Estimating nutrient concentrations and uptake in rice grain in sub-Saharan Africa using linear mixed-effects regression.

Context or problem: Quantification of nutrient concentrations in rice grain is essential for evaluating nutrient uptake, use efficiency, and balance to develop fertilizer recommendation guidelines. Accurate estimation of nutrient concentrations without relying on plant laboratory analysis is needed in sub-Saharan Africa (SSA), where farmers do not generally have access to laboratories. Objective or research question: The objectives are to 1) examine if the concentrations of macro- (N, P, K, Ca, Mg, S) and micronutrients (Fe, Mn, B, Cu) in rice grain can be estimated using agro-ecological zones (AEZ), production systems, soil properties, and mineral fertilizer application (N, P, and K) rates as predictor variables, and 2) to identify if nutrient uptakes estimated by best-fitted models with above variables provide improved prediction of actual nutrient uptakes (predicted nutrient concentrations x grain yield) compared to average-based uptakes (average nutrient concentrations in SSA x grain yield). Methods: Cross-sectional data from 998 farmers' fields across 20 countries across 4 AEZs (arid/semi-arid, humid, sub-humid, and highlands) in SSA and 3 different production systems: irrigated lowland, rainfed lowland, and rainfed upland were used to test hypotheses of nutrient concentration being estimable with a set of predictor variables among above-cited factors using linear mixed-effects regression models. Results: All 10 nutrients were reasonably predicted [Nakagawa's R2 ranging from 0.27 (Ca) to 0.79 (B), and modeling efficiency ranging from 0.178 (Ca) to 0.584 (B)]. However, only the estimation of K and B concentrations was satisfactory with a modeling efficiency superior to 0.5. The country variable contributed more to the variation of concentrations of these nutrients than AEZ and production systems in our best predictive models. There were greater positive relationships (up to 0.18 of difference in correlation coefficient R) between actual nutrient uptakes and model estimation-based uptakes than those between actual nutrient uptakes and average-based uptakes. Nevertheless, only the estimation of B uptake had significant improvement among all nutrients investigated. Conclusions: Our findings suggest that with the exception of B associated with high model EF and an improved uptake over the average-based uptake, estimates of the macronutrient and micronutrient uptakes in rice grain can be obtained simply by using average concentrations of each nutrient at the regional scale for SSA. Implications: Further investigation of other factors such as the timing of fertilizer applications, rice variety, occurrence of drought periods, and atmospheric CO2 concentration is warranted for improved prediction accuracy of nutrient concentrations.

Mineral-nutrient relationships in African soils assessed using cluster analysis of X-ray powder diffraction patterns and compositional methods.

Soil mineral compositions are often complex and spatially diverse, with each mineral exhibiting characteristic chemical properties that determine the intrinsic total concentration of soil nutrients and their phyto-availability. Defining soil mineral-nutrient relationships is therefore important for understanding the inherent fertility of soils for sustainable nutrient management, and data-driven approaches such as cluster analysis allow for these relations to be assessed in new detail. Here the fuzzy-c-means clustering algorithm was applied to an X-ray powder diffraction (XRPD) dataset of 935 soils from sub-Saharan Africa, with each diffractogram representing a digital signature of a soil's mineralogy. Nine mineralogically distinct clusters were objectively selected from the soil mineralogy continuum by retaining samples exceeding the 75 % quantile of the membership coefficients in each cluster, yielding a dataset of 239 soils. As such, samples within each cluster represented mineralogically similar soils from different agro-ecological environments of sub-Saharan Africa. Mineral quantification based on the mean diffractogram of each cluster illustrated substantial mineralogical diversity between the nine groups with respect to quartz, K-feldspar, plagioclase, Fe/Al/Ti-(hydr)oxides, phyllosilicates (1:1 and 2:1), ferromagnesians, and calcite. Mineral-nutrient relationships were defined using the clustered XRPD patterns and corresponding measurements of total and/or extractable (Mehlich-3) nutrient concentrations (B, Mg, K, Ca, Mn, Fe, Ni, Cu and Zn) in combination with log-ratio compositional data analysis. Fe/Al/Ti/Mn-(hydr)oxides and feldspars were found to be the primary control of total nutrient concentrations, whereas 2:1 phyllosilicates were the main source of all extractable nutrients except for Fe and Zn. Kaolin minerals were the most abundant phyllosilicate group within the dataset but did not represent a nutrient source, which reflects the lack of nutrients within their chemical composition and their low cation exchange capacity. Results highlight how the mineral composition controls the total nutrient reserves and their phyto-availability in soils of sub-Saharan Africa. The typical characterisation of soils and their parent material based on the clay particle size fraction (i.e. texture) and/or the overall silica component (i.e. acid and basic rock types) alone may therefore mask the intricacies of mineral contributions to soil nutrient concentrations.

Pre-treatment of soil X-ray powder diffraction data for cluster analysis.

X-ray powder diffraction (XRPD) is widely applied for the qualitative and quantitative analysis of soil mineralogy. In recent years, high-throughput XRPD has resulted in soil XRPD datasets containing thousands of samples. The efforts required for conventional approaches of soil XRPD data analysis are currently restrictive for such large data sets, resulting in a need for computational methods that can aid in defining soil property - soil mineralogy relationships. Cluster analysis of soil XRPD data represents a rapid method for grouping data into discrete classes based on mineralogical similarities, and thus allows for sets of mineralogically distinct soils to be defined and investigated in greater detail. Effective cluster analysis requires minimisation of sample-independent variation and maximisation of sample-dependent variation, which entails pre-treatment of XRPD data in order to correct for common aberrations associated with data collection. A 24 factorial design was used to investigate the most effective data pre-treatment protocol for the cluster analysis of XRPD data from 12 African soils, each analysed once by five different personnel. Sample-independent effects of displacement error, noise and signal intensity variation were pre-treated using peak alignment, binning and scaling, respectively. The sample-dependent effect of strongly diffracting minerals overwhelming the signal of weakly diffracting minerals was pre-treated using a square-root transformation. Without pre-treatment, the 60 XRPD measurements failed to provide informative clusters. Pre-treatment via peak alignment, square-root transformation, and scaling each resulted in significantly improved partitioning of the groups (p < 0.05). Data pre-treatment via binning reduced the computational demands of cluster analysis, but did not significantly affect the partitioning (p > 0.1). Applying all four pre-treatments proved to be the most suitable protocol for both non-hierarchical and hierarchical cluster analysis. Deducing such a protocol is considered a prerequisite to the wider application of cluster analysis in exploring soil property - soil mineralogy relationships in larger datasets.

Evaluating the utility of mid-infrared spectral subspaces for predicting soil properties.

We propose four methods for finding local subspaces in large spectral libraries. The proposed four methods include (a) cosine angle spectral matching; (b) hit quality index spectral matching; (c) self-organizing maps and (d) archetypal analysis methods. Then evaluate prediction accuracies for global and subspaces calibration models. These methods were tested on a mid-infrared spectral library containing 1907 soil samples collected from 19 different countries under the Africa Soil Information Service project. Calibration models for pH, Mehlich-3 Ca, Mehlich-3 Al, total carbon and clay soil properties were developed for the whole library and for the subspace. Root mean square error of prediction was used to evaluate predictive performance of subspace and global models. The root mean square error of prediction was computed using a one-third-holdout validation set. Effect of pretreating spectra with different methods was tested for 1st and 2nd derivative Savitzky-Golay algorithm, multiplicative scatter correction, standard normal variate and standard normal variate followed by detrending methods. In summary, the results show that global models outperformed the subspace models. We, therefore, conclude that global models are more accurate than the local models except in few cases. For instance, sand and clay root mean square error values from local models from archetypal analysis method were 50% poorer than the global models except for subspace models obtained using multiplicative scatter corrected spectra with which were 12% better. However, the subspace approach provides novel methods for discovering data pattern that may exist in large spectral libraries.

Spatial approach for diagnosis of yield-limiting nutrients in smallholder agroecosystem landscape using population-based farm survey data.

Adept use of fertilizers is critical if sustainable development goal two of zero hunger and agroecosystem resilience are to be achieved for African smallholder agroecosystems. These heterogeneous systems are characterized by poor soil health mainly attributed to soil nutrient depletion. However, conventional methods do not take into account spatial patterns across geographies within agroecosystems, which poses great challenges for targeted interventions of nutrient management. This study aimed to develop a novel population-based farm survey approach for diagnosing soil nutrient deficiencies. The approach embraces principles of land health surveillance of problem definition and rigorous sampling scheme. The advent of rapid soil testing techniques, like infrared spectroscopy, offers opportune avenues for high-density soil and plant characterization. A farm survey was conducted on 64 maize fields, to collect data on soil and plant tissue nutrient concentration and grain yield (GY) for maize crops, using hierarchical and purposive sampling. Correlations between soil test values with GY and biomass were established. The relationship between GY, soil NPK, and the tissue nutrient concentrations was evaluated to guide the setting up of localized critical soil test values. Diagnosis Recommendation Integrated System (DRIS) indices for total nitrogen (N), total phosphorus (P), and total potassium (K) were used to rank and map the prevalence of nutrient limitations. A positive correlation existed between plant tissue nutrient concentration with GY with R2 values of 0.089, 0.033, and 0.001 for NPK, respectively. Soil test cut-off values were 0.01%, 12 mg kg-1, 4.5 cmolc kg-1 for NPK, respectively, which varied slightly from established soil critical values for soil nutrient diagnostics. N and K were the most limiting nutrients for maize production in 67% of sampled fields. The study demonstrates that a population-based farm survey of crop fields can be a useful tool in nutrient diagnostics and setting priorities for site-specific fertilizer recommendations. A larger-scale application of the approach is warranted.

African soil properties and nutrients mapped at 30 m spatial resolution using two-scale ensemble machine learning.

Soil property and class maps for the continent of Africa were so far only available at very generalised scales, with many countries not mapped at all. Thanks to an increasing quantity and availability of soil samples collected at field point locations by various government and/or NGO funded projects, it is now possible to produce detailed pan-African maps of soil nutrients, including micro-nutrients at fine spatial resolutions. In this paper we describe production of a 30 m resolution Soil Information System of the African continent using, to date, the most comprehensive compilation of soil samples ([Formula: see text]) and Earth Observation data. We produced predictions for soil pH, organic carbon (C) and total nitrogen (N), total carbon, effective Cation Exchange Capacity (eCEC), extractable-phosphorus (P), potassium (K), calcium (Ca), magnesium (Mg), sulfur (S), sodium (Na), iron (Fe), zinc (Zn)-silt, clay and sand, stone content, bulk density and depth to bedrock, at three depths (0, 20 and 50 cm) and using 2-scale 3D Ensemble Machine Learning framework implemented in the mlr (Machine Learning in R) package. As covariate layers we used 250 m resolution (MODIS, PROBA-V and SM2RAIN products), and 30 m resolution (Sentinel-2, Landsat and DTM derivatives) images. Our fivefold spatial Cross-Validation results showed varying accuracy levels ranging from the best performing soil pH (CCC = 0.900) to more poorly predictable extractable phosphorus (CCC = 0.654) and sulphur (CCC = 0.708) and depth to bedrock. Sentinel-2 bands SWIR (B11, B12), NIR (B09, B8A), Landsat SWIR bands, and vertical depth derived from 30 m resolution DTM, were the overall most important 30 m resolution covariates. Climatic data images-SM2RAIN, bioclimatic variables and MODIS Land Surface Temperature-however, remained as the overall most important variables for predicting soil chemical variables at continental scale. This publicly available 30-m Soil Information System of Africa aims at supporting numerous applications, including soil and fertilizer policies and investments, agronomic advice to close yield gaps, environmental programs, or targeting of nutrition interventions.

Fertilizer response and nitrogen use efficiency in African smallholder maize farms.

Improving fertilizer recommendations for farmers is essential to increase food security in smallholder landscapes. Currently, blanket recommendations are provided across agro-ecological zones, although fertilizer response and nutrient use efficiency by maize crop are spatially variable. We aimed to identify factors that could help to refine fertilizer recommendation by analyzing the variability in fertilizer response (FR) and the agronomic nitrogen use efficiency (N-AE). A literature search for on-farm studies across Kenya and Sub-Sahara Africa (SSA), excluding Kenya, yielded 71 publications. The variability in FR was studied using a meta-analysis whereas key factors that influence FR and N-AE were studied with linear regression models. On average, the FR was 2, but it varied considerably from 1 to 28.5 (excluding outliers). In SSA, 18% of the plots were non-responsive plots with an FR < 1. The main factors affecting N-AE for Kenya were P-Olsen, silt content, soil pH, clay and rainfall, whereas only soil pH, exchangeable K and texture were important for SSA. However, our study indicates that available data on soil, climate and management factors could explain only a small part (< 33%) of the variation in FR and N-AE. Soil pH, P-Olsen, silt content, and rainfall had significant but low levels of power in explaining variation in FR and N-AE. Our findings indicate that strategies to refine fertilizer recommendation should include information on soil types and soil properties.

Mapping Soil Properties of Africa at 250 m Resolution: Random Forests Significantly Improve Current Predictions.

80% of arable land in Africa has low soil fertility and suffers from physical soil problems. Additionally, significant amounts of nutrients are lost every year due to unsustainable soil management practices. This is partially the result of insufficient use of soil management knowledge. To help bridge the soil information gap in Africa, the Africa Soil Information Service (AfSIS) project was established in 2008. Over the period 2008-2014, the AfSIS project compiled two point data sets: the Africa Soil Profiles (legacy) database and the AfSIS Sentinel Site database. These data sets contain over 28 thousand sampling locations and represent the most comprehensive soil sample data sets of the African continent to date. Utilizing these point data sets in combination with a large number of covariates, we have generated a series of spatial predictions of soil properties relevant to the agricultural management--organic carbon, pH, sand, silt and clay fractions, bulk density, cation-exchange capacity, total nitrogen, exchangeable acidity, Al content and exchangeable bases (Ca, K, Mg, Na). We specifically investigate differences between two predictive approaches: random forests and linear regression. Results of 5-fold cross-validation demonstrate that the random forests algorithm consistently outperforms the linear regression algorithm, with average decreases of 15-75% in Root Mean Squared Error (RMSE) across soil properties and depths. Fitting and running random forests models takes an order of magnitude more time and the modelling success is sensitive to artifacts in the input data, but as long as quality-controlled point data are provided, an increase in soil mapping accuracy can be expected. Results also indicate that globally predicted soil classes (USDA Soil Taxonomy, especially Alfisols and Mollisols) help improve continental scale soil property mapping, and are among the most important predictors. This indicates a promising potential for transferring pedological knowledge from data rich countries to countries with limited soil data.