Cassava (Manihot esculenta) Varietal Growth, Yield and Cyanide Content Performance in Three Sites in the South- Eastern Semi Arid Regions of Kenya

Study Sites

Field trials were established at three sites in South- Eastern Kenya, representing semi-arid agro-ecologies: Southeastern Kenya University (SEKU) Farm in Kitui County Located at ~1,100 m above sea level, with coordinates around 1.170°S, 37.770°E. The climate is semi-arid, receiving ~500–700  mm annual rainfall (bimodal). Soils are sandy loams of low fertility. The site experiences high temperatures (mean 24–30°C) and prolonged dry seasons (June–Sept) typical of ASAL conditions.

Lukenya University Farm in Makueni County Located near Mtito Andei (approx. 2.2559°S, 37.8937°E​ , 980 m asl). The area has well-drained light-textured (sandy loam) soils​. Rainfall is bimodal (~150–650 mm per season) with long rains in March–May and short rains in Oct–Dec, and distinct dry months​. Mean annual temperatures range from 14.3°C (min) to 35.1°C (max).

Scott Christian University (SCU) Farm in Machakos County Situated near Machakos town (~1.517°S, 37.263°E, ~1,600 m asl). Rainfall is bimodal but slightly higher than the other sites (~700–800 mm/year), and soils are reddish clay- loams. This site provided a comparatively more moderate semi-arid environment.

All three sites fall in Kenya’s Lower Midland to Inner Lowland agro-ecological zones, suitable for cassava. Prior to planting, the fields were cleared and ploughed. No cassava had been grown recently at these sites, minimizing disease inoculum. Each site provided ~5 acres for the trial, allowing ample area for replication and isolation of plots.

Planting Materials and Experimental Design Eight cassava varieties (treatments) were evaluated, comprising four local landraces and four improved varieties. The local varieties were selected based on farmer popularity and adaptability in Eastern Kenya: Kasukari (a sweet-fleshed landrace, name meaning “sugar”), Mzungu (also called Musungu, a local, white-fleshed cultivar), Kitwa (a locally preferred cultivar in Kamba region), and a Makueni local landrace (an unnamed farmer variety used as a check). The improved varieties included Migyera (an improved cassava introduced from Uganda, known for mosaic disease resistance​ ) and two high-yielding lines obtained from KALRO (Kenya Agricultural & Livestock Research Organization) Kiboko station, here referred to as KME1 and KME2. These KALRO varieties were bred for early bulking, drought tolerance and disease resistance​file-skxc2x5tsoghjx4txnyyqp. One had been propagated through tissue culture (denoted TC clone) to ensure disease-free planting material.

Stem cuttings of each variety were sourced as follows: Kasukari, Mzungu, Kitwa and the local landrace cuttings were obtained from farmers in Makueni/Machakos counties; Migyera cuttings came from KALRO Katumani; and the KME1, KME2 cuttings (including the TC clone) were obtained from KALRO Kiboko. Figure 1 all cuttings were ~30 cm long with 8–12 nodes, taken from healthy, disease-free mother plants. Before planting, cuttings were treated with wood ash to prevent fungal rot and sorted into uniform sizes.

Click to enlarge

At each site, the experiment was laid out in a Randomized Complete Block Design (RCBD) with four blocks (replicates). Each block contained plots for all eight varieties randomly assigned. Per variety per replicate, a plot of 10 m × 10 m (100  m²) was planted, containing 100 plants (10 rows × 10 plants) at a spacing of 1 m × 1 m. Planting was done on raised ridges about 30 cm high, spaced 1 m apart, as is recommended for cassava in semi-arid soils​. Cuttings were planted at a 45° angle with two-thirds of the cutting buried and one-third above ground​. This orientation encourages root development. The trials were planted at the onset of the long rains (late March 2022 for Lukenya and Machakos, early April 2022 for Kitui site).

Crop Management

Uniform agronomic practices were applied at all sites to minimize management variability. Basal fertilizer was not broadly applied because cassava can grow in low- fertility soils; however, each site received a moderate starter application in poorer patches (60  kg N/ha as urea), in line with recommendations for cassava on depleted soils​. Supplemental irrigation was provided at planting and during the first 3 months (up to ~June) to ensure good establishment since rains were erratic. Thereafter, crops relied on rainfall except at SEKU where a brief irrigation was applied during an extended mid-season dry spell. Weed control was done via hand weeding at 3, 8, and 24 weeks after planting​. Earthing up (hilling) was performed at 2 months to cover exposed tubers and improve root development​. No pesticides were applied; instead, fields were monitored for pests and disease. Only minimal pest incidence (occasional grasshopper defoliation) was observed and controlled manually. Likewise, no significant cassava mosaic virus symptoms were noted on the improved varieties; a few plants of the local landraces at SCU showed mild foliar mosaic which did not spread. This indicated the planting materials were disease-free and/ or varieties like Migyera have genetic resistance. The crops grew for a full season (12 months) to allow late-maturing varieties (bitter types) to bulk, as sweet types typically mature by 8–10 months. Harvest was done in March–April 2023 at all sites.

Data Collection

Phenotypic Growth Traits A sample of plants in each plot was measured to characterize growth and morphology. At 6 months after planting (mid-season), plant height (from ground to top of canopy) was measured for 10 representative plants per plot using a measuring pole. The number of primary branches per plant was counted (with growth habit classified as either erect (no branching) or branched). We also recorded foliage traits: for each variety, the color of leaves and petioles, and the stem cortex color and internode pattern were noted qualitatively, following descriptors. For example, petiole colors ranged from green or cream to reddish or purple among varieties​. The number of leaf leaflets was counted on fully expanded leaves (5th–10th leaf from the apex) for 5 plants per plot​. These morphological observations helped distinguish varieties and assess any stress effects (e.g. leaf retention).

Yield Parameters At harvest (12 MAP), all plants in each plot were harvested to determine yield. The number of marketable storage roots per plant was counted for 10 plants per plot (randomly selected). Total root weight from those plants was recorded using a field scale. From these, the average tuber weight and tubers per plant were computed. Yield in tons per hectare (t/ha) was then calculated for each plot by extrapolating the average per-plant yield over the planting density (10,000 plants/ha).

Yield/ha was also verified by weighing total harvests from each plot and converting to t/ha. Any unmarketable roots (diseased or <2  cm diameter) were excluded from yield. Throughout the growing season, notes were taken on any biotic stresses, notably, presence of cassava mosaic disease (CMD) symptoms, cassava green mite damage, or wildlife/herbivore damage, to complement yield data with pest resistance observations.

Hydrogen Cyanide (HCN) Analysis To assess cyanide levels, tissue samples were collected at 6  MAPS (when roots were developing but before full maturity). From each variety plot, we sampled the first fully expanded leaf from the top of a plant plus the next two leaves, and a piece of root (~10 cm³) from an actively bulking tuber​. These samples were immediately placed in labelled paper bags, sun-dried slightly in the field, and then oven- dried at 60°C in the lab to a constant weight. Dried samples were milled into powder for HCN testing. We followed the picrate paper method (Field cyanide test) as described by Bradbury (1999) for semi-quantitative analysis, with spectrophotometric confirmation. In brief, ground samples were mixed with phosphate buffer and linamarase enzyme to release cyanide, which was captured on alkaline picrate paper. The color change was compared to standards to estimate total HCN content​. Additionally, a subset of samples was analysed using a UV-Visible spectrophotometer (at 485  nm) via the acid hydrolysis/ninhydrin method for precise quantification​. Analytical blanks and known cyanide standards were run to calibrate the measurements. HCN concentration was expressed in mg HCN per kg fresh weight of the root or leaves (parts per million). Each variety’s root HCN was the average of three replicate samples.

Statistical Analysis Data from the three sites were initially analyzed separately and then combined for an overall analysis. Analysis of Variance (ANOVA) was performed using SAS (v9.4)​. For individual sites, one-way ANOVA tested varietal effects on growth and yield metrics. For the combined dataset, two-way ANOVA (General Linear Model) was used with variety and site as factors, to assess consistency of varietal performance and any variety×site interactions. Mean differences were evaluated using Fisher’s Least Significant Difference (LSD) at 5% significance level​. Traits recorded as counts (e.g. tubers/ plant) were square-root transformed before ANOVA to meet assumptions but are presented as back-transformed means. The cyanide data were analyzed in a split-plot manner with plant part (leaf vs root) as a subplot factor; a significant variety × plant part interaction on HCN distribution was observed, so root HCN was analyzed separately for varietal differences​. Pearson’s correlation analysis was conducted to determine relationships among key variables (leaflet number, height, tuber count, yield, and HCN) Maitha, et al. [6]. Results are reported as mean ± standard error. Significance is reported at p<0.05 unless stated otherwise.

Agronomic Performance Across Varieties and Sites

Plant Growth and Morphological Traits

The cassava varieties showed distinct growth habits and morphology. Petiole and stem colors varied by genotype, matching farmer descriptions: Kasukari had purple petioles, Kitwa cream, Makueni local dark brown, and Mzungu red​. Stem cortex color ranged from grey green in Kasukari/Kitwa to brownish in Mzungu/local​. All varieties had the typical palmate leaf shape, but leaflet counts differed significantly (p<0.001) among varieties​. Kasukari and Mzungu had the most leaflets per leaf (usually 7–9 leaflets), indicating robust foliage, whereas the local landrace had fewer (mostly 5–7). This corresponded with observed canopy density: Kasukari and Mzungu developed a dense leaf canopy, whereas the local variety and one KALRO clone had sparser leaves, partly due to some early leaf drop in the dry season.

Plant Height: Varied markedly by genotype (ANOVA p<0.001)​. Figure 1 illustrates the average plant height at 6 MAPS for each variety across sites. The best performers, Kasukari and Mzungu, were among the tallest plants, reaching 2.3–2.7  m in height by mid-season. Migyera was similarly tall (~2.5 m). In contrast, the tissue-cultured KALRO clone (KME1-TC) was significantly shorter, averaging only ~1.6 m​ at 6 MAP. Kitwa and KME2 were intermediate (~2.0–2.2 m). These differences persisted until harvest, as taller varieties maintained a height advantage. The branching pattern also differed: Kasukari typically grew erect with minimal branching until late in the season (often a single stem with a branchy crown at top), while Mzungu and Kitwa produced a few basal branches earlier (2–3 primary branches per plant on average). The local check had an intermediate habit (some plants branched; others did not). Despite these differences, all varieties established well. By 3 MAPS, survival was over 90% in all plots, indicating good adaptation and the effectiveness of supplementary irrigation in establishment.

Site Conditions: Influenced overall growth vigor, but the ranking of varieties was similar. The SCU site (Machakos) had slightly taller and more branched plants on average, likely due to better soil fertility. For instance, Kasukari reached ~3 m at SCU vs. ~2.5 m at the drier SEKU site. Meanwhile, the SEKU site (Kitui) experienced more moisture stress; all varieties were somewhat shorter and had smaller canopies there. Nonetheless, the top varieties (Kasukari, Mzungu, Migyera) maintained their superiority in height and leafiness at each location. No significant variety × site interaction was found for plant height (p>0.1), indicating that varietal differences in height were consistent across the different environments.

Yield and Tuber Characteristics: There were highly significant differences in cassava yield among the eight varieties (p<0.01). Table 1 summarizes the yield performance at each site and the overall means. Across all sites, Kasukari, Mzungu, and Migyera achieved the highest fresh root yields, each averaging 15–18 tons per hectare (t/ha)​. In individual site analyses, these three were top ranked with yields often double those of the lowest variety. For example, at Lukenya University, Mzungu produced 18.2 t/ha and Kasukari 17.5 t/ ha, significantly outperforming the local check (9.8  t/ha). Migyera, tested at SEKU and SCU, yielded similarly high (e.g. 16.0 t/ha at SEKU, the highest at that site). In contrast, the Makueni local landrace and one KALRO improved line recorded the lowest yields, between 8–11 t/ha. The tissue- culture propagated KME1 clone yielded poorly especially at Lukenya (only ~8 t/ha), corresponding to its stunted growth. Kitwa and the second KALRO line were intermediate, typically 12–14 t/ha. An LSD test (α=0.05) grouped Kasukari, Mzungu, and Migyera together as “a” (highest yield), significantly above the local check and TC clone which were group “c” (lowest), with Kitwa and the other improved line as “b” group.

Table 1: Fresh root yield of eight cassava varieties at three sites (tons per hectare). Different letters indicate significant differences among varieties per site (LSD, p<0.05). Source: Field trial data, 2022–2023. (Note: Migyera was not planted at Lukenya site; “–” indicates no data.) Overall, the Machakos site had the highest mean yield (14.1  t/ha), followed by Makueni (13.6) and Kitui (12.5). Yields at SCU were slightly higher for most varieties, likely due to better soil and possibly lower heat stress at the higher altitude. Kitui’s lower rainfall limited yields, though Migyera notably still achieved 16  t/ha there, reflecting its drought tolerance. The combined ANOVA showed a significant effect of site (p<0.05) and variety (p<0.01) on yield, but the variety × site interaction was not significant (p=0.34), indicating the yield ranking of varieties was relatively stable across the different environments.

In terms of yield components, Kasukari, Mzungu, and Migyera’s superior yields were mainly attributable to more and larger storage roots per plant. These varieties averaged 6–8 marketable tubers per plant, each weighing ~1.5–2.0 kg. In contrast, the local check had fewer tubers (3–5 per plant) and some were small (<1  kg). Kitwa and the KALRO lines had intermediate tuber counts (5–6) and weights. The differences in tuber number were statistically significant (p<0.001)​. For instance, Kasukari produced the most tubers per plant (mean 7.8) while the local produced 4.1. Tuber size differences were evident anecdotally (with Kasukari often called “Kasukari” because of its plump, sugar-sweet roots), but within the marketable category, average root weight did not differ as sharply as counts. All varieties produced elongate fusiform roots typical of cassava. Kasukari’s roots had pale cream pulp; Mzungu’s were brown-skinned; Kitwa and local had white pulp, and Migyera’s roots were yellow- cream inside (suggesting moderate beta-carotene content, though not measured).

No serious pest or disease damage was observed that could confound yield. Only a few plants (mostly of susceptible local variety) at SCU showed mild cassava mosaic virus symptoms; these plants had slightly lower yield, but because incidence was <5%, the overall yield trends by variety still reflected genetic potential rather than disease pressure. The improved variety Migyera (bred for virus resistance) showed no CMD symptoms at all, confirming its known resistance​. There were also no significant issues with wild animals or theft, as the plots were fenced and guarded, ensuring a fair assessment of varietal yield.

Hydrogen Cyanide (HCN) Levels in Roots: The cassava varieties differed significantly in cyanogenic potential (HCN content) of their tissues (p<0.001 for variety effect)​ . In all cases, cassava leaves had higher HCN than roots, but we focus on root HCN since that affects edible safety. The HCN concentration in fresh cassava roots showed by variety. Kasukari had the lowest root cyanide level, averaging 46.9 mg HCN per kg fresh root, which was significantly lower than most other varieties​. Mzungu was the second lowest at ~50.5 mg/kg, not statistically different from Kasukari (LSD

8.4 mg/kg)​. In contrast, the bitter landraces Kitwa and the Makueni local had the highest cyanide concentrations at 76.4 and 69.1 mg/kg respectively​. These were about 1.5 times the levels in Kasukari/Mzungu and were statistically grouped as the highest HCN (see, where Kitwa and Local are labelled “a” vs “b” for Kasukari/Mzungu). The improved variety Migyera showed moderate HCN content (~60  mg/kg) based on samples from two sites. The KALRO clones had intermediate cyanide as well: for instance, the TC clone in Makueni had ~55–60  mg/kg. Across all varieties and sites, root HCN ranged from a minimum of ~45  mg/kg to a maximum of ~80  mg/kg in this trial. It is noteworthy that all varieties exceeded the recommended safe limit of HCN for cassava flour (10 ppm, equivalent to 10 mg/kg)​ [7]. This means none of the cassava roots are safe to eat raw and must be processed (soaked, dried, or cooked) to reduce cyanide to safe levels. The varieties traditionally considered “sweet” (e.g. Kasukari) indeed had less cyanide than known “bitter” ones (Kitwa, etc.), but even the sweet ones had total HCN well above 10 mg/kg fresh weight. However, local preparation methods (peeling, drying, prolonged boiling) can reduce cyanide by 80–90%, making these roots safe for consumption after processing [8].

The distribution of HCN between leaves and roots differed by variety. Leaves generally had higher HCN than roots in most varieties, but interestingly, Kasukari showed a much higher leaf-to-root HCN ratio (its leaves had ~79 mg/ kg vs 47 mg/kg in roots)​, suggesting Kasukari retains more cyanide in foliage (perhaps as defense) and less in the edible root – a desirable trait. Mzungu showed a similar but smaller gap (69 in leaves vs 50 in roots) ​. In Kitwa and the local landrace, leaf and root HCN were both high (~80 in leaves and ~70–76 in roots)​, indicating these varieties are uniformly cyanogenic. The variety × plant part interaction was significant (p<0.001)​, confirming that the reduction of cyanide from leaves to roots was much greater in Kasukari and Mzungu than in Kitwa/local. This trait of partitioning cyanide more to leaves (and peels) is characteristic of so- called “sweet” cassava​file-fktguv8zzpama6twywmb8w. It means that in Kasukari and Mzungu, peeling and discarding leaves greatly lowers the cyanide hazard, whereas in Kitwa even the peeled roots contain substantial HCN. Migyera and the KALRO lines were also observed to have most cyanide in the peels (field taste tests found Migyera roots to be quite bitter in peels but less so in pulp, like a sweet cultivar). These findings align with laboratory analyses in similar environments: for example, Maitha, et al. also reported Kasukari’s root HCN as ~47 mg/kg and Kitwa’s ~76 mg/kg​, consistent with our results.

In summary, Kasukari, Mzungu, and Migyera can be classified as lower-cyanide (relatively sweet) varieties, whereas Kitwa, the local landrace, and to some extent the KALRO clones are higher-cyanide (bitter) varieties. All require processing, but the risk of acute toxicity is greater with the latter group if not properly processed. This has important implications for adoption – farmers and consumers may favor Kasukari and Mzungu not only for yield but also for taste and safety, as evidenced by Kasukari’s name (meaning “sugary” in local language, reflecting low bitterness).

Cross-Site Comparison: While the overall performance trends were similar at the three sites, some environmental influences were evident. Yields were on average ~10–15% lower at the driest site (SEKU, Kitui) compared to the wetter site (SCU, Machakos) Nonetheless, the top three varieties (Kasukari, Mzungu, Migyera) maintained high yields even in Kitui, indicating good drought tolerance. Migyera, for instance, showed remarkable stability – yielding ~16 t/ha in Kitui, virtually the same as in Machakos (where rainfall was higher). In contrast, the local variety’s yield dropped more sharply under drought (from 11.7 t/ha in Machakos to 9.8 t/ha in Kitui), suggesting it is less drought resilient. Kasukari and Mzungu, being landraces from the region, handled the stress well and even in Kitui yielded >13 t/ha, still outperforming others. Soil differences may have also played a role: the Machakos site’s slightly higher fertility likely boosted yields of moderate performers like Kitwa and KALRO clones (they yielded ~1–2  t/ha more at SCU than at the other sites). However, Kasukari and Migyera were less affected by soil differences, possibly due to better nutrient foraging or efficient use of available moisture. No significant site-by-variety interaction for yield was detected statistically, which means the ranking of varieties did not change drastically across sites – an encouraging result for recommending the best varieties widely in the region.

In terms of HCN, we observed anecdotally that plants grown under greater water stress (Kitui) had slightly elevated cyanide levels compared to more favourable conditions (this aligns with reports that drought can increase cassava’s cyanogenic glucoside production). For example, the local variety roots from Kitui were extremely bitter and likely had HCN at the higher end of the range (~75–80 mg/kg), whereas in Machakos the same variety’s roots, while still bitter, seemed a bit less so (~65–70  mg/kg). Kasukari’s cyanide stayed relatively low across sites, which is advantageous. Thus, selecting inherently low-CN genotypes like Kasukari is beneficial especially in stress-prone ASAL environments where environmental triggers might otherwise raise cyanide content.

No major differences in pest or disease incidence occurred across sites. All sites remained largely free of cassava mosaic disease, likely due to the use of resistant varieties and disease-free cuttings. At SCU (Machakos), a very slight presence of foliar mosaic on a few local plants under a shaded edge was recorded, whereas none was seen at Kitui or Makueni. The lack of widespread disease pressure meant that improved varieties like Migyera did not have a visible advantage in that aspect during this trial – but in areas with higher virus pressure, Migyera’s resistance would be critical.

Correlation Analysis: Pearson correlation coefficients were calculated pooling data across varieties and sites. There was a strong positive correlation between number of leaflets and plant height (r = 0.590, p<0.05) and likewise between leaflet number and tubers per plant (r = 0.562, p<0.05)​ . This indicates that vigorous vegetative growth (more leaflets, taller plants) translated into higher yield (more tubers), which is logical as a bigger canopy can produce and translocate more photosynthate to roots. Plant height was also positively correlated with tuber count (r ≈ 0.67, p<0.05, data not shown), reinforcing that taller, well-branched plants yielded more. These relationships were exemplified by varieties like Mzungu and Kasukari, which had both lush canopies and high tuber counts. On the other hand, the HCN concentration in roots showed a slight negative correlation with tuber yield (r ≈ –0.045, n.s.)​, hence not statistically significant, but a trend whereby high-cyanide varieties tended to yield less. For example, Kitwa and the local landrace (high HCN) were among the lower yielders, whereas Kasukari and Mzungu (low HCN) were top yielders. This inverse trend (though weak here) aligns with the idea that “sweet” cassava often results from breeding for human consumption traits which may coincide with better agronomics, whereas some bitter varieties prioritize defense (cyanide) at the expense of yield. The correlation between HCN and leaf traits was also examined: there was a positive correlation between leafiness and HCN in leaves (since vigorous plants have more cyanogenic glucosides overall), but importantly, high leaflet count correlated with lower proportion of cyanide in roots (since Kasukari had many leaves but low root HCN). These nuances suggest that selecting for strong vegetative growth and yield does not necessarily increase cyanide risk in the edible roots – as seen in Kasukari. In summary, the correlation analysis supports the agronomic findings that yield-related traits cluster together, and it hints that low cyanide and high yield can co-exist in certain genotypes.