Introduction
In most parts of sub-Saharan Africa, up to 90% of the food consumed is produced under small-scale rain-fed conditions. Southern Africa is generally considered to be most vulnerable to the devastating effects of climate variability (Rohrbach, Reference Rohrbach2003; Challinor et al., Reference Challinor, Wheeler, Garforth, Craufurd and Kassam2007; Mapfumo et al., Reference Mapfumo, Adjei-Nsiah, Mtambanengwe, Chikowo and Giller2013). Although projections are rather uncertain (Lobell and Burke, Reference Lobell and Burke2008; Lobell et al., Reference Lobell, Burke, Tebaldi, Mastrandrea, Falcon and Naylor2008), the negative effects of climate variability are likely to be increased temperatures and erratic rainfall (Challinor et al., Reference Challinor, Wheeler, Garforth, Craufurd and Kassam2007; Thornton et al., Reference Thornton, Jones, Ericksen and Challinor2011; Cairns et al., Reference Cairns, Crossa, Zaidi, Grudloyma, Sanchez, Araus, Thaitad, Makumbi, Magorokosho and Bänziger2013a ). Unreliable weather forecasts increase the uncertainty for smallholder farmers to respond to potential future climate threats.
Options are needed to respond to the increased risk associated with a variable climate to achieve food security (Cairns et al., Reference Cairns, Sonder, Zaidi, Verhulst, Mahuku, Babu, Nair, Das, Govaerts, Vinayan, Rashid, Noor, Devi, San Vicente and Prasanna2012). The underlying hypothesis of this paper is that a combination of conservation agriculture (CA) and locally adapted and drought-tolerant (DT) open pollinated varieties (OPVs) will enable farmers to better respond to a changing climate.
CA is a cropping system based on minimum soil disturbance, retention of living or dead plant material as soil cover and rotation or intercropping of different crop species (Kassam et al., Reference Kassam, Friedrich, Shaxson and Pretty2009). Plot-level benefits from southern Africa have shown that minimum soil disturbance and crop residue cover has the potential to maintain or increase high levels of water infiltration which leads to more available water in the soil and groundwater (Thierfelder and Wall, Reference Thierfelder and Wall2009, Reference Thierfelder and Wall2010). Mulch cover on the soil surface lessens evaporative losses and maintains a positive water balance (Roth et al., Reference Roth, Meyer, Frede and Derpsch1988; Findeling et al., Reference Findeling, Ruy and Scopel2003; Thierfelder and Wall, Reference Thierfelder and Wall2009; Rockström et al., Reference Rockström, Kaumbutho, Mwalley, Nzabi, Temesgen, Mawenya, Barron, Mutua and Damgaard-Larsen2009). The combined effects of minimum soil disturbance, crop residue retention and rotations may also lead to slow increases in soil carbon (Scopel et al., Reference Scopel, Findeling, Chavez Guerra and Corbeels2005) which further improves water-holding capacity, specifically in sandy soils (Chivenge et al., Reference Chivenge, Murwira, Giller, Mapfumo and Six2007; Thierfelder and Wall, Reference Thierfelder and Wall2012).
Besides adaptation to the effects of climate variability, CA has been demonstrated to reduce the emission of greenhouse gases (CO2 and CH4) (Reicosky and Lindstrom, Reference Reicosky and Lindstrom1993; Reicosky, Reference Reicosky2000; Johnson et al., Reference Johnson, Reicosky, Allmaras, Sauer, Venterea and Dell2005; O'Dell et al., Reference O'Dell, Sauer, Hicks, Thierfelder, Lambert, Logan and Eash2015) although there is controversy to what extent CA contributes to mitigation in tropical and sub-tropical environments (Powlson et al., Reference Powlson, Stirling, Jat, Gerard, Palm, Sanchez and Cassman2014). Nevertheless, CA is considered to be a ‘climate-smart agriculture system’ with the potential to reduce the negative effects of climate variability (Thierfelder and Wall, Reference Thierfelder and Wall2010).
Although the strategy of using CA shows promise, there are also concerns about CA as a solution to constraints of small-scale farmers in southern Africa (Bolliger, Reference Bolliger2007; Giller et al., Reference Giller, Witter, Corbeels and Tittonell2009; Baudron et al., Reference Baudron, Tittonell, Corbeels, Letourmy and Giller2012). While field-level benefits seem to be widely acknowledged, there are contestations on how CA fits at the farm and community levels (Thierfelder et al., Reference Thierfelder, Rusinamhodzi, Ngwira, Mupangwa, Nyagumbo, Kassie and Cairns2014).
Breeding for drought-tolerance could assist farmers as well to respond to the negative effects of climate variability in Africa (Cairns et al., Reference Cairns, Sonder, Zaidi, Verhulst, Mahuku, Babu, Nair, Das, Govaerts, Vinayan, Rashid, Noor, Devi, San Vicente and Prasanna2012; Cairns et al., Reference Cairns, Crossa, Zaidi, Grudloyma, Sanchez, Araus, Thaitad, Makumbi, Magorokosho and Bänziger2013a ). In the past, farmers in drought-prone areas grew crop species such as sorghum and millet which are more drought-tolerant than maize (Blum and Sullivan, Reference Blum and Sullivan1986; Muchow, Reference Muchow1989). In large areas of sub-Saharan Africa, sorghum and millet are therefore chosen as subsistence cereals (Rohrbach, Reference Rohrbach2003). However in southern Africa, farmers prefer to grow maize due to its high yield potential in seasons with normal rainfall distribution and the taste of the maize porridge, which is preferred over sorghum. Nevertheless, they risk losing 2 out of 5 years maize harvests due to drought (Nyamangara et al., Reference Nyamangara, Masvaya, Tirivavi and Nyengerai2013). Maize has become the predominant crop in Southern Africa (Dowswell et al., Reference Dowswell, Paliwal and Cantrell1996; Kassie et al., Reference Kassie, Erenstein, Mwangi, La Rovere, Setimela and Langyintuo2012) and strategies are required to better adapt maize to the effects of climate variability. The projected increased frequency of mid-season dry spells (Tadross et al., Reference Tadross, Suarez, Lotsch, Hachigonta, Mdoka, Unganai, Lucio, Kamdonyo and Muchinda2009) requires varieties that can withstand long periods of water stress.
Substantial progress has been made in increasing grain yields under drought stress (Cairns et al., Reference Cairns, Crossa, Zaidi, Grudloyma, Sanchez, Araus, Thaitad, Makumbi, Magorokosho and Bänziger2013a ). New varieties in CIMMYT's eastern and southern Africa drought breeding program are evaluated under three water treatments: (i) managed drought stress, with stress applied at flowering; (ii) random drought stress; and (iii) optimal, well-watered conditions. Promising hybrids and OPVs with improved grain yield under managed drought stress are further tested in regional trials managed by CIMMYT, national partners and private seed companies in eastern and southern Africa at more than 90 locations (Magorokosho et al., Reference Magorokosho, Vivek and MacRobert2009). This approach has been successfully used to develop new varieties (hybrids and OPVs) with improved grain yield under drought stress and higher yield potential (Setimela et al., Reference Setimela, MacRobert, Atlin, Magorokosho, Tarekegne, Makumbi and Taye2012). Recent on-farm trials in eastern and southern Africa of new drought-tolerant hybrids showed a 35 and 25% yield advantage against farmers, own varieties under low (<3 t ha−1) and high yield (>3 t ha−1) conditions, respectively (Cairns et al., Reference Cairns, Crossa, Zaidi, Grudloyma, Sanchez, Araus, Thaitad, Makumbi, Magorokosho and Bänziger2013a ). To facilitate farmers' access to these new varieties, extensive training and support is currently being given to small and medium enterprises within eastern and southern Africa to enable seed production and dissemination and variety registration (Cairns et al., Reference Cairns, Hellin, Sonder, Araus, MacRobert, Thierfelder and Prasanna2013b ).
The inclusion of legume crops such as pigeonpea (Cajanus cajan (L.) Millsp.) and cowpea (Vigna unguiculata (Walp)) in the cropping systems may further help to overcome climate related losses as both crops are drought resistant and often ensure food security when the main crop fails (Rusinamhodzi et al., Reference Rusinamhodzi, Corbeels, Nyamangara and Giller2012).
The objective of this study was to test different drought-tolerant maize varieties under CA and assess their performance under different agro-ecological environments. We tested both CA systems and drought-tolerant maize varieties on farmers' fields in central and northern Mozambique. Special emphasis was placed on possible synergistic effects of both improved management practices compared with the current traditional farmer practices. Several surveys were further commissioned to better understand farmers' ability to access improved maize varieties and also to capture farmer's perception and selection criteria for improved varieties.
Materials and Methods
Site descriptions
The study was carried out between 2007 and 2014 in five target communities of central and northern Mozambique (Table 1, Fig. 1). Mozambique is a sub-tropical country with a sub-humid and in some parts arid climate, situated between latitude 10° and 27°S and 30° and 41°E in south-eastern Africa. The selection of target communities was based on a rainfall gradient from north to south and the availability of an extension officer at the selected site. Further important selection criteria were accessibility of the farming community and relative uniformity of soil types within the community.
Figure 1. Geographic location of all experimental sites in central and northern Mozambique.
Table 1. Geographic location of target communities where maize varieties were evaluated under CA.
Note: Malomwe and Pumbuto were initiated in 2008; Nhamizhinga followed in 2009; Nzewe and Lamego followed in 2010.
The communities were selected in four districts of Nhamatanda, Gondola, Báruè and Angonia, located in the three provinces, Sofala, Manica and Tete (Fig. 1). The initial communities were Malomwe in Báruè and Pumbuto in Gondola districts where the study started in the 2007/2008 cropping season. In 2008/2009, Nhamizinga in Báruè district followed. The latest sites were established in Lamego, Nhamatanda district and Nzewe, Angonia district, which all started in the 2009/2010 cropping season. The sites are characterized by a unimodal rainfall distribution (i.e., one main rainfall season), which lasts from the end of October to April in Sofala and Manica. The site in Tete normally receives effective planting rains from December to May. Rainfall in Lamego is characterized by frequent droughts and a large rainfall range (233–902 mm). Pumbuto and Nhamizhinga have a more evenly distributed seasonal rainfall, while rainfall in Malomwe and Nzewe is characterized by strong fluctuations between years (Table 1 and Fig. 2). The sites have moderately sandy textured soils ranging from loamy sands to sandy loams (Table 1). Maize is the main food crop grown in all areas, often in monoculture but sometimes intercropped with pigeonpea, common beans (Vicia faba L.) and cowpea. In some areas groundnuts (Arachis hypogaea L.) and cassava (Manihot esculenta Crantz) are important food and cash crops.
Figure 2. Rainfall distribution in four target communities of central and northern Mozambique.
Field experiments
At each community, at least six farmers were selected to host a field experiment. The selection of actual trial sites was achieved through a participatory selection process in a community awareness meeting at the onset of the trial. Initial field experiments were laid out in Malomwe and Pumbuto with three general treatments comparing a conventional farming practice with two CA practices. The management at each site was by farmers with local supervision by an extension officer and scientific oversight by agronomists and socio-economists from CIMMYT. The treatments were as follows:
-
(a) Conventional tillage treatment (control) planted with maize after land clearing and burning of crop residues in the field. In sites of Manica and Sofala the maize was planted with a hoe in 90 cm rows and 50 cm in-row spacing with two living plants per station to achieve a plant population of 44,444 plants ha−1. In Tete, the rows spacing was 75 × 25 cm2, leaving one plant per station and an overall plant population of 53,333 plants ha−1 due to the higher yield potential in the north.
-
(b) CA practice with planting basins, maize seeded in pre-prepared planting basins with dimensions of 15 × 15 × 15 cm3 in 90 cm rows (75 cm in Tete) and 50 cm between basins. The target population was the same as in the conventional control. All crop residues from land clearing were left on the soil surface as mulch as well as after each consecutive harvest. The rate of mulch was aimed at 2.5–3 t ha−1 of crop residues. Where groundcover was not sufficient, locally found hyparrhenia grass (Hyparrhenia ssp.) was used to increase the overall groundcover to at least 30% cover.
-
(c) CA practice with a jab planter or pointed stick, maize planted directly without previous land preparation in 90 cm rows and 25 cm in-row spacing and leaving one plant per station (target population 44,444 plants ha−1, 53,333 ha−1 in Tete); all residues were kept on the soil surface as in the basin treatment.
In the 2009/2010 cropping season, a rotation component was introduced in Nhamizhinga, Malomwe and Lamego and all main treatments were split in half—one side was planted with maize, the other was planted with cowpeas. Maize plots were further subdivided into five sub-treatments consisting of a traditional open pollinated maize variety (OPV) commonly labeled as ‘Matuba’ and four improved open pollinated maize varieties (ZM309, ZM401, ZM523 and ZM625). The sites in Pumputo and Nzewe followed the same planting pattern in the 2011/2012 cropping season. Although there is no uniform Matuba variety in Mozambique as OPVs are developed from populations, this variety was the best available choice of a traditional control variety. We tried to minimize the experimental error by purchasing Matuba centrally and distributing it to the different on-farm locations. All ZM varieties were new products from CIMMYT's breeding pipeline which are selected under drought stress. ZM309 and ZM401 are short season, early maturing varieties (110–120 days to maturity), while ZM523 and ZM625 are medium- to long-season varieties (120–140 days to maturity).
Maize and cowpea was planted in full rotation in Lamego, Pumbuto, Nhamizhinga and Malomwe. In Nzewe, farmers rejected cowpea but opted for common bean (Phaseolus vulgaris L.) as their rotational crop.
Maize was fertilized with 58N:24P2O5:12K2O applied as basal dressing at planting and as top-dressing at 4 weeks after planting. The same amount of fertilizer was applied to all treatments. Weed control on all CA plots was achieved with an initial application of glyphosate (glyphosate [N-(phosphonomethyl) glycine] at a rate of 3 liters ha−1) and manual hand hoe weeding. In conventional systems, farmers used the hand hoe only for weed control. Legume treatments only received a basal dressing of 12N:24P2O5:12K2O at planting and no further mineral fertilizer application.
Pest control especially on the cowpea was achieved through regular (bi-weekly) spray of Dimethoate (O,O-dimethyl S-[2-(methylamino)-2-oxoethyl] dithiophosphate) as they were most affected by control aphids (Aphis ssp.) and elegant grasshoppers (Zonocerus elegans Thunberg).
Harvest procedures
Maize and legume yield was estimated from 10 sub-samples (9 m2 each) per main treatment but separated by sub-treatment to get individual yields of each variety. Harvest was done at physiological maturity and the fresh produce (cobs or pods) and biomass weighed in the field. Sub-samples of grain and biomass were air dried, shelled and a grain moisture taken. The grain yield data were expressed in kg ha−1 at 12.5% moisture content.
Seed survey
A seed survey was conducted in 2011 with farmers who live within the communities, are host farmers of trials or have geographic access to it. The survey was done to identify the importance of specific traits in maize varieties and to find out how improved maize varieties were perceived by farmers around the demonstration plots. A total of 145 farmers were interviewed in this survey. Only 48 farmers had distinct knowledge about the improved varieties (ZM309–ZM625), while others grew Matuba or different maize varieties. Primary data were collected from farmers using a structured questionnaire designed to capture maize varieties grown in the study area, the traits in maize varieties that farmers perceive as important or influencing their decision to adopt a given variety. The survey was done at a convenient time when maize was close to maturity and most of the traits could be directly evaluated (husk cover, ear size, ear aspect, kernel lines, kernel size color, lodging amongst others). Some maize crops of each variety were used for a pounding test and for roasting to find out how the poundability and taste would be.
Calculations and statistical analyses
Yield data were tested for normality and homogeneity of variance using the Kolmogorov-Smirnov test. The generalized linear model (GLM) in SAS 9.2 (TS2MO) of the SAS System for Windows © 2002–2008 was used to test the individual and interactive effects of tillage treatment (i.e., tillage), site, season, crop variety, fertilizer application on crop yields. When the Fisher-test was significant, a Least Significant Difference test (P ≤ 0.05) was used to separate the means.
Yield benefit of CA was calculated as the grain yield differences between CA and the conventional practice (CP). Relative yield of CA and CP were also plotted and the advantage of each treatment was evaluated through construction of a 1:1 line. When CA yields were larger than corresponding CP yields, the data point were above the 1:1 line and vice versa.
The Generalized Linear Mixed Models (GLMM) used for the two crops were:
-
(a) Maize model Y ijkm = α + βSS i + γTR j + δVR k + εSN m + ζ(SS i .TR j ) + η(SS i..VR k ) + θ(TR j .VR k ) + λ (SS i.SN m ) + ξ (TR j .SN m ) + π (VR k .SN m ) + ρ(SS i .TR j .VR k ) + ς(SS i .TR j .SN m ) + σ(SS i .VR k .SN m ) + τ(TR j .VR k .SN m ) + υ(SS i .TR j .VR k .SN m ) + R
-
(b) Cowpea model Y ijm = α + βSS i + γTR j + δVR m + ζ(SS i.TR j ) + η(SS i..VR m ) + θ(TR j .VR m ) + ρ(SS i .TR j .VR m ) + R
where SS i represents the ith site, TR j represents the jth tillage treatment, VR k is the kth variety and SNm represents the mth season, R is the residual and β, γ, δ, ε, ζ, η, θ, λ, ν, ξ, π, ρ, ς, σ and τ represent the fixed and random effects values.
Results
Maize grain yield
Average maize grain yields on both CA treatments (basin planting and direct seeding) and the conventional control treatment varied at different sites and years (Figs 3a–e). At Lamego (3a), the driest site, there was a complete crop failure in the first season due to drought but from year two onwards significantly greater maize yields were recorded on both CA treatments from 2011 to 2014. At Malomwe (3b), yield benefits were more variable but significant differences between CA and conventional treatments were recorded in four out of seven cropping seasons with the direct seeding treatment outyielding the conventional control in in 2009, 2011, 2013 and 2014. At Nhamizhinga (3c) the direct seeding treatment outyielded the control in four out of six seasons, whereas the basin treatment in two out of six seasons only. At Pumbuto (3d), significant difference between CA and control treatments were recorded after the second, sixth and seventh cropping seasons. At Nzewe (3e), significantly higher yields between the direct seeded treatment and the control plot were recorded in the fourth and fifth cropping seasons under treatment 2013 and 2014 where direct seeding outyielded the control. The basin treatment did not perform well in this area. Average yield increases between the basin treatments and the conventional control were 504 kg ha−1 and 800 kg ha−1 between direct seeding and the conventional control.
Figure 3. (a–e) Average maize grain yield in five target communities (Lamego (a); Malomwe (b); Nhamizhinga (c); Pumbuto (d); Nzewe (e)) in two conservation agriculture and one conventional crop management systems, 2008–2013. Error bars show the standard error of difference (SED) in a particular year; means followed by the same letter above the bar chart are not significantly different at (P ≤ 0.05) probability level.
Direct yield comparison between the conventional control treatment and the two CA treatments showed that most of the data points were above the 1:1 line (Fig. 6a). In about 83% of cases, there was a positive yield benefit for basin planting and in 89% of all cases, it was positive for direct seeding. In some few cases, CA treatments exceeded the 1:2 line implying that at this particular site and year the yield on CA treatments was twice as much as the conventional control treatment.
The specific performance of improved maize varieties by tillage treatment was analyzed from 2010 to 2013 (Fig. 5). The results showed that the improved varieties ZM523 and ZM625 had greater yields than the traditional variety Matuba under conventional agriculture. Under basin planting, all ZM varieties outyielded Matuba. The same trend was recorded under direct seeding (Fig. 5).
Results from the GLMM showed that cropping season, site, tillage management and crop variety had a strong significant effect (P < 0.001) on maize grain yield (Table 2). Site characteristics had the strongest effect on crop yield (F = 223.4), followed by season (F = 146.3) and treatment. The interactions (site×variety) and (season×variety) were highly significant, suggesting that crop varieties can be targeted to both sites and cropping seasons (wet versus dry) for improved crop productivity. The interaction site×tillage treatment was also significant suggesting that tillage was important in some sites and not in others. The interactions between season and tillage treatment was significant (P < 0.01) but the factor was less strong (F = 3.15) than the other factors. There was a strong significant effect in the interaction of season×variety indicating that the quality of the season had a strong effect on the variety performance. Also the interaction of season×site×variety was significant which implies that the quality of season and the difference in sites influences the variety performance at each site. However, there were no significant interactions between, treatment×variety; site×treatment×variety or season×treatment×variety (Table 2). The interaction between all factors (variety×tillage×treatment×season) was also not significant.
Table 2. Output of the GLMM procedures for explaining variability in maize grain yields due to tillage, fertilizer, season, site and crop variety under farmer conditions in central Mozambique.
Note: Trt, tillage treatment; Var, variety.
Legume grain yield
Cowpea grain yield was significantly higher at Lamego (Fig. 4a) on a direct seeded treatment in 2011 and between both CA treatments and the control in 2013. In 2014, the basin treatment outyielded the control only. Season 2010/2011 was heavily affected by drought and no yield was obtained in this season. In Malomwe (4b), the conventional treatment outyielded CA treatments in the first cropping season 2009/2010 but from there onwards CA had greater yields with the exception of 2012/2013. At Nhamizhinga (4c), greater yields were recorded between the basin treatment and the conventional control in 2010, 2011 and 2014 and the direct seeded treatment and the control in 2010 and 2014. At Pumbuto (4d), the direct seeding treatment outyielded the conventional tillage treatment in 2012 and 2013. At Nzewe (4e), the yield of common beans was significantly higher on direct seeding than the control in the first recorded cropping season (2011/2012) only.
Figure 4. (a-e) Average cowpea and beans grain yield in five target communities (Lamego (a); Malomwe (b); Nhamizhinga (c); Pumbuto (d); Nzewe (e)) in two conservation agriculture and one conventional crop management systems, 2010–2013. Error bars show the standard error of difference (SED) in a particular year; means followed by the same letter above the bar chart are not significantly different at probability level P ≤ 0.05.
Figure 5. Overall varietal performance in two CA and one conventional agriculture cropping system in Mozambique across sites and years, 2010–2014. The error bar shows the standard error of difference (SED) of varieties across cropping systems; means followed by the same letter above the bar chart are not significantly different at (P ≤ 0.05) probability level.
Figure 6. Yield comparison between different management strategies on maize varieties (a) and legume (b) crops in five target communities of Mozambique 2008–2013. Each dot represents a mean CA yield of farmer replicates in a target community in a particular year plotted against a conventional system. Dots above the 1:1 represent a benefit toward CA, dots below the 1:1 line favors the conventional system in the comparison.
Yield increases on legume crops averaged across sites and seasons were 156 kg ha−1 on the basin planted treatment and 175 kg ha−1 on the direct seeded treatment as compared with the conventional control. CA treatment response was positive with most comparisons displaying a positive trend toward CA (Fig. 6b) in the direct comparison of CA versus conventional treatments.
The GLMM of legumes showed significant effects (P < 0.001) of site, season and tillage on legume grain yield (Table 3). The site had the strongest factor on legume performance (F = 59.3) followed by season (F = 39.0). The interactions between treat×season, treat×site and treat×site×season was insignificant suggesting that the cowpea grain yield in each tillage treatment was not dependent on the site or season. However, the interaction of site×season was highly significant suggesting that some sites are more pronounced to adverse seasons than others.
Table 3. Output of the GLMM procedures for explaining variability in legume grain yields due to tillage, season and site under farmer conditions in central Mozambique.
Note: Trt, tillage treatment.
Maize traits considered as important by farmers
The targeted seed survey involving 145 farmers in target communities showed that farmers valued different traits differently (Table 4). Farmers rated good stable yield as most important (94%) followed by weevil resistance (81%) and resistance to specific diseases (e.g., ear rot 77%). Early maturity was stated as a factor very important to farmers in that particular area (73%). Not as important to farmers was the aspect (phenotypic appearance) of the cob (35%) and the color of the grain (39%).
Table 4. Maize traits considered important in maize varieties (in percent).
Farmers' ratings of varieties according to traits
Farmers were asked to evaluate the maize varieties that were on display on the demonstration plots (Table 5). The traditional variety, Matuba was generally liked because it had good husk cover (65%), low incidence of ear rot (62%), low rate of lodging (60%) and lack of weevils at harvest (59%). Traits that were not rated highly were white color (33%), large kernel size (28%), good yields (20%) and early maturity (15%).
Table 5. Farmer ratings of maize varieties (in percent) according to different traits.
For ZM309 the following maize traits were rated highly: early maturity (85%), no ear rot (67%), many kernel lines (63%) and 61% on diseases resistance. The worst observed traits in ZM309 were good yield (13%), no lodging (13%), no weevils at harvest (11%) and ear aspect cited by 8.7% of the farmers.
ZM401 variety traits were rated differently. Farmers liked its taste (53%), weevil resistance and early maturity (both 52%) and many kernel lines (51%). Less important traits to farmers were no ear rot (17%), early maturity (15.2%), no weevils at harvest (11%) and husk cover cited by 11% of the farmers.
For ZM523, the variety traits liked by farmers were good yield (63%), good taste (62%), large kernel size (61%) and nice ear aspect (57%). The worst observed traits in ZM523 were early maturity (too much time needed to mature) and no ear rot with 15% response and no lodging (11%).
Good traits on ZM625 were large ear size (76%) large kernel size (74%), good yields and good taste (63%) and nice ear aspect (54%). The worst observed traits in ZM625 were no weevils at harvest (35%), followed by ear rot (33%), good husk cover (28%) and no lodging (28%). As ZM625 is a long season variety, it was the characteristic of early maturity that was not rated here.
Discussion
Agronomic performance
In the majority of comparisons in the different years, CA treatments outperformed the conventional control treatment. However, due to variability between farmer replicates at each site, these were not always significant—only in 17 out of 30 maize yield comparisons and 11 out of 20 legume yield comparisons was a significant yield difference established. This result confirms previous findings from the region (Ngwira et al., Reference Ngwira, Aune and Mkwinda2012; Thierfelder et al., Reference Thierfelder, Chisui, Gama, Cheesman, Jere, Bunderson, Eash, Ngwira and Rusinamhodzi2013a , Reference Thierfelder, Mwila and Rusinamhodzi b ; Thierfelder et al., Reference Thierfelder, Matemba-Mutasa and Rusinamhodzi2015) and is in contrast with other studies from Pittelkow et al. (Reference Pittelkow, Liang, Linquist, Van Groenigen, Lee, Lundy, van Gestel, Six, Venterea and van Kessel2015) and Corbeels et al. (Reference Corbeels, Sakyi, Kühne and Whitbread2014), who looked at either incomplete CA systems (Pittelkow et al., Reference Pittelkow, Liang, Linquist, Van Groenigen, Lee, Lundy, van Gestel, Six, Venterea and van Kessel2015) or had a more variable dataset. Giller et al. (Reference Giller, Witter, Corbeels and Tittonell2009) stated that significant maize yields gains in CA cropping systems can only be reaped in the longer term which was also confirmed by Nyamangara et al. (Reference Nyamangara, Masvaya, Tirivavi and Nyengerai2013) on sandy soils in Zimbabwe. It was observed that yield differences between the conventional treatment with residue removal and significant soil disturbance and different CA treatments at some locations were evident in a very short time, which was valid for both cereals and legumes. However, a clear yield trend with increasing years of practice could not be established. This may be attributed to the relative short experimental period at each site. As significant improvement in soil quality with CA is rather unlikely to occur in a short time frame (Nyamangara et al., Reference Nyamangara, Masvaya, Tirivavi and Nyengerai2013, Reference Nyamangara, Marondedze, Masvaya, Mawodza, Nyawasha, Nyengerai, Tirivavi, Nyamugafata and Wuta2014), the main benefits from no-tillage and residue retention will have likely come from increased infiltration and better water-use-efficiency, which was previously observed by Rockström et al. (Reference Rockström, Kaumbutho, Mwalley, Nzabi, Temesgen, Mawenya, Barron, Mutua and Damgaard-Larsen2009) and Thierfelder and Wall (Reference Thierfelder and Wall2009) in Zambia as well as Thierfelder et al. (Reference Thierfelder, Chisui, Gama, Cheesman, Jere, Bunderson, Eash, Ngwira and Rusinamhodzi2013a ) and Ngwira et al. (Reference Ngwira, Thierfelder and Lambert2013) in Malawi.
Direct seeded treatments outyielded the basin treatment only occasionally (in Nzewe in 2013 and 2014 and in Pumbuto in 2014 on maize) indicating that both CA treatments were, in general, suitable systems in the drier areas, but direct seeding was more suitable in the high rainfall areas of northern Mozambique. In high rainfall areas, the basins, which are also considered as water harvesting systems, tend to accumulate too much water turning a potential benefit into a disadvantage. This highlights the need to adapt the right CA systems to different agro-ecologies and rainfall regimes.
Other studies do not show that legumes respond as quickly as maize to soil quality improvements as was previously stated in studies from Zimbabwe and Mozambique (Mupangwa et al., Reference Mupangwa, Twomlow and Walker2012; Rusinamhodzi et al., Reference Rusinamhodzi, Corbeels, Nyamangara and Giller2012; Thierfelder et al., Reference Thierfelder, Cheesman and Rusinamhodzi2012). This was not confirmed by our data. Maize comparisons had 83% and 89% positive responses when CA treatments were directly compared on-site with the respective conventional control treatment. On legumes, the yield benefit was 90% on both basins and direct seeding.
However, the decision by farmers to invest in CA in Mozambique and its limitations were also discussed previously (Rusinamhodzi et al., Reference Rusinamhodzi, Corbeels, Nyamangara and Giller2012). More awareness, training and knowledge exchange is needed to make farmers aware about the benefits of CA in years to come and to facilitate higher adoption rates. In large areas of Mozambique, shifting cultivation as well as slash and burn agriculture (Chitemene system) is still common. An investment in long-term soil fertility improvement such as CA is difficult under such circumstances because farmers will always change locations (if they can) once the soil fertility is depleted. Nevertheless, increased population pressure, future negative effects of climate variability expected for southern Africa and Mozambique in particular (Lobell et al., Reference Lobell, Burke, Tebaldi, Mastrandrea, Falcon and Naylor2008; Cairns et al., Reference Cairns, Crossa, Zaidi, Grudloyma, Sanchez, Araus, Thaitad, Makumbi, Magorokosho and Bänziger2013a ; Cairns et al., Reference Cairns, Hellin, Sonder, Araus, MacRobert, Thierfelder and Prasanna2013b ) and increased pressure to produce food from available land resources will force farmers to move away from traditional shifting cultivation to more intensive and permanent farming systems. CA could be one way of sustainable intensification in this environment.
Variety performance
Regardless of site and season, improved drought-tolerant varieties outyielded Matuba between 28 and 43% in a conventional system, between 26 and 46% in a basin planting system and between 26 and 43% in a direct seeding system. This confirms that there is a huge benefit for farmers to switch to new drought-tolerant germplasm (Bänziger et al., Reference Bänziger, Setimela, Hodson and Vivek2006; Cairns et al., Reference Cairns, Crossa, Zaidi, Grudloyma, Sanchez, Araus, Thaitad, Makumbi, Magorokosho and Bänziger2013a ) regardless of the way it is planted. However, planting the highest yielding variety in the dataset (i.e., ZM625) under direct seeding gave a 91% (2268 kg ha−1) yield benefit as compared with planting the traditional variety in the conventional systems.
The results from the GLMM indicate that variety was strongly influenced by season and site and not by tillage treatment as previously suggested by Gwenzi et al. (Reference Gwenzi, Taru, Mutema, Gotosa and Mushiri2008) for Zimbabwe. A direct interaction between tillage treatment and variety performance (e.g., treat×var, season×treat×var and season×site×treat×var) could not be established indicating that increases in maize grain yield due to treatment occurred similarly on different maize varieties. Breeding efforts to adapt varieties to site and season characteristics therefore, seem to be the most effective way to achieve breeding progress instead of selection for different tillage treatments.
Improved varieties are not a well-known part of the cropping systems in Mozambique. Most of Mozambique's maize growing area (90–99%) is sown to recycled or traditional maize varieties (Cavane and Donovan, Reference Cavane and Donovan2011; Kassie et al., Reference Kassie, Erenstein, Mwangi, MacRobert, Setimela and Shiferaw2013). The adoption rate in Mozambique of improved maize varieties is still very low (Langyintuo et al., Reference Langyintuo, Diallo, MacRobert, Dixon and Banziger2008); although there are increased efforts by the United States Agency for International Development (USAID) and the Alliance for a Green Revolution (AGRA) in Africa to increase access to and improved use of improved varieties. The lack of use by smallholder farmers may be explained by the lack of knowledge about their potential and the lack of access to a secure market that would help to facilitate selling of any additional surplus expected with improved varieties. Lack of knowledge and poor adoption of improved maize varieties by farmers may be due to the fact that there are very few seed companies in Mozambique and they mainly do not promote improved varieties (Kassie et al., Reference Kassie, Erenstein, Mwangi, MacRobert, Setimela and Shiferaw2013). Furthermore, the weak public extension services that are understaffed and underfunded cannot support variety promotion independently.
Farmer perception about improved varieties
This research added an important component to agronomic research results—the farmers' perception about varieties. Farmers considered it very important that the early season variety ZM309 was also resistant to weevils and common diseases, had little lodging and good husk cover, which they rated as higher than greater yields on other longer season varieties. The popularity of ZM309 is confirmed by farmers in other southern African countries, where it has also been released (DTMA, 2013). The longer season variety ZM625 was indeed acknowledged by farmers for being high yielding, having a large ear and kernel size, but it was not preferred by farmers due to the longer time needed for maturity. In an area of climate risk expressed by highly variable rainfall patterns (e.g., too little rainfall, too much rainfall, delayed rainfall and mid-season dry spells), it makes sense for farmers to aim for short season varieties, as was observed in target communities of Sofala and Manica. The perceived risk of crop failure makes farmers realize that an early season variety will probably not give the highest yield, but will provide at least some stable and reliable yield in most cropping seasons. Farmers also indicated that an early maturing variety will ‘give food on the table’ in the critical hunger months of February and March, when grain reserves from the previous cropping season are running low. The short maturity opens up the possibility of planting a legume (e.g., cowpea) into the standing maize crop (as a relay crop) or after the maize has matured (Dakora et al., Reference Dakora, Aboyinga, Mahama and Apaseku1987; Rao and Mathuva, Reference Rao and Mathuva2000; Tarawali et al., Reference Tarawali, Singh, Gupta, Tabo, Harris, Nokoe, Fernandez-Rivera, Bationo, Manyong, Makinde, Odion, Fatokun, Tartawali, Singh, Kormawa and Tamo2002). The use of short season varieties and a mixed use of short season and longer season varieties, as are often applied by farmers, need to be further explored and understood.
Conclusion
This study in central and northern Mozambique focused on the performance of improved DT maize varieties under different sites and crop management practices and included an assessment of the perception of farmers about improved maize varieties. The results show that CA systems outperformed the conventional tillage treatment in the majority of seasons; although the difference was not always significant due to variability between farmer replicates. This was valid for both maize and legumes in target communities of Mozambique. Average maize yield gains within one site and year ranged between 504 and 800 kg ha−1 on basins and direct seeded maize treatments, respectively, as compared with the control treatment. On legumes, the average yield gain was 156–175 kg ha−1 on basins and direct seeded treatments, respectively. The performance of improved maize varieties was on average greater than the traditional local variety Matuba. Maize yield was strongly influenced by the factors season, site, treatment and variety. However, a direct interaction between tillage treatment and maize variety performance (e.g., treat×var, season×treat×var and season×site×treat×var) could not be established, which indicates that the variety performance did not depend on the tillage treatment, but more on site and season characteristics. Legume yields were significantly affected by site and season, but all treatment interactions with site and season were insignificant.
Farmers lack access to improved varieties, have limited knowledge and face serious cash constraints to buy improved varieties. This has led to the widespread use of old recycled varieties and traditional landraces that are locally adapted to diseases and pests, but lack the capacity to perform under drought and heat stress.
Farmers’ rationale suggests that they prefer not only grain yields as the main deciding factor for growing an improved variety, but also value other important traits. Early maturity and resistance to pests were considered important traits to reduce the risk of crop failure as well as pre- and post-harvest losses. This often led to the preferred selection of the short season variety (ZM309), instead of longer season varieties.
The increased need to grow more food for smallholder farmers in Mozambique in light of future threats of projected climate variability requires more climate-resilient and permanent agriculture systems. Growing improved varieties under CA is one avenue to sustainable intensification which can provide multiple yield benefits for farmers in Mozambique.
Acknowledgments
We thank the USAID under the Platform for Agriculture Research and Innovation (PARTI) and SANREM CRSP Projects and the International Fund for Agriculture Development (IFAD) for having supported this work financially from 2008 to 2014. The work is embedded into the MAIZE and CCAFS CGIAR research programs. Their financial and logistical support is greatly acknowledged. Special thanks go to Stephanie Cheesman, Ivan Cuvaca, Antonio Rocha and Alberto Vura and numerous extension officers from Sofala, Manica and Tete, who assisted in managing trials and collecting field data.
