Geo-Powering Insights: The Art of Spatial Data Integration and Visualization

Summary

• Geospatial lakehouse architecture: The Databricks Data and AI platform enables seamless ingestion and transformation of vector, raster, and point cloud geospatial data with SQL and distributed processing.
• Multi-source integration: Connect to external geospatial systems like ArcGIS, then create custom visualizations with Databricks Apps, Streamlit, Flask, Tableau, or Power BI on the Databricks Data and AI platform.
• Spatial analysis at scale: Leverage H3 hexagonal indexing and advanced spatial SQL to perform complex geospatial analysis on massive datasets—supporting real-time insights.

Geo-Powering Insights: The Art of Spatial Data Integration and Visualization

Talk by Mathieu Pelletier, Specialist Solutions Architect, Databricks
In this presentation, we will explore how to leverage Databricks' SQL engine to efficiently ingest and transform geospatial data. We'll demonstrate the seamless process of connecting to external systems such as ArcGIS to retrieve datasets, showcasing the platform's versatility in handling diverse data sources. We'll then delve into the power of Databricks Apps, illustrating how you can create custom geospatial dashboards using various frameworks like Streamlit and Flask, or any framework of your choice. This flexibility allows you to tailor your visualizations to your specific needs and preferences. Furthermore, we'll highlight the Databricks Lakehouse's integration capabilities with popular dashboarding tools such as Tableau and Power BI. This integration enables you to combine the robust data processing power of Databricks with the advanced visualization features of these specialized tools.
Here's more to explore:
The best data warehouse is a lakehouse: https://www.databricks.com/product/databricks-sql
How to migrate legacy data warehouses: https://www.databricks.com/resources/ebook/migrate-your-legacy-data-warehouse-databricks
See all the product announcements from Data + AI Summit: https://www.databricks.com/events/dataaisummit-2025-announcements

Chapters

• 0:00 Introduction and mission
• 2:09 Geospatial lakehouse overview
• 4:55 Platform architecture
• 6:56 Data ingestion and pipelines
• 9:52 ArcGIS integration
• 14:20 Analysis and H3 indexing
• 18:16 Serving and visualization
• 20:53 Demonstration and conclusion

FAQs

What geospatial data formats does Databricks support?

The Databricks Data and AI platform ingests vector data (polygons, lines), raster data (images, satellite), and point cloud data using standard geospatial formats.

How do we connect to ArcGIS?

The Databricks Data and AI platform integrates with ArcGIS to retrieve geospatial datasets, enabling seamless data ingestion from enterprise GIS systems.

What is H3 indexing?

H3 is a hierarchical hexagonal indexing system that simplifies geospatial analysis on the Databricks Data and AI platform, enabling efficient spatial queries and aggregations.

Can we build custom geospatial dashboards?

Yes, create custom dashboards with Streamlit, Flask, Databricks Apps, or connect to Tableau and Power BI for advanced geospatial visualization on the Databricks Data and AI platform.

Full transcript

0:00 Introduction and mission

Welcome everyone to this next breakout session on the art of spatial data integration and visualization. I'm going to need be the speaker for this presentation. First a little about me. I'm a specialist solution architect here at datab bricks for two years now and before that I was a software engineer and I transitioned to a data engineer because I thought analytics and big data was way cooler than web development and regarding joial experience I worked at Expedia and also for a telical company in Xedia we were responsible for the uh backend the search for when you search for a point of interest hotels and region uh this is my team that was responsible For that we use postgis and Google maps.
So enough about me you know we're not here for that. Uh let's focus on I have a today's mission. I have something for you. Let's provide explanation on UFO sighting nonetheless. And how we going to do that? We're going to leverage geospatial lakeups to do that. And to do that we need to ingest the data first from multiple sources multiple format. All of these are zero reference and we're going to also try to read data from ARJS to talk how to integrate uh with RGis.
Then the f next step will be to analyze the data using special SQL and H3 indexing. If you're not familiar with H3 we're going to cover that later on. And also we're going to try GAI. You know why not? Let's try to find explanation reasoning for UFO sighting using GI. And at the end we need to serve the data publish the results. And if you are curious and you want to use this repo, this is the my my source code. I'm going to share it with you later on.
All right. I really like this quote by. Extraordinary claims require extra evidence. So let's try to find evidence in the data using geospatial. I'm going to first do an overview on the platform. Joial data comes in multiple shapes and form and I'm curious if you could raise your hand who is dealing with vector data lines polygons okay almost uh 60% raster data images satellite okay there's a bit less mobility mobility okay cool LAR okay okay yeah okay great so just to let you know we can ingest and transform this data using the lakehouse.

2:09 Geospatial lakehouse overview

We have best practices for all of them and let's focus more on the lakehouse architecture. So the first step it's the ETL layer to ingest data into the platform. We can ingest any special format vectorbased raster base point cloud. We have best processors for each one. We have patterns. We support spatial indexing uh gr um coordinate system transformation and we can also support batch mode and streaming mode depending on your use cases.
So this is for ETL. Regarding warehousing built into the platform, we have SD function, we have H3 indexing and what's going to be announced as well this week in the summit. We're going to have a native geospatial type geometry geography. So stay tuned on that. Our SQL platform is open and NC compliant. So if you are familiar with traditional RDMS databases like Oracle, SQL server possibilities regarding the storage layer.
This is pretty interesting. We are uh open data lake and that enables you to store your geospatial and nonspecial data at the same time at the same place collocate it. So you don't need like different platforms to do that know that creates silos and storing this data along your nonje files. You can develop application on top of that dashboard machine learning model. So your end user they don't need to work with different platform that are segregated and copy data back and forth and we support uh most geospatial uh format wtb and again something I mentioned we're going to have our own built-in spatial type coming soon this week actually and what is interesting what our storage layer uh we have predictive optimization through delta lake and you don't have to worry about partitioning indexing this is done for you you can focus on developing applications.
This is a higher level overview of our platform. We have the different sources at the left and then you can bring the data inside the platform using a specialized tool like FME by safe or using our own tool called loot pipeline. We're going to go over that in a minute. And the center you have the detance platform and what is interesting with this platform it's flexible. You can use our own special SQL product or you can also leverage specialized tool like cart toolbox, jopendas, xarray for any type of use cases and we can scale easily.

4:55 Platform architecture

You know I have customer that use datab bricks for small data set and I have customer that deals with pentabyte information. So you are you know we are you are priced accordingly. You can still use a platform for small CSV and small use cases. No worries about that. And regarding on the right, this is where you have your consolidated enterprise data where you can visualize, consume the data. Your data analyst can use SQL, can write SQL queries to interact with the data.
Your non-technical user, your business users can use Genie. This is our product to interact with native uh language, natural language. We support multiple languages uh English, Spanish, Latin. Uh yeah, maybe not Latin, but you know, you get the picture. using uh English you can uh interact with the data and the good thing about Genie you can even like display map in this product so you're not called user or or even me I'm using it you know I'm lazy sometime I just want to have an information and I just hit like I type it like in a simple English and GIS analyst as well can use all these feature but also they can use their own tools we support cartto rgis uh Postgis something we're going to go over later on.
BI tools as well. So we have integration as well with the external world. This is an overview. Now let's dig uh deeper. Let's ingest the data. This is our reference architecture. Is there anybody in the room familiar with Medion architecture? Does ring a bell? Not that many people. Okay. So in a nutshell you have different zones ra refine and rich sometime it's named differently uh bronze silver gold but it works the same way you refine the data layer you have your native files there gojson shape file you bring them in the platform we recommend to use delta format because you get all the benefits uh predictive optimization transaction statistics we have built-in readers to read uh vector data like JSON, XML, CSV and if you have any gap if you have if you're working with customized data set like really specialized format like ZAR LAR we have you can leverage a specialized library for that like raster IO XR so we are a flexible platform and then at each layer you validate you refine your data you apply your indexes at the end you have the final product the goal the single source of truth And at the goator this is where you can consume it using a external tool BI tools or using various interface uh special uh sorry SQL APIs Python APIs and also to bring data in you can leverage like I mentioned specialized tool like safe or you can also use we have our own ETL pipeline called legflow legflow pipelines it used to be named DT data life tables but it was rebranded for liquflow pipelines because it sounds way way cooler and the way it works it's uh declarative pipelines using uh python or or SQL and then datab bricks going to manage the full pipeline for you the deployment and also the um the maintenance and what is interesting with like full pipelines you can see like the different progress of your pipelines you see as well the matrix you can define data quality on top of your data.

6:56 Data ingestion and pipelines

So you can see like which record are dropped or are clean. If there's any task failure, you can resume this specific task. And also what is coming during this summit, but I'm going to analyze it to you as well. We're going to have a no code local solution called lakeflow designer. But it works the same way. It going to produce the same kind of uh output. And to architect orchestrate all your job, you can use workflows.
So workflows it's our orchestrator and it can organize and schedule all your jobs. It could be uh notebooks, pipelines or shell script. So this is to bring data into the platform. We recommend to use volumes to manage your non-tablar data. So if you have any uh shape files, tiff, lighter data, you can use volumes. This is a special kind of object unity catalog. You have you know you have your tractional databases table and volumes is a special kind of object to deal with nonabular data.

9:52 ArcGIS integration

What is interesting with uh unique catalog you get the lineage. So as soon as that data lands into the platform you can track it all the transformation logic that was applied on top of that to produce the final output you know the final goal layer either machine learning model uh your model or your your table everything is searchable credible indexable as well using UC let's focus on uh how to integrate with ARJS you know many many customer use ARJS Yes.
Is there any S3 customer here in the room? Yeah, you know. Yeah. Okay. Almost lost half of the crowd. Okay. Let's uh focus on the first option. There's S3 Gantics engine JE and it's pretty simple to use. It's a library you can install on your cluster directly and it support the latest and greatest version of DBR. What is interesting with this library uh out of the box you can use spark to uh query your your data that sits in ARJS pro online enterprise.
So natively they support integration with your RJS instances and they bring a lot of different option to do uh buffering clustering special join they even have like special analytic you know to find a hot spot similar location and so if you are dealing with large data set ma massive uh data set you can use that second option is to use a python RJS module again this is pretty easy to install if you're familiar with Python world you pip install in your notebook or even on a cluster.
Bear in mind this module is by default like for Python library it runs on a single server single node and but what is interesting with this module you can access your feature layer and your web map directly from RJS online in enterprise you can display the maps in the notebook using display HTML you can perform geocoding enrichment routing and also if you need to uh distribute the workload there's always an option to convert the data frame, the Python data frame from this library to Spark and then run your your code on on clusters.
This is another option. And also the last option is a Spark data source. So since DBR version 15, you can create your own data source using Pispark using Python. So this is something I did. I have wrote a custom data source to retrieve data directly from RGis REST APIs. And what is interesting with that you can develop APIs to do streaming to even write back to RJS and uh what is interesting with this solution you you you don't have to install anything it use spark data frame under the hood and you can combine it with the special SQL datab bricks solution to process larger data set efficiently.
So these are three option. If you're not familiar with arc gestures API, this is something that is available right away if you go on the feature server. This is the UI if you never seen it. And what is interesting with that you can query your instances live and there's a bunch of option you can define aware clouds you can define the output field. You can even redefine the uh coordinate system the output. So this is what I use under the the hood to retrieve the data in the lakehouse.
And also another option I wanted to share it with you. This is a work in progress if you want to uh also uh connect directly to the lakehouse from your RJS UI. S3 is currently working on a native uh CDW connector. Unfortunately I don't have any date for you but still I wanted to share the information. So uh stay tuned on that. they were waiting for the u datab bricks special SQL public preview which is coming pretty soon.

14:20 Analysis and H3 indexing

So we've ingested the data. The next step now is to analyze and transform it. And to do do that we have built-in job special function at this date. We have uh H APIs we have special SQL function as well. And the interesting thing about those function they are available everywhere in your injection pipeline electrol pipelines in notebooks in SQL and also you can um use them with serverless or with the classic compute as well and if you find any gap in the platform you know like I mentioned earlier it's a flexible platform we don't force you to use these special SQL function you can also install geopandas uh sedona uh raster and the good thing is most of the time they works well together.
There's no clash with the namespace. So you can combine them. No worries about that. If you're less familiar with H3 indexing, who ever heard of that H3? Okay. Oh, almost again after the crowd. I like that. So it's a grid index. Uh it's similar to S2 by Google or even geohash if you ever heard of that. And the way it works, it's a bit different than traditional indexes you find in databases which use a hierarchy. And the way it works, it breaks down the world in multiple tiles, hexagonal sty tiles like stop sign.
And you can define different resolution depending on your use case. One up to 12. One is is the coarser. It overlaps countries and continent like here and 12 it's really up to millimeter. So uh in inches it's I don't know a couple of inches. Um and then here in you see like for Manhattan we see the cell ID for the island and you can use that to perform your your special join. We're going to go over that in the next slide.
So this is our reference architecture to use these indexes. You start from the uh razone. You apply the medion architecture. You refine your data. You ingest. You have your polygons. You have your points. You uh they are all in delta lake format. You refine it. You clean it. You validate it. And then in your silver layer, this is where you can apply these H3 indexes. You can also specify multiple resolution depending on your use cases.
It really depends on if you need to be accurate or not. And then the join happen on those cell ID which are unique under the hood. They are longs. So that's why the the join is really efficient and fast because you're joining on too long. And also we have a a special kind of pattern. If you need to be super accurate, you can combine the join on industry index with geospatial function with geop geometry function like intersect or contains.
So this is a way to do it and then get the final output and also what is interesting with the platform you can use LM directly in your SQL so there's something you can use as well for your geospatial needs and the way it works you know you don't have to be a machine learning expert a big brain you just use AI query you specify the model you want to use either in-house model or a public one like cloud or openAI you define a prompt you can give variables and it will give you result right away in a plain SQL in a plain SQL statement and you can use that for to do add a query to do enrichment and what we have built in in the platform we have bunch of different function to do classification sentiment analyzers um masking so this is pretty interesting as well so we ingested the data we analyze it Now it's time to serve it to visualize it.
Again we have a bunch of option for that you can use map visuals. This is directly available in the lakehouse in datab bricks and at the moment we support markers. So latlongs points you can display that on on maps. We support corrop maps and also we are currently working to bring more complex types you know like polygons geometry geographies and also h3 cell ids and what is interesting with that it's available in SQL editor in genie so you know the feature to uh interact with the natural language and also in dashboards so you can create dashboards with maps and you can share that with your end user with your stakeholders And the interesting about dashboard your end user don't have to have an account in datab bricks to use it.

18:16 Serving and visualization

So you can embed these reports in your internet your externet. So this is an option. The other option is to use databicks apps. So if you have more complex use cases you need to create like interactive apps. This is an option. If you're not familiar with that, it's a self-contained web app that are hosted and managed by databicks platform. They are pretty easy to use. It's a straightforward process. There's a nice UI to create them.
For now, we support most Python framework such as dash, radio, stream, and shiny. So, you don't have to write any like JavaScript code or HTML. You can define everything in Python. And there's a bunch of option. And what is interesting with this solution is you can create interactive apps predictive analytics geni and obviously I mean it's a geospatial talk you can write and create data bricks geospatial apps and it's up to you like the uh which mapping widget you want to use you can use open street map Google API and what is interesting with apps it integrates with your data directly so there's no it's not hardcoded like the data here like you can send query to your warehouse and retrieve your data in real time.
Same thing for ji you can query your machine learning model in real time. So there's a it's not a hardcoded graph and if you are interested in that you can uh see there's a link there databics apps I hope I can share the content with you afterward and another option as well we integrate with BI tools I'm going to focus on PowerBI today but you know we support as well uh um Tableau and also uh RJS and also I have customer that use QGIS as well okay But for today, I'm going to focus on PowerBI.

20:53 Demonstration and conclusion

And the way it works, the PowerBI service can retrieve the data using the warehouse and then you can import it or use direct query if you're familiar with those concept from PowerBI. And it's again it's a pretty easy and straightforward process. Nowadays we release like I think it was this year we released a one-click button that is similar to uh buy now on Amazon and it does everything for you in the catalog in in UC you can specify which tables you want to publish to PowerBI and automatically it will create the semantic model for you and you can define which uh storage mode you you want to use but if you need like all the feature the you know the full-fledged experience you can still use the PowerBI desktop connector.
It's still supported. So if you need to uh have more control and when you are in uh PowerBI there's a couple of option to display uh geospatial data on maps they support correl using fill map and what is interesting Microsoft support different addresses in the world uh to define boundaries on the counties states um neighborhood and you have your go-to map this only support latl long points but you can uh do a cool thing can redefine tool tips and the size the colors and also what they support is RJS map for PowerBI what is interesting with this option there's a free version available but obviously if you are a street customer you can connect to your data directly from uh PowerBI and then you can have multiple layers you can uh superpose overlap your layers coming from RJS with the data coming from data bricks.
So this is a great option. All right, we are right on time. I like that. I'm going to try to do a live demo because to put it all together. That's great. I'm pretty sure it will work fine. All right, so we have uh our catalog. I have my my tables, my database for the summit. And uh the first thing I'm going to show is PowerBI. So it's pretty straightforward. We have this button here, publish to PowerBI. And then you can specify like the uh workspace you want to work with and you can define the query mode.
You can also select the tables. I already done that before the the meeting. So uh you can trust me it works. And what it does if you go in PowerBI service you will see your semantic model that was created automatically. So you can see your different relationship that match what you have in data bricks. So here I have my fact table, my UFO sighting because if you remember we are trying to find the uh reasoning before for sightings and we have my the dimension as well counties, cities and airports.
So this is for PowerBI and if I go back in my um my my repo, this is the repo I've shared with you uh previously. I have my source code. I have my notebooks. I have my data. The first step is we need to ingest data in the platform. Like we saw in the first section, the UFO sighting comes in a CSV format. It's geio reference. There's latl long. I define the schema because there's no header in the the for the file. I convert the latl longlong to a geometry because it's easier to work with.
And then we can explore the data right away using SQL. So most of the sighting the shape are light and triangle. And let's check like the moment of the day these sighting were reported using like uh this query and then we can see like most of these sighting happened during nighttime. So I don't have any explanation. Maybe like aliens sleep during the day. I don't know. I don't have a reasoning for that. But it's pretty interesting.
So I ingested the sighting there in the notebooks. There's different technologies as well to display um maps. I use volume. This is an option. And here I can display my latl longl long on on a map. You can uh show tool tips. You can redefine the marker. So this is great for analyzers. So I have my sighting. Now I want to bring data from RGis. So I use my custom data source that I created. It's pretty straightforward.
You see read format ARJS. You just need to define a URL and also you can override the wear clouds if you want to filter data. Same thing for output field. But for now I want to bring everything. So I've ingested couple of data set cities you know I want to check maybe there's explanation there if most sighting happen during in cities because because maybe it's because of lighting or uh skyscraper there's more object in cities so this is coming from my cities airports as well I want to check if uh maybe it's if it's close to airports that could explain sightings and counties counties because I want to maybe detect cluster is there any like can try to find patterns in the data.
If all these sighting happen in a specific region in in the states and this is the result you get when you use this uh API you get a unique ID for each record a feature type geometry as a joson and the properties as a map and then you can use that to persist your data set into the layouts in data format and also you can display polygons using volume like that. So we've ingested our data. Now we want to transform it, analyze, find patterns in the data using a spatial SQL.
Again, it's something we show earlier. So the first thing I do, I I perform a search within a radius to just give them an idea. Okay, there's a latl long a point. Now I want to retrieve like the sighting near that something interesting to explore. And then the next step is to do the join. So most of my data set they use latl longl longs. So I have latl longlongs in sighting in airports and cities. So it a bit it's a bit tricky to join them together because it's we can go naively and try to join all airports with all sighting but it won't be performant.
It will won't be ine efficient because I know for a fact my airports in New York uh won't match like my sighting in Texas. So it doesn't make sense to do that. So the way we're going to do it, we're going to do we're going to use Hindexing to perform this calculation. And I have like um an image maybe it will be easier to understand because it can be a bit tricky to understand the first time if you never done that. I convert my light lung to H3 indexes.
If you remember these are unique in the world. They are cell they are long and they are like in my my sighting are in yellow or light orange my airport is in um darker orange so as you can see here even if it intersect it's not a match even if these point are are near it they won't match like the cell ID so what I do using each tree we have function as well to retrieve the parent of these cell So in blue I have the parent of my UFO sighting and my airport.
And now we have a match. And then I'm going to show you the actual code. We have a match. We match on the parent. And then we can use a geometry ge function to be more accurate because I want to keep a a match within a distance of 10 kilometer. So there's something you can combine as well. You can combine H3 which are more coarser but are faster way to join and then you can combine it afterward with the intersect contains or in my case distance because I want to uh find airport that are close by and then we get the result we get if there's any match or not of an airport sometime we get even multiple airports so I do the same trick for cities because it's latl long as well now regarding ing counties.
It's a bit different because counties it's really like a a polygon with a boundary. So I'm going to use a different trick to do that. And also I wanted to show you that I've used pi spark to do like uh the first section of my notebook but also you can intertwine it with SQL. You know there's both option depending of which language you prefer. So I use SQL here. So the way it works is we're going to break down the polygon in multiple smaller tiles using this function test plate.
And what it does, I'm going to show it on a map. I'm not using folium anymore. I'm using Kepler. This is a really interesting library you can use in your notebook. And what is interesting with that you can it's interactive. It support bunch of format uh WKB, JSON, H3 and here we can see all the different counties in the states and you can see if you zoom in you can see all the different cell ids all the tiles and then we get the index the the cell ID.
So I've broken down my neighborhood my counties in multiple tiles. Now you're ready to join with the UFO sighting. And also what is interesting with that pattern is uh when you break down the original polygon, you can also like recover the original polygon. So you also have this information. So here I display the original polygon for uh Texas. So this is really interesting. You can keep you can have both of them. You can have the the polygon plus the H3 APIs at the same time depending on on your need.
And now I'm ready to join on the sighting and persist the information in a table. So I get all my matches siting matching to a county to a airport and to a city and I have my final table. And now we are ready to have fun to perform some computation. Let's try to find some trends. So here I found that most of my sighting happen in a urban um urban location close to 70 person. I don't have a reason but you know let's uh continue.
And the other one airport close by I get a 86% match. You know this is a crazy number. So maybe that could explain UFO sighting. You know they are close to our airports because maybe of lights. So this something I could share with official or government. Maybe I found finally I found an explanation for UFO sighting. Uh and also maybe we can try to find trend with uh population. So I've used I I group by county and sighting and then I get a trend and you can also display on on a scatter plot.
So as we can see below one mill 1 million of population the there's no like correlation but as soon as you hit this limit like 1 million there's more reports but again it makes sense maybe you know there's more sighting there's more people to see uh the the sighting and also what I'm trying to do as well in next section I'm going to try to find cluster aggregates and to do that you can use and hole it's a way to unclose all the points all the light long longs and it will give you like a convex polygon that unclose all these points.
So this is interesting operation if you want to detect cluster or aggregates and you can uh display that on a map. So here you can see the different clusters and you see this this pattern you know like in major cities in Vegas in in uh on the west coast here you have the mountains and you have uh less people in the middle but there seems seems to be a correlation between like large cities and number of sighting but again you know maybe there's a correlation with the there's more airports there there's more lights that could explain sighting and finally let's try with LLM.
So I'm going to use claude. I'm going to use a prompt. Can you explain what would be the best explanation for this UFO sighting with the description, the location, the detime and it come up with a possible explanation. So this is an other option as well. And what is interesting with the LLM, I mean it for this sighting in 98, it knows it couldn't be drones because there were less drones at that time. But still, you know, you have possible explanation as well.
So this is interesting as well for geospatial data. This is something we we saw in the presentation and like I mentioned in the presentation you can use these function to enrich the data or perform adock query. Okay. So the next step is to publish a result. I show you how to publish to PowerBI previously. Now let's focus on the viz. So couple of option here. You can display coret map. So here I display a chlorop with the number of sighting and again you can see like trend with the in Vegas in large city you can display like the number of sighting in a in a tool tip.
You can also display the raw data using a traditional map and you can use slicer to filter the data. I use direct query. So each time I modify a value, it will refresh the data from the warehouse. So let's pick another city. You can see, I don't know if you've seen it, there was a loader in the left corner. So this is another option. And the last option that we saw as well in the presentation is to use ARJS. So you can combine data from the lakehouse with the data from RJS.
Here what I did I did a layer USC airports just to give you an idea and I displayed the raw value as well. And again you know like we saw in the analyzers there seems to be a a trends with the airport close by. I don't know maybe this this is an explanation. So this is for PowerBI and another option as well for more complex use cases is to use a database apps. It's pretty easy to use. You go here and compute apps. You create an app.
You can define the the framework to use and then you get an app that is hosted in datab bricks. I've used Dash and Kepler to create this app. Um, Okay, it works. Okay, so I I use Kepler and again nothing is hardcoded. The data is refreshed live and using Kepler you can zoom in zoom out. I displayed a coruplet map. So again counties and you the you get a color scheme based on the number of sighting. So red more sighting uh light yellow less sighting and like I mentioned with Kepler you can define a bunch of things.
You can override the theme the colors and also you can specify a county if you want to look at a specific region. And again this is interactive. So here I display different information. I display the boundary of county and also I display the indiv individual sighting and with the tool tip you have the different information the description the date time airport close by and again everything is interactive and I've added as well a call to a club to give me a possible explanation for these sighting in these location.
So, as we said today, it's pretty easy to ingest, transform, and serve your special data. I hope you enjoy your presentation. Yeah.

Learn more about the Databricks Data and AI platform.