 Yes, recording now, thanks. And also please, I think we can, given that it will be a relatively small group, we would be happy to have a discussion at the end of this session, but feel free to introduce yourselves in the chat. We'll go through our demonstration first and then hopefully have time, and I'm certain we'll have time to share and discuss, but also to introduce ourselves properly towards the end of the session. So with that in mind, I will give a very brief introduction to SDSN and to the SDG Transformation Center, just two slides, since I know that you're mostly interested in looking at the data and how to extract it from our data repository. But by way of backgrounds, SDSN for those who don't know is a global network launched in 2012 to support the implementation of the SDGs. We operate under a mandate from the UN Secretary General, and we're directed by Jeffrey Sachs, an economist based in New York City. We have offices in New York, Paris, and Kuala Lumpur in Malaysia. And the SDSN consists of a few different offices, we have different goals. Some of what we do is about SDG policy analysis and support, and so that's what Gilamour and I work on is research into new indicators, new data sources, new ways of making sure that the data and analytics that we produce are used towards better policy making. So we help governments, we help UN institutions and other bodies to reinforce the science policy interface and to ensure that policies towards achieving the SDGs are measurable and actually entertainable. We're also a global network of knowledge institutions, so there's over 1500 universities and research institutions throughout the world that are linked to SDSN as members, so we share best practices, share the latest research on all of the SDGs, but some of our affiliate networks are specialized in certain areas, which we invite you to learn more about on our website. And finally we have an online education component that's to introduce SDSN at the SDG Transformation Center is the think tank piece of the puzzle of SDSN, and we work on producing data and analytics, including geospatial data and analytics, which Gilamour will be presenting. And we present this work through a number of indices, the SDG index, the Sustainable Development Report. We also work on international spillovers, but also increasingly on financing, the financing gap for SDGs at the country level, tracking SDG policies, so the extent to which governments are actually engaged in achieving the SDGs, and also looking at SDG implementation locally throughout the world. So with that very brief introduction out of the way, I invite you to look at our website, the SDG Transformation Center, and also to follow us in LinkedIn if you haven't, I'll share links in the chat, but I will now leave it to Gilamour to kick us off with the topic of our session, which is about the Rural Access Index, and he will take it from here. Thanks, Gilamour. Great. Thank you, Eamon. So hi, everybody, and happy GIS day to everyone. So I'm Gilamir Blomolsky. I am the SDG Transformation Center Geospatial Data Specialist or Geospatial Data Scientist, as one would wish, and my job essentially revolves around calculating and providing data sets for particular SDG indicators, so let me share my screen and we'll get right into it. Here you go. So I wanted to take a moment to talk about the Rural Access Index, but also about other work we've been doing in terms of geospatial SDG indicators, and I'll get to why that's relevant. So first of all, we are funded by Esri. So we have a bunch of our tools that are using our GIS tools, and if you go to our website, you'll notice that our flexure report, the SDR, is always available as a PDF, but also always as a map where you can filter by SDG indicator or by SDG go all together. So all of that is currently being developed with ArcGIS for JavaScript, and we also have a number of story maps that are done with ArcGIS online. So a lot of what we do is based on these tools. But the main focus of what we have been doing is creating geospatial data sets for SDG indicators, and that's where we might have more than one way to look at this. So you probably are aware of how things are done at the global UN, the hardcore UN ecosystem, where everything has to be either country-led or country-owned. So essentially much of the statistics that are being put out currently by UN stats is highly dependent on countries being able, so governments being able to run those methodologies and deploy those data sets themselves. We are fortunate not to depend on that ourselves, so the SDR is running on data that is sometimes calculated globally and not country by country. And for geospatial, this makes especially a lot of sense, because as we know, a lot of the data that we use in geospatial is coming from satellite imagery, which is collected very often at global scale. So we can run those calculations at global extent, and we can inform on those indicators across the entire globe at whichever scale in very comparable, we make everything comparable by calculating all at once. So we have been deploying for, so for our latest report, which was released last June, we have two new indicators that inform on STG indicators related to accessibility. So the first one is related to the 15-minute city. So we have run some Python scripts to calculate accessibility for pedestrians at all main large metropolises and also medium-sized cities in the entire globe. So we have used OpenStreetMap data for that and also open-sourced Python packages for those calculations. And more recently, we also did the work that I'll be showing today, which is the Rural Access Index, which is not a very difficult one to outline, but it's certainly a difficult one to calculate. It means a challenge. I think I have someone at my door, if you guys just excuse me for two seconds. I'll pick up, well, Guillermé is addressing his urgent issue. The end he'll probably speak about this more, but the methodology that we're using for the Rural Access Index was developed by a team at the World Bank. Perhaps you've, oh, you're back. I'm back, sorry. I'll let you pick it up. Okay, thanks. So essentially, there's good reason to run these calculations, because sometimes they have never been done at global scale following the most recent methodologies. But there's also a case for providing these calculations and deploy those methods to provide the data at a more granular level. So being able to provide not only country-level data, but also province, county, municipality-level data, which informs so much of what many, many cities are trying to do, which is producing voluntary local reviews, which are also based on the STGs, but need data that is relevant for those very localized locations. So just to comment on the tools we're using. So in many of these cases, we are using Python, a lot of Google Earth and Gene as well, and OpenStreetMap plays a big role also in providing data that you can exploit at global scale. So all of that is being made available through our website. We have an online library for the reports where those data sets are being integrated, but we also have a portal for the data we are producing. So you will find on our ArcGIS Hub, you will find the data in the format of a spreadsheet for all of the indicators that are being used in the report. But you will also find in its original format, so it's geospatial format, the indicators that we are producing with geospatial data. So you will find those two indicators that I showed and also some other ones. And just to mention, we are also partnering with the UN GGIM, which is the UN secretariat for geospatial data and all things geospatial in order to sort of break that barrier of not being able to use geospatial data at global scale for many of the reports because of how things are set up. So we're trying to break that barrier at the larger UN ecosystem. And so why is that relevant is we know that we are making some progress. We know that we are not making progress in many areas, but there's a rather large data gap for many of the indicators. If you take our latest report, you'll find that for the indicators we do have access to, we haven't really been making good progress since the pandemic started. For some other ones like infrastructure, it has been actually going upwards, but there's a lot for which we don't really have enough data to know. And many of those critical data gaps can be filled by geospatial data. There's a lot that exists, and it's currently not being leveraged into indicators for many reasons. And I'll cover the reasons why that was the case for the Rural Access Index. This slide is essentially to say all of this is useful because it provides better temporal and better spatial resolutions that can then be used in very desegregate scales for countries and states and municipalities, et cetera. So as I just said, those two accessibility-related indicators are part of the latest SDR. So we did on SDG11 the access to relevant services for pedestrian in urban areas, and then for rural areas, we did for SDG9, which is energy innovation and infrastructure. We measured populations access to our seasoned roads. So what's the story behind the Rural Access Index? It's actually rather sort of like a fruit salad of methodologies. A lot has happened. So in 2006, the World Bank created essentially the idea for an indicator that would measure the access that rural populations would have to roads and not any roads. So a road that provides access all throughout the year. And that created a lot of questions. So what is an all-season road? How do we delineate what that means? And that first iteration in 2006 was actually trying to measure poverty. So it was a proxy for poverty, and it was being measured through household service. So the responses were not really comparable. So different cultures, different locations would classify the roads they were using to get home as providing all-season access or not. And that provided a lot of variation. It was hard to get a glimpse of what's the situation for the entire globe. So it took 10 years for the World Bank to provide a new enhanced methodology, which they call measuring rural access using new technologies. And those technologies were actually GIS. So that's the point when the World Bank says, OK, we can't do service anymore, we need to use geospatial data, and then we can actually start comparing different locations. Otherwise, this doesn't really mean much. And so that happened. And so there have been some isolate applications of that method since, but no global implementation. And all of those implementations had slight changes. So slight differences differentiating them from one another, so making it hard to really create that comparison. And so in 2019, the World Bank endorsed and not commissioned a new supplemental guideline provided by TRL, which used to be a company from the UK, which provided some additional insight, especially on what makes a road all season. And so they took the expertise from a lot of road engineers saying what are the main risk factors that make a road impassable at some point throughout the year. So that 2019 methodology is the latest one that was endorsed by the World Bank and the World Bank is the custodian. So they have essentially the power to say, this is the latest and this is valid and the UN will endorse my endorsement. And so that has been sitting there in their website. And if you access the World Bank data catalog, you'll find that all of those methodologies and some data sets for some specific countries done following that. But you won't find any global data set. And that used to be entirely possible to just do. And so that's what we did. So we took that latest methodology and we applied it at global scale and at the most granular level we could get to. So you will find some global data sets like the one that is currently hosted on STGs today, which comes from the NASA CDAC. And you'll also find one that is mentioned by this TRL report, which was done by a company called Azaveya. But none of them actually follow the latest methodology and I'll get why that is critical because it really changes a lot of how you understand all season access. So this is a little diagram for us to understand what we're talking about when you talk about access. So in the definition, the World Bank stated that the RAI-RI access index is the proportion of the rural population who lives within two kilometers of an all season road. So in theory, it's actually pretty simple, right? So you just need to take all roads and all season roads and then create a buffer, a two kilometer buffer on both sides and seeing how much of the rural populations of any given place falls within that buffer. So in theory, it's really simple, but that actually raises a bunch of questions. So what do we consider to be rural and what to consider to be urban? Which roads can you consider to provide all season access? And considering that there's no time of database that contains information on all seasonness, how do we get close to that? What proxies do we use to get close to all season access? So all of those global datasets that I just mentioned did something that is actually pretty simple. They just equated all season access to mean road surface. So if a road is paved, it's considered to be to provide all season access. And if it's unpaved, then it doesn't. And so that's the whole point of this latest methodology is saying actually in many rural spaces, non-paved roads will provide all season access. There's nothing intrinsically wrong with an unpaved road. It can provide all season access. So all of those datasets were really providing understated or underassessed numbers for that access. So we were getting like 30% population having access to all season roads in countries like Colombia, which just didn't make a lot of sense. So that's the challenge we took on to define what is rural, what is all season and how we get at least close to that. So this is a diagram of the approach we took with all of the processes and all of the data sources that we use. So I'll cover this in bits. So the first thing, the first question was what is rural and what is urban? So if you were at the presentation from yesterday, from STGs today, they did a GIS Day event as well. They are also part of STSN. They were talking about the Degurba methodology that was created by GHSL, who are a part of the European Commission, which is essentially a new method for outlining urban areas and by extension saying what is rural in contrast. So we use that, that is the current best data we have on urban data, but we also use a previous and actually better delineated dataset that NASA put out a few years back. So we took those two and filtered for urban. So we took both of these at a global level and from that we derived the urban land cover, which is going to be subtracted from everything we're going to do following on. The second thing is trying to find a good dataset to make sure we have all of the roads mapped. So for that, we took three different datasets that we put together. So the first one comes from Global Bio. They have a dataset called Grip Road Network, which actually is taking data from authorities. So national governments are providing datasets to Global Bio and they put all of them together under the same data schema, which is great because many of those datasets actually have a field for saying does this provide all season access or not. And the issue with that is that its coverage is just not great. So many countries are not currently able to provide a good vision of what's the current road coverage. And this was the case for developing countries, but also for some developed countries where the official data was just not updated quickly enough to follow new roads that were built year after year. So for that, we also needed something that would provide that timeliness to this dataset. And that's why we use the Bing Road Network dataset. And that provides the best timeliness because the roads are being extracted from satellite imagery through neural networks automatically. So that doesn't depend on human being digitizing a road from plans or from imagery. So that is being released by Microsoft yearly. So from that, we can be sure that we have the full extent of roads. But that dataset doesn't come with any attributes at all. So it's just a road and nothing else. So we are obligated to consider all of these roads as being unpaved. We are not sure what's the pavement there. So we just assume the worst case scenario for the roads that exist exclusively on the Bing dataset. And then to make the bridge between those two, we are also using OpenStreetMap, which has for starters the hierarchy from which we can learn a few things. And in some cases, it also has the pavement information. So we took all of that together and then we created two different subsets from these datasets. So for all of the roads we consider to be paved, we created a buffer, a 2-limiter buffer around these roads. And for all of the roads, we consider to be unpaved within a separate buffer. So now we have two subsets, one for paved, one for unpaved. And then we have the buffers around these two subsets. So for the first one, where we consider roads to be paved, then we are going to consider all populations that are rural, that fall within that road buffer, have access to an all-season road. And that's perfectly fine. So everyone has been considered, if you have access to a paved road, then you have access to an all-season road. Then we can just run a clip operation. Unpaved roads are going to be more complicated because we know that some of them won't be able to provide all-season roads because we'll be finding potholes or they're going to be flooded at some point of the year. And some of them will be perfectly fine. Actually, it's completely fine to use an unpaved road to access your home. So for trying to understand which of these unpaved roads provide or not all-season access, we are going to create a passability index. So the passability index is actually being proposed from the 2019 methodology, where they say where those road engineers are telling us what is really critical for a road to be kept and to be kept passable is the amount of rain that road is going to be receiving and the type of terrain where that road is located. So a terrain where the slope is very high and where it rains a lot, that road has worse chances of providing access all throughout the year. And we are completely on board with that, except when we run the first time those calculations, we realized that very different places on earth would be presenting very similar amalgamations of those variables. So we noticed that in Scandinavia, so Sweden and Norway, we had some very slow peak rains and a lot of rain. It's one of the places where it rains the most on planet earth. And that would be very, very similar to places near the Congo basin. So Sudan, for instance, were presenting very similar results to that. And so we realized we needed to actually add an extra indicator that would tell us how are those roads being kept? Are those unpaved roads receiving any maintenance at all and by how much? And that indicator exists, which is the proportion of the GDP that is suspended on road maintenance, but that is not available all throughout all countries. So we needed to also use a proxy for that. And so we used GDP per capita to get closer to that maintenance status of roads. So we take that passability index that is calculated from those three sources and then we can just multiply that population that lies within the unpaved road buffer. And that will essentially tell us how much of that population we believe is going to have access to an all-season road. So that passability index is going to vary from zero to one. So if all of this is really intense, if the GDP is low, if it rains a lot and the rains are really sloppy, then this will be close to zero. And so we'll multiply that population at pixel level. Remind you that all of this is taking place in terms of raster algebra. So all of those calculations take place at the pixel level. And so if this is close to zero, I just multiply that population within that pixel by zero. And so I assume that 0% of people living there have access to an all-season road. But if all of this looks good, and it'll be closer to one, and then I can multiply all of that by one, and all of them are considered to have access. And so we put those two clipped population rasters together and then we get to the rural access index by dividing by the entire country's population, rural population. So this is sort of the larger picture of what we did. And of course, population is coming from World Pop, which provides the better resolution currently for population across the globe. So how does that passability index work? So we're not doing the same as other methodologies did in the past, where in some of those cases unpaved roads were just removed altogether from the network. We are applying that passability index to the unpaved road network. So what that is doing really is keeping the population in proportion to the likability of that rural being all season or not. So this is what I was just saying. So we're doing the pixel level through raster algebra, and population is capped at 100% if all of this looks good and to 0% if it doesn't. So this is a little matrix of what the first iteration looks like. So for the terrain, we are taking the minimum values up to the maximum values. So this is a very sloppy terrain, and this is a very plain terrain. And here in climate, you have this is on the contrary. This is a very sloppy terrain, and this is not. And this is a lot of rain, and this is a rather dry climate. So if those two conditions meet, so sloppy and rain, then the index will be close to 0.06. And if good conditions meet, then it's closer to 0.95. And if they're perfect, then it's 1. And then as I just said, Norway was presenting pretty similar results to South Sudan. And so we had to add that third indicator for the passability index. And this is what it looks when combined. So if I have, and this minimum then becomes very rainy and very sloppy, and this maximum becomes very dry and plain terrain. And then we also add GDP per capita here. So a very low GDP and bad conditions for terrain will result in 0.03. And if you get to the maximum here, and this was normalized, then you'll get one. So for whichever case, if you are at the maximum GDP, whichever terrain and climate conditions will be reverted back to 100%. So this comes from the idea that if a country has a really, really high per capita GDP, it doesn't really matter what conditions a road is exposed to. Because we have confidence that those roads are being capped with good maintenance. All right, so that was sort of... Before we move on, we have 20 minutes left. There was a question, is it okay if we address it now before going into the demonstration? Let's do it. Okay, so there was a question from one of our colleagues present about how do you distinguish between the rural population in different regions? Are there differences in the way that rural population is calculated in Asia versus Africa? For example, can you just explain a little bit how it's harmonized or not? Yeah, so you'll find all of that explained at the GHS Mod at the European Commission website. They have used, of course, satellite imagery and also samples from local experts in telling what is and what isn't, but essentially it's a trade-off between the number of people in a pixel and the density of occupation in those regions. So it's a threshold of quantity and density that is essentially telling you if that's urban or if it isn't. And in this picture, you can see those blobs here of gray areas. Those are urban areas, but this is actually what the grump dataset looks like. So grump, that is run by NASA, with a very similar proposition, they provide those as a vector. So it's much better delineated and I feel like they actually had someone look at pixels and then create clusters of these pixels, whilst GHS Mod is provided raw at pixel level. So in some cases, you might have some noise in the data because those conditions are met or they are not met. And that is a problem especially in northern India, where you have very sparse but very close together populations. So that region is the one that will vary the most between those two datasets, considering them to be rural or not. Because they're really close together, it's a lot of people, but it's still very much looking like a rural area. So that is one of those decisions you need to make and so we went with GHS as mod for that. But yeah, population and density. Thanks. So all of that was calculated in the cloud through Google Earth Engine. So luckily, most of those data sources are already available freely on Earth Engine. And so we don't have to actually run the process at once and create like this really gigantic raster for the entire planet at this very small pixel size. So that can be done interactively. And statistics are created when we extract zonal statistics by a country. But this is what the raw data looks like on Earth Engine. So you see here the buffer for the roads. And you see in yellow the populations that are within those buffers. In gray, you see urban areas, of course. And in red, you see populations that fall outside of the season road buffer. So all of those points in red are people living outside of the two kilometer buffer. So this code is made publicly available and it can be used to create statistics. But that is, you know, there's a knowledge barrier there to run things in that platform. So we make things available at country level on our own website. So you will find, and I'll get to that in a minute, you'll find visualizations and the complete methodology is also made available on the website. And you will find some more scientific checks of that approach. So we did two checks. So we checked for construct validity. So that is that we did by comparing our results to other similar, but still different methodologies for the same thing. So the Pearson coefficients, the correlation coefficients for that were pretty high. So we are not far from them, but we are not identical to them. And we also checked for convergent validity. So comparing our data set to others that we expect, we would vary accordingly. So we did that for GDP per capita and also for HTI and got pretty decent results. So we know they are not identical. But they are somewhat related. So we are pretty happy with that. So I want to take the time we have left to just show you how you can access that data. So if you go to the STGTC website, you will find a whole lot here. You can find our reports and our work on financing policies and also on spillovers. But we have a tab specifically for our geospatial data and tools. So if you go to about our geospatial tools, you'll find some explanations of what we are currently working on and what is available. So you can download data from these produce indicators. And you can also explore methodologies for some of them and also some maps. So if you click this explore button here, you'll get to see this very simple map that is hosted on ArcGIS online where you can check results for each country dynamically. You can also check the methodologies. So this is a PDF explaining all of what I just showed. And you can also download. So if you go to the download button, you will be taken to our ArcGIS Hub website. So this is where we keep all of the data we produce. And in here, we'll be able to not only get, for instance, a shape file. So here you have the metadata for this. So you have all of the explanations of how this was done from rich sources. You can use the API. So ArcGIS API for creating apps with this and whatnot. But you can also just plain click download here and get this in whichever format you want. So you can get this as CSV, a shape file, a geojson. You can do this as you wish with this. I had thought I would have the time to show how to make a dashboard out of this, but I don't think I have enough time. So I'll maybe just show you what I did prior to our meeting this morning. So you can find this data publicly published on ArcGIS online. So if you go to ArcGIS online.com and you create an account, I think you can still create a free account these days. You can just go to content. And if you search the entire catalog from ArcGIS online users, you will find the real access index available for use there. This has also been published to the ArcGIS 11 Atlas, which is Azure's catalog for curated datasets. So if you look for real access index here, you'll find it listed right here. So you can also access it through here. But what's fun and what I'm going to show is that you have a map here, which is also made available. So this is the feature layer that is being used by this map, which you can also access because it's all publicly available. You can create a web app with this. So if I want to create, for instance, a dashboard with this, I can just open, I'll just say this is GIS Day number two because I did one this morning with the same name. And then I can create some pretty simple but really effective visualizations with this map. So it'll load the map that already has a particular symbology to it. Then I can just go on and add some elements to go with it. So I think since we are measuring things from 0% to 100%, gauge is interesting. So for the gauge, I just need to select a layer, which is of course the only one we have there. And I want to show a single feature by Rai. So here you go. You can see it like this. It can also add colors to this. So from up to 50%, it's going to be a rather bad score. Up to 75 should be an okay score. So I'll just make this yellow. And up to 100 is going to be an excellent score. So I'll just make this green. And I can also add a title to this. And so the title should be just the country name. Here you go. And I'll also mention that this is the Viral Access Index for that country. I'll just centralize this and make this a little bigger. So here you go. And now I have a really big gauge here on my right. And I just need to create an interaction between those two. So if I go to my map and layer actions, I can add a filter action that goes to my gauge. So whenever I click a country here on my map, I need to go here actually. It's going to be under... No, actually, this is not needed. I just need to also make this work on... Here you go. Select feature. There you go. So whenever I select a country here, I see the little gauge on the right also change. I can also add different stuff. Like if I add a single bar chart here, it's in the same layer. But I'll just type a group by continent. And I'll be showing the statistics of Ubright. But I also need that to be averaged by continent. And there you go. You know, we have a question. At what scales did we aggregate the results only at the country scale or at the county and province or city scale as well? Right. So currently in our arch.js hub, you'll find this only at country, at national scale. But you have full access to the Earth Engine scripts where you can run these zonal statistics for whatever polygon you provide. But there you go. I just wanted to show you that this is all available on arch.js online and that you can do whatever arch.js online provides in terms of apps and maps. So this is also true for our different datasets. And for those other ones, we do have statistics by city and even intra municipal scale. If I watch the name of the story map, I think it's accessibility indicators for the STGs. So here you can see some maps that we put together for the urban accessibility indicator where we did that for every city in the planet using similar datasets but different methods. There you go. So Lagos is rendering here where you have accessibility at the intra municipal level. So these are points that look like a raster, but that's actually vectors. So all of that is also available, but this is outside of the rural realm. But we are thinking of making the rural access index available also at different scales. Shouldn't there be any interest in that? We have just a few minutes left. I think if there are any other questions, please feel free to ask. Otherwise, we will be providing our contact information so that you can reach out to us individually or after the presentation to follow up on specific questions about the data, about the methodology, or about how we're using this work. So as you've seen, I mean, this is an excellent example of the transformation centers and give me specifically his work on indicators for using geospatial data, but we're also exploring how this can be applied to policymaking. Also, how cities and local subnational governments are making use of geospatial data at their level, given the specificities of the local contexts and also trying to publicize and make people more aware of the usefulness, the timeliness, and the global scope of these indicators and of this data for measuring the SDGs globally. So on that, we have just a few minutes left. Once again, thank you. I see that someone asked or says data on cities, the local level will be great. Yes, I think that's, we've heard that in other contexts. And I think, again, you can within the data that's provided online, you can filter it to fit your needs, but I think that we can explore producing a more accessible format that's focused on cities, not for the ride, but for some of the other data. And on that, I think we have just two minutes left. We, the SDG Transformation Center and SDSN will be hosting events like this once a quarter, basically. So we invite you to, if you found this useful, to join us once again in early 2024. We'll be exploring some of the other work that we've been doing using geospatial data and some of our ESRI tools. And hopefully next time, we'll be able to involve some of our other colleagues, either at ESRI and or at UN, UN organizations, to compare and contrast how this, the data is being used and also to see some, some practical use cases of the, of its application. So on that, do you have any, any word of conclusion? No, thank you everyone who showed up. I'm looking forward for those next ones. And don't hesitate to drop me a line if you have any questions or if you just look to chat. So thank you very much. Once again, this meeting has been recorded, so we'll, we'll put it, we'll make it publicly available if you need to or want to look at it, watch it again, share it with colleagues. And in the meantime, feel free to, to follow us on LinkedIn and to check out our website, SDGTransformationCenter.org. And yeah, so on that, thank you very much. Have a good morning or evening. And we look forward to speaking to you again soon. Thank you. Thank you all. Bye-bye. Bye-bye.