How We Selected
Soil Samples for Soil Nutrient Analysis

The field covered an area of 66.36 hectares. The soil is primarily slightly degraded chernozems, but there are also parts of highly degraded and podsolized chernozems. The relief isn’t homogenous. The elevation difference here is 41 meters with a 6-degree slope.

As a reminder, Frendt and a third-party lab conducted the soil nutrient analysis.

1. They identified points where samples needed to be taken using GIS software. Then they uploaded the points to a GNSS receiver.
2. A total of 15 soil probes were made in a 9-meter radius from each point.
3. They put together mixed samples from each group of samples collected at one point.
4. Those samples were sent off to the lab to be analyzed.

A total of 236 samples were taken. That’s about 1 sample per 0.28 hectares.

The lab used the Chirikov method to identify mobile phosphorus and potassium.

Here’s how to understand that. As we see, the soil’s nutrient properties in the field vary greatly. For example, phosphorus content at different points differed over 30-fold. The difference was 9-fold for potassium. There were six groups of soil acidity in the field. These groups are defined by their pH level, from strong acidic to low acidic and neutral.

Now back to the problem at hand: money. On average, it costs $ 45 to analyze one sample. Add, on top of that, the cost of the sampling itself. With a 0.28-ha sample area, that turns into a hefty sum. And that’s just for one field.

Now let's pretend that Frendt doesn't exist. We want to calculate the fertilizer needed for this field, so we decide to conduct a soil nutrient analysis. How will we select soil samples?

Zone sampling. We divide the field into equal-sized zones, take samples from each zone (in a Z shape, diagonally, or randomly), mix them up, and then apply fertilizer based on the average in each zone.

Grid sampling and data interpolation. We take the samples using a grid and, based on the values in specific points, estimate the entire field's phosphorus and potassium level. This is interpolation. A special software program creates the estimates for us. I use ArcGIS, for example.

Now let's look at how that would work in our field.

Phosphorus and potassium distribution maps
Let’s say that we divide the field into 6.6-ha sample zones. We end up with 10 of those. We take a few samples in each one, mix them, and end up with averaged results. Let’s compare them with the reliable map that we got with a 0.28-ha sample grid.

Here’s how to understand that. There’s high variability in phosphorus and potassium in most field plots. The optimum difference in mineral content inside a plot is 50 mg. In 7 out of 10 plots, the difference between the minimum and maximum value for phosphorus and potassium is 4−5 times higher than the norm.

It’s risky to apply fertilizer based on this map. Yields can drop dramatically due to a lack or excess of nutrients in different parts of the field.

Potassium distribution maps
Now let’s imagine that we’ve conducted grid sampling and interpolated the data. These are how potassium distribution maps would look like with 0.4-, 0.9-, and 2-ha grids. I won’t show the phosphorus distribution maps. The trend there is the same.

Here’s how to understand that. The lower the grid’s density is, the more significantly the map will differ from the reliable one (i.e., the map from Frendt being used as an example) and the lower its accuracy is. For example, a map with a grid density of 2 hectares only captures general potassium distribution trends in the field. Again, it’s risky to apply fertilizer based on it or based on a map with a lower-density grid.

There’s no 'right' way to conduct soil sampling because it’s not about the method, but the sampling density.

The soil’s nutrient properties in the field vary greatly. To determine the exact nutrient content in the field, we have to conduct a detailed soil test. More details mean better quality… and a higher price tag.

A reliable soil nutrient analysis is expensive. On average, it costs $ 45 to analyze one sample. According to my calculations, using variable-rate fertilizer application is only worth it in this field if you’re using at least a 0.9-ha grid. In this case, a soil test would cost $ 3,000. The test will cost more in correlation with the field’s variability and area.

There are a few options.

Pull out your wallet and use a dense grid when selecting samples to get a reliable nutrient distribution map.

Hope you're lucky and do it by eye. But in the experience I talk about in the VRA section, there were 10−15 fields with higher variability. So I personally don't believe in luck when it comes to this approach.

Use OneSoil. We calculate how much fertilizer is needed based on productivity zones. We use 5 years of satellite images to identify them and determine relative yield over several years — all for free. Here's how that works, and here's the app itself.