Talk:Evidence for Flagging Exceptional Events
NFrank: Initial Feedback -- Rhusar 10:22, 24 April 2008 (EDT)
I have started to study the "report". There is a lot of great stuff here. (BTW, what happened to your google earth console to show the trajectories and other layers? I thought that format was really neat and added to our understandings.)
There are several aspects of your current report that can benefit from larger community input. In particular, there are a few points which are not currently clear to me. See item (1) below. I hope that you or others can clarify. I also think the conclusions need to be soften particularly for the more borderline exceedances and departures from "normal" and the outline of "affected" areas to be better delineated. Once those points are clearer, we should expose to a larger group.
(1) See the two phrases in blue below: Because of your "question mark," are your speculating or concluding that the trajectories reached Illinois and Minnesota? ... I also don't understand your connection between Ohio to Midlantic transport as basis to discount GA impact on OH. Please clarify :
"On May 24, the high concentration sites stretched from Alabama
northward into the Ohio River Valley. The backtrajectories for all
sites show a clockwise, anti-cyclonic circulation that brings marine
air from the Gulf and Atlantic into the eastern U.S. The western
bundle of trajectories, which result in high concentrations at
receptor sites from Georgia through Illinois/Minnesota? are all
passing through the Georgia fire region. These trajectories provide
evidence that the smoke has impacted eastern U.S. sites that stretch
all the way up to Minnesota/Illinois. On the other hand, the
backtrajectories for the air masses reaching the Ohio River Valley
are passing to the mid-atlantic states, well north of the Georgia
fire. Therefore, those sites in the Ohio River Valley are not likely
to be influenced by the smoke. "
(2) As you mentioned in earlier email, alternative moving medians or other baselines would be helpful.
In this example of this several week event, the 30-day median may
itself reflect the influence of the fire. Alternative indicators of
"normal" are therefore needed that look to (a) seasonal periods not
affected by the event, say from the prior 3 years and (b) different
baselines to more confidently distinguish exceptional from normal
(e.g. 1 or 2 std deviations above seasonal normals, or even 95th/98th
percentiles.)
Rudy, can the selection of normal be easily modified to permit user to consider such alternative indicators?
(3) More specificity is required to identify the boundaries of (and specific monitoring sites in ) the sub-areas for which there would not be an exceedance but for the event. The ones with high and low certainty should be id'd. This is where the alternative baselines could be helpful. It would also be helpful to extend these findings to other days claimed to be affected by the GA fires, and in particular, to isolate those exceedance days and locations with lesser amounts of evidence. The latter can include days without speciation data like May 25 with NE exceedances which is easy to examine using your
Analyst's Console. For these days, to what extend should data and
conclusions reached from adjacent days be used, and therefore with more
reliance placed on trajectories and perhaps with a higher baseline to
represent "normal" . Other recommendations for ways to help reach
conclusions are also desired. To help with the latter, we can possibly
call upon STI with whom we have a work assignment to evaluate
exceptional event days. For example, we can limit that task to those
days and locations which have already been proposed by the States as
exceptional. Let's see what we can accomplish internally first.