Getting an accurate forecast of the annual macadamia harvest has long been on the wish list for both farmers and the burgeoning macadamia industry itself, but it has been notoriously difficult to accurately predict. New research could be changing that, writes Bronwyn Herbert.
The Macadamia industry’s forecast yields currently rely predominately on weather forecasts.
However, in more recent years research attention has turned to satellite data to delve deep into the canopy of macadamia orchards to predict yields.
Associate Professor James Brinkhoff from the Applied Agricultural Remote Sensing Centre at the University of New England in Armidale, New South Wales, has been leading the yield modelling work since 2020, building on prior research by AARSC that evaluated the relationship between individual macadamia tree yield and canopy reflectance (measured from satellite) over a number of years and locations.
Earlier this year the Centre published research that showed accurate yield predictions could be obtained using spatio-temporal datasets including remote sensing, weather and elevation.
Using yield data from 2009 to 2019, forecast models were trained on free satellite imagery, as well as weather data sets and other inputs, to enable forecasts that could be produced for any orchard in Australia.
“We aim to produce forecasts in January for the upcoming harvest season,” he said.
“It’s well in advance of harvest starting (in March) and we supply (forecasts) to our collaborative growers as an interactive report to help them make decisions around marketing, finance and logistics.”
Ass Prof Brinkhoff said the work is also helping growers better understand the spatial variability of yields on their farms.
The forecasting could also open the door to offering growers further insights into the drivers of yield variability and enabling site specific management to optimise yields.
Currently the industry body’s regional yield forecasting program is run by the Queensland Department of Agriculture and Fisheries (QDAF), relying on predominately weather and soil information, with some market indicators such as price.
Ass Prof Brinkhoff said the key advantage of using satellite imagery is the scale and frequency of data available.
“It comes with a 30-metre resolution and the archive stretches back to before 1990 so we can really look at the historical trends,” he said.
The images are also taken every 16 days, so a dense time series is available.
Ass Prof Brinkhoff says investigations so far have shown the satellite data, which is picking up the overall vigour of the orchard canopy as well as water status, has a very good correlation with macadamia yields.
“It’s picking up the canopy reflectance -indicative of leaf area and biomass – which drives capacity for production of nuts,” he said.
The satellite also has some short-wave infrared bands, which are strongly related to water.
“Overall the particular block level errors are around 20 per cent,” he said.
“But one farm will have 10 or 20 blocks, so when you aggregate the block level predictions to the farm level, the predictions are stabilising at less than 10 per cent error.”
Currently 15 growers are involved in the trial forecasting project covering the industry, with orchards located across the eastern seaboard, stretching from Atherton in Far North Queensland to Macksville on the mid north coast of New South Wales, as well as major growing areas around Bundaberg and Ballina.
Ass Prof Brinkhoff believes there are still improvements that can be made to the model, including adding additional data such as irrigation inputs, pollination and variety related effects.
“Some varieties in some orchards, they have a huge yield one year and very low the next year,” he said.
“It’s an ongoing effort to look at improvements to models,” he said.
From here, the research work is looking to expand and increase its accuracy by getting information from additional orchards.
“We are always looking for more growers for the project – it really adds value to have more information from a wider variety of weather zone regions, different soil types and management practices,” Ass Prof Brinkhoff said.
Ultimately the goal is for this work to combine with the QDAF-based weather model predicting yields, so that the whole industry will benefit.
“We’ve been working closely with them, but it would be nice to bring these efforts together in an even closer collaboration in future years,” he said.
This research has been conducted as part of the Multi-Scale Monitoring of Tools for Tree Crops Phase 2 project, which led by the University of New England’s Applied Agricultural Remote Sensing Centre and is supported by Hort Innovation, the Australian Government Department of Agriculture and Water Resources as part of its Rural R&D for Profit program and by industry partners including the Australian Macadamia Society. More information can be found in this research article: https://doi.org/10.1016/j.agrformet.2021.108369
With acknowledgment and thanks to Bronwyn Herbert for her words and to the Society of Precision Ag for allowing us to share this article on our blog. This article was first published in the Society of Prevision Ag in late 2021.