Solving the problem starts with framing the right question. Dr Cristina Sargent of the Smith Institute believes that mathematics is a clear, objective way of resolving the type of complexity experienced in agriculture.
Improving forecasting for G’s Growers is one of the case-studies she will be discussing at the Big Data SIG ‘Bring Out Your Data’ on 21st March, which includes a brokerage workshop.
The Smith Institute completed over 100 projects in the last ten years across 12 high value sectors including aerospace, telecommunications, transport and defence.
What if we did this?
For agriculture it is work in prediction and creating models that will enable the analysis of various options using ‘what if?’ scenarios that is arguably the most interesting.
Dr Sargent says: “Often, it is not immediately apparent how we can use mathematical techniques to solve real life problems. This is why at the beginning we have in-depth conversations with a prospective client to understand the challenge they are facing and what they would like to achieve.
“We spend considerable time looking at current operations, models they may be using and data that they have access to. Once we are satisfied that we understand the problem we will design a solution. This may be a combination of models, algorithms and data. When creating a solution, data is just one piece of the puzzle.
“We need to select data that is relevant to the problem. It may come in different formats and collected at various frequencies. So, there is always a preparation stage, looking at the data, cleaning it up, checking for errors and deciding what is most pertinent.
“Once we have analysed it and have a feel for what is important, we devise a data strategy. This looks at relationships within the data that can reveal useful information and also the external data sources that might reveal new insights.
“By defining these relationships, it is possible to create a model that can be used for operation optimisation, scenario exploration and to help decision-making.
“Some clients have internal resources for data analysis and we can support this. Once we have identified the algorithms they need, the model and the most pertinent data, their staff are then on the right track and can adapt the model for their own needs.
“One example of this is in the area of animal welfare. For a large herd of cattle, it is important to identify signs of illness at an early stage, to stop disease spreading or to prevent contamination of the milk. Cattle that are unwell change their behaviours.
“By using various mathematical techniques it is possible to analyse and interpret the data collected from sensors attached to cattle to prevent the onset of bigger problems.”
How do I optimise inputs?
If parts of the field have a better yield than others how do you decide the optimum amount of inputs to apply? Is it better to put your resource into improving the poorer land or to maximise the quality on the good land?
Dr Sargent considers the question and explains that the first stage is to ‘define the problem’. She says: “The first stage is to understand the dynamics of the problem in order to get to a point where you have the right data to define the relationship between spend and yield.
“Using historical data, it is possible to explore the relationship and ground truth it to see what factors are the most influential. With a sensible model, it should be possible to forecast optimal spend.”
Forecasting a barbeque weekend
Forecasting is another area where a small improvement could make a huge difference in profitability for farmers and producers.
The Smith Institute is working with G’s Growers on a model that will reduce waste in the production of Iceberg lettuces.
Salad consumption is very variable depending on weather – a ‘barbeque weekend’ will boost demand and a wet one depresses it. The maturity of the lettuce head is another variable; there is only a short time when it achieves the quality requirements of the supermarkets, if the head is too developed it will not have the required shelf life.
G’s has employed sophisticated monitoring systems to measure the growth of the lettuces and also collates data on weather and microclimate. This has enabled it to identify key growth stages and amend sowing and planting schedules to mitigate against potential shortfalls in crop availability.
The Smith Institute has been helping it to use this data to develop optimal production schedules that can cope with uncertainty. Additionally, it is creating an engine capable of crunching the data and creating ‘what if scenarios’ which would allow the in-house team to consider different management strategies.
Bring out your data
Forecasting and prediction is one of the areas that Dr Sargent will be discussing at the Big Data SIG. She is very interested to hear about these types of challenge in agriculture.
There will be an opportunity at the meeting to have a one-on-one discussion about specific challenges with experts, so, do make sure to ‘Bring Out Your Data’ on 21 March 2017 – click here to register.