Dry spells assessment with reference to the maize crop in the Luvuvhu River catchment of South Africa

• Timing of dry spells relative to crop stages is key for crop production.
• Prior knowledge on dry spell distribution is momentous for rain-fed agriculture.
• No significant trends indicate consistency in the behavior of dry spells.

Agricultural productivity in South Africa is negatively affected by drought as a result of frequent periodic dry spells and increasing crop water demands resulting in poor crop development and low yields. Thus, we embarked on this study which aims at investigating dry spell occurrences in relation to growing season of maize in the Luvuvhu River Catchment. Daily rainfall data (1945–2014) from 12 stations which represent the catchment fairly well was utilized in this study. Three consecutive planting dates were staggered based on three consecutive onsets of the rainy season. Dry spells were categorized into three groups: short, medium and long dry spells. The data was then subjected to theoretical distribution fitting using the Anderson–Darling goodness-of-fit test; and probabilities of occurrence were computed using a probabilistic model that best fits the data. Trend analysis was performed on the frequency of dry spells per growing period using the non-parametric Spearman's rank correlation test. Out results indicated high probabilities (≥80%) of short dry spells at all the stations irrespective of the timing of planting. Further analysis revealed that a risk of yield reduction with planting following the first onset of rains was higher than that with planting following the second and third onsets. In order to minimize this risk, farmers can be advised to plant between mid-November to mid-December. Trend analysis indicated no trend for all the various dry spell lengths except for Thohoyandou with a decreasing trend and Sigonde with a weak increasing trend in long dry spells. Such findings can be used to describe drought conditions for improvement of agricultural productivity and food security, in a given area.