Temporal accumulations of extreme daily mean temperature anomalies

Statistical characteristics of accumulated extreme low and high daily mean temperatures have been investigated for Hungary. Temperature extremes over periods of a few weeks were studied using 96-year temperature records. Temperature is defined as extreme in the lowest or highest five percent of the empirical distribution of anomalies. Recurrence frequencies of extreme anomalies were computed at 1, 2,..., 75 days after the initial anomalies, and compared with the respective characteristics of an appropriate 2(nd) order autoregressive model. In accordance with the high 1-day lagged autocorrelation in Hungarian temperature time series, model-frequencies are large for low lags, and approach the value for independence (0.05) for periods longer than 10-15 days. Considerable similarity of recurrence frequency between observed and simulated series is found but with some important exceptions. After only 1 extreme cold winter-day or an extreme warm slimmer-day, the probability of the next extreme occurring with the same sign remains significantly high for longer than 1 month. Empirical frequencies an very sensitive to the calendar date of the initial extreme value occurrence. Discrepancies between real and model characteristics were tested for statistical significance by Monte Carlo simulation technique. Recurrence frequencies for the last 30 years have also been computed, and interesting similarities and differences were found relative to the earlier part of the 20(th) century. To identify possible relationships between droughts and hot spells extreme high daily temperatures and dry sub-periods during summers were selected. Recurrence frequencies of extreme positive anomalies in dry periods, more than 15-20 days after an initial occurrence, are surprisingly high during the second half of the summer.