Access to field-level daily seasonal weather forecasts allow UK growers to make better informed decisions that could minimise crop damage and maximise yield while limiting food and pesticide wastage. Weather Logistics has developed a forecast system built on numerical weather prediction data from the Copernicus Climate Data Store. This allows large grower co-operatives to optimise crop development that could help reduce the wastage of perishable crops with estimated savings of up to £20 million for lettuce growers. To refine and validate their forecasts, Weather Logistics’ objective was to run 24 years of historic climate data on the Hartree Centre’s high performance computing (HPC) facilities and optimise their existing codebase.
The team parallelised the existing code to make it run faster, handling larger datasets more easily and improving the quality of forecasts. Refactoring the code – reconstructing code without changing its external behaviour – made it easier to navigate, maintain and sustain. The refactored and parallelised code was used to run historical weather data through Scafell Pike - one of Hartree Centre’s supercomputers. This allowed the team to identify areas for improvement. They went on to containerise Weather Logistics’ serial code to run on cloud resources, providing flexibility and scalability to meet the demands of a growing business.
This work helped deliver more accurate weather forecasts by refining existing algorithms and demonstrating the added value of local short and long-term forecasts for farming communities. The software engineering work provided flexibility to update the codebase more easily in future and the modular approach reduced the risk of breaking code in future development cycles. As a result of this engagement, the team were able to offer a scalable solution to the company going forward enabling access to cloud computing
resources on demand.
"We can now deliver reliable forecast confidences to our clients, facilitating better decision making in the field."
"Working with the Hartree Centre has accelerated our technology to market without the need to recruit an experienced software team."
At a glance
- Conversion of serial code to optimised parallel code
- Created an easy to use container for scalable cloud deployment
- Complete refactor of a complicated codebase
- Ran 24 years of historical data through the Weather Logistics code. Saving at least 2 months of compute time compared to the run time without Hartree Centre engagement