20.04.2013 Views

Documentation of the Evaluation of CALPUFF and Other Long ...

Documentation of the Evaluation of CALPUFF and Other Long ...

Documentation of the Evaluation of CALPUFF and Other Long ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

The ETEX real‐time LRT modeling phase was performed in parallel with <strong>the</strong> tracer field<br />

experiment. When <strong>the</strong> release started, 28 modeling groups were notified <strong>of</strong> <strong>the</strong> starting time,<br />

source location, <strong>and</strong> emission rate. They ran <strong>the</strong>ir LRT models in real‐time to predict <strong>the</strong><br />

evolution <strong>of</strong> <strong>the</strong> tracer cloud, <strong>and</strong> <strong>the</strong>ir predictions were sent as soon as <strong>the</strong>y were available to<br />

<strong>the</strong> statistical evaluation team at JRC‐Ispra. The capability <strong>of</strong> providing <strong>the</strong>se predictions in<br />

real‐time was considered to be an important factor, as well as <strong>the</strong> model performance itself.<br />

Therefore, only those institutions that had access to a meteorological model or that received<br />

real‐time forecasts from a meteorological centre could participate.<br />

The analysis <strong>of</strong> <strong>the</strong>se calculations could not distinguish <strong>the</strong> differences between predictions <strong>and</strong><br />

measurements arising from dispersion model inadequacies as opposed to those arising from<br />

<strong>the</strong> meteorological forecasts used. Almost two years after <strong>the</strong> ETEX releases, <strong>the</strong> ATMES‐II<br />

modeling exercise was launched to evaluate <strong>the</strong> LRT models in hindcast mode. ATMES‐II<br />

participants were required to calculate <strong>the</strong> concentration fields <strong>of</strong> <strong>the</strong> first ETEX tracer<br />

experiment using ECMWF analyzed meteorological data as input to <strong>the</strong>ir own dispersion<br />

models. Any institution operating a long‐range dispersion model could now participate<br />

whe<strong>the</strong>r or not it had real‐time access to <strong>the</strong> meteorological data, <strong>and</strong> <strong>the</strong> number <strong>of</strong><br />

participants (49) was increased compared to <strong>the</strong> ETEX real‐time modeling exercise, even though<br />

not all <strong>of</strong> <strong>the</strong> original ETEX modelers took part in ATMES‐II.<br />

Contrary to ETEX, <strong>the</strong> differences between <strong>the</strong> measured <strong>and</strong> modeled concentration fields in<br />

ATMES‐II could be more directly related to <strong>the</strong> dispersion simulation, thanks to <strong>the</strong> use <strong>of</strong> <strong>the</strong><br />

same meteorological fields. However, even in this case, discrepancies between models were<br />

due not only to <strong>the</strong> calculation <strong>of</strong> dispersion, but also to <strong>the</strong> different ways in which <strong>the</strong><br />

meteorological information was used. Moreover, ATMES‐II modelers could also submit results<br />

obtained with a meteorological analysis different from that <strong>of</strong> ECMWF.<br />

As for <strong>the</strong> statistical analysis in ETEX real‐time modeling exercise, <strong>the</strong> analysis <strong>of</strong> ATMES‐II<br />

model results was divided into time, space <strong>and</strong> global analyses. The same statistical indices <strong>of</strong><br />

<strong>the</strong> first ETEX release were computed in <strong>the</strong> time analysis, while for <strong>the</strong> o<strong>the</strong>r two analyses<br />

some different indices were computed following <strong>the</strong> requirements <strong>of</strong> modelers, <strong>and</strong> <strong>the</strong><br />

experience gained during <strong>the</strong> two real‐time exercises.<br />

In a general, a substantial improvement in <strong>the</strong> models' performance in <strong>the</strong> ATMES‐II modeling<br />

was seen compared to <strong>the</strong> ETEX real‐time modeling phase for <strong>the</strong> common statistical indices.<br />

When comparing <strong>the</strong> results <strong>of</strong> <strong>the</strong> ATMES‐II statistical analysis with those for <strong>the</strong> real‐time<br />

simulation <strong>of</strong> <strong>the</strong> first ETEX release, a general improvement <strong>of</strong> <strong>the</strong> model performances for<br />

those who took part in both exercises is evident. This can be explained by <strong>the</strong> better resolution<br />

<strong>of</strong> <strong>the</strong> meteorological fields used, <strong>the</strong> availability <strong>of</strong> <strong>the</strong> measured values <strong>of</strong> tracer<br />

concentration that allowed participants to tune some parameters in <strong>the</strong>ir long‐range dispersion<br />

model <strong>and</strong> <strong>the</strong> time elapsed between <strong>the</strong> two exercises (2 years) during which improvements in<br />

model formulation <strong>and</strong> application procedures took place.<br />

Spatial Analysis: In ATMES‐II <strong>the</strong> spatial analysis consisted <strong>of</strong> <strong>the</strong> calculation <strong>of</strong> <strong>the</strong> Figure <strong>of</strong><br />

Merit in Space (FMS) at 12, 24, 36, 48, 60 hours after <strong>the</strong> release start. The FMS is <strong>the</strong> ratio <strong>of</strong><br />

<strong>the</strong> spatial distribution <strong>of</strong> <strong>the</strong> overlap <strong>of</strong> <strong>the</strong> predicted <strong>and</strong> observed tracer pattern to <strong>the</strong> union<br />

<strong>of</strong> <strong>the</strong> predicted <strong>and</strong> observed tracer pattern <strong>and</strong> is expressed as a percent (note that all<br />

statistical metrics are defined in detail in Section 2.4). A big improvement could be observed in<br />

<strong>the</strong> models' FMS compared to <strong>the</strong> ETEX real‐time exercise for <strong>the</strong> first release. For instance, at<br />

36 hours in ATMES‐II all <strong>the</strong> models had a non‐zero FMS, half <strong>of</strong> <strong>the</strong> models had FMS>45% <strong>and</strong><br />

10

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!