Back to Contents6.3.1 Soil Sampling, Soil Surveys, and Sensors for Precision Agriculture
Investigators: Tom Mueller, Scott Shearer, Anastasios Karathanasis, Ed Perfect, Mark Coyne, Ken Wells, Dick Barnhisel, Jim Thompson, Carl Dillon, and Michael Collins.
Precision agriculture technologies have allowed farm fields to be managed at finer and finer scales. Because the scale of measurement is related to the costs and profitability of grid soil sampling, it is important to understand the relationship between map quality and the intensity of sampling. The industry standard sampling intensity is 330-ft (2.5 acre) grids which may or may not be adequate for describing soil properties in Kentucky agricultural fields. Map accuracy is important because nutrient maps are the basis for variable rate fertilization.
Since a growing number of farmers across the United States have access to sensors (e.g. terrain modeling equipment, Veris® electrical conductivity sensor), methods should be developed that would allow this information to be used to improve the accuracy of nutrient maps. Additionally, this information could be used in conjunction with first order (1:6,000) soil surveys to improve site specific management and reduce its overall costs.
Grain yield maps are important because they can be used to identify regions of a farm field that may respond to changes in management. Therefore, it is important to understand how soil properties and topography affect grain yield. Because it is not currently economical to intensively measure soil physical properties in an agricultural field, it is important to relate soil properties to sensor measurements, digital elevation models (DEMs), and first order soil surveys. Further it is important to understand the relationship between DEMs and sensors and grain yield.
The first sub-objective of this study is to assess the accuracy of soil property maps for site-specific nutrient management. The second sub-objective is to determine if soil properties can be enhanced with terrain attributes and sensor information. The third sub-objective is to model the relationships between grain yield maps, soil properties, landscape properties, and sensor measurements. The fourth sub-objective is to assess the use of first order soil surveys for precision agriculture. A detailed description of each objective will be given in the subsequent text.
The first sub-objective of this study is to assess the accuracy of soil property maps for site-specific nutrient management. Specifically, interpolation procedures, scales of sampling, and sampling strategies (grid vs. zone sampling) will be compared. Three to five fields will be chosen that have a range of soil characteristics and that have several grain yield maps. Soil samples will be obtained using (i) three grid sampling scales (100, 200, and 330 ft grid increments), (ii) zones based on NRCS soil type and management history, (iii) and a random samples design. The random sampling approach will be used to assess the accuracy of the maps. Samples will be analyzed for pH, P, K, Ca, Mg, CEC, total carbon, soil water content, sand, silt, and clay content. In addition, for one of the fields, a microbial analysis of potential metabolic response (using BiologTM Microfilter Plates) will be conducted on extracts from each of the soil samples. BiologTM plates are used to assess the diversity of the soil microbial community by means of its substrate use, which reflects the community’s capacity to function. This capacity is of agronomic significance because it affects soil productivity. Also in this field, cores will be pulled and bulk density, saturated hydraulic conductivity, and water characteristic curves will be measured. All of the spatial data will be analyzed geostatistically and interpolated maps will be evaluated using spatial measures of map accuracy (root mean squared error, bias, prediction efficiency). An economic analysis will be performed to determine the efficacy of the different sampling and mapping strategies. Most of the sampling will be done by departmental technicians, the investigators, and county agricultural agents during the first year. Some of the soil physical property sampling will be done in the second year. Laboratory analyses will be conducted during all three years. This objective will be primarily addressed during the first and second years.
The second sub-objective is to determine if spatial estimates of soil properties can be enhanced with terrain attributes and sensors information (electrical conductivity, ground penetrating radar). In each of the fields, elevation maps will be created using a survey grade GPS system. The elevation data will be used to create digital terrain models (DEMs). A Veris® conductivity sensor will be used to create soil electrical conductivity maps. A number of geo-statistical techniques (co-kriging, kriging with an external drift, random field analysis) will be used to map soil properties. The validation data sets as described above will be used to determine how well secondary terrain and sensor data can be used to improve spatial estimates of soil properties. The GPS elevation and the Veris® electrical conductivity surveys will be conducted during the first two years. This objective will be mainly addressed during the second and third years. Terrain modeling will be conducted by Tom Mueller and Jim Thompson.
The third sub-objective of this study is to statistically model the relationships between grain yield map data, soil properties (e.g. pH, P, K, Ca, Mg, total carbon, NH4-N, NO3-N, soil water content, sand, silt, clay content, bulk density, and water characteristic curves), landscape properties (e.g. slope gradient, aspect, curvature, specific catchment area), and sensor measurements (e.g. electrical conductivity, ground penetrating radar). Multivariate techniques (e.g. multiple regression, principle components, and canonical correlations) will be used to model and describe the relationships between these properties. Computational analysis will be performed during each of the three years by Tom Mueller and Jim Thompson.
The fourth sub-objective has several parts: to determine if first order (1:6,000) soil surveys a) can be used to explain the spatial and temporal variability of grain yield, b) can be used in conjunction with minimal soil fertility information to improve their explanatory value, thereby reducing the scale and cost of soil fertility sampling, c) can be enhanced with digital terrain information and electrical conductivity maps, and d) used as a basis for delineating soil management zones.
Several NRCS soil scientists will be involved in making first order soil surveys of each of the three fields with and without the aid of digital terrain models and maps of electrical conductivity. Soil mapping units will be correlated to official series classification with the aid of complete and reference soil characterization. Selected validation data points in each field will be fully characterized and used to assess the accuracy of the first order soil surveys. Multivariate techniques will be used to determine the percent variability in grain yield explained by soil surveys and fertility data. Finally, these map units will be used as a basis for soil fertility sampling and management recommendations.
We expect to generate specific recommendations for grid soil sampling, interpolation, and mapping in Kentucky. And we hope to generate guidelines for creating first order soil surveys more efficient for agricultural management. These recommendations will be published in UK extension publications and refereed journal articles. What is learned from this study will also be incorporated into course material for a soil management class.