On-farm experiments and trials were conducted as mother-baby trials [14] in irrigated conditions during 2008-09, 2009-10, and 2010-11 in three villages (Maourey Kouara Zeno: 13˚35.61'N, 2˚39.03'E; Yerimadey: 13˚29.11'N, 2˚41.30'E; Bokossay: 13˚25.72'N, 2˚47.32'E) in the Fakara district, Dantiandou commune, Tillaberi region, western Niger, about 50 km northeast of Niamey, the capital (Figure 1). The villages comprised 61 households in Maourey Kouara Zeno, 90 in Yerimadey, and 51 in Bokossay. The Zarma, the principal ethnic group of this area, are agriculturists engaged mainly in the rainfed production of grains such as millet (Pennisetum glaucum (L.) R.Br.) and cowpea (Vigna unguiculata (L.) Walp.). The prevailing soil type in the Fakara district is Psammentic Paleutalf type of which 90% is sand fraction, with a pH of 5.3, 0.16 g・kg⁻¹ total nitrogen, 0.2% organic carbon, 4.1 mg・kg⁻¹ available phosphorus, and 1.2 cmol・kg⁻¹ effective CEC and typical characteristics of an infertile soil [16] - [18] . Rain falls from June until September, with a peak in August and total of about 550 mm [16] . From 2001 to 2008, the mean rainfall was 75.5 mm in September and 7.9 mm in October, little to none (<0.5 mm) to March, and 7.4 mm in April. The mean daily temperature ranged from 23.9˚C to 30.2˚C from September to February, 31.2˚C in March, and 33.9˚C in April.

Irrigated vegetable production starts in late September and ends in early April. The first season, thus, started from September in 2008 and ended in early April in 2009. The second season started from September in 2009 and ended in early April in 2010. Because of the location of available water resources and the need for daily management, especially irrigation, vegetables are grown on farms adjacent to each village (Figure 2). Wells are located at village center and used both for drinking and for irrigation. Metal fences protected the crops from animals at Mourey Koala Zeno and Yerimadey.

We used mother-baby trials to evaluate farmers’ selectivity of the technologies. We set up experiments and

demonstrations at one site in each village, which we regarded as the mother trial. Before setting up the mother fields, we conducted 1 to 3 days of farmer training in each village in March or April of 2008, 2009 and 2010. In this training, we presented the types and effects of organic and inorganic fertilizers, micro-dosing techniques [19] [20] for the application of mineral fertilizer, differences between traditional and advanced composts, characters of the exotic vegetables, sowing or planting methods for each crop, preparation and management of seedlings and so on. About 25 men and 25 women attended each meeting. After the training, we encouraged the farmers to practice the technologies of their choice on their farms, in what we regarded as baby trials. Groups of some farmers elected to conduct our trials exactly; we regarded these group baby trials as separate, and encouraged the farmers to apply fertilizer at half or quarter doses to avoid repetition and to test the effect of dose. Farmers who did not attend the training also conducted trials with reference of the activities of the baby-trial farmers; we regarded these as non-baby trials, and recorded them as well. The results of the mother trial in each year were presented to the farmers at a field day at the mother field in October to November in 2008 and 2009. The results of the baby trials were reviewed by the farmers at meetings in February 2009 and April 2010.

In each mother field, three mid-sized plots (3 m × 4 m) and ten small plots (3 m × 1 m) were established in 2008. Plots received one of four fertilizer treatments: manure, mineral fertilizer, both or neither. Thirteen vegetables were grown: bell pepper (Capsicum annuum L. var. angulosum), cabbage (Brassica oleracea L. var. capitata), carrot (Daucus carota L.), chili (Capsicum annuum “Acuminoum”), eggplant (Solanum melongena L.), lettuce (Lactuca sativa L.), melon (Cucumis melo L.), onion (Allium cepa L.), potato (Solanum tuberosum L.), pumpkin (Cucurbita maxima Duch.), sweet potato (Ipomoea batatas Poir), tomato (Solanum lycopersicum L.), and zucchini (Cucurbita pepo L.). The manure treatment comprised 9 t/ha of cow manure (110 kg・N/ha) with respecting the way of farmers in this region. The mineral fertilizer treatment comprised 3 g of compound mineral fertilizer (N:P:K = 15:15:15) at sowing and 2 g of urea (13.7 kg・N/ha) per hill after one month of sowing according to the methodology and the recommended doses of micro-dosing technique [19] [20] . The manure + mineral fertilizer treatment comprised 9 t/ha of cow manure, 3 g of compound mineral fertilizer (N:P:K = 15:15:15) at sowing, and 2 g of urea (123.7 kg・N/ha in total) per hill after one month of sowing. Although the application dose of nitrogen differed among the treatments, we respected the ways of micro-dosing technique and farmers in this region and thus we did not adjust application doses to ensure a standardized amount of nitrogen. The experiments were conducted in a split-plot design with three replications (one per village). Seedlings of bell pepper, cabbage, chili, eggplant, lettuce, onion, and tomato were transplanted into the small plots at 0.3-m × 0.3-m spacing about 2 weeks after sowing. Tubers or seedlings of potato and sweet potato were planted directly into the small plots at 0.3-m × 0.3-m spacing, and those of melon, zucchini, and pumpkin into the mid-sized plots at 1.0-m × 1.0-m spacing, and thinned to one plant per hill after 2 weeks. Seeds of carrot were sown in line in small plots. Crops were planted in September and harvested at physiological maturity from February to April. The plants were irrigated twice a day in morning and evening with a 24-litter of watering can from pond (Maourey Koala Zeno) or pumping wells. Weeding was conducted timely during the study period. The fresh biomass and yield were recorded in season 1 and 2.

In the group baby trials, in which we encouraged the farmers to apply fertilizers at half or quarter doses, all other details were the same as in the mother trials. The crops were sown and harvested with reference to the mother trials. For the baby trials, we encouraged the farmers to apply fertilizers at the same dose in the mother trials.

The differences between treatments were determined by ANOVA followed by Student’s t-test in JMP version 9.0.0 software (SAS Institute, Cary, NC, USA).

Following the first season of irrigated vegetable production, we combined all plots into one big plot (3 m × 24 m) in the mother fields. Each village had four such plots. In addition, we established another four such plots in fields with a history of rainfed grain cultivation. Each mother vegetable field and each mother rainfed grain field received one of the four fertilizer treatments, and a crop of millet (“Haini Tchirey”, a local landrace; 120 days to harvest) was sown in each. The fertilizers were applied as in the vegetable experiments. The experiment was conducted in a split-plot design with three replicates (one per village). Millet was sown at 1.0 m × 1.0 m in June 2009, thinned to three plants per hill after 2 weeks, and harvested at physiological maturity in October 2009. Millet plants were harvested and partitioned into ears and stalks. After drying, the ears were threshed and the total biomass was determined. All aboveground crop residues were removed from the plots at the end of the cropping season.

After training and before the baby trials in 2008, we interviewed the baby-trial and non-baby-trial farmers about them, their families, and their farms. After establishing the baby trials, we divided the data between the two groups.

To gauge the economic situation of the women, we visited an average of 24.3 female farmers in each village in 2008 and interviewed them about source of income and money in their possession. Additional farmers were interviewed if numbers were low.

Soil sampling and analysis were done according to Omae et al. [21] .

Both fertilizer application and vegetable type were similar between the mother and group baby trials (Table 1). The effects of both were significantly different when all results were averaged over seasons 1 and 2. Manure plus mineral fertilizer increased vegetable yields by 161% (P < 0.01), followed by mineral fertilizer at 119% (P <0.01). The yields of cabbage (11.4 t/ha) and onion (8.4 t/ha) were highest, followed by lettuce, carrot, potato, and melon. The yields of cabbage and onion were significantly greater than those of pumpkin, eggplant, tomato, sweet potato, zucchini, chili, and bell pepper. The interaction between fertilizer and vegetable was significant in the group baby trials in season 1.

In the group baby trials in season 1, mineral fertilizer and manure plus mineral fertilizer increased yields by 190% - 215% (P < 0.01) relative to no fertilizer; the yields of cabbage (10.2 t/ha) and carrot (9.7 t/ha) were highest, followed by lettuce and pumpkin. In season 2, the yield of onion (13.0 t/ha) was highest, followed by potato, melon, and eggplant.

Manure plus mineral fertilizer applied to mother vegetable fields increased millet biomass by 115% (P < 0.01) and yield by 118% (P < 0.01) relative to no fertilizer (Table 2). Manure plus mineral fertilizer applied at half

*P < 0.05, **P < 0.01, ***P < 0.001, n.s. not significantly different. Values followed by the same letter are not significantly different at P = 0.05 by Student’s t-test.

*P < 0.05, **P < 0.01. Values followed by the same letters are not significantly different at P = 0.05 by Student’s t-test. “Veg.” indicates mother rainfed vegetable fields. “Grain” indicates mother rainfed grain fields.

rate to group baby vegetable fields produced a similar increase in millet biomass (n.s.) and increased millet yield by 195% (P < 0.05). Mineral fertilizer applied to mother vegetable fields increased millet biomass by 213% (P < 0.01) and yield by 160% (P < 0.05) relative to that applied to mother rainfed grain fields. Similarly, manure plus mineral fertilizer applied to mother vegetable fields increased millet biomass by 112% (P < 0.01) and yield by 124% (P < 0.05) relative to that applied to mother rainfed grain fields. In the group baby fields, millet biomass and yields showed increasing trends with increasing fertilizer dose from quarter to full (Table 2).

Total N, organic C and Bray I P of no application plot in season 1 reflected initial soil condition of Fakara very well [18] except for pH, which showed higher value due to be exposed to reducing condition by irrigation (Table 3). The no-fertilizer treatment significantly increased pH compared with fertilizers in season 1. Manure increased total N by 137% (P < 0.01) relative to no fertilizer in season 1. It also increased total N by 83% relative to mineral fertilizer alone in season 1, and by 55% (P < 0.05) in season 2. It increased organic C by 150% (P < 0.01) relative to no application and mineral fertilizer alone in season 1. Manure plus mineral fertilizer increased Bray I P by 53% - 829% (P < 0.001) relative to the other treatments, and increased total P by 18% - 32% (P < 0.05) relative to no fertilizer and mineral fertilizer alone.

There were no significant differences between baby-trial and non-baby-trial farmers in percentage of women, farmer age, cultivation area, numbers of plots, number of vegetable types, number of plants, fertilizer application, percentage of area cultivated, or years of experience (Table 4). In contrast, 74.8% of baby-trial farmers belonged to a farmers’ group, versus only 13.3% of non-baby-trial farmers (P < 0.01). In both groups, about 75% of farmers were women. Both groups grew 3 to 4 vegetable types on 200 - 235 m². Around 40% of both groups applied manure plus mineral fertilizer, while 33.7% - 49.1% applied none. Both groups preferentially grew pumpkin (38.2% - 57.5%), followed by cabbage, onion, lettuce, and watermelon (Table 4(b)). Both groups had the longest experience with pumpkin (2.4 - 5.8 years), followed by watermelon, tomato and eggplant, zucchini, lettuce, onion, potato, and cabbage (Table 4(c)).

*P < 0.05, **P < 0.01, ***P < 0.001, n.s. not significantly different. Values followed by the same letter are not significantly different at P = 0.05 by Student’s t-test.

^aMean of 43.3 farmers. ^bMean of 40.5 farmers. **P < 0.01, n.s. not significantly different. Values followed by the same letters are not significantly different at P = 0.05 by Student’s t-test.

^aMean of 43.3 farmers. ^bMean of 40.5 farmers. n.s. not significantly different.

^aMean of 43.3 farmers. ^bMean of 40.5 farmers. n.s. not significantly different.

The number of vegetable types grown increased by 105% (P < 0.05) in season 1 relative to before the trials, returned to the basal level in season 2 (P < 0.05), and decreased by 85% in season 3 (n.s.) (Table 5). Before the trials, farmers grew pumpkin, onion, cabbage, lettuce, watermelon, tomato, and eggplant. Throughout the trial, they regularly grew pumpkin, cabbage, lettuce, onion, and potato. The percentage of farmers who applied manure plus mineral fertilizer in season 2 increased by 55% relative to before the trials, and those who applied

none decreased by 85%, although the differences were not significant. The farmers earned 244 142 to 287 125 CFA francs per village per year.

Yields of cabbage and onion were highest, and yields of pumpkin, eggplant, tomato, sweet potato, zucchini, chili, and bell pepper were lowest, regardless of fertilizer application (Table 1). Manure plus mineral fertilizer increased overall vegetable yields by 161%, followed by mineral fertilizer at 119%. The biggest vegetable crop in Africa in 2013 was potato (30 499 kt), followed by sweet potato (20 131 kt), tomato (18 649 kt), onion (9650 kt by dry weight), cabbage (4114 kt), green chilies and bell peppers (2874 kt), melons (2068 kt), and pumpkin (1990 kt) [22] . The production potential of potato, onion, cabbage, and melon in our trials is thus parallels vegetable production in Africa, but that of lettuce and carrot does not. Because lettuce and carrot can be grown more easily and harvested sooner than potato, onion, cabbage, and melon, they may offer an advantage in increasing yields by increasing plantings. Carrot, in particular, can be harvested early by thinning. This may be compatible with the minimal experience of the farmers with those crops (Table 4(c)).

The fresh biomass and yields of vegetables tended to increase as the rate of fertilizer increased (Table 2). This means that an upper limit on application rates was not reached. Excessive application of fertilizer has led to annual positive nutrient balances of 843 kg・N/ha, 70 kg・P/ha, and 200 kg・K/ha in urban and periurban agricultural systems in Niamey [23] . Similarly, annual surpluses of 85 - 882 kg・N/ha, 109 - 196 kg・P/ha, and 20 - 306 kg・K/ha were applied in small-scale periurban vegetable farming in Hanoi, Vietnam [24] . These rates pose high risks of soil pollution [25] . The rates applied in our study (110 kg・N/ha as manure and 13.7 kg・N/ha as mineral fertilizer) are less than the above reports. Further study will be needed to investigate the effects of fertilizer application and rates on soil pollution, and to determine ways to take up excess fertilizer such as through continuous cropping.

The women earned 244,142 to 287,125 CFA francs per village per year by irrigated vegetable production. This equates to 10 to 12 bags of millet per village, or ?.2 to ?.9 per woman (656 CFA franc = ?). Almost 90% of farmers (n = 73) stated that money is the biggest limitation on activities. Small businesses account for the largest proportion of income generation by women (30.5%), followed by sales of agricultural products (17.2%), labor (6.9%), sales of legumes (4.5%), sales of animals (3.4%), non-local work (2.2%), and sales of wild medicinal plants (1.1%); the rest (34.0%) work in all other areas. The women farmers have only ?.8 each in their pockets. In many parts of Africa, women and men operate separate income and expenditure, with women carrying the primary burden for the financing of children’s welfare [6] . Women and children are rarely the beneficiaries of the income generated by cash crops; their present wellbeing is founded rather in subsistence farming. For women to supplement their earnings, small-scale-income-generating activities must be supported.

Around 40% of farmers, both baby-trial and non-baby-trial, were already familiar with applying manure plus mineral fertilizers (Table 4(a)). This practice improved soil fertility (Table 3) and grain production in the following rainy season (Table 2). In mother vegetable fields also, manure plus mineral fertilizer increased millet biomass and millet yields.

In contrast to the low nutrient inputs and the negative NPK balances that generally prevail in West Africa [26] [27] , some farmers apply very high rates of nutrients, leading to large nutrient surpluses [24] [28] [29] ; high inputs of irrigation water and fertilizer have been recorded in Benin, Ghana, and Niger, in West Africa [23] [30] [31] . Continuous cropping is thus beneficial for the management of total soil fertility in the region, and can give women farmers, who lack access to sources of income in West Africa, as in Benin [9] , more chance to earn incomes.

There was no significant different between baby-trial and non-baby-trial farmers in percentage of women, age, cultivation area, number of plots, number of vegetable types, number of plants, fertilizer application, percentage of area cultivated, or years of experience (Table 4). Because the location of water resources is limited, and the farmers gather every day to draw water for irrigation (Figure 2), the opportunity to chat may promote technical exchange and assist in the dissemination of farming technologies.

Although the use of manure plus mineral fertilizer fluctuated over time, there were no significant differences in usage (Table 5). Nevertheless, the farmers may see the advantage of this practice from the results of the mother trials (Table 1) and the baby trials (Table 2).

From the results of seasons 1 and 2, the women farmers were able to select the five vegetables that they preferred: pumpkin, cabbage, lettuce, onion, and potato (Table 5).