Honey M&M

Inspiration

Bees (and other pollinating insects) are an essential part of agriculture around the world and climate change affects their ability to polinate the plants that end up in our plates.

The advantage of studying honey bees is that we have a metric to measure the amount of pollination that was done by bee colonies over a year, that metric being honey production. The more plants are polinated, the more honey is produced.

By evaluating the honey produced by hives and matching it to meteorological data, it is possible to find the extreme weather events that affect the bees' ability to polinate

What it does

Our project aims to model honey production for hives in Quebec (and/or other regions) by evaluating the impact of historical weather data on past honey yields.

The predictive model is then used with future climate predictions to in turn to predict honey yields.

How we built it

We started by trying to grasp the data as best we could. Initially we assumed the given data for the hackathon was representing the honey yields of the Dawson College Hives, without visualizing it on a map. We then realised it was actually the yield of apiaries around the city of Quebec.

Since the data is limited in numbers, the inital idea was to maximise feature generation (Number of days above X degrees (per season), Longest Heat Wave, Average solar radiation (per season), Peak solar radiation, Atmospheric pressure, Temperature, Rainfall, Pollen levels, Smog levels, Pesticide use, etc.)

This way we could model honey production based on A LOT of features, and determine which features were the most talkative.

Challenges we ran into

The challenges of developping features is getting useful data from which it is actually possible to build a predictive model.

Indeed, a lot of data was not openly accessible (such as historical pollen rates, specific bee hive production (vs. provincial averages), and more). Therefore, a lot of the science backed hypothesies were not possible to verify with data, since the data was not available.

It is so complicated to explain the reason behind high and low yield years. It all depends of the environment around the hives (5km radius) there are usually many factors involved which makes it hard to pin point causes. Having general weather data is therefore not so explicative. Last year in Montreal, it was identified that not cutting the dandelions due to covid helped the young bees to start the season well (dandelions have lots of pollen that helps babies to grow).

However the data of honey production by Montreal hives is private in most cases. Some companies have sensors for bee hives that may be useful to gather data specific to hives, such as Alveole and Nectar.

Accomplishments that we're proud of

It was our first time gathering Environmental Data from large databases such as Environment Canada and ClimateData.ca The learning curve for navigating these databases and cleaning up the data that was found was pretty steep and extracting valuable variables was complicated at times ! But we we're able to develop a proof of concept and some data visualization tools from the available data.