Understanding yield-limiting factors for sorghum in semi-arid sub-Saharan Africa: beyond soil nutrient deficiency

ABSTRACT

Even in semi-arid regions of sub-Saharan Africa, low crop yields are often attributed to a lack of soil nutrients and limited use of chemical fertilizers. However, in areas where soils with low effective soil depth are common, water availability may be a more significant factor that limits crop production than a lack of soil nutrients. In Sudan Savanna, the effective soil depths of three dominant soils, namely Pisoplinthic Petric Plinthosols (PT-pt.px), Petric Plinthosols (PT-pt), and Ferric Lixisols (LX-fr), are < 30, 30–50, and ~ 100 cm, respectively. Consequently, a lack of soil water is anticipated for both Plinthosols types. This study investigated the limiting factor in sorghum cultivation for each dominant soil in Sudan Savanna by creating plots with and without irrigation and fertilization. The results indicated that neither rainfall nor irrigation significantly affected sorghum yield under unfertilized conditions in all soil types, suggesting that even in very shallow PT-pt.px, a lack of soil water was not the limiting factor under unfertilized conditions. However, under fertilized conditions, reduced rainfall significantly decreased sorghum yield in PT-pt and PT-pt.px. In conclusion, in addition to the lack of soil nutrients being the limiting factor in sorghum cultivation under unfertilized conditions, regardless of soil type, lack of soil water becomes a limiting factor in soils with an effective soil depth of < 50 cm (e.g., Plinthosols) under fertilized conditions. These findings underscore the importance of knowing the effective soil depth when developing efficient fertilizer application methods to achieve food security in semi-arid regions of sub-Saharan Africa.

KEY WORDS:

• Effective soil depth
• limiting factor
• Plinthosols
• sorghum
• sub-Saharan Africa

1. Introduction

The population of sub-Saharan Africa (SSA) has seen a considerable increase, from 0.4 billion in 1980 to 1.2 billion in 2020, and is estimated to increase by another billion by 2050. This means that one in four people globally will reside in SSA (World Bank Health Nutrition and Population Statistics 2023). Considerably, enhancing agricultural productivity in SSA is vital to meeting the escalating food demand. However, the growth rate of crop productivity is incongruent with the population increase in SSA. From 1980 to 2020, maize and sorghum productivity (yield per unit area) increased by only 46% and 20%, respectively (FAOSTAT 2023). This low productivity is primarily attributed to a lack of soil nutrients (Liu et al. 2010; Sanchez 2002) and the limited use of chemical fertilizers (Morris et al. 2007; Mueller et al. 2012). The importance of chemical fertilizer application for improving crop productivity (mainly maize) and maintaining soil organic matter in SSA has been emphasized (Vanlauwe and Giller 2006). Contextually, the Abuja Declaration on Fertilizer for the African Green Revolution was adopted to promote the use of fertilizers in SSA.

While the lack of soil nutrients and limited use of chemical fertilizers are often considered major constraints to crop production in SSA, these factors may oversimplify the low crop production issues and their solutions at the field level. Previous studies (Kihara et al. 2016; Vanlauwe et al. 2011; Zingore et al. 2007) have reported that the large number of fields in SSA (more than half in Kihara et al. (2016)) show little or no response to fertilizer application. Therefore, simple fertilization is unlikely a universal solution for low productivity at the field level in SSA. Kihara et al. (2016) concluded that this low response to fertilization was due to low soil pH and a lack of secondary/micronutrients, and potentially a lack of soil water. Identifying location-specific limiting factors in crop production and developing tailor-made technologies to address these factors is crucial to improving low productivity (Giller et al. 2011).

At the regional level, studies have suggested that the lack of soil water could be a limiting factor in crop production in drylands (Reynolds et al. 2015; Waddington et al. 2010). Given the correlation between rainfall amount and net primary production (Liu et al. 2021; Pan et al. 2015), it is undeniable that a lack of soil moisture limits crop production in drylands on a regional scale. However, farmers have adapted their practices over time to suit the environmental conditions of drylands, so a lack of soil water may not limit crop yields at the field level in the current cropping systems. From the on-farm trials in southern Zimbabwe, Twomlow et al. (2010) have reported that, even in semi-arid regions, a lack of soil nutrients (especially nitrogen (N)), rather than a lack of soil moisture, is the limiting factor in crop production. However, these trials were conducted in areas dominated by Chromic Luvisols with deep effective soil depth (EU 2013), so the findings may not apply to other SSA areas where Plinthosols or Leptosols, with shallow effective soil depth, are predominant. Simulations of maize yield potential in soils with different effective soil depths under rain-fed conditions suggest that soil water storage capacity in the root zone could be particularly important for yield potential (Guilpart et al. 2017). This implies that the lack of soil moisture can be a major limiting factor in the shallow soils widely distributed in SSA. According to Eswaran et al. (1997), soils with an effective soil depth of < 25 and 25–50 cm account for 33% and 23% of total land in Africa. Therefore, to develop tailor-made soil management practices that can be implemented by local farmers at the field level, limiting factors in crop production should be evaluated carefully in shallow soils, especially in semi-arid climates.

Sudan Savanna, a semi-arid region with 600–900 mm annual rainfall in West Africa, stands as the study site, being the largest sorghum-producing area in Africa. Sorghum ranks as the second most important upland crop in Africa after maize (FAOSTAT 2023), making Sudan Savanna crucial for the continent’s food security. In this region, soils with very shallow effective depth are widespread across the shield on West African Craton (Ikazaki et al. 2018b). Their study revealed that Pisoplinthic Petric Plinthosols (PT-pt.px; effective soil depth <30 cm) dominate the upper slope of rolling hills, Petric Plinthosols (PT-pt; effective soil depth 30–50 cm) are prevalent in the middle slope, and Ferric Lixisols (LX-fr; effective soil depth ~ 100 cm) are found in the lower to toe slope. Based on soil surveys from 61 representative soil pits in watersheds (200 km²), it was determined that PT-pt.px, PT-px, and Lixisols occupy 39%, 23%, and 31% of the total land area (unpublished data). The effective soil depth is constrained by the petroplinthite, which is an impermeable, consolidated soil layer. The sorghum yield without fertilizer application is proportional to the effective soil depth, in the following order: LX-fr (1.1 Mg ha⁻¹) > PT-pt (0.6 Mg ha⁻¹) > PT-pt.px (0.2 Mg ha⁻¹) (Ikazaki et al. 2018b). However, the limiting factor in sorghum production, whether a lack of soil nutrients or soil water, remains unclear in Sudan Savanna.

To determine if the limiting factor in crop production is solely a lack of soil nutrients, even in the shallow soils of semi-arid regions of SSA, this study explored the factors restricting sorghum production under unfertilized and fertilized conditions across three dominant soils in Sudan Savanna (PT-pt.px, PT-px, and LX-fr).

2. Materials and methods

2.1. Site description

A two-year field experiment was carried out at Saria station (12°16’ N, 2°09’ W; 300 m above sea level) of the Institute of the Environment and Agricultural Research (INERA), Burkina Faso. The climate was BSh according to the Köppen system. The mean annual rainfall in the area was 821 mm (1978–2015), with a mean annual temperature of 28°C (2013–2015). Rainfall in this region follows a unimodal pattern, with the dry season spanning from November to April and the rainy season from May to October. The mean annual potential evaporation ranges from 1,700 to 2,000 mm, resulting in 0.40–0.47 aridity index (Ouattara et al. 2006). Three experimental fields representing the dominant soils (LX-fr, PT-pt, and PT-pt.px; Supplementary Figure S1) were selected. Soil physical and chemical properties in the soil layers at 0–10, 10–25, 25–50, 50–75, and 75–100 cm deep are shown in Tables 1 and 2. Since the consolidated petroplinthite layer, which cannot be taken for analysis, started at a depth of 50 cm in PT-pt and 25 cm in PT-pt.px, the characteristic values for the soil layers deeper than 50 and 25 cm in PT-pt, and PT-pt.px are blank. Tables 1 and 2 were derived from a weighted average of each characteristic value reported in Ikazaki et al. (2018a). Additionally, highlighting the significant presence of coarse fragments (iron nodules >2 mm diameter) in PT-pt.px and PT-pt is crucial, while LX-fr exhibited a lower content (Table 1).

Table 1. Physical properties of the soils studied.

Layer	Coarse fragment ≥ 2 mm	Particle size distribution				Bulk density	Volumetric water content
		Coarse Sand	Fine Sand	Silt	Clay		pF 1.6	pF 2.0	pF 2.5	pF 3.0	pF 3.2	pF 3.8	pF 4.2
		0.2–2 mm	0.02–0.2 mm	0.002–0.02 mm	< 0.002 mm
(cm)	(% weight)	- - - - - - - - - (% weight of fine earth <2 mm) - - - - - - - - -				(Mg m^−3)	(%)
Ferric Lixisols (LX-fr)
0–10	1.5	39.8	45.1	6.5	8.7	1.6	22.5	17.3	13.1	9.7	8.6	5.6	4.4
10–25	1.2	35.5	37.3	6.3	20.9	1.7	23.2	20.1	17.2	14.4	13.3	10.8	9.3
25–50	2.3	30.6	31.5	5.5	32.3	1.6	25.9	23.6	21.3	18.3	17.2	15.0	13.6
50–75	5.5	29.7	31.7	5.6	33.0	1.6	26.6	24.3	21.8	18.5	17.4	15.3	14.0
75–100	25.1	29.3	33.7	6.4	30.6	1.7	25.6	23.5	20.6	17.1	16.0	14.4	13.0
Petric Plinthosols (PT-px)
0–10	27.8	43.1	41.9	4.9	10.1	1.7	20.7	15.0	10.1	8.2	7.5	4.4	4.0
10–25	39.9	40.2	36.5	4.9	18.4	1.6	20.2	15.9	12.1	10.3	9.7	7.0	6.3
25–50	58.0	35.1	31.4	5.6	27.9	1.7	18.7	16.5	14.5	13.1	12.6	10.9	10.0
Pisoplinthic Petric Plinthosols (PT-pt.px)
0–10	54.1	49.9	40.5	5.2	4.4	1.8	16.3	13.6	11.1	8.9	8.1	5.9	4.9
10–25	58.4	50.2	39.3	5.0	5.4	1.9	16.0	14.4	12.2	9.7	8.9	6.6	5.4

The characteristic values were obtained by the weighted average of each value reported in Ikazaki et al. (2018b).

Table 2. Chemical properties of the soils studied.

							Exchangeable bases					CEC
Layer			EC	OC	TN	C/N^†	Ca^2+	Mg^2+	K^+	Na^+	Al^3+	per fine earth	per clay	EBS^‡	Bray-1 P
(cm)	pH (H2O, 1:5)	pH (KCl, 1:5)	(mS m^−1)	(g kg^−1)			- - - - - - - - - - - - - - - - (cmolc kg^−1) - - - - - - - - - - - - - -							(%)	(mgP kg^−1)
Ferric Lixisols (LX-fr)
0–10	4.9	4.0	4.3	2.2	0.3	8.8	0.5	0.2	0.1	0.0	0.1	1.7	20.1	85.6	3.7
10–25	5.3	4.3	4.2	2.4	0.3	8.3	1.4	0.6	0.0	0.0	0.0	2.9	15.2	95.6	1.2
25–50	5.2	4.3	2.9	2.4	0.3	7.9	1.7	0.8	0.1	0.0	0.0	4.7	14.4	98.6	0.4
50–75	5.3	4.3	2.5	2.2	0.3	7.6	1.6	0.8	0.1	0.0	0.1	4.5	13.5	97.8	0.5
75–100	5.5	4.3	2.4	1.5	0.2	6.6	1.4	0.8	0.1	0.0	0.1	2.7	8.9	96.8	0.6
Petric Plinthosols (PT-px)
0–10	5.3	4.2	2.8	3.1	0.3	10.5	0.7	0.2	0.0	0.0	0.1	1.2	13.3	90.8	2.9
10–25	5.8	4.6	3.3	3.2	0.3	10.3	1.4	0.3	0.0	0.0	0.0	1.6	8.8	98.8	1.5
25–50	6.2	5.0	3.4	2.9	0.3	9.1	2.1	0.5	0.0	0.0	0.0	3.6	12.1	100.0	0.7
Pisoplinthic Petric Plinthosols (PT-pt.px)
0–10	5.5	4.5	2.8	2.5	0.2	10.3	0.5	0.1	0.0	0.0	0.1	0.8	18.7	88.6	2.1
10–25	5.3	4.3	2.6	1.9	0.2	10.1	0.3	0.1	0.0	0.0	0.1	0.7	12.8	77.1	1.6

EC: electrical conductivity; OC: organic carbon content; TN: total nitrogen content; CEC: cation exchange capacity; EBS: effective base saturation.

†Ratio of OC to TN; ^‡ratio of exchangeable (Ca + Mg + K + Na) to exchangeable (Ca + Mg + K + Na + Al) defined in IUSS Working Group WRB (2015).

The characteristic values were obtained by the weighted average of each value reported in Ikazaki et al. (2018b).

2.2. Experimental settings

In 2016, 25 plots, measuring 6.0 × 4.0 m², were established in each field. Five treatments outlined in Table 3 were assigned with five replicates, considering two factors: irrigation and fertilization (N application rate). In Treatment 2 (Table 3), hand irrigation using well water was applied to prevent water stress until harvest time. Specifically, after 4 rainless days, on the 5^th day, 20-mm irrigation was applied in 2016. In 2017, after 3 rainless days, on the 4^th day, a 10-mm irrigation was applied. The irrigation method was altered in 2017 as the 20-mm irrigation sometimes led to waterlogging, especially in LX-fr. The total irrigation amounted to 120 mm in 2016 and 80 mm in 2017.

Table 3. Treatments other than soil types.

Treatment number	Irrigation	Nitrogen application rate (kg ha^−1)
1 (control)	No	0
2	Yes	0
3	No	37
4	No	74
5	No	111

Following the long-standing blanket recommendation in Burkina Faso (IRAT 1978), the N application rate was set at 37 kg N ha^‒1 (along with 23 kg P₂O₅ ha^‒1 and 14 kg K₂O ha^‒1), double (74 kg N ha^‒1), and triple (111 kg N ha^‒1). Notably, only N was evaluated, as it is considered the most crucial limiting factor for sorghum production in SSA (Tonitto and Ricker-Gilbert 2016). The planting density was 3.1 hills m^‒2. The improved sorghum varieties, namely Kapelga, were utilized.

Each plot was encircled by 8 cm high earthen mounds to prevent water and fertilizer runoffs from neighboring plots. Sorghum was sown at the end of June in both years. Two weeks after sowing (WAS), the number of plants in each hill was reduced to three. In the fertilized plots, 100 kg ha^‒1 of compound fertilizer (14% N, 23% P₂O₅, and 14% K₂O) was applied at 2 WAS. The variations in the N application rate were accommodated by adjusting the urea quantity (46% N): a total of 50, 130, and 211 kg ha^‒1 urea was applied for 37, 74, and 111 kg N ha^‒1 treatment. According to INERA’s recommended method, half of the urea was applied at 4 WAS and the remaining half at 6 WAS as a top dressing. Weed control was conducted using a hand hoe 2–3 times per cropping season. Sorghum was harvested at the end of October in 2016 and at the beginning of November in 2017.

2.3. Measurements

2.3.1. Weather

Meteorological data were collected at 10-min intervals using an automated weather station. The station was equipped with temperature and relative humidity sensor (HygroVUE™5, Campbell Scientific, Logan, UT, U.S.A.), rain gauge (TE525MM-L, Campbell Scientific, Logan, UT, U.S.A.), and albedo meter (CHF-SRA01, Hukseflux, Delft, Netherland).

2.3.2. Soil

The volumetric water content (VWC) was measured using 30-cm-long TDR probes (CS616; Campbell Scientific, Logan, UT, U.S.A.) in a control plot of each field starting on 25 July 2016. These probes were installed diagonally at various depths: 0–10, 10–25, 25–50, and 50–75 cm in LX-fr; 0–10, 10–25, and 25–50 cm in PT-pt; and 0–10 and 10–25 cm in PT-pt.px. This was due to the consolidated petroplinthite layer starting at 50 and 25 cm depth in PT-pt and PT-pt.px. Accounting for temperature effects on probes, soil temperature was monitored in each soil layer using thermistor probes (108; Campbell Scientific, Logan, UT, U.S.A.) in accordance with Campbell Scientific (2020). The estimated VWC was then calibrated in situ using the gravimetric method as in Campbell Scientific (2020).

The soil pF (soil matrix potential) was determined through the following steps: first, the soil moisture characteristic curve was constructed for each soil layer (Table 1) by fitting the van Genuchten model using SWRC Fit (Seki, Toride, and van Genuchten 2023). The model demonstrated a strong fit (R² >0.99, p < 0.001 for all soil layers). Second, VWC was converted to the matric potential using the fitted model and then to the soil pF.

Using the established relationship between total available water (TAW in mm) and readily available water (RAW in mm) defined by FAO (1998), we proceeded to calculate the VWC at the lower limit of RAW (θ_LL in %) in the following manner:

$\mathit{TAW}=\mathit{\hspace{0.17em}}\left({\theta }_{\mathit{FC}}-{\theta }_{\mathit{PWP}}\right)/100\mathit{\hspace{0.17em}}\times \mathit{Z}$

$\mathit{RAW}=\mathit{\rho }\times \mathit{TAW}=\mathit{\hspace{0.17em}}0.55\mathit{\hspace{0.17em}}\times \mathit{\hspace{0.17em}}\left({\theta }_{\mathit{FC}}-{\theta }_{\mathit{PWP}}\right)/100\mathit{\hspace{0.17em}}\times \mathit{Z}=\mathit{\hspace{0.17em}}\left({\theta }_{\mathit{FC}}-{\theta }_{\mathit{LL}}\right)/100\mathit{\hspace{0.17em}}\times \mathit{Z}$

therefore, ${\theta }_{\mathit{LL}}=\mathit{\hspace{0.17em}}0.45\mathit{\hspace{0.17em}}\times {\theta }_{\mathit{FC}}+\mathit{\hspace{0.17em}}0.55\mathit{\hspace{0.17em}}\times {\theta }_{\mathit{PWP}}$

where θ_FC and θ_PWP represent VWC values at field capacity and permanent wilting point (%). Z stands for the thickness of soil layer (mm), and ρ denotes the average fraction of TAW that can be depleted from the root zone before moisture stress occurs (0.55 for grain sorghum). The pF values for field capacity and permanent wilting point were assumed to be 2.5 and 4.2, respectively. If VWC of a soil layer falls below the lower limit of RAW (θ_LL), it indicates that the soil layer lacks RAW for sorghum. The pF values at θ_LL were ~ 3.2–3.3. The available water for sorghum was calculated by subtracting θ_LL from VWC at that specific time in each soil layer. Negative values were considered zero in this calculation. Then, the total amount of water readily available for sorghum within the effective soil depth was calculated in each soil type.

2.3.3. Sorghum

During the harvest, all sorghum plants, except those at the plot borders, were harvested and subjected to yield surveys. The panicles were divided into rachizes and grains, and the moisture content of the grains was measured using a portable grain moisture tester (MT-16; Agratronix, Streetsboro, OH, U.S.A.). The grain yield was expressed in terms of dry weight.

2.4. Statistical analysis

Statistical analysis was performed using SPSS ver. 21 (IBM, Armonk, NY, U.S.A.). The effects of rainfall amount, irrigation (comparing treatments #1 and #2 over 2 years), and N application rate (comparing treatments #1 and #3–5 over 2 years) on sorghum yield were assessed using a generalized linear model (GLM). Rainfall, irrigation, and N application rate were fixed factors, and replication (block effect) was a random factor because a split-split design was employed (Montgomery 2012 for details). The post hoc Tukey Honestly Significant Difference test was conducted for statistical analysis among groups. Significance was defined as p < 0.05.

3. Results

3.1. Weather

During the growing period (June 20–October 20) in 2016, the total rainfall was 632 mm, the average daily minimum and maximum temperatures were 22.4°C and 32.3°C, and the mean solar radiation was 21.8 MJ m⁻² day⁻¹. In 2017, these values were 523 mm for total rainfall, 22.2°C and 33.0°C for average daily minimum and maximum temperatures, and 22.6 MJ m⁻² day⁻¹ for mean solar radiation. When compared to the average rainfall during the same period over the last 38 years (1978–2015), 663 ± 110 mm (mean ± standard deviation), the rainfall in 2016 was close to average, while it was lower than average in 2017.

3.2. Soil water

VWC values across all soil layers were typically higher in 2016 than 2017 (Figure 1), reflecting the disparity in rainfall. In 2016, despite average rainfall amount and rainfall patterns, VWC values dropped below θ_LL ~ 100 days after sowing (DAS) (Figure 1). This pattern was similar to 2017, characterized by lower rainfall, except for VWC at 10–25 cm layer of LX-fr. Despite the two soil layers in PT-pt.px having only 3%–5% clay and an abundance of coarse fragments, leading to a low water holding capacity, VWC values in these layers consistently surpassed θ_LL until 100 DAS. This pattern was also noted in LX-fr and PT-pt. The total amount of water readily available for sorghum generally correlated with the effective soil depth (LX-fr > PT-pt > PT-pt.px), except in 2016 with average rainfall, where PT-pt showed similar trends to LX-fr (Figure 2). However, the soil pF, especially in 10–25 cm layer, showed an inverse relationship with the effective soil depth in both years; it was higher (indicating drier conditions) in LX-fr than in PT-pt and PT-pt.px (Figure 3).

Figure 1. Volumetric water content (VWC) in each soil layer of control plot.

Black and gray lines represent VWC in 2016 and 2017; LX-fr, Ferric Lixisols; PT-pt, Petric Plinthosols; PT-pt.px, Pisoplinthic Petric Plinthosols; dashed, dotted, and chain lines represent VWC at pF values of 1.6, 2.5, and 4.2; solid line represents VWC at lower limit of RAW (θ_LL), calculated according to FAO (1998).

Figure 2. Amount of water readily available for sorghum within the effective soil depth for each dominant soil type.

LX-fr, Ferric Lixisols; PT-pt, Petric Plinthosols; PT-pt.px, Pisoplinthic Petric Plinthosols

Figure 3. Soil pF for each dominant soil type.

LX-fr, Ferric Lixisols. PT-pt, Petric Plinthosols; PT-pt.px, Pisoplinthic Petric Plinthosols. Volumetric water contents shown in Figure 1 were converted to soil pF.

3.3. Effects of irrigation and fertilization on sorghum production

The impact of rainfall and irrigation on sorghum yield was not statistically significant across all soil types under unfertilized conditions (Table 4). Conversely, fertilization exhibited a notable increase in sorghum yield across all soil types (Table 5). Yield saturation was observed at 74 kg N ha^‒1 in LX-fr and PT-pt, whereas it occurred at 37 kg N ha^‒1 in PT-pt.px. When considering fertilized conditions, reduced rainfall had a significant negative effect on sorghum yield in PT-pt and PT-pt.px, but not in LX-fr. Moreover, an interaction indicating a synergistic effect between rainfall and fertilization was observed solely in LX-fr.

Table 4. Effects of rainfall and irrigation on grain yield for each dominant soil type.

	Grain yield
	LX-fr	PT-pt	PT-pt.px
Rainfall	ns	ns	ns
Irrigation	ns	ns	ns
Rainfall × Irrigation	ns	ns	ns

LX-fr, Ferric Lixisols; PT-pt, Petric Plinthosols; PT-pt.px, Pisoplinthic Petric Plinthosols.

ns, not significant (p > 0.05).

Comparison between treatment #1 and #2 (Table 3) over 2 years.

Fertilizer was not applied in both treatments.

Table 5. Effects of rainfall and fertilization on grain yield for each dominant soil type.

	Grain yield (kg ha^−1)
	LX-fr	PT-pt	PT-pt.px
Rainfall
Lower (2017)	1023 (104) a	521 (82) b	524 (59) b
Average (2016)	1334 (139) a	796 (96) a	846 (88) a
Fertilizer (kg ha^−1)
0	522 (52) c	325 (60) c	287 (71) b
37	1040 (121) b	520 (144) bc	800 (79) a
74	1535 (120) a	971 (107) a	808 (105) a
111	1596 (142) a	784 (101) ab	780 (116) a
Rainfall	ns	*	**
Fertilizer	***	*	**
Rainfall × Fertilizer	*	ns	ns

LX-fr, Ferric Lixisols; PT-pt, Petric Plinthosols; PT-pt.px, Pisoplinthic Petric Plinthosols.

***P < 0.001, **P < 0.01, *P < 0.05, ns, not significant (p > 0.05).

Comparison between treatment #1 and #3–5 (Table 3) over 2 years.

Values in parentheses indicate standard errors.

4. Discussion

Irrigation at 120 and 88 mm in 2016 and 2017, representing 19% and 15% of the respective recorded rainfall amounts during the corresponding growing seasons, did not improve the grain yield under unfertilized conditions across all soil types with different effective soil depths. Furthermore, low rainfall did not adversely impact the grain yield in any soil type. These results were consistent with the observation that VWC values remained predominantly higher than θ_LL until ~ 100 DAS, even in the year with low rainfall. Previous studies have consistently highlighted that a water deficit during the booting to half-blooming stages (~ 60–80 DAS) significantly reduces grain yield (Inuyama, Musick, and Dusek 1976; Lewis, Hiler, and Jordan 1974; Salter and Goode 1967). In contrast, water deficit occurring from the soft dough stage to maturity (after ~ 80 DAS) does not have a detrimental effect on grain yield (Eck and Musick 1979; Mastrorilli, Katerji, and Rana 1995; Stichler, Mcfarland, and Coffman 1997).

VWC values in the two soil layers of PT-pt.px exceeding θ_LL until ~ 100 DAS, considering their typically low water holding capacity, was unexpected. This anomaly can be attributed to the presence of the petroplinthite, an impermeable layer, and coarse fragments on the soil surface. The impermeable petroplinthite possibly hindered deep water percolation, while the coarse fragments covering 15%–40% of soil surface (Ikazaki et al. 2018b) possibly suppressed evaporation from the soil surface. A comparison of the layer just above the petroplinthite in PT-pt.px (10–25 cm deep) with a layer of the same depth in LX-fr revealed that the soil pF was consistently lower (indicating wetter conditions) in PT-pt.px than LX-fr. This result would provide evidence that the petroplinthite inhibited deep water percolation in PT-pt.px. In addition, Kemper, Nicks, and Corey (1994) reported a drastic reduction in evaporation from the soil surface due to coarse fragment mulching. Similarly, since the soil pF in 10–25 cm layer was lower in PT-pt than in LX-fr, the petroplinthite in PT-pt may have contributed to soil water retention in 10–25 cm soil layer. Undoubtedly, the petroplinthite restricts rootable soil depth and negatively affects crop yields when comparing yields across different soil types (Ikazaki et al. 2018b). However, when comparing inter-annual differences in crop yields in PT-pt.px or PT-pt, the petroplinthite may mitigate yield loss in low rainfall years by retaining moisture in the surface layer. To our knowledge, this is the first study to describe the positive aspects of petroplinthite in crop production, despite its general association with suppressing crop yields.

Another reason for the lack of soil water not being a limiting factor in all soil types under unfertilized conditions could be the very low nutrient contents in topsoils caused by low clay and organic matter contents and soil erosion by water (Ikazaki et al. 2018a, 2020). Water erosion selectively washes away fine soil particles, rich in nutrients (Pierce and Lal 1994). The total N and available phosphorus contents in 0–25 cm layers of PT-pt.px, PT-pt, and LX-fr are only 0.2–0.3 g kg^‒1 and 1.2–3.7 mg kg^‒1, and PT-pt.px and PT-pt have a large number of coarse fragments that contain few nutrients.

In this study, the increase in grain yield across all soil types with fertilization strongly suggests that the lack of soil nutrients, especially N, as previously suggested by Twomlow et al. (2010), is the primary cause of reduced sorghum yield in the present Sudan Savanna. However, under fertilized conditions, reduced rainfall notably decreased grain yield in PT-pt and PT-pt.px, implying that while the limiting factor is solely the lack of soil nutrients in Lixisols (LX-fr), the lack of soil water could also be a limiting factor under fertilized conditions in Plinthosols (PT-pt and PT-pt.px). This difference may be attributed to the higher amount of water readily available for sorghum in LX-fr compared to PT-pt and PT-pt.px in 2017. The lower VWC within the effective soil depth in PT-pt and PT-pt.px in 2017 may not have been adequate to support the increase in biomass resulting from fertilization.

The observed synergistic effect of rainfall and fertilization was only evident in LX-fr. In addition, within Plinthosols soil type, yield saturations were observed at 74 kg N ha^‒1 in PT-pt and 37 kg N ha^‒1 in PT-pt.px. These differences are likely attributable to difference in the effective soil depth, with LX-fr having the greatest depth, followed by PT-pt, and PT-pt.px. Given sorghum’s natural characteristic of being a deep-rooted crop (Blum and Arkin 1984; Stichler, Mcfarland, and Coffman 1997), its response to fertilizer application tends to diminish with the root zone being restricted. Assefa and Staggenborg (2011) and Flores, Clark, and Gourley (1988) have correlated root mass with sorghum yield, particularly under well-fertilized conditions. However, since this study did not assess root mass, it remains uncertain whether the low response to fertilization in Plinthosols, particularly PT-pt.px, was primarily due to low root mass or a reduced total amount of water readily available for sorghum within the effective soil depth. Regardless, whether the effective soil depth exceeds 25 cm might be an important indicator in determining optimal fertilizer application rates aligned with soil conditions. Vogel (1993) similarly reported varying responses to fertilization in a semi-arid region in Zimbabwe when comparing soils with 25 and 45 cm rooting depths, albeit in maize.

5. Conclusion

Even in semi-arid Sudan Savanna, where Plinthosols with low effective soil depth are predominant, the primary factor contributing to reduced sorghum yield today appears to be a lack of soil nutrients rather than soil water. Plinthosols may still retain water in the root zone due to the presence of the petroplinthite and surface coarse fragments, potentially alleviating water shortage in years with low rainfall under unfertilized conditions. However, with the application of chemical fertilizers, the lack of soil water becomes a limiting factor for sorghum yield in Plinthosols (but not in Lixisols). This complexity indicates that limiting factors in crop production are multifaceted for soils with <50 cm effective soil depth. It emphasizes the importance of knowing the effective soil depth in advance when developing efficient fertilizer application methods in semi-arid regions of SSA. Additionally, the observation of yield saturations at different fertilizer application rates for PT-pt and PT-pt.px suggests that whether or not the effective soil depth exceeds 25 cm may be a crucial indicator in optimizing fertilizer application strategies.

Acknowledgments

We thank Dr. Adama Kaboré, Dr. Barthélémy Yelemou, and Mr. Simporé Kouka for their support.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This study was conducted under the JIRCAS-INERA collaborative project, “Development of watershed management model in the Central Plateau, Burkina Faso (2016–2020).”