INTRODUCTION
Napier grass (Pennisetum purpureum Schumach.) is one of the most productive perennial tropical grasses. It is currently used with minimal inputs and is the main fodder crop that underpins the small scale intensive dairy cattle industry in the relatively moist areas (750–2500 mm rainfall) of East Africa including at high altitudes of tropical and subtropical ecological zones (Lowe et al., Reference Lowe, Thorpe, Teale and Hanson2003). It is only moderately tolerant of drought and so its cultivation is limited by low rainfall, but primarily by low temperatures and frosts (Tessema, Reference Tessema, Mihret and Solomon2010; Tudsri et al., Reference Tudsri, Jorgensen, Riddach and Pookpakdi2002). Both growth and regeneration following harvest are known to be constrained by transient and prolonged water stress (Mwendia et al., Reference Mwendia, Yunusa, Whalley, Sindel, Kenny and Kariuki2013). Low temperatures and rainfall, therefore, are the main environmental constraints in the ongoing expansion of this grass into marginal environments. In Kenya, where rain-fed agriculture is the mainstay of the economy (Jaetzold et al., Reference Jaetzold, Schimidt, Hornetz and Shisanya2006), Napier grass cultivation has expanded into new areas, including semi-arid environments, and by 1998 accounted for 15% of the total arable land (Staal et al., Reference Staal, Chege, Kenyanjui, Kimari, Lukuyu, Njubi, Owango, Tanner, Thorpe and Wambugu1998) compared with just 4% in 1983 (Stotz, Reference Stotz1983). Further expansion would have since occurred following ongoing promotion of this grass in East Africa (ILRI, 2013) and introduction of improved lines from Brazil (Liya, Reference Liya2013).
Productivity amongst Napier grass cultivars differs depending on environmental conditions, especially rainfall. Wijitphan et al. (Reference Wijitphan, Lorwilai and Arkaseang2009) reported DM yields of 2.6–10.2 t ha−1 per harvest, in over 11 harvests of fertilized Napier grass with supplementary irrigation in a region of Thailand where daily mean temperature is in the range 16.0–36.2 °C; whereas an average DM yield of 6.8 t ha−1cut−1 was produced on fertilized sandy soil after 20 weeks in a wet (2430 mm annual rainfall) and warm (25–30°C) environment in Vietnam (Man and Wiktorsson, Reference Man and Wiktorsson2003). DM yields in the range of 5.8–16.6 t ha−1 were produced over two cuts during the season by two Napier grass hybrids in South Africa where annual rainfall averaged 750 mm (Pieterse and Rethman, Reference Pieterse and Rethman2002). In a Ghanaian environment with high rainfall and temperature (1194 mm; 21.0–34.0 °C), four Napier grass provenances produced 24.8–45.0 t ha−1 of DM per cut over three harvest periods of 30, 60 and 120 days, on sandy loam soils (Ansah et al., Reference Ansah, Osafo and Hansen2010). Further, DM yields of up to 17 t ha−1 per cut were reported on the fertile vertisols of the Ethiopian plains where annual rainfall averaged 625 mm with a cool daily temperature range of 5–26°C (Tessema et al., Reference Tessema, Mihret and Solomon2010). DM yields of up to 30.6 t ha−1 yr−1 have been reported in Brazil (Jank et al., Reference Jank, De Lima, Simeão and Andrade2013). The high yields in these studies were associated with a number of plant characteristics including profuse tillering and low leaf to stem ratio (Tessema et al., Reference Tessema, Mihret and Solomon2010), tall plants with high leaf biomass (Ansah et al., Reference Ansah, Osafo and Hansen2010; Man and Wiktorsson, Reference Man and Wiktorsson2003) and leaf area index (LAI) (Kubota et al., Reference Kubota, Matsuda, Agata and Nada1994).
Prevailing environmental conditions also modify herbage quality of Napier grass both in terms of DM partitioning between the shoots and roots and between the stem and the leaves, and in terms of digestibility. Allocation of biomass to roots is increased when/where soil–water supply is low to facilitate exploration and uptake of water and improve persistence (Yunusa et al., Reference Yunusa, Zolfaghar, Zeppel, Li, Palmer and Eamus2012). The leaf/stem ratio is associated with fibre contents and, hence, forage digestibility. This ratio is also sensitive to management, and DM in the leaf as percentage of shoot biomass reportedly fell from 80 to 60% when cutting intervals were increased from 4 to 10 weeks, while NDF increased from 64 to 75% and crude protein dropped from 15.5 to 6.8% (Man and Wiktorsson Reference Man and Wiktorsson2003). A narrow range (1.0–1.3) in the leaf to stem ratio coupled with a narrow range (72.5–74.5%) for NDF was reported for the four cultivars studied in Ghana (Ansah et al., Reference Ansah, Osafo and Hansen2010), while Tessema et al. (Reference Tessema, Mihret and Solomon2010) reported a reduction in the ratio by 33%, without significant change in NDF that remained constant (around 55%), when cutting interval was extended from 60 to 90 days. Leaf to stem ratio therefore, generally declines with continued growth especially when there is adequate soil water as more allocation goes to the stem.
Expansion of the dairy industry into dry lowlands in eastern Kenya would be more likely if suitable drought tolerant cultivars can be identified from the wide genetic pool of Napier grass provenances currently available (Lowe et al., Reference Lowe, Thorpe, Teale and Hanson2003). This study was therefore undertaken to identify accessions with high yield and quality potential for marginal agroecological zones, by examining growth in 10 accessions at two contrasting sites. The aims were to (1) characterize productivity of the accessions, and (2) identify differences in forage quality, at the two contrasting sites.
MATERIALS AND METHODS
Site descriptions
Field trials were undertaken on Kenyan dry warm eastern lowlands at Katumani (1° 35ʹ S; 37° 14ʹ E; 1600 m above sea level), and on the cooler wetter central highland at Muguga (1° 13ʹ S; 36° 38ʹ E; 2052 m above sea level). Katumani experiences mean daily temperature of 24.7 °C and an annual rainfall of 655 mm, while Muguga has a mean daily temperature of 17.6 °C and annual rainfall of 878 mm (Jaetzold et al., Reference Jaetzold, Schimidt, Hornetz and Shisanya2006). The soil was acidic sandy-clay at Katumani and clay at Muguga (Table 1).
Table 1. Textural and selected chemical properties and bulk density of the topsoil at the trial sites at Katumani and Muguga, Kenya.
OC, organic carbon, EC, electrical conductivity and each value is a mean of 4 measurements.
Experimental design
The trials were planted in November 2011 using canes of the 10 accessions in separate 4 × 4 m plots in four replicates. The plots were arranged in a randomized complete block design. The Napier grass canes used had different original sources, but in this study, all were obtained from the field genebank at the International Livestock Research Institute (ILRI) in Ethiopia, where they were assigned accession numbers: Swaziland (16,790, 16,791), Tanzania (16,783, 18,448), USA (16,806, 16,808, 16,809) and Zimbabwe (16,796); two of the lines included were hybrids of unknown parentage (16,835, 16,837) (Lowe et al., Reference Lowe, Thorpe, Teale and Hanson2003). All these accessions were selected because they had been reported as tolerant to Napier grass smut disease (Kabirizi et al., Reference Kabirizi, Muyekho, Mulaa, Msangi, Pallangyo, Kawube, Zziwa, Mugerwa, Ajanga, Lukwago, Wamalwa, Kariuki, Mwesigwa, Nannyeenya-Ntege, Atuhairwe, Awalla, Namazzi and Nampijja2015), which is a major constraint on Napier grass production in east Africa.
Trial establishment and maintenance
Land preparation was undertaken in mid-October 2011 and it involved ploughing with a tractor-drawn disc plough to 0.25 m depth followed by harrowing. Each plot was supplied diammonium fertilizer (18:20:0 of N, P, K) at an equivalent rate of 26 kg P ha−1. Canes with at least three nodes were pushed into the soil at an angle of about 45° to bury two nodes with 16 canes per plot. Planting was done in 1×1 m grids. The plots were kept weed-free manually with a hoe when needed. Calcium ammonium nitrate fertilizer (26% N) was top dressed at a rate of 13 g per stool (plants derived from one cane) and equivalent to 33.8 kg N ha−1, annually. No symptoms of disease or pest infestation were observed during the trial period. The grasses were harvested in week 23 when they attained an approximate height of 0.5 m at Muguga and 0.2 m at Katumani, for standardization. Subsequent harvests occurred at shorter intervals of about 8 weeks after regeneration. There were seven growth cycles at each site and at the end of each cycle several growth variables were measured.
MEASUREMENTS
Plant height and number of tillers
Plant height was measured and number of tillers counted a day before harvest at the end of each of the first five growth cycles on two stools that were randomly selected from each of the 40 plots. Plant height was measured from the base to the apical meristem on two randomly selected tillers per stool.
Leaf area index (LAI)
The LAI was estimated by first harvesting the youngest fully expanded leaf from every plot a day before harvesting the whole trial. They were placed in a cooler box and immediately taken to the laboratory and fresh weight determined. Leaf area was estimated by adding the rectangular section of the leaf (length × width) and the triangular section (½ × width × length). The samples were then oven dried at 60°C for 48 hrs. The relationship between the leaf area and corresponding DM weight was used to estimate total leaf area in a plot from total leaf dry weight. Plot leaf dry weight was estimated from the leaf to stem ratio obtained from the whole plot (described below) during cycles 1 to 4. The LAI was then established as the ratio of total leaf area (one-sided) to land area (Larcher, Reference Larcher2003)
DM yield and leaf to stem ratio
At each harvest, four stools were removed by cutting at 20 mm above ground level from the middle of each plot. The sampled plants were weighed fresh; then subsamples of about five tillers were taken from each sample and weighed fresh after which they were separated into leaves and stems, oven dried and weighed. These weights were used to calculate total leaf area, leaf to stem ratio and total DM yields for each plot. All remaining stools were harvested and removed from the plots to allow regrowth.
At the end of the 4th growing circle, 389 days after planting, two sets of root samples were collected from all plots with an auger (55 mm internal diameter equivalent to 0.00475 m2) at 0.2 m depth increments to 1.0 m depth adjacent to two cut stools in the plot. The two sets of samples from each plot were pooled and then washed to recover all roots from each sample. After washing the roots were oven-dried and weighed.
Nitrogen (N) content and NDF
The dry shoot samples (from leaf to stem ratio determination) were ground to pass through a 1 mm sieve and then used to determine NDF using an Ankom Fiber Analyzer (Ankom Technology Fairport, NY, USA) following the AOAC procedure (AOAC, 1975). Nitrogen content was determined according to AOAC (1980) with a segmented flow analyzer (Skalar segmented-flow auto analyzer, VW Scientific, York). Relative feed value (RFV) was estimated as described by Jeranyama and Garcia (Reference Jeranyama and Garcia2004). Briefly, this involves ranking forages against lucerne (Medicago sativa) that has a RFV of 100 at bloom. Values are obtained as a product of digestible dry matter (DDM) and dry matter intake (DMI). The two parameters are calculated from Acid Detergent Fibre (ADF) and NDF. The ADF was estimated from NDF measured as given by Bayble et al. (2007).
Weather data
Key weather variables consisting of temperature and humidity were logged every 3 hours and rainfall every 24 hours at 2 m height using solar powered self-logging weather stations (Weather station ICO348, Jaystar electronics, Sydney, Australia). The weather stations were installed during the first regrowth growth cycle at both sites. The vapour pressure deficit (VPD) was calculated from the humidity and temperature (Mwendia et al., Reference Mwendia, Yunusa, Whalley, Sindel, Kenny and Kariuki2013), and the growing degree days (GDD) as GDD = ∑(mean daily temperature − 5°C). Rainfall productivity was derived as DM yield per unit mm of rainfall over the seven growth cycles.
Data analyses
To enable more ready comparison among the accessions and identify typologies of Napier grass, cluster and principal component analysis were done to group accessions based on LAI, leaf: stem ratio, total DM and leaf yield using Minitab version 15 statistical software (Minitab 2007). Correlation coefficients (r) over selected variables (Table 2) were computed in Minitab (2007). Data on total DM yields, plant height and number of tillers were subjected to analysis of variance using Genstat version 14 (GenStat, 2011). Data were first checked for normality by doing a Shapiro–Wilk test and were linear transformed where necessary. Significant differences between means were assessed using the least significant difference test (Quinn and Keough, Reference Quinn and Keough2002).
Table 2. Correlation coefficients (r) amongst measured Napier grass attributes at Katumani and Muguga from pooled data over four growth cycles.
Coefficients are significant at *p < 0.05 or ***p < 0.001; n = 300.
RESULTS
Weather
The VPD ranged from 0.45 to 0.90 kPa with an overall mean of 0.63 kPa between the 1st and 7th growth cycles at Katumani; the GDD range was 711–948°Cd with a total of 5574°Cd over the same period. The range in mean minimum temperature was 12.9–16°C while the mean maximum temperature range was 22.5–29.6°C. The least rainfall of 1 mm was recorded during the 3rd growth cycle while most rain (337 mm) fell during the 1st growth cycle. The amounts of rainfall (mm) recorded for the following growth cycles were 14, 138, 66, 40 and 181, respectively, for growth cycles 2, 4, 5, 6 and 7. Total rainfall received over one year from November 2011 to the end of October 2012 was 766 mm which was within the expected annual range (500–800 mm/year) reported by Njarui et al. (Reference Njarui, Gatheru, Wambua, Nguluu, Mwangi and Keya2011).
At Muguga the VPD ranged from 0.28 to 0.82 kPa, and GDD was from 565 to 810°Cd. The VPD averaged 0.47 kPa and GDD totalled 4843°Cd over the seven growth cycles; while the range in mean minimum temperature was 11.0–14.0°C, and in the mean maximum was 20.7–27.4°C. The least rainfall recorded during the period was 22 mm that fell during the 6th growth cycle, while the highest of 630 mm occurred in the 1st growth cycle. Rainfall received during the remaining growth cycles was 68, 40, 291, 211 and 503 mm, respectively, for growth cycles 2, 3, 4, 5 and 7. Total rainfall recorded between November 2011 and the end of October 2012 was 1148 mm, higher than the 954 mm average annual rainfall reported by Jaetzold et al. (Reference Jaetzold, Schimidt, Hornetz and Shisanya2006).
Napier grass grouping
Cluster analysis produced three yield clusters designated as HYC, MYC and LYC (Figure 1). The HYC comprised of three ILRI accessions, numbers 16,809 (USA), 18,448 (Tanzania) and 16,791(Swaziland); the MYC consisted of 16,783 (Tanzania), 16,796 (Zimbabwe), 16,806 (USA) and 16,735 (hybrid of unknown parentage); and the LYC comprised of 16,790 (Swaziland), 16,808 (USA) and 16,837 (hybrid of unknown parentage). Productivity grouping allowed us to identify common traits within each cluster, specifically degree of tillering. This information on Napier grass typologies is useful when selecting accessions for particular uses given the large number of accessions available.
Figure 1. Cluster analysis for the Napier grass accessions productivity grouping: (a) Dendrogram of Napier grass groupings, and (b) ordination of the accessions from Principal Component Analysis. The analyses were based on dry matter yields (DM), leaf area index (LAI), leaf dry matter yield (Leaf DM) and leaf to stem ratio (LS ratio) pooled over the first 5 growth cycles at both Katumani and Muguga. Ordination loadings are based on pooled values of the above variables for the 10 Napier accessions at both Katumani and Muguga over the first 5 growth cycles. Low yielding cluster (LYC), Moderate yielding cluster (MYC) and High yielding cluster (HYC).
Leaf area index and leaf to stem ratio
The LAI at Katumani was lower for LYC than the other two clusters in growing cycle 2 and 4, and similar for both the other two cycles (Figure 2a). The lowest LAI was in cycle 3 when the index was about 1.0 and the maximum was about 4.0 observed in growth cycle 1. The leaf to stem biomass ratio at this site increased from about 0.8 –1.0 for all the clusters in growing cycle 1 to 2.5–3.5 in cycle 4. The leaf/stem ratio was often in the order HYC < MYC < LYC, except in growing cycle 1 when it was higher for MYC than the other two clusters in which it was similar (Figure 2c).
Figure 2. Mean values (± se) for leaf area index (LAI) (a, b) and leaf to stem ratio (c, d) for high yielding (HYC), medium yielding (MYC) and low yielding (LYC) clusters of Napier grass during the four growth cycles at Katumani (a, c) and Muguga (b, d). Each growth cycle lasted for 8 weeks.
The LAI at Muguga was mostly in the order HYC ≈ MYC > LYC (Figure 2b), and there were dips in the index for all three clusters in cycles 2 and 3, followed by increases in cycle 4 when it reached a value of about 6.0 for HYC and MYC. The leaf/stem biomass ratio increased from about 1.0 in growing cycle 1 to a maximum of about 2.5 in cycle 2, before falling to about 1.5 in cycle 4 (Figure 2d). The leaf/stem ratio was higher for MYC than for the other two clusters, except in growing cycle 2 when this ratio was similar for MYC and LYC.
The LAI was generally higher at Muguga than at Katumani, the difference being more than double during growing cycle 4. At the two sites, LAI for HYC and MYC was generally higher than for LYC. In contrast, the leaf/stem biomass ratio was often higher at Katumani than at Muguga, especially in growing cycle 4.
Dry matter yields
The leaf DM produced at Katumani was in the order HYC ≈ MYC > LYC in all the cycles except in cycle 2 where all clusters values were similar (Figure 3a). The leaf biomass reached a maximum of about 2.8 t ha−1 in growing cycle 1 then declined to and stabilized between 1.5 and 2.0 t ha−1 in the following two growing cycles. The total above ground DM yields, except in growing cycles that showed no clear trend (1 and 7), was always HYC ≈ MYC > LYC (Figure 3c). Minimum DM yield was measured in growing cycle 6 and the maximum of about 6.5 t ha−1 in growing cycle 1.
Figure 3. Mean values (± se) for (a, b) leaf dry matter yields and (c, d) total above ground DM for high yielding (HYC), medium yielding (MYC) and low yielding (LYC) clusters of Napier grass during seven growth cycles at Katumani (a, c) and Muguga (b, d). Each cycle lasted 8 weeks.
At Muguga, LYC had lower leaf DM than the other two clusters, which often had similar yields (Figure 3b). The leaf DM was lower for all clusters in growing cycles 2 and 3 than in cycles 1 and 4. The total above ground DM yields at this site were also generally in the order HYC ≈ MYC > LYC, with the yield for LYC being up to 50% lower than for the other two clusters, especially in cycles 1 to 4 (Figure 3d). The yields were particularly low in the growing cycles 2, 3 and 6 when biomass produced did not exceed 5 t ha−1, compared with up to 7.0 t ha−1 produced in other growing cycles.
Both the leaf DM and total DM for the three clusters at Muguga were generally higher than at Katumani, except in growing cycles 1 and 2 for total DM, when they were similar, and growing cycle 2 for leaf DM. The leaf DM at Katumani was in the range of 0.5–3.0 t ha−1 and total DM 1.0–6.0 t ha−1, compared with 0.6–4.0 t ha−1 for leaf DM and 2.0–7.5 t ha−1 for total DM at Muguga.
The accessions in the HYC were characterized with numerous tillers and large leaves. Although the sources of some of the accessions were not known, there was no obvious pattern in terms of geographical sources with respect to productivity clustering (Table 3).
Table 3. Key yield variables for Napier grass accessions during the study period at Katumani and Muguga in 2011–12 in Kenya: mean tiller density at harvest, mean plant height at harvest, and total (shoot and root) dry matter yields.
* Napier accession number as given by International Livestock Research Institute (ILRI). Productivity clusters are based on pooled standardized values for DM yield, leaf dry matter yield, leaf to stem ratio and leaf area index at Katumani and Muguga. Dry matter yield is cumulative for the seven growth cycles. Means followed by different letters with a column differ significantly (P < 0.05).
Distribution of root DM in the top 0.8 m soil profile was similar under the three clusters, but was higher at 0.2–0.4 m for the LYC than the other two clusters, at Katumani (Figure 4a). Almost 60% of the root biomass was in the top 0.4 m of the soil. The total root DM accumulated during the 13-month study period at Katumani was generally less than 1.0 t ha−1 at any 0.2 m depth level (Figure 4a). Distribution of root biomass at Muguga (Figure 4b) was similar to that observed in the soil profile at Katumani except that the root DM was higher for HYC than the other two clusters in the top 0.2 m of the soil, while MYC had lower root DM than the other two clusters in the rest of the soil profile. The total root biomass at the end of the 4th growth cycle was highest for LYC (1.6 t ha−1) at Katumani and HYC (2.6 t ha−1) at Muguga (Figure 4a, b).
Figure 4. Mean values (± se) for (a, b) root dry matter yields in the soil profile, and (c, d) root to shoot ratio, for high yielding (HYC), medium yielding (MYC) and low yielding (LYC) clusters of Napier grass measured after five growth cycles (389 days after planting) at Katumani (a, c) and Muguga (b, d).
The root/shoot ratio, based on the sum of biomass from the initial four harvests, was in the order LYC > HYC ≈ MYC at Katumani (Figure 4c), and was LYC > HYC > MYC (Figure 4d) at Muguga. The root/shoot ratio was similar at both sites for LYC and HYC, but was higher for MYC at Katumani than at Muguga. There was no significant correlation observed between root biomass and above ground biomass at either of the two sites even during water stressed periods (harvests 2 and 3).
Forage quality
At Katumani, N in the shoot was slightly depressed in growth cycles 1 and 3, when there were no differences between the clusters, compared with growth cycles 2 and 4 when it was higher in LYC than the other two clusters (Figure 5a). At this site, the average N concentration for the three clusters ranged between 2.2 and 2.8%. At Muguga (Figure 5b), the average shoot N concentration for the three clusters increased from 1.8% at the start to a maximum of 3.6% in growing cycle 2 and declined afterwards to about 2% in growing cycle 4. There were no differences in the shoot N concentration between the clusters during the first cycle, but the trend was LYC ≈ HYC > MYC during the following three growth cycles. The NDF showed a reverse trend to that of N. It was lower when N was higher and higher when N was lower (Figure 5c, d). The NDF generally declined during successive growing cycles from about 68% in cycle 1 to about 60% in cycle 4, with a slight increase in cycle 3 at Katumani (Figure 5c); the NDF trend was mostly HYC ≈ MYC > LYC at this site. NDF at Muguga declined from about 65% in cycle 1 to 60% in cycle 2 before rising to 68% in cycle 4 (Figure 5d). The RFVs of the Napier clusters over the seven growth cycles at Katumani were 66.8, 66.4 and 70.5 for HYC, MYC and LYC respectively while the values at Muguga were 68.2, 67.3 and 70.5.
Figure 5. Mean percentages (± se) for (a, b) total nitrogen (N), and (c, d) neutral detergent fibre (NDF) in the shoot for high yielding (HYC), medium yielding (MYC) and low yielding (LYC) clusters of Napier grass determined at the end of each of the four growth cycles at Katumani (a, c) and Muguga (b, d). Each cycle lasted 8 weeks.
Correlations
There were significant positive correlations between total DM and LAI, total DM and leaf yield, number of tillers and LAI, NDF and LAI, number of tillers and total DM, and total DM and NDF. Negative correlations included those between total DM and leaf to stem ratio, plant height and leaf to stem ratio, NDF and N, number of tillers and leaf to stem ratio, leaf to stem ratio and NDF, and total DM and N (Table 2). There was no precise correlation in the shoot biomass produced and rainfall observed during the seven growing cycles at either site (Figure 6a, b), though growth was generally higher at times of higher rainfall and vice versa with lower rainfall. Also total DM was higher at the site with higher rainfall (Figure 7).
Figure 6. Mean dry matter yields pooled for the three clusters and rainfall during the growth cycles in 2012/2013 at Katumani (a) and Muguga (b).
Figure 7. Total DM (shoot + root ± se). Values are sums for each accession from the 1st to 7th growth cycles then averaged for accessions within a cluster. HYC is high yielding cluster, MYC is moderate yielding cluster and LYC is low yielding cluster.
DISCUSSION
Forage productivity
There was a seasonal trend in biomass yields that was common to the three clusters and both sites (Figure 3). Biomass yield was higher at the beginning and the end of the calendar year when conditions were warmer and wetter than in the mid-year. The relative productivity potential of the three clusters in these marginal environments can be discerned from their yields during this mid-year dry period, growth cycles 2 and 3, in which rainfall was less than 5% of that received during cycle 1. At Katumani there was virtually no rainfall in that period of June to September. Relative to the biomass produced in growth cycle 1 at Katumani, yields during growth cycles 2 to 4 were reduced by 70% for LYC and 60% for HYC and MYC (Figure 3); reductions in average biomass yields between growth cycle 1 and the following growth cycles 2 and 3 at Muguga were about 40% for HYC, 36% for MYC and 60% for LYC. The LYC was thus far less able to maintain productivity during the transient drought periods than the other two clusters. Poor productivity of the LYC during the dry period was despite this cluster partitioning proportionally more biomass into the roots compared with the other two clusters (Figure 4). Although a high root/shoot ratio does not always translate into more soil-water uptake (Hamblin and Tennant, Reference Hamblin and Tennant1987; Yunusa et al., Reference Yunusa, Zolfaghar, Zeppel, Li, Palmer and Eamus2012) and/or yield (Mwendia et al., Reference Mwendia, Yunusa, Whalley, Sindel, Kenny and Kariuki2013), it has been associated with enhanced drought tolerance in container-grown temperate tall fescue (Festuca arundinacea) (Karcher et al., Reference Karcher, Richardson, Hignight and Rush2008). However, Mwendia et al. (Reference Mwendia, Yunusa, Whalley, Sindel, Kenny and Kariuki2013) found in a Napier grass cultivar that improved tissue water status under water limited supply was achieved through stomatal closure that consequently reduced carbon assimilation. The LYC in the current study had only marginal advantage in its leaf water potential or stomatal conductance over either HYC or MYC during the dry period at both sites (Mwendia, Reference Mwendia2015). This suggested mechanisms other than tissue water-status and/or water-use may be contributing to the poor performance of the LYC such as rates of carbon assimilation and/or respiration (Yunusa et al., Reference Yunusa, Thomson, Pollock, Youwei and Mead2005), which is beyond the scope of this paper.
Forage quality
Productivity in terms of biomass yield was inversely correlated with forage quality, and thus the LYC had lower NDF, despite its poor biomass yields, than either MYC or HYC. More leaves were produced as a proportion of shoot biomass by LYC than the other two clusters, especially at Katumani (Figure 3). The leaves are known to be a larger sink of N, and contain lower fibre content, than stems (Tudsri et al., Reference Tudsri, Jorgensen, Riddach and Pookpakdi2002), and tissue N is a good measure of crude protein (Moran, Reference Moran2012). Tissue N was consistently higher by up to 20%, while NDF was lower by almost a similar magnitude for LYC that produced lower biomass compared with either HYC or MYC at both sites (Figure 5). As such, the RFV was higher for LYC (70.5) compared with 66.8 for HYC or 66.4 for MYC. The inverse correlation between NDF and biomass accumulation (Table 2) reflected different growth rates amongst the clusters. The high NDF and the associated poor digestibility for HYC and MYC was probably due to an elevated accumulation of complex poorly digestible structural carbohydrates or fibres such as cellulose and lignin needed to support the observed high rates of growth (McCuistion et al., Reference McCuistion, Bean and McCollum2010; Sheaffer et al., Reference Sheaffer, Petersen, Hall and Stordahl1992).
Reduced NDF during the mid-year drought period was consistent with similar responses by two Napier grass cultivars subjected to water stress (Mwendia et al., Reference Mwendia, Yunusa, Whalley, Sindel, Kenny and Kariuki2013). More than double N in the soil at Muguga at the start of the trial (Table 1) could explain the generally higher tissue N and lower NDF (Figure 5) at this site compared with Katumani. Digestibility would have been further improved by the cooler conditions at Muguga relative to the warmer Katumani since lignin concentration is reported to increase in C4 grasses such as Napier grass under elevated temperatures (Ford et al., Reference Ford, Morrison and Wilson1979).
Site differences
Rainfall was the primary determinant of the yield differences between the two sites, as the greater energy input (°Cd) at Katumani was not reflected in high biomass. Although the relationship between rainfall received and biomass yields in each of the seven growth cycles (Figure 6a, b) was not always consistent, the cumulative biomass yields during the study period reflected total rainfall received at the sites. The average total (shoot + root) yield (total for all growth cycles per accession and then averaged per cluster) at Katumani of 22.4 t DM ha−1 was about 60% of the 36.8 t ha−1 obtained at Muguga (Figure 7) consistent with rainfall at the former site being 44% of the 1765 mm at Muguga. This resulted in rainfall productivity of 28.8 kg ha−1mm−1 at Katumani that was 39% higher than the 20.8 kg ha−1mm−1 at Muguga. Although these are approximations, the differences were probably the result of less frequent rainfall at Katumani that would have minimized evaporation from bare soil. For instance, between May 2012 and May 2013, there were 43 rainfall events at Katumani compared with 119 at Muguga. Loss of precious rainfall through wasteful evaporation appeared to have subsumed any impact of the differences in VPD, considered as a major determinant of water-use efficiency (Sinclair et al., Reference Sinclair, Tanner and Bennett1984); the difference in VPD between the sites was narrow (average of just 0.16 kPa) during the study.
There was little difference in biomass yields between HYC and MYC at either site, especially at Muguga where their yields were within 3% of each other. The yield advantage of these clusters over LYC was much larger at Muguga where the yield difference averaged 29% compared with just 17% at the dry Katumani site (Figure 7). However, the more fertile soil at Muguga produced more digestible forage that had lower NDF than at Katumani as discussed above.
CONCLUSIONS
The 10 accessions evaluated in this study differed in their productivity, but could be clustered into high, medium and low yielding groups. High productivity was associated with tillering and leaf production, both of which could serve as guides for cultivar selection. Biomass accumulation during the mid-year drought also showed that the low yielding cultivars suffered greater yield reductions. It is possible however, that in having larger root systems that the low yielding cultivars may be more persistent over the long term compared with the higher yielding clusters but this requires further investigation. The LYC had superior forage quality arising from the leaves, which had lower fibre content, accounting for more of the total biomass than in the other two clusters. Productivity at the two sites was largely driven by rainfall. In conclusion, the consistent high yields from accessions in HYC and MYC during the various growth cycles at both sites make them obvious preferred materials for future plantings at the two sites. Although LYC is disadvantaged by its much lower biomass yields, it would be worthwhile to investigate its persistence in future studies in the light of its high root/shoot ratio.
Acknowledgments
The authors would like to acknowledge financial support from the Australian Agency for International Development (Ausaid), and East African Agricultural Productivity Programme (EAAPP) for facilitation and logistics through centre Director, Kenya Agricultural Research Institute (KARI) Muguga South. We also thank the two anonymous reviewers for their much appreciated and useful comments.
