Introduction
Declining soil and crop productivity are widespread in smallholder farming systems as resource constrained farmers face increasing prices of agricultural inputs (Jayne et al., Reference Jayne, Mather and Mghenyi2010), limited quantities of manure because of dwindling livestock numbers (Homann Kee-Tui et al., Reference Homann Kee-Tui, Bandason, Maute, Nkomboni, Mpofu, Tanganyika, van Rooyen, Gondwe, Dias, Ncube, Moyo, Hendricks and Nisrane2013), highly variable seasonal rainfall patterns (Tadross et al., Reference Tadross, Hewitson and Usman2005; Mupangwa et al., Reference Mupangwa, Walker and Twomlow2011) and limited production of soil fertility enhancing legumes largely due to seed shortages (Ncube, Reference Ncube2007). Continuous conventional tillage with limited cereal/legume crop associations and use of organic soil fertility amendments has contributed substantially toward declining soil productivity in Southern Africa (Twomlow et al., Reference Twomlow, Steyn, du Preez, Petersen, Unger and Payne2006; Thierfelder and Wall, Reference Thierfelder and Wall2012). Soil erosion is widespread on smallholder farms and nitrogen (N), phosphorus (P) and potassium (K) balances are negative in smallholder systems (Smaling et al., Reference Smaling, Nandwa, Janssen, Buresh, Sanchez and Calhoun1997). Increased incidences of parasitic weeds such as Striga are now common on smallholder farms (Kagot et al., Reference Kagot, Okoth, Kanampiu, Okoth and Mageto2014). Household food security is further threatened by use of inappropriate agronomic practices such as poor weed control, delayed planting and use of crop varieties that are not adapted to agro-ecoregional conditions (Cairns et al., Reference Cairns, Hellin, Sonder, Araus, MacRobert, Thierfelder and Prasanna2013; Wall et al., Reference Wall, Thierfelder, Ngwira, Govaerts, Nyagumbo, Baudron, Jat and Graziano de Silva2013). It is imperative therefore that smallholder farmers adjust their way of farming by using more sustainable and resilient cropping systems.
Conservation agriculture (CA), a farming system that aims at improving soil and crop productivity, has the potential of increasing and stabilizing yields of major cereal and legume crops, and improving profitability of smallholder cropping systems under varying agro-ecoregional conditions (Ngwira et al., Reference Ngwira, Thierfelder and Lambert2012; Thierfelder et al., Reference Thierfelder, Chisui, Gama, Cheesman, Jere, Bunderson, Ngwira, Eash and Rusinamhodzi2013). CA focuses on minimum soil disturbance, use of soil cover through mulching and crop associations in the farming system (FAO, 2002). The three CA principles, when implemented together, result in increased soil carbon and fertility, improved infiltration of rainwater, soil water conservation and higher crop productivity compared with conventional agriculture (Hobbs et al., Reference Hobbs, Sayre and Gupta2008; Govaerts et al., Reference Govaerts, Sayre, Goudeseune, De Corte, Lichter, Dendooven and Deckers2009; Kassam et al., Reference Kassam, Friedrich, Shaxson and Pretty2009; Thierfelder and Wall, Reference Thierfelder and Wall2009; Mupangwa et al., Reference Mupangwa, Twomlow and Walker2013). The CA systems introduced for smallholders allow timely planting at the onset of the cropping season due to land preparation in the off-season (e.g., planting basins) (Twomlow et al., Reference Twomlow, Urolov, Jenrich and Oldrieve2008) or labor saving animal traction (AT) direct seeding techniques (Johansen et al., Reference Johansen, Haque, Bell, Thierfelder and Esdaile2012), thereby making effective use of soil water during the crop establishment stage. Crops can withstand intra-seasonal dry spells better because of improved soil water retention when all CA principles are implemented (Govaerts et al., Reference Govaerts, Sayre, Goudeseune, De Corte, Lichter, Dendooven and Deckers2009; Mupangwa et al., Reference Mupangwa, Twomlow, Walker and Hove2007). Economic returns and profitability of cropping enterprises have been found to be greater under CA systems compared with conventional agriculture under smallholder conditions (Ngwira et al., Reference Ngwira, Thierfelder and Lambert2012; Ram et al., Reference Ram, Singh, Saini, Kler, Timsina and Humphreys2012). Technology adoption by farmers is driven by profitability of the new practice in the short term, reduction in production risk in the farming system, and an intervention that is compatible with the farming objectives and skills of the farmer (Defoer et al., Reference Defoer, De Groote, Hilhorst, Kante and Budelman1998; Ojiem et al., Reference Ojiem, de Ridder, Vanlauwe and Giller2006).
Planting basins are shallow hand-hoe dug pits roughly 15 cm long, 15 cm wide and 10–20 cm deep that are maintained each season in the same place (Twomlow et al., Reference Twomlow, Urolov, Jenrich and Oldrieve2008). There are variations in the depth of basins promoted in Zimbabwe depending on the organizations that provided technical support to the non-governmental organizations (NGOs). The depths vary from shallow (<10 cm) meant to minimize soil disturbance, up to 20 cm deep meant to break the plow pan that developed during many years of conventional tillage (Nyamangara et al., Reference Nyamangara, Nyengerai, Masvaya, Tirivavi, Mashingaidze, Mupangwa, Dimes, Hove and Twomlow2014). Rip-line planting is done with a Magoye ripper attached to a beam of a conventional plow. The ripper tine opens planting furrows for most crops grown in Southern Africa and a seed is placed into the furrows by hand. An AT direct seeder (DS) opens the planting furrow, places basal mineral fertilizer and seed into the furrow, and covers the furrow in one pass (Thierfelder et al., Reference Thierfelder, Rusinamhodzi, Ngwira, Mupangwa, Nyagumbo, Kassie and Cairns2014).
These CA systems targeting farming households of different resource endowment have been introduced in Southern Africa over the past decade (Mazvimavi and Twomlow, Reference Mazvimavi and Twomlow2009; Wall et al., Reference Wall, Thierfelder, Ngwira, Govaerts, Nyagumbo, Baudron, Jat and Graziano de Silva2013; Thierfelder et al., Reference Thierfelder, Rusinamhodzi, Ngwira, Mupangwa, Nyagumbo, Kassie and Cairns2014). In Zimbabwe, as in other parts of Southern Africa, hand dug planting basins were widely promoted for households with limited access to draft animal power as a food security intervention (Twomlow et al., Reference Twomlow, Urolov, Jenrich and Oldrieve2008; Mazvimavi and Twomlow, Reference Mazvimavi and Twomlow2009). Farming households with access to draft animals were initially excluded in these programs. When planting basins are used, producing a maize crop on 0.25 ha is possible given the available family labor in most rural households of Zimbabwe (Ndlovu et al., Reference Ndlovu, Mazvimavi, An and Murendo2014). However, when the area under planting basins is increased, more labor is required for land preparation and weeding (Mazvimavi and Twomlow, Reference Mazvimavi and Twomlow2009; Ndlovu et al., Reference Ndlovu, Mazvimavi, An and Murendo2014), making manual CA systems unattractive to farming families especially the ones that already own draught animals and farm on landholding of 1–5 ha.
AT tine ripping and direct seeding have shown great potential for sustainably increasing cereal and legume productivity in smallholder cropping systems compared with the conventional practices (Nyamangara et al., Reference Nyamangara, Nyengerai, Masvaya, Tirivavi, Mashingaidze, Mupangwa, Dimes, Hove and Twomlow2014; Wall et al., Reference Wall, Thierfelder, Ngwira, Govaerts, Nyagumbo, Baudron, Jat and Graziano de Silva2013; Thierfelder et al., Reference Thierfelder, Rusinamhodzi, Ngwira, Mupangwa, Nyagumbo, Kassie and Cairns2014). The AT CA systems have widened the choice of sustainable crop production practices for smallholders in different agro-ecoregional conditions of Southern Africa. However, information on cereal–legume productivity and the associated socio-economic benefits derived from planting basins and AT CA systems compared with conventional practices is still limited for the different agro-ecoregions (AEs) of Zimbabwe.
Although CA technologies have been widely promoted in Southern Africa, adoption rates have remained low and often partial, and the benefits for smallholders remain highly debated (Baudron et al., Reference Baudron, Andersson, Corbeels and Giller2012; Andersson and D'Souza, Reference Andersson and D'Souza2013; Arslan et al., Reference Arslan, McCarthy, Lipper, Asfaw and Cattaneo2013). The suitability of CA systems for mixed crop/livestock systems have been questioned given the resource endowment and farming objectives to smallholders (Giller et al., Reference Giller, Witter, Corbeels and Tittonell2009; Palm et al., Reference Palm, Blanco-Canqui, DeClerck, Gatere and Grace2014; Pittelkow et al., Reference Pittelkow, Liang, Linquist, van Groenigen, Lee, Lundy, van Gestel, Six, Venterea and van Kessel2014). In countries such as Zambia and Zimbabwe, partial adoption of CA principles is common because of challenges associated with mulch and legume seed availability (Baudron et al., Reference Baudron, Mwanza, Triomphe and Bwalya2007; Mazvimavi and Twomlow, Reference Mazvimavi and Twomlow2009; Umar et al., Reference Umar, Aune, Johnsen and Lungu2011). Partial adoption may occur because a new technology may not exhibit overwhelming advantage over the traditional system (Dixon, Reference Dixon1980). In Southern Africa dissemination of CA technologies has further been hampered by the heterogeneity of smallholder farming households. Thus, a number of studies have argued that CA is appropriate for certain households and AEs, further emphasizing that it cannot be a ‘one-size-fits-all’ development strategy for smallholder crop production (Giller et al., Reference Giller, Witter, Corbeels and Tittonell2009; Umar et al., Reference Umar, Aune, Johnsen and Lungu2011; Nyanga, Reference Nyanga2012; Baudron et al., Reference Baudron, Jaleta, Okitoi and Tegegn2013).
Smallholder farmers' adoption of an agricultural innovation depends on the riskiness of the technology (Guto et al., Reference Guto, Pypers, Vanlauwe, de Ridder and Giller2011). Smallholders in developing countries are often hesitant to adopt agricultural innovations even if economic analysis shows clear high returns to investment (Mazvimavi and Twomlow, Reference Mazvimavi and Twomlow2009; Guto et al., Reference Guto, Pypers, Vanlauwe, de Ridder and Giller2011). These farmers are more concerned about the dispersion of these net benefits and downside risk related to trade-offs between the short-term costs of investing in the new technology compared with expected long-term economic benefits in the future (Ngwira et al., Reference Ngwira, Thierfelder, Eash and Lambert2013). To enhance adoption of new agricultural technologies farmers require comprehensive information about the economic impacts and the skewness of the downside risk. Hence stochastic dominance analysis, a non-parametric tool, can be used to determine economically optimal cropping systems for the heterogeneous smallholder farmers in contrasting AE niches. Stochastic dominance ranks alternatives according to their risk characteristics (Hien et al., Reference Hien, Kabare, Sansan and Lowenberg-DeBoer1997). It classifies technologies that are predominant, those that might be acceptable to risk-neutral farmers and those that could be used by risk-averse farmers (Hien et al., Reference Hien, Kabare, Sansan and Lowenberg-DeBoer1997). Stochastic dominance analysis focuses on the distribution of the mean and the variance but not the parameters themselves.
Stochastic dominance compares the cumulative distributions of the net benefits (outcomes) based on two rules. The first degree stochastic dominance (FSD) rule states that if one cumulative distribution is to the left of another cumulative distribution for all levels of outcome, the technology with the distribution to the right is dominating the technology whose distribution is to the left. This type of dominance is called ‘first degree stochastic dominance’. The second stochastic dominance rule (SSD) assumes that human beings are risk averse and they prefer to avoid lower outcomes. Graphically, a technology is dominating if the area under its cumulative probability curve is smaller at every outcome level than that of the alternative. Thus, the alternative with the smallest area under the curve at any given outcome level has the lowest probability of low value outcomes.
We hypothesized that planting basins, rip-line and AT direct seeding systems will outperform conventional plowing in the different AEs of Zimbabwe. The second hypothesis was that planting basins, rip-line and AT direct seeding systems will have a similar effect on crop productivity and the overall profitability of smallholder farming in the different AEs. The objectives of the study were therefore to determine: (a) the effect of conventional, planting basin, rip-line and direct seeding systems on maize (Zea mays L.), cowpea [Vigna unguiculata (L.) Walp] and soybean [Glycine max (L.) Merrill] yields, and (b) the economic performance of planting basins, rip-line and direct seeding systems relative to conventional plowing at field scale under smallholder conditions in three AEs of Zimbabwe.
Materials and Methods
Description of study sites
An on-farm study was conducted in 26 target communities spread across three AEs of Zimbabwe. Zimbabwe is divided into five AEs based on rainfall, soil quality and farming systems (Vincent and Thomas, Reference Vincent and Thomas1961), and the characteristics of each AE are summarized in Table 1. The 26 target communities were located in 13 districts namely Chivi, Gokwe, Insiza, Masvingo, Mzingwane and Zaka (low rainfall AE); Kariba (medium rainfall AE); Goromonzi, Kadoma, Makonde, Madziwa, Murehwa and Zvimba (high rainfall AE). Soils in high rainfall AE were granitic sands while in medium rainfall AE soil varied from sandy loams to loams. In low rainfall AE, the predominant soil texture was sand with pockets of sandy loams in Insiza and Zaka districts. In each target community, five farmers were selected by villagers to host the trials for a minimum of three cropping seasons and two target communities were selected in each district giving a total of ten farmers per district. The selected farmers in each community received training on calibration and operation of CA equipment, calibration of knapsack sprayers and application of glyphosate (480 g l−1 active ingredient), seeding maize and legumes in different CA systems, pest and disease control and socio-economic and biophysical data collection. Training sessions were facilitated by CIMMYT researchers, Extension Officers and NGO partners before the onset of each cropping season.
Table 1. General characteristics of the five agro-ecoregions (AEs) of Zimbabwe.
Source: Adapted from Vincent and Thomas (Reference Vincent and Thomas1961).
The smallholder systems in the three AEs are based on mixed crop–livestock production with maize (Z. mays L.) and sorghum [Sorghum bicolor (L.) Moench] as the major cereals. Grain legumes grown for food and income include: groundnuts (Arachis hypogaea L.), cowpea [V. unguiculata (L.) Walp] and soybean [G. max (L.) Merrill]. The major cash crops grown by smallholders are cotton (Gossypium hirsutum L.) and tobacco (Nicotiana tabacum L.). The area of each cash crop varies between seasons depending on the previous year's market prices for the different crops. Traditionally, smallholders plow their land at the onset of the rainfall season. Seeding is done by third furrow plow-planting or opening furrows after plowing, placing the seeds into the furrows and covering manually. Farmers with no access to draught animals open planting stations manually using hand hoes and seed into the basins by hand. General management of the crops include: weeding, fertilizer and manure application, and pests/disease control during crop growth. Crop residues from cereals contribute feed toward livestock well-being, while the latter is a source of draught power, manure and income for the mixed crop–livestock smallholder farming system.
Experimental treatments and design
The treatments consisted of four cropping systems: (a) ox-drawn conventional moldboard plowing; (b) planting basins; (c) ox-drawn tine ripping (rip-line seeding) and (d) ox-drawn direct seeding. An experiment was established with five farmers being the trial replicates at each target community. The trial was ongoing for three consecutive cropping seasons (2010/11–2012/13). In the first season (2010/11) maize was the only crop grown on the trial at each farmer's field. The total trial area of 50 × 40 m2 was subdivided into four equal portions (50 × 10 m2) where the main treatments were established. In 2011/12 and 2012/13 seasons the main treatments were halved (each 50 × 5 m2) and a cowpea and/or soybean rotation with maize was introduced. The description of each treatment is summarized below:
-
(1) Conventional moldboard plowing: an ox-drawn V100® conventional plow was used and plowing operation was conducted after the first effective planting rains each season. There were no other operations such as disking or harrowing after plowing. Maize was spaced at 90 × 50 cm2 with two seeds per station, 45 × 10 cm2 for cowpea and 45 × 5 cm2 soybean with one seed per station for both crops.
-
(2) Planting basins: basins measuring 15 × 15 × 15 cm3 and spaced at 90 × 50 cm2 were dug in November each year. For maize, three seeds were planted per station and later thinned to two plants per basin 2 weeks after crop emergence. For cowpea and soybean, six seeds were placed in a planting basin and these were not thinned.
-
(3) AT ripping (rip-line seeding): A Magoye ripper attached to a beam of a VS100® conventional plow was used for opening furrows in one pass to a depth of 15–20 cm.
Ripping was done at seeding after receiving effective rains and ripped furrows were spaced at 90 cm for maize and 45 cm for cowpea and soybean. In-row spacing was 25 cm for maize with one plant per station, 45 × 10 cm2 for cowpea and 45 × 5 cm2 soybean with one seed per station.
-
(4) Animal drawn direct seeding: A Brazilian made Fitarelli DS (Irmãos Fitarelli, Brazil, model #12) was used at all sites and seeding was done after receiving effective planting rains. Maize rows were spaced at 90 cm while 45 cm was used for cowpea and soybean. In-row spacing for maize was 25 cm with one plant per station, 45 × 10 cm2 for cowpea and 45 × 5 cm2 soybean with one seed per station.
In basin, rip-line and DS treatments mulch cover was applied every year at seeding. Farmers used hyparrhenia grass [Hyparrhenia filipendula (L.) Stapf] for mulching (2.5–3 t ha−1) whenever crop residues from the previous maize crop were grazed in-situ by livestock during the dry season. Crop residues from the harvested crop in the previous season were removed. All experimental plots had maize in the first season (2010/11) but each plot was sub-divided into two in order to accommodate both maize and cowpea/soybean crops in the second (2011/12) and third (2012/13) seasons of the study.
Trial management and harvesting procedures
At seeding, the four systems received an equal amount of basal compound D (8 N:14 P2O5:7 K2O) fertilizer applied at 165 kg ha−1 (13 N, 9.9 P and 9.6 K kg ha−1) for maize and cowpea each season, and 165 kg ha−1 soybean blend (6 N:27 P2O5:20 K2O) fertilizer for soybean. The soybean blend supplied (9.9 N, 19 P and 27 K kg ha−1) to each cropping system. In the conventional plowing system, basal fertilizer was placed in furrows opened after plowing. The planting basin system received 7.4 g of basal fertilizer per planting basin at seeding. At seeding basal fertilizer was dribbled by hand along the rip furrow in the rip-line system. In the direct seeding system, basal fertilizer was applied by the machine at seeding in each season. Soybean planted in the four systems was inoculated at seeding each season. In each season, maize in all treatments was top dressed at 200 kg ha−1 using ammonium nitrate (34.5 % N) and the 200 kg ha−1 was split applied at 4 (100 kg ha−1) and 7 (100 kg ha−1) weeks after crop emergence.
At seeding, weeds in the CA plots were controlled by spraying glyphosate (480 g l−1 active ingredient) at 2.5 liters ha−1 and thereafter weed control was done manually whenever weed plants were about 10 cm tall or in radius for crawling species, and the decision on when to weed was the responsibility of the host farmer. In the CP tillage system, weeds were removed by using hand hoes whenever they reached about 10 cm tall or in radius for crawling species. Aphids (Aphis craccivora L.) in cowpea were controlled using carbaryl (1-naphthyl methylcarbamate) (85% active ingredient) whenever necessary.
Harvest procedures
All crops were harvested at physiological maturity. Maize ears and above-ground biomass were collected from ten sample plots of 2 rows × 5 m (9 m2) for maize. For cowpea and soybean pods and biomass were collected from ten sample plots of 4 rows × 5 m (9 m2) in each cropping system. Cob, pod and biomass samples were weighed in the field before taking sub-samples for determining grain and biomass moisture content. Grain and stover sub-samples were air dried for 5 weeks before determining dry grain and biomass weight. Grain moisture content was measured using the mini GAC® moisture tester (DICKEY-John, USA).
Analysis of crop yield data from the four tillage systems
Maize, cowpea and soybean yield data were subjected to test for normality using the Shapiro–Wilk Normality test in STATISTIX 9 for personal computers (Statistix, 2008). Maize, cowpea and soybean yield data were then subjected to analysis of variance (ANOVA) using the randomized complete block design (RCBD) with tillage system as treatment factor and farmer as a replicate (Arslan et al., Reference Arslan, McCarthy, Lipper, Asfaw and Cattaneo2013). Crop yield data collected in each season from each AE were analyzed separately because the three cropping seasons experienced different rainfall distribution patterns. The start and end of growing seasons were also quite variable in each AE. Where the F-test was significant means were separated by least significance difference (LSD) at P < 0.05.
To generate cumulative probability distribution functions (CDFs) for maize grain yield, data were pooled over the years and sites in each AE. To generate CDFs for each cropping system in an AE, pooled maize yield data were ranked in an ascending order and a serial rank number was assigned to each value to derive a plotting position (Kipkorir et al., Reference Kipkorir, Raes, Bargerei and Mugalavai2007). Each plotting position corresponded with the frequency of a given maize yield value on the probability scale. Plots of probability against maize grain yield were developed using Sigma Plot (version 10.0). The cropping system with the least maize production risk, graphically, lies below and to the right of distribution functions of the other cropping systems tested in each AE (Tsubo et al., Reference Tsubo, Walker and Ogindo2005; Foti et al., Reference Foti, Mapiye, Mutenje, Mwale and Mlambo2008). From the CDFs maize yield less than 20 percentile was used to represent low production range while yield greater than 80 percentile was used to illustrate the high maize production levels on the smallholder farms from each AE.
A stability analysis of the maize grain yield from the four tillage systems was conducted for each AE using the Ebernart and Russell model (Ebernart and Russell, Reference Ebernart and Russell1966). Daily rainfall was recorded manually using a rain gauge located at each trial site during the three seasons of experimentation. To assess the relationship between seasonal rainfall and maize grain yield, regression analysis was conducted for each tillage system in the three AEs used in the study. Seasonal rainfall totals and maize grain yield were used as the independent and dependent variables, respectively.
Economic analysis of the four tillage systems
The data used for the economic analysis were collected during 2010/11, 2011/12 and 2012/13 cropping seasons. Gross margin analysis was used to assess the potential net benefits of growing maize, cowpea and soybean under the four tillage systems. The net benefit (returns) from the four systems was compared using the pair-wise t-test. All variable costs of the four systems were recorded by the farmer using standardized protocols with the help of the resident agriculture extension officer. All family labor resources were standardized using the adult male equivalents to minimize the quantity, quality and customs dimension following recommendations (Fathelrahman et al. Reference Fathelrahman, Ascough, Hoag, Malone, Heilman, Wiles and Kanwar2011). Labor was valued at prevailing local market prices in order to avoid distortions when farmers used family labor. The value of crop residues or other plant materials used as soil cover and the effects of crop rotation on crop yield were taken into consideration in the economic analysis. The shadow price of the crop biomass was incorporated in the economic analysis. The gross return, total variable costs (TVC) and net benefits were calculated as follows:
$${\rm G}{{\rm R}_{{ijk}}}{\rm =} \left[ {{\rm (}{{Q}_{i}} \times {{P}_{i}}{Q}{\rm ) + (}{{Y}_{i}}\;{{P}_{i}}{Y}{\rm ) +} \left( {\displaystyle{{{\rm NL1}} \over {a}} \times {{P}_{i}}{N}} \right)} \right]$$
where GR ijk is the gross revenue for farmer I, technology j in agro-ecological region k, Q i is the quantity of grain, P is the prevailing market price.
In this analysis, labor and financial capital were considered as the most limiting factors of production. The returns to labor for the different tillage systems were therefore, calculated as gross receipts less than the other material cost rather than just dividing labor by the labor cost;
$$\eqalign{&{\rm Returns}\;{\rm to}\;{\rm labor}\;({\rm USD}) \cr & = \displaystyle{{{\rm Gross}\;{\rm receipts} - ({\rm TVC} - {\rm labor\,\, costs})} \over {{\rm labor}\;{\rm costs}}}}$$
Similarly, the return to every dollar invested was calculated by dividing the gross margin by the total variable cost;
$${\rm Returns}\;{\rm to}\;{\rm TVC}\;{\rm (\% ) =} \displaystyle{{{\rm Gross\,\, margin}} \over {{\rm TVC}}} \times {\rm 100}$$
In this study, FSD and SSD were used simultaneously to classify the different tillage systems into three groups, namely, (1) the dominated strategies, (2) strategies that could be acceptable to risk-neutral farmers and (3) those that could be used by risk-averse farmers (Hien et al., Reference Hien, Kabare, Sansan and Lowenberg-DeBoer1997). The CDF of the alternative tillage system were generated by Stata 13 statistical package (http://www.stata-journal.com/software, 2013) using plot level data generated from three cropping seasons of on-farm experimentation. Risk was evaluated only in terms of the distribution of net returns across seasons.
Results
Rainfall patterns in the three AEs varied from season to season and the highest rainfall was recorded in medium rainfall AE (Table 2). Rainfall averaged 686 mm in 2010/11, 717 mm in 2011/12 and 572 mm in 2012/13 across the trial sites used in high rainfall AE. In medium rainfall AE, rainfall averaged 1061 mm in 2010/11, 612 mm in 2011/12 and 730 mm in 2012/13. Rainfall averaged 597 mm in 2010/11, 503 mm in 2011/12 and 686 mm in 2012/13 across the trial sites in low rainfall AE. The longest dry spell lasted 24 days and was experienced in medium rainfall AE during the February–March period of 2012/13 season. In medium and low rainfall AEs, the longest dry spell was 13 days and the shortest was 21 days across the trial sites used in the two AEs.
Table 2. Mean seasonal rainfall received at sites in the districts that hosted the on-farm trials in 2010/11, 2011/12 and 2012/13 cropping seasons.
Maize grain yield
Maize responses to the four tillage systems depended on the AE where trials were conducted and also varied from season to season due to variable rainfall patterns (Table 3). In high rainfall AE, maize plant populations were similar in 2010/11 and 2011/12 seasons. In medium rainfall AE, rip-line and DS systems had higher maize populations in the first and second seasons. Plant populations in low rainfall AE also varied from season to season across the four tillage systems. In high and low rainfall AEs, the four tillage systems had similar grain yield in each cropping season. However, in medium rainfall AE maize grain yield was significantly influenced by tillage system in the third season (Tables 3 and 4). In 2011/12 season the planting basins, rip-line and DS systems had 541, 703 and 1504 kg ha−1 more grain compared with the conventional system in medium rainfall AE. In 2012/13 season the planting basins, rip-line and DS systems outperformed the conventional practice by 553, 393 and 1875 kg ha−1, respectively.
Table 3. Maize plant population and grain yield from four tillage systems in different agro-ecoregions (AEs) of Zimbabwe during three seasons (2010/11–2012/13).
nd, not determined in 2012/13 season; ns, not significant.
Means with the same letter in each column under each AE are not significantly different (P < 0.05).
Table 4. ANOVA for maize grain yield responses to four tillage systems in three agro-ecoregions (AEs) of Zimbabwe in 2010/11, 2011/12 and 2012/13 cropping seasons.
1 DF = degrees of freedom; SS = sum of squares; MS = mean square.
Pooled maize yield data from high rainfall AE suggested that the conventional and DS treatments were superior compared with basins and rip-line systems across the trial sites used (Fig. 1). In medium rainfall AE, the DS treatment outperformed the conventional, planting basin and rip-line systems at all maize yield levels (Fig. 1). In low rainfall AE, the CP tillage system was superior at low production maize levels while the rip-line and DS treatments were superior for high maize production purposes (Fig. 1).
Figure 1. CDF for maize grain yield derived from experimental sites used in 2010/11, 2011/12 and 2012/13 cropping seasons in high, medium and low rainfall AEs of Zimbabwe.
Stability analysis indicated that the CP treatment was more a stable maize production system in the high rainfall AE compared with the CA practices (Table 5). The AT DS was stable in medium rainfall AE, while basins were more stable in the low rainfall AE of Zimbabwe. In the high rainfall AE, maize yield had a significant linear relationship with seasonal rainfall totals in CP (P = 0.0217), basins (P = 0.0270), rip-line (P = 0.0053) and DS (P = 0.0119) (Fig. 2). Rainfall had a significant (P < 0.001) quadratic relationship with maize grain yield achieved across the experimental sites in the medium rainfall AE (Fig. 3). In the low rainfall AE, maize yield had a significant linear relationship with rainfall in CP (P = 0.009), rip-line (P = 0.0001) and DS (P = 0.0001) systems (Fig. 4). Maize yield had no significant relationship with seasonal rainfall in the basin tillage system.
Figure 2. Relationship between maize grain yield and seasonal rainfall recorded across trial sites in high rainfall AE of Zimbabwe.
Figure 3. Relationship between maize grain yield and seasonal rainfall recorded across trial sites in medium rainfall AE of Zimbabwe.
Figure 4. Relationship between maize grain yield and seasonal rainfall recorded across trial sites in low rainfall AE of Zimbabwe.
Table 5. Ebernart and Russell stability parameters for maize grain yield produced from four tillage systems in high, medium and low rainfall agro-ecoregions (AEs) of Zimbabwe across 2010/11, 2011/12 and 2012/13 cropping seasons.
Stable tillage system has b i = 1 and S 2 d i = 0.
Cowpea and soybean yield from the three AEs
Planting basins had the lowest cowpea population except in the first cropping season in high rainfall AE (Table 6). Cowpea grain yield was not significantly influenced by tillage system in high rainfall AE during 2011/12 and 2012/13 seasons (Table 6). In low rainfall AE, the rip-line system outyielded (P = 0.002) the CP system by 428 kg ha−1 in 2012/13 season. As observed with cowpea, planting basins had the lowest soybean population in medium rainfall AE (Table 7). Soybean grain yield was significantly (P = 0.019) influenced by tillage systems with DS system yielding 544 kg ha−1 more than the CP system in 2011/12 season. In 2012/2013, the rip-line and DS treatments yielded 973 and 1341 kg ha−1 more grain than the CP system.
Table 6. Cowpea plant population and grain yield responses to different cropping systems in high and low rainfall agro-ecoregions (AEs) in 2011/12 and 2012/13 seasons.
ns, not significant.
Means with the same letter in each column under each AE are not significantly different (P < 0.05).
Table 7. Soybean plant population and grain yield responses to different tillage systems in medium rainfall AE in 2011/12 and 2012/13 seasons.
Means with the same letter in each column under each AE are not significantly different (P < 0.05).
Economic benefits of the four tillage systems in different AEs
Although the economic analysis results showed that the DS system yielded the highest net benefits (US$ 350 ha−1) relative to the CP system (US$ 258) in high rainfall AE (Table 8), it was dominated by the other three cropping systems tested in our study (Fig. 5). The stochastic dominance analysis results revealed that the DS system has higher risks relative to the conventional system.
Figure 5. CDFs of net benefits from the four tillage systems in high, medium and low rainfall AEs of Zimbabwe (2010–2013).
Table 8. Gross Margin Analysis (US$ ha−1) of four tillage systems in high rainfall AE (2010–2013).
Gross margin means sharing the same letter in a row, are not different at the 5% level of significance (t-test).
These results suggest that in this AE, risk-averse farmers would prefer the conventional system despite the higher returns from direct seeding. The economic analysis for medium rainfall AE showed that among the four treatments tested, the DS system has the most preferred risk characteristic for both risk-averse and risk-neutral farmers in this agro-ecology (Table 9 and Fig. 5). The DS consistently dominated all the other three systems using both the FSD and SSD selection criterion. In low rainfall AE, though the DS system had the highest returns per hectare (US$ 699 ha−1) (Table 10), it was dominated by the other three tillage systems (Fig. 5). Planting basin system would be the most preferred system by risk-averse smallholder farmers.
Table 9. Gross Margin Analysis (US$ ha−1) of four tillage systems in medium rainfall AE (2010–2013).
Gross margin means sharing the same letter in a row, are not different at the 5% level of significance (t-test).
Table 10. Gross Margin Analysis (US$ ha−1) of four tillage systems in low rainfall AE (2010–2013).
Gross margin means sharing the same letter in a row, are not different at the 5% level of significance (t-test).
Discussion
Seasonal rainfall patterns in different AEs
Rainfall pattern was variable in each AE during 2010/11, 2011/12 and 2012/13 cropping seasons. In high rainfall AE below average seasonal rainfall was recorded in 2011/12 season and this can be attributed to a number of 7–15 day dry spells experienced during that cropping season in some districts such as Madziwa, Makonde and Kadoma. On the other hand, normal to above normal rainfall patterns experienced in NRs medium and low rainfall AEs were due to even rainfall distribution experienced during the three cropping seasons. Associated with the below normal rainfall pattern was the high rainfall variability across trial sites located in high rainfall AE. Some districts of the high rainfall AE (e.g., Kadoma, Madziwa and Zvimba) recorded seasonal rainfall of less than 800 mm, while others (e.g., Murehwa) received more than 1000 mm in the same cropping season. High variations in site-to-site seasonal rainfall in high rainfall AE is not consistent with general observations in the sub-humid parts of Southern Africa (Nicholson, Reference Nicholson2000; Kalognomou et al., Reference Kalognomou, Lennard, Shongwe, Pinto, Favre, Kent, Hewitson, Dosio, Nikulin, Panitz and Buchner2013). Such rainfall variations at trial sites within 1–2 km radius within a target community highlight the fact that smallholder farmers ought to have a wide range of resilient and sustainable crop production options. Rainfall variability in low rainfall AE is consistent with findings from previous studies conducted in semi-arid areas of Southern Africa (Tadross et al., Reference Tadross, Hewitson and Usman2005; Mzezewa et al., Reference Mzezewa, Misi and van Rensburg2010; Mupangwa et al., Reference Mupangwa, Walker and Twomlow2011; Kalognomou et al., Reference Kalognomou, Lennard, Shongwe, Pinto, Favre, Kent, Hewitson, Dosio, Nikulin, Panitz and Buchner2013). Dry spells of varying duration experienced in the three AEs are a common phenomenon in Southern Africa and occurred during the mid-season period when soil moisture demand by crops was high (Nicholson, Reference Nicholson2000; Usman and Reason, Reference Usman and Reason2004; Kalognomou et al., Reference Kalognomou, Lennard, Shongwe, Pinto, Favre, Kent, Hewitson, Dosio, Nikulin, Panitz and Buchner2013).
Maize, cowpea and soybean yields from different AEs
Maize and cowpea responses to the four tillage systems varied across AEs and rainfall seasons, and the cropping systems gave similar maize yields in high and low rainfall AEs. With similar soil fertility and agronomic management the conventional system could perform as much as the CA systems in the short term, a trend consistent with results reported in other studies conducted in Southern Africa (Mupangwa et al., Reference Mupangwa, Twomlow and Walker2012; Thierfelder et al., Reference Thierfelder, Chisui, Gama, Cheesman, Jere, Bunderson, Ngwira, Eash and Rusinamhodzi2013). In high and low rainfall AEs, maize and cowpea yields were variable from site to site and this is consistent with the rainfall patterns experienced in the two agro-ecologies. In medium rainfall AE, maize and soybean yields in the four cropping systems were also variable in all seasons. High variability in crop yields can also be attributed to differences in trial management by the host farmers in each target community, and soil moisture differences across sites because rainfall was not evenly distributed within the study areas. Higher maize yield in medium rainfall AE compared with high and low rainfall AEs can be attributed to the fact that trial sites in medium rainfall AE were located in areas with inherently more fertile soils. Soils in high and low rainfall AEs were predominantly granitic sands of low fertility and these soils are inherently deficient in major and micro nutrients (Mapfumo and Giller, Reference Mapfumo and Giller2001; Thierfelder et al., Reference Thierfelder, Cheesman and Rusinamhodzi2012). Low cowpea and soybean populations and yields in planting basins can be attributed to the fact that these legumes grow better when planted in rows at higher densities than in planting basins.
A comparison of the superiority of the four tillage systems in the different agro-ecologies showed that the CP system can outperform the CA systems under the sub-humid conditions of Zimbabwe. This observation can be attributed to the fact that with CP practice a clean seed bed is established by the plowing operation as weeds are incorporated into the soil while in CA systems effective weed control relies on the correct application of pre-emergence herbicide (i.e., glyphosate used in our study). Effective early weed control in CA treatments varied from farmer to farmer in each target community used in our study. The inappropriate application of glyphosate at seeding in some target communities was compounded by continuous rains experienced during some cropping seasons and this made manual weed control difficult in CA systems. In the 700–1000 mm rainfall AE, our results suggest that the DS system is a more viable option for smallholder farmers targeting higher production. In medium rainfall AE, the DS system was the dominant treatment regardless of the rainfall pattern, highlighting the fact that it is an appropriate option for smallholder farmers with commercial production orientation in their farming system.
Economic benefits of the four tillage systems in different AEs
Identification of economically optimal cropping system for the heterogeneous smallholder farmers in the contrasting agro-ecological niches is necessary to enable the development of appropriate recommendation domains for smallholder farmers (Zingore et al., Reference Zingore, Murwira, Delve and Giller2007).
Although the economic analysis results showed that DS cropping system with maize and legume rotation gave the highest net benefits (US$ 350 ha−1) relative to the conventional system (US$ 258) in high rainfall AE, it was dominated by the other three tillage systems tested in our study. These results suggest that in this agro-ecology, risk-averse farmers would prefer the conventional system despite the higher returns from the DS system. Consistent with findings from other studies it can be deduced that resource constrained farmers may maintain a traditional cropping system when there is risk even when economic incentives indicate that farmer practice is not sustainable (Hien et al., Reference Hien, Kabare, Sansan and Lowenberg-DeBoer1997; Ngwira et al., Reference Ngwira, Thierfelder, Eash and Lambert2013). The economic analysis for medium rainfall AE showed that among the four treatments tested, the DS system has the most preferred risk characteristic for both risk-averse and risk-neutral farmers in this region. The DS consistently dominated the other three systems using both the FSD and SSD selection criterion. According to Tittonell (Reference Tittonell2014) farmers in this region are emerging commercial farmers who require agricultural technologies that will enable them to intensify production for profit maximization whilst reducing downside risks. Thus, the implication of these results is that for commercial-oriented smallholder farmers in a land abundant farming system, mechanized CA is the most viable option.
In low rainfall AE, although the DS system had highest returns (US$ 699 ha−1) it was dominated by the other cropping systems. Planting basin system would be the most preferred alternative by risk-averse farmers. These results concur with findings from other studies that crop yield benefits of CA systems such as planting basins may outweigh the potential disadvantages such as increased labor costs especially in semi-arid environments (FAO, 2000; Twomlow et al., Reference Twomlow, Urolov, Jenrich and Oldrieve2008). Thus these results suggest that in farming systems similar to those found in low rainfall AE where farming is predominantly for household food self-sufficiency (Tittonell, Reference Tittonell2014), planting basins have less risk and provide stable yields irrespective of the labor cost.
Conclusions and Recommendations
Four cropping systems were tested under smallholder farming conditions in three agro-ecological regions of Zimbabwe. Maize, cowpea and soybean productivity under the four cropping systems varied across agro-ecological regions, being influenced largely by seasonal rainfall patterns and inherent soil fertility of the experimental sites. In the short term, conventional practice and the CA systems gave similar maize and grain legume yields under low (450–650 mm) and high (750–1000 mm) rainfall AEs. Under moderate rainfall and better inherent soil fertility conditions, illustrated by medium rainfall (500–800 mm) AE, CA systems out-performed the conventional practice with the DS treatment being superior to the other CA systems. Results from this study also indicated that for higher maize production purposes the DS system is a preferable option compared with conventional practice, planting basins and rip-line seeding.
Economic analysis results revealed that comparing and evaluating technologies based on mean values of gross margin or net benefits is necessary but is not a sufficient condition to prescribe a technology for smallholders operating in risky environments. It is important to examine variations in the distribution of economic parameters particularly where risk influences decision making. The results showed that the DS system consistently outperformed the other systems based on gross margin analysis across all agro-ecologies although it was dominated in high (750–1000 mm) and low (450–650 mm) rainfall AEs. In instances with limiting biophysical (i.e., rainfall and soil) and socio-economic conditions the results revealed that the DS system is dominated by other cropping systems and is therefore less attractive for risk-averse farmers relative to conventional practice, planting basins and rip-line seeding options. Despite the lack of short-term maize yield benefits from the CA systems in some seasons, economic analysis results show that the high returns derived from these systems can be an incentive for some farming households in the different agro-ecologies of Zimbabwe. This can be used as a stepping stone toward scaling out of the different CA systems that are being promoted for sustainable intensification of smallholder farming systems of Southern Africa.
Because short-term biophysical and socio-economic assessments of the four cropping systems are not suitable, it is necessary to evaluate them in the longer term to fully understand how these systems perform under different agro-ecological and socio-economic conditions. Further studies also need to focus on water and nutrient use efficiencies in these cropping systems under the different agro-ecological conditions and smallholder farmer circumstances in Southern Africa.
Acknowledgements
The authors acknowledge the funding from DFID through the Protracted Relief Program (Grant ZWPRPIF2010002) and FAO (Grant OSRO/ZIM/402/UK, SAFR/091/07). This study was embedded in the Strategic Initiative (SI) 2 of the MAIZE CRP. Farmers, AGRITEX and NGO officers from the project communities put in a tremendous effort to make this work possible, their effort is greatly appreciated. Research technicians from CIMMYT namely Jefias Mataruse, Sign Phiri and Herbert Chipara worked tirelessly in the implementation of project activities and data collection from the trials, we acknowledge their contribution to this work. The authors are also grateful to Rumbidzai Matemba-Mutasa for conducting stability analyses on the yield data.
