Merlin in Japan, March 2026: the new age of bird watching III

 

Grrrrrrrrrrrrrrrrrrrrrrrrrrrrrrr TAX! … So sorry skipping last week for telling my forest adventure with AI. Anyway, this week is for my impression from Merlin in the nearby forests. First, you download the App from Apple Store or Google Play Store on your cellphone. To use the apps fully, please create your account in eBird and tell Merlin you’ll use your eBird account for Merlin. (More to it later.) And don’t forget you keep logging in both eBird and Merlin when you bird-watch with these two AI app. The starting page of Merlin would look like this (Sorry, in Japanese again).

The three icons in the middle of the page tell you the options we can choose to identify a bird somewhere near. From the left,

The icon is to identify bird (single or plural) with flow-chart. You report where you find this bird (address, or GPS data your cellphone report), time (calendar and time the cellphone report to the system), and, if you can see the bird, the size, the most prominent color of it, and finally the activity the bird is doing. When you report them, and tap Next, Merlin replies back the possible candidates based on your saying like this

A part of the list from my reporting.
I saw Brown-eared bulbul and
Merlin reported back this.
An apparent mistake.

When none of the suggestions fit your sighting, Merlin gives you opportunities to adjust your reporting at the end of the suggestion page, like this

They are the items you reported.
Maybe the size I told Merlin was not exact …

You continue this exercise until you’re sure the bird you’re watching is “This,” just like searching in the pages of reference books. If you think none matches of your finding in the end, you click the bar (directed by the red arrow in above photo), and AI will remember the problem which will be utilized by the next identification exercise in its artificial brain.

The right logo (above) is to identify a bird by photo. I think the way of this approach is basically the same as the usage of picture reference books. AI Merlin sees the picture, searches its huge picture data book, and decides the most likely species for the image. In the training by Wild Bird Society, we tried it with ppt slide showing professionally taken phot of birds like this.

i.e. the bird shown was already identified by pros, and the photo was taken by a pro. Merlin replied to us exactly the “right” answer. For the above it was Eastern Buzzard (Buteo japonicus). The problem is, can we take the above level photo with our cellphone in a field when a bird suddenly comes out in front of us? That is the question, oh yeah. Maybe you have iPhone 17 Pro Max, and the machine will give you the best shot … lucky you. I’ve never tried it in the field. In any case, seeing a bird in a forest and taking a nice photo is difficult at best. The best way would be to sit down quietly, hold your cellphone steady (that would be difficult unless you use supporting gears), wait for a long long time until a bird comes in the frame, and press the shutter button of your cellphone at the right moment. Good luck.

So, the most useful approach to ask help from AI in forests should be the middle icon, which is recording the voice of a bird and Merlin identifies it from its audio data. We tap the icon, and Merlin starts recording.

Recording and thinking …

When AI recognizes the voice matching to the data from its database, it will suggest the identification result like

It surely is a useful tool when the condition is right. I mean, when we’re surrounded by noises, not only from transportation traffic, chatting with our friends, etc., but also of winds and the resulting creaking of swaying bamboos of Niiharu, or babbling stream of Tanzawa, Merlin cannot identify the voice. I also tried the audio data recorded by TASCAM, cleaned the noise with Audible to extract birds’ voice, and played the adjusted data in front of my Merlin. Nah. The artificial brain requires the best sound condition to function … There are much room for Merlin to improvement, I guess …

It says, in Japanese,
“No suggestion.”
At least Merlin is honest
telling us it has no idea
for the owner of the voice.

In any case, unless enough data is stored in its database, even if Merlin stays in the best condition to listen to birds’ voices it cannot identify the species correctly. We must help by trying many times and reporting the data. I suppose Merlin’s main database is eBird, and so we need to keep signing-in both. The process goes like this. When we think we’ve done the recording session, we stop the mike. Merlin will ask whether to store the recording in your cell phone. We tap “Store the data” button. Your sound data is stored in your phone as .wav file. Mind you. “.wave” file is large. Merlin certainly has a function to send this data to your email or cloud storage with “share” icon. But your storage will quickly be occupied by the recording, and you may find problems with the other tasks.

“Oh I know.
I think it is a voice from a long tailed tit,”
Merlin said.
Stop button is that red one
 in the middle of the screen.

“Save” button.

So, I just stored the data in my pixel and moved the data to eBird manually. I open my eBird account to report my findings with sound data. The way to upload sound data to eBird offers us two options. On is upload the data as “Soundscape,” and another is to designate a particular data for a specific bird. Even if we have only one recording, we can upload the data for all the cases like this.

But it is cumbersome way for reporting the result. Instead, we can upload the data just once for “Soundscape,” then edit the reporting memo like

We’ve done the recording, and keep crossing our fingers Merlin’s ability to identify birds will be improved using eBird database. I hope reporting process from Merlin to eBird becomes less burdensome someday soon. Let’s keep crossing our fingers. Playing with Merlin is not useless, I wish. But … not all agree with me, it seems.

Last month there was an annual symposium held by Tanzawa-Oyama Nature Restoration Committee. This year, the theme of the talks was to promote Nature Positive by mobilizing people power for data collection. The panellists were upbeat to use cell phones and aggregating reports in GBIF and iNaturalist. (eBird is feeding the data to GBIF.) But many audience (who were regulars for such gatherings, and so, it seems to me, veterans, aka senior citizen males) were VERY sceptical in the usage of AI and collaborate with the movement. One leader of an organization in Ebina City 海老名市 proudly said “We don’t care what’s going on in the other forests, and do not see the point of sharing our data globally by taking our time from learning whateva of AI and reporting our findings. What’s wrong with writing a private memo of our flowering, and keep it private?” The other attendees said “To begin with, we don’t trust such AI or something. We do not find the merit of using such. Our familiar way of using reference books must be the best.” The speakers of the meeting were all professors or senior researchers of Tanzawa who were staying in their laboratory normally and do not come out much from there, I guess. They looked a bit embarrassed by reactions from the congregation.

The occasion held a panel session as well.

Probably, it is just a matter of generation gap, and the incidence at the symposium is common when a new technology enters a scene. I simply hope Merlin will learn more and more, and someday soon we can use it more easily in our forest nearby. Let us see how things turn out.

If you have any questions about Yokohama’s Green Tax and Green Up Plan, please make a contact with

Strategic Planning Division, Green Environment Bureau, City of Yokohama
横浜市みどり環境局戦略企画課

Phone: 045-671-2712
FAX: 045-550-4093

Email: mk-kikaku@city.yokohama.lg.jp