2.1. Study Area
The study was conducted in the Soudano-Sahelian zone of southern Mali, covering the commune of Tominian (13.2857° N, 4.5908° W) and Yorosso (12.3548° N, 4.7782° W). The rainy season lasts from June to October with rainfall peaks in August. In 2016, mean seasonal rainfall in the region was 1005 mm while in 2017 seasonal rainfall was 861 mm. The number of rainy days was 51 in 2016 compared to 45 in 2017. The dry season includes a relatively cold period from November to February and a hot period lasting from March to May. The mean maximum temperature is 34 °C during the rainy season and 40 °C during the hot dry period.
Vegetation in the region is savannah with trees and shrubs, mainly from a natural regeneration system, and cultivated lands are mainly characterized by parks of Vitellaria paradoxa (shea nut tree), Parkia biglobosa (néré), and Adansonia digitata (baobab). The mean population density is 16.4 inhabitants per km
2 with a mean of 8 persons per household [
19].
Cropping land is spatially dispersed and the largest share is allocated to cereal production. Sorghum (Sorghum bicolor (L.) Moench) and millet (Pennisetum glaucum (L.) R.Br.) are the main crops, representing, respectively, 38% and 32% of the cultivated area, but maize (Zea mays L.) is also important, covering 12%. Cereals are grown in a two- or three-year rotation with cotton (Gossypium hirsutum L.). Fertilizer and pesticides are mainly applied to cotton and maize. Millet and sorghum usually do not receive fertilizer but benefit in the crop rotation from previous fertilizer applications to cotton or maize.
Cattle is a key component of the mixed crop-livestock farming systems in the study area. Eighty per cent of farmers own at least one pair of oxen, a cultivator, and a seeder, and use animal traction for soil preparation, weeding, and sowing [
3].
The soils are mainly Ferric Lixisols with low clay content (<10%) in the topsoil. Soils are in general moderately acidic with a pH of around 5–6 [
20] and with low nutrient holding capacity and low organic matter content [
21]. The fertilizer application rates recommended by agricultural research and extension services have generally proven too costly for smallholder farmers. In addition, they involve a high financial risk, which is a major factor driving decision making for smallholder farmers [
22].
2.2. Field Experimentation
Experimentation was the last phase in a series of four activities focusing mainly on biophysical characterization of farm fields, farmer’s dialogue on the cereal-cowpea intercropping system, technical organization, and cropping system selection by the respective farmers.
A total of 159 trials including 76 with millet and 83 with sorghum, both intercropped with cowpea, were conducted in 2016 and 2017 in 108 villages. The experimental design for each trial was arranged in a randomized block design with 4 treatments based on 3 improved cowpea varieties and the farmers’ local cowpea variety. The selection of cowpea varieties was oriented towards the farmers’ objectives, which were mainly based on earliness of production and availability of biomass for animal feeding. The same varieties were simultaneously tested at all sites.
The intercropping system was designed by the community based on a previous study [
15] and on the farmers’ experience. The implemented intercropping system consisted of 2 rows of cereals (millet or sorghum) followed by 2 rows of cowpea varieties (
Table 1). Each farmer selected either millet or sorghum in combination with cowpea varieties.
Field plot size for each treatment was 100 m2. For cereals (millet and sorghum) and cowpea varieties (wilibali and korobalen), the inter-row distance was 0.75 m, with a within-row plant distance of 0.4 m because of erected stem character and small space occupation rate. For sangaranka and local varieties, within-row plant distance was 0.8 m and the inter-row distance was 0.75 m. These varieties are creeping crops with large space occupation rates. The distance between adjacent cereal and cowpea rows was 0.75 m. All crops were thinned (2 plants/hole) at 15 days after planting to achieve the recommended planting densities.
Planting dates mostly occurred in June. Weeding was carried out before 20 days after planting and again between 30 and 40 days after planting (
Table 2), i.e., weeding was completed twice for each field. Based on the national recommendation and the farmers’ common practice, an average of 100 kg/ha of diammonium phosphate (18-46-0) was applied between 15 and 20 days after planting. To protect crops from enemies, particularly cowpea, water-based Neem [
23] was spread between 35 and 45 days after planting (DAP) for the first application and between 50 and 55 DAP for second application.
Rotation effect was determined based on the crop cultivation calendar for the previous three years. In total, the effects of three types of rotation, i.e., cereal-cereal, cereal-legume, and cereal-cotton, on yield were analyzed using an unbalanced design regression model. The cereal consisted of millet, sorghum, or maize while the legume consisted of groundnut or cowpea.
2.3. Measurement
The timing of different operations including planting, weeding, harvesting, and fertilizing was recorded by field technicians. Crop physiology status such as flowering and maturity dates was also collected. At crop maturity, farmers harvested the total area of the plot with the assistance of the researchers. Mature millet and sorghum plants were harvested following the local practice of cutting the panicles and bagging. Legume pods were harvested when mature. Biomass of all crops was weighed at the plot, and a sub-sample was taken for weighing. Millet ears, sorghum panicles, and legume pods were dried on a clean floor at the homestead and were threshed and hand-winnowed; legume pods were shelled by hand. Grains were weighed and grain sub-samples were taken and weighed as well. All sub-samples (grain and biomass) were dried and re-weighed to determine dry weights in kg/ha.
2.4. Statistical Analysis
Because of the varied number of experiments per village, across villages, and per year, we used an unbalanced design using the GenStat regression model for the variables mean separation. Firstly, ANOVA was performed to separately evaluate the simple effect of cowpea grain yield with cowpea biomass yield under intercropping with millet and sorghum. Secondly, for the purpose of economic analysis, we compared yield of grain to biomass under intercropping with each of the two cereals and their respective interactions with varieties. Treatment structure consisted of either grain or biomass variables for varieties per crop and their respective interactions with the year, representing the annual rainfall effect. Villages were considered as replicate. Significant means were separated using average standard error of difference (SED). We also used Box plots for capturing the distribution of variables.
2.4.1. System Gain
To determine system economic gain per hectare, we used a gross margin (GM) analysis model that is equal to the difference between total revenue (TR) and total variable cost (TVC) and is expressed as follow:
Total revenue means the total market price of production per hectare multiplied by the crops’ yields (grain or biomass) while TVC includes mainly input costs such as insecticide, fertilizer, and ploughing. The system economic gain was expressed in West African CFA franc (XOF).
2.4.2. Farmers’ Ranking of Cowpea Varieties
A total of 30 farmers (18% of the total) participated to the prioritization of the cereal-cowpea intercropping systems using a paired comparison scaling method. Each farmer was requested to provide a weighted score for cowpea grain, biomass, and total income. The respective scores were multiplied by the number of scores for each cowpea variety to obtain a total weighted score that was then divided by the total number of respondents to obtain the weighted mean score (WMS). Rank order was given according to the WMS values.