Premium
Using big data to predict the likelihood of low falling numbers in wheat
Author(s) -
Williams Richard M.,
Diepeveen Dean A.,
Evans Fiona H.
Publication year - 2019
Publication title -
cereal chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.558
H-Index - 100
eISSN - 1943-3638
pISSN - 0009-0352
DOI - 10.1002/cche.10140
Subject(s) - falling number , falling (accident) , preharvest , relative humidity , equating , range (aeronautics) , humidity , environmental science , statistics , mathematics , chemistry , meteorology , geography , horticulture , food science , biology , postharvest , medicine , materials science , environmental health , gluten , rasch model , composite material
Background and objectives Preharvest sprouting in wheat reduces quality and impacts farmer profitability. The international recognized falling number test can be used to measure that damage. Trying to understand the complex interactions that cause a reduction in wheat quality, equating to low falling number levels, is challenging. An alternative research approach to replicated experiments was to use a multiseason dataset of load‐by‐load delivery information to investigate whether correlations between falling number levels and 40 climate measurements could be identified. Findings This study used over 250,000 falling number data points from individual truckloads tested during seven harvests in Western Australia. The analyses identified relative humidity measured at the maximum temperature and daily temperature range as having consistent correlations with falling number levels over multiple seasons. Other climate measurements were also observed to have significant correlations with falling number, but these were less consistent within and between seasons. Conclusions The linkage of humidity and temperature range levels in the period before harvest commences to the occurrence of low falling number levels helps to further understand the complex interactions that change starch quality. Significance and novelty The findings demonstrate that value can be obtained from the use of a large, nonexperimentally designed dataset.