https://orcid.org/0000-0002-5015-7837
Figures
Abstract
Rice production in sub-Saharan Africa (SSA) is restricted by low water availability, soil fertility, and fertilizer input, and phosphate rock (PR) application is expected to increase production. Soil water conditions and soil types affect the efficacy of phosphorus fertilization in improving productivity. However, these factors are rarely discussed together. In this study, we aimed to investigate the soil types and soil water conditions in the fields, as well as their effects on rice productivity after phosphorus fertilization, and optimize the findings using remote sensing techniques. A soil profiling survey, followed by a field experiment in seven farmer fields, was performed in the Central plateau of Burkina Faso. The following treatments were applied: nitrogen and potassium fertilization without phosphorus (NK), PR application with NK (NK+PR), and triple super phosphate (TSP) application with NK (NK+TSP). Submergence duration and cumulative water depth were recorded manually. The inundation score, estimated using a digital elevation model, explained the distribution of soil types and soil water conditions and correlated negatively with sand content and positively with silt and clay content, indicating an illuvial accumulation of fine soil particles with nutrient transportation. The field experiment showed that although grain yield was significantly restricted by phosphorus deficiency, the increase in yield after phosphorus fertilization was higher in Lixisols and Luvisols than in Cambisols because of the low Bray-2-phosphorus content of Lixisols and Luvisols. The inundation score correlated positively with grain yields after NK+PR and NK+TSP treatments. In conclusion, soils with low inundation scores (mainly Lixisols and Luvisols) showed a drastic increase in grain yield after TSP application, whereas those with high inundation scores showed comparable yields after PR and TSP application despite the low phosphorus fertilization effect. Our findings would help optimize fertilization practices to increase rice productivity in SSA.
Citation: Iwasaki S, Kanda T, Nakamura S, Uchida S, Saïdou S, Barro A, et al. (2024) Integrating field surveys and remote sensing to optimize phosphorus resource management for rainfed rice production in the Central plateau of Burkina Faso. PLoS ONE 19(10): e0312070. https://doi.org/10.1371/journal.pone.0312070
Editor: Paulo H. Pagliari, University of Minnesota, UNITED STATES OF AMERICA
Received: December 12, 2023; Accepted: September 30, 2024; Published: October 25, 2024
Copyright: © 2024 Iwasaki et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and its Supporting Information files.
Funding: This work was financially supported by the Science and Technology Research Partnership for Sustainable Development (SATREPS) project No. JPMJSA1609, Japan Science and Technology Agency (JST), and Japan International Cooperation Agency (JICA) (Project on the establishment of the model for fertilizing cultivation promotion using Burkina Faso phosphate rock, No. JPMJSA1609). There was no additional external funding received for this study. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Rice demand and production have increased in Sub-Saharan Africa (SSA) over the years. From 1991–2001, rice cultivation accounted for only 1.76% of agricultural production in the region; however, this figure rose to 3.96% from 2002–2013 [1]. Despite the increase in cultivation, the average grain yields in SSA were reported to be 4.0 Mg ha−1 for irrigated lowlands, 2.6 Mg ha−1 for rainfed lowlands, and 1.6 Mg ha−1 for rainfed uplands, significantly lower than the potential rice yield that could be obtained from the region [2]. Limited water availability, soil fertility, and fertilizer input are the factors that hinder rice production in the region [3, 4]. Rainfed agricultural production in SSA is highly sensitive to spatial and temporal variability in precipitation [5]. Additionally, highly weathered soils with low phosphorus (P) contents are prevalent in SSA and severely limit rice production [6]. Although the application of P fertilizers is the most common approach for overcoming soil P deficiency [7], its adoption by local farmers is limited because of the high cost of imported fertilizers [8] and high variability in their fertilization efficacy. A potential alternative source of P fertilizers is the widespread deposits of phosphate rocks (PRs) throughout SSA [9]. For example, Burkina Faso alone has 100 million metric tons of low-grade PRs that can serve as an alternative P fertilizer source [10, 11]. However, the efficacy of P fertilizers is impacted by several factors, such as P solubility of fertilizer, mean submergence duration and groundwater level (GWL) [12–15].
Soil type is another critical factor impacting crop production and fertilization efficacy in SSA [16]. A study in SSA has investigated the relationship between the soil type and rainfed lowland rice productivity [17]. In lowland soils, the distribution of soil types and soil water conditions is closely related, as both these factors are affected by the topography [18–21].
The Central plateau of Burkina Faso is a flat region situated in the upper areas of the Black and White Volta Rivers, with extremely gradual slopes. The river bottom zone of this region is characterized by seasonal or occasional flooding, largely determined by topographical and environmental factors. The flooding pattern in SSA has been estimated using deterministic, hydrodynamic, and probabilistic models [22–25]. However, as these models cannot be applied directly to assess flooding levels in the Central plateau of Burkina Faso because of the lack of hydrometric data, a remote sensing approach has been suggested to be relatively more practical for evaluating flooding tendencies in this region [26]. Although maps have been created to indicate the appropriate type and amount of fertilizer application in Africa in recent years [27, 28], lowland rice-producing areas are yet to be included in these maps. Specifically, the factors that impact P fertilization efficacy in lowland rice-producing areas remain to be investigated.
Therefore, in this study, we aimed to evaluate the soil types and soil water conditions in rice cultivation areas in Burkina Faso and analyzed their useability in recommending P fertilizers for rice production. Herein, we developed a new strategy for assessing flooding levels in the Central plateau of Burkina Faso using a digital elevation model (DEM) in combination with the geographic information system (GIS) to overcome the limitation of the lack of hydrometric data for this region. Our findings would contribute to the development of fertilization practices to help increase rice productivity in the upper reaches of rivers in Burkina Faso and SSA.
Materials and methods
Study site
This study was conducted in the Central plateau of Burkina Faso, specifically in the seven rainfed lowland rice cultivation fields (Nassoulou, Siguinvouse, Ramongo, Nandiala, Villy, Poa, and Sissene, named after the name of the village wherein they are located) around the Institute of Environment and Agricultural Research (INERA) Saria Station (12°16ʹ N, 2°0ʹ W) (Fig 1). These seven fields were selected to cover the entire watershed region upstream to downstream. Permission for the survey was obtained by INERA from the farmers who owned these seven rice cultivation fields. The study area receives a monomodal rainfall pattern between May and October, with the mean annual precipitation and temperature values of 800 mm and 28°C, respectively, as recorded at the INERA Saria station [14]. The mean annual potential evaporation in the study area is estimated to be between 1,700 and 2,000 mm [29].
Soil profiling survey and analysis
A soil profiling survey was conducted between October and November 2018 and March 2019. Composite soil samples were collected from each horizon to determine their physicochemical properties. The soil profiles were described and classified following the World Reference Base for Soil Resources system [30].
Soil particle size distribution was determined by degrading soil organic matter with H2O2 and dispersing it with sodium hexametaphosphate. Coarse and fine sand fractions (0.2–2 mm and 0.02–0.2 mm, respectively) were measured by sieving, whereas silt (0.002–0.02 mm) and clay (<0.002 mm) fractions were measured using the pipette method. Soil texture class was determined based on the textual triangle (30).
Total carbon (C) and nitrogen (N) contents were determined using the dry combustion method using an NC Analyzer (NC 220F; Sumika Chemical Analysis Service, Tokyo, Japan). The pH and electrical conductivity of the 1:5 soil: solution suspensions in deionized water and in 1M potassium chloride (KCl) were measured using a pH electrode (LAQUA F-72; HORIBA Ltd., Kyoto, Japan) and conductivity meter (ES-51; HORIBA Ltd., Japan), respectively.
Available P was determined using the Bray1 and Bray2 methods [31]. Phosphorous concentrations were determined calorimetrically using a spectrophotometer (UV-1800; Shimadzu Corp., Osaka, Japan). Exchangeable cations were extracted using 1.0 M ammonium acetate (pH 7.0), and the cation concentrations were determined using an inductively coupled plasma atomic emission spectrophotometer (ICPE-9000; Shimadzu Corp., Osaka, Japan).
After the exchangeable base extraction, the residues were washed with 80% (v/v) ethanol to determine their cation exchange capacity (CEC). Saturated NH4+ ions were extracted 4 times with 1 M KCl (pH 7.0), and the concentrations were determined using the salicylate method [32] and continuous-flow analysis in an Auto Analyzer III (BL-TEC KK, Tokyo, Japan).
The base saturation (BS) was calculated as the ratio of exchangeable basic cations to the CEC. The effective BS (EBS) [30] was calculated as the sum of exchangeable base cations (Ca2++Mg2++K++Na+) divided by the effective CEC (Ca2++Mg2++ K++Na++Al3+).
Field experiment
We conducted a field experiment to assess the efficacy of directly applying PR in rice production. The experiment was performed over 3 years, starting in 2013. The experimental plots were managed following local farmer practices, with rainfed cultivation utilizing floodwater, and no puddling or leveling was performed. Each plot was 25 m2 in size and was enclosed in a 30-cm-high bund to prevent fertilizer overflow and collect water. The following four treatments were used in this study:
- No fertilization (CT)
- Application of N and potassium (K) without P fertilization (NK)
- Application of PR with NK (NK+PR)
- Application of triple superphosphate (TSP) with NK (NK+TSP)
The application rates for N, P, and K were 90 kg ha−1, 60 kg ha−1, and 50 kg ha−1, respectively. The source of P used was PR from the Kodjari deposit (12°1′ N; 1°55′ E) in Burkina Faso, which contained 113 g P kg−1. The solubility of P in the PR for water, alkaline ammonium citrate, and 2% citric acid were 0.2, 2.5, and 31.1%, respectively. Moreover, N and K were applied as ammonium sulfate and KCl, respectively. In 2015, P was not applied. Rice seeds were directly sown at a spacing of 20 cm × 20 cm at the beginning of July 2015, and the plants were harvested in mid-October 2015. The duration of submergence and water depth were measured manually.
Inundation score estimation using a DEM
The flow of inundation score estimation and the following analysis are shown in Fig 2. In this study, the ALOS World 3D-30m DEM with 5 m resolution was used (https://www.eorc.jaxa.jp/ALOS/en/aw3d30/data/index.htm). The dataset provides a digital surface model, including anthropogenic construction and large trees that may show higher elevation values than ground-level ones. However, as the ground surface in this region is usually covered with bare soil or low vegetation during the dry season, the difference between the DEM and digital surface model was assumed to be negligible.
DEM, digital elevation model; AW3D30, ALOS World 3D-30m DEM (https://www.eorc.jaxa.jp/ALOS/en/aw3d30/data/index.htm); P. phosphorus.
The inundation score was estimated based on water accumulation scores and convexities. First, the water accumulation score was calculated using ArcGIS® v. 10.7.1 software to represent the accumulated number of cells on the upstream side and the amount of water from rainfall. The DEM data in geographical coordinates were converted to the Universal Transverse Mercator system. Subsequently, a low-pass filter with a window size of 5 × 5 was used to convert the integer values to floating values. Logarithmic transformation was applied to the water accumulation score, and a low-pass filter with a window size of 25 × 25 was applied twice. As a 25 × 25 spatial filter was used to eliminate the insignificant effect of microtopography, the longitudinal and cross-sectional convexities were estimated using ENVI® v. 5.5.1. A negative convexity indicates concave topography, inducing water stagnation. Finally, the inundation score was calculated by multiplying the water accumulation score by convexities. In this calculation, scores are shown as absolute values, and when convexity was positive, water was assumed not to stagnate, and the inundation score was treated as zero.
Statistical analyses
The R version 4.2.1 software [33] was used to process and visualize the data. Considering that the data was not normally distributed, Spearman’s correlation coefficient was used to demonstrate the relationships between soil physicochemical properties. A one-way analysis of variance followed by Tukey’s honestly significant difference multiple comparisons test was utilized to analyze the differences in grain yield between the fertilization treatments for three soil groups (Lixisols and Luvisols, Cambisols, and all sites separately). The differences in the grain yield of the same treatment across different soil types were analyzed using non-paired t-tests. Principal component analysis (PCA) was applied to understand the variability of the data and identify the characteristics of each site and the soil type by reducing the dimension of the data while minimizing the loss of information [34, 35]. After the standardization, cumulative precipitation, water accumulation score, longitudinal and cross-sectional convexity, submergence duration, cumulative water depth, and soil particle distribution data were subjected to PCA. Single regression models were applied to investigate the relationship between inundation scores and soil particle distribution, submergence duration, cumulative water depth, and grain yield in each treatment. The possibility of substituting TSP with PR was demonstrated by illustrating the relationship between the inundation scores and the relative grain yields obtained after the NK+PR and NK+TSP treatments. An alpha level of significance of 0.05 was used throughout the study.
Results
Soil classification and physicochemical properties
Table 1 describes the physicochemical properties of the three soil groups. Supplementary materials (S1 Table and S1 Fig) show soil morphological properties and photographs of the soil profiles. The seven soil profiles exhibited gleyic conditions from the middle to lower horizons; however, reducing conditions were not detected using dipyridyl solution testing. Therefore, no soil was classified as a Gleysol. The soil types were classified based on the presence of argic horizons, CEC per clay in argic horizons, and ferric horizons. Soils at Nassoulou, Siguinvouse, Ramongo, and Poa sites had argic horizons starting ≤100 cm below the soil surface. The CEC per clay in the argic horizons of Siguinvouse, Ramongo, and Poa was <24 cmolc kg−1 clay, whereas that in the argic horizon of Nassoulou was >24 cmolc kg−1 clay. The EBS value was >50% for all horizons between 50 and 100 cm. Notably, Bmvg (46–73 cm) and Bmvg2 (73–100 cm) at Ramongo and Btwgc (42–55 cm) and Bmv (55–90 cm) at Siguinvouse were ferric horizons (detailed definitions of the horizons are provided in [30]). Thus, the soils at Ramongo and Siguinvouse were classified as Ferric Relictigleyic Lixisols, whereas the soil at Poa was classified as Abruptic Relictigleyic Lixisol. The soil at Nassoulou was classified as Abruptic Relictigleyic Luvisol. Cambic horizons (Bwg1 and Bwg2 in Sissene, Bwg1 in Nandiala, and Bwg1 in Villy) were present in the soils at Sissene, Nandiala, and Villy, as confirmed by their distinct color and clay content compared with that of their underlying horizons. Additionally, EBS was ≥50% throughout the soil at these sites. Sideralic Bwg1 (21–31 cm), Bwg2 (31–40 cm) at Sissene, and Bwg1 (19–35 cm) at Villy were present, as their CEC per clay was <24 cmolc∙kg−1 clay. Therefore, the soils were classified as Eutric Sideralic Relictigleyic Cambisol at Sissene and Villy and Eutric Relictigleyic Cambisol at Nandiala. The mean elevation of the Lixisols and Luvisols group and Cambisols group was 312.8 and 296.7 m, respectively. The surface horizon of the Cambisols group had a finer texture with higher silt and clay contents than those in the group containing Lixisols and Luvisols (Table 1). Total C and N content, CEC per clay were higher for the Cambisols than for the Luvisols and Lixisols in the entire soil profiles. Sand content correlated negatively with total C and N content, exchangeable cations, and CEC (Table 2). In contrast, silt content showed a positive correlation with total C and N content and Bray2-P. Clay content correlated positively with the concentrations of exchangeable cations and CEC.
Soil water conditions and rice grain yields under different fertilization treatments
Soil water conditions and grain yields and a comparison of these variables across the different soil types are presented in Table 3. The results indicate that Cambisols had longer submergence durations and greater cumulative water depths than Lixisols and Luvisols. In Lixisols and Luvisols, the application of NK did not significantly increase grain yield. However, the NK+PR and NK+TSP treatments resulted in significantly higher grain yields than those in the CT and NK treatments. In Cambisols, the only significant difference in grain yield was observed between the CT and NK+TSP treatments. The mean grain yield across all years and sites ranged from 1.24 ± 0.85 Mg ha–1 for the CT treatment to 4.15 ± 1.94 Mg ha–1 for the NK+TSP treatment. The application of NPK (NK+TSP and NK+PR) significantly increased grain yield, whereas NK application without P did not. Generally, grain yield was higher in Cambisols than in Lixisols and Luvisols, and a significant difference in grain yield between soil types was only observed in the CT and NK treatments. Furthermore, the increase in grain yield over that in the control treatment (CT) was significantly higher in the NK+TSP and NK+PR treatments compared with that in the NK treatment, although no significant differences were observed between the treatments in Cambisols.
Geographical characteristics of the seven observation points and inundation score estimation
The water accumulation scores and convexities are summarized in Table 4 (additional topographic features are shown in S2 Fig). Water accumulation scores generally corresponded with the positions of the sites on the slope and elevation, as shown in Fig 1. Nassoulou and Siguinvouse exhibited small water accumulation and positive longitudinal convexity. In contrast, the other five sites exhibited water accumulation values >0.75, with varying longitudinal and cross-sectional convexities. Longitudinal convexity generally matched water accumulation, whereas cross-sectional convexity did not (Fig 3).
Relationships between (A) water accumulation and longitudinal convexity and (B); water accumulation and cross-sectional convexity. Lixisols, Luvisols, and Cambisols: Soil type in the World Reference Base for Soil Resources (WRB) system [30].
Furthermore, water accumulation scores and longitudinal convexity explained the distribution of the soil types and soil water conditions. Specifically, Cambisols were found on sites with water accumulation scores >0.76 and longitudinal convexity <−0.0052 (Fig 3A and Table 4). The duration of submergence and cumulative water depth were longer and deeper on sites with large water accumulation scores and large negative longitudinal convexity (Fig 3B).
The PCA demonstrated the characteristics of the different soil types and soil water conditions (Table 5 and Fig 4). Notably, PC1 accounted for 54% of the total data variance, whereas PC2–PC4 explained 10–16%. Variation between the two soil groups was explained by PC1, whereas variation within the same soil group was explained by PC2.
PC, principal component; Rain, cumulative precipitation during the growing season; WA, water accumulation score; LC, longitudinal convexity; CC, cross-sectional convexity; Submerge, submergence duration; Depth, cumulative water depth.
Furthermore, the water accumulation scores and longitudinal convexities strongly contributed to PC1. Based on these findings, we used the water accumulation scores and longitudinal convexities to estimate the inundation score. Cambisols had larger inundation scores (0.73 on average) than Lixisols and Luvisols (0.18 on average) (Table 4).
Integration of field observations and remote sensing data
The inundation score exhibited a significant negative correlation with sand content but a positive correlation with silt and clay content (p <0.05) (Fig 5). If Nandiala, which exhibited an exceptionally long duration of submergence and deep cumulative water depth, was excluded, the inundation score accounted for 76 and 84% of the total variation in the duration of submergence and cumulative water depth, respectively (Fig 5). There were significant positive relationships between the inundation score and grain yield in the CT and NK treatments, with almost similar slopes in the regression equation (Fig 6A). However, this relationship was not observed in the NK+PR and NK+TSP treatments. Moreover, there was a significant positive correlation between the inundation scores and relative yields of the NK+PR treatment over the NK+TSP treatment (Fig 6B).
Points represent the mean ± standard error for the three years (2013–2015). Soil particles represent the weighted average sand, silt, and clay content in the 0–20 cm soil layer. If the relationship was significant, regression lines and equations were drawn (p <0.05). R2adj, adjusted coefficient of determination; gray region, 95% confidence interval.
Relationships between inundation score and rice grain yield in the different fertilization treatments (A) and between the relative grain yield of PR+NK over TSP+NK (B). Points represent the mean ± standard error for the three years (2013–2015). If the relationship was significant, regression lines and equations were drawn (p <0.05). CT, without fertilization; NK, without P fertilization with N and K fertilization; NK+PR, PR application with N and K fertilization; NK+TSP, TSP application with N and K fertilization; N, nitrogen; P, phosphorus; PR, phosphate rock; K, potassium; R2adj, adjusted coefficient of determination; gray region, 95% confidence interval.
Discussion
Soil type distribution and the estimation of soil water conditions using inundation scores
The seven soil profiles were classified as Lixisols, Luvisols, and Cambisols (Table 1). The soil horizons were predominantly classified based on the presence of gleyic properties caused by groundwater movement and the presence of an argic horizon. Gleyic properties, without reducing conditions, were observed in Siguinvouse, Nassoulou, and Poa, indicating temporal groundwater impact. In the study area, the GWL gradually rises at the start of the rainy season, saturates with water up to the soil surface level during the rainy season, and sinks to ≥100 cm below the soil surface during the dry season [13]. Hence, both groundwater and stagnant water impact the soil profile morphology and create an albic horizon in soils.
The main differentiator among the soil groups was the presence of an argic horizon starting within 100 cm below the soil surface. Soils with argic horizons, e.g., Lixisols and Luvisols, are generated on stable landforms and have an advanced age of pedogenesis [36]. Topography is important in determining soil distribution in SSA [18, 20, 37]. Ikazaki et al. [18] showed that the soil type in this region changes even within a 12 m elevation. Hence, the local soil types in this area differ, even on a small scale. Here, Cambisols were located at relatively low altitudes (an average of 296.7 m), whereas Lixisols and Luvisols were located at comparatively higher elevations (an average of 312.8 m).
The main physicochemical soil property distinguishing the Lixisols, Luvisols, and Cambisols was the surface horizon soil texture (Table 1). The silt and clay content of the surface horizons of Lixisols and Luvisols (Ramongo, Siguinvouse, Poa, and Nassoulou) were low to medium (silt: 10.8–25.4%; clay: 6.9–19.6%). In contrast, the silt and clay content were high for the three Cambisol profiles (silt: 20.6−30.3%; clay: 30.5–48.3%). The high silt and clay content in Cambisols is attributed to soil erosion and silt and clay illuviation from the upper and middle slopes to the lower slope and river bottom [20]. In Burkina Faso, intense rainfall occurs during the rainy season, causing surface water runoff without underground penetration. Wakatsuki et al. [38] reported that surface runoff deposits clay particles in the surface soil of managed rice fields in Burkina Faso. Ikazaki et al. [18] stated that the effective soil depth thickness for crop production increases in the lower part of the slope. In contrast, the clay in the argic horizon in the Lixisols and Luvisols is attributed to illuvial accumulation from the overlying horizon [18, 20]. Deressa et al. [20] reported that in western Ethiopia, the argic horizon depths in the profiles correlate with the wetting front depths and subsequent drying.
In this study, we utilized DEM and GIS to estimate the inundation scores and explain the distribution of the soil types and soil water conditions (Fig 5). It was found that Cambisols were only present in areas with an inundation score >0.40. The inundation score is determined by the water accumulation score and longitudinal convexity, key factors that affect soil erosion and clay illuviation [39, 40]. Therefore, the inundation score can be utilized in predicting the distribution of soil types on a watershed scale. The significant correlation between the inundation score and soil particle distribution emphasizes the illuvial accumulation of fine soil particles with water movement (Fig 6A).
Notably, Lixisols were also found in areas with an inundation score of 0.41. This can be partly explained by the fact that Lixisols have a wider range of soil characteristics compared to Cambisols, as demonstrated by PC1 (Fig 4). For example, soil in Poa was classified as Lixisols. However, the distribution of soil particles in the profiles showed a different pattern (Table 1). Although the argic horizon was observed in the soil profile of Poa, the silt and clay content in the surface horizon was relatively higher compared with that at the other Lixisol and Luvisol sites. Additionally, the argic horizon in Poa was found at a depth of 9.0 cm from the surface. These findings suggest that the soil in Poa is situated in the transition zone between Lixisols and Cambisols. It is essential to conduct surveys at multiple locations covering the watershed in future studies to normalize the results.
The inundation score explained the field observations of submergence duration and cumulative water depth, indicating that the inundation score can be utilized for further analysis to understand P fertilization efficacy (Fig 6). However, we needed to exclude Nandiala, with its extremely long submergence duration and deep cumulative water depth, which may be attributed to the impact of groundwater fractionation or inundation from natural water basins, important factors determining water conditions in this region [12, 13]. Additionally, based on the field observations, the site is located near a road, which obstructs the flow of water and has an impact on the submergence conditions.
Inundation score as an indicator of the fertilization effect
In this study, the grain yield for the CT and NK+TSP treatments may reflect the original soil productivity and the potential yield of rice cultivation, respectively. The grain yield in the NK treatment (average 1.93 ± 1.05 Mg ha–1) was significantly lower than that in the NK+TSP treatment (average 4.15 ± 1.94 Mg ha–1) (Table 3). Considering the proposed critical limit of the Bray1-P content based on previous studies in SSA, which is 7–9 mg P kg−1 [41], the Bray1-P content in this study were substantially low, ranging from 0.9–6.1 mg P kg−1 (Table 1). These results indicate that P deficiency was a critical limiting factor for rice productivity, as reported in SSA [6, 42]. However, grain yield in NK showed an increasing trend compared with CT. In addition, recent studies in SSA showed that cases of P and K deficiency alone did not frequently occur during rainfed lowland rice cultivation; however, their interactions with N deficiency are an important limiting factor in rice production [6, 17]. Furthermore, Dossou-Yovo et al. [43], compiled 652 field experiments in SSA and showed that the soil total N content was not invariably related to soil N mineralization and rice yield without N application. Based on these results, P is the critical yield-limiting factor in rainfed lowland rice cultivation in this region, along with an overall nutrient deficiency. Notably, 3.52 ± 1.23 Mg ha–1 of grain yield in the NK+PR treatment was far larger than the 2.6 Mg ha–1 of the average yield of rainfed lowland rice cultivation in SSA [2], indicating the potential utilization of low-grade PR to increase rice production in SSA. However, there are conflicting data regarding the effect of the direct application of PR on lowland rice cultivation in SSA. According to previous studies, soil water conditions are critical factors in controlling the effect of PR and PR-based local P fertilizers effect [12–14]. Although soil type is often considered an important factor that impacts the efficacy of fertilization, in this study, the soil water conditions and soil type distribution were closely related and explained based on inundation scores, as discussed previously. Therefore, these two factors can be addressed together.
Significant positive relationships were observed between the inundation scores and grain yield for the CT and NK treatments but not for the treatment with P fertilization (Fig 6A). Furthermore, the regression slopes for the CT and NK treatments were almost parallel, with values of 1.73 and 1.71 for the CT and NK treatments, respectively, and the y-intercepts differed at 0.56 and 1.26 Mg ha−1 for CT and NK, respectively. These relationship trends suggest that N and K limit grain yield, regardless of the differences in inundation scores. Therefore, grain yield increased after NK application in Lixisols and Luvisols, and Cambisols showed similar values. The significant positive relationships between the inundation scores and grain yield for the CT and NK treatments can be explained by soil particle distribution and chemical properties (Table 2). Clay and silt particles contribute to the physical and chemical stabilization of soil organic matter, which affects soil C and N accumulation [44, 45], and clay particles exhibit a high CEC [46]. The findings of previous studies support significant positive correlations between silt or clay content and total C and N content, exchangeable cations, and CEC, with high correlation coefficients. Moreover, water erosion can transport macro- and micronutrient elements, as reported globally, including in Africa [47, 48]. Although there was no significant correlation between soil particles and Bray1-P content, Bray2-P content positively correlated with silt content (Table 2). Thus, the illuvial accumulation of silt particles and the transportation of P at sites with high inundation scores resulted in high grain yields in the CT and NK treatments. Moreover, yield limitation due to low Bray2-P content was relatively more extensive in soils with low inundation scores, resulting in a drastic increase in grain yields after P application in Lixisols and Luvisols.
Finally, we found that the relative grain yields of the NK+PR treatment over the NK+TSP treatment increased at sites with high inundation scores (Fig 6B). Iwasaki et al. [13] revealed that rice plants in the central plateau of Burkina Faso could utilize the relatively soluble P fraction in fields with high GWLs. In this study region, the rice cultivation area is characterized by alternative inundation and drying, even in the lowland (corresponding to the sites with high inundation scores in this study), particularly in the early growing stage [13]. These results were probably due to the positive effects of alternative drying and rewetting on the acceleration of root architecture development and P diffusion in the soil [49] at the sites with high inundation scores.
Furthermore, there was a large variation in yields under P fertilization conditions at similar flooding scores. Thus, concerns remain regarding the relatively small area under evaluation in this study. Therefore, it is desirable to increase the number of study sites in future studies and validate the results obtained in this study.
Conclusions
Inundation scores can be used to estimate the distribution of soil types in lowland areas along rivers and the effects of different sources of P fertilization while considering the interactions between soil water conditions, the accumulation of soil particles through illuviation, and the movement of nutrients. Regardless of the soil type, rice grain yields were significantly limited by P deficiency. However, the effect of P fertilization and the optimal selection of P fertilizer types differed based on the inundation score. At sites with low inundation scores (mainly Lixisols and Luvisols), a drastic increase in grain yields can be expected after TSP application. Conversely, at sites with high inundation scores, although the P fertilization effect was relatively small, the PR application showed a performance comparable to that of the TSP application. These results provide basic information for determining optimal P fertilization in the upper reaches of rivers in Burkina Faso and SSA. However, further soil profiling surveys and observations for soil water conditions are essential to bridge the gap between remote sensing data and field-scale observations.
Supporting information
S1 Fig. Topographical characteristics of the study sites.
https://doi.org/10.1371/journal.pone.0312070.s001
(DOCX)
S2 Fig.
Photographs of the soil profiles (upper panel) and soil texture distributions (lower panel). Horizon names were set following the World Reference Base for Soil Resources [30].
https://doi.org/10.1371/journal.pone.0312070.s002
(DOCX)
S1 Table. Soil morphological properties based on the World Reference Base for Soil Resources [30].
Mottling is described in the order of abundance, contrast, shape, and species. The mottled color is expressed in parentheses. Abbreviations: N, None; V, Very few (0–2%); F, few (2–5%); C, Common (5–15%); M, Many (15–40%); A, Abundant (>40%). Contrast: F, Faint; D, Distinct; P, Prominent. Shape: RO, Root-like; FI, Filmy; IR, Irregular; SP, Speckled; CL, Cloudy. Species: Fe, Iron; Mn, Manganese. Concretions are described based on abundance and species. The abbreviations are the same as those used for mottling. The concentrations of Fe2+ ions were related to the strength of the reaction with the 0.2% α-dipyridyl solution in 10% acetic acid. Strength of reaction:–, none; ±, very weak; +, weak; ++, strong; +++, very strong. Rocks are described in the order of abundance, weathering state, size, and shape. Abbreviations: N, None; F, Few (0–5%); C, Common (5–10%); M, Many (10–20%); A, Abundant (20–50%); D, Dominant (>50%). State of weathering: F, Fresh; SL, Slightly weathered; W, Weathered; ST, Strongly weathered. Size: FC, Fine gravel (0.2–1 cm); G, Gravel (1–5 cm); S, Stone (5–10 cm); LS, Large stone (10–20 cm); UN Unknown. The structure of the rocks is described in the order of their grades, sizes, and types. Grade: W, Weak; M, Moderate; S, Strong. Size: VF, Very fine (<5 mm); F, Fine (5–10 mm); M, Medium (10–20 mm); C, Coarse (20–50 mm); VC, Very coarse (>50 mm). Type: SB, Subangular block; PL, Platy; M, Massive; N, None.
https://doi.org/10.1371/journal.pone.0312070.s003
(DOCX)
Acknowledgments
The authors thank the local farmers and technical staff members for their support with field trials and the Soil Team at JIRCAS for soil chemical analysis.
References
- 1.
Chauhan BS, Jabran K, Mahajan G. Rice production worldwide. Cham, Switzerland: Springer International Publishing; 2017. http://link.springer.com/10.1007/978-3-319-47516-5
- 2. Tanaka A, Johnson J-M, Senthilkumar K, Akakpo C, Segda Z, Yameogo LP, et al. On-farm rice yield and its association with biophysical factors in sub-Saharan Africa. Eur J Agron. 2017;85:1–11.
- 3. Callo-Concha D, Gaiser T, Webber H, Tischbein B, Müller M, Ewert F. Farming in the West African Sudan Savanna: Insights in the context of climate change. Afr J Agric Res. 2013;8(38):4693–4705.
- 4. Niang A, Becker M, Ewert F, Dieng I, Gaiser T, Tanaka A, et al. Variability and determinants of yields in rice production systems of West Africa. Field Crops Res. 2017;207:1–12.
- 5. Sanfo S, Barbier B, Dabiré IWP, Vlek PLG, Fonta WM, Ibrahim B, et al. Rainfall variability adaptation strategies: An ex-ante assessment of supplemental irrigation from farm ponds in southern Burkina Faso. Agric Syst. 2017;152:80–89.
- 6. Saito K, Vandamme E, Johnson JM, Tanaka A, Senthilkumar K, Dieng I, et al. Yield-limiting macronutrients for rice in sub-Saharan Africa. Geoderma. 2019;338:546–554.
- 7. Chien SH, Prochnow LI, Tu S, Snyder CS. Agronomic and environmental aspects of phosphate fertilizers varying in source and solubility: an update review. Nutr Cycl Agroecosyst. 2011;89(2):229–255.
- 8.
Kelly VA. Factors affecting demand for fertilizer in sub-Saharan Africa. Agriculture and rural development discussion paper 23. The World Bank: Washington DC. 2006;23:89.
- 9.
Food and Agriculture Organization of the United Nations. Use of phosphate rock for sustainable agriculture. FAO Fertil Plant Nutr Bull. Rome, Italy. 2004;13, 148. Available from: https://www.fao.org/4/y5053e/y5053e00.htm
- 10. Nakamura S, Fukuda M, Nagumo F, Tobita S. Potential utilization of local phosphate rocks to enhance rice production in sub-Saharan Africa. Japan Agric Res Q: JARQ. 2013;47(4):353–363.
- 11.
van Straaten P. Rocks for crops: Agrominerals of sub-Saharan Africa. ICRAF, Nairobi, Kenya: Guelph, Canada; University of Guelph. 2002. p. 338.
- 12. Fukuda M, Soma DM, Iwasaki S, Nakamura S, Kanda T, Ouattara K, et al. Site-specific responses of lowland rice to acidulated and calcined phosphate rock fertilizers in the Center-West region of Burkina Faso. PLoS ONE. 2021;16(4):e0250240. pmid:33872338
- 13. Iwasaki S, Fukuda M, Ikazaki K, Nakamura S, Ouattara K, Nagumo F. 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. Soil Sci Plant Nutr. 2021;67(4):460–470.
- 14. Nakamura S, Saidou S, Barro A, Jonas D, Fukuda M, Kanda T, et al. Kodjari phosphate rock for rain-fed lowland rice production in the Sudan Savanna, Burkina Faso. Trop Agric Dev. 2020;64(2):97–106.
- 15. Santos WO, Mattiello EM, Barreto MSC, Cantarutti RB. Acid ammonium citrate as P extractor for fertilizers of varying solubility. Rev Bras Ciênc Solo. 2019;43:e0180072.
- 16. Lal R. Erosion-crop productivity relationships for soils of Africa. Soil Sci Soc Am J. 1995;59(3):661–667.
- 17. Zingore S, Adolwa IS, Njoroge S, Johnson J-M, Saito K, Phillips S, et al. Novel insights into factors associated with yield response and nutrient use efficiency of maize and rice in sub-Saharan Africa. A review. Agron Sustain Dev. 2022;42:82.
- 18. Ikazaki K, Nagumo F, Simporé S, Barro A. Soil toposequence, productivity, and a simple technique to detect petroplinthites using ground-penetrating radar in the Sudan Savanna. Soil Sci Plant Nutr. 2018;64(5):623–631.
- 19. Baldwin D, Naithani KJ, Lin H. Combined soil-terrain stratification for characterizing catchment-scale soil moisture variation. Geoderma. 2017;285:260–269.
- 20. Deressa A, Yli-Halla M, Mohamed M, Wogi L. Soil classification of humid Western Ethiopia: A transect study along a toposequence in Didessa watershed. CATENA 2018;163:184–195.
- 21. Kyebogola S, Burras LC, Miller BA, Semalulu O, Yost RS, Tenywa MM, et al. Comparing Uganda’s indigenous soil classification system with World Reference Base and USDA Soil Taxonomy to predict soil productivity. Geoderma Reg. 2020;22:e00296.
- 22. Bates PD, De Roo APJ. A simple raster-based model for flood inundation simulation. J Hydrol. 2000;236:54–77.
- 23. Di Baldassarre G, Schumann G, Bates PD, Freer JE, Beven KJ. Flood-plain mapping: a critical discussion of deterministic and probabilistic approaches. Hydrol Sci J. 2010;55(3):364–376.
- 24. Ettritch G, Hardy A, Bojang L, Cross D, Bunting P, Brewer P. Enhancing digital elevation models for hydraulic modelling using flood frequency detection. Remote Sens Environ. 2018;217:506–522.
- 25. Komi K, Neal J, Trigg MA, Diekkrüger B. Modelling of flood hazard extent in data sparse areas: a case study of the Oti River basin, West Africa. J Hydrol: Reg Stud. 2017;10:122–132.
- 26. Huang L, Mao X, Wang J, Chen X, Wang GH, Liao ZW. The effect and mechanism of improved efficiency of physicochemical pro-release treatment for low grade phosphate rock. J Soil Sci Plant Nutr. 2014;14(2):316–331.
- 27.
Leenaars, JGB, Ruiperez González M, Kempen B. Extrapolation of fertilizer nutrient recommendations for major food crops in West Africa; a proof of concept (with dataset). In: Project report (draft) for IFDC. USAID–West Africa Fertilizer Program, Accra. ISRIC–World Soil Information, Wageningen, the Netherlands. 2018. pp. 7–81.
- 28. Szilas C, Semoka JMR, Borggaard OK. Establishment of an agronomic database for Minjingu phosphate rock and examples of its potential use. Nutr Cycl Agroecosyst. 2007;78(3):225–237.
- 29. Ouattara K, Ouattara B, Assa A, Sédogo PM. Long-term effect of ploughing, and organic matter input on soil moisture characteristics of a ferric lixisol in Burkina Faso. Soil Tillage Res. 2006;88:217–224.
- 30.
IUSS Working Group WRB. World reference base for soil resources 2014, update 2015. International soil classification system for naming soils and creating legends for soil maps. World Soil Resources Report No. 106. FAO, Rome; 2015.
- 31. Bray RH, Kurtz LT. Determination of total, organic, and available forms of phosphorus in soils. Soil Sci. 1945;59:39–46.
- 32. Nelson DW. Determination of ammonium in KCl extracts of soils by the salicylate method. Commun. Soil Sci. Plant Anal. 1983;14:1051–1062.
- 33.
R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria; 2022. Available from: https://www.R-project.org/
- 34. Armenise E, Redmile-Gordon MA, Stellacci AM, Ciccarese A, Rubino P. Developing a soil quality index to compare soil fitness for agricultural use under different managements in the Mediterranean environment. Soil Tillage Res. 2013;130:91–98.
- 35. Bilgili AV, Aydemir S, Altun O, Sayğan EP, Yalçın H, Schindelbeck R. The effects of biochars produced from the residues of locally grown crops on soil quality variables and indexes. Geoderma. 2019;345:123–133.
- 36.
Fanning DS, Fanning MCB. Soil morphology, genesis, and classification. New Yok: John Wiley & Sons Inc.; 1989. pp. 177–178.
- 37. Adhikari K, Minasny B, Greve MB, Greve MH. Constructing a soil class map of Denmark based on the FAO legend using digital techniques. Geoderma. 2014;214:101–113.
- 38. Wakatsuki T. Paddy agriculture can prevent soil erosion in West Africa. Res. Environ. Disrupt Toward Interdiscip. Coop. 1988;18:20–27 [in Japanese].
- 39. Gunal H, Ransom MD. Clay illuviation and calcium carbonate accumulation along a precipitation gradient in Kansas. CATENA 2006;68(1):59–69.
- 40. Olorunfemi IE, Komolafe AA, Fasinmirin JT, Olufayo AA, Akande SO. A GIS-based assessment of the potential soil erosion and flood hazard zones in Ekiti State, Southwestern Nigeria using integrated RUSLE and HAND models. CATENA 2020;194:104725.
- 41. Bado VB, Djaman K, Mel VC. Developing fertilizer recommendations for rice in Sub-Saharan Africa, achievements and opportunities. Paddy Water Environ. 2018;16:571–586.
- 42. Vandamme E, Ahouanton K, Mwakasege L, Mujuni S, Mujawamariya G, Kamanda J, et al. Phosphorus micro-dosing as an entry point to sustainable intensification of rice systems in sub-Saharan Africa. Field Crops Res. 2018;222:39–49.
- 43. Dossou-Yovo ER, Vandamme E, Dieng I, Johnson J-M, Saito K. Decomposing rice yield gaps into efficiency, resource and technology yield gaps in sub-Saharan Africa. Field Crops Res. 2020;258:107963.
- 44. Bronick CJ, Lal R. Soil structure and management: a review. Geoderma. 2005;124(1–2):3–22.
- 45. Six J, Conant RT, Paul EA, Paustian K. Stabilization mechanisms of soil organic matter: Implications for C-saturation of soils. Plant Soil 2002;241:155–76.
- 46. Gruba P, Mulder J. Tree species affect cation exchange capacity (CEC) and cation binding properties of organic matter in acid forest soils. Sci Total Environ. 2015;511:655–62. pmid:25596350
- 47. Alewell C, Ringeval B, Ballabio C, Robinson DA, Panagos P, Borrelli P. Global phosphorus shortage will be aggravated by soil erosion. Nat Commun. 2020;11(1):4546. pmid:32917863
- 48. Bashagaluke JB, Logah V, Opoku A, Sarkodie-Addo J, Quansah C. Soil nutrient loss through erosion: Impact of different cropping systems and soil amendments in Ghana. PLoS ONE. 2018;13(12):e0208250. pmid:30566517
- 49. De Bauw P, Vandamme E, Lupembe A, Mwakasege L, Senthilkumar K, Merckx R. Architectural root responses of rice to reduced water availability can overcome phosphorus stress. Agronomy 2018;9(1):11.