The Canadian operational procedure for forecasting total ozone and UV radiation

In the spring of 1992 Environment Canada started daily forecasts of ultraviolet-B radiation reaching the ground for locations in Canada. The forecast quantity is a ‘UV Index’ based on an empirical formula relating UV flux at the ground to solar zenith angle, total depth of ozone in a vertical column, and the time of year. The primary subject of this paper is the procedure for forecasting total ozone that has been used operationally since June 1993. The UV forecast is produced in three steps. First, total ozone it estimated over all of the northern hemisphere using a regression relationship between upper atmospheric variables and total ozone. Second, a correction is made to the forecast based on current measurements of ozone over Canada. Third, the total ozone estimates are fed to the formula to obtain the UV Index. The procedure runs in a ‘perfect prog’ real-time forecast mode, with values of the atmospheric variables taken from predictions of the Canadian global spectral model. The predictand data consisted of total ozone observations for the northern hemisphere for the period 1987-1991, measured by the Total Ozone Mapping Spectrometer (TOMS) instrument aboard the Nimbus-7 spacecraft. The predictors were obtained from an assortment of upper-tropospheric and lower-stratospheric meteorological variables, analyses of which were processed into a set, of predictors known from previous work to be related to total ozone. To these variables were added climatological total ozone and various functions of latitude to represent the observed large-scale latitudinal and seasonal variations in total ozone. Six equations were developed, one for each of three latitude bands in each of two seasons. Selection of the latitude bands toot designed so that separate regressions would be applied in tropical and extra-tropical latitudes in winter and summer, since there are significant differences in total ozone patterns. It was found that up to about 82% of the variance in total ozone could be explained with six predictors. Further tests with independent data suggested root mean square error (RMSE) levels of about 16 Dobson units in regression forecasts for Canadian locations for the summer of 1992 and about 23 Dobson units for the summer of 1993, which was an abnormal period of persistent record-low stratospheric ozone. Although the forecasts show considerable skill, they are improved by the application of a correction procedure before the UV Index formula is applied. The correction procedure uses observations from a network of 13 Brewer spectrophotometers in Canada and a standard successive correction objective analysis procedure to obtain a field that is used to correct the next day's forecast. The correction procedure successfully removes the bias and reduces the RMSE of the ozone forecasts to about II Dobson units in both summers of 1992 and 1993.