Abstract. Water productivity in smallholder rain-fed agriculture is of key interest for improved food and livelihood security. A frequently advocated approach to enhance water productivity is to adopt water harvesting and conservation technologies (WH). This study estimates water availability for potential in situ WH, and supplemental water demand (SWD) in smallholder agriculture in South Africa's Thukela River Basin (29 000 km², mean annual precipitation 550–2000 mm yr⁻¹). The study includes process dynamics governing runoff generation and crop water demands, quantification of prediction uncertainty, and an analysis of the reliability of in situ WH.

The agro-hydrological model SWAT (Soil and Water Assessment Tool) was calibrated and evaluated with the Sequential Uncertainty Fitting algorithm against observed discharge (at ten stations) and maize yield (the dominant crop type) for the period 1997–2006. The water availability was based on the generated surface runoff in smallholder areas. The SWD was derived from a scenario where crop water deficits were met from an unlimited external water source. The reliability was calculated as the percentage of years in which water availability ≥SWD. This reflects the risks of failure induced by the temporal variability in the water availability and the SWD.

The calibration reduced the predictive uncertainty and resulted in a satisfactory model performance. For smallholder maize yield, the Root Mean Squared Error was 0.02 t ha⁻¹ during both the calibration and the evaluation periods. The width of the uncertainty band was reduced by 23% due to the calibration. For discharge during the calibration (evaluation) period, the ten-station range in the weighted coefficient of determination (Φ) was 0.16–0.85 (0.18–0.73), and in the coefficient of determination (R²) 0.42–0.83 (0.28–0.72). The calibration reduced the width of the uncertainty band by 25% on average.

The results show that the smallholder crop water productivity is currently low in the basin (spatiotemporal median: 0.08–0.22 kg m⁻³, 95% prediction uncertainty band (95PPU)). Water is available for in situ WH (spatiotemporal median: 0–17 mm year⁻¹, 95PPU) which may aid in enhancing the crop water productivity by meeting some of the SWD (spatiotemporal median: 0–113 mm year⁻¹, 95PPU). However, the reliability of in situ WH is highly location specific and overall rather low. Of the 1850 km² of smallholder lands, 20–28% display a reliability ≥25%, 13–16% a reliability ≥50%, and 4–5% a reliability ≥75% (95PPU). This suggests that the risk of failure of in situ WH is relatively high in many areas of the basin.