Saving Bees with Maths

Posted by Dr Martine Barons on 05 December 2016

The bees are dying

The bees are dying, and without pollination services, human food supplies will be severely threatened. We already face a grave challenge to produce enough food to feed the world's projected population of 9 billion by 2050 and any loss of ecosystem services, particularly pollination, will exacerbate this problem.

Insect pollinators not only support food production but are also important in the reproduction of many flowers and other plants, besides being attractive and interesting themselves. Our enjoyment of life would be significantly deteriorated on a number of levels if we lost these creatures.

The pressures facing insect pollinators are multi-faceted, dynamic and not completely understood. Therefore there is a need for decision support to underpin the policies we put in place to preserve them.

Recent advances in the networking of probabilistic systems for decision support, funded by EPSRC, allow us to develop an integrating decision support system (IDSS), bringing together information from many different areas of expertise in a robust, coherent and defensible way to produce scores for candidate policies so they can be compared by policymakers as an aid to decision-making.


Having established the structure and dependencies that exist within the system, the complex models and the data used by the disparate expert panels providing information to the IDSS needs to be identified. In some cases, experimental and observational evidence can be insufficient or missing altogether. In these cases, we can use structured expert elicitation to obtain the quantities we need. A panel of experts is brought together and asked to provide their subjective estimates in response to very specific and well-defined questions. Depending on the method used, there may also be calibration questions - questions to which the experts do not know the answer but which will become known soon. These help to determine not only how the individuals vary in their levels of expertise, but also how well they can provide quantitative estimates, especially probabilities.

For the pollinator IDSS, there is no experimental evidence to provide some much-needed conditional probabilities about the likely abundance of honey bees, wild bees and other insect pollinators (e.g. hoverflies) given high or low disease prevalence and average or unusual weather and good or poor forage availability. There are 16 combinations here and we need estimates for each one.

Top experts in the field, many of whom had given evidence to parliament as the matter was discussed and the National pollinator Strategy formulated, made up the panel. The IDEA protocol for structured expert elicitation was used to gain estimates for the probability of high pollinator abundance for the three categories of pollinator separately under the different disease, weather and forage conditions. The calibration questions came from about-to-be-published research papers and gave us a way to weight the experts according to their ability to estimate probabilities.

The IDEA protocol

The IDEA protocol asks the experts to give private, individual estimates of the probabilities required, beginning with the lowest plausible value, then the highest, and then the best estimate. The order is to avoid one of the known biases, towards a mid-point. These probabilities are then displayed in anonymised form for all to see, and a facilitated discussion explored the reasons why the scores varied. This ensured that all experts have access to the same information and checks experts' assumptions. Finally, a second, private estimate of the probabilities is given by each expert. The calibration exercise follows the same protocol and is used to provide weights for the combination of the judgements.

These can then be used within the IDSS to support policymakers to select policies to ameliorate adverse conditions for insect pollinator abundance, helping to secure their (and our) future.

To see more on the IDSS approach see the BBC feature.


In the following table, contact information relevant to the page. The first column is for visual reference only. Data is in the right column.

Photo of Dr Martine Barons
Name: Dr Martine Barons, CMath MIMA AMInstP
Job title: Director
Section / Team: Applied Statistics & Risk Unit
Department: Department of Statistics
Organisation: University of Warwick