Speak for the Trees: a weekly newsletter about your ecosystem (using iNat + 15 other data sources)

Third-party Tools and Apps using iNat API
1

Goal of the tool:
Give ecosystems (a town, watershed, forest, coastline, etc.) a continuous voice by turning public environmental data into a weekly newsletter. You pick a place, subscribe to the newsletter for that location’s ecoregion, and each week the platform surfaces something noteworthy from recent data with a concrete call to action (commenting on proposed legislation, volunteering at a local event, submitting an iNat observation).

Niche it fills in the iNaturalist ecosystem:
We’re a consumer of iNat data, not a data source. We pull recent observations for a subscriber’s ecoregion via the API and combine them with 16 (and growing) other public environmental datasets, such as EPA clean water act violations, harmful algal bloom sightings, drought conditions, and more, in order to write a “state of the ecosystem” narrative for that place. iNat observations are what tie that narrative back to specific living things people can actually go see: we use them to celebrate recovering species populations, flag increases in invasive observations, and draw hypotheses about how other observed conditions might be affecting local species. When an observation is featured, we credit and DM the observer.

Is there a commercial component, or do you plan into include one in the future? Are donations requested? (yes/no, explain):
Currently, there is no product for sale. We do ask for voluntary donations (link removed by moderator) to cover hosting/API costs (currently targeting $1,000/month) which supports running the platform itself. Separately, some posts link out to donating to third-party conservation orgs relevant to a location; that money goes to those orgs, not to us.

What sort of data (if any) does your app collect from its users?:
Email address (passwordless sign-in), and first/last name only if you sign in with Google or file a public regulatory comment through the site. If you follow a location, we store the coordinates/place you searched and its matching ecoregion. Standard product analytics. From iNaturalist specifically: nothing beyond what’s already public via the API for observations we cite (observer username, taxon, location, date). We don’t store iNat login credentials, and observer thank-you messages are currently sent manually by a team member through iNat’s own messaging UI. Full policy at speakforthetrees.com/privacy .

Link to your iNaturalist profile:

https://www.inaturalist.org/people/acyanlight

Description:
Speak for the Trees uses public environmental data to send a weekly newsletter about your local ecosystem. Each week, we fetch recent data for your location and publish a post about something noteworthy, with a call to action to get involved. iNat observations are one of 16+ data sources feeding these posts; we also monitor facilities in violation of clean water act discharge permits, harmful algal bloom sightings, drought conditions, and more.

Discussion questions or areas seeking feedback:

  1. We use iNat observations to celebrate recovering species populations, flag increases in invasive observations, and draw hypotheses about how other conditions might be affecting locally observed species. What other topics would you be interested in learning about?
  2. Are there other aspects of iNat data we should be including?
  3. Feedback on the report format itself: what do you think about the content we send about your local ecosystem?
1 Like
2

This sounds like it might be some kind of AI-generated synthesis which I would definitely not be interested in. Raw iNat data really isn’t appropriate for assessing

unless done in a very careful, rigorous way.

5 Likes
3

Reiterating my reply from when you first posted this yesterday: if there is any machine generated text, then I for one am not interested spending my time reading it.

3 Likes
4

yup. especially considering:

i don’t know exactly how back end is structured, but this does seem like a very large cost if you’re just pulling and hosting data. i wonder how much of this cost would go to AI?

5 Likes
5

Unless you actually plan to hire real human writers to cover each local area - and I don’t see that happening - this idea seems very dangerous.

Especially letting AI urge people to take specific political action. Even if your own intentions in creating it were good, that could go very very wrong fast.

3 Likes
6

I appreciate your position here. We don’t claim to make scientific or statistically significant findings here. Our primary goal is to build local ecological literacy and point folks towards opportunities to take action. The data used in each post is mapped and cited.

7

We group subscribers by EPA level 4 ecoregion. We produce one newsletter per week per ecoregion, and all subscribers within an ecoregion receive the same newsletter. To minimize AI consumption, we do not support user chat. Our costs scale with per ecoregion with >0 subscribers.

We currently have subscribers in ~100 of the ~900 level 4 ecoregions. Our current AI consumption is approximately $150/mo. Scaling to full U.S. coverage would be a little over $1k/mo. If we can help increase ecological literacy of thousands of people and help connect them to local opportunities to take action, I think that is an efficient use of capital.

1 Like
8

Thanks for your perspective. We are speaking with conservation groups about helping to support their work. Several have mentioned that their teams are small and spread thin, and it is difficult to monitor all of the places they steward. We believe there is opportunity to support these conservation efforts by being a first-draft tool that helps identify topics for them to investigate further and then write about. This is something we are still exploring.

Another line of thinking is having our system flag anomalies to local journalists. For example, last week, our ecosystem agent noticed a spike in spotted lanternfly observations around Springfield, Massachusetts. Our system flagged the post as worth sharing with local reporters for further investigation. A reporter wrote back the next day. After reviewing the data we shared, speaking to a few local experts, and interviewing Rob, they ran the story.

9

paywalled

10

I don’t think this

is an out. If you make claims, you need to be reasonably certain that they are true/correct. If the newsletter is just posting statements that an AI model thinks are correct based upon iNat data that it hasn’t been specifically trained to evaluate, and a trained human hasn’t vetted, then it’s irresponsible in my opinion to make those statements and just link to the data. That’s not going to

if the actual content isn’t correct or reliable. I’m not interested in that kind of content.

7 Likes
11

I mostly agree with you. We have built tests to validate the information being published where possible. It isn’t perfect, and there will be some mistakes. Too many mistakes and it erodes trust in all of the content.

If we are able to minimize those mistakes, I believe there is a role we can play here.

12

i read some of the posts, and most of what i see is just gobbledygook. if you want to present opportunities for action, you can just post a calendar of important events. if you want to present what’s happening, it’s better to present a dashboard of important metrics. maybe show where there might be interesting things happening on the ground for people to go see. if you want to present a small section at the end that contains some AI verbiage that tries to tie it all together, that’s fine, but right now, the main thing is the AI stuff, and i think what you’ve got right now is going to point people to act on information that isn’t going to help them act in a useful / meaningful way..

you also erode trust when you use other people’s content without attribution when their licensing explicitly requires it:

image
image1920×1080 425 KB

3 Likes
13

Here’s a non-paywalled link: https://www.masslive.com/westernmass/2026/08/spotted-lanternfly-putting-mass-in-sticky-spot.html?gift=6f96cb8b-fcff-45b9-b183-9eb58b8b63f9

14

We have built some raw data views that you may find useful, and these data sets are starting to help inform the content that is being written:

  • This view displays the compliance status of the ~550k facilities registered through the Clean Water Act to discharge into a body of water. This data is provided through EPA Echo and is available there, but we are working to improve the legibility of that information.
  • This view displays the locations of thousands of conservation organizations across the country. For those large enough to be require to publish financial records, we are also displaying that information.
  • This view shows the USGS National Land Cover Data grouped by ecoregion, with an option to show y/y change.

As for attribution, the points on the map all link to the specific observations on iNaturalist. But you’re right that the pop-up should attribute the source as well — I will add that.

15

in that particular case, there is no observation in iNaturalist. you’re just using iNaturalist as the source for a taxon photo. in my mind, that’s bad karma. even though this particular photo is hosted in the AWS iNatuarlist Open Dataset bucket, if there’s another one that’s hosted by iNaturalist, that would be traffic that iNaturalist would have to pay for.

1 Like
16

The screenshot you shared shows an observation of that species on the map. Click to expand the map, then click that point.

I’m adding attribution to the pop-up as we speak.

17

the photo that i pointed to earlier isn’t from any observations on the map. it’s just the primary photo for the taxon in iNatuarlist. it comes from an observation more than a hundred miles away from the map area from 20 years ago.

as i noted, you haven’t attributed the photographer according to their license terms, and you’re parasitically sourcing images from another organization’s servers.

2 Likes
18

With respect, that does absolutely nothing to address my concern about an AI “newsletter” potentially urging subscribers to support misleading local legislation, and/or hallucinating upcoming legislation that does not actually exist.

If your goal is to surface possible leads for journalists to follow up on, please skip the direct newsletter part.

1 Like
19

Ah, now I understand. You’re correct that the species photo is not from any of cited observations. It is instead provided by the /taxa endpoint as a default photo for the species. Double checking those docs and I now see that it also provides the user, so I can update the citation to mention them.

20

This sounds good and looks good - I did actually take a look at a few African locations and hopped out before signing on. What I did see looks promising. However - you are using terrestrial ecozones only and excluding marine ecozones completely, a seen from the few coastal ones I typed in and have personal experience of and having visited those actual zones. This will eventually lead to skewed data, especially if others have mentioned, you are not managing the use of the AI back end.

Also, are

USA or North American or EU centric only? Meaning of course that if your only data for Putsonderwater is USA sourced from the EPA, WWF etc, then you have not actually managed to pull data for the region from local sources. It might be that only one of the data sources you have is dealing with the region mentioned, which will not make for good data or good AI. If you cannot manage to get the bare minimum of data points for a specific region, I would urge you to rather not make that region available. An unavailable region is better than an AI trained report showing something from 1 or 2 data sets only which AI trained itself as the be all and end all of data for that region. You will of course realise that a call to action on one dataset is not a good thing.

Further - (and this is actually about a tree) iNat data is not always what it seems. The CNC2026 results for Cape Town included a mention that the https://www.inaturalist.org/taxa/53421-Pinus-radiata is endangered - which it might in its native region but it hardly is endangered in South Africa where obs took place and was included in the CNC results as endangered. In fact it is classed as a category 1b invader which should be eradicated except in controlled plantations (https://invasives.org.za/wp-content/uploads/2022/05/South-Africa-Listed-Invasive-Species-A5-Booklet.pdf). The concern here is of course that you call people to action when local government starts ripping these out, which should hardly be the goal. You cannot possibly be flagging an increase in this population as a cause for celebration merely because iNat says it is endangered. So ultimately you should disclose your data sources and ideally perform an audit of the data you have pulled before publishing the report.

As for drawing hypothesis on data reported, uhm, no definitely not. If your target audience are not scientifically minded, the difference between wild guess, hypothesis and theory will be about as clear mud. This again endangers the conclusions drawn and the call to action that might occur from it.

This is a good idea - but you need to do some work and some more actual research before letting another AI bot loose on the world - nice graphics is the same as fashionable clothes - it wears thin after awhile.