed to optimise observation systems and retrievals for different chemical species.
  • Access to observations should be improved to facilitate use by the modelling and scientific communities.

What activities can we identify to move forward? (Group 2: Reanalysis)

  • Development of the observation system including:
    • satellites and their sampling strategies (GEO, LEO);
    • ground-based networks with super sites (GAW etc …);
    • regular aircraft flights (IAGOS etc …).
  • The establishment of joint consolidated data bases driven by users.

Assimilation, modelling and forecasting

There are a number of data assimilation techniques and many models in use in this fast developing area for both Global Climate Models (GCMs) and Chemical Transport Models (CTMs). One crucial point of understanding, however, is that the interests of chemical modellers differ from those in numerical weather forecasting. One cannot simply apply the expertise and techniques developed for classical meteorological variables by numerical weather forecasting research to chemical data assimilation, There is much research needed before chemical data assimilation methods can be applied routinely.

At present, most chemical data assimilation work is carried out by small research groups, yet it is a very large task to put models and data together. The experience from numerical weather forecasting suggests that, at some stage, there must be greater effort toward chemical data assimilation, either through distributed research projects or the establishment of a research institute .

Forecasting and data assimilation (Group 1: Forecasting)

  • Chemical forecasting can cover a range of timescales from hours for air quality to months for ecosystems. The value of Chemistry Modelling and Data Assimilation should be explored for both reanalysis (see Group 2: Reanalysis) and for seasonal forecasts.
  • The differences in priorities between chemical forecasting and weather forecasting should be recognised and appreciated. Chemical data assimilation should concentrate on processes for which there is both a paucity of knowledge and a perceived importance, such as:
    • emission rates,
    • wet and dry removal, and
    • planetary boundary layer structure and mixing.

Modelling needs (Group 1: Forecasting)

  • With the vast amount of work associated with a mature meteorological data assimilation configuration, we recommend that some means be found to facilitate cooperation in research on chemical data assimilation.
  • Some fundamental aspects of modelling itself, such as numerical methods and boundary conditions, still require detailed attention.
  • There are still diverse views on the minimum chemical configuration required for adequate forecasting – attention should be given to defining such a minimum configuration for practical forecasting.
  • The optimum trade off between complexity, resolution and speed needs definition for practical forecasting.

How good are the CTMs or GCCMs? (Group 2: Reanalysis)

  • Intercomparisons of Global Climate Models (GCM) and Chemical Transport Models (CTMs) from the point of view of tropospheric chemistry are required.
  • Integration of boundary layer, land surface and meteorology within the models should be a high priority.
  • In the context of data assimilation, there is a need to assess the value of treating the chemical problem with online or offline approaches.
  • Currently CTMs are limited by realistic temporal and spatial emissions, and unexpected release of pollutants (e.g. forest fires); there is a need to develop better representations of emissions for models and forecasting.
  • There is a need to develop strong collaboration between the communities involved in numerical weather prediction and atmospheric modelling

How good are chemical assimilation methods and tools? (Group 2: Reanalysis)

  • Despite recent progress, there is a need to advance and assess the usefulness of different assimilation methods in the context of atmospheric chemistry. This assessment will also depend on the application, the targeted species and the time scale.
  • Multivariate chemical error co-variances should be developed to address the problem of chemical balance in analyses.
  • Emissions error co-variances should be developed and assessed for inverse.
  • Since a large fraction of the observations consists of retrievals, there is a need to identify systematic and random errors (i.e. precision and accuracy) in such observations, in forward model parameterizations and in the assumptions inherent in retrieval algorithms.
  • The assimilation of multi-sensor observations should be explored as a way to identify systematic error.
  • The usefulness and reliability of simultaneous estimation of source (boundary condition) and initial condition should be assessed by data assimilation methodology.

Inverse model development (Group 3: Inverse Modeling)

  • Intercomparisons of results from the various inverse modelling techniques are required.
  • The use of multi-species inversions should be advanced.
  • Use of coupled surface-atmosphere models for species with strong surface exchange should be further explored.
  • The development of the use of statistical tests for error assessments of the components of the inverse model system is required.

The role of transport models (Group 3: Inverse Modeling)

  • A key problem is the uncertainty in the prediction of the height and dynamics of the planetary boundary layer (PBL). Consideration must be given to tracer model development, in particular for PBL dynamics, parameterization of convection, and removal of particles and gases by wet and dry transformation and scavenging processes. Errors from tracer transport computations must be quantified.
  • Model intercomparisons supported by a well documented experimental database are recommended.
  • Tracer release experiments are strongly recommended for providing validation data for Chemical Transport Models (CTMs), and to support development of improved models.

Validation (Group 1: Forecasting)

  • Validation should be carried out regularly as models and skills develop. Suggested approaches include the following:
    • controlled inter-comparison exercises with agreed forecast protocols, and verification metrics;
    • a transposed approach similar to that used by AMIP which is a standard experimental protocol for global atmospheric general circulation models; it enables scientists to focus on the atmospheric model without the added complexity of ocean-atmosphere feedbacks in the climate system
    • case studies of extreme cases, where campaign data collected by NILU and others serves as a basis;
    • research quality data, especially profile data, is essential (sondes, lidars, aircraft).

What activities will move us forward? (Group 2: Reanalysis)

  • The encouragement of strong collaboration between operational numerical weather centers and science groups.
  • The provision of adequate, dedicated computing resources to carry out chemical data assimilation
  • Continued model integration into earth system models to improve:
    • stratosphere-troposphere coupling;
    • gas-phase and aerosol coupling; and
    • global and regional coupling.

Emissions

Tropospheric chemical models rely heavily on adequate emission inventories, and the paucity, age and unreliability of some of these has been a weak point for all models in the past. While data assimilation can assist in obtaining a better picture of emissions this boot-strapping process, still requires a reliable and up to date emission inventory.

Emissions (Group 1: Forecasting)

  • Global and regional emission inventories need to be reassessed, not only with respect to anthropogenic sources but also for the biogenic sources. Two particular research directions to pursue simultaneously are:
    • assessment of emissions in an extended reanalysis would be valuable;
    • definition of initial conditions and bottom-up estimates of emissions.

Emission fluxes (Group 3: Inverse Modeling)

  • The propagation of errors from a-priori emission assumptions must be quantified since emissions are a determining factor for the quality of the results from inverse modelling.
  • Better communication is needed between inverse modellers and the research emission community, so that better data with sensible error estimates can be provided.