A Simple Graphical Technique for Conditional Long Range Forecasting of Below-Average Rainfall Periods in the Tuvalu Islands, Western Pacific

For the Tuvalu Island group in the western Pacific, a simple graphical method is proposed as a means of forecasting whether rainfall totals for the next 1, 2,…,6 months will be below average. The method is based on scatter plots where the points are color-coded as above- or below-average rainfall, with the plot axes being lag-1 and lag-2 NINO4 sea surface temperatures. Within the scatter plots there are reasonably clear data fields with higher frequencies of below-average rainfalls associated with cooler precursor NINO4 temperatures. These data fields are defined by subjectively emplaced separation lines which bifurcate the scatter plots into “predictable” and “unpredictable” fields. If two precursor NINO4 temperature values define a point located in a predictable field then a warning would be issued for below-average rainfall over the next n-month period, depending on the time scale of the scatter plot. A long rainfall record at Funafuti in Tuvalu indicates that success in predictable-field forecasting of below-average rainfalls would range between 68% for 1-month rainfall totals and 89% for 6-month totals. The forecasting success derives from persistence of cooler NINO4 sea surface temperatures which are associated with lower rainfalls in Tuvalu. However, many dry periods are also located in the unpredictable field and cannot be forecast by this method.