Seasonal forecasting of the Kenya coast short rains, 1901-84
✍ Scribed by Farmer, Graham
- Publisher
- Wiley (John Wiley & Sons)
- Year
- 1988
- Weight
- 565 KB
- Volume
- 8
- Category
- Article
- ISSN
- 2314-6214
No coin nor oath required. For personal study only.
✦ Synopsis
A rainfall anomaly time series for the Short Rains (September to December) on the Kenyan Coast (KCS) is derived for the period 1901 -84. On the year-to-year time scale a high degree of association can be seen between high/low extremes of a Southern Oscillation Index and negative/positive anomalies in the KCS series.
The SOI-rainfall relationship is investigated seasonally with KCS lagging the SO1 series by one, two and three seasons. A decay of the correlation with increasing lag can be seen. When the time series is divided into 1901-42 and 1943 -84 the relationships are stronger in the more recent period. Some forecasting skill does seem possible using the June to August (JJA) SO1 to predict the September-December rains. While the 1901 42 period shows a skill not much improved from using a simple climatological forecast, the 1943-84 period shows a much stronger relationship. Possible reasons are discussed for the disparity between the two time periods, also evident in results from others parts of Africa and Asia. There may wcll have been a change in some underlying climate mechanism between the early decades of this century and the more recent decades.
KEY woRi)s ENSO Kenya Rainfall Regression Seasonal forecasting Southern Oscillation Index
G . FARMER
season to the next. The aim here is to see if that persistence can be utilised to form a seasonal rainfall forecast. We are considering, therefore, forecasting within the year previous to the short rains occurring (September to December), rather than from one short rains to the next.
DATA
Rainfall data for this analysis have been assembled from three sources. The first two sources are combined by the National Center for Atmospheric Research (NCAR) at Boulder, Colorado, USA. NCAR archive data that eventually appear in the publication Monthly Climate Data For The World. They have also incorporated the very large amounts of African rainfall data collected by S. E. Nicholson (e.g. Nicholson, 1980).
Thc NCAR data used here end in 1983, while those of Nicholson generally end around 1973. No Kenyan rainfall data are on the NCAR tape for the year 1928. There arc some (undocumented) site changes in the NCAR data series for the four stations used here. However, the distance and altitude changes involved are not great. The third source of data used here, and obviously the primary source for all the data, was the Kenya Meteorological Department (KMD), who kindly made data available for updating the NCAR/ Nicholson database.
The Southern Oscillation Index (SOI) has been taken from Ropelewski and Jones (1987). Gaps in the Tahiti pressure series were filled by Jones (personal communication) using regression equations between Tahiti and Apia, Western Samoa, and Tahiti and Santiago, Chile. The SO1 is derived using monthly sea lcvel pressure at Tahiti and Darwin. Each station is first normalised with respect to its 1951 80 mean and standard deviation. The value for Darwin is then subtracted from that of Tahiti, and the result is normalised again, using the mean and standard deviation of the normalised pressure difference, again over the 1951-80 period. This paper uses averagc seasonal SO1 values; December to February (DJF), March to May (MAM), June to August (JJA) and September to November (SON). See Chen (1982) for a discussion of different indices of thc Southern Oscillation.