Optimal P fertilization using low-grade phosphate rock-derived fertilizer for rice cultivation under different ground-water conditions in the Central Plateau of Burkina Faso

ABSTRACT

Previous studies have reported two major methods of increasing PR solubility; calcination and partial acidulation. In addition, soil water condition would influence the solubility of P fertilizers. However, the effects of local P fertilizers, namely calcined PR (CPR) and partially acidulated PR (PAPR) on rice yield under different soil water conditions have not been explored comprehensively. The aims of the present study were to evaluate the effects of local P fertilizers produced from PR in Burkina Faso under different soil water conditions and to propose local fertilizers with optimal application rates for different soil water conditions. The field experiments were conducted at four farmers’ fields with different ground-water levels (GWL). CPR, PAPR, and superphosphate were applied at rates of 0, 7.6, 15.3, and 30.5 kg P ha⁻¹, respectively. Superphosphate mostly consists of water-soluble P fraction (WP), PAPR of WP and alkaline ammonium citrate-soluble P fraction (SP), and CPR of SP and 2% citric acid-soluble P fraction (CP). The solubility is in the order of WP > SP > CP. The GWL was monitored during the growing season, and yield components were observed. Results of multiple regression analysis showed that WP influenced grain yield under all soil water conditions, whereas SP only influenced grain yied at mean GWL > −24.7 cm. Therefore, PAPR with high WP has an advantage over CPR in the field with GWL > −18.7 cm, and both CPR and PAPR are effective in fields with GWL > −6.5 cm. The optimal application rate was 14.4 kg P ha⁻¹ as WP in the field with low GWL (mean −29.2 cm), 15.2–15.9 kg P ha⁻¹ as WP + SP in the field with middle GWL (mean −24.7 to −18.7 cm), and 11.1 kg P ha⁻¹ as WP + SP in the field with high GWL (mean −6.5 cm). According to the results, the optimal fertilizer types and application rates differ according to the soil water conditions.

KEY WORDS:

• Lixisols
• Ground-water level
• Rice
• Local phosphate fertilizer
• Optimal fertilization

1. Introduction

Rice is the most rapidly expanding food commodity, both in production and consumption, in sub-Saharan Africa (SSA) (USDA 2018). However, the average rice production in SSA is only around 2.1 Mg ha⁻¹, which is far below the potential yield (USDA 2018). Low phosphorus (P) content in the highly weathered soils of SSA is a major constraint for rice production (Nishigaki et al. 2019; Saito et al. 2001). According to (Bouwman, Beusen, and Billen 2009), global soil P balance will increase slowly, and soil P depletion may become a major problem in agricultural land. Consequently, many studies evaluated ways to increase rice production via P fertilizer application (Sahrawat et al. 2001; Savini et al. 2016; Vandamme et al. 2016). However, few farmers can access and utilize imported fertilizers, mainly because of the high cost incurred from long-distance transportation (Kelly 2006). Thus, local farmers may increase fertilizer input only if the fertilizers are available, affordable, and profitable for rice production (Tsujimoto et al. 2019). Nevertheless, global phosphate rock production has steadily increased (approximately 6 MMT per year) from 139.3 MMT in 2000 to 207.5 MMT in 2012 (Chowdhury et al. 2017).

Geological phosphate rock (PR) deposits occur throughout SSA (FAO 2004). According to the Burkina Faso Government, there is 100 MMT of PR in the Kodjari region of Burkina Faso (Van Straaten 2002). which could be exploited as a source of P as an alternative to expensive, imported P fertilizers. However, most PR deposits in SSA have been rarely utilized because of their low P content and reactivity (FAO 2004). Nakamura et al. (2013) summarized the potential utilization of local PR for rice production and showed that direct application of local PR could improve rice yield in lowland fields. In contrast, other studies have stated that the direct application of local PR in SSA is ineffective (Fukuda et al. 2013; Sale and Mokwunye 1993). These conflicting results indicate that various factors influence the fertilization effect of PR, such as PR solubility, climate, and environmental conditions (Chien and Menon 1995a; Sale and Mokwunye 1993).

Another approach to improve rice production using local PR is to enhance P solubility (Nakamura et al. 2019). Partial acidulation of PR by sulfuric acid has been promoted promoted by the International Fertilizer Development Center as a solubilization method. Acidulation of PRs with sulfuric acid produces the phosphate component monocalcium phosphate, small amounts of dicalcium phosphate, and residual apatite in the amount dependent on the degrees of acidulation (Mizane and Rehamnia 2012). The positive effects of partially acidulated PR (PAPR) on rice production have been reported (Chien and Menon 1995b; Rahman 2018). The degree of solubility in PAPR increases with the addition of more sulfuric acid; however, the viscosity and acidity of the product also increase because of unbound sulfuric acid (Akiyama, Tsumita, and Wada 1992; Frederick and Roth). Therefore, to secure adequate solubility, it is necessary to minimize the amount of sulfuric acid added. In contrast, Akiyama, Tsumita, and Wada (1992) proposed the calcination procedure for PR in a temperature range of >900°C with alkaline carbonates to convert fluoride apatite into α-tricalcium phosphate and rhenania phosphate. Nakamura et al. (2019) applied a modified version of this method to the low-grade PR from the Kodjari region and achieved more than 90% solubility in 2% citric acid. These authors Nakamura et al. (2019) also demonstrated that calcinated PR (CPR) significantly increased rice production in pot experiments. However, the effects of PAPR and CPR on rice cultivation under different soil conditions remain poorly understood, especially in West Africa.

Along with increasing the grain yield on existing lands favorable for rice production, it would be necessary to expand rice cultivation into surrounding marginal lands to attain rice self-sufficiency in many West African countries (Van Oort et al. 2015). The major problem in these lands is poor soil water conditions. Hömberg and Matzner (2018) and King et al. (2015) have stated that soil water conditions are closely related to the solubilization, transportation, and hence availability of P in soil. Despite the negative effect permanent flooding has on lowland rice cultivation (De Bauw et al. 2019), P fertilization is generally considered to be more effective under flooded conditions than under upland conditions on similar soils (Dobermann et al. 1998); consequently, optimal local fertilizer application should differ according to soil water conditions.

Thus, this study aims to evaluate the fertilization effects of PAPR and CPR under different soil water conditions. We hypothesize that P fertilization would be more effective under optimal water conditions for rice cultivation. The results could facilitate the selection of optimal fertilization methods under different soil water conditions. If PAPR and CPR improve rice yield in fields with low ground-water level, they could facilitate the expansion of rice cultivation areas into more marginal lands of the Central Plateau of Burkina Faso.

2. Materials and methods

2.1. Study site

The field experiment was conducted in 2019 in Nassoulou Village (12°21′12.0ʺ N 2°07′37.4ʺ W), Central Plateau of Burkina Faso, West Africa. The village is located in the Sudanian Savanna and characterized by mid-latitude steppe and desert climate (BSh) according to the Köppen climate classification. In this region, rice cultivation is generally conducted in lowland riverine areas (bas fond in French). Using a digital surface model constructed based on ALOS World 3D-30 m images (JAXA 2018), a line transect was placed along a gentle slope, starting from the river bottom, with an average gradient of 0.8% (190 m in length). On the line transect, four farmer fields were selected, which were named Middle slope, Lower slope (1), Lower slope (2), and Lowland. The Middle slope was located 165–190 m from the river bottom, and the two fields on the lower slope were located 60–135 m from the river bottom. The Lowland was located on the river bottom. The Lowland has been used as a paddy field for a long time. The two sites on the lower slope were occasionally used for rice cultivation. The field on the Middle slope was mainly used for upland crop cultivation.

A soil profile survey was conducted in October 2018 in the Lowland (Table 1). The soil was classified as Ferric Lixisol (IUSS Working Group WRB 2015), which is distributed on the lower slopes extending into the river bottom, and has higher productivity mainly because of the deeper effective soil depth (Ikazaki et al. 2018a). The topsoil was characterized by high sand content and low carbon content, resulting in a weak soil structure. Clay content gradually increased with soil depth. The texture class changed from sandy loam at the surface to heavy clay at the bottom. Available P content was determined by both Bray-1 and Bray-2 methods in the profile and ranged from 0.26 to 1.79 and 1.00 to 3.25 mg P kg⁻¹, respectively. Similar to other areas of the Sudan Savanna, the availability of P was quite low (Ikazaki et al. 2018b).

Table 1. Physico-chemical properties of lowland soils

Horizon	Depth	Soil texture			Total carbon	Total nitrogen	Available P		pH
		Sand	Silt	Clay			Bray-1	Bray-2	H2O	KCl
	(cm)	(%)			(g C kg^−1)	(g N kg^−1)	(mg P kg^−1)
Apg	0–6	77.0	11.1	11.9	4.40	0.39	1.79	3.25	5.1	4.1
ABg	6–24	50.8	14.3	34.9	4.93	0.48	0.63	2.00	6.0	4.6
Bg1	24–35	43.2	12.9	43.9	4.06	0.49	0.26	1.00	6.2	4.7
Bg2	35–50	39.2	11.5	49.3	2.84	0.36	0.31	1.75	6.4	5.0
BCg	50–70	39.5	10.5	50.0	2.39	0.31	0.44	1.75	6.5	5.2
Cg	70–80+		-	-	2.85	0.35	0.59	4.75	-	-
Horizon	Depth	EC	Exchangeable cations					CEC	Base saturation
			Na^+	K^+	Ca^2+	Mg^2+	Al^3+
	(cm)	(S m^−1)	(cmolc kg^−1)						(%)
Apg	0–6	5.7	0.04	0.12	1.42	0.57	0.42	3.8	56.2
ABg	6–24	2.2	0.09	0.17	5.29	2.05	0.08	10.1	75.0
Bg1	24–35	1.9	0.09	0.18	5.67	2.11	0.07	11.2	72.0
Bg2	35–50	2.2	0.12	0.28	6.43	2.70	0.07	17.6	54.2
BCg	50–70	2.4	0.12	0.34	6.90	3.05	0.07	16.3	64.0
Cg	70–80+	-	0.16	0.50	7.99	3.71	-	18.0	68.5

Soil horizons were classified according to WRB (IUSS Working Group WRB 2015). EC: electrical conductivity; CEC: cation exchange capacity.

2.2. Preparation of calcinated and partially acidulated phosphate rock

Two types of fertilizers were produced by partial acidulation and calcination method using PR from the Kodjari deposit (12°1′ N; 1°55′ E) in Burkina Faso. The PR contained 113 g P kg⁻¹, and its 2% citric acid solubility was 31.1% of the total P. Total P content and solubility were similar to those documented previously (Nakamura et al. 2019; Zapata 2004). To prepare PAPR, we followed the acidulation method described by Frederick and Roth (). The addition rate of sulfuric acid was determined based on the mineral composition of PR; 326.8 mL H₂SO₄ was used to 1 kg BPR to produce 100% acidulated phospate rock (PAPR100). Therefore, 245.1 mL H₂SO₄ was added to prepare PAPR75. In the present study, we employed PAPR75 as PAPR for the experiments. The calcination procedure described by Nakamura et al. (2019) was used to prepare CPR. Fine powdered PR was mixed well with K₂CO₃ at a rate of 166 g K per kg PR. The components were mixed with distilled water, and then pressed to form coin-shaped pellets. The pellets were calcinated at 900°C for 10 min using a muffle furnace (FP32; Yamato Scientific Co., Ltd., Japan). Commercially available single superphosphate (SSP) was used.

The chemical properties of CPR, PAPR, and SSP were analyzed following the procedure outlined by the Food and Agricultural Materials Inspection Center (FAMIC 2013). The water-soluble P fraction (WP) corresponded to water solubility. The alkaline ammonium citrate soluble P fraction (SP) was defined as alkaline ammonium citrate solubility subtracted by water solubility. The 2% citric acid-soluble P fraction (CP) was defined as 2% citric acid solubility subtracted by alkaline ammonium citrate solubility. The residual P fraction (RP) was the P fraction not soluble in 2% citric acid.

2.3. Experimental design

A total of 10 treatments were established, namely P unfertilized control (CT) and P fertilized at 7.6, 15.3, and 30.5 kg P ha⁻¹ using CPR, PAPR, and SSP, respectively. The application rate was set based on the recommendation of INERA (P1: 15.3 kg P ha⁻¹), and 7.6 (P0.5) and 30.5 (P2) kg P ha⁻¹ corresponded to a half and twice the recommended rate, respectively. Four replicates were set up based on a complete randomized block design. The experimental subplots measured 16 m² (4 m × 4 m) in size with 1-m spacing between neighboring plots. A bund of about 30 cm height was established around each subplot to avoid any contamination by fertilization. For all treatments, 74 kg N ha⁻¹ and 16.6 kg K ha⁻¹ (correspond to 40 kg K₂O ha⁻¹) were applied as urea and potassium chloride (KCl), respectively. Urea, P fertilizers, and KCl were mixed well and applied about one week before seeding by broadcasting. Rice (Oryza sativa) was directly seeded on 8 July 2019 with 20 cm × 20 cm planting space. We used ‘FKR19,’ an improved variety registered by INERA, to adapt to the lowland rice cultivation in West Africa. The plants were harvested around the first week of October 2019. Total biomass, grain yield, and harvest index were measured.

2.4. Ground-water level measurement

Pipes (150 cm) were used for ground-water level (GWL) measurements and holes drilled at 120 cm from the top of the pipe. The pipes were then installed to a depth of 120 cm from the land surface. The section of the pipe above the ground was covered with a cap to protect it from incoming rain. The GWL was manually measured inside the pipe using a measuring stick on Monday, Wednesday, and Friday mornings each week.

2.5. Soil sampling and analysis

Soil was collected at a depth of 0–20 cm in each subplot before rice cultivation to determine initial soil chemical properties. Composite samples were air-dried and passed through a 2 mm mesh sieve to remove stones, roots, and other plant residues. A mixture of soil and distilled water at a ratio of 1:2.5 was used to measure soil pH (H₂O) with a LAQUA pH/ION F-72 (Horiba, Japan) and electrical conductivity (EC) with a COND meter ES-51 (Horiba, Japan). Available P was extracted by Bray-1 and Bray-2 extracting solutions (Bray and Kurtz 1945). The concentration of P in the filtrate was determined by a colorimetric method (Murphy and Riley 1962) using a UV-1800 spectrophotometer (Shimadzu, Japan).

2.6. Data processing

2.6.1. Evaluation of the P fertilization effect

The effect of P fertilization was evaluated based on yield relative to CT and relative agronomic efficiency (RAE) of local P fertilizers for SSP (hereafter RAE). RAE was calculated as:

(1) $RAE\left(\mathrm{\%}\right)=\left(Y_{localPfertilizers}-Y_{CT}\right)/\left(Y_{SSP}-Y_{CT}\right)\ast 100,$(1)

where Y_{local P fertilizers}, Y_CT, and Y_SSP indicate rice grain yields (Mg ha⁻¹) in the corresponding treatments at P2 application level.

2.6.2. Optimal fertilization rate and required price of local P fertilizers

The optimal fertilization rate was considered as the minimum P application rate that produced grain yields comparable to saturated grain yield. Specifically, the comparable yield was defined as Y_SSP at the P2 level subtracted by one standard deviation (SD) of Y_SSP at the P2 level. For example, Y_SSP and SD at the P2 level in Lowland were 3.99 and 0.37, respectively, and thus the comparable yield was calculated as 3.99 − 0.37. The optimal fertilization rate was calculated for each site from the relationship between the application rate of the P fraction and grain yield.

To evaluate the availability of local P fertilizers to local farmers, the required price of local P fertilizers to achieve a cost benefit ratio (CBR) of 1 and 2 under optimal fertilization rate conditions was estimated. CBR was calculated as:

(2) $CBR=\left[\left(\mathrm{\Delta }Y\times P_{rice}\right)-\left(C_{urea}+C_{KCl}\right)\right]/\left(P_{fertilizer}\times A{R}_{opt}\right),$(2)

where ΔY is the expected increase in grain yield under optimal fertilization compared with the expected increase without fertilization (Mg ha^–1), P_rice is the farm gate price of rice (€258.2 Mg^–1 according to Dr. Koide [personal communication]), C_urea and C_KCl are the costs of urea and KCl, respectively (€ ha^–1), P_fertilizer is the required price of P fertilizers (€ 100 kg^–1) to achieve a certain value of CBR, AR_opt is the optimal fertilization rate of three types of fertilizers (kg ha^–1). C_urea and C_KCl were calculated by multiplying the market price of urea and KCl (€45.7 and €160.1 100 kg^–1, respectively) by the application rate (74 kg N ha^–1 and 16.6 kg K ha^–1, respectively). Net benefit (€ ha^–1) was estimated as:

(3) $Netbenefit(\text{euro}h{a}^{-1})=\left(\mathrm{\Delta }Y\times P_{rice}\right)-\left[C_{urea}+C_{KCl}+C_{P-fertilizer}\right]$(3)

where ΔY is the expected increase in grain yield under optimal fertilization compared with the expected increase without fertilization (Mg ha^–1), P_rice is the farm gate price of rice, C_urea, C_KCl, and C_P-fertilizer are the costs of urea, KCl and P fertilizer application, respectively (€ ha^–1). C_P-fertilizer was calculated using the optimal fertilization rate of three types of P fertilizers and the required price of P fertilizers to achieve a CBR = 2.

2.7. Statistical analysis

Statistical analysis was performed using R version 4.0.0 (R Core Team 2020). The difference in the soil chemical properties between the sites were analyzed by Tukey HSD method. The effects of the site, fertilizer type, P application rate, and their interactions were analyzed by the three-way analysis of variance (ANOVA), followed by multiple comparisons using Shaffer’s modified sequentially rejective Bonferroni procedure. The effect size of the source was evaluated by eta squared (η²). The difference in the RAE between the PAPR and CPR was analyzed by paired t-test. Factors controlling grain yield and yield relative to CT were analyzed at each site using a stepwise multiple regression analysis. The application rate of the P fractions (WP, SP, CP, and RP) and water-soluble sulfur were used as explanatory variables. Mean GWL, soil pH (H₂O), EC, soil available P content, and the application rate of P fractions were used as variables to analyze yield relative to CT. Standardized partial regression coefficients were estimated. The selection of explanatory variables was based on the Akaike’s information criterion (AIC). A simple regression analysis was conducted using the explanatory variables that significantly contributed to grain yield to simplify the contribution. The alpha level was set to 0.05.

3. Results

3.1. Chemical properties of fertilizers

Chemical properties of P fertilizers are provided in Table 2. The partial acidulation and calcination methods increased the solubility compared with raw PR. The residual P fraction (RP), which is not soluble in the 2% citric acid, decreased from 68.9% in PR to 25.8% and 32.7% in CPR and PAPR, respectively. SSP included a high water-soluble P fraction (WP) and contained a small quantity of alkaline ammonium citrate-soluble P fraction (SP). PAPR mainly consisted of WP and SP, whereas CPR contained almost equal amounts of SP and 2% citric acid-soluble P fraction (CP). The pH (H₂O) at the solid:liquid extraction of 1:10 for CPR, PAPR, and SSP was 12.3, 2.8, and 3.0, respectively.

Table 2. Chemical properties of phosphate rock and fertilizers used in this study

Solubility		BPR	CPR	PAPR	SSP
(a): Water solubility	% of TP	0.2	2.4	28.9	91.6
(b): Alkaline ammonium citrate solubility	% of TP	2.5	36.8	56.9	96.8
(c): 2% citric acid solubility	% of TP	31.1	74.2	67.3	96.8
P-fractions		BPR	CPR	PAPR	SSP
WP [= (a)]	% of TP	0.2	2.4	28.9	91.6
SP [= (b) – (a)]	% of TP	2.3	34.4	28.0	5.3
CP [= (c) – (b)]	% of TP	28.6	37.4	10.4	0.0
RP [=100 – (c)]	% of TP	68.9	25.8	32.7	3.2
TP	(g P kg^−1)	11.6	6.8	9.2	8.3
pH (H2O)		7.4	12.3	2.8	3.0

BPR: Burkina Faso phosphate rock from the Kodjari deposit, CPR: calcinated phosphate rock, PAPR: partially acidulated phosphate rock; SSP: single super phosphate; WP: water-soluble P fraction; SP: alkaline ammonium citrate-soluble P fraction; CP: 2% citric acid-soluble P fraction; RP: residual P fraction; TP: total P. The pH (H₂O) was measured in the 1:10 solid: liquid extraction.

3.2. Seasonal variation in ground-water level

During the growing season, GWL ranged from +9 cm (in the Lowland on September 16) to −120 cm (in the Middle slope on July 5) (Figure 1). The mean GWL in the Middle slope, Lower slope (1), Lower slope (2), and Lowland were −29.2 ± 32.8, −24.7 ± 18.5, −18.7 ± 23.2, −6.5 ± 21.1 cm (mean ± SD), respectively. GWL in the Lowland was significantly higher than that on the Middle slope and Lower slope (1) (p < 0.01, respectively).

Figure 1. Precipitation (a) and seasonal variation in ground-water level (b). Panel (a) shows the monthly precipitation and mean temperature

3.3. Soil chemical properties

The results of the soil chemical analysis are shown in Table 3. A significant difference (p < 0.01) between sites was observed for soil pH (H₂O) and available P, but not for EC. Soil pH (H₂O) was highest in the Middle slope (7.10), followed by Lower slope (2), Lower slope (1) (7.07 and 6.21, respectively), and Lowland (5.95). Available P ranged from 0.49 mg P kg⁻¹ in the Lower slope (2) to 1.21 mg P kg⁻¹ in the Middle slope.

Table 3. Soil chemical properties before the experiment

		pH (H2O)		EC		Available P (Bray-1)
	n			(S m^−1)		(mg P kg^−1)
Middle slope	45	7.10 ± 0.85	c	4.64 ± 4.20	ns	1.21 ± 0.58	c
Lower slope (1)	48	6.21 ± 0.47	b	4.52 ± 9.81		0.71 ± 0.27	b
Lower slope (2)	48	7.07 ± 0.30	c	4.07 ± 0.88		0.49 ± 0.23	a
Lowland	48	5.95 ± 0.23	a	3.97 ± 1.48		1.12 ± 0.22	c

Mean values with different letters are significantly different (p < 0.05). ns: not significant. Values indicate mean ± standard deviation.

3.4. Grain yield, biomass, and harvest index

Both biomass and harvest index highly correlated with the grain yield (Table A1). Table 4 shows the results of the three-way ANOVA for the grain yield, biomass, and harvest index. The main effects of site, fertilizer type, and P application rate on grain yield were strongly significant (p < 0.001). Grain yield corresponded well to the site position on the slope. Grain yield was lowest in the Middle slope (2.12 Mg ha⁻¹) and highest in the Lowland (2.81 Mg ha⁻¹). Among the fertilizer types, grain yield was significantly higher for SSP (2.78 Mg ha⁻¹), followed by PAPR (2.43 Mg ha⁻¹) and CPR (2.15 Mg ha⁻¹). Higher P application rates significantly improved grain yield until the P2 level. However, a significant interaction between the site and P application rate was observed, which indicated diverse P fertilization effects depending on site-specific conditions. Therefore, grain yield was examined under different P application rates at each site [Figure 2 (a)]. For PAPR and SSP, grain yield increased with increasing application rate. Maximum grain yield was recorded at the P2 level in all sites. In contrast, the trend for CPR differed depending on site. The maximum grain yield was obtained at P2 in the Middle slope and Lowland; however, grain yield peaked at P1 in the Lower slopes (not statistically significant).

Figure 2. Comparison of grain yield (a) and relative agronomic efficiency (RAE) (b) of local P fertilizers between treatments. Error bars in (a) and (b) indicate the standard deviation and standard error, respectively (n = 4). CPR: calcinated phosphate rock; PAPR: partially acidulated phosphate rock; SSP: single superphosphate;*: p> 0.05

Table 4. Effects of the treatments on grain yield, biomass, and harvest index

	Site				Type of fertilizer				P application level
		Grain yield	Biomass	Harvest index		Grain yield	Biomass	Harvest index		Grain yield	Biomass	Harvest index
		(Mg ha^−1)	(Mg ha^−1)	(%)		(Mg ha^−1)	(Mg ha^−1)	(%)		(Mg ha^−1)	(Mg ha^−1)	(%)
	Middle slope	2.12 ± 0.91 a	5.18 ± 1.47	41.6 ± 11.5 a	CPR	2.15 ± 0.92 a	4.71 ± 1.46 a	43.9 ± 11.7	P0	1.37 ± 0.45 a	3.37 ± 0.70 a	37.6 ± 7.3 a
	Lower slope (1)	2.59 ± 1.07 c	5.59 ± 2.07	46.0 ± 9.0 a	PAPR	2.43 ± 1.03 b	4.97 ± 1.50 a	47.3 ± 12.2	P0.5	2.36 ± 0.68 b	4.93 ± 1.08 b	48.1 ± 11.5 b
	Lower slope (2)	2.29 ± 0.76 b	5.35 ± 1.41	42.6 ± 7.4 a	SSP	2.78 ± 1.08 c	5.93 ± 1.81 b	45.0 ± 8.8	P1	2.80 ± 0.87 c	5.80 ± 1.47 c	48.2 ± 10.9 b
	Lowland	2.81 ± 1.26 d	5.16 ± 1.73	51.7 ± 13.3 b					P2	3.29 ± 0.98 d	6.72 ± 1.39 d	47.8 ± 10.7 b
Grain yield					Biomass				Harvest index
Source	df	F-ratio	p-value	η^2	Source	F-ratio	p-value	η^2	Source	F-ratio	p-value	η^2
S	3	10.3	***	0.06	S	1.7	0.177	0.01	S	8.6	***	0.11
F	2	14.4	***	0.06	F	20.9	***	0.10	F	1.8	0.170	0.02
P	3	72.5	***	0.46	P	71.3	***	0.49	P	11.1	***	0.15
S * F	6	1.1	0.367	0.01	S * F	0.5	0.819	0.01	S * F	1.6	0.142	0.04
S * P	9	3.1	**	0.06	S * P	2.3	*	0.05	S * P	1.7	+	0.07
F * P	6	2.0	+	0.02	F * P	2.5	*	0.03	F * P	0.4	0.871	0.01
S * F * P	18	0.7	0.785	0.03	S * F * P	0.7	0.847	0.03	S * F * P	0.8	0.672	0.06

Mean values with different letters are significantly different (p < 0.05). df: degree of freedom; S: site; F: type of fertilizer; P: P application level; +: p < 0.1; *: p < 0.05; **: p < 0.01; ***: p < 0.001; CPR: calcined phosphate rock; PAPR: partially acidulated phosphate rock; SSP: single super phosphate; P0: non-fertilized (control); P0.5: 7.6 kg P ha⁻¹; P1: 15.3 kg P ha⁻¹; P2: 30.5 kg P ha⁻¹.

RAE was low in the Middle slope for CPR and PAPR (62.8% and 62.8%, respectively), and markedly high in the Lowland for CPR and PAPR (97.3% and 99.8%, respectively). In the Lower slopes, PAPR had higher RAE (86.2% and 93.8% in Lower slope (1) and Lower slope (2), respectively), whereas CPR did not (24.3% and 43.1% in the Lower slope (1) and Lower slope (2), respectively).

Both biomass and harvest index were highly correlated with grain yield (Table A1; p < 0.01). Site did not affect biomass but significantly changed the harvest index (Table 4; p < 0.001). In contrast, fertilizer type significantly affected biomass (p < 0.001) but not the harvest index. The P application rate significantly affected both the biomass and harvest index (p < 0.001, respectively).

3.5. Factors controlling grain yield and fertilization effect

Table 5 presents the results of multiple regression analysis. Regression equations were significant at all sites (p < 0.05 in the Middle slope and p < 0.01 for all other sites). The adjusted coefficient of determination (R²) was lowest in the Middle slope (R² = 0.45) and highest in the Lower slope (2) (R² = 0.81). WP persistently and strongly contributed to grain yield (p < 0.01). However, the contribution of SP was site-dependent. SP was rejected in the Middle slope, but weakly contributed to the Lower slope (1) (p < 0.1), and significantly contributed to the Lower slope (2) and Lowland (p < 0.05 and p < 0.01, respectively). Mean GWL and pH (H₂O) negatively contributed, whereas, available P (AP), WP, and SP positively contributed to yield relative to CT. A regression equation with high R² (0.85, p < 0.01) was obtained; in this equation, the variance inflation factor of variables was >1.63, suggesting no multicollinearity problem in the multiple regression analysis.

Table 5. Factors controlling grain yield and yield relative to the control

Site	Equation		p-value	R^2	SE	AIC
Middle slope	Grain yield =	0.882 WP** + 0.338 CP + 1.665**	< 0.05	0.452	0.46	−12.23
Lower slope (1)		0.765 WP** + 0.742 SP* – 0.594 CP + 2.031**	< 0.01	0.735	0.441	−13.48
Lower slope (2)		0.814 WP** + 0.820 SP* – 0.427 CP + 1.763**	< 0.01	0.806	0.275	−22.94
Lowland		0.731 WP** + 0.698 SP** + 2.053**	< 0.01	0.705	0.471	−12.64
All sites	Relative yield for CT =	(b): −0.506 GWL** −0.352 pH** + 0.170 AP** + 0.538 WP** + 0.156 SP** + 543.9**	< 0.01	0.852	29.03	247.96

Equations were obtained by stepwise multiple regression analysis. WP: water-soluble P fraction; SP: alkaline ammonium citrate-soluble P fraction; CP: 2% citric acid-soluble P fraction; GWL: mean groundwater level; AP: soil available P; SE: standard error; AIC: Akaike’s information criterion; +: p < 0.1; *: p < 0.05; **: p < 0.01; ***: p < 0.001.

3.6. Determination of optimal fertilization rate and expected increase in grain yield

Simple regression analysis showed an evident relationship between the application rate of P fractions and grain yield (Figure 3). The optimal application rate of the P fractions was 14.4 kg P ha⁻¹ of WP in the Middle slope, and 15.9, 15.2, and 11.0 kg P ha⁻¹ of WP + SP in the Lower slope (1), Lower slope (2), and Lowland, respectively. The optimal fertilization rate for each fertilizer, expected increase in grain yield, the required price of local P fertilizers, and net benefit at CBR = 2 are summarized in Table 6. Under optimal fertilization, grain yield was expected to increase by 1.08–2.00 Mg ha⁻¹. The required price to achieve CBR = 1 or 2 were lowest in the Middle slope and highest in the Lowland. Net benefit was estimated to range from 59.3 in the Middle slope to 177.8 in the Lowland.

Figure 3. Relationship between the application rate of P fractions and grain yield. The Sum of effective P fractions that were significantly selected in Table 5 was employed as X-axis. The horizontal and vertical lines indicate the grain yield of SSP treatment with an application rate of 30.5 kg P ha−1 subtracted by its standard deviation and optimal fertilization rate. Error bars are standard deviation (n = 4). The gray areas represent the 95% confidence interval. CPR: calcinated phosphate rock; PAPR: partially acidulated phosphate rock; SSP: single superphosphate; WP: water-soluble P fraction; SP; alkaline ammonium citratesoluble P fraction

Table 6. Optimal P fertilization rate and expected increase in grain yield

Site	Optimal rate	Optimal fertilization rate			Expected grain yield		Expected increase	Required price to achieve CBR = 1			Required price to achieve CBR = 2			Net benefit
		CPR	PAPR	SSP	Without P	Optimal P		CPR	PAPR	SSP	CPR	PAPR	SSP
	(kg P ha^−1)	(kg P ha^−1)			(Mg ha^−1)		(Mg ha^−1)	(€ 100 kg fertilizer^−1)			(€ 100 kg fertilizer^−1)			(€ ha^−1)
Middle slope	14.4 as WP	600	50.0	15.7	1.78	2.86	1.08	1.30	21.6	61.7	0.70	10.8	30.9	59.3
Lower slope (1)	15.9 as WP ＋ SP	43.2	27.9	16.4	1.55	3.51	1.96	54.3	114.4	176.2	27.2	57.2	88.1	172.9
Lower slope (2)	15.2 as WP ＋ SP	41.3	26.7	15.7	1.57	2.98	1.41	33.3	70.0	108.1	16.6	35.0	54.1	101.9
Lowland	11.0 as WP ＋ SP	29.9	19.3	11.4	1.62	3.62	2.00	80.7	170.2	262.3	40.4	85.1	131.2	177.8

Optimal fertilization rate is defined as the minimum P application rate that produced grain yields comparable to those under saturated ones. Values are estimated in Figure 3. CPR: calcinated phosphate rock; PAPR: partially acidulated phosphate rock; SSP: single super phosphate; WP: water-soluble P fraction; SP: alkaline ammonium citrate-soluble P fraction; CBR: cost benefit ratio. €1 = 655.957 FCFA (fixed rate).

4. Discussion

4.1. Effects of PAPR and CPR on grain yield under different GWL

Grain yield was significantly affected by the site, fertilizer type, and application rate (Table 4). Based on η², P application rate explained 46% of the variance in grain yield. Grain yield increased by 168%, 204%, and 251% under P0.5, P1, and P2, respectively. Thus, the limited plant growth due to low P in soils in this region (Gebrekidan and Seyoum 2006) can be successfully mitigated with P fertilizer application to significantly improve rice grain yield.

However, the effects of P fertilization differed across sites (Figure 2). RAE variation closely reflected the contribution of the P fractions [Figure 2 (b) and Table 5]. In the Middle slope with low GWL (Figure 1), only WP contributed to grain yield; consequently, PAPR and CPR had a disadvantage over SSP with high WP. In comparison, in the Lowland with high GWL (Figure 1), WP and SP had an almost equal contribution (Table 5), with both PAPR and CPR performing well (Figure 2). At the other two sites with middle GWL, both WP and SP contributed to grain yield, as observed in the Lowland (Table 5). However, more WP + SP was required to reach comparable yields (15.2–15.9 kg P ha⁻¹) than that required for the Lowland (11.0 kg P ha⁻¹) (Figure 3). Therefore, PAPR with higher total P content (9.2 g P kg⁻¹) and WP + SP (56.9% of total P) showed an advantage over CPR (6.8 g P kg⁻¹ and 36.8% of the total P, respectively). Thus, PAPR is useful in fields with a mean GWL > −24.7 cm, and both CPR and PAPR are effective in fields with mean GWL of – 6.5 cm.

Site and fertilizer type also significantly affected grain yield (Table 4). Site mainly reflected differences in GWL in the current study (Figure 1), while fertilizer type reflected differences in P solubility (Table 2). The post-hoc test showed that GWL contributed to grain yield by increasing the harvest index. Emam et al. (2014) also reported that the harvest index significantly decreased in fields subjected to water stress compared with the well-watered conditions in Egypt. In contrast, fertilizer type (i.e., only SSP) contributed to grain yield by increasing biomass. Song et al. (2019) reported that application of SSP with high WP significantly enhanced root biomass and structure, thereby allowing plants to access and solubilize P sources in the soil and increase above-ground biomass. Our results demonstrate that WP is essential for enhancement of biomass compared with SP and CP. CPR showed a low performance at the P2 application level in the two sites on a lower slope. We were not able to explain this phenomenon, which requires further research.

4.2. Effect of GWL and soil pH on fertilization effect

Multiple regression equations for yield relative to CT revealed that soil water condition as well as soil chemical properties affected the fertilization effect (Table 5; Figure A1). In particular, the mean GWL and soil pH (H₂O) had a large contribution. Multicollinearity was not found in the multiple regression analysis. Therefore, it could be inferred that the GWL and the soil pH independently affected the yield relative to CT. Moreover, the high R² (0.85) indicated that the variables selected in this equation well explained the variability at the four study sites and the small effect of the other soil environmental factors.

Soil water condition is generally related to the accessibility and solubility of P fertilizer in the soil. Consistent with previous studies (Fukuda et al. 2013; Balasubramanian et al. 2007), the effect of P fertilization was more substantial at the site with high GWL, because it increased solubilization, transportation, and subsequent availability of P in the soil (Hömberg and Matzner 2018; King et al. 2015).

The mechanisms of the effects of pH on P availability are diverse (Haynes 1982). Robinson and Syers (1990) evaluated how pH affected the solubilizing rate of PR using the incubation method, and showed that P solubility in water increased under low-pH conditions because dissolution of the apatite relies on the net supply of protons. However, lowering soil pH also increases the P fixation and Al toxicity (Margenot et al. 2016). Haynes (1982) has shown that the availability of P in soil generally increases as pH rises from 4.0 to 7.0 because of the stimulation of mineralization of soil organic phosphorus. Carbon content and exchangeable Al were low in the current study (Table 1), while the minimum pH was relatively high (Table 3). Therefore, Al toxicity was not a problem, and the direct effect of low soil pH on P solubility would be greater than the other effects.

4.3. Optimal fertilization and expansion of rice cultivation area

The availability and profitability of fertilizers influence local farmers’ decisions to increase fertilizer inputs (Tsujimoto et al. 2019; Yanggen et al. 1998). Although local P fertilizers were inferior to SSP in terms of solubility and optimal application rate (Table 6), their availability and profitability would be superior to that of imported commercial fertilizers. The use of local PR deposits could reduce transportation costs and might significantly reduce the price of P fertilizers in SSA (Kelly 2006). The reduced price of local P fertilizers would increase cost benefits and profitability. Buah and Mwinkaara (2009) emphasized the importance of economic analyses to calculate the net benefit to farmers.

Guèdègbé and Doukkali (2018) showed that the local price of P fertilizer in the African market considerably exceeded that in the international market because of high transportation costs, which represented 43% of the local P fertilizer price. The authors estimated that, by increasing domestic fertilizer production and by reducing transportation costs, it would be possible to reduce the price of P fertilizers from €74.6 to €42.5. The required price of CPR and PAPR to achieve CBR = 1 was comparable or higher than €42.5 in fields with mean GWL > −24.7 cm (Lower slope (1), Lower slope (2), and Lowland). Therefore, the net income of farmers could be enhanced by using local P fertilizers. Moreover, the required price of PAPR to achieve CBR = 2 could be realized.

In this region of the Central Plateau of Burkina Faso, most areas of rice cultivation are distributed in lowland riverine sites. Our result indicated that PAPR could help expand rice cultivation to the lower slopes with GWL > −24.7 cm (~135 m from the river bottom in the line transect), while SSP could even facilitate rice cultivation on middle slopes with GWL > – 29.2 (~190 m from the river bottom in the line transect).

5. Conclusions

This study demonstrated the effectiveness of two types of local P fertilizers made from low-grade PR products in Burkina Faso. Only WP contributed to grain yield in fields with low GWL, but showed an equivalent contribution in fields with high GWL, WP, and SP. Therefore, PAPR with high water-solubility has an advantage over CPR in fields with low GWL. Both PAPR and CPR performed well at sites with high GWL. The optimal application rate was determined to be 14.4 kg P ha⁻¹ as WP in the field with low GWL (mean −29.2 cm), 15.2–15.9 kg P ha⁻¹ as WP + SP in the field with middle GWL (mean −24.7 to −18.7 cm), and 11.1 kg P ha⁻¹ as WP + SP in the field with high GWL (mean −6.5 cm). Optimal fertilizer type and application rate differed according to the water condition. Our results provided the basic information toward expansion of the rice cultivation area into the surrounding land on the Central Plateau of Burkina Faso.

Disclosure of potential conflicts of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

The authors thank local farmers and technical staff members (Mr. Mohamed Ouedraogo and Mr. Kafando Serge Placide) for supporting our field trials, as well as JIRCAS’s Soil Team for soil chemical analysis. Soil sample was imported with plant phytosanitary certificate of Ministry of Agriculture, Forestry and Fisheries.

Additional information

Funding

This work was financially supported by the Science and Technology Research Partnership for Sustainable Development (SATREPS), Japan Science and Technology Agency (JST)/Japan International Cooperation Agency (JICA) (Project on establishment of the model for fertilizing cultivation promotion using Burkina Faso phosphate rock).

’

Table A1 Pearson’s correlation matrix of grain yield, biomass, harvest index, and yield components

	Grain yield			Biomass			Harvest index			Number of panicle			Number of grain			100 grain weight
	(Mg ha^−1)		n	(Mg ha^−1)		n	(%)		n	(panicle ha^−1)		n	(grain panicle^−1)		n	(g)	n
Grain yield	1
Biomass	0.809	***	171	1
Harvest index	0.598	***	171	0.049	0.528	171	1
Number of panicle	0.302	***	189	0.396	***	171	0.080	0.298	171	1
Number of grain	0.778	***	186	0.544	***	171	0.526	***	171	−0.221	***	186	1
100 grain weight	0.285	***	189	0.296	***	171	0.128	0.096	171	0.168	*	189	0.117	0.112	186	1

*: p < 0.05; **: p < 0.01; ***: p < 0.001

Figure A1. Effect of ground-water level (a) and soil pH (H2O) (b) on the yield relative to the control. Error bars indicate standard deviation (n = 4). Gray shading corresponds to the points of other fertilizers. CPR: calcined phosphate rock; PAPR: partially acidulated phosphate rock; SSP: single superphosphate; P0: nonfertilized (control). P0.5: 7.6 kg P ha-1; P1: 15.3 kg P ha-1; P2: 30.5 kg P ha−1