Abstract
In recent years, the sources of data on farming practices have increased exponentially. On-the-go fuel consumption, grain moisture and yield per hectare are now easily stored by farmers. Fine resolution (5 m) wheat yield forecast is presented here using two machine learning approaches: (1) bootstrapped regression trees (BRT) where predictions are pixel-wise; and (2) convolutional neural networks (CNN) where predictions use neighbouring pixels. This study made use of publicly available data (e.g. Sentinel 2 imagery) and farm-owned data (e.g. yield data). Results showed better performance of CNN over BRT, and when higher resolution data was included.
Abstract from the paper’s OpenAlex record, shown for indexing and research. Copyright remains with the authors or the publisher. Read at the source ↗
The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.
Claims 2
Claim · model-written summary
CNN performed better than BRT for wheat yield forecasting.
- Taxon
- Wheat
Triticum aestivumt0 - Outcome
- performancee0
- Direction
- bettere0
- Comparator
- BRTe0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
t0“Wheat yield forecast”
e0“Results showed better performance of CNN over BRT”
Claim · model-written summary
Including higher resolution data improved wheat yield forecast performance.
- Taxon
- WheatTriticum aestivumt0
- Treatment
- higher resolution datae0
- Outcome
- performancee0
- Direction
- bettere0
- Comparator
- BRTe0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
t0“Wheat yield forecast”
e0“Results showed better performance of CNN over BRT, and when higher resolution data was included.”