Futuristic rainfall projections are used in scale and various climate impact assessments. However, the influence of climate variability on spatial distribution patterns and characteristics of rainfall at the local level, especially in semi-arid catchments that are highly variable and are not well explored. In this study, we explore the influence of climate variability on the spatial distribution and rainfall characteristics at a local scale in the semi-arid Shashe catchment, Northeastern Botswana. The LARS-WG, Long Ashton Research Station Weather Generator downscaling method, three representative scenarios (RCP 2.6, RCP 4.5, and RCP 4.5), three trend detection methods (Mann-Kendall, Sen’s slope, and innovative trend analysis) and L-moment method were used to assess climate change impacts on rainfall. Two data sets were used; one with 40 years of observed data from 1981–2020 and the other with 70 years from 1981–2050 (40 years of observed and 30 years of projected data from 2021–2050). Generally, the study found trend inconsistencies for all the trend detection methods. In most cases, Sen’s Slope has a high estimate of observed and RCP 2.6, while ITA overestimates rainfall totals under RCP 4.5 and RCP 8.5. The trend is increasing for annual total rainfall in most gauging stations while decreasing for annual maximum rainfall. The catchment is homogeneous, and Generalized Logistic distribution is the dataset’s best-fit distribution. Spatial coverage of a 100-year rainfall between 151–180 mm will be 81% based on observed data and 87% based on projected data under RCP 2.6 scenario when it happens. A 200-year rainfall between 196–240 mm under RCP 4.5 and 8.5 has high spatial areal coverage, at least 90% of the total catchment. The outcomes of this study will provide insightful information for water resource management and flood risk assessment under climate change. There is a need, however, to assess the transferability of this approach to other catchments in the country and assess the performance of other advanced modelling systems, such as machine learning, in this region.

No abstract