Turning the iNaturalist API into a Spatiotemporal Service

Using SeerAI's Boson data mesh technology

Presented By SeerAI

In this post, we will describe how we were able to convert iNaturalist’s REST API into a fully-fledged spatiotemporal data service that can be used in GIS workflows. Here, we have the iNaturalist data served on an ArcGIS map for the Denver area, which we accomplished by using SeerAI's Geodesic platform - specifically, the geospatial data mesh we call Boson.

Powered by Esri

This map is dynamically serving data from the iNaturalist API by using SeerAI's Boson data mesh technology. Try clicking on a feature to see more information about the observation!

Boson and iNaturalist

SeerAI’s Geodesic platform has several powerful components. One of them is Boson, a geospatial data mesh. Boson allows users to stream data from wherever it lives natively, into the Geodesic platform, where it can used in conjunction with other data, served to maps, and otherwise incorporated into GIS workflows.

In this example, we show how we were able to do this with the iNaturalist API. iNaturalist is a website and app that powers a crowd-sourced nature observation effort — users can upload photos of plants or animals, and the iNaturalist community identifies them. iNaturalist maintains a database of these observations and identifications, along with a wealth of supplementary information.

Open link

Description of iNaturalist

Basic explanation of how iNaturalist works

Users can retrieve these data on the iNaturalist website, or by interacting with their API. These methods allows users to query the data like any ordinary database, but without geospatial functionality. The data can’t be served to maps or incorporated into GIS workflows without significant additional effort.

How Boson is better - the ordinary way

Here's how we would access the data using Python's requests library in the absence of the Geodesic platform. In this example, we search for all observations in January, 2024 within a bounding box around Denver, CO. 

import requests

api_url = "https://api.inaturalist.org/v1/observations"

search_parameters = {
    "nelat": 39.938698,
    "nelng": -104.698334,
    "swlat": 39.579232,
    "swlng": -105.225677,
    "d1": "2024-01-01",
    "d2": "2024-01-31"
}

res = requests.get(api_url, params=search_parameters)
data = res.json()

This give us the data in json format. That's great, but if we want to use this as geospatial data, we will need to parse the responses, create geojson features, and iterate over those features to make a feature collection. Even then, all we have is a static dataset. If the underlying data change, or if we want to do a different query, we have to start over at the beginning. Additionally, the size of our queries will be limited by memory or speed or both.

How Boson is better - the Boson way

That’s where SeerAI’s Boson technology comes in. Using the Boson Python SDK, we create a remote provider, which serves as the intermediary between the iNaturalist API and Boson. The remote provider is what turns the iNaturalist API into a spatiotemporal data service in Boson. With the remote provider hosted in Google Cloud run, we simply add iNaturalist to Boson using the remote provider's url. We perform the same query as above with the search method, this time using a standard geospatial bounding box and datetime range. Unlike in the example above, the "data" variable in the final line is already a feature collection.

import geodesic
import datetime.datetime as dt

ds = geodesic.Dataset.from_remote_provider(
    name='inaturalist', 
    url='https://inaturalist-remote-provider-azwzjbkrwq-uc.a.run.app'
    )

result = ds.search(
    bbox=[-105.225677,39.579232,-104.698334,39.938698],
    datetime=[dt(2024, 1, 1), dt(2024, 1, 31)],
    limit=None
    )

data = result['features']

Once this API has been converted into a Boson dataset, it can by dynamically queried in a completely scalable way. Furthermore, all datasets in Boson behave in the same way, regardless of whether they come from STAC catalogs, ArcGIS Online Living Atlas, Google Earth Engine, flat files (geojson, geoparquet, etc.) - all of which are easy to add into Boson. What this enables is easy, seamless, and scalable data fusion.  For an example of how easily you can work with multiple datasets from multiple sources using Boson, see this example. 

Going beyond feature collections

Additionally, Boson allows you to serve your data out to a variety of geospatial services. Instead of plotting a feature collection on a map, we will now serve the dataset as an ArcGIS feature layer. Geodesic has a web app that makes this easy.

First, we open Entanglement, Geodesic's knowledge graph. When we created the dataset, it automatically gets added to the knowledge graph as a dataset node. We can now find the dataset with a simple search.

We can find the dataset in our knowledge graph by searching in the Geodesic web app

Clicking the share button will generate a share token of our preferred type. Here, we've shared as an ESRI geoservice. URLs to share as a feature service, feature layer, tile layer, etc. are shown.

We copy the feature layer URL and use it to add a new feature layer in ArcGIS online.

We give the layer a name, and save it.

Finally, we add that new layer to an ArcGIS web map, and we have the iNaturalist data on full display! Try clicking a feature to see additional information.

Conclusion

We have shown how Boson enables users to stream data from REST APIs as if they were honest-to-goodness geospatial data services. This allows users to seamlessly query these APIs in their GIS workflows and to do so at scale. Furthermore, datasets in Boson behave in a uniform manner, regardless of their source,  enabling data fusion with unprecedented ease . To learn more about how to harness the power of Boson for spatiotemporal data fusion, please check out these  cookbooks  on our  documentation page . If you have further questions, please  contact us.