A Double machine learning trend model for citizen science data

Just saw this through The Wildlife Society, the focus was on eBird, but seems it would be transferrable to iNat?
With close to 60,000 observations for the City of Surrey where I work, I would love to have better tools to analyze the incredibly valuable data that our community science efforts represent.

“Citizen science can give researchers remarkable data, but the findings are subject to the accuracy of the participants. Since human behavior can be even more confounding than wildlife behavior, researchers at the Cornell Lab have had to wrestle with changes in how birders record their findings.”
https://doi.org/10.1111/2041-210X.14186

From a brief glance, it seems like a general method that could be adapted to iNaturalist data. That said, iNat data is a bit more complex than ebird because it has so many different species, with different patterns of observations. For instance, they specifically trained their models on checklist data:
" To help control for variation in observation process, we analysed the subset of the data for which participants report all bird species detected and identified during the survey period, resulting in complete checklists of bird species."
which iNat doesn’t have. I think adapting this approach would be possible, but require a lot of time, effort, and expertise.

But there’s nothing stopping you (or anyone else) from going for it! It would probably lead to a paper in a good journal if done well.

1 Like

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.