Understanding growth of East Africa highland banana: experiments and simulation

Key words: leaf area; radiation interception; QUEFTS model; fertilizer recovery fractions; nutrient mass fractions; crop growth; calibration; validation; radiation use efficiency; sensitivity analysis East Africa Highland banana yields on smallholder farms in the Great Lakes region are small (11−26 Mg ha−1 cycle−1 in Uganda, 21−43 Mg ha−1 cycle−1 in Burundi and 25−53 Mg ha−1 cycle−1 in Rwanda). The major causes of poor yields are declining soil fertility and soil moisture stress. In order to improve production, knowledge on highland banana physiology, growth patterns and response to fertilization is important, to establish the potential yield of the crop, to quantify the yield gaps between potential and actual yield, and to explore options for closing the yield gaps. Measurements of plant morphological characteristics, radiation interception and biomass (by destructive harvesting) were taken in experimental fields in central and southwest Uganda. Results showed that total leaf area can be estimated by using height and girth (used to estimate middle leaf area) and number of functional leaves. The light extinction coefficient, k determined from photosynthetically active radiation (PAR) measurements over the entire day was 0.7. Banana plants partitioned more dry matter (DM) to the leaves during first phase of vegetative growth, with the pseudostem becoming the dominant sink later with 58% of total DM at flowering, and the bunch at harvest with 53% of the total DM. Changes in dry matter partitioning influenced the allometric relationships between above-ground biomass (AGB in kg DM) and girth (cm), the relationship following a power function during the vegetative phase (AGB = 0.0001 (girth)2.35), and exponential functions at flowering (AGB = 0.325 e0.036 (girth)) and at harvest (AGB = 0.069 e0.068 (girth)). This thesis shows that allometric relationships can be derived and used to estimate biomass and bunch weights. In fertilizer trials, yield increases above the control (13.0 Mg ha−1 yr−1) ranged from 2.2−11.2 Mg ha−1 yr−1 at Kawanda, to more than double at Ntungamo, 7.0−29.5 Mg ha−1 yr−1 (control 7.9 Mg ha−1 yr−1). The limiting nutrients at both sites were in the order K>P>N. Differences in soil moisture availability and texture resulted in higher yields and total nutrient uptakes (K>N>P) at Ntungamo, compared with Kawanda. Per unit dry matter yield, highland bananas take up a similar amount of N (49.2 kg finger DM kg−1 N), half the amount of P (587 kg finger DM kg−1 P), and five times the amount of K (10.8 kg finger DM kg−1 K), when compared with cereal grain. Calibration results of the static nutrient response model QUEFTS using data from Ntungamo were fair (R2 = 0.57, RMSE = 648 kg ha−1). The calibrated QUEFTS model predicted yields well using data from Mbarara southwest Uganda (R2 = 0.68, RMSE = 562 kg ha−1). A new dynamic radiation and temperature-driven growth model, LINTUL BANANA 1 was developed to the compute potential yields of East Africa highland banana. The model considers (i) the physiology of the highland banana crop; (ii) the plant dynamics (i.e. three plant generations, Plant 1, 2 and 3 at different stages of growth constituting a mat); and (iii) three canopy levels formed by the leaves of the three plants. Average computed potential bunch dry and fresh matter were slightly higher at Ntungamo (20 Mg ha−1 DW; 111 Mg ha−1 FW), compared with Kawanda (18.25 Mg ha−1 DW; 100 Mg ha−1 FW), and values compared well with banana yields under optimal situations at comparable leaf area index values (20.3 Mg ha−1 DW; 113 Mg ha−1 FW). Sensitivity analysis was done to assess the effects of changes in parameters (light use efficiency, LUE; the light extinction coefficient, k; specific leaf area, SLA; the relative death rate of leaves, rd; relative growth rate of leaf area, RGRL; and the initial dry matter values) on bunch dry matter, leaf dry matter and leaf area index (L) at flowering. Sensitivity results for Kawanda and Ntungamo showed that changes in LUE1 resulted in more than proportional increase in bunch DM (1.30 and 1.36), a higher leaf DM (0.60 and 0.67) and L at flowering (0.60 and 0.67). Changes in rd1 values reduced bunch dry matter, leaf dry matter and L at flowering. Changes in SLA1 reduced only leaf DM, whereas both leaf DM and L at flowering were reduced by changes in k1 at both sites. Initial dry matter values had a small effect (sensitivity < 0.0263) for bunch DM, leaf DM and L at flowering. Based on the model results, it is clear that the potential yield of East Africa highland bananas is more than 18 Mg ha−1 DW. Management options that increase LUE and reduce the relative death rate of leaves, and improvements in parameters related to light interception (SLA and k) are important to increase yield.