Forcasting NYCDOH Announcements
Of West Nile Virus Activity

Successful Forecasting

Initial public forecasts began in mid-summer of 2000. See these early attempts to forecast WNV activity before the method had been fully developed. See latter attempts also.

For reference, here is the summer of 2000, ozone, and WNV activity for NYC. Ozone data is very precise timewise. West Nile virus activity is loose for these reasons: Date found for dead birds is somewhat loose. Date collected for mosquito pools can be loose and doesn't mean that the mosquito was infected with WNV on the collection date. Human hospitalizations were usually preceeded by illness.

Note that both WNV activity and ozone levels oscillate at the same frequency. From 7/15/00 onward, every peak and valley in the ozone level line (red) is reflected in the WNV activity line (blue). 9/12/00 is a reasonable and minor exception.

Though initial forecasts have been a bit clumsy, they have been sucessful re all medium/major peaks and valleys for WNV activity during the summer of 2000. I had no station monitor data until 9/27/00, after the ozone/WNV season, and relied on AirNow maps and weather reports.

Of those outside of medical orthodoxy, mosquito control agencies, and emergency management, I am one of the few that can look forward to the next West Nile virus epidemic.

See below, a comparison of comments and predictions, extracted from the timeline:

Prediction dates are after the fact, but before the NYCDOH announcements of test results.

Date Timeline2000 "Diary": Predictions, and AirNow map, weather, monitor observations After-thoughts
7/18/00 At 7am, I noticed extreme smog throughout NYC, obviously overnight NJ refinery emissions, sulfur compounds. I searched in vain for a mention of this in,,, and others. So, for the record, I emailed NYgreens and received confirmation. 4 days later, 8/22/00, see the 1st occurrance human WNV disease for the summer.
7/21/00 This ozone pattern (1 hour average) and the ozone values are nearly identical to 8/2/00. These values and the 8/2/00 values are the highest ozone values of the summer, both preceding major human hospitalization (the first hospitalization of the summer and the second group of hospitalizations. Similar: 7/21/00 and 8/2/00
7/22/00 This is the first WNV hospitalization for year 2000. 1st human WNV.
8/2/00 These ozone patterns and WNV consequences are similar to 7/21/00. Similar: 7/21/00 and 8/2/00
8/4/00 Staten Island Borough Pres. Guy Molinari stated 8/4/00, "Our Helpline unit has fielded... reports of more than 500 dead birds..." [Other boroughs had reported few or no dead birds.] The epicenter for 2000
8/7/00 Prediction #1 (by email to 5 persons on 8/16/00)==> There will be reports of higher WNV activity for 8/7-8/11. This proved right, the peak WNV activity for the year occurred. I also predicted human death, which did not occur.
8/12/00 Prediction #2 (emailed to 5 persons on 8/16/00)==> "There will be a lull 8/12-8/20 in terms of onsets of WNV disease" -- based on my 8/16/00 review of AirNow and weather predictions of cool weather, which usually means less ozone. This prediction worked well, however, I could have better predicted the lull if I had weather or monitor staion data. I had not yet began reviewing NYCDOH press releases. Fine prediction, but hampered by no station monitor data, at the time.
8/13/00 Prediction #3, emailed on Aug 20: I contradicted NYT (8/8/00) quote of W.B. Stone (DEC) who stated that WNV activity would "amplify" from "mid-August until Labor Day." I stated that 8/12/00 was the peak of the epidemic and that WNV activity would diminish. I didn't foresee AirNow not portraying haze during the following days. If I had been watching the weather I might have taken into account the high temperature and haze of 8/15-8/16 and modified the prediction. I need the latest data from Stone, but he won't reply to date (1/7/01). Forecast re the peak proved true, but the forecast re diminishing WNV proved not true.
8/14/00 Predicting a lull for this area was risky because of the potential for belated findings and collections, brought about by the 8/11/00 smog. No station monitor data at the time of prediction.
8/16/00 Prediction #2 re lull in WNV activity cut short by abberative WNV activity (in terms of AirNow map correlations). These bird findings could be from 8/11-8/15 smog. However, weather report reveals haze 8/15, also 8/16/00 during 2am-9am (refinery emissions?) AirNow covers only 10am-12pm). AirNow can be grossly revised. Monitor station data indicates ozone.  
8/20/00 This WNV activity confounds the toxin theory (need NYCDOH spreadsheet and EPA monitor station data, for accuracy). It could be explained by 8/16-8/18 haze and Monday calls being previously deferred calls (gov offices closed over weekend). Though 8/20 shows no sign of haze on the hourly weather report, the daily indicates haze. [Later: EPA monitor station data gave sense to all of this.]  
9/9/00 Prediction to myself (on 9/15): NYCDOH should be announcing WNV activity for these high ozone days through 14th.  
9/20/00 Prediction (10/1/00): This ozone should be indicating WNV activity.  
9/22/00 I forecasted to the Connecticutt Post (Marian Brown) at 11 a.m. by email, and to Ward Stone (NYSDEC, with summary report) at 1:24pm by confirmed delivery priority mail -- that high WNV activity would be announced by NYCDOH for the period 9/9-9/15/00. A few hours later, this forecast was confirmed by NYCDOH with PR101922 and the next day with PR102923. (No reply from Connecticutt Post to date. Ward Stone replied and denied that the correlation indicated any causation). With no weather or monitor station data for September I relied only on AirNow maps.  
9/30/00 Last AirNow map. As of 10/1/00, AirNow maps are suspended until May, 2001.  


The Fundamentals

West Nile Virus occurs in less than 8% of massive bird death during the summers of 1999-2000 in the NYC region. Therefore it could not be causative for the novel, massive bird death. Nevetheless, it occurs concurrently and therefore is a measure of disease in animals and humans.

It is my thesis that the phenomena known as "WNV acivity" correlates with unprecedented levels of air toxics, which became a novelty as of 1995 when MTBE levels were increased, and unique refinery dumping/formulating began in the NYC region. WNV activity is also concurrent with the decision to begin testing for WNV in this region as of the summer of 1999 and not to test in other regions.

Air toxics poisoning became publically obvious as of the June of 1999 due to record-level air toxics (not seen since the 1980s) and the overwhelmingly obvious signs of disease found in the massive bird deaths and unusual human death.

Since it takes 5 to 30 days for the DOH to announce bird/mosquito collections and human hospitalizations, this theory of toxic causation can forecast the official announcements by several days, being accurate within hours of air toxic indicators. It appears that the DOH announcements usually take 5 to 12 days.

Forecast Method

Goal: To forecast increased WNV activity (bird death, WNV-positives, hospitalization, illness, WNV-positive mosquito pool collections).

Note that ozone and other air toxics are marginal in their effect, and take a days to accumulate tissue damage, to bring a bird or human to a critical threshold where disease is imminent. Disease events occur during sustained exposure and immediately after high spikes.

It is assumed that industrial pollutants, auto traffic, MTBE, and/or air traffic are present in the epidemic area.

1) Wait. The haze season (ozone) begins in May. The forecaster should wait for the first regular reports of dead bird reports to occur, usually in late June or July. This is the signal that accumulated tissue damage has reached a threshold and epicenters are defined.

2) Observe the indicators at the epicenters.

Indicators of Increased WNV activity
Reliability Indicator Comment
First Personal observations of haze (site and smell). Necessary, because AirNow maps can be grossly revised for unpublicized reasons and they are wide-focus, interpolated views of large areas. AirNow maps do not include overnight readings. Weather stations may not suffice either, as they may not be near enough to the epicenter, or fog mistaken for haze, or fog concealing or incorporating haze.
Second EPA air monitor station data of epidemic area. Very clear and obvious. These are often already located at the disease epicenters. Their very existence is indictive of air emission dangers. Unfortunately, the printed reports may contain large drop-outs of data due to maintenance or faulty equipment.
Third Weather Observations of haze. These often show haze when AirNow maps fail. A haze index formula with timeline are provided.
Fourth AirNow map indicating ozone in an epidemic area. Very dramatic colored views of interpolated ozone levels in large areas. Unfortunately, lacking in resolution and subject to gross downward revisions.
Fifth High temperature in epidemic area. Air Toxics are more present at ground level during high temperatures. Convection effect is less effective in pushing stack emissions into upper atmosphere. Photochemical effect, necessary for the generation of smog is more effective during hot and sunny days.

If all is working, then these indicators should act concurrently.

Forecasting a Lull

If all these indicators are low then a lull in WNV activity is forecasted, though a faulty indicator, or indicator undergoing maintence, or other lack of data, can defeat a forecasted lull. Sometimes air emissions are not shown, or don't show up because they are intense but local, not affecting station monitors. Sometimes, the data is revised by the EPA/DEC for unpublicized reasons.