Machine Learning-Based Growth Prediction in Autumn-Sown Spring Oats under a Maritime Climate

Highlights Spring-type oat sown in autumn demonstrates high yield potential under the Irish maritime climate; however, it is challenged by limited variety choice and a better understanding of yield physiology. The work presented here aimed to understand the source-sink balance and identify suitable vegetation indices for predicting green area index, light interception and above-ground biomass. Information generated from this study has the potential to be used in oat crop modelling under climate change scenarios and high-throughput screening of varieties utilising vegetation indices, derived from multispectral sensors. Abstract Ireland has one of the highest average oat (Avena sativa L.) yields among major oat-growing countries, thanks to the cooler maritime climate and longer growing season. Despite the potential, oat production in Ireland is challenged by a lack of diverse varieties and limited research into oat growth, development and yield physiology. This study aimed to compare two machine learning (ML) approaches in predicting different growth parameters. The field experiments were conducted on autumn-sown spring-type cv. Husky at three sites with contrasting latitudes over three years in Ireland. We assessed canopy development, light interception, biomass accumulation, and nitrogen (N) dynamics alongside hand-held multispectral sensor data. An ML framework utilising the RReliefF algorithm for feature selection was employed to identify optimal VIs for predicting the different growth and canopy parameters. The grand mean of 10.3 t/ha of grain yield and 28,562 for grain no./m2 demonstrated a high yield potential. Source-sink analysis reveals that autumn-sown oat is primarily sink-limited by grain no./m2 at the pre-anthesis stage and influenced by growth rate and nitrogen uptake rate. Conversely, grain weight was occasionally co-limited by source-sink at the post-anthesis stage and strongly influenced by grain-filling rate and duration. Preliminary ML regression models indicate that spectral VIs can reliably estimate growth parameters. A comparative analysis of Random Forest (RF) and Partial Least Squares (PLS) regression revealed that the optimal modelling framework is dependent on the target variable. Both RF and PLS predicted GAI and LI with high accuracy (R2 > 0.80; RMSE < 1; RPD > 2.28), indicating reliable quantitative predictive capacity. These findings provide a framework for oat crop modelling under climate change and demonstrate the potential of high-throughput phenotyping under a maritime climate.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-17
DOI
https://doi.org/10.5281/zenodo.24319595
Primary Topic
Remote Sensing in Agriculture
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Machine Learning-Based Growth Prediction in Autumn-Sown Spring Oats under a Maritime Climate

Tapiwa Zengeza, Atikur Rahman, Farjana Akter
Zenodo (CERN European Organization for Nuclear Research)
Remote Sensing in Agriculture
article

Machine Learning-Based Growth Prediction in Autumn-Sown Spring Oats under a Maritime Climate

Tapiwa Zengeza, Atikur Rahman, Farjana Akter
article en

Abstract

Highlights Spring-type oat sown in autumn demonstrates high yield potential under the Irish maritime climate; however, it is challenged by limited variety choice and a better understanding of yield physiology. The work presented here aimed to understand the source-sink balance and identify suitable vegetation indices for predicting green area index, light interception and above-ground biomass. Information generated from this study has the potential to be used in oat crop modelling under climate change scenarios and high-throughput screening of varieties utilising vegetation indices, derived from multispectral sensors. Abstract Ireland has one of the highest average oat (Avena sativa L.) yields among major oat-growing countries, thanks to the cooler maritime climate and longer growing season. Despite the potential, oat production in Ireland is challenged by a lack of diverse varieties and limited research into oat growth, development and yield physiology. This study aimed to compare two machine learning (ML) approaches in predicting different growth parameters. The field experiments were conducted on autumn-sown spring-type cv. Husky at three sites with contrasting latitudes over three years in Ireland. We assessed canopy development, light interception, biomass accumulation, and nitrogen (N) dynamics alongside hand-held multispectral sensor data. An ML framework utilising the RReliefF algorithm for feature selection was employed to identify optimal VIs for predicting the different growth and canopy parameters. The grand mean of 10.3 t/ha of grain yield and 28,562 for grain no./m2 demonstrated a high yield potential. Source-sink analysis reveals that autumn-sown oat is primarily sink-limited by grain no./m2 at the pre-anthesis stage and influenced by growth rate and nitrogen uptake rate. Conversely, grain weight was occasionally co-limited by source-sink at the post-anthesis stage and strongly influenced by grain-filling rate and duration. Preliminary ML regression models indicate that spectral VIs can reliably estimate growth parameters. A comparative analysis of Random Forest (RF) and Partial Least Squares (PLS) regression revealed that the optimal modelling framework is dependent on the target variable. Both RF and PLS predicted GAI and LI with high accuracy (R2 > 0.80; RMSE < 1; RPD > 2.28), indicating reliable quantitative predictive capacity. These findings provide a framework for oat crop modelling under climate change and demonstrate the potential of high-throughput phenotyping under a maritime climate.

Zenodo (CERN European Organization for Nuclear Research)
Teagasc - The Irish Agriculture and Food Development Authority (IE), Dundalk Institute of Technology (IE)
Remote Sensing in Agriculture
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon