Location: Dale Bumpers Small Farms Research CenterTitle: The anatomy of uncertainty for soil pH measurements and predictions: Implications for modellers and practitioners
|LIBOHOVA, ZAMIR - Natural Resources Conservation Service (NRCS, USDA)|
|SEYBOLD, CATHY - Natural Resources Conservation Service (NRCS, USDA)|
|ADHIKARI, KABINDRA - University Of Arkansas|
|WILLS, SKYE - Natural Resources Conservation Service (NRCS, USDA)|
|BEAUDETTE, DYLAN - Natural Resources Conservation Service (NRCS, USDA)|
|PEASLEE, STEVE - Natural Resources Conservation Service (NRCS, USDA)|
|LINDBO, DAVID - Natural Resources Conservation Service (NRCS, USDA)|
Submitted to: European Journal of Soil Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/10/2018
Publication Date: 11/24/2018
Citation: Libohova, Z., Seybold, C., Adhikari, K., Wills, S., Beaudette, D., Peaslee, S., Lindbo, D., Owens, P.R. 2018. The anatomy of uncertainty for soil pH measurements and predictions: Implications for modellers and practitioners. European Journal of Soil Science. 70:185-199. https://doi.org/10.1111/ejss.12770.
Interpretive Summary: Data is available in multiple forms and data is used to create soil maps. All data have some errrors in measurement between samples or between labs using different techniques. This paper focused on the errors and how large the errors could be when making predictions from maps. To evaluate this issue, the measurement of pH was used as an example. Overall, the error found between laboratories varied by about 0.5 pH units. The pH varied depending on the method used and the greatest difference was between the pH measured in water compared to pH measured using a weak calcium chloride solution. When making liming recommendations, most states have rates based on 0.1 pH units. This research indicates that liming recommendations can be affected by the errors in the measured data. The research is continuing to develop methods to manage the differences found when translating the laboratory measurements to field maps.
Technical Abstract: Spatial predictions of soil properties require the assessment of error predictions compared to measured values. The objective of this study was to assess the magnitude of error for soil pH associated with different sources and the implications for management using the U.S. National databases. Error sources included measurement methods, laboratory conditions, pedotransfer functions, database transections, positional accuracy, and aggregations methods. The mean soil pH 1:1W and 1:5W combined was 6.3 and significantly higher than the mean soil pH 1:2CaCl2 and 1:5CaCl2 combined (5.7). The Root Mean Square Error (RMSE) for mid infrared (MIR) was 0.40 for pH 1:1W and 0.32 for soil pH 1:2CaCl2. The RMSE for between laboratory reproducibility (R) varied from 0.50 (pH 1:1W) to 0.68 pH units (pH 1:2CaCl2), and was greater than within laboratory reproducibility (RL) (RL-pH 1:1W, 0.34 pH units) and repeatability (r - pH 1:2CaCl2, 0.04 pH units). The RMSE for relationships of soil pH 1:1W vs. pH 1:5W and pH 1:2CaCl2 vs. 1:5CaCl2 varied from 0.27 to 0.43. The RMSE for the soil profile depth aggregation (weighted mean vs. equal-area spline) was 0.36. The RMSE for soil pH 1:1W between locational position comparing Global Positioning System (GPS) and Public Land Survey System (PLSS) was 0.57. Spatial aggregation methods had the highest RMSE varying from 0.58 to 1.3 pH units. The error size influences model predictions and associated management recommendations. Soil liming recommendations based on 0.1 pH increments do not reflect error measurements and/or spatial prediction uncertainty. Recognizing the error and utilizing this knowledge can aid users of model output to make better recommendations based on model predictions.