Our world’s population is projected to grow. This has sparked international food strategies to support new crop technologies that improve crop output efficiency. Yet limited land available for farming in the UK has been compounded by evolving land use. Combined with food; future farms will be exploited for biofuels, solar and wind power to help meet emissions targets.

Weather Logistics, a recent Nottingham start-up in the field of long-range forecasting, aims to tackle food production issues through better management of UK agricultural risks. At present it produces fine scale predictions on 25km space-scales, at challenging 1-6 month ‘seasonal’ timescales. Its products will include metrics to help predict the growth and health of crops. The vision is to inform grower groups and food procurers on the risks to crop health posed by extreme seasonal weather.

Crop health is affected by climate and timely decisions are required. This includes crop spraying programmes, selecting plant varieties, drilling, sowing, harvesting, and planning summer irrigation systems. Weather Logistics strives to address these risks at the monthly to seasonal timescales. Different crops are susceptible to different pests and diseases at different stages of the growth lifecycle. These threats are a growing concern in a warmer and moister climate. For instance, aphids weaken fruit trees and prefer mild spring conditions. Botrytis mould is most prevalent during ‘damp squib’ summers; a recent trend that is detrimental to the quality of soft-fruits such as strawberries.

To downscale forecasts to farming scales, Weather Logistics uses a statistical model that combines space-derived data, seasonal forecast output from European weather centres, and historical weather and climate data. The company manages a large amount of post-processed climate data, producing forecasts that describe the average weather conditions.

The company’s current development is a product to predict the number of growing days. These seasonal forecasts will be produced in conjunction with output uncertainties, applying relationships that combine independent datasets from around the globe. This method differs from conventional nested models, with a local forecast produced from its regional parameters. The ultimate goal is to quantify the benefit of preventive spraying and crop planning for the management of arable crops, and to allow food procurement and logistic companies to better market fresh produce to food retailers and decide where best to locate their factories.

Consequently, Weather Logistics recently attended the European Study Group with Industry (ESGI107), hosted by the University of Manchester, where it presented its seasonal forecast problem with supplementary climate data. Its goal was to discover new avenues to both refine the seasonal forecast and to better quantify its uncertainties. Presenting the problem entitled “How to best combine statistical-empirical relationships to downscale seasonal forecasts?”, Weather Logistics attracted a mixed group of enthusiastic industrial mathematicians and were fascinated to discover ways to better select, represent and combine forecast predictors to improve the seasonal forecast output.

Several areas of focus were identified early in the workshop; and methods were suggested to tackle each. The study group first set out to understand the forecast concept in more detail, such as the selection of jet stream parameters to form a new diagnostic climate index. Some members of the study group embarked on a mechanistic study to assess the validity of the seasonal forecast process, and devised new techniques to refine and lower uncertainties in numerical predictions. This included applying machine learning techniques to train the forecast with past observations.

Correlation coefficients of individual and combined temperature predictors

This figure shows the two monthly correlation coefficients (smoothing out oscillations seen in one month correlation coefficients) of the temperature predictors. The correlations of the individual predictors (strength, standard deviation and position of the jet stream close to the UK) are shown by the dashed lines and the summer, winter and all season predictors are shown by the solid lines. The summer predictor is a combination of strength and standard deviation of the jet stream close to the UK, and the winter predictor is a combination of strength and position of the jet stream close to the UK. The all season predictor is formed by taking a weighted sum of summer and winter predictors.

The study group outcome exceeded expectations, providing the company with a comprehensive framework of methods through a final report to improve the seasonal forecasts. This will no doubt allow Weather Logistics to make better decisions and perform unbiased calculations of risk for the companies customers.

Weather Logistics is now collaborating with the University of Manchester’s School of Mathematics and with study group members from across Europe to further develop products, using some of the proposed techniques. Research and development is essential for business to grow and the company is looking out for anyone interested in collaboration, particularly through Knowledge Transfer Partnerships.

This Guest Blog was contributed by:
Christopher Nankervis, Weather Logistics Ltd

In March 2015, Natural Environment Research Council (NERC) funded a problem at the UK’s European Mathematical Study Group with Industry presented by Weather Logistics Ltd. The problem was funded as part of the Probability, Uncertainty and Risk in the Environment (PURE) Research Programme, which aims to improve understanding and management of risk and uncertainty for natural hazards. The Study Group in the UK is organised by the Smith Institute for Industrial Mathematics and System Engineering and brings together people from a wide range of backgrounds to focus on issues of real importance to industry that can be addressed by mathematical modelling.

Please get in touch to find out more about Study Groups, and browse the ESGI and the Smith Institute Network pages for further information, including other problems from Study Groups. The next UK edition of the European Study Group (ESGI116) will take place at the University of Durham, from 11-15 April 2016.