ISA Newsroom

Agricultural news farmers want to know.

Partnership to develop better imagery products for crop assessment

Article cover photo
Figure 3. Example of ISA/SIUE calibrated temporal NDVI imagery sequence from 2016, true comparisons over time. The imagery was calibrated using calibration tarps. The NDVI scale is shown on the side.

By: Joshua Pritsolas and Randy Pearson, Southern Illinois University Edwardsville and Peter Kyveryga, ISA Analytics

Background: Lately, the agricultural industry has witnessed a dramatic increase in vendors providing aerial imagery to farmers. The latest trend is to develop different types of “vegetative plant health maps” based on the aerial imagery to help growers and agronomists better identify stress factors within their fields.

While various “vegetative plant health maps” provide a significant insight into spatial variability within fields, they have some inherent limitations, leading to the misuse and misunderstanding of the utility of these images. This is especially true when the vegetative indices or “crop health maps” are compared from one field to another and across time—within a growing season and from year to year. The problem is in the lack of imagery calibration. The information shown on the “crop health map” last August cannot be compared to another image of a field taken last August or the same field taken the next day or this year. Farmers will end up with a series of imagery that have just one time use.


NDVI – Normalized difference vegetation index is the ratio of two quantities, one- difference between near-infrared and red, and the other the sum of near-infrared and red. NDVI ranges from -1 to +1 and correlates well with plant biomass and leaf greenness.

Georeferencing-assigning real geographic coordinates to imagery.

Mosaicking-combining or stitching multiple images from the same field to create a single image.

Research study

In 2015, the Iowa Soybean Association (ISA) Analytics team, in response to the increased misuse and misunderstanding of vegetation indices by the agricultural community, posed the following questions:

  1. Is NDVI the most commonly used vegetation index, appropriate for the agricultural community?
  2. What is the reliability of vegetation indices (such as NDVI) if they are not computed from the calibrated digital data (converted from digital numbers into percent reflectance)?

In the summer of 2015, ISA and Southern Illinois University Edwardsville (SIUE) teamed up to develop a plan to address the above questions. A team from Iowa State University joined the effort in 2016. The result was a 190-acre calibration research site located a few miles northeast of Ankeny, Iowa, with multiple fields of both corn and soybean. Two farmers agreed to help with the project.

This site was used to test various imaging companies’ digital data to understand its potential for producing calibrated vegetation indices. Calibration tarps (Figure 1) with known percentage reflectance values were deployed and images were collected every two weeks over the growing season.

Figure 1. Six tarps in gray scale array deployed before an August 27, 2015 flight. From background to foreground: 3, 6, 12, 22, 44, and 56 percent reflectance

During the inaugural 2015 season, a host of data quality issues were identified, including poor georeferencing, poor mosaicking, and most importantly, improper generation of vegetation indices. These issues were communicated to the various imaging companies and modifications to their systems ensued. As a result, the 2016 growing season resulted in much improved data from the companies, enabling the computation of vegetation indices (and products derived from vegetation indices) that could be compared from field-to-field and across the growing season. The following illustrates the successes from the 2016 season.

Results from Mavrx, Inc.

In 2016, Mavrx Inc. developed a calibration method for their aerial imagery and began prototype production of a vegetation index called “Absolute Crop Performance” (ACP), which is the same as NDVI. They both use a ratio of the difference between near-infrared and red, and the sum of near-infrared and red. Because of this similarity, we conducted a comparison test with the Mavrx prototype ACP and ISA/SIUE’s calibrated NDVI on an imagery captured on August 3, 2016. The test results were as follows:

  1. A high linear relationship between ACP and ISA/SIUE calibrated NDVI (R² = 0.78) with a slope of the equation close to one. Linear relationships are needed to extrapolate calibration equations over multiple fields.
  2. When all pixel values were examined, more than 95 percent of pixel values had an index value within +/- 0.03 (Figure 2); a difference of 0.03 in NDVI value, which ranges from -1 to +1, is very small.
  3. Although the two indices were similar, subtle differences were identified. These differences will be addressed by Mavrx for producing crop health maps in 2017.
Figure 2. The range of differences between Mavrx ACP crop health map and ISA/SIUE calibrated NDVI map.

These results also showed Mavrx has begun to develop calibration techniques that can produce reliable vegetation index products over time (for example, see Figure 3). The improved imagery will enable spatial and temporal comparability of crop assessment products for growers.

Figure 3. Example of ISA/SIUE calibrated temporal NDVI imagery sequence from 2016, true comparisons over time. The imagery was calibrated using calibration tarps. The NDVI scale is shown on the side.

Results from GeoVantage, Inc. & Aeroptic, LLC

For the 2015 and 2016 growing seasons GeoVantage Inc collected 15 images from the ISA calibration site. While the digital quality, spatial accuracy and mosaicking processes were quite good, their post-processing procedures introduced a nonlinearity in their data making large-scale calibration of their imagery impossible.

 The recent purchase of GeoVantage by Aeroptics, LLC has resulted in a renewed interest in the conversion of their systems and post-processing procedures, which should facilitate a more efficient calibration process in the 2017 growing season. Their overall objective is to provide highly calibrated multispectral imagery by the 2018 growing season. During the 2017 growing season ISA/SIUE will be assessing their initial modification, which is supposed to provide linear data that can be easily calibrated. The linear calibration equations can be extrapolated over multiple fields. This is a welcome change from an imagery provider that has been a mainstay in agricultural imaging for many years.


The overall goal of the joint ISA and SIUE image calibration project was to make the image acquisition industry and technology provide more reliable and standardized imagery products that can be compared from one field to another and across time. As a result, new imagery calibration products will help produce more reliable crop health maps, identify better management zone maps and assist in crop modeling.

In 2017, ISA and SIUE will continue work with image acquisition industry which should ultimately lead to the adaptation of industry products that can increase productivity and/or minimize cost.

For permission to republish articles or to request high-res photos contact Aaron Putze at

©2017 Iowa Soybean Association On-Farm Network®. All rights reserved. On-Farm Network® is a registered trademark of the Iowa Soybean Association, Ankeny, IA.Portions of some On-Farm Network trials are paid for in total or in part by the soybean checkoff.

Iowa Soybean Association | On-Farm Network | 1255 SW Prairie Trail Pkwy | Ankeny | IA | 50023 | US

Podcast Series See more

Podcast Library Graphic

Newsletter Signup

By becoming a member, you'll automatically receive the ISA member newsletters including: ISA eWeekly, ISA Research Advance, Farm Wealth Update, Soy Transportation News and our printed newsletter, ISA Monthly.

Sign Up

Iowa Soybean Review

August 2017 Contact Ann Clinton for past publications.