Parameterisation and evaluation of the FAO-AquaCrop model for a South African taro (Colocasia esculenta L. Schott) landrace

• AquaCrop was successfully parameterised for a South African taro landrace.
• Calibration and validation of AquaCrop were successful.
• AquaCrop model was able to simulate growth and yield of taro under varying water and environmental conditions.
• The model is simple and could be easily used for technology transfer.

Promotion of taro, a neglected underutilised crop, as a possible future crop under water-limited conditions hinges on availability of information describing its yield responses to water. Therefore, AquaCrop was calibrated and validated for the first time for an eddoe type taro landrace from South Africa, using data from pot, field and rain shelter experiments conducted over two seasons (2010/11 and 2011/12) at two locations (Pretoria and Pietermaritzburg) representative of semi-arid climates. Observed weather and soil physical parameters for specific sites together with measured crop parameters from optimum experiments conducted during 2010/11, were used to develop climate, soil and crop files in AquaCrop and to calibrate the model. Observations from the 2011/12 growing season and independent data were used to validate the model. Model calibration showed a good fit (R2 = 0.789; d-index = 0.920; RMSE = 2.380%) for canopy cover (CC) as well as good prediction for final biomass (RMSE = 1.350 t ha−1) and yield (RMSE = 1.205 t ha−1). Model validation showed good simulation for CC under irrigated conditions (R2 = 0.844; d-index = 0.998; RMSE = 1.852%). However, the model underestimated CC under rainfed (R2 = 0.018; d-index = 0.645; RMSE = 20.170%) conditions. The model predicted biomass (R2 = 0.898; d-index = 0.875; RMSE = 5.741 t ha−1) and yield (R2 = 0.964; d-index = 0.987; RMSE = 1.425 t ha−1) reasonably well for pooled data [field (RF and FI) and rain shelter (100, 60 and 30% ETa)]. The model also predicted biomass (R2 = 0.996; d-index = 0.986; RMSE = 1.745 t ha−1) and yield (R2 = 0.980; d-index = 0.991; RMSE = 1.266 t ha−1) well for the independent data set.