An efficient method for generation of full climatological records from daily rainfall

Abstract

Three methods were compared for providing data to accompany a daily rainfall record: (i) a seasonal mean calculated from smooth curves; (ii) a seasonal mean modified by rainday effects; and (iii) actual data with missing observations filled in by method (ii). Existing plant growth models developed for the Australian tropics were used to compare the three data generation methods. These models were used to predict the number of growth days (a measure of how much plant growth would be expected in a season), green days (a measure of the duration of weight gain by rangeland cattle), runoff, and germination days. Three stations were used, one each from the monsoonal tropics, the coastal dry tropics, and the coastal subtropics. Comparatively little bias was introduced from methods (i) and (ii). Differences were greater for episodic events such as runoff and germination, than for more continuous events such as growth. There were no significant effects of generation method on calculated growth days or green days. Runoff calculated from method (i) averaged 7% lower than when actual data were used, compared with only a 3% underestimate from method (ii). The number of germination days were underestimated by between 4 and 17% when method (i) was used, compared with between a 3% underestimate and a 6% overestimate with method (ii). It was concluded that method (ii) should be used in preference to the seasonal mean alone, because of the more accurate model predictions from method (ii). A stochastic weather generator would however be required where events such as frosting would be important in risk analyses.