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

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

C.2.2 HYYSPLIT<br />

GLOB BAL STATISTIICS<br />

FOR CAPPTEX<br />

RELEASSE<br />

3<br />

Figures CC‐2<br />

<strong>and</strong> C‐3 display d <strong>the</strong> gglobal<br />

statisttics<br />

for <strong>the</strong> HHYSPLIT<br />

senssitivity<br />

tests w<br />

<strong>and</strong> C‐3 ccontaining<br />

th he statistical<br />

metrics whhere<br />

<strong>the</strong> bestt<br />

performingg<br />

model has<br />

respectivvely,<br />

lowest <strong>and</strong> highest score. For t<strong>the</strong><br />

FOEX meetrics,<br />

INITD140<br />

scores t<br />

nearly 0% %, followed closely c by INNITD1<br />

<strong>and</strong> INNITD2.<br />

INITDD3<br />

scores <strong>the</strong>e<br />

poorest w<br />

score followed<br />

by INITD4.<br />

The twwo<br />

puff conffigurations<br />

hhad<br />

<strong>the</strong> poorrest<br />

NMSE a<br />

performaance<br />

metrics s (with valuees<br />

<strong>of</strong> approximately<br />

127 <strong>and</strong> 130 pg m<br />

1.57 for FFB).<br />

The fou ur puff‐particcle<br />

model coonfiguration<br />

options (IN<br />

<strong>the</strong> best overall score es for both NNMSE<br />

<strong>and</strong> FB.<br />

INITD1 annd<br />

INITD2 ex<br />

score witth<br />

30% <strong>and</strong> 31% 3 respecttively,<br />

with <strong>the</strong><br />

poorest pperforming<br />

b<br />

For <strong>the</strong> wwithin<br />

a facto or <strong>of</strong> 2 <strong>and</strong> 5 metric (FA22<br />

<strong>and</strong> FA5, FFigure<br />

C‐3, to<br />

configuraations<br />

INITD3<br />

<strong>and</strong> INITD4<br />

<strong>and</strong> <strong>the</strong>ir ccounterpart<br />

particle‐puf<br />

<strong>and</strong> INITDD104<br />

perform<br />

slightly beetter<br />

than thhe<br />

o<strong>the</strong>r connfigurations.<br />

Figure C. 2.2‐2, botto om left), all o<strong>of</strong><br />

<strong>the</strong> HYSPLIT<br />

configurations<br />

show a<br />

ranging ffrom<br />

‐0.04 to o ‐0.09.<br />

‐3 with Figuress<br />

C‐2<br />

<strong>the</strong>,<br />

<strong>the</strong> best withh<br />

ith a 21% FOOEX<br />

<strong>and</strong> FB statistical<br />

for error<br />

<strong>and</strong> 1.56 <strong>and</strong><br />

ITD3,4,103,1104)<br />

exhibiteed<br />

xhibited <strong>the</strong> best overall<br />

KSP<br />

being INITD33<br />

with 49%.<br />

op), <strong>the</strong> hybrid<br />

puff‐partticle<br />

ff configurattions<br />

INITD103<br />

For <strong>the</strong> PCCC<br />

metric (PCCC,<br />

a slight negaative<br />

correlation<br />

Figure C‐ ‐2. Global model m perforrmance<br />

statistics<br />

for ninne<br />

HYSPLIT INITD<br />

sensitivity<br />

tests foor<br />

CAPTEX RRelease<br />

3.<br />

3

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

Saved successfully!

Ooh no, something went wrong!