Cost-effective sampling strategies for soil management

Abstract

Efficient fertiliser management requires information about the nutrient status of each management area or field. This information can be gathered by observing soil nutrients at a number of sites in the field. The quality of this information is dependent on the sampling strategy that is employed. The sampling strategies suggested in fertiliser recommendations such as RB209 are generally based on anecdotal evidence regarding the number of soil cores required or are designed to ensure that the errors in estimating soil nutrient concentrations are less than an arbitrarily defined threshold. Such strategies do not directly link the sampling effort to the consequences of erroneous soil nutrient information, which may include reduced profitability or the long term development of nutrient excess or deficiency. We develop a quantitative framework to study the effectiveness of different sampling designs so that rational sampling recommendations for phosphorus (P), potassium (K) and nitrogen (N) can be developed.

For all nutrients, current recommendations suggest that measurements should be regularly spaced on a 'W' design which covers the field. Four alternatives to the 'W' are tested: an optimized sample configuration, stratified random sampling, rank set sampling and a clustered or bad practice design. We quantify the errors associated with each design, determine the management decisions that will be made by the farmer based on this erroneous information and then model the effects of these decisions. Thus we are able to relate the resources devoted to sampling to the expected profitability or long-term nutrient status of the field.
Our study shows, that for a particular sampling effort, sample designs can be optimized to give smaller errors than the 'W' design. However we also find that the errors from estimating soil-nutrient status with a 'W' are not large enough to substantially affect the quality of soil nutrient management. This is because once a certain accuracy in estimating soil-nutrient concentrations has been achieved, the quality of the management recommendations are limited by other sources of uncertainty in predicting the amount of nutrients the crops will access from the soil. Therefore the benefits of using optimized designs do not outweigh the extra complexity which they entail. If in the future fertiliser recommendations are more sensitive to soil information, say for example if nitrous oxide emissions had to be carefully controlled, then the use of optimized sample designs should be re-explored.

We find that in the scenarios explored in this project, decisions regarding K require less accurate information than P. A bulked sample every four years of 10 soil cores is sufficient to maintain both soil P and K stocks within a target range. This is less than half of the number of cores which is currently recommended. For N, rational sampling effort varies according to the expected SNS in the field, and the field size. Bulked samples of 10-15 soil cores are adequate for most fields. Including more than 10 cores in the bulked sample is warranted when fields are larger than 20 ha or if SNS is expected to be high (>160 kg/ha). The largest financial benefit from sampling occurs when soil nitrogen supply is around 175 kg/ha since at these concentrations the yield is most sensitive to sampling errors and erroneous decisions. There is a smaller benefit when the expected SMN is much larger or much smaller since in these circumstances it is clear that either a small or large amount of N fertiliser should be added. We determine the circumstances in which SMN measurement and the use of barometer fields are cost effective in comparison to prior knowledge of SMN. We do not consider alternative methods of estimating soil nitrogen supply such as the Field Assessment Method.

The framework developed in this project is not only suitable to assess the cost-effectiveness of different sampling designs under current fertiliser recommendations but also to develop and assess the cost-effectiveness of modifications to these recommendations.