Transcript large ( 10,4 MB) - LCRS - Laboratory for Climatology and Remote

EU COST Action 722: The Understanding of Fog Structure, Development and Forecasting
Presented on behalf of the 13 participant countries
1. Background.
In September 2001, the EU initiated a European Concerted Research Action designated as COST Action 722, and entitled "Short range forecasting methods of fog, visibility and low clouds".
Currently thirteen European countries participate, Austria, Cyprus, Denmark, Finland, France, Germany, Hungary, Poland, Spain, Sweden, Switzerland, UK. The main objective of the Action is to
develop advanced methods for very short-range forecasts of fog, visibility and low clouds, adapted to characteristic areas and to user requirements. This overall objective includes:
the development of pre-processing methods of the necessary input data,
the development of appropriate forecast models and methods,
the development of adaptable application software for the production of the forecasts.
A first phase report, COST 722 (2003), an “Inventory”of current science and techniques was completed by Autumn 2003 and is to be followed the second, “Research and Development”, phase by
2005, with completion of “Development”and “Dissemination”phases by 2006/7.
The development of visibility, fog and low clouds is influenced by many parameters (e.g. cloudiness, temperature, humidity, wind speed, topography, vegetation and radiation). Fog is especially
very sensitive to small changes in the relevant meteorological parameters in the lowest layers of the atmosphere, so that very high-quality measurements are required. Therefore, high forecast
skill is dependent on accurate knowledge of the vertical distribution of the relevant input parameters.
2. Phase 1 Conclusions.
The conclusions emphasize that a deeper understanding of the relevant processes of the development of fog, visibility and low clouds is required for improved forecasting. NWP models are able
to represent the large scale forcing on many occasions, but are not sensitive to the local meteorological and topographical parameters. 1-D models can be used for local forecasts, but only fine
scale 3-D models have the potential to consider all relevant processes. The “inventory”includes a need for increased spatial and temporal data, and an observation that statistical methods can
be useful for prediction purposes. The full COST 722 Phase 1 report entitled “Very short range forecasting of fog and low clouds: Inventory phase on current knowledge and requirements by
forecasters and users” is available on the website: http://137.248.191.94/cost/publications.html. The analysis includes a study of all commonly used forecasting models in the EU.
3. Current Work, Phase 2.
This phase includes three main areas:
“Initial Data”.
This area will include a study of incorporating all available climatologies, surface and remote sensing data, and the ability to differentiate between low clouds and fog.
“Models”.
The objective is to examine the potential of different models, investigate the mechanisms and improve model physics, with the possibility of developing probability
forecasts
and method validation.
“Statistical Methods”.
Limitations and potential of existing techniques, data requirements, improvement and portability are important areas.
An observational programme based on the collection of data from the airport Paris CDG, will be used for testing and comparison of different techniques.
It is not possible to include all the current activities, but a sample of the undergoing work is presented below.
Results from use of the ALADIN model for winters 2002/3 and 2003/4 have shown that the systematic` underprediction of the sharp inversions and the use of a cloudiness scheme which produces too little cloud. Figures 1(a)
Observational experiment at Paris CDG airport
Meteorological tower of 30m :
and 1(b) show how an improved sub cloud scheme can help.
T / Hu%
Since December 2002
Ground measurements :
T / W inside the soil (between ground
Airport terminal 1:
and –
50cm)
T / H%
short- and long-wave radiations
Radiation fluxes
Fig 1(a)
Fig 1(b)
Figure 1 : Predicted (red) and observed (blue)
soundings in Vienna during a typical stratus episode.
Results obtained (a) with reference cloudiness
scheme, (b) with improved sub-inversion cloudiness
scheme. Because of cloud-radiation feedback the
improved scheme leads to a sharper inversion
structure, in better agreement with observations.
Comparison of results using the UK Met Office model for different resoltions of 12km, 4km and 1km.
Figure 2. Comments:
Model with 1km grid qualitatively better and more consistent
with observation.
1km model quantitatively better for 10.12.2003 case study
Models with 1km and 4km grids produce and dissipate fog
more quickly, fog also denser.
Models with 1km and 4km grids more similar than 12km grid
model.
Clear need for improved microphysics and radiation schemes.
Figure 3. Current results for Roissy, for
the hit rate and false alarm rate, using a
combination of the Cobal / Ispa models and
observations
Working Group 3 is evaluating the use of statistical methods and model evaluation criteria.
Fig. 2 Verification statistics for 10.12.03 case study.
The detection of fog and false alarms
The pseudo-ROC diagram shows the Hit Rate (HR as in the
ROC) vs the False Alarm Ratio (FAR = proportion of
warnings that are not justified) for a deterministic forecast
(one single point) or for a number of probability thresholds
(one curve). Good performance is indicated by the closeness to
the upper left corner. The diagram also shows the bias of
deterministic forecasts: there is no bias where HR+FAR=1, i.e.
Along the diagonal.
Figure 4 shows the performance of an operational Model Output Statistics probabilistic forecast of low
visibility thresholds in Dijon (NE France) at 06UTC. The statistical model runs from 15UTC. Lead time is
+18h. The method is a linear discriminant analysis. Predictors are elaborated from the output of the French
ARPEGE model (run at 12UTC) and from observations of different variables (available at 15UTC). Scores
have been computed from 2 consecutive winter seasons (2002-2003 and 2003-2004). The pseudo-ROC
diagram shows an encouraging performance even for lower thresholds.
The COBEL 1-D model has been
successful at reproducing fog events
in the past and it is being modified for
future use:
1-Dimensional Modeling
1-D (Column)
Boundary Layer Model
Improvements are needed.
Gravitational settling scheme
Include vegetation parameters
Soil model coupling process
Coupling the model to various high
resolution 3-D models and the usage
of an assimilation technique should
provide interesting results
High Vertical
Resolution
A Detailed Look at
the Physical Processes
Radiative Transfer
Turbulent Mixing
Soil-Atmosphere
Interactions
- momentum
- heat
- humidity
Other groups in the COST action are working on areas not included in this poster.
These include projects to
• improve the 2m visibility predictions using a DMI development of the HIRLAM model
• modify the turbulence representation and inclusion of special snow tiles for the HIRLAM / SMHI
model
• develop the Hirlam model , modified by ground observations, use of probabalistic calculations and
comparison for Madrid - Barajas region
• characterization of factors which control the generation of radiational and advectively forced fog
Further details are available from the COST web site,
http://137.248.191.94/cost or www.lcrs.de or from the poster presenters at the conference: …
...