One strategy to feed the human population in a sustainable manner is to increase the agricultural productivity through improved management of locally available resources. To address the challenges of high rainfall variability and low soil fertility in Sub-Saharan Africa, low-cost soil moisture and soil fertility enhancing technologies, such as water harvesting (WH) and ecological sanitation (Ecosan), are particularly favoured. In this dissertation, crop growth constraints in smallholder rain-fed agriculture are analysed on river basin and regional scales in Southern Africa. The dissertation proceeds with a quantification of the potential impacts of WH, Ecosan, and other irrigation and fertilization technologies on maize yield (the local staple food crop), evaporation, transpiration, and river flow regimes. Building on this, it indentifies locations and conditions where such technologies could be suitable, and analyses their potential to improve agricultural productivity and food security.
A scenario analysis is used together with the SWAT agro-hydrological computer model (Soil and Water Assessment Tool). The Sequential Uncertainty Fitting algorithm is applied to calibrate the model against river discharge and maize yield, and to quantify prediction uncertainty. The dissertation consists of four studies detailing baseline conditions, model performance, data processing methods, and potential impacts of WH and Ecosan technologies in the Thukela River basin (29,000 km2) and throughout Southern Africa (1.8x10⁶ km²).
The first study quantifies the model performance and estimates the water availability for in situ WH, supplemental water demand (SWD) in smallholder agriculture, and the reliability of in situ WH in the Thukela River basin. On average, 0–17 mm of runoff water is available for in situ WH each year (95% prediction uncertainty range (95PPU)). This may meet some but not all of the SWD (median: 0–113 mm a⁻¹, 95PPU). The reliability of in situ WH (the percentage of years in which the water availability ≥ SWD) 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). Hence, the risk of failure of in situ WH is relatively high in many areas. These results build on a satisfactorily performing model. The Root Mean Squared Error for smallholder maize yield was 0.02 t ha⁻¹ during both the calibration and evaluation periods. The range in the coefficient of determination (R²) for discharge during the calibration (evaluation) period was 0.42–0.83 (0.28–0.72).
The second study explores the potential impacts of application of in situ WH and fertilization with stored human urine (Ecosan) in the Thukela River basin. Based on the model results, the impacts on smallholder maize yields are likely to be small with in situ WH (median change: 0%) but significant with Ecosan (median increase: 30%). The primary causes for these effects were high nitrogen stress on crop growth, and low or untimed soil moisture enhancement with in situ WH. However, the impacts varied significantly in time and space, occasionally resulting in yield increases of up to 40% with in situ WH. Full fertilization increased yields by a factor of five, whereas full irrigation reduced spatial yield variability by a third. Fertilization significantly improved the productivity of the evaporative fluxes by increasing transpiration (Ecosan median: +2.8%, 4.7 mm season^-1) and reducing soil and canopy evaporation (Ecosan median: −1.7%, −4.5 mm season^-1). In situ WH did not generally affect the river flow regimes. Occasionally, significant regime changes occurred due to enhanced lateral and shallow aquifer return flows, such as higher risk of flooding and increased low flows.
The third study focuses on improving model performance through a more efficient use of sparse climatic data. Two approaches to obtain climatic time series for SWAT were compared: the conventional centroid method and time-dynamic Voronoi tessellation. Climatic time series were estimated with each method for each sub-basin in Southern Africa. The Voronoi method utilized all available precipitation and temperature data, but the centroid method used only 14.5% and 82.5%, respectively. With the centroid method, sub-basin time series were on average 42% and 63% incomplete, respectively. With the Voronoi method, all sub-basin time series were complete. SWAT was fed with each climate dataset. Each model setup was independently calibrated and validated against 213 discharge stations and smallholder maize yields from eight provinces in South Africa. A similar performance was obtained for maize yield with both methods, but discharge simulations improved significantly with the Voronoi method. The R² between observed and simulated daily discharge increased from 0.23 to 0.39 in the calibration period and from 0.41 to 0.48 in the validation period on average. The Voronoi method improved the simulation of the river flow regime (i.e. the magnitude, frequency, timing, and flashiness of flow). The largest improvements were obtained in data scarce situations, and at relatively high spatial and temporal resolutions. In essence, the Voronoi method provides a more robust climatic dataset for assessing impacts of agricultural management alterations with the SWAT model in Southern Africa.
The fourth study quantifies the potential impacts of in situ WH, external WH, and Ecosan in Southern Africa. Ecosan significantly improved the maize yields and crop water productivity across Southern Africa, while in situ WH and external WH were only beneficial in specific seasons and locations. On average, Ecosan increased maize yields by 12.4% and transpiration by 2.2% (p-values: <2.2x10⁻¹⁶). WH did not significantly affect the yield, transpiration or river flow regimes on the Southern Africa scale (pvalues: >0.7). However, external WH more than doubled the yields in specific situations, with a supplemental irrigation amount of less than 40 mm season⁻¹. The areas where in situ WH and external WH were most beneficial are located in the Limpopo River basin, the central Free State, and along the southeast coast in KwaZulu-Natal and Eastern Cape. WH was particularly beneficial for the lowest yields. External WH generally had a larger impact than in situ WH. Significant water and nutrient demands remained even with Ecosan and WH management. By fully addressing these demands, the yields increased by a factor of four. Fertility enhancements raised yield levels but also the yield variability, while soil moisture enhancements particularly improved the temporal yield stability.
In conclusion, enhanced management of locally available water and nutrient resources with WH and Ecosan technologies can increase smallholder agricultural productivity and food production: moisture enhancements in specific conditions (typically for the lowest yields), and fertility enhancements throughout Southern Africa. To use the available water more productively, it is paramount to increase soil fertility. To reduce spatial and temporal yield variabilities, it is necessary to address crop water demands. Simultaneous approaches are the most promising.