—but I'll try to be brief.
Data quality is the core challenge of citizen science, and there are multiple steps that we employ to try to look at this.
The first step is that there are regional filters that are set by experts within a region that basically say, for this date and this location, what is the maximum number. If you exceed that, people are going to raise their eyebrows. In that case, because most sightings are coming in through mobile, you're presented at that time with something that says, “This is an unusual observation”, and you're basically asked for photos or for sound documentation.
The next step, which we often think less about.... False positives are something that I think we immediately jump to, but the bigger challenge is that often there are things that are vocalizing that people miss. No matter what type of sensor networks you have, you're going to have error rates associated with failures to understand this, so what we're doing is basically calibrating a sensor network.
Ours is a sensor network of people. People have quite a bit of variation in their level of expertise. If you go out with Geoff LeBaron, he's going to detect and identify just about all the birds you find. If you go out with my mom, she may detect the chickadees and some of the common birds, but not the other ones.
There's a lot of work that we also do in looking at expertise and classifying how many species per unit time people find under a variety of circumstances. That then allows us to counterweight different sensors and basically build out a more standardized sensor network.
There are several other approaches that we use as well, but those are two of the high themes.