Forecasting suitable breeding conditions for the red-billed quelea Quelea quelea in southern Africa.
Red-billed queleas Quelea quelea are major pests of small-grain crops throughout sub-Saharan Africa. When conditions permit, birds breed in vast colonies which are targets for control operations. Queleas are intra-African long-distance migrants whose complex movements vary annually according to variations in rainfall patterns. An ability to forecast where and when colonies could be established will greatly improve the efficiency of control measures. We describe such a forecasting model for the southern African subspecies Q. q. lathamii. The model is based on and provides a partial test of the conjectured rainfall-migration model of Ward (1971), whereby quelea movements are determined by rainfall patterns and grass seed availability. We consider that a threshold quantity of rainfall (shown here to be c. 60 mm) within any 2 weeks at the start of the wet season is necessary to cause annual grass seeds to germinate, forcing queleas to emigrate. Only if rainfall exceeds another threshold (240 mm within a 6-week span and at least 6 weeks elapsed after the 60-mm threshold had been exceeded), allowing production of fresh seed, can queleas return and commence breeding. Birds may breed again in the same season by undertaking further long-distance 'breeding migrations' to areas where sufficient rain had fallen even later. Using only estimates of rainfall derived from satellite data, we describe a simple spatio-temporal model for southern Africa to show at weekly intervals those areas where (a) the wet season has not yet started; (b) the early rains migration threshold has been exceeded; (c) the threshold to permit breeding has been exceeded; and (d) conditions are no longer suitable for new colonies to be established. Synthesis and applications. The model was run for three seasons and performed well compared to an expected null distribution of breeding colonies among quarter-degree grid squares. In 2002/2003 and 2004/2005, but not in 2003/2004, quelea colonies were reported significantly more often in grid squares where the model had predicted breeding to be possible. However, in a more refined analysis of a smaller subset of colonies for which precise dates of their establishment were known, their spatio-temporal distribution matched predictions in 95%, 85% and 99% of grid squares, respectively, over the three breeding seasons. These success rates show that predictions from the model, the first of its kind for any African bird, can aid the planning of quelea control strategies.