Saving the world with geospatial data: Sustainability analytics on Google Cloud

foreign [Music] everybody my name is Chad Jennings and I wanted to say thank you very much for spending a little bit of time with us I realize that we occupied that critical hour between the end of a long day and happy hour so I won't say we're gonna go really fast but I will say we're going to be really interesting so setting the bar high for our speakers um our goal with the talk we have three and they are we're gonna review the geospatial analytics Suite inside of Google cloud and what our investment Theses and philosophy philosophies have been then we have resulting from those Investments we have two really outstanding things to launch which Emily is going to take care of and then we have not one but two customers to come up and talk to you about their Journey so if we're successful with all three of these goals what I'd like to leave you with is a feeling of both inspiration and reassurance that as you tackle these really gnarly questions around sustainability and climate change that Google cloud and our ecosystem of Partners and the other experienced customers out there have got you covered so that's what we're on about let's go ahead and get started so to accomplish those three goals we have four people right that's an over constrained set so we should be good um Emily and I'll run you through the product stuff and then Jeremy and Martin will come up and talk about what their companies have done using the things that we're launching in like you know 10 minutes or as soon as I stop talking um if you are watching this online you can use this Glide to navigate the video for the rest of us here we really don't need it because you all are stuck so let's go ahead and get started here I'm going to set the stage just a little bit and so all of us as data analytics minded folks have been totally consumed with one topic over the last year and that is generative AI right it's everywhere it's everywhere in this constant in this conference and to be honest it's right it came and it disrupted much of the work and much of our priorities for the first half of this year however that's for the analytics industry if you search for Google Trends and you look at generative Ai and all the related subtopics and climate change and all the related subtops subtopics you'll see that climate change is 6 to 10 times more prevalent than anything around generative AI so while we are focused on gen AI large language models how do we incorporate these effectively how do we incorporate these responsibly the rest of the world like planet Earth and most of the people on it are concerned about other stuff they're concerned about climates and it's not unreasonable to see why because it's getting hot it's getting hotter and I'm I won't spend a lot of time motivating this here but I did want to show you these two pieces of data so the chart on the right um speakers right here your left I guess um shows the daily Global air temperature since 1979 through to the first half of 2023. so the light gray lines are 1979 through 2021 the dark gray line is 2022 which was the hottest year on record up until anyone up until probably 2023 right we look like we're gonna take that record away the article on the left side here or sorry your uh viewer is right um you know this isn't just a result of the natural stochasticity of you know of our planet right we are causing this change and so as we are changing the world businesses have to adapt to this new environment and so businesses priorities are changing in response so governments are trying to figure out what new ESG reporting metrics are important and when are they due and when do those metrics get businesses the proper considerations in their Industries insurance companies are trying to figure out how to manage the new and higher magnification of hurricanes right the one coming to Florida right now the one that just happened in La hurricanes in La really yeah so insurance companies are having to Grapple with this financial institutions are trying to calculate the portfolio investment risk due to climate change public sector is trying to figure out how to manage road infrastructure Bridges now that they're actually being submerged in and drying out far more often than is used to happen so really the question of where is sustainability and where is climate change driving need for new analytics the answer is pretty much everywhere and so if you have a platform your platform needs to have three specific qualities to answer this challenge those three qualities are oh sorry I got ahead of myself um when you've got sorry let me back up a little bit wrong slide um when you've got needs that are this prevalent in this omnipresent in the world you start to get input and demands from two places one is the very top of your funnel right governments are setting up new um you know Governors are setting up new uh reporting processes and your customers are asking for new answers and so when you're getting squeezed from both the top and the bar both the top and the bottom your entire company needs to respond and if your entire company needs to respond you can bet the board will want to respond and if the board wants to respond then you'd better have a good strategy so here you go the strategy around sustainability really has two aspects to it one of them first and foremost it's a data challenge right we've never been through this before so we have to do research to figure out what we're up against right that's a data challenge the second one is a cultural Challenge and when you put these two together Google feels like it is very well positioned to have significant impact here because if you have the right data infrastructure and if your data infrastructure has the right capabilities then you can build the right culture to be collaborative and to be Innovative and to be fast so here are the three fundamental characteristics that your data infrastructure really needs you need massive and ready to go geospatial data catalogs you need immense storage capability and a lot of computes and fortunately the geospatial analytics Suite that I'm going to talk about here is really built to highlight these three particular capabilities as well as it's designed around the two major data types that geospatial analytics uses so here's our analytics Suite the um the Suites built on four products the pillars are primarily four products bigquery for Vector analysis Earth engine for raster analysis vertex AI for access to all the Fantastic apis that they're producing and then Google Maps platform for sharing and communication of the results that you generate using the other tools and so let's dive in a little bit to the real or the ones that we're focusing on the session today so bigquery geospatial actually a little bit of a Nostalgia trip for me I was actually lucky enough to be the product manager who launched bigquery geospatial at this very conference I think In This Very Room six years ago since then we've continued to invest to expand more functions and more capabilities inside of bigquery geospatial so that it does more and more for you so we were the first Cloud at that time to have geospatial support and obviously we continue to lead with our continued investment the thing that that investment in by implementing this in bigquery what that really gets you is data catalog scalable storage and this is what massive compute gets you things just go really fast especially geospatial queries inside of bigquery so this is an example of a pretty significant Geo join and so this join is basically asked the question how many land Parcels in Texas have impervious surfaces so you're taking a 12 million row table joining it again with a 12 million row table the answer has 4 million rows and Cloud SQL this takes more than half a day it's 12.1 seconds in bigquery we see this kind of stuff all the time this is not unusual I did not cherry pick this result as a matter of fact I think Jeremy is gonna Jeremy is going to talk about some of his experience with this kind of performance boost so that's what you get out of bigqueries capability here let's dive in about as briefly to Earth engine Earth engines got these three primary pillars of differentiation the first one I mentioned before the data catalog so Earth engine's data catalog is more than 70 petabytes big and more than a thousand different geospatial data sets that are analytics ready so for example let's say you wanted to do an analysis over landsat images over the last 40 years that's including all of the relevant images is one line of code in Earth engine the Earth engine team has built a pipeline that collects it from the satellite does all of the pre-processing ready so that's just ready to go in your analysis um storage and computation we'll talk about that one a little bit when Martin and Jeremy come up the thing I wanted to also dive into is that Earth engine has been in the market for non-commercial use for more than a decade and over that time they have built an outstanding and impressive user Community user Community gets together in events like this but specifically focused on Earth engine and there are 91 000 monthly active users in their community that means if you're new to it there is a big repository of willing helpers to help get you rolling okay the last one the last pillar that I wanted to talk about is one that's actually growing here a cup this has this pillar of Partners has grown remarkably in the last two years and so if you want to do some of this work if you don't have a PhD in geospatial analytics that's fine you don't need it but you might want to have somebody with that qualification on speed dial and that's what you can get from the partner Network so Partners help with expert Consulting a number of our partners uh so carto climate engine acclima woza several of them up here have also built bespoke Solutions on top of the earth engine and bigquery and when they've they've done that customers like yourselves and the partners sorry just waiting for the truck to go by um so customers and partners have asked us for a couple primary things over the last half decade of these pieces being out in the marketplace number one is more capabilities inside the products and number two is better interoperability because when a when a partner or a customer Builds an application generally it starts with whatever the problem definition is that middle circle there is signifying tying several different pieces of the Google Cloud platform together to generate the solution and then exposing the solution either through a map Google Cloud Google Maps platform or through an API but that process of tying several things together can be difficult and then it can be difficult to maintain so we heard that feedback and our investments over the last couple years have been to really dive into interoperability between all of these pillars in addition to adding new capabilities so covered the suite covered the investment thesis now I get to hand it off for the really fun part where Emily gets to come up here and tell you what fruits those Investments are bearing so much Chad so I'm Emily Schechter I'm a product manager on Google Earth engine theme so Chad mentioned that one of the ways we think about building an industry leading sustainability platform is by taking an integrated view across multiple geospatial Platforms in TCP so I'm really excited today to announce two new connectors that make using these tools together even easier so first I'll talk about using bigquery and Earth engine together and what we've done to make that easier so as Chad mentioned Earth engine and bigquery share a common goal to make large-scale data processing accessible and as usable by as wide range of people and applications Earth engine is fantastic for roster data processing like imagery whereas bigquery is optimized for processing large tabular data sets and is already a critical part of many cloud data analytics workflows but by using bigquery and Earth engine together customers can accomplish tasks that are impossible on any other system so to help people use these tools together we built an easy to use function to export tables from Earth engine to bigquery and we're thrilled to announce that this is now generally available to use and there's a great blog post out on the cloud blog that describes how to use it so the new connector is our first major step towards deeper interoperability between Earth engine and bigquery improving ease of use for workflows that use both Services enabling new analyzes that combine raster and tabular data so many of you may have tried to do this before and it involved a few steps so by creating the data connectors between bigquery and Earth engine we collapse these steps into a single command and since those middle steps were really data engineering data wrangling and not analysis we've not only made this process easier but we're also enabling you to spend more of your time on your valve Bill science and Analysis so for example let's say you were doing some kind of crop yield prediction for agriculture so your workflow might now look something like this first you use Earth engine to calculate biomass from some satellite imagery or maybe entire continents you bring that data over to bigquery use a new connector and you now can continue your analysis in bigquery for example by doing a geospatial join with some other data that you already have in bigquery or by using something like bqml to create a model that predicts next year's crop yield so we're really excited that this feature is now generally available and we can't wait to see how you drive sustainability impact with it so next I'll talk about what we've been focusing on in terms of machine learning and interoperability with Earth engine and vertex AI so while machine learning has been in use for remote sensing applications for decades and it's been a core part of what people do in Earth engine for a long time thinking about that integrated view of geospatial analytics on GDP meant we really had a critical opportunity to bring together Earth observation imagery from Earth Engine with the easy ability to train and deploy deep learning models in vertex AI so Earth engine previously had an integration with Cloud AI platform for deep learning and since vertex AI is superseding Cloud AI platform we wanted to both integrate with the new platform and also solve a number of pain points that we heard about the old Earth engine and Cloud AI platform connector so we're super excited to launch our new vertex AI connector to public preview so you can now take your pixels from Earth engine and fire them into all types of models on vertex so the way this works is that you take imagery from Earth engine out to cloud storage you train your model on vertex AI post your model and then do predictions with this new Earth engine and vertex connector we have a technical blog post we have documentation and we have example notebooks that are all posted online for you to check out and get started so we're really excited about both new ways that you can use geospatial data to unlock sustainability insights across our geospatial Cloud products and we have today two of our awesome customers to talk more about their experiences using these products to generate sustainability impacts so first we'll have Martine from Bayer crop science talking about how they've applied Earth engine and bigquery together to power r d changes which is especially important to understand how food production needs change as our climate changes and then we'll have Jeremy from Clear Channel talking about using geospatial insights with AI to drive sustainability Focus impact so without further Ado here is Martin from bear crop science thank you Emily um hi thanks everybody for taking the time I know there's a lot of sessions and um it might be a little bit hard to um hear but I'll do my best um so I'm Martin Mendez Constable I'm the vice president of analytics and data for uh mentioned those around the crop science and what I want to do today is to show you uh how we're using these products this technology is uh likely as child was saying are starting to be more connected across each other right um maybe as a little bit of a background when we talk about vegetables r d in club science one of our core missions is around sustainability deliver digital solutions that influence sustainable products products meaning vegetables ripe tomatoes peppers cucumbers and so on that we sell um globally as well so in order to do that we obviously have a very robust r d pipeline that requires us to generate a lot of data geospatial data in nature most of it um and in a global nature as well um I'm not intending to bore you do that with you know breeding concept here but you you can think about it this way so you have if it's a tomato you have the parents and you're gonna breed them you're gonna come up with a new type of tomato and you're gonna put that seed of a tomato in the hands of Growers all over the world where you're gonna test them right so are those uh better than the seeds that we're selling now are they more sustainable sustainable uh environmentally friendly do they consume more or less water do they have better flavor better shape better colored you know a million things that you're gonna characterize and measure right and a lot of that is is truly sure special data and data that is collected once you put the seeds on the ground in Trials around the trials all over the world thousands of plots that are planted all over the world in about 20 different vegetable crops in an annual basis where you're going to collect more and more data and then hopefully use that data to uh Advance some of those products to the commercial phase right so one of the challenges we had on in any big corporations this is a unique obeyer is how fragmented our ecosystem is in terms of data and when we started talking to Emily and chat about using uh Google Earth engine and bigquery together this was one of our our premises right can we actually stop moving data around so much and um disclaimer our club science Warehouse is built on bigquery already so that was a decision the Bayer Club science made a few years back where we are bringing all our data into a single Warehouse built on top of bigquery so um same as Chad maybe I'll share a little bit of a trip into memory lane for the last 15 20 years I've been um chasing down the product team to understand how we can use Google retention more connected to Google Earth engine right so how can we integrate this data now that we have the connector that Emily uh presented we've been using it already as part of this project and that will advance our mission uh greatly right we will stop basically moving data from all over the place we have now the native integration between Google retention and bigquery and then we can use bigquery ML and advanced analytics on top of both um so in a way you can think about the data catalog that Chad and Emily presented and we definitely like to leverage all the data source data layers that are there already um and in conjunction with um bigquery and Google attention right so we run all our models train them validate them and then using the Google Earth engine visualization options we can enable our end users breeders members of our testing teams all over the world to see the result of this analytical models right so this isn't just you know our stock in our IDs or like that or models or whatever is people actually seeing the insights coming out of those models in this case the mo the map that you see there on that side basically a representation of different environments within California we have used basically all the environmental layers that are available in Google retention and we have brought all our data from bigquery and we have a split basically different environments according to how suitable they are for multiple different crops let's say tomatoes peppers cucumbers and so on Now by using these platforms to charge points that are scalable you can do that real time globally right so no longer you can do it in California but you can do it across the globe on every single country where we are doing business where we're testing this vegetable seeds um am I doing so you can imagine the advantages it used to be very time consuming and very paper based right now It's All Digital mom on top of that there's a lot of different features that we like about Google retention I'll just mention one of them which is data Fusion capability in the days of uh you know 10 10 years ago for maybe some of you that are as old as me you'll remember having to put all the data together and start to agree that to some sort of common cell size and then creating all these derivatives of data storing them and then coming up with an output of it right so that that's no longer needed because with this data Fusion feature now we can do real time on the flight greeting we don't we don't end up creating so much more data that nobody's going to use and this actually takes two lines of code in Google retention you used to take you know a data scientist hours and hours and hours of coding and you end up creating a lot of data that was going to waste anyway now with a handful of lines of code we can do this real time and have this output all um we like to use the Uber concept of a hexagon grids so we like to use the H3 indices to um to grid all these data to a common a common cell size now let's look at it how the impacts our business right because this isn't just r d for the sake of the doing data science but we do want to end up impacting our business so we are we have a very robust number one pipeline in in the world when it comes to vegetables r d we had to create more products that Growers would want whether it's better tasting tomatoes or you know more sustainable um bell peppers because they use less water or they're more resistant to a particular Pest and disease you know we have to keep advance in our pipeline right putting all these things together now we have the advantages of the data catalog the processing the modeling and all this visualization now this colors that you see on this map of Italy are the same concept of environmental clustering as we call it applied to a very different geography before it was California now it's Italy now also you can look across countries right so you can see what tomatoes are better suited for Italy versus California and maybe there's a particular cultivar or hybrid that applies to both that can perform very well in both that's not always the case but just to give you an example now and you can run it like I said before across every single country in the world very very fast then you can influence Your shermoplasm Design right so in the example you have cauliflower you know some people like white cauliflower yellow cauliflower purple cauliflower a smaller head a bigger hair you can use all that data that is collected across all the trials that we're running and inform basically our pipeline Advanced products that are not just more sustainable but the consumers are seeking more right and ultimately you can also use digital phenotype in all of these Technologies put together influence where we deploy sensors on the field on our r d trials do not just not look at things and access color but now measure color with sensors on the ground right whether it's a UAV or a sensor mounted on on a small robot inside a greenhouse or an open field trial you can put it all together and you can mine all this data in ways that we could never dream of I'm gonna stop there and I'm gonna give now um the word to share me so he can walk you through how they are implementing Google Earth engine and bigquery thank you hi everyone can you hear me hi everyone my name is Jeremy Flynn I'm the senior vice president of data products and strategy at Clear Channel outdoor where I oversee audience and measurement solutions to run effective out of home advertising campaigns so yes you get to hear from a billboard guy out of cloud conference at Clear Channel I build audience solutions that match the right location with the right audiences to build the best performing out of home audience plans possible you probably have seen our ads no matter how you got to this conference whether you landed at SFO or Oakland or San Jose where you drove up the 101 or outside the Moscone Center where you walked in today we have thousands of advertising ad units in the city of San Francisco and my hope for today is you ever you never not notice them after this session right now our industry the adipo industry is making deep commitments towards becoming more sustainable whether that is putting solar panels on top of our bus shelter ad units to reduce our Reliance on the grid or making the slow conversion to a hundred percent digital Billboards we're deeply committed to reducing the time effort and materials that go into getting an out of home ad campaign live and and for all of you that might mean the world becoming a little bit more like Minority Report or Blade Runner but hopefully with a happier ending today we're starting to see much more of an intersection between Brands promoting a sustainable message and the marketing and business outcomes they're looking to drive at the board level that Chad referenced earlier CMOS are in those board meetings and they're being asked to promote the sustainable efforts of the brands that they work for and I believe that effective advertising when done right through the power of influence could actually build a more sustainable world this happens when you can drive sustainable consumer behavior and action over a long period of time um I think a lot of you are about to remember this but I'm a child of the 90s and I really remember that got milk advertisement that I saw literally all over the place on TV on billboards in my school library and now as adults I I've realized how powerful that simple message was in driving towards mass action and how we need to start doing that today with respect to sustainability and climate action so for advertising to be most effective that means aligning the creative or advertising message with the right place the right time the Right audience and that right place normally is the right location so an effective advertising strategy um means that like we have an ability to understand where to place the right ads at the moment of families deciding to make a more sustainable choice we also do this for governments as Chad mentioned earlier it's hot we actually partner with a lot of local governments towns municipalities to actually promote sustainable messages and call to actions during times of natural disasters whether that be advising people to evacuate during floods or during upcoming heat waves we are there and at the ready because we exist in the Public Square on the other hand with respect to Consumer purchase we actually work with Brands to promote the sustainable products that they're bringing to Market and increasingly we see a statistically significant increase in household purchases of these more sustainable products that can mean meatless uh meatless uh Brands where we promote them right outside of the grocery stores where they could be purchased or more sustainable Automotive vehicles that we'd all love to see more of out on the roads today at Clear Channel we productize this geospatial data through our radar analytics Suite one to modernize literally the oldest advertising medium in the world and two to make the world more sustainable that might seem like a Bitty Pretty bold Proclaim from a billboard guy but I'll tell you how we're doing with that data one of the products in our suite is our radar view product which I'm about to demo and we this is an insights platform that is built on top of Google bigquery Google vertex and one of our our strategic Partners cardo that that combines demographic data behavioral data audience location location and proximity data to build the best performing out of home campaigns and something that Chad also mentioned earlier speed actually drives performance enablement not just for the successful out of home campaigns that we're building but to truly revolutionize the business practices or the workflows of our sales teams trying to sell those out of home ad campaigns but a major problem with all of that data is that there are millions and millions of possible combinations and the options to build a plan then become Limitless and so that creates bottlenecks it creates bottlenecks for people who we want to use our products to turn around media plans that we want to be selling to Brands and advertisers so that's where Google and cardo come in using their technology we're able to create even more self-service based campaigns where we've introduced conversational artificial intelligence and natural language of processing to actually increase the speed and turnaround times of the types of media plans that we're trying to create this combination of Google bigquery with vertex AI helps us produce actionable location intelligence and audience insights that ultimately helps us build and and drive more successful out of AD campaigns let's uh look at it in practice so I'm gonna take a second to run the video by adding Google bigquery and vertex running native within our radar View application we can produce geospatial Analytics where we can first show all of our Billboards on the map and then we can actually add in an audience in this case we want to drive an audience uh who might be in the market for buying an electric vehicle stand out to the side and then for today's purposes we also can isolate that exact media plan in the city of San Francisco for hyper location-based targeting to reach consumers in the city who might want to change out their car for a more sustainable brand model and so as this runs you could see the locations or the ad units that we could sell updating in real time and finally we're able to add in a final proximity filter which is the proximity of our ad units to contextually relevant locations in this case electric charging stations that actually exist in the city of San Francisco and here there are three major ones and as we zoom in on the map in a minute you'll actually start to see those locations pop up so now we understand the ad units in context to the proximity of those Vehicles understanding the audience that are most likely to actually purchase the product that we're trying to advertise okay that's the video and I was taught how to flip to the next slide here we go you saw how fast that works and again speed is a performance enabler but by aligning the stack and having our radar view products sit on top of cardo sitting on top of Google vertex Ai and sitting on top of Google bigquery we are processing hundreds of millions of records of data to surface the actionable audience insights to build the most performative out of home plan and before we actually migrated to bigquery um we were working with multiple managed cloud service providers and there was a lot of timeouts Maps weren't rendering and our sales team were not using data on their media campaigns so we had weaker campaigns after that we actually saw a significant reduction in the processing time like the example that we saw earlier and now with over 300 sales people on Clear Channel who've been selling Billboards longer than I've been alive across 30 markets we're helping advertisers large and small locally nationally internationally actually run higher more performative campaigns and hopefully as those campaigns become more sustainable in nature and those calls whether they be calls to action or more sustainable products being brought to Market like Martin is doing we're helping to use geospatial data to help make the world a better place even if it's for advertising so now I'd like to invite Martin uh oh Chad chat up first and a few others gotta leave it to the advertising guy to have the perfect tagline for his uh for his talk there so we're gonna we're gonna switch to the wrap up here and then we'll have time looks like we'll have time for a couple of questions um so I want to I just wanted to uh say that for Emily and me right from the product side the googlers who are involved in sustainability either from the product management engineering go to market sales teams our jobs we view our jobs to deliver exceedingly performed and exceedingly capable products that interoperate exceedingly well together so that you all can go build you so that you all can go grapple with the sustainability and geospatial problems that you need to solve what's really cool about that is that stuff that was impossible three four years ago is now totally possible so I encourage you all to go back and rethink some of the things that you've been you know have that you've had on the shelf that you haven't been able to do because the tooling will allow that to work now um and so as you grapple with these problems I'm really really excited to see what Solutions you go ahead and develop um and then speaking to a couple of Solutions in particular foreign side we're so excited that so many of these Solutions um especially a couple of them up here really demonstrate the incredible sustainability impact that customers are having with Earth engine um that's really driving change across these critical problems of climate change mitigation and climate resilience all right so we are pretty much at the end grab the app let us know what you thought thank you very much foreign

2023-12-23 16:37

Other news