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.

⎛ b ⎞<br />

TS = ⎜ ⎟×<br />

100%<br />

(2‐5)<br />

⎝ a + b + d ⎠<br />

2.4.3.2 Temporal Analysis<br />

In Section 2.4.1 temporal statistics related to <strong>the</strong> timing <strong>of</strong> when <strong>the</strong> predicted <strong>and</strong> observed<br />

tracer arrives at a monitor or arc <strong>of</strong> monitors, its residence time over a monitor (or arc) <strong>and</strong><br />

when <strong>the</strong> tracer leaves <strong>the</strong> monitor (or arc) were discussed. Ano<strong>the</strong>r temporal analysis<br />

statistics is <strong>the</strong> Figure <strong>of</strong> Merit in Time (FMT), which is analogous to <strong>the</strong> FMS only it is calculated<br />

at a fixed location ( x ) ra<strong>the</strong>r than a fixed time as <strong>the</strong> FMS. The FMT evaluates <strong>the</strong> overlap<br />

between <strong>the</strong> measures (M) <strong>and</strong> predicted (P) concentration at location x <strong>and</strong> time tj. The FMT<br />

is normalized to <strong>the</strong> maximum predicted or measured value at each time interval <strong>and</strong> is<br />

expressed as a percentage value in <strong>the</strong> same manner as <strong>the</strong> FMS (Mosca et al., 1998).<br />

The FMT is sensitive to both differences between measured <strong>and</strong> predicted <strong>and</strong> any temporal<br />

shifts that may occur.<br />

2.4.3.3 Global Analysis<br />

Following Draxler et al. (2002), four broad categories were used for global analysis <strong>of</strong> model<br />

evaluation. These broad categories are: (1) scatter; (2) bias; (3) spatial distribution <strong>of</strong><br />

predictions relative to measurements; <strong>and</strong> (4) differences in <strong>the</strong> distribution <strong>of</strong> unpaired<br />

measured <strong>and</strong> predicted values. One or more statistical measures are used from each <strong>of</strong> <strong>the</strong><br />

four categories in <strong>the</strong> global analysis. These include <strong>the</strong> percent over‐prediction, number <strong>of</strong><br />

calculations within a factor <strong>of</strong> 2 <strong>and</strong> 5 <strong>of</strong> <strong>the</strong> measurements, normalized mean square error,<br />

correlation coefficient, bias, fractional bias, figure <strong>of</strong> merit in space, <strong>and</strong> <strong>the</strong> Kolmogorov‐<br />

Smirnov parameter representing <strong>the</strong> differences in cumulative distributions (Draxler et al.,<br />

2002).<br />

Factor <strong>of</strong> Exceedance: In <strong>the</strong> scatter category, better model performance is observed when <strong>the</strong><br />

Factor <strong>of</strong> Exceedance (FOEX) measure is close to zero <strong>and</strong> FA2 (described next) has a high<br />

percentage. A high positive FOEX <strong>and</strong> high percentage <strong>of</strong> FA5 would indicate a model’s<br />

tendency towards over‐prediction when compared to observed values.<br />

⎡ N<br />

FOEX = ⎢<br />

⎣ N<br />

( P > )<br />

i Nii<br />

⎤<br />

− 0 . 5⎥<br />

× 100%<br />

⎦<br />

Where, N in <strong>the</strong> numerator is <strong>the</strong> number <strong>of</strong> pairs when <strong>the</strong> prediction (P) exceeds <strong>the</strong><br />

measurement (M) <strong>and</strong> <strong>the</strong> N in <strong>the</strong> denominator is <strong>the</strong> total number <strong>of</strong> pairs in <strong>the</strong> evaluation.<br />

In FOEX, all 0‐0 pairs are excluded from <strong>the</strong> analysis. FOEX can range from ‐50% to +50% with a<br />

perfect model receiving a 0% value.<br />

17<br />

(2‐6)<br />

(2‐7)

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

Saved successfully!

Ooh no, something went wrong!