You're listening to episode 767 of a very spatial podcast.

August 17th, 2025.

Hello and welcome to a very spatial podcast.

I'm Jesse.

I'm Sue.

I am Barb and this is Frank.

And this week we're going to talk to the folks over at World Pop about data sets that are coming out in the near future, and of course the work they've been doing for, uh, a while now already.

Be sure to stick around after the interview for our regular news web corner and events.

Happy to be joined today by Professor Andy Tatum, who is, , director of World Pop and Heather Chamberlain, senior Enterprise, fellow of World Pop, , which world Pop is at the University of Southampton.

Thank you for joining us today.

Thank you.

Great to be here.

Let's begin with what is World Pop?

Well, , world Pop is a, an applied research group.

Here at University of Southampton and we are focused on using geospatial data to fill gaps in, in small area population data.

So working out methods co-developing them with, with data users from governments and UN agencies, and yeah, trying to improve.

Our ability to estimate and map populations.

Now, whenever we look at, at population, of course, we have a variety of sources that, uh, we'll talk about.

I'm sure as we go through the conversation that are everything from global, uh, un, which of course you're helping with some of that as well.

Uh, all the way down to subsections of of countries.

Let's talk a little bit about your methodologies of bringing all these disparate sources, , together.

I guess our methods already cover sort of.

Two main, uh, broad scenarios.

So we focus, , on a, a sort of subset of methods that we call top-down methods, where we're really looking at national or subnational, generally census-based data or projections and disaggregating those down to to smaller areas.

So our focus really is on providing, , as granular and, , a granular small spatial.

, Areas of estimates as possible.

And so when we do, when we work on methods that are based on census data, that's disaggregated, that works if we have good census data, but we know that in a lot of countries that data may not be available or it may be highly outdated.

It may be many decades since the census was conducted, and so that's when we also work on methods that we call bottom up methods, which are more focused on looking at alternative sources of population data and working out if we can extrapolate or model those in some way to try and fill in data gaps to go ahead and continue with that.

Can you give some examples of some of the.

Bottom ups, uh, sources that you've been using.

Sure.

So, , some of the really early work we did in this context was with, , Afghanistan, , where the last census was conducted.

The last national census was conducted in 1979.

So there's lots of uncertainty there in terms of population numbers and obviously a lot of population movement and conflict, , have occurred since that, since that time.

So.

That was one of the early instances in which some of these methods were developed, and that was making use of, , survey data for specific provinces and then using modeling to try and fill in some of the gaps.

Now as we see this combination of the bottom up and the top down examples that you're, or data sets that you're utilizing, how do you deal with some of the border issues as well?

So you have the ability to look at a country and, and I think everybody can understand how we.

Can conflate different parts of a country through various surveys and trying to understand that.

, But how do you deal with some of the border issues whenever you have somewhere like, you know, Pakistan and Afghanistan where you do have more recent, actual national census in one country versus dealing with the bottom up survey based on the other side of the border?

Okay.

Yeah, it's a great, great question.

And it very much depends, I think, on on who, who and how these types of data are going to be used.

So we're, we are generally producing.

Two types of population data.

One where we're trying to produce a consistent set of estimates across the entire world for multiple time periods and trying to do the best we can.

And there we are making use of projections and estimates for one country, and then next door we may be using a census on.

What we're trying to do is turn all of those into a.

Grided set of estimates, so numbers of people per one kilometer or 100 by a hundred BTA grid square.

But in other situations, we may be working directly with a government who are interested in producing more recent estimates for a very specific time period and a very specific purpose.

And in some of those cases, it can be around.

Border issues where populations have changed substantially because of conflict, because of, uh, natural disaster.

And that means populations have changed a lot since the last census.

And so they're interested in using things like satellite imagery and mapping buildings to get a better idea of how many people that are likely to be and where they're likely to be.

That brings up whole other questions for areas that are depopulating and, and number of structures, but that's a separate set of conversations.

I think whenever we look at.

This, and you did mention that you're looking at one kilometer pixel size.

What's the scale that you kind of consider this to be at?

'cause of course, we have a resolution, but that doesn't necessarily, , infer the scale that you consider this to be appropriate at.

Do you wanna start with Yeah, I mean, I think so The, the finest resolution that we pro we produce estimates at is three arc seconds, about a hundred meters at the equator, and we tend to recommend that.

You shouldn't use a single grid cell to estimate just the population in that grid cell and consider it as an actual representative boundary of that being an exact number in that grid cell.

But as you aggregate these grid cells up to larger areas, , say to one kilometer or, or larger areas, then we're gonna have more confidence in those numbers being reflective of the, of the actual spatial distribution of population.

I don't think we have a hard and fast rule of exactly the minimum size of a unit, but yeah, the, the advantage of this gradient data sets is.

Sure.

You know, the flexibility to be able to summarize it by different administrative units or different decision making units or to delineate the outside outskirts of a city and work out an estimate of how many people are in that city, or to integrate it with other types of data like the location of a health facility, and try and work out how many people are living within five kilometers of that health facility.

Very difficult to do if you are using administrative count.

It's within.

Boundaries that are all the different shapes and sizes, but the gridded data offers that, , opportunity to, to summarize and aggregate.

And it's at those kind of aggregate levels that we we're more confident in those estimates and that that tends to be how our data are typically used, rather than someone going to one grid cell and saying, there's nine people here.

Definitely, , which we don't really recommend.

It is.

It is one of the.

Positives and negatives of RAs is that they provide a sense of fuzziness as long as you understand fuzziness, but a lot of the population sees it as like a, an image pixel where this.

Color means it must be this in this area.

Whereas a lot of our, our raster data is whether we're talking about soils or population, it's a representation of what we think it is.

So yeah, there's this whole other conversation about fuzziness that exists with, with these type of data sets.

But again, they're important to have, and I think the fact that we have been moving away from some of our traditional vector hard boundaries.

Two Raster has helped to hopefully confer a little bit more of the, the idea of continuous boundaries.

Compared to what we did back in the, the nineties and early two thousands.

Yeah.

And it's, I guess a big focus of our statistical modeling teams is to try and not just produce a single estimate, um, because we know there's a lot of uncertainties in.

Population data, but to, uh, they're developing a lot of different types of Bayesian models so that we are producing for each grid cell a full posterior distribution of population estimates so that those can be summarized in different ways to give you, uh, not only a.

Best estimate, but a confidence interval to say, we're, we're pretty sure there's not more than this many people in this, this area that you're looking at or, or should not be less than this many people.

But, but when you're looking at an individual grid cell, those confidence intervals can be huge.

But when you start to aggregate up, you yeah, you're, you're getting actually perhaps useful.

Information in terms of an estimate and its confidence interval, but there's always also the challenges with trying to communicate that uncertainty Yes.

Associated with that data.

Yeah, definitely.

Whenever we say population, we kind of assume we're just talking about.

Density perhaps, or sheer numbers, but there's a lot of different data that you do actually provide through World Pop.

Can you go through a few of the different aspects besides just general population that you're providing?

Yeah, so the, the our, our main focus is, I guess primarily population counts.

But also broken down by age and sex classes.

That's the kind of second thing that people tend to want to know in terms of who, who is there in terms of children, women of childbearing, age, the elderly.

But we also do work, , making use of household survey data to try and estimate.

Things like, , rates of poverty, vaccination coverage, literacy, access to sanitation, travel times to healthcare.

So again, it very much depends on who is the end user and the kind of information they want to know and the kind of decisions they want to make.

And yeah, there were a variety of projects going on at at World Pop in terms of producing these.

These types of outputs with an ultimate aim of trying to build up, I guess, some kind of demographic atlas so that we are trying, trying to understand who those populations are for, making sure that nobody's is left behind and that those, uh, decisions can be tailored to those populations most at need.

You also mentioned that you're looking at different times, so in also included in this is things like urban change, global settlement growth.

So can you talk a little bit about how you're dealing with this as a temporal issue as well as a spatial issue?

Yeah, so, , I mean this is something where the data that's available on settlements and down to the level of individual buildings has really rapidly progressed in the last five plus five to 10 years.

, And so we're constantly sort of trying to adapt and integrate that.

Data into our workflows.

And increasingly, , as you said, it, data that is, , temporarily explicit that's allowing us to see these changes in settlements, , allow us to integrate this into, , where we're, .

Distributing population across the surface, but also look at potentially projecting this into the future as well.

So some of the, the, the new, , data that we're having that, that we've, , recently developed that will be, , published and made openly available launched, , soon is integrating that data at an an annual time step for every year from 2015 up to 2030.

, And so that is integrating the settlement growth for.

, Every country globally.

, And then using, , estimates of population, , down to the individual grid cell level as well.

And of course, people wanna be able to utilize this in their existing workflows, um, and technologies.

So can you talk a little bit about the integrations that you have?

Everything from, of course, the API all the way through to some of the, uh, extensions that you have for ArcGIS, uh, QGIS, those type of things.

A very large group of things that I've thrown out for you to talk about now.

Uh, yeah, yeah.

This is, this is where we should bring in our more spatial data tech people.

But, , but yeah, we, we trying to make.

The, the types of data we produce as accessible as possible to all kinds of different users out there.

, And because it's population data, it underpins so many, so many bits of work that academics, commercial companies, governments, UN agencies are doing.

So we have the API, we have, uh, a QGIS plugin, as you mentioned, that enables users to.

To automatically download and access those data and bring them into QGIS.

Our data are available within Esri's Living Atlas.

They are part of the humanitarian data exchange that un Cher lead a part of the un NFPAs population data portal.

Uh, and they're also within tools like health.

Information systems that about half the world's government uses.

So that DHIS two system, they're, they're part of those to, to enable access to small area population estimates and integrated into to Google Earth engine as well.

Yes.

Yep.

So hosted on that system and accessible to that, uh, data community as well.

I think the list is probably quite long.

That's a small flavor.

Yeah.

Whenever we look at this, you also mentioned, uh, some of the UN organizations that.

You're partnering with as well as other countries.

Can you talk a little bit more about how, uh, especially at the UN level, different portions of the UN are, are utilizing these data sets?

Yeah, so it very much depends on the agency, of course, but for instance, whenever.

There is any kind of earthquake, flood, other, other type of natural disaster.

Then, , UNOCHA unsat, , make use of our data sets to overlay them with, with things like the extent of a flood, the track of a hurricane.

To estimate numbers of people likely to be exposed to those.

So it enables that kind of rapid assessment and more precise assessment than if they were using census data and match the boundaries.

We work with UNICEF who are interested obviously in the, the health and wellbeing of children.

So in that case it's things like.

Estimating numbers of children who are unvaccinated, , so vaccination coverage rates, but also where those children are, , to be able to reach them.

We work with U-N-F-P-A who have a mandate to support countries produce, , a robust and rigorous census, , and.

To produce population data itself.

And so we've worked with them for a long time to co-develop methods to, to fill gaps where countries cannot do a census or they can do one but cannot reach everywhere in the country.

And we work with World Health Organization for, for things like.

Access to, to health services and again, for, for vaccination campaigns, and there may be others as well.

I think the, the Food and Agriculture organization, I think we use our data in their, in their portals as well for looking at vulnerability, for looking at, um, uh, access to, to.

Agricultural, uh, produce.

And of course it's not just, uh, un that's, that's working on projects.

If you go to your website, of course, world pop.org, uh, there's a whole list of current past, uh, projects as well as other materials including course access to the data.

So whenever people are going to the site to look at the data, um, what are some of the things they should keep in mind in terms of.

Questions they should ask themselves before they start looking at the data?

Or are there tools to help them find data on your site?

There are, I mean, there are basic tools to help you find, find the data.

There's a, there's a page we put up to, to ask people, basically to get people thinking about why they are.

Why they want the data, , what are their applications?

Because as you say, we produce population data sets for different purposes, , for different time periods, , and for different spatial extents.

, And so there's, yeah, there's a, there's a need for people to think about.

Why, , what, why are they needing population data and what, what they want to use it for?

Is it for a single recent time period for a single country, , or is it to look at changes over time across multiple countries and that as that.

Those kind of things are, are things to think about, but also we try as much as possible to put documentation on there, on the website so that people can understand what goes in, what the types of modeling that's done involve and what comes out.

, Because that's pretty important to understand.

Uh, in, in some countries there hasn't been a census for 20 or 30 years, and therefore we are producing estimates that are.

Uncertain.

, There may be, there may be better than anything else that's out there at the moment, but they still can be incredibly uncertain.

And so people should keep in mind what's, what's going in.

And it varies from country to country.

So while we make available global estimates of each person in each grid square across the entire world.

Those, uh, outputs have different levels of uncertainty because of the different types of inputs that are going in in some countries.

It involves just taking a very recent, very well, uh, undertaken census and.

Desegregating it to grid squares.

In other cases, it involves a lot of estimation when there hasn't been a census since the 1980s or 1970s.

, And though they may look the same on the grid, they of course are very different in terms of their, their levels of uncertainty.

I think that, I, I hope, but I know is not the case that people think of with any kind of census, is that the results of this census is still a statistical product.

Uh, we, we can't literally count every person.

Um, so we take what we do get so we have more confidence in those that are better prepared and more recent, but it's still a statistical product.

So while, yeah, I I it is different.

I think it's the confidence as much as anything else, as I'm sure the, the statistics in the dataset would show.

Um, and of course that's the choosing the right world pop population data for you page that's linked right off your front page, that, that helps get people, um, to that point of understanding the data.

Is there anything else you would like to highlight about World Pop and especially the product that's coming out soon?

I mean, we can Yeah.

Highlight the, the new global data and the, the old data that we're calling Global One, uh, was a, a, a data set that covered.

The year each year from 2020 20.

It was, we finished producing it in 2018, and, and it has since found its way into hundreds of different uses, uh, and applications, , all across the world.

But it's, , it's become quite outdated.

It's using.

The, the 2010 round of national censuses as its basis, it's using geospatial data sets that were around in the, in 20 17, 20 18, which as we know are now massively advanced in terms of our ability to map even individual buildings across continents, whereas we didn't have that ability before.

So the new set of data covers 2015 to 2030.

It's funded by the Gates Foundation.

It.

Brings in both the 2010 round of censuses and the 2020 round of censuses, , new types of demographic models to fill, fill those gaps.

, And estimate those populations, new sets of covariates to be able to model and predict those populations in that spatially.

We're bringing in those building maps that come from satellites that, that are produced by, by groups like Google and Microsoft.

Uh, and then yes, a, a, a new.

Master grid, new sets of coastlines, new water bodies.

So everything has, has been, I guess, given a, an upgrade in terms of, , those, those output data sets.

Anything else you'd like to highlight about World Pop in general?

On the one side, there's this new data that's covered that are globally and uh, and I kind of.

Co-developed with other data providers and research institutes, but the other half of our work is very much co-developing individual country data sets with national statistics offices, ministries of health in lower middle income countries.

And perhaps, yeah, Heather has good example of working with Zambia, for instance, on, I guess there's quite a lot of different examples we could, we could talk about.

Some of this stems back to this early work that we, we talked about earlier in, in Afghanistan when we were working with the statistics agency there back in 20 18, 20 17.

And since then we've worked with a lot of different statistics agencies around the world.

And this might be in context where they haven't had a census for a long time, or they're preparing for a census, or they've conducted a census, but then have issues with gaps in coverage in particular geographic regions.

So in each of those cases, it's a.

It's a case of understanding what the data gaps are that exist, and then trying to develop methods that aim to address those gaps, fill those gaps in using the available data as best we can.

And there's very much the focus on these estimates always being co-developed.

It's not, we're going in there and using a, a specific, , a standard set of methods.

It's.

Working very closely with staff in national statistics agencies to understand the problem and then come up with a solution together with a big focus on capacity strengthening as well.

So we don't want to.

Just do this piece of work once and then leave.

It's a case of trying to improve the statistical understanding, the geospatial understanding around these methods and so that hopefully further down the line if these estimates need to be updated, then actually there's more capacity within statistics agencies to do at least some of this work themselves.

And I think one of the, one of the contexts in which this has been really.

Sort of quite well established, I think is in working with particularly U-N-F-P-A to bring together regional workshops where multiple countries have come together.

There's been training and then U-N-F-P-A have gone on to support those, um, national statistics offices with their country offices to try and enable these methods to be more widely applied and supported.

Yeah, and it's quite, uh, quite important this.

Co-development for actual use and impact of the data.

'cause these are kind of new types of methods.

This for, although all of us here are quite familiar with the world of geospatial and the abilities and, and what's possible with satellite imagery and what's possible with statistical and spatial methods nowadays in busy, low resourced statistics offices, it's, it's very much something new and, and population data is.

Uh, underlies so much of decision making.

It's, it's your GDP, it's your allocation of resources.

It's the voting, it's a representation in parliament so they can be incredibly sensitive and important data to get right and.

To, to switch from, uh, the methods that have been used for years and years of just sometimes just a straight line projection from the last census 10 or 20 years ago to something that is very alien to the, to, uh, the government.

It takes a lot of work to ensure that those, those statistics offices understand and, and are convinced in the methods enough that they can convince their president, they can convince the public that this, these types of methods are producing more reliable and accurate.

Population numbers that people can trust and that they have the ability and understanding to be able to replicate those methods and, and explain them to, to the public so that these types of data can be adopted.

And yeah, as a, as a result of some of these co-development, we've, we've seen these types of geospatial methods be.

Used and, uh, the outputs adopted by governments in places like Papua New Guinea in South Sudan, and we've seen them be used for vaccination campaigns, uh, in, in Nigeria, in Zambia, uh, in Afghanistan.

And so, yeah, if.

It takes, it takes some pain and it takes more time than if we were just sitting here at the university and producing those estimates ourselves.

But it really brings, I think, a lot more impact and, and yeah, really great to see those numbers actually changing.

Changing how resources are allocated and reaching populations who've sometimes never been counted before.

Can you, uh, mention very quickly as we're heading out, uh, some of the information about the upcoming coming launch of the product?

Uh, yes.

So we are launching, , these new global data, uh, through a webinar on the 4th of September will be.

, Giving details on that data where you can access it, how it was put together.

, We will also be having contributions from the Director General of Statistics, Sierra Leone.

We will have contributions from regional advisor from U-N-F-P-A, um, from our Gates Foundation funder.

Um, we will have a, a technical.

Question and answer session, and it's free to sign up to.

There's, uh, an event right link that you can access that through our website and we'll be sure to include a link to it in our show notes as well.

Well, it is important work and it is good.

Uh, in this current time when data is sometimes disappearing to see new data sets, uh, making its way onto the, the world stage.

Um, both to support global initiatives as well as local initiatives in, in various countries.

So thank you for, uh, doing this work.

And of course, thank you for joining us today.

Great.

Thank you.

Thank you.

Thanks for the time.

Now, kicking off with, uh, one that means absolutely very little to most people, but uh, at the recent.

What conference was that?

Psychograph Conference, which still exists I had forgotten about.

Kronos announced, which is an open, uh, source group, highlighted that they were working with the OGC and Niantic, spatial Anes and ri, I guess, to put out Gian Splats in WebGL transmission format or GLTF for short.

Um, and this isn't.

A huge thing for most people.

But in the space where Gian Splats are replacing a lot of the 3D models that we were creating with point clouds, it's relevant.

So if you are someone who is doing that, basically creating 3D data and wanna look how gian splats are working, and you're probably already using applications that are generating gian splats, check that out.

And if you're someone who just wants to get things, uh, in 3D onto the web.

This is another group that should be looking at that, but we're not gonna go into details about what it's though the too long didn't read is if you're doing 3D stuff, there's a new way to do it.

Uh, kind of new in the open format.

Yeah.

Something that you should comment on is that the draft part 1 0 8 rules for the FAA, uh, this is the beyond visual line of sight.

Rules.

Uh, these are ones that, uh, were already mandated to be created in the near future by the last administration in Congress.

And then a couple of months ago, the current administration said You have to have this done right now.

'cause of course, it's a moneymaking thing, so they're gonna be pushing it really hard.

Um, and, uh, there are some things in there that are, you know, less rule than I would like to see for beyond visual line of sight.

So go through, take a look at it, see what you think of it.

Yes.

If you are already flying, UAVs, drones, whatnot, or are thinking about it.

And a term you may hear a lot, which you may not necessarily put together as beyond visual line of sight is bev loss.

That's the word for it.

Which, because government has to have acronyms that sound like words.

Well, I mean, to be fair beyond visual line of sight is a long phrase.

It is one.

But um, again.

If you heard Bev loss, you go, what the heck is that?

That's what that is.

Nobody else any has anything to say about Bev loss?

I, I do.

I have a lot of things to say about Bev loss, and I agree with you.

It's not as rule based as I would like, and the secretary said, we need to do these things for these industries that will grow using drones.

And I understand the point of view, but I think beyond visual implies like.

At least in my head, not too far beyond, but the way they're talking about it is like way be beyond visual, you know?

Like don't worry about it, it out there doing a thing, it'll be fine.

Which I'm a little, I know we'll eventually get there.

I'm just not sure we're quite ready for that yet.

Uh, especially whenever we also could talk about how it seems that maybe DJI is already just saying whatever.

We're not gonna deal with the stuff that's going on and letting drones run out in the US and not releasing other products just because they don't wanna deal with it.

So, yeah, one of the largest manufacturers is, is kind of calling it.

Somewhat a pause for the us.

So that's a question mark.

Um, you have, I mean, we can go all the way back to, what was it, dark Angel back in the late nineties where you had, uh, military slash police drones being common cause uh, there, and those were always doing questionable things.

And that's from the 1990s before we even had a lot of these.

That's say cool, but underappreciated TV show for those of you who aren't familiar.

Yeah, I was like that, that is a pop culture minutia.

Callback.

Yeah.

But that's awesome though.

But it was also one of the earlier like, um, aspirations of the, the drone industry originally.

Um, at least in the industry I was in, and to go even further the callback, um, I never, I can never remember there was a movie and a TV show and one was called Blue Thunder and I can't remember which one is, is, and they couldn't call it Blue Thunder.

For the other thing, whatever.

But the helicopter itself was, oh, the helicopter thing.

It was just called Blue Thunder.

But for legal reasons, they couldn't call.

I think the TV show, they weren't allowed to call Blue Thunder.

Anyway, it's a whole thing, but it had the ability, why am I making this up with Airwolf, one of those two had the ability to, to put out drones that did this and bring it back into like the motherships, which is, if that is older, this, it's not much older.

No, it is about 10 years older.

But you're right, it's not much older.

I, I just have to say though, in their list of industries, this is going to impact, or they have, you know, very big ones that, that use this, and I know they have, you know, standards within the industry for them, like agriculture, aerial surveying, public safety, recreation.

But then you get into package delivery and other areas where I, I don't think they internally within the industry have as much, you know, outlined.

Best practices.

But what I wanted to tag onto there also was wedding photography, because that's what I see a whole lot of, you know, beyond lot, you know, Bev laws would be going on without knowing what the rest of the industry is saying to do.

Um, I will say that, you know, whenever we talk about deliveries, Amazon and lots of other companies, uh, everything to Wal, even Walgreens, I think.

Uh, has gotten in on it, have been testing these, so they are building at least internal best practices, if not industry best practices, uh, because they want and have wanted this to happen since what, 2017 or whatever it was.

Uh, whenever we first saw, uh, the first part, 1 0 7 come through.

Yeah.

So I, I think there's some of that.

Um, but yes, it's, I think.

It goes back to what Frank said in his mind, beyond visual line of sight is just, it's, it's just over the hill.

I can't see it for a second.

It's gonna come back anyway.

Versus this is really, I would use the word autonomous.

Yeah.

Autonom.

I mean it really is a better term.

Autonomous.

Yeah, that's what they're talking about.

There's just like, send this thing out and then it'll come back later or not, you know, in a few days, whatever.

It doesn't matter 'cause it's autonomous and I'm like, that's, that's a different thing.

And again, it's not necessarily a bad thing.

We have in agriculture, for example.

A lot of autonomous systems out there, but they're, you know, kind of rooted in the ground usually.

And, but there's a lot of that out there.

So this is something that is inevitable, but I think it's really fundamentally different than, I mean, you know, you can make the argument that I've sit here in West Virginia and if I launch a drone in, uh.

Mongolia.

That's beyond my visual line of sight.

Right.

That's, well, since it's in Mongolia, it doesn't fall under FAA anyway, so whatever.

Well, I know, but, but the point being is that, you know, it feels like there should be some limits there when I say that.

Yeah, I, I mean, to be fair, I almost never actually fly my drone.

I use it autonomously, but I use it autonomously within my line of sight.

So I set up the expectation of where I want it to go.

I say, you know, grab images in this area.

I want to use it for creating a 3D model, or just creating an ortho photo and launch.

And I just sit there, watch it, watch the screen, basically making sure that if something does happen, if it loses a propeller or something, I'm there to hopefully.

Maybe be able to guide it a little bit away from anything that it might hit.

But really if it, if something like that does happen, it's, you said anyway, like I said, we do a whole podcast on this, but yeah, then maybe we should, I don't know.

We never really have done a big, these, our threats on drones, but, uh, something a little bit less to talk about though.

If there are reason to talk about it, uh, we're seeing more natural language searches in the MAP apps.

So instead of having to just say, find X or directions to X, uh, you can say, you know, find the best sandwiches that are nearby.

Or, you know, how do I get from here to there instead of having to, just the fact that we are seeing more use of LLMs.

In how we talk to the software and how they give us information back.

Um, and I highlight this because, uh, one, it's already in Google Maps to some extent, but uh, it's definitely a feature in iOS 26.

So I've been testing that and it, it's there.

It works.

Is it any good?

And, and the reason I ask this is because I'm perennially frustrated.

So my car has CarPlay, which means that you, or Apple, I can't remember the terms, uh, yeah.

CarPlay, uh, which means that I, and I have like a Hud.

But I can only get the symbols in the hood if I use apple maps.

So basically I use Apple maps everywhere I go because that's just the least pain in the neck to do.

But when I use the the voice thing and I say, plot me of course to blah, blah, blah, I find the system to be way worse than I expect it to be, considering it's supposed to be location based.

So for example, um, there's a place that I like to go to.

It's a guitar store and it's called Empire Music.

Now, that's probably not a unique name in North America.

I know it's not a unique name in North America, but when I say plot me a course to Empire Music, why is it showing me places in Montana?

No, I'm not driving to Montana.

I'm in, I'm in a location that, why is it not smart enough to go, well, let's start with the closest one and then we'll work our way outward, you know?

Did you mean this one?

Yes.

That's the one I met.

Anyway, that's my little pet peeve.

What, what I'm wondering is, is it any good at actually discerning things and making those connections, particularly with regard to location, get me the best sandwich.

Near me for God's sakes.

I think the answer to that is it's kind of okay, but I think that that is a larger contextual conversation about AI and Geo, is that it understands language and it can understand nearby as long as there's some agent in there that then goes out, checks to see what your nearby is and brings it back.

To the LLM and gives you that answer, uh, or he gives it that answer so it knows where to search.

But right now it, it's, it's a loosely coupled model.

And, um, so yeah, I think there's, I think there are issues and there are gonna continue to be issues.

Right now, of course, iOS 26 is still in beta, but even once it's out of beta, this isn't gonna really change the way the LLMs themselves work.

And I think that's one of the big questions for us with geo.

And AI is finding a way to branch that area between location and language.

Because right now for most of us, AI is deep learning, machine learning, uh, tools as opposed to the LLMs.

Uh, even with Esri's recent rollout, that focus was on, you know, how do you use LLMs to help you find.

How to do things or where things are in the dataset.

So, yeah.

Or not, sorry, not in the dataset, but in the, the interface.

So I don't know.

I, I just want to make a caveat there for anyone who may be Googling it.

I don't actually remember if the other empire is in Montana.

I just remember it's out west somewhere, but it's like really it's days away for me to drive and I can't imagine there's more than, or there, there's only two.

I mean, that's, yeah, there's, there's multiple ones and, and in fact Empire music.

Um, one of them is a record store, which I love record stores, don't get me wrong, but it's, no, again, one of the ones that's maybe California or someplace is a record store, and I'm looking specifically for a guitar store, so.

It.

It's just those type of things.

Another place where it happens is when we're in Pittsburgh, a lot of times we'll go to Whole Foods because there's no Whole Foods near us and there's some things we can get to Whole Foods that we like.

So we'll be in a place and I'll say, plot me of course, to Whole Foods, and it will automatically pick the one that is furthest away, almost always, which is.

Weird, like why I wanna wanna go further and longer because I'm driving to a city.

It's a very odd, the, the model as tied to location to me is very frustrating.

I find it not as robust as I want it to be.

Do you think you find it more frustrating because you do have a sense of what it's capable of and you know the terms you should be able to use?

Or do you think that people who don't have that background and use it are even more frustrated?

That's a really good question.

I mean, I have some conceptualization of, even in a place like Pittsburgh where if you've never navigated Pittsburgh, it's, it's ridiculous.

It's unique.

I was gonna say unique.

It is a very weird place to navigate.

Even with that weird navigation.

I do have a sense of roughly where and how to get places very roughly.

Um, I would imagine anyone.

Navigating Pittsburgh that's not familiar with it, which just takes the default.

I think you're right.

They would just go, okay, and then not really.

So maybe it is, my frustration is a partially a function of having some idea of where these non uh, unique places in an area can be or should be.

I've just gone on for mental that tangent.

This is another good podcast episode.

We should do continue with things that I don't think anybody's gonna say anything about, but they might.

Uh, OSM has released vector tiles.

So if you're using OSM and you've wanted to be able to use vector tiles instead of the uh, OSM stream, you now have access to some layers, I think.

I don't think they're all in there.

I was gonna say like, how much so far?

Yeah.

If you're not using Vector telles, you should play with them a bit.

I mean, they're a very nice way to, to lower your footprint.

In terms of how much data comes across and how much stuff you're generating on the backend and all that stuff, so, yeah.

Yeah.

And this is based on their entirely new backend.

Um, so it's one of the reasons it can be so quick.

And another one, which I rearranged so that everybody knows, is now Instagram maps.

If you've used Snap Maps before, apparently it looks the same.

It's been a long time since I looked at Snap Maps.

It's been a long time.

Or it's been a week since I looked at Instagram maps and I only looked at it to make sure I wasn't posting to it, even though I rarely post.

Yeah, I'm safe from that since I literally haven't posted anything on my Instagram since I joined it whenever ago.

Yeah, I, I use it a lot and uh, I'll say I find it a weird, confusing app anyway, and I find that I don't.

Uh, I, I don't, I wouldn't have run across this on my own.

'cause I, I purposely go get your garbage outta my way and let me just do the thing that I want to do and leave me alone on Instagram.

That's how I feel about Instagram.

And that's, that's kind of problematic because apparently for some people, uh, depending on what your settings were for privacy anyway, you were automatically posting.

So your, your posts were automatically being mapped, um, to everyone.

And so if you were posting things from home.

They were there.

And of course if you were yeah, posting things, uh, you know, and you're someone who people are likely to try to find, then they would know where you were when you posted.

So it's, it's.

I mean, there's some of that there that's always been there, that you've been, have been able to get to the location information, just not as easily, and not just, Hey, here's a map of where it is.

It does look like it's mostly for friends as opposed to followers, which is, yeah, some level of niceness, but also you know it since it's tied to your Facebook friends.

You know, you may have friend at a whole lot of people that you don't really want to know where you're at at any given moment or anything like that.

So then it may have changed.

They may have changed the default after the backlash whenever it first rolled out.

Because at beginning it was kind of a, here's the map, everybody can see it unless you have it marked to be only friends or to be private.

Um, so yeah.

And you still, I think you do still have the option to, at least that's a couple of days ago, to market as everybody if you want to.

Yeah.

Which, you know, if you wanna live your life out loud, go ahead.

But yeah, no thank you.

Well, it's just really interesting 'cause I know I saw a lot of.

You know, people that were on Instagram, you know, saying this is going on, and it feels like the rollout was that they, they didn't realize this was happening.

And it's really interesting that a, a rollout like this could happen where it's, uh, you know, an opt out instead of an opt in, um, for the Instagram map.

Yeah.

Um, and hopefully they've, you know, adjust.

Future rollouts like that, um, with any location based, um, changes.

But we have seen many examples over the years where that is not the case.

It, it's, yeah.

No, it's interesting.

Like, it, it feels like a, a throwback to, uh, all those things like Google attitude and all those kinds of things where, and also a reminder again to so many people that don't realize that the metadata in your phone.

Or the metadata that's being processed from your phone has location in it.

And so, 'cause some people I know don't really think about Instagram as being something that's necessarily recording place, unless they specifically caption it or say, Hey, this is me at the beach, or whatever.

It just, it doesn't, it does not register that, that.

You know, every, every photo has a place.

There's, I mean there's whole questions about things from meta.

This is one of the few that have been MAP related this summer, but there have been WhatsApp questionable things.

There have been just Facebook and meta in general, questionable things.

Uh, Sue and I both have the meta glasses where we have essentially disabled the AI portions of it just.

So I'm now into the world of, of localized LLMs more than anything else, I guess, in my mind.

But that said, I also then will randomly go to copilot and ask it random questions.

No, it will offer me random answers before I ask it.

I'm like, get away from me.

It's clippy on steroids.

It's it, yeah.

Copilot is a bit aggressive and, uh, what, what I find annoying, I don't know about anyone else's.

Office 365 is I used to go to Office 365 and I would had the Chevron, uh, what are the dots?

I always forget what they called.

Yeah, the waffle, the nine dots, that's, I will always just call it that.

Yeah.

But yeah, you could go to that and you could go to like, I need to get a PowerPoint, or I need to get a accelerator.

You know, whatever.

Go to what you needed to go to.

Oftentimes I would use that as my easy way to get to OneDrive for work, but now like co-pilot's in the way, and I'm just like, I just want to go to my OneDrive, man.

Just get out of the way.

Here's my response to that.

I had removed the link bar from all browsers like years, like decade or more ago.

A month ago when this happened, I re-added the link bar just so I could have a link that's direct to the apps page so I don't have to go through copilot to then click two more times just to get to the list of apps.

I don't even know how to do it.

I'm not sure where you go.

Uh, over on the left hand side at the bottom, yeah.

It says Apps click that.

And that is essentially the same as clicking the waffle menu before we have a, we have, because of the.

At, at work, at our higher education institution, we have these, this card page that has a link to all those.

And lately I just wanted that 'cause it's quicker, but I've never, I'm just like, where the hell the thing go?

I'm not gonna draw this.

Go to the card page and I click on it.

Yes.

I don't, I don't understand why they had to say, okay, well you know this portal office.com used to take you to what you wanted.

Go to office.com, but now it's gonna take you to M 360 five.microsoft dot.

Slash whatever.

I don't know.

It's just, and and, and it's really annoying 'cause I got, I got snarky about it.

I was like, I typed into co-pilot, how do I open office?

And it was like, oh, well you just go to the start menu and you click on it.

I'm like, I know how to do that stupid machine.

Anyway, sorry.

Yeah, it's just the dumbest, it's the dumbest thing.

The, the machine is training us and our syntax.

So rather than the other way around, uh, which I, I think kind of goes to the last one a little bit, I guess.

No, not really.

Nah.

But, uh, Google has released Alpha Earth Foundations, which is, uh, in my mind a dataset.

More than anything else.

Um, the link in the show notes will say that it's an ai, what do they call it?

The, the first line, an advanced AI model, but it's the results of a deep learning model to create essentially 10 meter pixel representations of the earth that they call the 10 by 10 embeddings as opposed to a 10 by 10 meter thing.

Yeah.

So it's.

Something we're just gonna point you to.

There's information about it.

Uh, it has set, quote, unquote, 64.

Well, it depends on where you look at it.

Uh, on the same page it says components, dimensions, and bands.

It's referred to all three ways, and one place.

It's in two of those in the same sentence.

No idea what any of those things mean, whether they're different colors.

Don't worry.

So it's been, it's like land use, uh, land cover data set that's already been pre-processed for you.

And I assume that's what it is, but I'm not sure what those 64 are.

'cause I haven't been able to find that page yet.

Uh, yeah, I'll let you guys talk about it.

I have questions about it.

Yeah, don't worry about it.

It's Google.

It'll be shut down in two years, so if you, so it is, the idea is that this is, uh.

A data set that could be used to look at essentially land cover and other types of elements related to that.

And it is available, uh, what they've done so far.

So I think that covers the year 2017 to 2024.

And I don't know if it's one data set.

Uh, I guess the idea in the beddings that you can get to the, the data sources, but it's called the satellite embedding data set.

And you have to use it in Google Earth Engine, which is, which is free if you're using it for research and education.

And a couple other things.

So it, it doesn't seem that it's easily able to be used in other platforms right now, but, uh, anyway, I think you should check it out because I mean, one of the things it's trying to address is sparsity of data.

Uh, at certain resolutions, uh, which when you're using remotely sensed, uh, data sets for understanding, uh, agricultural issues or ecosystems or things like that, the resolution's really important and some of our most complete data sets don't have necessarily the best resolution for what we can do now.

So anyway, uh, if you're interested, uh, check it out.

Um, there's a, a writeup in an article about kind of the, the technology, math and everything behind it.

But, um, if you want to see the dataset, you can, uh, download Google Earth Engine and check it out.

It's called the satellite embedding dataset and just to, to kind of, it's just what they think of it.

They're, they're thinking of it as a virtual satellite, right?

So to cover things.

That the sensors and data sets we have, uh, just can't quite capture.

That's the idea anyway.

So do either of you use, do any of you use Google Earth Engine?

I, I don't.

That's why I'm asking.

Not used it since before.

In the Before times.

Yeah.

And thats you.

It's been a really long time.

There was some reason why maybe somebody asked me or something.

I downloaded it and played with it for like a half an hour.

So I do not use it.

No, Jesse.

Not, like Sue said a while.

Yeah, I, I played with it and I was going to use it, but I remember running into the issue of there was a cost association associated with it.

Um, so maybe I just didn't figure out.

Yeah, it's a full stuff now.

Maybe they just like with, with Google Earth maybe.

'cause there used to be premium stuff there too.

But yeah, I haven't, I haven't tried to use it in a long time.

I was just curious 'cause I feel like it's, uh, I mean it was the bee's knees at one time, but it, it, I'm not sure it's made a ca Well, okay.

In fairness, probably because I have access to all the ESRI stuff, that's the dominant reason I have other op other things that I can use to answer the questions I want to answer.

So I was just kind of curious is it is just not a tool set that I turned to.

None of us are doing research on a large area.

Yeah, so I think that's true.

That's, that's where Google Earth Engine, um, shines a little bit more.

A lot of us do a, at the table, do a lot more of very, uh, large scale, small area, uh, kind of research.

And so, you know, being able to map the whole world is, is less useful for us.

We just want to be able to focus on certain areas.

We want lidar, we want, you know.

Drone data for imagery, we want, you know, someone who's gone out with a GPS unit and collected data for us more than anything else.

So it, it, I think it is a scale issue between what we do and what Google Earth Engine offers.

I, I think that's what I see usually when I, when I, um, see different forms where people are discussing it, is they tend to be working with, um, those coastal and, and large scale world environments.

But we want people to know about it, even though.

I who started talking about it then kind of had questions about it, I still want you to go out and look at it.

Just be aware that it, the, the metadata is not as.

Transparent as I would like it to be, but maybe in Google Earth engine it's a little bit better.

And that's it for the news.

All right, so, and onto the web corner.

Alan Carroll has a book about his journey telling stories with maps, um, lessons from a lifetime of creating place-based narratives.

Um, it brings together the stories of a.

How he thought up this idea with his background at National Geographic as the chief cartographer there.

Um, the fundamentals of storytelling for humanity, um, talking about best practices and how maps help us to tackle who, what, when, where, why, and how.

The connections between maps and memory, and also gives you a way to learn about approaches to storytelling and place-based topics through story maps.

If you've read any of a Aaron Carroll's very well written.

Articles, they, they tend to be.

Just like story maps, they flow really well.

They tend to be succinct, but they give you what you need to know to, to tell that story.

Um, so I'm looking forward to getting this book and using in it in classes.

He also focuses on practical features and functions for story maps, um, so that he can inspire you to apply them to your own stories.

Um, so I had the opportunity to go to a workshop he taught virtually online several years ago, and it was for primarily museum.

Geographers people working with exhibits inside and outside and how to make that connection to the stories with the exhibits and the places.

Um, a lot of the Smithsonian and the, the cherry blossoms, which, you know, is one of the earlier and still most impactful story maps out there about the story of the, the cherry blossoms in DC Yeah.

If you're interested in either Island Carroll's career or.

Story maps, go check it out onto the events corner.

As always, we encourage you to go check out events such as State of the Map, October 3rd through fifth in Manila, and there is a call for posters.

Geo Week 2026 will be February 16th through the 18th.

So now we're getting into 2026 in uh, Denver, Colorado.

Uh, of course if you'd like us to add your event to the podcast, send us an email to podcast@veryspatial.com.

If you'd like to reach us individually, I can be reached at you@veryspatial.com.

I could be reached at barb@veryspatial.com and you can reach me atFrank@veryspatial.com.

I'm available at kind of spatial and of course you can find to.

All of our contact information over at very spatial.com/contact.

As always.

We're the folks from very spatial.

Thanks for listening, and we'll see you in a couple weeks.