How I used technology this summer to learn about the wild birds in my neighborhood

BirdNet is a cell phone app that identifies birds from their songs. Here are my observations after using it for 3 months.

---

Introduction

Back in June, I posted a SteemLinks article, How I used a Raspberry Pi to automate birdwatching - ZDNet. The underlying source article suggested use of two applications, Merlin, and BirdNet. According to the article, Merlin helps with identification of birds by sight and by sound. BirdNet, on the other hand, only assists with identification by sound.

I immediately downloaded both applications with the intent of trying them out. It turned out that Merlin required email verification, so I terminated my relationship with that one, but kept experimenting with BirdNet for the next few months when I was out in the yard with the dog. Here's what I found.

Usability

Getting started was very intuitive. There are basically only three options to choose on the main screen: "Record", "Select", and "Results". And the options are numbered so that the order is obvious.

After choosing the "Record" option, the user is presented with an image of a microphone (which starts recording) and a pause button. So, to get started, you just launch the app, go where the birds are singing, then choose "Record" and activate the microphone. The display has a visual spectrogram which makes it easy to see when the application has picked up bird sounds.

Once the application has picked up the bird songs, you press the "pause" button and move on to step 2, "Select". To select the bird sounds that you want to analyze, you simply drag your fingers across the relevant portion of the spectrogram until you have the bird sounds bracketed. With that done, you can pick from two options: "Play" to listen to the recording or "Analyze" to have the AI make a guess at the bird that's doing the singing.

Once the analysis is done, the screen automatically flips to the third option, "Results". In this section, it will tell you the bird species that it thinks it heard, as well as the estimated likelihood that it's correct. These values range from "Highly Uncertain" to "Almost Certain".

After analysis, each of the observations is saved in a list, and there's also an option to save it for later analysis. It's also possible to replay, share, or delete the saved observations in the list. The saved observations also include the locations (including a map), time, and date.

For each identified species, it also links to the wikipedia page where you can see a picture of the bird and learn more about it.

In addition to the main screen, settings are available to customize the following settings:

• Language for bird names
• Time of year for species selection (Weekly or Year Round)
• Detection Sensitivity
• Audio input source
• Spectrogram colormap
• Spectrogram length (15-30 seconds)
• Amplitude gain
• Max. frequency
• Spectrogram contrast

I have not played around much with those settings, as the default settings seemed to work pretty well.

Accuracy

Many times, I couldn't actually see the birds that the app was recording, so I don't really know for certain how accurate it is. I will say that it was pretty consistent at linking similar sounding bird songs to the same species of birds. Also, the species that it identified were all native to the region. So, I can't say for sure, but I believe that it was pretty accurate on the guesses that indicated a high level of certainty.

Something interesting which I hadn't realized until scrolling through the list of observations while writing this article is that the species that it detected seemed to change during the course of the summer.

Practical Results

I began observations on June 20, and from then until September 30, I recorded 110 bird observations.

The main challenges that I had were that, in summertime, people in the neighborhood were frequently running lawn equipment that drowned out the birds. Similarly, I live close to a highway, and highway noise also overwhelms the singing birds. Finally, most of my observations were made while walking the dog, and the dog likes to bark when he sees other animals, so many bird sounds were overwhelmed by a yapping German Shepherd Dog.

My first observation on June 20 was a Fish Crow and it was listed as "Highly uncertain". Over the course of the following week, I also observed the Veery or Wilson's Thrush, Tufted Titmouse, and the Carolina Wren, the American Crow, American Robin, and the Baltimore Oriole. Of those, the Carolina Wren and the Tufted Titmouse were the only ones logged as "Almost Certain".

I'm not going to list all the different birds I recorded (in this post), but it's interesting to compare them with the birds I recorded in the last weeks of September. In the last part of September, I was still recording Carolina Wrens, but most of the other birds from June were absent. Other birds that I recorded in September included the Blue Jay, the Northern Flicker, and the Gray Catbird. By this time, I had learned what sounds the app was likely to pick up, so most of the observations scored as "Likely", "Highly Likely", or "Almost Certain".

To be honest, I had no idea that I'd find this many varieties of birds in my own yard. It was quite surprising. Every time I noticed a new species, I'd follow the link to wikipedia to see if I'd found a rare bird or not, but it turns out that every one I checked was listed as "least concern" under their conservation status.

Two other things surprised me, just because I had never really paid attention to the bird songs outside before. One is that they were most active - by far - early in the morning, say around 6am or so. Second is that during the hot months, late June, July and August, there were a lot of birds singing late at night, even midnight or later. Scanning through my list of observations, it looks like the latest ones I recorded were a Tufted Titmouse, a Baltimore Oriole, and a Northern Mockingbird all between 1:20 and 1:30 in the morning on June 28. In September and October, I haven't noticed as much nighttime singing, and I also haven't been using the application as frequently.

Finally, I'm still not good at recognizing the songs without help from the app. Developing that skill would, apparently, take a good deal of practice.

Conclusion

So there you have it, if you're interested in Citizen Science or bird watching, I definitely recommend the BirdNet application, and if you're willing to register your email address, it looks like Merlin would be fun, too.

Thanks for your attention!

---

Thank you for your time and attention.

As a general rule, I up-vote comments that demonstrate "proof of reading".

---

Steve Palmer is an IT professional with three decades of professional experience in data communications and information systems. He holds a bachelor's degree in mathematics, a master's degree in computer science, and a master's degree in information systems and technology management. He has been awarded 3 US patents.

---

^{Pixabay license, source}

Reminder

Visit the /promoted page and #burnsteem25 to support the inflation-fighters who are helping to enable decentralized regulation of Steem token supply growth.