Research using satellite data has been conducted to monitor in-season mango tree vigour and to improve the accuracies of pre-harvest yield forecasting for Australian mango growers.

A recently developed ‘18-tree calibration’ methodology using high resolution satellite data (WorldView-3) provided very accurate yield predictions at tree, block and orchard level at fruit-set period.

This has been validated for three consecutive seasons (2019/20/21), encompassing 13 farms (>250 individual orchard blocks) across three growing regions, eight mango varieties, with various tree ages and management practices.

On average, an overall accuracy of >90% was achieved in fruit count estimation using satellite data – a significant improvement on traditional manual yield estimation methods.

As well as the high accuracies achieved, the use of only 18 trees for in-field calibration is significantly less than the 2–3% of trees currently counted by growers, offering significant labour and time savings in the establishment of yield estimates.

The further derivation of yield variability maps has allowed growers to identify not only spatial variability in tree health across their orchards, but also to quantify that variation in yield.

However, WorldView-3 satellite data does incur a cost to purchase. To increase the adoption of this methodology, the accuracies of both low cost (Planet) and freely available (Sentinel-2) satellite data were also examined. The 2020/21/22 yield predictions were found highly comparable to those from WorldView-3.

Time-series mango yield prediction

As a further extension of research, ‘Freely available’ satellite imagery (Landsat) was used as time-series to develop a model that can provide yield forecasts much earlier in the season and with no in-field fruit counting requirement. This will further reduce the labour costs and time to estimate yield manually. This method identifies historic growth patterns at farm and block levels that are influenced by external factors such as weather conditions, management practices, pests, and diseases.

The results for 2021–22 season were found highly accurate at both farm and block level, with significant improvement over the traditional method of yield estimation. Further validation will be undertaken over additional farms across growing regions, varieties, and seasons.

The above outcomes provide mango growers with a choice of satellite platform (spatial, spectral and temporal resolutions) at a range of price points. This methodology and the associated high yield forecasting accuracies it provides is continuing to attract more Australian mango growers every season.

*Written by Priyakant Sinha and Andrew Robson, researchers at the Applied Agricultural Remote Sensing Centre (AARSC), University of New England, Armidale NSW. Email: psinha2@une.edu.au