Soil tests are not perfect, so a soil-test value should be considered not as a single value, but as one within a range. These are some of the reasons why a soil test is imperfect: 1) it represents a one-point-in-time measurement, whereas the crop takes nutrients through an extended period of time and typically under very different soil-water and temperature conditions than at the time of sampling; 2) the information generated typically comes from a sample in the plow layer, but crop roots extract nutrients below that layer; and 3) laboratory precision, which is typically within 5 percent to 10 percent of the true value. Despite these imperfections, soil testing is the most important guide to profitable application of P, K and lime because it provides a framework to determine the fertility status of the field.

Of course a soil-test report can be only as accurate as the sample sent for analysis. In fact, the spatial variability of available nutrients in a field makes soil sampling the most common and greatest source of error in a soil test. To collect samples that will provide an accurate measurement of the fertility of an area, be sure to determine the sampling distribution, collect samples to the proper depth, collect samples from precisely the same areas of the field that were sampled in the past, and collect samples at the proper time.

Sampling Distribution
The number of soil samples taken from a field is a compromise between what should be done (information) and what can be done (cost). The most common mistake is taking too few samples to represent a field adequately. Shortcuts in sampling may produce unreliable results and lead to higher fertilizer costs, lower returns, or both. For whole-field uniform fertilizer applications, typically sampling at the rate of one composite (at least five soil cores taken from within a 10-foot radius around the sampling point) from each 2-1/2-acre area gives a general indication of the fertility of that field. For site-specific applications or for fields where large variations in test values over a short distance are suspected, it is better to collect one sample from each 1.1-acre area. For conservation tillage fields where fertilizers are applied in a band, there is no universally accepted method to collect soil samples. Typically it is suggested that if the location of the band is unknown, the number of samples to make a composite should be increased to 20 or 30 and the sampling position should be varied in relation to the row so the band does not bias test results. If the position of the band is known, then it is typically suggested to stay a few inches away from it or to take a large number of samples off the band for each sample taken in the fertilizer band. Another method of soil sampling a field is to collect samples from zones with common characteristics. This is a directed sampling approach (also known as smart sampling). This method integrates precision agriculture data (yield maps, crop canopy data, soil data, past management history information, etc.) and defines sampling zones with common characteristics that may be influencing crop productivity.

Source: Fabian G. Fernandez, Illinois Agronomy Handbook