How did we forecast the future of our trails?

In out latest ‘Whiteboard in the Woods’ video, Data Manager Heather shared a brief insight into how we created our ‘data forecasts’ of the impacts of the UK’s forthcoming Deposit Return Scheme for drinks containers, and why we’re such staunch supporters of its introduction. On the post we got a comment saying ‘we love maths!’ so it only seemed right we’d invite Heather to take a deeper dive into the process behind the data.
‘How do you even work this stuff out?!’
If you’ve had a peak at our latest State of Our Report I can understand you wondering where on earth all those numbers come from. As a scientist it is always hard to work with ‘uncontrolled’ data: everything is usually rigid, with strict guidelines, undertaken by the researcher themselves, with planned locations and timeframes.
This is not how TFT’s data is collected: it’s gathered by our citizen scientists - you lovely people counting what you find in the places you care about and sharing that with us. We have some locations with hundreds of submissions and lots of locations with only one or two. This doesn’t mean we can’t use it to do science though - the magic of statistics is that you can use a sample to understand the whole. And we have a lot of samples (yes data submissions)!

At the start of working on each SoOT Report I first try to understand the patterns of our data: how many Deposit Return Scheme (DRS) eligible items are there on a UK scale? Does that vary by county? By type of trail? By type of clean? I usually start by plotting some graphs, getting a visual overview of our data before deep diving. It’s exciting to be able to ask so many questions of numbers that come directly from the hands of volunteers.
This year is the first time we’ve had enough data to start looking at a time-series: how our data evolves through time. I was finding all sorts of patterns, but mostly I couldn’t understand why the amount of items we were finding varied so much.
So it got me wondering: can I do some statistics to understand what influences how much single-use pollution we are finding? After a bit of research I decided to do a negative binomial regression which gave me some coefficients (numbers) and a word to indicate what that coefficient meant in relation to our data. By looking at how long we spent cleaning, how far we cleaned and how many volunteers took part the results overwhelmingly pointed to the number of people being the biggest factor in how much single-use pollution we remove.
Now, this has to be the best result we could celebrate. The most important thing in what we do, is you. Your time, your effort, your love for the trails and your communities that rally with you. This is the factor that has the greatest influence on making change.

It does also raise some further questions which now need investigating. At what point does the number of people not influence how much SUP we find? If a trail is clean you should be able to have 1 volunteer or 10 volunteers and find the same amount of SUP. Is it that when we have more volunteers it’s because we are targeting a location known for high levels of SUP? What does this mean for our items per km of trail value?
This is what I love about science - you find out something interesting and it shakes the foundation of all the things you thought you knew!




Comments