| description abstract | The rainfall anomaly index (RAI) has been widely used to study variations over time in Sahelian rainfall. Its interpretation is often complicated by excessive missing data and changes in station network, both of which prevent a precise quantification of the significance of any given RAI estimate. Also, unless the time series are filtered, high interannual variability often obscures important rainfall fluctuations. Here we apply a simple method for calculating approximate confidence limits of areal rainfall estimates to annual data from two constant network configurations in Sudan and West Africa. These cover, respectively, the periods 1920?88 (13 stations) and 1922?85 (12 stations) and contain only 3 and 5 missing annual totals out of 897 and 768. The resulting annual RAI estimates, and 95% confidence limits, were subjected to a 9-point binomial filter, and a 30-point retrospective uniform filter (i.e., an annually updated WMO reference period), also called a running mean. By combining the RAI and confidence-level estimates with different filters we develop a technique that should be useful for interpreting RAIs and assessing, the impact of climate on natural resources. This technique can be used to construct quantitative indicators of the terms climate anomalies, climate fluctuations, and climatic change. We illustrate this by applying tentative indicators to the two Sahel series, and also, by way of contrast, to an annual RAI for southern Sweden (15 stations covering the period 1861?1988 with 3 missing annual totals out of 1920). For example, recent individual anomalous years occurred in Sudan in 1978 and 1988 (wet) and 1984 (dry), and for both West Africa and Sudan a climatic change compared to century-mean rainfall had nearly occurred by the late 1980s. Southern Sweden has witnessed two recent climate fluctuations in the early 1970s (dry) and in the mid-1980s (wet). In conclusion, we hint at some refinements to the technique, but stress that for climate monitoring purposes the need for station networks of high quality and consistency over time will remain undiminished. | |
