On short-term total-column-ozone-forecast errors

The accuracy of short-term predicted total-column ozone amount, Ω, is reported to be 10–12 DU or about 3–4%. This result was achieved by several authors independently using linear multiple regressions with different predictors. The operational procedure of the German Weather Service to predict Ω uses as predictors the ozone amount of the preceding day and temperatures at some pressure levels at the forecast day. The previous day ozone is the most informative predictor. It explains 30–55% of the variance. The reduction of variance, achieved totally, varied from 64–83% with maximum values in spring (March–May) and minimum values in November. Forecast errors of the predictors have in general no substantial impact on the Ω forecast. On average, an error of 1 K in T 55 = T 50(i) − T 500(i) causes an error of 3 DU in ozone prediction. An error of 1 DU in Ω (i − 1) results in 0.3–0.6 DU forecast errors. The replacement of Ω (i − 1) by the climatological ozone amount is thus generally not critical. Errors in T 55(i) and (T 200 (i)-T 200(i − 1)) have about 8 and 4 times more weight, respectively, than errors in Ω of the preceding day, when the parameters are related to K and DU. The limit of 20 DU maximum deviation, found for Ω forecast and diagnosis, is exceeded only for large errors of the predictors, e.g. more than 5 K for T 55 and 7 K for T 200.