Understanding productivity of East African highland banana

Abstract
Drought stress, potassium (K) and nitrogen (N) deficiencies are major constraints to East African highland banana (Musa spp. AAA-EA; hereafter referred to as ‘highland banana’), a primary staple food crop for over 30 million people in East Africa. This study explored the main and interactive effects of water, K and N on growth and yield of highland banana. The aim was to build a crop growth model geared towards a decision support tool for managing the crop water and nutrient requirements across agro-ecological zones. Individual plant data from three on-station fertilizer response trials in central and south-western Uganda were used to quantify the effect of drought stress on banana production and explore possible interactions with nutrient availability. Cumulative rainfall within 12 months to the date of harvest (CRF12_H) was computed from daily rainfall records for each plant harvested at maturity. Average bunch weight ranged from 8.0 to 21.9 kg between trials and cycles and was 8–28% less in dry (CRF12_H ≤ 905 mm) than in normal (905 < CRF12_H ≤ 1365 mm) rainfall periods. Linear relations were observed between CRF12_H and maximum bunch weight over the whole range of observed CRF12_H (500 – 1750 mm), whereby every 100 mm decline in rainfall caused maximum bunch weight losses of 1.5 – 3.1 kg or 8 – 10%. The drought-induced yield losses in areas with annual rainfall < 1100 mm are perhaps as high as 20 – 65%. To evaluate the highland banana dry matter allocation in response to drought stress and deficiency of K and N, individual plant measurements at harvest from two fertilizer response field trials in central and south western Uganda and the cumulative rainfall received 365 days from sucker emergence (CRF12_E) were analysed. Plants that received CRF12_E < 1100 mm were considered to have grown under dry conditions; otherwise they were considered to have grown under wet conditions. The impact of drought stress on dry matter accumulation and its partitioning between above- and below-ground biomass at harvest stage was evaluated. The main findings were verified through allometric analysis of pre-harvest stage plants sampled from farms of known K and N nutritional status and plants from a screenhouse drought stress pot trial in Uganda. Fresh bunch weight for plants that received K was about 15 kg plant‒1 irrespective of whether the plant grew under dry or wet conditions. Dry conditions without K application reduced fresh bunch weight by 50% compared to when the plant grew under wet conditions. Drought stress had no effect on DM allocation but enhanced DM allocation to below-ground biomass due to K deficiency. The phenology of highland banana was evaluated to test whether highland bananas’ flowering is independent of site (Kawanda vs. Ntungamo in central and south-western Uganda, respectively) effects. A growth analysis was also done to evaluate the relative contribution of morphological (specific leaf area and leaf mass ratio) vis-à-vis physiological (net assimilation rate) components of growth rate (RGR) to mitigation of growth reduction in response to limiting supply of water, K or N. Physiological age at flowering was delayed by 739 °C d at Kawanda compared with that at Ntungamo whose chronological age at flowering was in turn 51 d older. At both sites a threshold total dry mass of 1.5 kg per plant was required for flowering. Net assimilation rate contributed at least 90% to RGR increase due to wet conditions at both sites. A soil water balance model was adapted to the highland banana cropping system from an annual cropping system simulation framework. A sensitivity analysis was done on the adapted model to identify input parameters for calibration. The model output variables were most sensitive to the rooting depth and soil water contents at field capacity and permanent wilting point. The adapted soil water balance model was linked to the model for potential highland banana production (LINTUL-BANANA1) to explore the water-limited growth and yield of highland bananas in central and south-western Uganda. The model accurately predicted total dry weight (RMSE 0.85). The drought stress yield gap across sites was 55% and was higher (74%) at Kawanda than at Ntungamo (41%). Self-mulch reduced drought stress yield gap by 10% at Kawanda, but had no effect at Ntungamo. The model needs further calibration of the soil water balance parameters defining available water capacity and crop parameters defining dry matter partitioning between the plant structures for accurate simulation of yield components. Adding nutrient limitations, especially potassium, to the model is also recommended for comprehensive evaluation of the contribution of mulch to drought stress mitigation and sustaining highland productivity in the low-input production systems of Uganda.