Fertilizer Recommendation Support Tool

Fertilizer Recommendation Support Tool (FRST) User Manual, v. 1.0.0.0

Table of Contents

  1. FRST Background
  2. FRST User Instructions
  3. Understanding Soil Test Correlation
  4. Registration Option
  5. Dictionary of Terms
  6. References
  7. Acknowledgements
  8. FRST Programming Information
  9. FRST Model Citation

1. FRST Background

The Fertilizer Recommendation Support Tool Project, or “FRST”, is a national initiative to modernize fertilizer recommendations by pooling expertise and soil test correlation and calibration data from across the country into an accessible decision support tool for phosphorus and potassium.

We expect FRST will provide additional soil test information to augment and refine existing recommendations. The FRST decision aid consists of inputs, an interactive map, and outputs. The user specifies criteria, and the decision aid’s output is the critical soil test value (CSTV) or the soil test value above which you do not expect a yield increase from fertilization (correlation). We are working to add frequency of response information, a fertilizer nutrient-rate calibration component for phosphorus and potassium output and an additional nutrient, sulfur, to the tool. Users may report any problems with the tool by contacting us at https://soiltestfrst.org/contact/ .

This FRST project has been created to:

  • Increase the transparency of soil test evaluation by promoting clear and consistent interpretations of crop nutrient soil test results and yield responsiveness.
  • Remove political and institutional bias from soil test interpretation.
  • Store and manage crop soil fertility data and assist in the advancement of science related to soil testing.
  • Enhance end-user awareness, confidence, and adoption of soil-test based crop nutrient management recommendations.
  • Provide a collaborative environment for discussing and communicating the issues, needs, and science of soil-test-based nutrient recommendations.

The FRST decision aid is a browser-based tool that can be successfully used on desktop and notebook computers. The decision aid will work on mobile devices but there may be some display issues depending on screen resolution and orientation. Work is continuing to design the decision aid so it can be used across a broad range of devices. There are no known browser-specific issues.

2. FRST User Instructions

The FRST decision aid can be accessed at frst.scinet.usda.gov/tool. The tool display provides seven data selection criteria inputs:

  • Soil Nutrient (phosphorus or potassium)
  • Crop
  • States/Territories
  • Mapped Soil Series
  • Years
  • Soil Sample Depths
  • Soil Test Method

An interactive map of states and counties in the United States and Puerto Rico is displayed (Figure 1).

A map of the united states

Description automatically generated

Figure 1. FRST decision aid home screen with input selectors, county-level map of the United States and Puerto Rico, and location of the trials.

Selection order does not matter for the Soil Nutrient, Crop, States/Territories, Mapped Soil Series, and Years. Years will reset when soil nutrient, crop, states, and mapped soil series selections are changed. Select any one of them and only the relevant information in the other three input selectors will be available. For instance, the Mapped Soil Series selector auto-fills without selecting any other inputs. All soil series in the database are available for selection. Selectable Soil Series are reduced as Soil Nutrient, Crop, and/or States/Territories are selected. Selecting a Soil Series along with the other inputs will provide the fewest number of data points and often provides little soil test correlation information. The greatest amount of usable information will be provided when only selecting Soil Nutrient and Crop.

Most selectors have a pull-down menu or list of check boxes. State/Territories can be selected either using the drop-down menu or by using the lasso tool (Table 1). The lasso tool also allows you to select counties within and across states. The lasso tool is in the map tool bar and is the button with a yellow circle. To use the lasso, first select the lasso button by clicking on it. A blue border will appear around the button. Next, click on the map to identify the start location of the lasso. While holding down the mouse button circle the area of green counties you want to analyze. This will populate the States/ Territories input box with the relevant states and counties. It should be noted that only data from the counties within the lasso area will be used. When the lasso clear button is clicked, all counties with data within the selected states will be highlighted green and all data from the selected States/Territories will be used for data analysis.  A description of each map tool button can be found in Table 1.

Table 1. Map toolbox buttons .

Reduce map size (zoom out)

Increase map size (zoom in)

Center map in map view window

Lasso state/county selection tool

Clear lasso selection

Deselect county and close county data information window

 

When the lasso tool is not active, there is no blue border around the lasso button, you can click on any county to display information on data availability for the selected county. To clear the selected county and data information window, click on the county deselect button in the map toolbar.

Soil Nutrient, Crop, and Mapped Soil Series entries can be cleared by selecting the blank entry at the top of the drop-down menu, whereas the “Clear States and Territories” must be used to clear this category. Years will default based on the inputs selected, listing the available data years. All input entries will clear by selecting the “Reload Current Page” symbol  at the top of your web browser.

At a minimum, Soil Nutrient and Crop must be selected for the tool to run. That selection will auto-populate Years and display options for Soil Sample Depth. Once Soil Sample Depth(s) is/are selected, the Soil Test Methods will be available for selection as radio buttons, single selection. Multiple Soil Sample Depths can be selected but only a single Soil Test Method can be selected for soil test correlation. Soil Sampling Depth selections can also be deselected. Anytime there is a change in the Soil Sampling Depths selected, the Soil Test Method options will be updated, and your selection will be cleared. Note the tool begins collecting the trial data for soil test correlation modeling as soon as the Soil Test Method is selected.

After selecting the Soil Depth and Soil Test Method input categories, FRST processes the available data and displays the following series of notes as the tool runs:

  • ”The FRST tool is running soil correlation using trial data matching the selected criteria. This notification will close when the correlation analysis is finished.”

  • “Retrieving trial data from FRST database.”

  • “Completed trial data retrieval.”

Finally, a progress bar will be displayed while the data is analyzed.

Once the analysis is complete, several pieces of information are provided: a detailed map with data metrics, the soil test correlation graph and associate information, a trial yield maximum distribution histogram, and a summary table that allows you to review data for individual trials. Output information is accessed by scrolling down the browser window on the right-hand side. If you want to change Soil Depth and/or Soil Test Methodology, simply re-click the boxes with your new choices and the program will rerun.

Once all the desired criteria are selected, the results will be displayed including:

  1. FRST output map
  2. Soil test correlation graph
  3. Site year maximum yield distribution graph
  4. Listing of trial data

FRST Output Map

The FRST Output Map consists of a map based on the selected inputs. In the example presented below, Soil Nutrient (K), Crop (Soybean), State (LA), Soil Depth (0-6 & 0-8 inches), and Soil Test Method (Mehlich-3) were selected, and the data processed; the interactive output displays a map of the U.S. and territories with trial location (county) and lists the number of states, counties, and trial count per soil test depth (Figure 2).

A map of the united states

Description automatically generated

Figure 2. FRST output map illustrating the location and number of trials meeting the specified criteria for nutrient, crop, state, soil sample depth, and soil test method.

Soil Test Correlation Graph

Below the map is the Soil Test Correlation Graph (Figure 3). The soil test correlation page displays the relative crop yield on the y-axis and the soil test concentration on the x-axis; the critical soil test value (CSTV) and relative yield are indicated with a green dashed line. The critical soil test value (CSTV) identifies the level above which additional application of the nutrient is unlikely to increase crop yield. Values for the critical soil test value and corresponding relative yield are displayed at the bottom of the graph above the x-axis and next to the green dashed line. Determining the critical soil test value requires calculating relative yield and then applying an empirical model to the data. The Quadratic Plateau is the empirical model selected for soil test correlation by FRST. The critical soil test value is set at 95% of the maximum predicted relative yield. Detailed information about selecting the relative yield and empirical model can be found in the next section, Understanding Soil Test Correlation.

A screenshot of a computer screen

Description automatically generated

Figure 3. Soil test correlation graph and accompanying information.

Sometimes there is insufficient data (12 or fewer trials) to perform soil test correlation analysis and provide output. While there is no established minimum number of trials required for soil test correlation, the FRST uses 12 site-years of data as its minimum. In other cases, data simply shows no trend, and the following message will be displayed,

  • “FRST was unable to calculate an acceptable correlation from the selected data. This may occur because of poor data distribution, lack of positive crop yield response to fertilization, or several other factors.”

In these cases, only the trial data used for the analysis will be displayed so you can still review the data that is currently available through FRST.

There is other information on the soil test correlation graph starting with the right-hand side. The upper right-hand side displays the list of the inputs that were selected. The lower right-hand side provides the number of runs needed using a well-established method, bootstrapping, and a key of symbols. For more information on bootstrapping, see the Dictionary of Terms & Understanding Soil Test Correlation sections. The date of the analysis is in the lower right-hand corner of the screen.

The R2 statistic (upper left-hand corner) describes the proportion of the variation in the dependent variable (relative yield) that is predictable from the independent variable (soil test level). The R2 and the confidence interval of R2 are listed.

At the bottom of the graph, the quadratic plateau model parameters determined from bootstrapping and data fitting are presented. These values are used to determine the critical soil test value (CSTV). The first parameter Soil Test Value Joint Point (STVJP) is the nutrient soil test value (ppm) above which relative crop yield is constant. This is commonly referred to as the joint point for the quadratic plateau model where the plateau section of the model starts. The next parameter is Relative Yield Max or RYMax, which is the relative crop yield (%) maximum value. The RYMax value can never be greater than 100% and is often lower. The quadratic plateau intercept (Int) is the last parameter displayed. This is the predicted relative crop yield when the soil test concentration for the given nutrient is 0 ppm.

Site Year Maximum Yield Distribution Graph

Below the soil test correlation graph is the graph showing the Site-Year Maximum Yield Distribution histogram (Figure 4). Site-year count is on the y-axis and maximum yield is on the x-axis. This histogram provides a quick overview of the trial yield data.

Both the soil test correlation and site-year maximum yield distribution figures can be downloaded as a png file using the “Download” button in the lower right-hand corner of each of the panels.

A screen shot of a graph

Description automatically generated

Figure 4. Site-year maximum yield distribution histogram

Listing Of Trial Data

The final output section is a Listing of Trial Data from which data was used to determine the critical soil test value (CSTV). If multiple depths were selected, trials are grouped into separate tables by soil sample depth and can be viewed by clicking on the appropriate tab at the top. Trials are listed from lowest to highest soil test value for the nutrient of interest. Clicking on an individual trial title line displays trial treatment information and additional data.

Each trial header identifies the trial year, the county and state, the beginning soil test nutrient value in parts per million (ppm), the relative yield (RY) for the unfertilized treatment, and the maximum yield (Figure 5a). To view detailed information for each trial, click the down “v” arrow or any portion of the header. Information provided for each trial includes the nutrient of interest, treatment nutrient rates, source of nutrient, actual yield, and relative yield; an analysis of variance may or may not be available based on trial data reported (Figure 5b). Soil information for each trial can be obtained by clicking the “View Soil Information” button (Figure 5c). The amount and type of information returned depends on what other soil characteristics the researcher(s) measured and reported. Simply click the “Close” button to return to the individual trial information. To close the trial information, click the “^” arrow or any portion of the trial header. Also, the currently selected trial will automatically close when another trial is selected.

(a)

(b) A screenshot of a computer

Description automatically generated

(c) A screenshot of a test

Description automatically generated

Figure 5. Trial output information including (a) notice regarding the data and then a listing of trials by soil test depth, (b) expanded individual trial data showing the experimental details and output, including statistical information when available, and (c) measured soil characteristics.

3. Understanding Soil Test Correlation

Soil testing provides the backbone for nutrient management in modern agricultural production systems. Soil test correlation and calibration answers two questions:

  1. At what soil test value can we stop applying a fertilizer nutrient and not limit yield? (correlation)
  2. If we do need fertilizer, how much fertilizer do we need? (calibration)

The soil test value determined by correlation is referred to as the critical soil test value (CSTV). The critical soil test value is the point below which a crop is expected to respond positively to fertilization and above which fertilizer application is unlikely to increase crop yield. Correlation is the first step of analyzing data that has a critical mass of field trial information. At a minimum, data must include soil test nutrient availability and some measurement of crop yield response to fertilization expressed as the relative yield of the control (unfertilized)-plot treatment relative to the crop yield receiving an optimal (or non-limiting) rate of fertilizer. Soil test correlation determines whether a soil test method (e.g., Bray 1, Mehlich-3, Ammonium Acetate, etc.) quantitatively extracts a fraction of soil nutrients that can be utilized for developing crop fertilizer recommendations. One important component of the critical soil test value is that a 1-unit change (30 ppm vs 31 ppm) affects the fertilizer rate recommendation. This is called the “Boundary Effect”.

Determining the critical soil test value (CSTV) requires calculating relative yield and then applying an empirical model to the data. Detailed information follows for these components.

Relative Yield

Directly comparing crop yields from multiple trials and researchers is difficult since site-specific crop yield potential is affected by spatial, temporal, management, and environmental factors. Therefore, the individual site-year yield data are typically normalized to a relative yield value. Relative yield is a ratio of two crop yields and is often converted to a percentage by multiplying by 100. Relative yield can range from 0-100% or exceed 100%, depending on the method for determining maximum attainable yield used in the denominator. We analyzed six different relative yield calculation methods and found they did not significantly change the critical soil test value (CSTV). The calculation method shown in Equation 1 was selected as the most appropriate method for calculating relative yield in the FRST decision aid where the numerator or control treatment yield represents the yield when the nutrient of interest is not applied, and the denominator represents the highest yielding treatment in the trial which could be the control treatment. Thus, the relative yield is always ≤100%.

Equation 1: % relative yield = (control yield ÷ highest mean treatment yield) × 100

Critical Soil Test Value (CSTV)

Soil test correlation involves fitting an empirical model to a dataset and interpreting the model to identify the critical soil test value above which fertilization is unlikely to increase crop yield. Model fitting requires enough site-year observations to document crop response to fertilization across a wide range of soil-nutrient availability values, calculating the relative yield of each trial’s control treatment (receiving no-fertilizer for the nutrient-of-interest), and applying a model to the site-year data to describe the relationship between soil test and relative yield response. Many mathematical models and interpretations have been used to determine critical soil test value (CSTV), and the best model for FRST was unclear so we evaluated different models with the goal to select a single model for the critical soil test value interpretation of the FRST decision aid.

Many models were considered by FRST collaborators, and some were excluded due to well-known model disadvantages. Four models [Arcsine-log Correlation Curve (ALCC), Exponential (EXP), Linear Plateau (LP), & Quadratic Plateau (QP)] were selected for review and fit to three example datasets using field trial data from multiple states.

The three datasets included corn response to Olsen-P from the Midwest USA, soybean response to Mehlich-1 K from Virginia, North Carolina, and Maryland, and Mehlich-3 K from Arkansas. The critical soil test level was identified as 95% of the maximum predicted relative yield for the ALCC, EXP, and QP models and the joint point (100% of the maximum relative yield) for the LP and QP models for a total of five soil test correlation modeling approaches. The joint point is where the response line or curve meets the horizontal line (plateau where relative yield is constant and equals maximum relative yield). The predicted critical soil test values (CSTV) were compared using relative yield as the dependent variable. The five critical soil test values ranged from 46-66 ppm for the Mehlich-1 K dataset, 115-165 ppm for the Mehlich-3 K dataset, and 7-16 ppm for the Olsen P dataset. The Quadratic Plateau model tended to have the narrowest and most stable ranges of critical soil test values.

The response frequency of significant yield increases to fertilization was also examined as a companion metric to communicate the frequency of false positive and false negative error predictions above and below the critical soi test value (CSTV). After reviewing all models thoroughly, the FRST collaborators selected the Quadratic Plateau as the model to use in the FRST decision aid. The critical soil test value (CSTV) is set at 95% of the maximum predicted relative yield at the joint point where the model plateau occurs reflecting the diminishing returns to fertilization as the response curve nears the joint point. The curve shown on the FRST correlation output, and the associated values are based on the Quadratic Plateau model selected through this process of model comparison.

FRST Soil Test Correlation Modeling

The FRST decision aid uses bootstrapping to perform soil test correlation to estimate critical soil test value (CSTV). Bootstrapping is a technique where a dataset is randomly resampled multiple times to generate new datasets. Generated datasets will have the same number of data points as the original dataset. Within each generated dataset, some original data points may not be included while others may be included more than once. This is commonly referred to as resampling with data replacement. This technique is used to better quantify the variability associated with a given data population which the original dataset represents.

The FRST decision aid requires data from a minimum of 12 site-years to perform soil test correlation using bootstrapping. The decision aid attempts to fit the quadratic plateau model to the original dataset and then to each of the bootstrapped datasets. Currently, the decision aid requires 1000 successfully fitted bootstrap datasets to complete correlation and estimate a critical soil test value for the original dataset. For any additional insights into bootstrapping and soil test correlation, see https://adriancorrendo.github.io/01-index.html.

Fitting individual bootstrap datasets may fail for the following reasons.

  • Model parameters cannot be determined because of insufficient data in the quadratic or plateau portion of the model.
  • Data does not follow a quadratic plateau response model and parameters cannot be determined.
  • Soil test value at the model joint point is below the lowest soil test value or above the highest soil test value found in the dataset.
  • Maximum relative crop yield (plateau relative crop yield) is greater than 100.
  • Intercept, relative crop yield at 0 ppm soil test value, is greater than maximum relative crop yield.

The FRST decision aid will continue soil test correlation execution as long as the decision aid is able to successfully fit the model to one dataset out of every ten bootstrapped datasets. In other words, when the decision aid is unable to fit the model to ten consecutive bootstrap datasets, soil test correlation will be stopped, and correlation information will not be presented.

The number of unsuccessful bootstrap datasets is reported on the right side of the correlation graph. High counts for unsuccessfully fitted datasets indicate that soil test correlation results based on the original dataset are relatively weak and additional data should be considered.

4. Registration Option

Registration is not needed to use the FRST decision aid and obtain the critical soil test value (CSTV) graphics and trial data output as described in the previous sections. Users who would like graphics that provide i) the standard output plus trend lines for the successful bootstrap runs, and ii) 75% and 95% confidence intervals for the median joint point can register for login access. Additional soil test correlation and calibration features may be added with time.

5. Dictionary of Terms

This document provides a dictionary of terms to understand soil-test correlation and calibration terminology used in the United States.

  1. Arcsine-log Correlation Curve Model (ALCC) – begins by transforming relative yield by the arcsine of its square root requiring that all relative yield values be ≤ 1 (i.e., ≤ 100%) transforming the soil test values by natural log, flipping the axes so that relative yield becomes the independent variable, and centering the relative yield data by subtracting the transformed sufficiency level target [i.e., x – arsine(√0.95)]. The next step is fitting a linear regression between the new relative yield and soil test data. The curve always peaks at 100% relative yield, and never has a positive y-intercept, making it somewhat unique from conventional nonlinear models often used in soil test correlation research. For more information see Bolster et al (2023) or Correndo et al (2023).
  2. Bootstrapping is used to measure uncertainty of the critical soil test value (CSTV) estimate. Bootstrapping involves resampling the original data many times to create many new datasets with the same number of data points as the original dataset. Because the original data is resampled randomly with replacement, a new dataset, or bootstrap, may contain multiple occurrences of some points while other points are not included. As a result, the soil test correlation analysis may or may not be successful using the new dataset. FRST requires the original dataset to be resampled and successfully analyzed 1000 times to calculate CSTV, provided there was a limited number of unsuccessful runs.
  3. Boundary Effect – describes the phenomena when a change in 1 soil test nutrient unit (ppm, pound per acre, or unitless index) results in a major change in the recommended fertilizer rate or definition of the soil test level (from Low to Medium).
  4. Calibration (soil test calibration) – the second step in developing soil-test-based fertilizer recommendations that define the average rates of fertilizer-nutrient needed to produce maximal yields across the range of deficient to sufficient soil test values. Soil test calibration can only be performed if the soil test correlation is successful and sufficient data exists to empirically predict the decision aid rate of a fertilizer nutrient required to achieve optimum yield.
  5. Control-Treatment Relative Yield – the relative yield of the control treatment in a soil test correlation-calibration trial. The treatment receives all nutrients and crop management inputs as other treatments except for the nutrient of interest.
  6. Correlation (soil test correlation) – the first step in developing soil-test-based fertilizer recommendations that defines the relationship between crop relative yield of the control treatment and soil test values for multiple site-years of data to identify the critical soil test value (CSTV). The soil test correlation may also use other parameters (besides relative yield) such as Response Frequency.
  7. Critical Soil Test Range (CSTR) – the soil test values that identify the transition from where a crop yield benefit is expected from fertilization to soils (with higher values) where no yield benefit from fertilization is expected. The CSTR includes the Critical Soil Test Value (CSTV) and may be based on the error associated with the CSTV or a range of predetermined relative yield values. The FRST tool does not currently include a CSTR.
  8. Critical Soil Test Value (CSTV) – the soil test value that separates soils (with a lower value) where a yield benefit is expected from fertilization from soils (with higher values) where no yield benefit is expected from fertilization.
  9. Exponential Model (EXP) – in the context of soil test correlation, EXP is a regression model that describes a rapid increase in response followed by diminishing increases towards a maximum (asymptote).
  10. False Negative Error – a fertilizer rate recommendation error in soil test validation research where a benefit from crop fertilization is expected but no yield benefit occurs.
  11. False Positive Error – a fertilizer rate recommendation error in soil test validation research where no benefit from crop fertilization is expected but a yield benefit occurs.
  12. Legacy Data – data included in the FRST database that was obtained from research results published in peer-reviewed papers, reports, or archived in electronic or physical files that describe replicated soil test correlation field research and include the beginning soil test values, soil test method, soil sample depth, individual treatment yield means, and research origin.
  13. Linear Plateau Model (LP) – a segmented regression model where the linear increase breaks at a certain point beyond which the response remains flat, or constant. The point where the function flattens is referred to as the joint point.
  14. Maximum Yield Treatment – the definition used in FRST identifies the highest yielding treatment in a trial and the treatment yield is used as the denominator for calculating relative yield as defined by Pearce et al (2022).
  15. Minimum Dataset – the minimum data determined to be necessary for soil fertility trials by the FRST project (Slaton et al., 2022).
  16. Quadratic Plateau Model (QP) – a segmented regression model with two phases: a curvilinear response followed by a flat plateau. The point where the function flattens is referred to as the joint point. The quadratic plateau model was selected as the soil test correlation model as described by Slaton et al. (In Review)
  17. R2 (Coefficient of Determination) – a goodness of fit statistic measuring the proportion of variation in the dependent variable (e.g., relative yield) that can be predicted by the independent variable (soil test value) from fitting a regression model to a dataset. The R2 values range from 0 to 1.0 with 0 indicating no relationship and 1.0 indicating a perfect relationship. Most soil test correlation datasets have R2 values between 0.2 and 0.6 indicating a weak to moderately good relationship.
  18. Relative Yield (RY) – the crop yield of a treatment expressed as a percentage that is “relative” to the yield of another standard treatment. For FRST, the highest-yielding trial treatment is the standard by which relative yield is calculated and all treatments are compared, Pearce et al. (2022).
  19. Response Frequency – the percentage of times that a statistically significant yield increase occurs from fertilization with a specific nutrient. Response Frequencies should be associated with a soil test level or a range of soil test values.
  20. Site-Year – a field trial that represents a single location and year of data.
  21. Soil Test Level – a range of soil test values or concentrations represented by a title that communicates whether soils with this level of nutrient should respond to fertilization. The title and its meaning should be intuitive to a wide range of expertise on soil testing. A lot of different soil test level terms have been used over time, but typical titles include Low, Medium, Optimum, and High.
  22. Soil Test Nutrient – refers to a relatively rapid test that assesses a portion of soil nutrients that are available to plants and should be considered as an “availability index”. The soil test nutrient values are usually expressed as a concentration (parts per million or pounds per acre) but are sometimes expressed as an index (unitless). Regardless of the units, the value is neither the total nutrient content of the soil nor the total amount of plant available nutrients in the soil.

6. References

Fertilizer Recommendation Support Tool Website. (2024). www.soiltestfrst.org.

Bolster, C. H., Correndo, A. A., Pearce, A. W., Spargo, J. T., Slaton, N. A., & Osmond, D. L. (2023). A spreadsheet for determining critical soil test values using the modified arcsine-log calibration curve. Soil Science Society of America Journal, 87, 182–189. https://doi.org/10.1002/saj2.20498

Correndo, A.A., Pearce, A.W.. Bolster, C.H. , Spargo, J. T., Osmond, D.L., Ciampitti, I. A. (2023). The Soiltestcorr R Package: An Accessible Framework for Reproducible Correlation Analysis of Crop Yield and Soil Test Data. Software X. https://doi.org/10.1016/j.softx.2022.101275       

Lyons, S.E., Osmond, D.L., Slaton, N.A., Spargo, J.T., Kleinman, P.J.A., & Arthur, D.K. (2020). FRST: A national soil testing database to improve soil fertility recommendations. Agricultural & Environmental Letters, 5:e20008. https://doi.org/10.1002/ael2.20008

Lyons, S.E., Arthur, D.K., Slaton, N.A., Pearce, A.W., Spargo, J.T., Osmond, D.L., & Kleinman, P.J.A. (2021). Development of a soil test correlation and calibration database for the USA . Agricultural & Environmental Letters . 6, e20008. https://doi.org/10.1002/ael2.20058

Pearce, A.W., Slaton, N. A., Lyons, S. E., Bolster, C. H., Bruulsema, T. W., Grove, J. H., Jones, J. D., McGrath, J. M., Miguez, F. E., Nelson, N. O., Osmond, D. L., Parvej, Md. R., Pena-Yewtukhiw, E. M., & Spargo, J. T. (2022). Defining relative yield for soil test correlation and calibration trials in the fertilizer recommendation support tool. Soil Science Society of America Journal , 86, 1338–1353. https://doi.org/10.1002/saj2.20450

Slaton, N.A., Lyons, S.E., Osmond, D.L. Osmond, Brouder, S.M., Culman, S., Drescher, G., Gatiboni, L.C., Hoben, J., Kleinman, P.J.A., McGrath, J.M., Miller, R., Pearce, A., Shober, A.M., Spargo, J.T., & Volenec, J.J. (2022). Minimum dataset and metadata guidelines for soil-test correlation and calibration research. Soil Science Society America Journal, 86, 19–33. https://doi.org/10.1002/saj2.20338

Slaton, N, Pearce, A., Gatiboni, L.C., Osmond, D., Bolster, C., Miguez, F., Dhillon, J., Farmaha, B., Kaiser, D., Margenot, A., Moore, A., Ruiz-Diaz, D., Sotomayor, D., Spackman, J., Spargo, J.T., & Yost, M. (2024). Soil test correlation model and sufficiency interpretation selection comparison. Soil Science Society America Journal . (In review)

7. Acknowledgements

Funding for the FRST project was made possible by the USDA-ARS National Programs for Natural Resources and Sustainable Agricultural Systems (Grant 58-8070-8-016), the USDA-NRCS ( 69-3A75-17-45,
NR203A7500010C00C and NR233A750011G016), Hatch Funds provided by the National Institute of Food and Agriculture, U.S. Department of Agriculture, and Hatch and Smith-Lever funds that support collaborator research and extension projects, and OCP North America.

8. FRST Programming Information

FRST Database: PostgreSQL 14

Application: C# Blazor Web Application (C#, HTML, CSS, Razor, JavaScript)

Application Components:

  • Bootstrap 5.x (frontend toolkit used for responsive web design)
  • D3.js (JavaScript library used for maps and graphs)
  • EPPlus (Excel spreadsheet library for .NET Framework/Core)
  • html2canvas (used to create images for download)
  • MailKit (used for application emails)
  • MathNet.Numerics (mathematics and statistics library)
  • Microsoft.AspNetCore.Identity.EntityFrameworkCore (identity and authentication library)
  • Microsoft.EntityFrameworkCore (database support library with LINQ)
  • Microsoft.JSInterop (used for .Net and JavaScript inter-operations)
  • NpgSQL.EntityFrameworkCore.PostgreSQL (PostgreSQL database support library)
  • PuppeteerSharp (application headless browser used for APIs)
  • TopoJSON (extension of GeoJSON used with US county map)

9. Model Citation

Buol, G., Osmond, D., Slaton, N., Spargo, J., Lyons, S., Pearce, A., Uthman, Q., Yost, M., & Kaiser, D. (2024). Fertilizer Recommendation Support Tool, V 1.0.0.0. frst.scinet.usda.gov/tool.