The field trials were located at Kpalga (latitude 09° 26′ 45.8″ N and longitude 000° 57′ 49.7″ W with an elevation of 170 m above sea level) and Kpachi (latitude 09° 25′ 48.5″ N and longitude 000° 58′ 28.0″ W with an elevation of 181 m above sea level) in the Northern region of Ghana, and also at Nyagli (latitude 10° 08′ 54.1″ N and longitude 002° 23′ 15.9″ W with an elevation of 173 m above sea level) in the Upper West region of Ghana (Supplementary Figure S1). The two locations in the Northern region are herein referred to as Nyankpala. The soils in the Nyankpala and Nyagli were Acrisols and Leptosol, respectively. The study sites have a unimodal rainfall distribution with an annual rainfall of 1000–1200 mm and mean temperature between 26 and 30°C with little variation throughout the year. Weather data and rainfall pattern for the growing season at the study sites were downloaded from http://www.awhere.com. The experiment was conducted between June and November, 2014. The fields had no known history of Rhizobium inoculation and were planted to maize (Zea mays L.) in the previous cropping seasons. Unless otherwise stated, all laboratory analyses reported were carried out on soils from both soybean and cowpea fields following standard protocols. Seven soil core samples were taken from each plot, thoroughly mixed and composite samples taken into transparent polythene bags and kept in the refrigerator at 4°C prior to laboratory analysis. The soil parameters analyzed were particle size (hydrometer method), soil pH (1:2.5; H₂O), organic carbon [modified Walkley and Black procedure as described by Nelson and Sommers (1982)], total nitrogen [Kjeldahl method as described by Bremner and Mulvancy (1982)], available soil phosphorus [Bray No. 1 solution as outlined by Olsen and Sommers (1982)], and exchangeable potassium [ammonium acetate (NH4OAc) extract; Table 1]. Calcium and magnesium were determined in 1.0 M ammonium acetate (NH4OAc) extract (Black, 1965). Iron (Fe) and Nickel (Ni) were determined by using Diethylenetriamine pentaacetic acid (DTPA) extractant. The values were read on Atomic Absorption Spectrophotometer (Buck Scientific model 210 VGP).

The enumeration of rhizobia in soils and inoculants were carried out using the most probable number count method (Vincent, 1970) where cowpea was grown in plastic growth pouches (Mega International, USA) and inoculated with serial dilutions of soils from the experimental sites and inoculants (Woomer et al., 1997). Soybean was not included as a trap host after several failed attempts to raise healthy seedlings due to poor growth in growth pouches compared to the better growth of cowpea. The soybean being promiscuous could be substituted with cowpea which is also promiscuous without compromising the results of bacterial counts in inoculum or soil. Also, because each of the inoculants, Legumefix and Biofix, has only one strain of Bradyrhizobium japonicum, it is expected that cowpea and soybean will have equal chance to be infected by each strain. Thus, cowpea was then used to estimate nodulating bacterial numbers in these inoculants as well as in the soils. Uniform clean cowpea seeds of good viability were surfaced sterilized with 95% alcohol for 10 s and 3% hydrogen peroxide for 3 min and rinsed in six changes of sterilized distilled water as described by Somasegaran and Hoben (1994). Ten steps, 10-fold dilutions and six steps, fivefold dilutions were prepared for the Bradyrhizobium inoculants (BR 3267, BR 3262, Biofix and Legumefix) and the soil samples, respectively. Pattern of nodulation was assessed after 28 days based on the presence or absence of root nodules. Population estimates were assigned using MPNES software (Woomer et al., 1990) (Table 2).

The BR 3267 and BR 3262 are industrial strains imported from Brazil as slant cultures. The strains were further cultured on yeast mannitol agar (YMA) incubated at 28°C. The cultured strains were then looped into yeast extract mannitol broth and placed in an orbital incubator at a temperature of 28°C at 125 rpm until it become turbid. Peats imported from IITA, Nigeria, were bagged (50 g peat/bag) and gamma radiated at Ghana Atomic Energy Commission (GAEC). Using a 20 ml sterile syringe with 18 gauge needle, 50 ml of the B. japonicum broth cultures were withdrawn from the broth and introduced into 50 g peat aseptically under the laminar flow cabinet as described by Somasegaran and Hoben (1994). The bags were then aseptically sealed, labeled accordingly and gently massaged until the inoculum was absorbed evenly by the peat. The freshly prepared inoculants were then incubated 28°C for 2 weeks to cure (Somasegaran and Hoben, 1994). Direct cell count by the drop plate method was done to verify the colony forming units in the cured inoculants.

The field was plowed and harrowed to a depth of 15 cm and divided into plots before planting. Each plot measured 6 × 3 m with an alley of 2 m between plots and 3 m between blocks. Soybean cultivar, Jenguma, and cowpea cultivar, Songotra, were used for this study.

Five grams of each of the Bradyrhizobium inoculants was added to 1 kg of seeds using the two–step method which involves adding a sticker to the seeds before the inoculant. Gum Arabic was used as a sticker in this study at a ratio of 1.5 g to 15 ml clean lukewarm water. Inoculated soybean and cowpea seeds were air dried for 30 min and manually sown at a spacing of 75 cm × 10 cm.

The study was laid out in a randomized complete block design with five replications and four treatments for each crop. The treatments were; Legumefix with B. japonicum strain 532 C (Becker Underwood, UK), Biofix with B. japonicum strain USDA 110 (MEA, Kenya), uninoculated control and uninoculated with applied N in the form of urea at a rate of 100 kg N ha^-1 for soybean and strains BR 3262 and BR 3267 (EMBRAPA, Brazil), uninoculated control and uninoculated with applied N at a rate of 100 kg N ha^-1 in the form of urea for cowpea. The urea was split applied; 50 kg N ha^-1 at 1 week after planting and the other half at 50% flowering (R₃ growth stage). Each treatment received a basal application of 30 kg P ha^-1 and 30 kg K ha^-1 as triple super phosphate and muriate of potash, respectively.

Nodulation and shoot biomass were assessed at the R₃ stage for both soybean and cowpea. The plants were cut at about 5 cm above the soil level. The roots of the plants were carefully dug out and collected into polythene bags, together with detached nodules and transported to the laboratory. The roots were put in a 1 mm mesh sieve and washed under running tap water to remove adhered soil. The nodules were gently removed, washed, and counted. Shoot and nodules were oven dried at 60°C for 72 h. Shoot dry matter was measured after harvesting the pods at maturity (R₈ stage). Plants were harvested from the inner rows excluding the border rows and oven dried at 60°C for 72 h. The grains were weighed with a standard electronic balance and recorded.

Return on investments for using the Bradyrhizobium inoculants were calculated using value cost ratio (VCR). The VCR was calculated based on the adopted equation from Nziguheba et al. (2010):

Where YB is the grain yield from treated plots, YC is the grain yield from uninoculated control plots, PG is the unit price of grain yield, PB is unit price for inoculant or fertilizer, and QB is the quantity of inoculant or fertilizer. The dollar to cedi exchange rate as at the time of this study was USD $ 1 to GH¢3.60. An inoculant with a positive VCR was considered to be economically viable. A VCR value greater than a threshold of 3–4 was considered profitable (Dittoh et al., 2012).

The data obtained from Kpachi and Kpalga was pooled together (herein referred to as Nyankpala) in order to assess the interaction between strains and locations. The data was transferred into GenStat statistical software version 12 for Analysis of Variance (ANOVA). Significant differences were assessed at 5% (P = 0.05) level of significance. Where there was significant difference, means were separated using the Fishers protected least significant difference (LSD) procedure. Orthogonal contrast was used to compare individual treatments. Multivariate Analysis of Variance (MANOVA) was used to assess the contributions of the strains in the inoculants to grain yield.

The physical and chemical properties of soils of the experimental sites are presented in Table 1. The soils of the study locations were predominantly sandy loam except Nyagli which was sandy. The pH recorded was near neutral. Total N and organic carbon were very low. Available P ranged from low to very low. Potassium, Calcium, and Magnesium were adequate while DTPA-Fe was low. The ratings were done according to Landon (2014).

The daily accumulated and number of rains received at the study sites are highlighted in Supplementary Figures S2 A and B. There were short periods of dry spells between 10 and 20, 30 and 40 days after planting. In general, however, the rainfall pattern at Nyankpala was much better than Nyagli. There were 10 days of short dry spells in Nyagli with an average daily rainfall of 0.74 mm during the flowering stage of the plant (Supplementary Figure S2A) whereas in Nyankpala average rainfall was ≥10 mm during the same growth stage (Supplementary Figure S2B).

Number of Bradyrhizobium cells g^-1 inoculant and the population sizes of the indigenous rhizobia per location are presented in Table 2. The study sites had low numbers of indigenous rhizobia (<10 rhizobia cells g^-1 soil). However, the population of Bradyrhizobium in the various inoculants ranged from 10⁶ to 10⁸ cells g^-1 inoculant.

The results of the effects of Bradyrhizobium inoculation on soybean and cowpea in the Nyankpala are presented in Table 3. Bradyrhizobium inoculation significantly (P = 0.001) increased soybean nodulation over the uninoculated plants with or without nitrogen in Nyankpala (Table 3). Percentage increases in nodule number due to Biofix and Legumefix were 91 and 118%, respectively. Inoculation increased nodule weight by more than two- and three-folds compared to uninoculated control and N-fertilized plants, respectively (Table 3).

Inoculation did not significantly (P = 0.22) increase shoot biomass of soybean (Table 3). Inoculation of cowpea with BR 3262 and BR 3267 significantly (P = 0.001) increased shoot dry weight by 38 and 37%, respectively, compared to the uninoculated control in Nyankpala (Table 3). Inoculation of cowpea with BR 3262 produced a higher shoot dry weight but statistically similar to the weight produced by plants inoculated with BR 3267 and uninoculated control which received 100 kg N ha^-1 (Table 3). Both strains were equally effective in significantly (P = 0.001) increasing nodule number and weight (Table 3).

Inoculation of soybean with Biofix and Legumefix significantly (P = 0.001) increased grain yield by 19 and 12%, respectively, over the uninoculated control. Soybean plants that were inoculated with Biofix and Legumefix produced grain yields that were 97 and 92%, respectively, of the potential yield of Jenguma which is 2.5 tons ha^-1. The orthogonal contrast analysis revealed that the inoculated plants performed better than the nitrogen treated plants. The BR 3267 inoculated cowpea gave significantly (P = 0.001) higher grain yield than the uninoculated control. On the other hand, the effect of Bradyrhizobium strain BR 3262 on grain yield of cowpea was not significant (Table 3). Cowpea plants inoculated with BR 3267 and BR 3262 produced grain yields that were 46 and 37%, respectively, of the potential yield of Songotra which is 2.5 tons ha^-1.

The results of Bradyrhizobium inoculation on cowpea in Nyagli are given in Table 4. Bradyrhizobium inoculation significantly (P = 0.013) increased cowpea nodule number over the uninoculated plants with nitrogen at Nyagli in the Upper West region (Table 4). Nodule numbers produced by the inoculated plants were not significantly different from that of the uninoculated control. The BR 3267 inoculated plants produced significant (P = 0.004) increase in nodule weight by 62% compared to the uninoculated control (Table 4). The nitrogen-fertilized plants produced the lowest nodule number and nodule dry weight (Table 4).

Shoot biomass was not significantly (P = 0.135) affected by the treatments (Table 4). Nonetheless, the inoculated plants consistently recorded higher shoot biomass than the control. The grain yield of cowpea in Nyagli in the Upper West Region was generally low and there was no significant (P = 0.433) difference among the treatments (Table 4). No significant differences were observed between the two strains in all measurements in this location (Table 4). The contrast analysis did not show any significant difference among all the treatments in shoot dry weight and grain yield (Table 4). Wilks lambda values from MANOVA for soybean (0.067, P = 0.00) and cowpea (0.039, P = 0.00) showed that 93.3 and 96.1% of the variations observed in soybean and cowpea, respectively, were due to inoculation (Table 5).

Returns on investments based on VCR were positive and therefore indicated that all the applied inoculants were economically viable but Biofix/Legumefix, and BR 3267 were profitable for soybean and cowpea, respectively. However, considering VCRs of the inoculants; Biofix (8.7), Legumefix (4.1), BR 3267 (4.6), and urea (2.0), the use of inoculants would be more profitable in Nyankpala than urea (Figure 1). At Nyagli in the Upper West region, the VCR for cowpea inoculants and urea were BR3267 (1.8), B3262 (1.3), and urea (0.9). Unlike, Nyankpala, the use of inoculant in Nyagli would not be economically viable (Figure 1).