A novel approach to mapping areas suitable for rainfed production of bambara nut and cowpea.

Abstract

South Africa is a water scarce country, highlighting the need to cultivate water use efficient crops under rainfed conditions. Certain neglected and underutilised crops (NUCs) exhibit greater drought resistance, water use efficiency and nutrient content when compared to conventional crops (e.g. maize and soybean), and thus have potential to enhance food and nutrition security at the smallholder scale. To increase production of NUCs in southern Africa, small-scale farmers first need to know where these crops can be cultivated, which they typically obtain from existing land suitability maps. Various methods have been used to develop such maps, which are classified as either traditional or modern methods. Previous studies have used simulated data from a crop model to validate, but not develop, land suitability maps, especially in South Africa. Utilising modelled data to develop a land suitability map for rainfed production of taro was first conceptualised and tested in 2022, which was then modified and applied to orange flesh sweet potato (OFSP) and taro in 2024. This study represents a continuation of this work, where the methodology was further refined and improved, then applied to other NUCs (namely bambara nut and cowpea). Specific objectives include: (i) assessing the impact of four planting months (October, November, December and January) on simulated crop yield, (ii) determining the impact of frost occurrence on crop cycle length and yield, (iii) accounting for frost occurrence when mapping land suitability, and (iv) partially automating the mapping process using the Python programming language. The AquaCrop model was run to simulate 49 seasons of data (from 1950/51 to 1998/99) for over 5 800 relatively homogenous response zones across southern Africa. For each AquaCrop output variable (e.g. biomass, yield and crop water use), various statistics were then derived from the modelled data and stored in a crop database. For each NUC, the crop cycle length was calculated using thermal time, i.e. when sufficient heat units have accumulated for the crop to reach physiological maturity. Thereafter, the crop cycle length was reduced to the first frost date when the daily minimum air temperature dropped to 5.4°C or below. Hence, AquaCrop was run for two main scenarios, namely with and without frost impacting the crop’s growing season. Simulations were performed for a single planting density (representing smallholder farming systems) for each planting month. In total, 16 national scale model runs were completed. Python code, together with a geographic information system, were then used to analyse and select AquaCrop output variables, which were then used to develop each land suitability map. The land suitability maps highlight the coastal regions of KwaZulu-Natal and the Eastern Cape as being highly suited to rainfed production of bambara nut and cowpea, whilst inland areas are moderately to marginally suitable. Findings also indicate that planting date affects crop production, with bambara nut and cowpea showing optimal suitability as well as higher yields at earlier and later planting dates, respectively. Frost occurrence reduced the crop cycle length, resulting in lower crop yield simulations, which was also influenced by planting date. Areas deemed suitable for crop production were substantially reduced by (i) frost occurrence, and by (ii) excluding unsuitable land uses (e.g. protected areas and urban areas). This study illustrates how planting date affects land suitability mapping, especially in frost-prone regions. This study assessed land suitability for crop production using an approach that was further enhanced by incorporating frost occurrence in crop simulation modelling, which was achieved for the first time, and thus is considered novel. Python automation expedited map development, producing numerous land suitability maps for each crop in reduced time.