Characterizing the growing period using seasonal rainfall onset dates in the semi-arid region of Tanzania

Rainfall characteristics such as onset and cessation dates, seasonal rainfall amount, and distribution significantly impact agricultural production in rainfed systems. Studies have found that timely crop planning is necessary to maximize crop production and increase the resilience and sustainability of the rain-fed system. Thus, timely and accurate prediction of seasonal rainfall characteristics is crucial to enhance effective crop planning and minimize climate-induced crop production risks. The present study used seasonal rainfall onset dates computed using a long-term dataset, i.e., 1935–2020, acquired from the Tanzania Meteorological Authority (TMA) using Liebmann’s method to characterize the growing period in the semi-arid region of Tanzania—Kongwa district. Liebmann’s method was used due to its proven suitability in both hydrological and agronomical applications. We further used the well-known climate indices, i.e., the SOI (Southern Oscillation Index), the IOD (Indian Ocean Dipole), and NINO 3.4 averaged over the July–September period in the decision tree model, to predict the onset dates and characterize the growing period. We found the late-onset seasons—two weeks after the 7th of December—had lower rainfall (17% less than the climatological mean) and were at least 15 days shorter than the climatologically normal growing period. Moreover, the variability in seasonal rainfall in the late-onset season (CV = 28%) was found to be at least 5% higher than in the early-onset season. Late-onset seasons had a 40% chance of receiving the minimum amount of rainfall required for high-water-demand cereals like maize (450 mm). We also found SOI to be a good predictor of onset dates compared to NINO 3.4 and IOD. The SOI predicted well both normal and late-onset infections—50% and 68% precision (hit rate), respectively—compared to the IOD and NINO 3.4, whose precision was less than 10% in predicting the late onset and about 63% in predicting the normal onset. Although our results are useful to guide crop planning before the season, we recommend further studies to examine the agronomical and economic impacts the onset dates would have on crop productivity.