Soil nutrient maps of Sub-Saharan Africa: assessment of soil nutrient content at 250 m spatial resolution using machine learning

Abstract

Spatial predictions of soil macro and micro-nutrient content across Sub-Saharan Africa at 250 m spatial resolution and for 0-30 cm depth interval are presented. Predictions were produced for 15 target nutrients: organic carbon (C) and total (organic) nitrogen (N), total phosphorus (P), and extractable-phosphorus (P), potassium (K), calcium (Ca), magnesium (Mg), sulfur (S), sodium (Na), iron (Fe), manganese (Mn), zinc (Zn), copper (Cu), aluminum (Al) and boron (B). Model training was performed using soil samples from ca. 59,000 locations (a compilation of soil samples from the AfSIS, EthioSIS, One Acre Fund, VitalSigns and legacy soil data) and an extensive stack of remote sensing covariates in addition to landform, lithologic and land cover maps. An ensemble model was then created for each nutrient from two machine learning algorithms- random forest and gradient boosting, as implemented in R packages ranger and xgboost-and then used to generate predictions in a fully-optimized computing system. Cross-validation revealed that apart from S, P and B, significant models can be produced for most targeted nutrients (R-square between 40-85%). Further comparison with OFRA field trial database shows that soil nutrients are indeed critical for agricultural development, with Mn, Zn, Al, B and Na, appearing as the most important nutrients for predicting crop yield. A limiting factor for mapping nutrients using the existing point data in Africa appears to be (1) the high spatial clustering of sampling locations, and (2) missing more detailed parent material/geological maps. Logical steps towards improving prediction accuracies include: further collection of input (training) point samples, further harmonization of measurement methods, addition of more detailed covariates specific to Africa, and implementation of a full spatiotemporal statistical modeling framework.

Comparing Machine Learning Models and Hybrid Geostatistical Methods Using Environmental and Soil Covariates for Soil pH Prediction

“…The introduction of kriging interpolation of the ML models' residuals did not really improve the results. This is in accordance with the findings by Henglet al [41], stating that ". .…”

Quantifying Variability in Maize Yield Response to Nutrient Applications in the Northern Nigerian Savanna

“…Most Nigerian soils are highly weathered with low activity clays (such as kaolinite) which makes them more vulnerable to fertility degradation under continuous arable use with poor nutrient replenishment [10][11][12]. Deficiencies in soil primary macronutrients (particularly N, P and K) are widespread and have been reported in most parts of the country [9,[13][14][15]. In addition, deficiencies of S and some micronutrients have also been reported in some Nigerian savanna soils [16][17][18].…”

Standardised soil profile data to support global mapping and modelling (WoSIS snapshot 2019)

“…sand, silt, clay and pH). Having a soil (taxonomic) classi- (Hengl et al, 2017b). b Such profiles (geo-referenced solely according to their country of origin) may be useful for developing pedotransfer functions.…”

“…It is well known that "soil observations used for calibration and interpolation are themselves not error free" (Baroni et al, 2017;Cressie and Kornak, 2003;Folberth et al, 2016;Grimm and Behrens, 2010;Guevara et al, 2018;Hengl et al, 2017b;Heuvelink, 2014;Heuvelink and Brown, 2006). Hence, we provide measures for the geographic accuracy of the point locations as well as the accuracy of the laboratory measurements for possible consideration in digital soil mapping and subsequent earth system modelling (Dai et al, 2019).…”

“…Quality-assessed data provided through WoSIS can be (and have been) used for various purposes. For example, as point data for making soil property maps at various spatialscale levels, using digital soil mapping techniques (Arrouays et al, 2017;Guevara et al, 2018;Hengl et al, 2017a, b;Moulatlet et al, 2017). Such property maps, for example, can be used to study global effects of soil and climate on leaf photosynthetic traits and rates (Maire et al, 2015), generate maps of root zone plant-available water capacity (Leenaars et al, 2018) in support of yield gap analyses (van Ittersum et al, 2013), assess impacts of long-term human land use on world soil carbon stocks (Sanderman et al, 2017), or the effects of tillage practices on soil gaseous emissions (Lutz et al, 2019).…”