Nova Scotia pilot

1
Nova Scotia pilot evaluation of CHRONOS AQHI
Doug E Steeves Atlantic Storm Prediction
Centre February 13, 2007
2
Contents

• AQHI Forecasts to Nova Scotia Pilot
• Technical Issues
• Lessons Learned
• Chronos AQHI Verification
• Conclusions

3
AQHI Forecasts

• Forecast support July 8 Sept 30
• Sent once per day at approx 0800 ADT
• Maximum AQHI for today and tomorrow for three
  sites Halifax, Dartmouth (Lake Major), and
  Sydney
• Manually extracted from FQ (Chronos) files on CMC
  website.

4
Sample Forecast

• FHCN99 CWHX ddhhhh
• 2006-05-10 070000,Sydney,1,2
• 2006-05-10 070000,Halifax,2,3
• 2006-05-10 070000,Dartmouth,1,2

5
(No Transcript)
6
Chronos Guidance

• 20061031
• AQHI from/de CHRONOS (CTM)
• Pour/for Sydney
• P0000Z
• A / At 0 m
• Hres 1-6 7-12 13-18
  19-24 25-30 31-36 37-42 43-48
• /Hrs
• Maximum 1.81 1.64 1.82
  1.88 1.88 2.49 3.22 3.28
• Moy/Mean 1.79 1.62 1.74
  1.86 1.80 2.16 2.79 2.10
• AQHI from/de CHRONOS (CTM)
• Pour/for Halifax
• P0000Z
• A / At 0 m
• Hres 1-6 7-12 13-18
  19-24 25-30 31-36 37-42 43-48
• /Hrs
• Maximum 2.03 2.02 2.47
  3.15 3.33 2.86 3.32 3.46
• Moy/Mean 2.00 1.99 2.27
  2.74 2.92 2.38 3.21 2.26

7
2 t yr-1

• Lake Major

Tufts Cove 400 t yr-1
Shearwater
Halifax
7 t yr-1
Refinery 250 t yr-1
8
(No Transcript)
9
Technical Issues

• July 8-11 -- Forecasts were not properly ingested
  by NS EL web site.
• July 14-15 -- Wrong date stamp on the data.
  Likely the form had not been properly updated by
  the forecaster. The forecast values were likely
  correct.
• Aug 11 Identified error in manual input of
  AQHI. Dartmouth forecast high risk, Halifax
  forecast low risk.
• Aug 23-24 -- AQHI guidance from Chronos
  unavailable. Forecast sent for day 1on Aug 23
  (based on previous days Chronos run) but not for
  day 2.
• Sept 1 -- No forecast. Reason unknown.
• Sept 6 -- Forecast late due to network problems.
  Sent around 1200 pm.
• Sept 9-11 -- No forecast. Likely forecaster
  error.
• Sept 16-17 -- AQHI guidance from Chronos
  unavailable. Forecast sent for day 1 on Sept 16
  (based on previous days Chronos run). No forecast
  sent on Sept 17.
•  

10
Lessons Learned

• Forecasters should work with pollutant
  concentrations before calculation of AQHI.
• For the purposes of this pilot, automation of
  product from Chronos would have been better.
• Manual extraction introduced errors.
• Feedback suggests some confusion over the
  forecasts because actual AQHI was presented as a
  number and a risk factor while no number was
  associated with the forecast.

11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
Conclusions

• Error scores (RMSE and bias) are not large but
  persistence clearly outperformed CHRONOS
• There was no clear winner wrt forecasts by
  category
• Diurnal variation to CHRONOS RMSE and bias
• CHRONOS had positive bias at L Major and negative
  bias at Halifax and Sydney.
• Possibly explained by gradient of ozone at coast
  or representation of NOx.
• Performance measurement of AQHI provides minimal
  feedback to the forecast process
• Evaluation should be done on the pollutant level

22
Acknowledgements

• NS Department of Environment and Labour
• Barb Bryden, Kamila Tomcik, Fran Di Cesare, Jenn
  Hutchins, Johnny McPherson
• CMC
• Veronique Bouchet, Jacinthe Racine, Stephane
  Gaudreault
• ASPC
• Bruce MacDonald, Geoff Livingstone
• Air Quality Prediction
• Mike Howe

23
(No Transcript)