The principle of citizen science is that, for each piece of data, we gather answers from several volunteers to calculate the “average” answer to the question. We call this process the “aggregation” of the results.
In Solar Jet Hunter, we have two workflows, and the aggregation is different for both workflows.
Aggregation in “Jet or Not”
In the first workflow, “Jet of Not”, volunteers answer a yes/no question. This is a simple task, and the aggregation is rather simple: the aggregated answer is the most chosen answer. But we can also calculate an agreement level to quantify how sure we are about the answer: for instance, if 100% of the volunteers answered that there is a jet, we are extremely confident that there is a jet in the movie. But if we got only 55% of volunteers answering “yes”, then we are not as confident that there is a jet.
Aggregation in “Box the Jet”
In the second workflow, “Box the jets”, the aggregation is more complex. In this workflow, we have base points and boxes for each jet that was found by the volunteers, and sometimes there is more than one jet in the subject! Clustering algorithms are used to find the aggregated points and boxes. Below is an example of the result of the aggregation for a given subject:
Jets detected in successive subjects might belong to the same solar jet!
We wanted to show short movies to volunteers during the project, and as a result, sometimes a solar jet is cut into different, successive movies. One the aggregation is done for the workflow “Box the jets”, we look at all the neighboring subjects in time to see if a jet in subject 1 continues in subject 2, and so on. After this comparison, we cluster jets from different subjects together in a final list of solar jets.
So, how many jets did the solar jet hunters find?
The original data selected was divided into 9,665 subjects, with observations of regions of the sun that were previously reported as containing jets, on data taken between 2011 and 2016. Solar jet hunters confirmed the presence of jets in about 21% of the subjects – more than 2,000 subjects. The volunteers reported that there were more than two jets in more than 300 subjects. After aggregation and clustering across subjects, we obtained a catalogue of 883 jets!
We will soon post more information about this catalogue on this blog!
And what now?
A catalogue of 881 jets with precise information on their timing, position, and extent, is an amazing resource for solar physicists interested in jets! It’s also a great training set to see if we can use machine learning to detect jets in extreme ultraviolet observations. We will keep you posted on the way the catalogue is being used to study solar jets and solar activity.
But Solar Jet Hunter is not done!
We have been working on a new version of the project that uses subjects in a video format – this should help both volunteers to spot jets and the team to aggregate the time information and provide finer time information in upcoming jet catalogue. We are entering a period of beta-test tomorrow: stay tuned, we need your help to analyze the remaining data that is available!