Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: DEVELOPING ANALYTICAL AND MANAGEMENT STRATEGIES TO IMPROVE CROP UTILIZATION OF .... AND REDUCE LOSSES TO THE ENVIRONMENT

Location: Crop Systems & Global Change

Title: A career in agriculture and NIR: What I've learned all too often still needs to be learned. Or "The need for caution"

Author
item Reeves Iii, James

Submitted to: NIR news (Near Infrared Reflectance News)
Publication Type: Popular Publication
Publication Acceptance Date: March 4, 2011
Publication Date: N/A

Interpretive Summary: In August 2010 I gave a talk at the 2010 International Diffuse Reflectance Conference for winning the Tomas Hirschfeld Award for achievements in near-infrared spectroscopy (NIRS). Using light in the visible and near-infrared spectral ranges, NIRS has over the last several decades become one of the primary methods used for the rapid analysis of agricultural materials. The objective of my talk was basically to try and pass along some of the observations I have made based both upon my own research and as a reviewer. For the long time, expert users of NIRS much, if not all, will be old hat, but hopefully for some it will be helpful. The first conclusion I have come to is that a great deal of education about NIRS is still needed. Many new users are not utilizing the large base of knowledge readily available and thus are repeating the same errors such as using too few samples for calibration development. The second main conclusion is that even for long term users assumptions are made about why we do certain things, or do things in a certain way, which are often not based on any sound foundation. For example, it is held by most that spectral data needs to be transformed in specific formats (log transformations) for calibrations to be accurate, yet recent work has indicated that this is not true and that spectral format is relatively irrelevant. Basically, we all do it because it was first done that way, seems logical, and thus has not been tested. Similarly, the methods used to develop calibrations have changed over the last two decades with new developments in computers and programs, but the change in methods used appears to be due more to commercial interests than based on studies showing them to be scientific better. Thus, it can be seen that there is still a great need for education, but how do we get the word out? While all have to learn there is a great loss of time, money and effort if the learning occurs at the manuscript review stage or if something better is not even examined because of assumptions based on false premises. I have wondered if the problem is that NIR has become too main stream and thus the potential problems are not even considered. In 2006 (last year I did a report for), there was no PITTCON NIRS users meeting because it was stated “There was no near-infrared (NIR) there anymore”, but while no session other than the Hirschfeld award session was labeled NIR (one for chemometrics), there were 65 talks (40 oral) on NIR in 26 different oral sessions alone, 2.7% of total PITTCON presentations were NIR based. While NIR may now be thought of as just like other more traditional methods, the education level needed for successfully using NIRS does not appear to be there in far too many cases. Also, many, if not most, of the errors occur in areas where abundant information is available which makes getting the word out by traditional methods such as publications even harder. This leaves the final question as: What is the best method to help disseminate knowledge already known to stop wasted efforts, and is it even necessary, or just my soapbox issue?

Technical Abstract: This is a summary of the talk I gave for the 2010 International Diffuse Reflectance Conference for the Tomas Hirschfeld Award for achievements in near-infrared spectroscopy (NIRS). The objective of the talk was basically to try and pass along some of the observations I have made based both upon my own research and as a reviewer. For the long time, expert users of NIRS much, if not all, will be old hat, but hopefully for some it will be helpful. The first conclusion I have come to is that a great deal of education about NIRS is still needed. Many new users are not utilizing the large base of knowledge readily available and thus are repeating the same errors such as using too few samples for calibration development. The second main conclusion is that even for long term users assumptions are made about why we do certain things, or do things in a certain way, which are often not based on any sound foundation. For example, it is held by most that spectral data needs to be transformed in specific formats such as log (1/R) of Kubelka-Munk for calibrations to be accurate, yet recent work has indicated that this is not true and that spectral format is relatively irrelevant. Even using single beam, non-background corrected spectra works fined. Basically we all do it because it was first done that way (Karl Norris – It just worked, Personal Communication, seems logical (Beer’s Law), and thus has not been tested. Similarly, the methods used to develop calibrations have changed over the last two decades from multi-linear regression to partial least squares, neural nets, etc., with new developments in computers and programs, but the change in methods used appears to be due more to commercial interests than based on studies showing them to be scientific better. Thus it can be seen that there is still a great need for education for all of us, but how do we get the word out? While all have to learn there is a great loss of time, money and effort if the learning occurs at the manuscript review stage or if something better is not even examined because of assumptions based on false premises. I have wondered if the problem is that NIR has become too main stream and thus the potential problems are not even considered. In 2006 (last year I did a report for NIR News), there was no PITTCON NIRS users meeting because it was stated “There was no near-infrared (NIR) there anymore”, but while no session other than the Hirschfeld award session was labeled NIR (one for chemometrics), there were 65 talks (40 oral) on NIR in 26 different oral sessions alone, 2.7% of total PITTCON presentations were NIR based. While NIR may now be thought of as just like other more traditional methods such as gas-chromatography the education level needed for successfully using NIRS does not appear to be there in far too many cases. Also, many, if not most, of the errors occur in areas where abundant information is available which makes getting the word out by traditional methods such as publications even harder. This leaves the final question as: What is the best method to help disseminate knowledge already known to stop wasted efforts, and is it even necessary, or is it just my soapbox issue.

Last Modified: 7/31/2014
Footer Content Back to Top of Page