ISRIC Report 2011/01: Overview of soil phosphorus data from a large international soil database

isric_report_2011_01.pdf (pdf, 3.80 MB)
XLS-dataset (ISRIC Report 2011/01) (, 14.53 MB)
Year of publication
Niels H Batjes
Document tags
This exploratory study presents an inventory of extractable soil phosphorus (P-Bray, P-Olsen, P-Mehlich and P-water) and P-retention (P New Zealand) data held in a publicly available version of the National Cooperative Soil Characterization Database (USDA-SSL). The primary aim is to assess whether representative P-values by broad soil group (FAO system) can be determined for each of these analytical methods. In the affirmative, such derived data may be linked to global soil geographic databases, that consider the FAO Legend for demarcating major soil groups, in order to be used for model-based assessments of food security.
For many soil properties, FAO soil groups have been shown to be useful carriers of soil information. The primary data set, containing over 35 000 soil profiles, was screened to identify profiles suitable for this assessment. These profiles should be geo-referenced, classified at Great Group level, according to USDA Soil Taxonomy, to permit correlation with major soil groups of the FAO system.
Further, they should include P-data, analysed according to defined analytical methods. Following soil correlation and screening of the primary data for possible inconsistencies, depth-weighted soil properties (0-20, 20-40, 40-60, 60-80 and 80-100 cm) were stored in a secondary data set. Some 4920 of the original profiles, from 48 different countries representing 25 FAO major soil groups, have P-data for some horizons. In accordance with pedological practice, different analytical procedures were used by USDA-SSL largely depending on soil acidity and mineralogy.
For 0-20 cm depth, 2733 profiles have data for P-retention, 2680 for P-Bray1, 289 for P-Mehlich3, 189 for P-Olsen and 85 for P measured in water. Generally, there a fewer observations for the deeper layers. Median and mean P-values are presented for each FAO soil group and depth layer, further stratified according to soil textural class.
Coefficients of variation (CV) in excess of 100% are observed for extractable-P, irrespective of the extraction method, for most FAO soil groups. For P-Bray1, the average CV is 134% for 0-20 cm, with extremes of 71% and 224%. CVs for P-retention range from 20% to 72% (for 0-20 cm) for the major soil groups under consideration. When considering the upper 60 cm (i.e., 0-20, 20-40 and 40-60 cm) for mineral soil units, average P-retention (wt%) decreases in the following sequence: Andosols (74 (Avg.) ± 28 (SD)), Ferralsols (57 ±19), Acrisols (48 ± 22) and Podzols (46 ± 27). Overall, this pattern is in accordance with soil P-fixing properties reported in earlier studies. The present set of median P-values, by major soil grouping and depth layer, is not considered representative for any specific geographic area or country; as such, it should not be used for spatial extrapolation using the ISRIC-WISE derived soil property geodatabase and similar.
Similarly, best subset regressions that consider extractable-P (determined according to a given analytical method) as a multilinear function of several independent predictor soil variables had low predictive capability with most adjusted R2 values being < 0.3.
In view of these findings, and the fact that actual amounts of plant available-P will be affected by crop type, farmers‘ nutrient management and soil conservation
practices, a practical solution may be needed for defining available P-levels, for specific soil units, as required for exploratory assessments of resource scarcity and food security at a broad scale. One approach could be to propose P levels (for defined analytical methods) below which a marked fertilizer response is likely to occur
according to published sources.
Using such simple assumptions as a first approximation, "windows- of-opportunity" for fertilizer application in inherently unfertile and nutrient depleted regions could be estimated, in combination with auxiliary GIS layers on regional P-fertilizer use, subject to the availability of a regionally validated model. The ultimate purpose of such an exploratory analysis would be to improve nutrient input strategies for raising food production in nutrient-limited cropping systems.
Keywords: data set, soil phosphorus, P-Bray, P-Olsen, P-Mehlich, P-Water, P-retention, uncertainty, extrapolation