Wednesday, December 17, 2008

Interview with Brian Sullivan of eBird

I'm ashamed of the pun, but eBird kills two birds with one stone.  On the one hand are birders who are tired of keeping complicated lists of their sightings, and are looking to simplify.  On the other hand are scientists that are frustrated with extrapolating bird population data from banding stations and estimations and are looking for more complete data. 

This is where eBird comes in.  In addition to providing birders with a format to track and manage their lists and observations, it also allows scientists access to vast amounts of continent-wide data on bird populations.  After starting slowly, eBird now records millions of individual sightings each year.  In addition to being a great success story in the citizen science movement, eBird has become an indispensable tool for many birders, including myself.  

And guess what: there's a new version in the works.  I recently talked with Brian Sullivan, one of the three guys in charge of eBird, about the site's past, present and future.

Birdist: Let's start out by talking a little bit about your role, and the history of eBird.

Brian Sullivan: Well, eBird was launched in 2002 by the Lab and Audubon as a joint project, and the goal was generally to harness the power of the observations of birders, and put them into an archive where they could be used and accessed by scientists and conservationists.  So, that was sort of the overarching goal.  Internally, our big hypothesis was that observations made by birders could be used by scientists for scientific purposes, and we're still in the process of proving that, because eBird is relatively young.  You know, birders have collected data and made huge contributions to our knowledge of ornithology from the start, and this was sort of a no-brainer idea in terms of pulling all that data together into one useful format.

And so on your end, on the science end, it's very useful in terms of gathering data and those things.  On my end, as a casual birder, it's very useful as a way to keep track of what I'm doing and what I'm seeing.  So, when you guys started it up, did you really start from the scientific end of things or did you also think that this would be something that casual birders are going to need to keep track of their sightings?

Well, when eBird was first launched, it was built by scientists, for scientists.  It was less birder-friendly than it is now.  And that was fine, and a certain number of people will participate when they just want to contribute their data for the good of science and conservation and things like that.  The flaw was that there was no reward to the user back at the beginning.  So, basically you submitted your data into the vacuum and it was sucked in and put together with everyone elses, but you didn't get any reward back, even simple things like keeping a life-list, and things like that.  

So we saw a decent amount of participation in the beginning, but then it plateaued.   In 2005, they brought in Steve Kelling, who is the Director of Information Science at the Lab, and John Fitzpatrick, the Lab Director, decided to come at it from a different perspective and try to hire some people who would turn eBird on its head, essentially, and develop it from a birding perspective.  So, late 2004, early 2005 they brought me and Chris Wood on to basically try to increase participation and redevelop eBird in a way that would be more widely used in the birding community.  Chris and I basically vied for the same job, and then they decided to hire both of us because our ideas were so similar about how to get eBird to get to the next level. 

The big piece that we brought in were some simple user rewards, like various lists that birders keep.  So, when you input data, it automatically updates your lists based on the location where you submitted your data, things like that.  Simple stuff.

Simple, but necessary.  For someone like me, who didn't have a long history of keeping records on paper, and had grown up being very comfortable with computers, eBird was a no-brainer.  I think the very first day I discovered eBird, which was the middle or end of 2005, I signed up and started inputting.  And so, if it was your intention to get people like me into eBird, it certainly worked.  

Yeah, our goal at that point was to make eBird a tool for birders that they would want to use and not feel like they had to use.  That was you begin create participation but also sustain participation because the people who use it, like it, and continue to use it.  And so that was the major change in 2005 that took us from, you know, three to five thousand checklists a month up into the thirty and forty thousand checklist a month range.  And its grown steadily since.  We've had some months with over one-hundred thousand checklists, which is over one million observations.  

So we're happy with that.  This nice thing about the growth is that it allows us to leverage more development, funding and personnel, to build Version Three.  Version Three is essentially going to continue with the same line of thinking that we had before, but we'll sort of beef it up a little bit so that birders have more control over how they want to keep their lists in eBird.  One of the big aspects that we're developing is the ability to build a community around birds, and open things up in a way so that it's less anonymous.  Birders love to be congratulated when they find a great bird, they love the reward they get for their hard work.  We're trying to figure out how to built mechanisms so that other users can give each other a pat on the back, and can also communicate with them when something is wrong.  

So are you looking toward social networking sites like Facebook and MySpace in terms of figuring out how these types of online communities operate or look?

Yeah, absolutely.  Facebook's a great model for how that would work.  At this point we're trying to figure out how to tie in to some of those community sites by, basically, bringing pieces of eBird out so they can be embedded into your Facebook page or things like that.  So, if you wanted to display your Maine list on your Facebook page you could put a little gadget into your page so people could see it.

Which is like the gadget I know of now, which is the pull-down menu by state, where you can see notable sightings broken down by state, correct?

Exactly.  We're trying to, with 3.0, built a very sophisticated home listing base, where you can enter data, manage your data, and manage your lists like you were never really able to before.  But also we're really interested in pushing eBird out to some of these other websites.  You know, when you're a birder you want people to see what you've reported, and you want to know what other people have seen.  So, we're trying to figure out ways to make that happen.  Say, like on Twitter, where you're following what a friend is doing, well maybe on your Facebook page you could set up a feed when you could follow some of your local Maine birding friends and see what they're reporting immediately, either on Facebook or Twitter or your cell-phone.  We're trying to make a more open atmosphere to eBird and also a more immediate atmosphere to it.  Now it takes a long time to get information back...birders want to know about a rare bird yesterday, not a couple of days from now.

If I was going to offer a suggestion for something I'd like to see in a new version, it would be better searching features.  One thing I sometimes do is sit around in Maine and plan trips for myself, or think about what birds I could see if I were somewhere else.  That involves looking at eBird and figuring out, you know, what birds I could see in Oregon in December.  Or April.  And that has been something that's been a little bit difficult to do on eBird.  Are you working on anything for Version Three in terms of searching?

Yeah, absolutely.  In the next few weeks, actually, we're going to put out a Google-map, where you will be able to search for a species and it will give you a Google-map of, say, California with all the reported sightings of a certain species.  Then you can click on those points and get data for each sighting.  That'll be a big step in the right direction in terms of data-out.

One of the other things we've talked about, that we'd really like to develop for Version Three, is, you know when you go the data entry and you get a Google-map and it's got all the hotspots on it?  Well we're thinking about turning that around and making it and input, so that you could go to a map of Oregon, click on a hotspot and then get all the barcharts with everybody's data together.  So if you clicked on a species there you might get the most recent sightings of that species at that hotspot.  That's what we're looking at in terms of hotspot output.

Those hotspots are recommended by users, right?  Then you guys go in and approve the ones that you agree are actually hotspots?  Because one of the other problems with searching is that people label different sites with different names.  So, when someone has seen a certain bird at, say, Bridge 4, it could be hard for me to tell what that location is if it's not a commonly-known hotspot.  Is that someone you guys could address?

Well, hotspots have always been sort of a thorn in our side.  The hotspot concept was essentially a "wiki" concept before "wikis" were even developed.  This was back in 2002, and the idea was that we would let the community create and define hotspots.  So we have this system in place where people can suggest hotspots, but we still have to manually approve them.  And that is a bit clunky.  One thing we have on the table, and I'm not sure it's going to happen in 3.0, is moving the entire hotspots process over into a wiki framework, where the users can provide information about hotspots and manage them.  All of the things we do now we would add to that mapping output.  

That would be helpful because hotspots are really something that is done on a local level.  If five or six people in Maine visit the same site, but call it something different, then the data doesn't back to you and doesn't get back to us as well as it could.  

Right, yeah, and we're painfully aware of issues with hotspots.  What we struggle with all the time is that we have a certain amount of time and resources we can put towards development and we have to prioritize what we think will give us the biggest bang for the buck.  Hotspots is definitely on the list, and it has been for some time, we just have to figure out how to tackle it the right way, and that is probably to put most of the work in the hands of the birders.  You know, birders are fanatical about their birding locations, and it would be an amazing resource to click on, you know, Higbee Beach in Cape May and figure out exactly where to go to find a certain species from the locals who bird there every day.  

Plus, it's really the only way you could do it.  You guys are good birders, but you can't be in every single spot.  If you want to make a comprehensive site, you have to leave it with locals, the people who know places the best.  Are there any other Version Three tricks up your sleeves?

Well we're planning on launching eBird worldwide for Version 3.0.  It'll be great because you could enter data from anywhere.  We get asked for that alot.  We're spending a lot of time and energy now to develop the network we'll need to launch worldwide.  It's going to be hard in some places.  The first step is just allowing people to enter data.  Then it's going to know, the nice thing about eBird is that you can always re-run the data through finer filters.  So, the data in Bangladesh may not be squeaky-clean right off the bat, but when we have better filters we'll be able to reprocess that data through those filters to make sure they're good.  But the first step is to allow people to enter data so they have it on their lists.  

It takes awhile for any community to give polished data.  I remember when I started using eBird, I was putting all kinds of "Xs" and "sp." in, and I wasn't reporting all the birds...I think after awhile you learn the sort of etiquette of eBird, or simply which inputs produce the most useful results.  

That's interesting that you'd say that, that's something we're very interested in at eBird is knowing how eBird has changed the way people go birding.  We've always designed and developed eBird with that goal in mind.  We don't just want people to use it without learning anything.  We want them to know why it's important to report all birds, or to provide some effort information with your records.  It's great to hear that you learned those things without much prompting, really with just interaction with the program.

This goes to the question of, you know, eBird was started with scientific interests in mind, what have you learned in the five or six years that eBird has been around?

The amount of data we're collecting at this point - you know, eBird is a young project - but the amount of data we're collecting really allows us for the first time to look at the distribution and abundance and, sort of overall patterns and dynamics of bird populations at the continental scale.  We can start to look at the movements of Nashville Warblers all the way from their breeding grounds all the way down to their wintering grounds.  It's just amazing to look at an entire species' range as it shifts across the continent.  We're really excited about that.

Right now we have a group of scientists that the Cornell Computers Science Department that are using eBird data in a number of ways to try to model bird abundance at the daily level, on a continent scale.  It's pretty amazing.  They've actually succeeded in doing that for species like Tree Swallow and Eastern Pheobe, species where we have a lot of data.  The scientists are taking that data and tuning it up so we can start to look at some of the stuff for rarer species, obviously the goal being things like: what do these species need, where do they migrate, what is the connectivity look like between breeding grounds and wintering grounds in terms of habitat available, how do you prioritize conservation of those habitats?  Basically making models that help identify species-habitat relationships and fill gaps where there are huge bits of information that are lacking.  Like, where are LeConte's Sparrows during migration?  Where are they between Michigan and Mississippi?

I think that as we continue to collect data, and as birders continue to learn that providing effort-based information and giving checklists with complete species helps us in terms of analysis, we'll be able to learn even more.  Beyond eBirds organizational use as a database, which is great also for sort of the day-to-day record keeping on a regional scale, it's a great tool for organizing information for local journals, state journals and even national journals.  All of those things are being tied in now.  From an analysis standpoint, we're just at the tip of the iceburg with what we can do with the information.

On the same end, you've learned a lot about birds, have you learned anything about birders that you didn't expect? 

Well, Chris Wood and Marshall Iliff and I are what they call the "three-headed monster" that runs eBird now, and all three of us have been birding for a good portion of our lives.  We're pretty dialed in about what birders want.  I think some of the scientists at Cornell have learned a lot about birders by seeing the results we've got with eBird by making some simple tweaks that appeal to birders.  It's been a great learning process for citizen-science in general to learn that when you build a project you need to consider the needs of your audience instead of the needs of the scientists.  Or, I should say, as well as the needs of the scientists.   You need to not only develop a project that will gather useful data but it needs to be a project will result in some sort of useful reward for the user.  That's been the biggest thing that's been learned from this entire process, and I think a lot of citizen-science projects in the future will be based on that model.  ø


About Us | Site Map | Privacy Policy | Contact Us | Blog Design | 2007 Company Name