Intra‐seasonal rainfall characteristics and their importance to the seasonal prediction problem

Abstract Daily station rainfall data in South Africa from 1936 to 1999 are combined into homogeneous rainfall regions using Ward's clustering method. Various rainfall characteristics are calculated for the summer season, defined as December to February. These include seasonal rainfall total, region‐average number of station rain days exceeding 1 and 20 mm, region‐average of periods between rain days at stations >1 and >20 mm, region‐average of wet spell length (sequential days of station rainfall >1 and >20 mm), correlation of daily station rainfall within a region and correlation of seasonal station rainfall anomalies within a region. Rank‐ordered rainfall characteristic data generally form an s‐shaped curve, and significance testing of discontinuities in these curves suggests that normal rainfall conditions in South Africa consist of a combined middle three quintiles separated from the outer quintiles, rather than the traditional middle tercile. The relationships between the various rainfall characteristics show that seasons with a high total rainfall generally have a higher number of heavy rain days (>20 mm) and not necessarily an increase in light rain days. The length of the period between rain days has a low correlation to season totals, demonstrating that seasons with a high total rainfall may still contain prolonged dry periods. These additional rainfall characteristics are important to end‐users, and the analysis undertaken here offers a valuable starting point for seeking physical relationships between rainfall characteristics and the general circulation. Preliminary studies show that the vertical mean wind is related to rainfall characteristics in South Africa. Given that general circulation models capture this part of the circulation adequately, seasonal forecasts of rainfall characteristics become plausible. Copyright © 2002 Royal Meteorological Society.