In about half of all forecasting situations, however, the oscillating frequency of the trend itself does not even remain constant in the short time of three days, which often leads to a cusp point in the trend curve in the time-range being forecast (example d). Univariate techniques naturally fail here, resulting in the Type 1 and Type 2 errors discussed in Chapter 10. At this point forecasting accuracy can only be improved by taking into account the causal interplay of the atmospheric state parameters. Such situations are in principle univariately unpredictable; multivariate study is essential.