Using precision agriculture to improve soil fertility management and on-farm research

(This is an expanded version of the article which appeared in the printed edition.)

Introduction

Precision agriculture technologies have potential to improve soil fertility management and on-farm research or demonstrations. Two major projects are being developed. One project focuses on the field-scale study of relationships between the phosphorus (P) and potassium (K) contents of soils and plants with grain yields and on the evaluation of soil sampling methods. No treatments are applied to the fields in this project. It is designed to address the expectation of many producers and agronomists in that grid sampling will adequately describe soil nutrient supplies better than the traditional "sampling by soil type" method and that variation in nutrient levels will explain much of the yield variability within a field. The other project focuses on the evaluation of variable-rate fertilization and on adapting new technologies to traditional on-farm strip trials. In this project, treatments are applied to long strips replicated several times across the fields. Intensive grid soil sampling is conducted before and after applying the treatments. The treatments compared vary between fields and include fertilizer placements (starter, deep-banding), interactions of herbicides and fertilization, variable-rate fertilization or manuring, and others. In both projects, yields are measured with calibrated yield monitors and, in some strip trials, the yield monitors are checked by weighing the yield of each strip.

Soil sampling for precision agriculture

The results of sampling many fields show that the spatial variability of P and K and other nutrients in soils is complex and that variability patterns differ markedly among fields. The causes for variability on a large scale are different from the causes of variability on a smaller scale. Soil types, landscape characteristics, previous crops, or proximity to feeding lots usually create variation over a scale of several acres. Practices such as tillage, fertilization, and manure application also create large variability on a scale of a few feet or even inches. Although in some fields the spatial variations of P, K, and pH tend to follow the distribution of soil types or other landscape characteristics, the variability of P or K (and sometimes of pH) usually do not follow the distribution of soil types and the patterns differ among fields. In many fields, the variability over many acres often is similar to that over a few hundred square feet.

Attempts to find an optimum soil sampling scheme valid across all fields have been largely unsuccessful. The maps in Fig. 1 show examples of the different answers a producer may get when different sampling schemes are used. The maps show assessments of soil P for three fields using three sampling methods. The data show that no general rule applies, and similar results have been observed for soil K and pH. Sometimes, intensive grid sampling results in a useful description of nutrient supplies. Often, however, sampling by soil type was as useful and it should make more economic sense. Data from most fields suggest that sampling of large cells three to four acres in size does not represent soil nutrient levels appropriately in many fields because the variation within those areas is as large as the variation over the entire field and cell borders usually do not follow soil mapping units or landscape. Sampling of large areas often overestimates nutrient levels and pH of significant areas of the fields. Increasing the number of soil cores collected for each composite sample will not solve this problem. Reducing the cell area or increasing the number of points sampled may not be economically viable, however. It is likely that a targeted grid sampling scheme that considers landscape characteristics or other field information is the most economical alternative. This procedure is flexible enough to adapt to different field characteristics and different intensities of sampling. Digitized soil maps, soil test data, yield maps, and aerial photographs (of bare soil and crop canopy) can be used to plan an efficient sampling scheme.

Figure 1. Distribution of soil-test P values in three fields assessed by three soil sampling methods.

The observation that economically feasible soil sampling procedures may not describe the variation in soil nutrients with as much detail as agronomists want is not new. Even traditionally recommended soil sampling methods have always compromised detail for economic feasibility. In spite of deficiencies, however, soil testing for P and K has proved successful as a method in which to base fertilizer recommendations. This is not different for grid soil sampling.

Soil sampling and variable-rate fertilization

The impact of grid sampling and variable-rate fertilization on soil fertility management and the profitabilty of crop production depends on several factors. Some include the nutrient levels found in relation to crop needs, the nutrient variability patterns, the fertilizer recommendations used, the expected responses to fertilization, and the additional costs. Results of four trials with corn and soybean in fields that tested optimum or high on average showed no major yield advantage of variable-rate P fertilization (or of uniform fertilization) in most fields because there was little response to P. This should not be surprising because surveys show that more than half of Iowa fields test optimum or above in P and K. High variability in a field with predominantly optimum or high values is likely to be irrelevant because the probability of yield responses in soils testing optimum or above is small and the proportion of low-testing areas is also small. Given the likelihood of small responsive areas in many Iowa fields, the most likely benefit of intensive grid sampling and variable-rate fertilization will be accomplished through savings in maintenance fertilization. Producers who will also benefit from these practices are those who realize that most parts of their fields test above-optimum and do not need maintenance fertilization until levels decreased to the optimum range. These outcomes will also be beneficial from environmental and sustainability perspectives. Without grid sampling, uncertainty usually leads producers to apply a uniform high maintenance rate over the field when it may not be needed.

Interpretation of soil-test and yield maps

Many producers believe that variation in nutrient levels will explain much of the yield variability within a field and that much can be learned from comparisons of several layers of information. Statistical analyses and visual observations of maps of soil-test values, soil types, and yields showed that only part of the yield variability in each field could be explained by the fertility measurements. This result could be expected because crop yields are influenced by a variety of factors. The field measurements that were related to yields varied among fields and high nutrient variation not necessarily explained highly variable yields. Although in a few instances low yields could be reasonably explained by the soil-test or soil type maps, use of tests and data management methods that cannot normally be used by producers showed that apparent correlations often were misleading and would lead to wrong conclusions. Questions concerning what nutrients limit yields cannot be answered with certainty unless treatments with and without fertilization are used. This concept lead to efforts in developing better methods for comparing management practices on a field scale.

On-farm comparison of management practices

Results of many on-farm comparisons show that precision agriculture technologies can be successfully adapted to on-farm, field-scale evaluations of alternative management practices. A commonly used method of on-farm research is based on the applications of treatments to long strips replicated across the field and on weighing large loads of grain. Use of grid sampling, differential global positioning systems (DGPS), yield monitors, and data management with geographical information systems (GIS) computer software allow for a more detailed evaluation of treatment differences for different parts of a field and for estimating interactions between response to fertilization and other growth factors.

Figure 2. Example of possible field measurements and data management when using precision agriculture technologies for on-farm evaluation of starter fertilization and herbicides for corn.

The data in Fig. 2 shows, as an example, the type of data handling that can be done by using grid sampling, DGPS, and yield monitors to study the response of corn to starter fertilization and two herbicides. This methodology allows for better evaluations of practices because commonly used statistical analyses can be improved by accounting for the spatial variation of yields. It also allows the study of treatment differences for different parts of the field having different soil test values, soil types, and other field characteristics. A minimum set of quality control procedures must be followed, however, to reduce errors due to DGPS problems, use of yield monitors, and border effects of the treatments compared. Common errors are inappropriate calibration of the yield monitor, wrongly georeferenced yield points and field strips, errors due to field borders, waterways or grass strips, and changes of the width of the harvested swath (a frequent problem with soybeans). It must be emphasized that valid conclusions concerning differences between treatments applied to strips are possible only when treatments are replicated across the field, independently of the length or the width of the strip used.

General recommendations

The results up to this time show that no general recommendation is valid for all fields. The variability of soil nutrients is field-specific and, ideally, each field should be sampled and fertilized differently. Informed estimates of the benefit of grid soil sampling and variable-rate fertilization require knowledge of each field. Intensive grid soil sampling and variable-rate fertilization will probably result in significant economic benefits when variation within a field is such that large proportions of the field test below and above optimum. Unfortunately, the optimum scheme and the merits of variable-rate fertilization can be reasonably estimated only after conducting an intensive expensive preliminary sampling. A reasonable alternative is to use available field information to improve the traditional "sampling by soil map unit" method and conduct a more intensive but still cost-effective sampling. Previous soil-test data, yield maps, aerial photographs, and field histories can be used to target specific areas for intensive sampling. This approach is likely to increase economic benefits to producers and will still exploit the potential of new precision farming technologies.

Precision agriculture technologies can be used to evaluate and demonstrate alternative fertilization or other management practices on the basis of on-farm strip trials. This methodology is useful for new research, to adjust recommendations for local conditions, and to understand the causes of yield variability better. Producers should interpret with caution maps of soil-test values, soil types, other soil or plant measurements, and grain yields. Questions concerning which nutrients limit yields cannot be answered with certainty unless fertilization treatments are used. Use of the variation usually observed within fields (in soils, diseases, yields, and other factors) in combination with replicated applications of treatments is useful to identify nutrient deficiencies and to study the interactions of fertilization with other growth factors. This possibility does not preclude the need for replicating the comparisons in various fields and years. The preliminary results show that a minimum set of quality control procedures must be followed to reduce errors, however. Completion of ongoing studies will result in detailed protocols that producers can follow to conduct successful on-farm comparisons.

This article originally appeared on pages 12-14 of the IC-480 (4c Precision Ag Edition) -- April 9, 1998 issue.

Updated 11/10/2006 - 11:57am