There is a place for NIR in avocados

By Massimo Nettis and Terry Rudge, Rudge Produce Systems

NIR (near infrared) technology is very useful for measuring dry matter in Shepard avocados and you do not need to be a scientist to build a model.

According to Massimo Nettis of Rudge Produce Systems, you just need to read the instructions and be very careful about how you take measurements.

“A device such as a Felix F750 can take hundreds of non-destructive readings in the field in the time it takes to do one or two oven dry matter readings,” Massimo said.

“This makes NIR a great tool for growers to predict harvest maturity and to understand the variability of fruit dry matter. It also allows growers to understand how weather conditions and growing practices affect fruit maturity.”

There is widespread use of NIR to measure dry matter of mangoes, but the place for this technology in the avocado industry has been less obvious. This year, Rudge Produce Systems needed to take multiple dry matter measurements of avocados without destroying fruit. This was required as part of Hort Innovation project Implementing best practice of avocado fruit management and handling practices from farm to ripening (AV18000). This project is managed by the Queensland Department of Agriculture and Fisheries.

“We found the Felix F750 to be an ideal tool for this task but needed to develop a ‘model’ so the F750 had the smarts to estimate dry matter of avocados,” Massimo said.

“A model is like a language and our devices were fluent only in mango. They could understand ‘Hass speak’ because our Sydney representative Nasser Abdi had already built a model, with help from Kerry Walsh’s team at Central Queensland University.

“We needed to build a new model for Shepard, and we needed it urgently. I was given the task even though I had never heard of an F750,” he said. Massimo’s experience has useful lessons.

Lesson 1 – read the manual

On the Felix website there is a link to a YouTube about Felix’s Modelbuilder software.

“The video was useful, but I did not understand the jargon. I had no idea what terms like ‘training sets’ and ‘reference data’ meant,” Massimo said.

“I did not really understand the process until I read the ‘Data Viewer & Model Builder User Manual’. I read the step by step the instructions in the User manual at least three times.”

Lesson 2 – collect reference data

“I must have checked at least 50 trays of Shepard to collect fruit with a wide enough range of maturity,” Massimo said.

“I finished with a sample of 40 fruit ranging from 20-26% dry matter. I would have liked to include fruit with readings down to 18 and 19%.

“Next time I will start earlier and try to include some less-mature fruit.”

Lesson 3 – Be precise

Massimo said he did this part “pretty well”, marking the skin to ensure Felix readings were taken from exactly the same place as flesh plugs were to be taken.

“We used scales with readings to 0.001 grams and a very good dehydrator,” he said.

Lesson 4 – use firm fruit for model building

Massimo said the fruit he used to build the model was at hard green stage.

“This was because I wanted to build the model as early as possible in the Shepard season,” he said.

“It was easy to draw plugs of flesh for drying and I did not have the problem of flesh sticking to the side of the corer and knife.

“I plan to add data from ripe fruit to my model so I can be sure it works on eating ripe fruit. This will be a much dirtier process and I will need to be very careful in handling flesh samples.”

Lesson 5 – Using the Modelbuilder tools to optimise the data

Massimo said he entered scan readings from the F750 and matched them up to the destructive Dry Matter readings.

“I pressed ‘build’ and it took only minutes to come up with a model. I then used the in-built tools to see how good it was,” he said.

“To start with, my scan data did not correlate strongly with dry matter, but it improved considerably after I took out one or two outlying data points.

“The biggest weakness of my model was that it was based on only 40 fruit (80 readings). Next time I will take more readings and add them to the 80 that I already have.”

Lesson 6 – Contact Felix support for any problem

“I copied the new model on to an SD card and inserted it into the F750, but our device could not detect the new file,” Massimo said.

“I wasted a lot of time reviewing what I had done. Eventually I swallowed my pride and contacted Felix support, and it turned out that the SD card was corrupted.

“I followed the few easy steps that Felix suggested, and the model worked perfectly.”

Lesson 7 – Test it in real life

Massimo said his Shepard avocado model worked.

“The Dry Matter values that it estimates are close to the actuals. I will continue to add additional data to the model to make it more reliable.”

Massimo’s recommendations

  • The F750 can be used to take non-destructive measurements of dry matter in Shepard, and you do not need to be a scientist to build a model. You do need to read the instructions and be particularly careful about how you take measurements during the model building process.
  • The F750 has a real place in helping avocado growers predict harvest maturity and understand the variability of fruit dry matter.
  • A grower can take hundreds of NIR readings in the time it takes to do one or two oven dry matter readings. It is easy to see how it can be used to investigate how cultural practices influence fruit maturity.

More information

For further information contact Massimo Nettis on 0432 765 459 or


The Implementing best practice of avocado fruit management and handling practices from farm to ripening (AV18000) project has been funded by Hort Innovation, using the avocado research and development levy, co-investment from Department of Agriculture and Fisheries and contributions from the Australian Government.

Hort Innovation - Strategic Levy Investment (Avocado Fund)

This article was produced for the Winter 2021 edition of Talking Avocados.

Author: Lisa Yorkston
Date Published: 15/09/2021