Investigating factors affecting efficiency and quality of sweetpotato sprout production from on-farm, nursery plant beds

Sweetpotato has spread from its origins in the central American region, to become the sixth most important food crop globally, after rice, wheat, potatoes, maize, and cassava. It is grown in more countries than any other roots crop, and produces more biomass and nutrients per hectare than all other food crops. Sweetpotato production in many parts of the world has shifted to systems using planting materials that are practically disease (especially virus) and pest-free. Commonly, nursery plant beds generate sweetpotato vine segments on-farm, with the harvested cuttings subsequently used for field plantings. Commercial sweetpotato farms in countries including USA, China, and Australia install pathogen/pest ‘free’ sweetpotato bedding roots in 0.5-1.5 m wide beds, covered with soil. Emerged sprouts are cut when they are 10-50 cm long, and planted out in production fields, to grow sweetpotatoes for processing or fresh market sales. 

Previous studies have demonstrated that sprouts at least 14 cm long, with an intact growing tip, and with at least three nodes planted beneath the soil surface, had greater potential sweetpotato yields than sprouts missing one or more of these attributes. Longer sprouts were usually more productive. Sprouts more than 40-50 cm in length can however cause problems with handling and planting, particularly with partially mechanised planting systems. 

Australian growers use the most input (cost) intensive plant-bed systems in global commercial sweetpotato production. These growers use 100% pathogen-tested (PT) bedding roots directly from a supplier, rather than using bedding roots that have been through one or more generations of commercial field production. Bedding roots are hand-arranged on the plant beds, with one-two cm between roots, avoiding root overlap. After covering with soil, once sprouts have emerged, and are at the desirable length, they are hand-cut, sorted (with variable levels of grading intensity), and planted out within 24-48 hours. The cutting process may be repeated four-nine times during a nine-month sweetpotato planting period. This contrasts with systems, such as in the USA, where installation of bedding roots and cutting are often mechanised, with only one-three cuttings undertaken.

In the Australian sprout production system, the generation of large numbers of high-quality sprouts, over an extended growing season, is critical for subsequent successful commercial sweetpotato production. 

This thesis explores how the size and shape of bedding roots influenced their arrangement on plant beds, and subsequent sprout production (quantity and quality) over time. Data analyses based on Bayesian workflows and regression were employed to facilitate building and testing of theory-based regression models of size/shape attributes of bedding roots, for both individual sweetpotatoes, and their arrangement in plant beds. The thesis also explored how fertiliser nitrogen levels impacted sprout production for different sweetpotato cultivars. Plant bed productivity was characterised using both sprout quantity and quality attributes. Bayesian regression modelling was used as the main analytical method throughout the investigations.

Approximating sweetpotato bedding root shape as an asymmetric ellipsoid generated simple models connecting linear dimensions (longitudinal length, maximum width) to plan area, volume, and weight. Plan area is the projected longitudinal area of a resting sweetpotato, viewed from directly above. Bayesian modelling, with theoretically derived, informative prior distributions for model parameters, successfully predicted sweetpotato plan areas from measurements of length and width, or weight and width. For sweetpotatoes within the standard bedding root size ranges (maximum widths 4-9 cm), the parameterised models predicted plan areas within ±20 cm2, and weights within ±80 g, for 95% of sampled roots. 

Models for individual sweetpotatoes were expanded to predict relationships between size/shape characteristics of sweetpotatoes in a bedding root lot (e.g., 500 kg of bedding roots), and their arrangement in an installed plant bed. I defined the Bedding root Area Index (BAI) as the summed plan areas of the individual bedding roots, divided by the raised area of plant bed on which the lot was installed. In two experiments, the size of bedding roots 3-8 cm diameter did not affect BAI, with 95% of lots giving 40-50% bed coverage. Wider, or distorted bedding roots, produced lower BAI, due to less efficient arrangement at the edges of the raised plant beds. The weight of bedding roots (kg) required to install one square metre, was defined as the Bedding Root Requirement (BRR), and was approximated as three times the median width of the bedding roots. 

In Australia, sweetpotato sprouts are planted horizontally, 5-6 cm below the soil surface. Acceptable sprouts are 20-28 cm long, with premium sprouts 28-40 cm long. Production of acceptable and premium sprouts/m2 increased with bedding root size for the first four cuts. Crooked roots produced the same number and quality of sprouts as straight roots. For three out of four experiments, plant bed death, due to fungal and/or bacterial diseases, markedly reduced sprout production during the season. Death was associated with excessive/extended rain periods. Plant beds harvested at the optimal time produced 90% premium sprouts. Early cutting reduced total sprout numbers and the proportion of premium sprouts. Correctly timing sprout cutting had the greatest agronomic impact on sprout production and quality. In two experiments involving three sweetpotato cultivars (Beauregard, Orleans, Bellevue), changing the amount of nitrogen fertiliser added after each sprout cutting (10, 30 or 60 kg N/ha) did not affect total sprout production, or the proportion of premium sprouts. The amount of N removed at each sprout harvest was five times the maximum amount of N fertiliser added.

The thesis discusses implications of research techniques and modelling options for experimental results and interpretation. I explored conceptual models of bedding root installation and sprout production in Australian plant beds, with consequences for practical sweetpotato plant bed management. Bayesian, theory-based regression models were shown to be superior predictors of sweetpotato plan area and weight than the polynomial models used by other investigators. These models successfully predicted BAI and BRR, which can aid bedding root supply and pricing negotiations in the bedding root supply chain. My investigations showed minimising the risks of plant bed death, and optimising the time of sprout harvesting, were the key factors influencing sprout production and quality. The optimum bedding root width was 5-8 cm; narrower roots produced fewer acceptable/premium sprouts, while wider roots were more expensive to install, with no sprout productivity benefit. Within the optimal bedding root size range, each centimetre increase in width increased bedding root requirement by 3 kg/m2, and sprout production/m2 by 10-15%. Future research into practices to mitigate plant bed diseases, models to predict optimum sprout cutting time, and investigations into sources of sweetpotato N (particularly potential endophytic microorganisms), would be valuable for industry and are recommended.