Turfgrass soil sampling, part 6 of 7

Now this series gets interesting. I’ve reviewed what I do, what a Twitter survey says people are doing, what Rutgers and Penn State recommend, and what an intensively sampled lawn in Virginia suggests would be an appropriate number of subsamples to combine in composite samples for turfgrass.

The article Guiding soil sampling strategies using classical and spatial statistics: A review by Lawrence et al. recommends doing things differently. In short, they recommend that “soil samples should not be composited.” I recommend reading the full article if you are interested. The article is about testing agricultural fields. It is quite relevant to turfgrass sites too. Below are some quotes from the article, with notes from me interspersed. Any italicized or bold text is mine.

They get right to the purpose of soil nutrient analysis in the first sentence saying that it is “a key practice to increase the efficiency of nutrient management.” As an aside, I’ll take the opportunity here to point out that that is how soil testing is supposed to work, but in the turfgrass industry, soil testing seems to actually decrease the efficiency of nutrient management because of misinterpreted soil tests. But it is easy to fix this with a method such as MLSN.

In their review, Lawrence et al. looked at results from other articles, and calculated “an estimate for the range of soil sampling densities that would be required to achieve a margin of error of 10% of each study’s mean at a 5% precision level.” For K, the median sampling density is less than 5 samples per hectare; for P the median sampling density is 8.4 samples per hectare. I think it is interesting to consider these sampling densities in comparison to what is standard in the turfgrass industry today. To achieve that aforementioned level of precision with P, for example, one would need on average a single sample for every 1,190 m2 (12,809 ft2). With K, fewer samples would be required.

“Compositing soil samples before analysis is a common method for reducing soil analysis costs … Compositing, which effectively calculates the mean of a number of soil cores, is unlikely to represent the true population parameter of interest and will typically present a higher estimate of nutrient concentration than is accurate … the distribution of a nutrient would be positively skewed … For fertilizer management, this may cause the farmer [or the turfgrass manager] to apply an inadequate amount of fertilizer

Then in a section on the “Current State of Extension Soil Sampling Recommendations,” this:

“54% of extension sources suggested ‘zig-zag,’ ‘Z’, or ‘W’ sampling, with 44% mentioning the need to take ‘representative’ samples. This suggestion is at odds with the requirements of a design-based approach in which random selection of locations is paramount. As a result, it is highly likely that these methods often result in biased results, especially if the samples are composited before analysis and the soil property is lognormally distributed.

In the general suggestions and conclusion to the article, more things to think about:

“If no prior information on the soil property is available, log-normality should be assumed, and soil samples should not be composited. If the collected cores display log-normality, the geometric mean or median should be used instead of the arithmetic mean. Only in instances in which normality has previously been established should soil samples be composited.

“… it is not surprising that extension recommendations often suggest practices such as compositing and Z-sampling, which do not have a strong foundation in peer-reviewed literature. Fortunately, there are theoretically simple ways to correct these recommendations, such as suggesting soil cores to be individually analyzed.

Micah Woods
Micah Woods

Scientist, author, consultant, and founder of the Asian Turfgrass Center