Sweetpotato is grown throughout the year in Mozambique but drought affects storage root yield and biomass productivity. The objectives of this research were to estimate the impact of genotype × environment interactions (G × E) in sweetpotato and select genotypes based on drought indices such as geometric mean, percent yield reduction, drought sensitivity index and harvest index. A total of 58 clones were evaluated during the dry season of 2006, 2008 and 2009. Two treatments were applied for this multi-year trial: full irrigation and without irrigation at the middle of root initiation growth stage. The field layout was a randomized complete block design with three replications. ‘Jonathan’, ‘Resisto’ and ‘Tanzania’ were the check cultivars in each treatment. Storage root and vine yields were recorded at harvest in the trials. Harvest index was computed from the yield data. The analysis of variance, regression and the additive main effects multiplicative interaction (AMMI) analyses, plus phenotypic coefficient of variation and ecovalence were used for dissecting the G × E and assessing the stability of each clone. Treatment, genotype × treatment and genotype × year (G × Y) interactions had highest contributions to the variation in storage root yield observed among clones. The stability of harvest index was significantly correlated with the absolute AMMI’s IPCA1 and IPCA2 values for storage root yield. Cultivar performance varied within treatments. Four clones had significantly higher storage root yield (t ha−1) than ‘Tanzania’, the best check cultivar under drought. In conclusion, storage root yield (t ha−1) was negatively affected by drought and G × Y interaction. Harvest index stability and the geometric mean may be key to identify clones with storage root yield stability and high storage root yield under both treatments. At least two environments should be used at early breeding stages to consider harvest index in the early breeding cycle.