Jarrod Miller and James Adkins, University of Delaware
Variability in soil land landscape characteristics reduces yield response to management techniques, particularly regarding seeding rates and fertilizer additions. Yield maps provide a spatial map of yield, which can be associated with drainage issues, soil nutrient holding, or nutrient concentrations. One method to uncover soil variability and crop response is to use precision soil sampling, including either grid or zone methods. Both increase the cost of taking soil samples, and each have their value depending on the desired outcomes.
For grid sampling, prior work in other states has shown advantages of grids no larger than 2.5-acres, with increased accuracy at 1-acre. To observed how accurate grids must be on the Delmarva peninsula, the Maryland Grain Producers sponsored a project at the University of Delaware Warrington Irrigation Research farm. This farm has variable rate pivot and variable rate linear irrigation (Figure 1a). A Veris was used to map soil EC on 90-foot centers (Figure 1b), and then the field was sampled on a 0.25 acre grid spacing. These grid samples match up with the center of our 90 by 90 foot treatment plots, outlined on our center pivot and linear plots (Figure 1). Soil samples were taken from the upper 8-inches of soil from the center of each grid in April 2022 and submitted to the University of Delaware Soil Testing lab for analyses.
Grid Sampling Results
Grid Sampling Correlations
The densest sampling scheme (¼ acre) was strongly correlated with soil characteristics and nutrient contents at the next densest sampling of ¾ acres. The ¼ acre sampling was not as strongly related to 1 ½ or 3-acre sampling schemes, particularly for organic matter and CEC measurements. This is important, considering that organic matter and the CEC are unlikely to change over the long term, so a more accurate map could last a longer period. The 3-acre grid (Table 1) also had moderate to weak correlations to the ¾ and 1 ½ acre grids. In some cases correlations were not significant (CEC, OM, and Mg), so the relationships to general soil characteristics are again better represented by a denser grid, while nutrients like P and K have moderate relationships based on grid size.
Grid Sampling Average Plot Levels
Using our 90×90 foot plot maps (Figure 1), we averaged soil nutrient levels based on each grid sampling density (Table 2). It is important to know that grid density was changed by deleting points from the ¼ acre grids, and different placement of points would also change these results. For example, the average CEC was greatest based on 1.5 acre grids but was similar across the other treatments. However, if the sampling point location changes, these averaged may also change.
Organic matter also averaged higher in our plot maps in the 1 ½ and 3-acre grids, but was lower in ¼ and ¾ acre grids. So in this study, lower resolution (>1.5 acre grids) increased OM estimates when averaged across small areas, similar to setting up zones in the field. There were no differences observed in plot pH based on sample size, but K, Ca, and Mg were all higher in the 1 ½ acre grids, and Ca was lowest in the 3-acre grids (Table 2). However, the differences were minor in terms of determining fertilizer application.
Interpolated Maps of Different Soil Properties and Nutrient Concentrations
Maps of soil properties can be visually striking, without any additional analyses. When comparing CEC by sampling density (Figure 2), the loss in resolution is clear when you drop to grids above ¾ acre. On the coastal plain, where CEC can range from extremely low values (< 4 meq/100g), accuracy could help with variable rate management of leachable nutrients, particularly potassium in these soils.
Results for soil pH are similar (Figure 3), where pH is most accurate at ¼ density, but with a large loss in resolution when grids reach 1 ½ acres. What is particularly striking about the pH map for the ¼ grids is the blocked pattern on the lower half of the pivot, following the shape of the research plots, indicating past management. This pattern is not present in the lower density sampling. Grid sampling is often described as best at finding past management issues but may miss the smallest details above ¼ acre grids. The high density sampling also produces the most accurate lime rate map (Figure 4), which has a range of 1-6 tons per acre, based on UD recommendations.
The ¾ acre grid had similar relationships to the ¼ in both nutrient correlations and plot average extractions. This project reveals interesting patterns and corroborates that grid sizes should be less than 2.5 acres, and preferably less than an acre.
Future analyses will include the response to variable rate lime and K application, as well as variation in micronutrients.
This research was sponsored by the Maryland Grain Producers Utilization Board (http://www.marylandgrain.org/).