An investigation of daily maximum ground-level ozone concentration forecasting using CART statistical techniques / Doug McCollor. : En57-45/93-11E-PDF

"CART (Classification and Regression Trees), a relatively new statistical analysis and prediction technique in the field of meteorology, was used to study ground-level ozone concentration data obtained during a forecasting experiment in the summer of 1992. Data from six ozone monitoring sites throughout Vancouver and the Lower Mainland were used in the study. It was found that the CART techniques were comparable, but not more favorable, with standard meteorological statistical techniques employed during the original experiment. This study also used time-series representation of the CART results and data. The analysis shows that the CART technique as used in this case relies strongly on the previous day's maximum ozone concentration level as a predictor, resulting at times in a one-day time lag response. The forecast ozone concentration values are also a result of smoothing by the CART statistical technique, diminishing extreme-valued events in the observed data set. As a result, two "poor" air quality events which occurred in the experiment were not predicted by any of the statistical methods studied to date"--Abstract, page [1].

Lien permanent pour cette publication :
publications.gc.ca/pub?id=9.882693&sl=1

• Format MARC XML
• Format MARC HTML

Demander des formats alternatifs