Take home messages

• Targeting crop flowering date to the optimum time maximises yield potential and minimises the risk of yield loss due to frost, heat and water stress.
• Imagine an online tool that can provide growers with the optimum sowing date for new crop cultivars or regions based on the validated genetic controls in the plant.
• This project used genomic analysis of recent and near-release cultivars to build and test APSIM’s ability to predict flowering dates observed in National Variety Trials (NVT).
• The resultant Crop Flowering Calculator assists growers, agronomists and breeders to select the most suitable sowing date for specific cultivars to achieve optimal flowering dates to maximise yield across all Australian grain-growing areas.

Introduction

Changing seasonal conditions, better fallow management, improved seeding equipment, and the availability of crop varieties with greater vigour and adaptability have prompted a re-evaluation of sowing date recommendations for wheat and canola. Expanded cropping (particularly canola) into new environments, such as high- and low-rainfall zones, has created the need to identify optimal flowering periods (OFP) and phenological response of new cultivars for these regions.

The OFP is the range of dates when flowering is most likely to maximise yield. Flowering too early risks frost damage or insufficient biomass, while flowering too late increases the chance of heat and water stress. To optimise yield, growers need to match sowing dates and crop varieties to ensure flowering occurs within the OFP for their specific region.

Crop phenology, or the timing of crop development stages, is influenced by temperature, day length, and cold exposure, all controlled by genetic factors. These influences mean sowing date recommendations vary depending on the crop’s genetics and the regional climate. However, when new cultivars are released, their phenological response to Australia's diverse environments is not always well understood.

This project aims to provide timely and accurate predictions of flowering periods for new wheat, barley, and canola cultivars, delivered through an online tool. This would allow growers to confidently match sowing dates with new cultivars which will flower in the OFP to minimise stress and maximise yield.

Methods

We identified the OFPs for wheat, barley, and canola at over 3 000 sites in the Australian cropping zone using the APSIM NextGen crop models run over a continuous 66-year period (1957–2022). Climate data were sourced from BOM weather stations, and regional soil profiles were used for each site. Importantly, the analysis incorporated newly developed frost and heat damage functions for APSIM NextGen applied to a range of cultivars from slow winter types to fast spring types in each crop.

APSIM uses phenology parameters to describe how a cultivar responds to temperature, vernalisation (cold exposure), and photoperiod (day length). Traditionally, these parameters are derived from observations in controlled environments and field experiments at various sowing dates and locations.

This project used genomic analysis and machine learning to streamline this parameter development for APSIM. Because phenological responses are ultimately driven by genetic controls, we can correlate the genetics with observed responses. We used machine learning models trained on phenology observations and genetic data for a diverse range of varieties (49 wheat, 26 barley, and 208 canola cultivars).

Material Transfer Agreements with breeding companies enabled genomic analyses of commercial and pre-release cultivars included in National Variety Trials (NVT) for 2022–2024. In 2024, genomic models were used to predict phenology parameters for 30 wheat, 5 barley, and 27 canola cultivars. These were validated by running APSIM simulations at NVT locations with available phenology data. Only cultivars with at least two years of observed data across five locations, and with acceptable prediction accuracy, were included in the Crop Flowering Calculator. Insufficient phenology measurements were collected at NVT sites in 2024 for inclusion in the study.

A web-based interface was developed to deliver the Crop Flowering Calculator. APSIM simulations were run for 66 years across all sites and validated cultivars for sowing dates between 1 March and 9 August. The results, including predicted flowering dates, were compiled into a dataset that underpins the tool.

Results

Validation of phenology parameters from the machine learning model showed strong agreement with 2022 NVT data for wheat cultivars, achieving a correlation (r²) of 0.81 and a root mean square error (RMSE) of 9 days for 19 cultivars. While this is slightly less precise than phenology-focused observational trials such as those in the National Phenology Initiative, which achieved r² = 0.92, it was acceptable given phenology observations were made every 1–2 weeks at NVT sites rather than 3–4 times a week. For canola, the correlation from machine learning models using 2022 NVT data was r2= 0.79, again slightly less precise than r2= 0.90 achieved from phenology observation experiments. Notably, if more, high-quality phenology observations at 3-4 targeted environments are made available the model predictions (a) can be validated more widely, and (b) are likely to have improved accuracy.

The Crop Flowering Calculator tool allows users to put in their location and crop type, then ask: (see Figure 1)

• which variety should I sow (if I know my sowing date)
• when should I sow (if I know my variety).

Figure 1. User interface of Crop Flowering Calculator, showing pages for a) location selection, b) crop selection and c) cultivar selection, where sowing week is selected from dropdown menu, or d) sowing date selection, where up to three cultivars are chosen.

Users can apply filters to refine the results, such as selecting up to three varieties or filtering by herbicide group or breeding company. Results are presented as a marker on a timeline (see Figure 2) showing the range of predicted flowering dates relative to the OFP for each site.

Figure 2. Results page in Crop Flowering Calculator, showing a) cultivar selection output example where filters can be set for categories, such as herbicide group and breeding company, and b) output example where up to three cultivars can be matched with the OFP for a range of sowing dates.

Conclusion

This project used long-term APSIM simulations and genomic analysis to define optimal flowering periods for wheat, barley, and canola at more than 3 000 sites across the Australian cropping zone. The results were used to develop the Crop Flowering Calculator, an online tool designed in collaboration with consultants, researchers, and breeders. Crucially, the methods underpinning the tool are scalable, applicable for almost any environment and variety, and can be deployed at a fraction of the cost compared to methods that underpinned crop flowering calculators of the past. The tool provides growers and agronomists with reliable, evidence-based insights to guide sowing decisions for new varieties and regions, helping them to achieve better yields with reduced risk.

Acknowledgements

The research undertaken as part of this project is made possible by the significant contributions of growers through both trial cooperation and the support of the GRDC, the authors would like to thank them for their continued support. This research was also supported through funding from CSIRO (‘National Phenology Initiative - Phase 2’ (NPI2)). Thanks to the APSIM Initiative, which supports APSIM modelling software.

Contact details

Julianne Lilley
CSIRO Agriculture and Food
GPO Box 1700, Canberra ACT 2601

Jeremy Whish
CSIRO Agriculture and Food
306 Carmody Rd, St Lucia QLD 4067