AARSC Blog
Select Page

Satellites predict mango yields with 90% accuracy

As the 2020/21 harvest season has shown, obtaining accurate yield forecasts for mango prior to harvesting is difficult. This can lead to failures in meeting forward selling quotas, or miscalculating harvesting requirements such as labour, packaging and transport.

Prior research undertaken by the UNE Applied Agricultural Remote Sensing Centre (AARSC) and several collaborative partners, has identified yield estimation and mapping derived from satellite imagery can provide accurate results. This has been further validated over many farms during 2019/20/21 seasons, with the results further confirming the commercial viability of this method.

During the 2019/20/21 growing season AARSC, Central Queensland University, DAFQ, NT DPIR and Australian Mangoes further evaluated the accuracies of very high-resolution satellite imagery (WV3) for measuring tree health and yield variability within commercial mango orchards grown across four growing regions (NT: Acacia Hills and Katherine, Nth Qld: Dimbulah and South East Qld: Bundaberg).

For the 2019-20 season, yield estimation was carried out for 95 individual orchard blocks across 10 farms in NT, NQLD and SEQLD. For the 2020-21 season this number increased to 278 blocks (from 13 farms), to further include different mango varieties (Calypso, R2E2, KP, HG, Keitt), tree ages and management.

Fruit count prediction

Through the on-ground measure of fruit yield per individual tree (18 trees per block), the evaluation of differing canopy spectral responses (Vegetation Index (VI)) and the inclusion of canopy area for each tree (VI*CA), yield predictions (total fruit count) were derived for each sampled block. The forecasts were compared against visual assessment from the participating growers (current commercial practice) as well as actual pack-house yield recorded after the commercial harvest.

Improved delivery of image productspictured is the yield map that farmers receive to interpret their yield predictions. The image used is a satellite image shot from above the mango crops and has a spatial layer applied to demonstrate the yield variability. this is represented in a colour coded legend.

To complement the numerical forecast of yield for each orchard, the algorithms developed were used to derive a forecasted yield for all individual trees, provided to participating growers through a web app.

The derived yield variability maps better inform growers on where within and across season variability occurs in tree health and yield and as such informs on where to undertake targeted agronomy to determine potential limiting constraints on production (ie soil, water, nutrient or other).

The web apps do not require the growers to have any existing spatial software, with maps viewable on any device e.g., tablets and mobile phones. With minimal instructions, the grower can access tree health and yield information at individual tree – to block – up to the orchard level.

With options to zoom-in/out, panning and other touch-screen commands, growers can better access the spatial variability of their farms. The web app development, design, look and feel is currently being assessed by participating growers with their feedback used to make further improvements.

This research is led by the UNE AARSC – in collaboration with Central Queensland University, DAFQ, NT DPIR and Australian Mangoes and is funded by Department of Agriculture, Water and the Environment – Rural Research and Development for Profit program, Horticulture Innovation
and project partners.

Prof Andrew Robson and Dr Priyakant Sinha are researchers at the Applied Agricultural Remote Sensing Centre (AARSC), University of New England, Armidale, Australia. 

Prof Andrew Robson: arobson7@une.edu.au

Dr Priyakant Sinha psinha2@une.edu.au

Article first published in Australian Tree Crop Magazine June/July 2021