Panama Live: From Observations to Visualizations

June 30, 2020

Panama is home to hundreds of tropical bird species, many of which we know very little about. The 24/7 Cornell Lab of Ornithology’s Panama Fruit Feeder Cam, located at the Canopy Lodge in the Anton Valley, gives us a window into that world and a chance to see these species up close! While researchers have documented feeder visitation behavior of many bird species in the Northern hemisphere, they haven’t done so for many bird species in the Southern hemisphere.

To learn more about these incredible birds, the community that watches the Panama Fruit Feeder Cam teamed up with scientists to design an investigation. After weeks of watching the cam and discussing potential questions, we settled on these three:

1. When do focal species arrive at the feeder?
2. How does this vary from day to day?
3. Does the timing of food affect when birds arrive?

With the questions in hand, we voted on studying six easily recognizable species: Clay-colored Thrush, Crimson-backed Tanager, Gray-headed Chachalaca, Gray-cowled Wood-Rail, Rufous Motmot, and Thick-billed Euphonia. We also determined that the easiest way to record information about food would be to record the moment that food is put out on the feeder platform.

For the first time ever, we collected data in real time using our new live data collection tool (see how it worked in this short video that opens in a new tab). Over 60 people contributed to the effort by watching the Panama Fruit Feeder Cam live and clicking buttons whenever they saw one of the six species arrive or when food was put out. After collecting data for about two weeks in mid-February, we amassed a total of over 11,000 observations.

Our next challenge was to decipher what the observations meant and consider different ways to interpret and summarize the data. Multiple people could have watched at the same time, for different periods of time, at different times of the day. Even if people were watching at the same time, their clicks will probably not match up exactly because there will always be a delay between when something happens and when each of us can click the button to report it. Additionally, we had to account for times when the camera was offline and eliminate those observations from the dataset.

First, we limited the dataset to the times that the live cam was streaming. Next, to be absolutely sure when the cam was and was not being watched, we limited the data set to only include participant sessions with a clear “stop,” time indicated by clicking the “End Data Collection” button.

We then looked at this dataset to answer our first research question: when do focal species arrive at the feeder? We had to figure out what observation time interval would make the most sense. The time interval needed to be short enough that most of it would have been watched by participants, but not too short that visualizing it would be impossible. An hour was too long while five minutes was too short. In the end, we decided half-hour time intervals from 6:30 a.m. – 6:30 p.m. were the most appropriate.

For each half-hour interval, we determined how much of it was watched by one or more participants. If 15 minutes or more was watched, we considered that interval “watched.” Then, we assigned a “1” if the species was observed arriving at the feeder (present) and a “0” if it wasn’t (absent), creating a “presence-absence” dataset. Once we had that, we calculated the probability, or percentage chance, that a species would arrive during each half-hour time interval. For example, for 6:30 – 7:00 a.m., if the Rufous Motmot was there for four days and the total number of days participants watched that interval was eight days, then the probability would be 4/8 or 0.50. To turn that number into a percentage we multiply by 100.

We followed a similar method to answer the second question about day-to-day variation, but this time calculated the percentage chance that a species would arrive during a half-hour time interval for each day. For instance, on February 24, if the Rufous Motmot was at the feeder for 2 out of the 22 half-hour time intervals watched that day, the probability would be 2/22 or 0.09, a 9% chance.

To answer the third question about the influence of food on arrival, we followed a similar method with an added first step. We needed to figure out when food was put out because sometimes multiple participants clicked at different but similar times when that happened. If observations of food being put out were within five minutes of each other, we “binned” them together and used the first recorded time as the one to indicate when food started to be put out. Then, we excluded eight times food was put out because it was within one hour of a previous time and we wanted to document species arrival within the time span of one hour.

To figure out the percentage chance that a species would arrive within a certain time period of food being put out we chose a time interval of five minutes. This allowed us to break up the one hour or 60 minute interval into 12 distinct intervals (from 0-5 to 55-60 minutes) and figure out when participants did and didn’t watch. We considered an interval “watched’ if three minutes or more was watched by at least one person. As before, for each species we assigned a “1” if it was present and a “0” if it was absent for each time interval. Before we calculated the probability, or percentage chance, we grouped the food being put out into three separate groups to account for any daily variation: morning (6:00-10:00 a.m.), midday (10:00 a.m. – 2:00 p.m.), and evening (2:00 – 7:00 p.m.). With the presence-absence data in hand we calculated the probability, or percentage chance, that a species would arrive during each five-minute interval since food was put out in the morning, midday, or evening.

Once we had these data summaries we set to work creating interactive graphs for the community to explore. As in any scientific investigation, it’s an opportunity to explore the data and see what they can tell us; we can then do statistical analyses to see if the patterns appear to be meaningful. We encourage everyone to join us and explore the interactive graphs. Please share your thoughts, observations, and questions in the forums below each graph. We’d love to consider your observations and conversations as we summarize insights in the final report for Panama Live.