Under irrigated conditions, plant height at harvest was significantly influenced by both genotype (P ≤ 0.002) and location (P ≤ 0.001). The interaction between the genotype and site was not significant (Table 2). QQ74 grew to 108 cm tall at Bunda, followed by Bio-Bio (98 cm) (Figure 3). The plants were significantly shorter at Bembeke, with a mean height of 47 cm compared to 87 cm at Bunda (Figure 3). Bembeke is on average 5°C cooler than Bunda (Figure 2), and receives approximately 500 mm more precipitation per year. It is likely that the cooler temperatures were responsible for the shorter plant height, as cultivation occurred during the dry season and therefore irrigation would have ensured approximately equal access to water across both locations. Low temperatures similar to those experienced at Bembeke slow down enzymatic activity and leading to slow plant growth (Adams et al., 2001). At Bembeke, Bio-Bio grew to a height of 66 cm, followed by Brightest Brilliant Rainbow (64 cm) (Tables 2, 3). Plant height was strongly correlated to seed yield (0.74), biomass (0.80), and days to maturity (0.76), but showed no association with the harvest index (Table 2).

Panicle length at harvest under irrigated conditions was different across environments (P ≤ 0.005), but did not vary across genotype (Table 2). The quinoa grown at Bunda had a mean panicle length of 35.5 cm and those at Bembeke had a mean panicle length of 29 cm (Table 2 and Figure 3). No genotype × environment interaction was found for panicle length. Panicle length is a yield component in quinoa and significant variation in this parameter may entail significant grain yield differences in quinoa (Long Nguyen, 2016). Genotypes with longer panicles are generally expected to give more grain yield than those with shorter panicles. In this study panicle length differences were attributed to genotype differences rather than site effect.

Under irrigation, a significant genotype × environment interaction (P ≤ 0.001) was found for days to maturity. The genotypes and the site were also significantly different independent of each other at (P ≤ 0.001) and (P ≤ 0.001) respectively. Maturity was strongly correlated to plant height (0.80), plant height (0.76), and seed yield (0.76), but did not show any relationship with harvest index (Table 3). Jacobsen et al. (1996) also reported a positive association of quinoa grain yield with maturity period (plant cycle), plant height, length and diameter of inflorescence. This association is expected as the genotypes that were early to flower, were taller at harvest and accumulated the most biomass which translated to the highest dry grain yield per ha.

The variety Ecuadorian was the slowest to mature, taking 119 days at Bunda and 112 days at Bembeke. Inca Red (Pasankala) was also slow to mature (119 days) at Bunda (Table 2 and Figure 3). These results are consistent with those of Spehar and Santos (2005), who reported that variation in days to maturity, was found to be similar among quinoa genotypes, independent of sowing dates when evaluating them across four sites in Brazil. In these results they observed that sensitivity to photoperiod and temperature in quinoa was a function of origin. Cultivars originating from the equatorial tropics were more sensitive to photoperiod and had a longer vegetative phase. Cultivars adapted to the altiplano of Peru and Bolivia were increasingly less sensitive to photoperiod the farther they originated from the equator, with the shortest vegetative phase occurring in genotypes from the highest latitudes, such as from southern Chile, northern United States, or northern Europe. This evidence indicates that to characterize growth and development of quinoa, it is necessary to analyze the response to temperature and photoperiod in all developmental phases and using a large number of genotypes (Bertero et al., 2004).

A significant interaction was found between genotype and environment (P ≤ 0.001) for grain yield under irrigation (Table 2). Both genotype and environment effects were significantly different (P ≤ 0.001). Higher yields were obtained at Bunda (237–3019 kg/ha; Table 2) compared to Bembeke (62–692 kg/ha) (Figure 3). This could be the result of lower temperatures at Bembeke compared to Bunda during the time the experiments were conducted (Figure 2). Temperature is the main abiotic factors affecting quinoa growth, germination and productivity. Hirich et al. (2014) showed that low temperatures negatively affected the productivity of quinoa by increasing length of growing period. According to research conducted by Jacobsen and Bach (1998), the optimal temperature of growth for quinoa is approximately 22°C, and Bunda was closer to this ideal than Bembeke.

The highest yielding variety at Bunda was Titicaca (3019 kg/ha), exhibiting exceptional yield potential for optimizing sustainable intensification strategies in mid-elevation locations where irrigation is available. Multi-Hued was the highest yielding variety (692 kg/ha) at Bembeke (Figure 3). Multi-Hued was developed in Canada, where cooler temperatures are common during the growing season, therefore it is not surprising this variety performed well in the lower temperatures found at Bembeke. The lowest yielding variety at both sites was Ecuadorian which yielded 237 kg/ha at Bunda and 62 kg/ha at Bembeke (Table 2 and Figure 3). Ecuadorian also was slowest to reach maturity at both sites (112–119 days) as compared to the mean of the rest of the genotypes (101–103 days). This delayed maturity is likely a primary reason for the low yields shown in Ecuadorian. Seed yield was strongly correlated to biomass (r = 0.87; p < 0.001), plant height (r = 0.74; p < 0.001) and days to maturity (r = 0.76; p < 0.001) under irrigation. The faster maturing varieties were typically among the highest yielding. Yield differences among quinoa varieties are commonly observed, often due to differences in heat or drought tolerance (Pulvento et al., 2010; Gonzalez et al., 2012). For example, a 2 years study in Italy reported that Regalona Baer performed better than KVLQ520Y under rainfed conditions, most likely due to higher tolerance to high temperatures associated with water stress (Pulvento et al., 2010).

When grown under rainfed conditions at Bunda, differences (P ≤ 0.001) were found for grain yield, plant biomass and seed size among the 11 varieties (Table 4). The highest yielding genotype was Black Seeded (2050 kg/ha) followed by Multi-Hued (1603 kg/ha) and Bio-Bio (1446 kg/ha). Similar results of these two varieties were obtained from the trials under irrigated conditions where Multi-Hued was the highest performer in Bembeke (624 kg/ha) and fourth highest performer in Bunda (2916 kg/ha). This indicates potential for broad adaptation in the variety Multi-Hued, whereas others are perhaps more narrowly adapted to specific environments. For example, Bio-Bio and Black Seeded were the 2nd and 4th highest performers in Bembeke, but among the lowest performers in Bunda. Further testing in multiple years across multiple locations in Malawi is needed to better understand the adaptation potential in these varieties. In general, varieties grown at Bunda under rainfed conditions were lower yielding compared to the same varieties at the same location under irrigated conditions. This was due in part to the poor establishment of the crop in the rainy season, as some plants were heavily affected by heavy storm damage post-emergence. These results support other studies that show mitigation of plant stresses may help to increase yields. Geerts et al. (2008) found that well-planned deficit irrigation can stabilize quinoa yields between 1.2 and 2 Mg/ha in the central Bolivian Altiplano. For maximum production, drought stress should be mitigated by irrigation during plant establishment, flowering and early grain fill (Geerts et al., 2008).

Multi-Hued had the highest biomass yield (3612 kg/ha) followed by Black Seeded (3163 g/ha) and Bio-Bio (2942 g/ha). Interestingly, varieties with the highest grain yield also produced a lot of biomass compared to the low yielding genotypes. This could be an important consideration when developing agronomic strategies for sustainable intensification of farming systems which include quinoa as a rotational crop. There were also significant differences among varieties for seed weight. Cherry Vanilla (2.4 g) and QQ74 (2.6 g) had biggest seeds whilst Ecuadorian (1.9 g) had the lowest seed weight. Bertero et al. (1999) found that seed size was reduced under long day and hot temperatures 21 days after anthesis. Seed size was reduced by 73% in the long day hot temperature treatments (16 h at 28°C) compared to the short day and cool-temperature treatments (10.25 h at 21°C) (Bertero et al., 1999).

A significant interaction between genotype and environment was found for the harvest indices (P ≤ 0.05). Genotype was also significantly different. Titicaca had highest harvest index (0.555) at Bunda, whereas the QQ74 had high harvest index (0.503) at Bembeke (Figure 4). The varieties had a harvest index range of 0.099–0.555 at Bunda and 0.203–0.503 at Bembeke. Harvest index is an excellent parameter to assess the dry matter partitioning and efficiency of plants for mobilization of photo assimilates. Most of the genotypes showed to have high efficiency in their ability to partition dry matter into photo assimilate as their harvest index ranged from 0.3 to 0.5; this resulted into high yields.

Spehar and Santos (2005) reported that positive association among dry matter production, plant height and grain yield is translated to maturity period of the quinoa plants; where later maturing genotypes grew taller than the ones that matured early, and were also shown to be superior in other yield components. However, the study results also showed exceptions for harvest index where there were low values for late and high values for early maturing genotypes, which suggest a possibility to develop quinoa for high grain and biomass production to suit different farming systems.