Yearly Predictions Based on the Past 10 Years of Weather Data

This graph shows the flowering response curves for the first cultivar selected in the left panel. Each colored curve represents predictions based on a set of weather data from the past 10 years.

Helpful Tips

• Click the year in the legend to toggle the lines on or off.
• Double-click the year in the legend to isolate a line.
• Hover over the top right of the plot for a range of useful tools.

FlowerPower predicts the Zadoks flowering stages of cereal crops with a statistical model. The model uses DPIRD trial data and predicts the following growth stages:

The model predicts the growth stages using weather data from the past 10 years. That is, it will produce 10 predictions (one for each year). The predictions are presented on this webpage.

From 2025 onwards, the Oats Z71 module will not be updated with new trial data. The module will still be updated with new weather data, but no new cultivars will be added. We recommend using Oats Z49 instead.

Getting started

1. Type or select your crop, port zone, sites, and cultivars from the drop-down menus in the left panel. Click “Add Cultivar” to include additional cultivars.

2. There are two ways to visualize the predictions. Select the tabs at the top for the desired formats:

   • Comparison: Compare flowering response curves between cultivars. The predictions of the past 10 years are aggregated.

   • 10-Year Variation: Show the variation in the 10 flowering response curves as predicted with the weather data from the past 10 years. Only the predictions of a single cultivar are shown.

3. Advanced users may want to export the predictions in a CSV file and visualise the predictions locally with spreadsheet software (e.g., Excel).

   The exported file will contain both the predicted day of year and the growing degree-day (GDD) to reach the growth stage for the past 10 years. We use the following formula to compute GDD:

   $$ GDD = \max\left( \frac{T_{\max} - T_{\min}}{2}, 0 \right), $$

   where \(T_{\max}\) and \(T_{\min}\) are the daily maximum and minimum temperatures, respectively. We use the SILO weather dataset hosted by the Queensland Department of Environment and Science.

Technical Details

The statistical model underlying FlowerPower uses penalized splines to model the growing degree-day (GDD) response curves. The GDD predictions are then transformed to flowering day of year using daily temperature data from the SILO database. This approach is similar to Sharma and D’Antuono (2011). We use the implementation of penalized splines in the mgcv R package (Wood 2011).

For each site and cultivar, we compute 10 predictions based on weather data from the past 10 years (one prediction for each year). These predictions are then either aggregated (in “Comparison”) or shown individually (in “10-Year Variation”).

Reference

Sharma, D.L., and D’Antuono, M.F. (2011). Predicting flowering dates in wheat with a new statistical phenology model. Agronomy Journal, 103(1), 221–229.

Wood, S.N. (2011). Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 73(1), 3–36.

Limitations

• The model offers an indication of how different cultivars compare in flowering dates based on historical data. While not 100% accurate, it provides a reasonable approximation. The model has a margin of error of about 4–5 days on average.

• There is a bug that may resurrect “removed” rows of variety-dates after changing sites or port zones. The current workaround is to avoid removing rows that are already present when starting the app. Removing any extra rows created by users is fine.

We hope you find FlowerPower helpful despite its limitations. If you come across any bugs (e.g., long loading times, frequent timeouts) or have any feature suggestions, we would really appreciate it if you could let us know by emailing [email protected] with “FlowerPower Bug Report” or “FlowerPower Feature Request” in the subject line, as appropriate.

Disclaimer

The Department of Primary Industries and Regional Development (DPIRD) and the State of Western Australia accept no liability for any use or release of information in or linked to this website, or any errors or omissions therein.

Acknowledgement

FlowerPower is a flowering predictive tool that evolved from an Excel program into a web-based system. The development was led by Mario D’Antuono (until 2018) in collaboration with Darshan Sharma. Since 2019, Kenyon Ng has been developing and maintaining the model and webpage with support from Dean Diepeveen and Kawsar Salam. The model was built in close consultation with DPIRD agronomy researchers, who also provided trial data (some co-funded by GRDC).

Weather data was obtained from the SILO database hosted by the Queensland Department of Environment and Science under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.