NASA ARSET NASA ATL24 Bathymetric Data for Coastal and Near-Shore Applications

0:00 Speaker 1: Thank you for joining today's r SET training NASA satellite laser altimetry for coastal and nearshore bathymetry. I'm Sean McCartney from NASA Goddard Spaceflight Center in Maryland, serving as your RSET trainer. I'm excited to begin this first part of our comprehensive two part satellite bathymetry training with all of you. For newcomers to our SET trainings, the following slides present a short program introduction. The Applied Remote Sensing Training Program or r SET, part of NASA's Earth Action Capacity Building program, offers free, accessible training on remote sensing technologies, methods, and applications.

0:41 Speaker 1: Arset's mission is to increase Earth science, remote sensing and model data use in decision making by providing training for professionals in the public and private sectors, environmental managers, and policymakers. On the right of the slide you can see all the different thematic areas in which we conduct trainings. Our SET offers both online and in person training sessions designed for beginners through advanced users. All trainings are provided at no cost through live instruction, instructor led sessions, or self paced formats like the fundamentals of remote sensing course on arset's Learning management system.

1:20 Speaker 1: Materials are available in multiple languages and can be freely used and adapted for educational purposes. Users of our set, methods and data should acknowledge the NASA applatter Remote Sensing training program. Additional information is available on our website. The following slides provide an overview of the two part webinar series NASA satellite laser altimetry for coastal and nearshore bathymetry. So why would somebody want to take this training? Well, nearly eighty percent of the world's oceans are unexplored and unmapped.

1:55 Speaker 1: Successful coastal management depends on monitoring, and coastal bathymetry is essential for navigational hazards, for vessel operations, tidal modeling and prediction, tsunami risk assessment and forecasting, underwater cultural heritage preservation, and environmental change monitoring. Satellite derived bathymetry enhances capacity for collecting high resolution, accurate depth data supporting varied applications. By the end of this two part training, participants will be able to identify NASA bathymetry data used for global, coastal and nearshore bathymetry mapping for risk reduction relating to shipping and navigation.

2:38 Speaker 1: Identify the applications and limitations of ice AT two bathymetry data ATL twenty four for coastal and nearshore bathymetry mapping, Plot and download i AT two bathymetry data using slide roll web client, and analyze i AT two bathymetry data using slide roll Python climate Cli client. The prerequisites for the training are two are set courses Fundamentals of Remote Sensing course and mapping and monitoring lakes and reservoirs with satellite observations. From December second to December fourth, there will be two one and a half hour sessions which will include presentations, demonstrations, and question and answer sessions.

3:22 Speaker 1: All materials and recordings for each session will be available from the training web page. If you are not able to attend one part, a recording will be made available within forty eight hours on the ARSET training web page. Homework opens on December fourth and will be due on December thirty first. You will be able to access the homework from the training page on Thursday, December fourth. A certificate of completion will be awarded to those who attend all live sessions and complete the homework assignment by the given due date.

3:57 Speaker 1: Part one of the training is focused on NASA ei AT twenty four bathymetric data for coastal and nearshore applications. The objectives of Part one of the training are as follows. By the end of today, participants will be able to identify nasabathymetry data used for global coastal and nearshore bethymetry mapping for risk reduction relating to shipping and navigation. Identify the applications and limitations of i AT two bethymetry data ATL twenty four for coastal and nearshore bethymetry mapping and plot, and download IAT two bethymetry data using slider roll web client.

4:37 Speaker 1: Please put your questions in the questions box and we will address them at the end of the webinar. Feel free to answer your questions as we go. We will try to get to all the questions during the Q and A session. After the webinar, the remainder of the questions will be answered in the Q and A doc which will be posted to the training website about one week after today's training. It is not my pleasure to introduce the guest trainers for Today's webinar Doctor Lauria Magruder, doctor Dian Fritz, and Joseph Paul Swinsky.

5:07 Speaker 1: Laura Magruder serves as Associate professor in Aerospace Engineering at the University of Texas at Austin, with previous roles at Jet Propulsion Laboratory Ann JOHNS. Hopkins Applied Physics Laboratory. Following nine years of science team leadership on NASA's i SAT two mission, encompassing mission development, implementation, guidance, and operational instrument pointing strategies, Magruder has transitioned to leadership positions on additional remote sensing science and instrument teams. Dian Fritz holds a position of Associate Scientist at the National Snow and Ice Data Center NSIDC, located in Boulder, Colorado, and serves as a Remote sensing educator at the University of Colorado.

5:50 Speaker 1: Her academic credentials include a doctorate in geological Sciences. J. P. Swinskey is a twenty four year NASA Goddard software developer who spent most of his career creating flight software for spacecraft and science instruments. He led flight software development for the Atlas instrument from twenty ten to twenty sixteen. In twenty eighteen he transitioned to cloud based data processing solutions, and in twenty twenty co developed slide rule with the ice AT two Project Scientist and University of Washington, which is a public web service providing fast cloud access to ice AT two data.

6:26 Speaker 1: Laurie over to you.

6:31 Speaker 2: Thanks Sean. I'm so happy to be here and talk about IAT to the new bathymetry product that has just come out earlier this year. I can honestly say that I SAT too is my favorite satellite, and I worked on it for many, many years, and so this is just a really exciting time for me to be here and provide an overview of the mission and introduction to the ATL twenty four Coastal and nearshore bathymmetry product. So let's start with the mission overview. The Ice Cloud and Land Elevation Satellite two is an Earth observing laser altimetry mission that launched in twenty eighteen, so we've been on orbit for a little more than seven years.

7:16 Speaker 2: It completed its prime mission and successfully satisfied all its science requirements in twenty twenty one, but all operations are still nominal and it's continuing to give us amazing data. I SAT two carries a single instrument called ATLAS, the Advanced Topographic Laser Altimeter System. ATLAS is a photon counting laser altimeter, meaning it's sensitive to single photon reflections from the Earth's surface. This state of the art technology to date has produced over two trillion laser pulses. Turns out, that's a whole lot more than the two billion from the original IAT mission in two thousand and three to two thousand and nine.

7:57 Speaker 2: Right now, the prediction is that IAT two has all consumables on board, meaning laser energy and fuel to continue operating until twenty thirty three. So ATLAS uses a neodymium doped yttrium orthovanidate laser to produce six individual beams that's configured into three beam pairs that you can see in the schematic here on the slide. Each pair contains a weak and a strong beam with respect to laser energy at a one to four energy ratio. The pair components are separated by ninety meters in the cross track direction, and the pair to pair separation is three point three kilometers.

8:46 Speaker 2: Each of the six beams provides an elevation profile of the illuminated surface with around a twelve meter diameter footprint, so these are just creating pencil lines of elevation across the Earth's surface. Atlas output wavelength is a five to thirty two nanometer which is in the visible green part of the spectrum, and a pulse width of one nanoseconds with one nanosecond which speaks to the high vertical resolution of the system. Photon counting altimetry allows some unique advantages in terms of data collection.

9:22 Speaker 2: So since the receiver is sensitive to single photons, it allows the system to operate at a lower energy level compared to previous or other types of light our technology. In turn, that means that the system can pulse at a faster repetition rate, and in this case, Atlas has a ten killer repetition rate that at the altitude, which is about a five hundred kilometer average altitude, it can produce a laser footprint or puts down a laser footprint every seventy centimeters in the long track direction.

9:55 Speaker 2: The final thing I'll mention here in terms of the collection strategy is the fact that the orbital plane is inclined at ninety two degrees from the equatorial plane, which means that the coverage for I set two goes from eighty eight degrees south latitude to eighty eight degrees north latitude. So I think it's important to get oriented with respect to what the IST two data looks like. So I'm showing you an example elevation profile over West Greenland west Central Greenland. The horizontal axis is latitude and the vertical axis is the elevation.

10:31 Speaker 2: So each of these black dots indicates the Atlas detection a detected photon. By Atlas, every photon has its own latitude, longitude, and elevation, So the challenge with photon counting is the need to separate the signal photons from the noise photons. So although in this case it's pretty easy for us to look at the profile and identify the surface which is this dense black line, it's a lot more challenging to automate the process of signal finding. Signal finding also differs greatly with surface reflectants, surface type, and surface feature length scales.

11:07 Speaker 2: So I set two provides data products for specific surface types that have been produced with unique algorithms optimized for the surface characteristics. So when we start to explore these data over the last seven years on orbit. You can see through aggregation all of the impact that I set to has had with respect to the specific data products that provides. We see sea ice measurements both in the Southern Ocean and the Arctic. We see the land ice and mass change for Greenland and Antarctica. And next comes the brown or the terrain, the land surface elevations, and then the vegetation canopy heights, which will be in green here in just a minute.

11:52 Speaker 2: Additionally, we have data products over the ocean and inland water surfaces. However, what was somewhat unexpected was the ability of YET two to measure beneath the water surface or measure underwater topography, which is also called bathymetry. It's common knowledge that the green laser wavelength can penetrate water, but it was not fully anticipated that a photon counting system from three hundred miles in space would perform as well as it does. You can see here an example of the bathymetry that we've been seeing for quite some time.

12:32 Speaker 2: It's a sea surface which is this dark black line horizontal on the elevation profile, and then underneath is the underwater topography or bothymmetry. In this case we're at the Great Barrier Reef. Studies of this performance or the capability of bothymmetry has happened in various locations, and ultimately we know that ATLAS can measure bathymetry up to forty meters in depth for clear water conditions. Certainly, it varies depending on where you are geographically. However, until this year, we've not had a dedicated withymmetry data product from the mission.

13:11 Speaker 2: I thought it'd be important to just show you a snapshot of the data products for iat IO. The level zero and Level one products are telemetry and converted telemetry products. The level two products are ATLO three, which is the geolocated point cloud, and then ATILO four, which are normalized backscatter profiles. ATLO three is the input to all of the Level three A products, and those are those blue boxes horizontal or in the middle of the slide. These are a long track products produced from these optimized algorithms that I mentioned previously.

13:53 Speaker 2: These Level three A blue boxes include those data we saw from Land ice Heights. C ice cis free board land and vegetation inland, water and ocean. But then you can see this this newly added box here ATIL twenty four, which is the along track near shore coastal bathymetry product which I'm happy to introduce to you today as an introduction. ATL twenty four was developed by the collaborative team. My team here at University of Texas at Austin Center for Space Research Jeff Perry and Matthew Holwell, and our teammates at Oregon State University Chris Parrish, Keana Keef, JP Swinsky at slide ROLL and all of the mission or Project Science Office partners Jeff Lee, Tom Newman, Dennis Felixon.

14:48 Speaker 2: We're all working together and released a TAIL twenty four on April first, earlier this year. Ultimately, ATAIL twenty four is designed to provide the community signal finding and photon classification for sea surface and seafloor along global coastlines. We had the goal of creating a robust algorithm that could be successful at the global scale, provide appropriate corrections to the data and confidence values of the signal classifications, and then also a per point uncertainty. So these were the goals of ATAIL twenty four when we first initiated development, which was a couple of years ago.

15:30 Speaker 2: What I'd like to emphasize here is the significant challenge with developing a single algorithmic approach that could satisfy all of those product goals. However, to meet this challenge, we use a novel machine learning approach called an ensemble. The ensemble creates a robust capability in the environment that is highly variable when you consider that everywhere there's a different bottom type, a different surface reflection, changing water turbidity levels, and sea surface conditions. So AHL twenty four also provides a confidence value for each photon that indicates an estimate of the classification fidelity, and then additionally we correct this with an index of refraction correction.

16:19 Speaker 2: The inputs of the ensemble are classification predictions from several base algorithms, and when I say classification predictions, it is that photon a sea surface, is it the theymetry or is it unknown right now? The ensemble gathers input from six distinct methods, and since the errors of each classification based algorithm are uncorrelated, the ensemble is able to predict the symmetry from the combination of predictions from the base algorithm input. So this creates a scenario of the whole is greater than the sum of the parts, meaning that the ensemble performs at least as well as the best performing algorithm for a given location.

17:02 Speaker 2: Ultimately, we're taking advantage of the fact that the single that a single signal finding algorithm cannot perform equally everywhere. So what I want to point out here in the upper right is just a granular a transect that we pulled out, and it shows the different colors just below the area where the turquoise photons are those classified as sea surface and the brighter blue photons are those that the ensemble predicts as the symmetry. And then I'll point out here this plot on the bottom right is each column is the F one score of the base algorithms and the ensemble for predicting sea surface and seafloor.

17:44 Speaker 2: So the ensemble is the column in the middle, and then you have three base algorithms on each side to the left and to the right. F one score is a metric that we often use in combination with a lot of other metrics to understand how good we're doing in terms of accurately classifying the photons as sea surface or c floor. So the green labels are each algorithm of how they predict C surface, and you can see that each of those values is almost they're pretty equal and almost to one, which would be a perfect classification score F one score.

18:21 Speaker 2: But the brown values are the the F one scores for classifying C floor. And you can see here that there's variable in this location. Each algorithm performs differently based on those variations of values. But the ensemble the column in the middle shows the highest F one score for classifying sea surface. And like I said that the ensemble is created to perform at least as well as the highest performing base algorithm. Another key benefit to this ensemble, which is an XG boost, is that it can look at its decision trees and create a probability of classification correctness if you will.

19:06 Speaker 2: And that's what we've turned into a confidence score. This is a really good visualization, and then I'll get to the confidence score of the comparisons of classifications of the base algorithms. So you can see here the top is manually labeled reference data. So this is what we say is the quote ground truth of sea surface and sea floor for this particular profile. Then each profile moving down from the manually labeled reference data is the output. So the top one is the ensemble, and you can see as you go through we have the six input I think we just have five because I couldn't sit in on the slide, but you can see the variation, Like if you look at the bathy pathfinder algorithm that over here towards the right, it gets a little confused and grabs some of the water column and labels it as the symmetry, whereas maybe if we skip down to the media and filter, which is second from the bottom profile, you can see that for whatever reason, that algorithm misses some of those bathymetry points on the right.

20:10 Speaker 2: So each of these algorithms has a different type of error associated with it or capability in correctly signal finding, but we can use all of them in the ensemble to create the best guess of classification. So the confidence as I mentioned, is an output of the XG boost ensemble, and it's kind of internally generated, but it has a lot of variables, and it's also based on this current composition of base algorithms. So what I want to emphasize here is that right now for version one release, when we feel that the confidence level of zero point six or higher is a good way to users can filter the data for high confidence classifications.

20:59 Speaker 2: However, that's based on the input algorithms, as I said, and so next versions might have a different threshold recommended of confidence, so be mindful of that as you look to versions two release excuse me, version one, Release two, and version two that comes out probably next year, that those confidence thresholds are filtering might change. This shows you those C floor classification photon classifications from about point six too point one, and you can see it is just by visually inspecting that it seems pretty accurate and there's very few outliers in terms of what the ensemble is predicting.

21:45 Speaker 2: So in terms of the full A tail twenty four workflow, I have a version here. So let's start in the upper left box that shows the intersection of the ATL of three granules and then our ATL twenty four search mask. The search mask is a global mask that uses a derived retrievability score for probable I SAT to bothymmetry detection, and this is based on water turbidity and maximum depth of a given region. And then we take that mask and we widen vertically and horizontally to ensure that we don't miss anything.

22:26 Speaker 2: And the mask guides the processing of ATL twenty four to just our coastline near shore environments, so moving down along the less left side, we process the data and use a function we created called costnet to make a first pass at identifying water surface photons. This is because we have these six independent algorithms for classifying bathymmetry photons, but some of them need a starting point of where the water surface is, and that's why we do a first approach to water surface classification.

22:59 Speaker 2: Each of these these algorithms use a different approach to signal finding. Some of them are histogramming approach or density evaluations, some of them are machine learning approaches, but again all of these in all of these predictions from these base algorithms are fed into the ensemble. Once the ensemble provides the classification predictions again which is supposed to be the greater than the some of the parts, we perform some blunder detection and outlier removal, which are if we go down the gray boxes in the middle of the flow chart, and then we do a final sea surface finding and provide a correction for the index of refraction of water for any of the photons that occur beneath the sea surface.

23:47 Speaker 2: The uncertainty calculation is calculated from a subaqueous uncertain D lookup table, which is the parallelogram at the bottom, and then it all so uses the coefficient of diffusion diffusion the diffuse coefficient excuse me from veers the KD four ninety value. So then that's our uncertainty value on per point level. And then our data product is generated with a variety of parameters. Some of them are listed here on the far right C floor, ellipsoid height, C floor orthometric height, and uncertainty estimates along with the confidence value.

24:30 Speaker 2: And I should also mention for the refraction correction. As you can see up in this top parallelogram, we do have our own refractive index layer that's created from global level temperature and salinity values to create the most accurate correction in time and space. So we kicked off the full production of a TAIL twenty four back in February. It took about thirteen days to produce twenty seven point six terabytes of data. We processed this in AWS.

25:08 Speaker 3: I thought these.

25:09 Speaker 2: Metrics were pretty interesting to say that a total linear coverage of bathymmetry was thirteen point seven million kilometers and then you can see here the number of our symmetry points, which we do follow the ASPRS convention of classification, so ourthymmetry is class forty and our c surface is class forty one. We collected a little over seven billion bathymetry points, and if we just filtered those by our confidence value of point six, then we had more than a half identified as high confidence pathymmetry.

25:43 Speaker 2: And you can see here in the figure the just the coverage in yellow of where we have but symmetry within the product. We chose eight sites that had that we had pretty consistent reference data to compare to to understand what how accurate not the classification confidences, but how accurate the actual bathymmetry elevations were. And you can see here this is on the left column, are the accuracy test results for all of the symmetry points that those eight sites. And you can say we have an RMS of point six eight meters and then if we threshold those values for high confidence on the right plots showing a reduction in RMS of down to forty three centimeters, and so it does improve the overall quality of the symmetry elevations by just using high confidence, but you also have to weigh the impact of removing some of those symmetry points that could very well be bothymmetry, but not all of the algorithm's predictions supported that.

27:00 Speaker 2: One of the things that is I'm sure would be interesting. Are some strengths and limitations to the current version of a TAIL twenty four, And I just want to emphasize that the ATAIL twenty four is a global coverage of the Earth's coastlines, and we have yet to have a space based laser bathymetry system that provides the wealth of data that I set to is providing. I think our approach for signal finding and classification is pretty robust because it can leverage the advantage of each individual algorithm or base algorithm where it works well and we don't lose any quality of the output data and signal finding.

27:45 Speaker 2: It's accessible in a variety of ways. You can download the data from several methods, which sure is part of this training. The architecture of the algorithm is also configured for continual improvement, meaning that we can right now we have six base algorithms, but we could up that to twenty four base algorithms, and it's all set up the structure is set up to continue to add algorithms as methods are identified and implemented. And then also the accuracy of this product is pretty great, and it has been proven suitable for calibration and validation of spectral satellite derived but symmetry in most cases.

28:27 Speaker 2: Some of the limitations, though you can mention here, are that some of the conditions do have results in variable data quality, as you can imagine, and then the current uncertainty estimates are overly optimistic, but we're still working on refining that and making them a little bit more realistic. The data exists only on the coastlines and coincident with the current search masks, so the ATIL twenty four algorithm is not run anywhere except the intersection of the ATL three grain rules and the search mask.

28:59 Speaker 2: And then the photon classification. Conflicts between a TAIL twenty four sea surface and AT eight terrain surface do exist. So if you are interested in seamless you know topobathy profiles along our coastlines and you pull in at eight terrain heights, there's some confliction in the in the photon classification, so it's not there's not that continuity yet in those two along track products,

29:30 Speaker 2: and I wanted to point out a few resources that might be helpful in using a TIL twenty four. Understanding more about a TEL twenty four. The first two bullets are papers that were recently published this year in Earth and Space Science. The first one perish at All. It provides more detail about the accuracy analysis that those eight sites that I was showing earlier. And then the second one, my Magruder at All, is in depth description the base algorithms and the ensemble and talks a little bit more about the metrics associated with understanding the performance of the classification.

30:09 Speaker 2: And then the algorithm. Theoretical basis document is the last bullet, and that gives you even more information about the algorithm. It includes the refraction correction approach. It also includes uncertainty evaluation and all of the other details about the product.

30:30 Speaker 2: And I think that's all I have. I'm so happy to be here today with you, and I will throw it back over to Sean.

30:42 Speaker 1: Thank you Laurie for the excellent overview of I two and the ATL twenty four bathymetry product. We will now transition to Diane Fritz from the National Snow and Ice Data Center to present on ATL twenty four data resources and access. Diane over to you.

31:00 Speaker 4: Hi.

31:01 Speaker 5: I'm Diane and I work at the National Snow and Ice Data Center NSIDC, and I am going to do a quick segment to show you ATL twenty four data, that bathymetry product and the resources and access that we have for you at the DECK or Distributive Active Archives Center. So what is the role of NSIDC. We preserve many, many data sets, including all of the products from i SAT too. Although we are the National Snow and Ice Data Center focused on the criosphere. You'll notice from Professor mcgrider's previous video that she showed I SAT two is a global data set, all of the ones that get derived from that mission, and so we have all of those products at NSIDC.

31:58 Speaker 5: We have a lot of products from other satellite missions, airborne surveys, field observations, all sorts of different things related to the Christ fear and we are one of eleven DAX and you can see on the map here where some of those are located. We are in Boulder, Colorado, but we are accessible all over the globe by emailing us at NSIDC. At NSIDC dot org. Data is a free for anyone. You just need an Earth Data log in to be able to access it. But it's free to create that as well, and it's helpful to be able to work with geotips, net CDF, HDF files, thing other kinds of science files like that.

32:49 Speaker 5: But we have user guides also to help you, and my job and the User Services office at NSIDC is to help anybody use the data that we host at the deck. So this is an example of one of our data set landing pages. In particular, this is the ATL twenty four Bathymetry Product landing page, and I just want to show you some useful things that if you go here, you'll be able to find a user guide that talks a little bit more about what the data is. We have a citation so you can easily give credit to the people that produce this data, but also know who they are and be able to look up the papers and other research that they've done.

33:40 Speaker 5: And we also have a subscribe possibility here, so if you want to know more about a particular data set, you can sign up for email updates of that data. So if there is an update, something that is a new parameter or something like that gets added to a new version, you can get a notification about that. But on our data set landing pages. In addition to these icons at the top, we also have a menu for being able to find things on the side that right menu that you see there, So this gets you access tools, documentation, and help articles.

34:27 Speaker 5: What I have on this screen right here is an example of two different access tools that we have for ATL twenty four. One is the NASA Earth Data Search, which I'm going to do a quick little demo of how you can use that to find ATL twenty four data and actually subset it. There's also slide rule Earth as one of the tools, and many more, and you'll get to hear more about Earth later. We also have documentation, these tech documents if you want to dive further into the algorithms that we're created to get the bathymetry product out, data dictionaries and user guides, and then we also have help articles that are going to help you maybe just understand how to work with a certain kind of platform with HDF files for example, or know a little bit more about the data.

35:27 Speaker 5: So all of these resources are available from that data Set landing page. Now, if I go to the tool NASA Earth data search. I will see some kind of visualization like this, and I'll demonstrate this. You can search for data by mission, so I can put in IAT too and see what's available. I can look for keywords. If I know the data set short name ATL twenty four, I can put that in and look for it. So this is a way to access data from all of the NASA docks, not just NSIDC.

36:12 Speaker 5: There are customization services for select data sets in here as well, and so I'm going to show you what it's like to just pull as small spatial chunk of an ATL twenty four file and visualize that. So this is a picture of panoply. It is a nice kind of visual browse platform and it does some more things as well. But NASA developed this and it's a really nice way to quickly look at a ATL twenty four granule and see what is available in it. As Professor mcgrader mentioned, there are classes of photon classifications in this data set, and that variable is class underscore pH for photon, and forty represents a bathymetry photon and forty one represents those sea surface photons.

37:14 Speaker 5: So on the left here and the small side, I've got a full granule download, and I've zoomed in and changed the scale so that forty the bathymetry photons show up in yellow. And you can see that in here that we have some pretty good returns in the area of the Bahamas for bathymetry. So I'm going to show a little demo using NASA Earth Data Search to grab one of these ATIL twenty four granules near the Bahamas and show you what it's like when we open it in paniply. So I'm going to dive over to my browser now, and this is the dataset landing page for ATL twenty four.

38:09 Speaker 5: The URL is an SIDC dot org slash data slash ATL twenty four. You don't have to type in the rest, the version will self populate. But if I scroll down, you'll see a map that shows that the ATL twenty four coverage is global, and you'll see these different data access tools that you can get slide rules down there. We're going to go to NASA Earth Data Search and click on this and it's going to take us to a page already with ATL twenty four loaded into this. Now I can hover over these different granules.

38:53 Speaker 5: I'm going to click on this one and it'll take me to a place where this one exists. So this is that one that I showed on the slide deck. I can do spatial search here. I can also do temporal search with these controls. I'm going to just draw a little box and grab a little bit of this granule in this upper part. But in doing that spatial search constraint, I still have all of these other granules here as well, so you can see that I've I still have two hundred and twelve that go through this over the course of the mission.

39:37 Speaker 5: I'm for the purpose of this demo, I am just going to copy this granual name and search by it as well and reduce it down to one. But you can use wild cards in here, or you can do a temporal search to narrow it down a little bit more.

39:57 Speaker 5: So with this, I'm going to say I want to download all which is down to my one granule at this point, and it's going to take me to this platform where it's going to give me an option to say do you want to just download this data straight or do you want to customize it? And Harmony is the tool that we have for doing spatial and temporal subsetting on some IAT to data sets. So I'm going to choose a service. The only option here for this data set is the Harmony trajectory subsetter, and once I've selected that, I can scroll down and see that had I had more data sets or a temporal constraint, I could use that, but I still have just that one granule.

40:55 Speaker 5: So I have my spatial subsetting here that looks good to me. I'm going to clear done, go to click download data, and it's going to start putting in that order for me. With that, taking that bounding box and actually trimming my granule just to that area. I can go to my order status and see how Harmony is doing, and it looks like that job it's going to look at that job code it was already successful. This takes a little bit of a lag time to show that, but that's where you can go and get that link directly.

41:40 Speaker 5: It can also notify you when everything is finished processing and you can pull data from here. So I'm going to hop over to panoply. I've already downloaded a subsetted file before this, and this is what it looks like when you open that and you can see the three beams here and when I expand one of those from the This is the ATL twenty four product, and we have our class of photons. We have heights of different kinds and flags and different kinds of information, and then I can plot these and if I'm looking at my class of photons, and again I've changed my scale here in panoply from zero to forty one to thirty nine to forty one, just so I can see that bathymetry photons in the side, so I can see that I've got some good response there.

42:45 Speaker 5: I can also bring up the height data as well. And so if you recalled the way that I grabbed that data, there was a little bit of an island going across here, and you can see some of that height shown in this area. So that's how panoply can help you inspect these HDF files pretty quickly. I'm going back to our data order and see that it's here. I can actually click on this link and get this subseted file directly downloaded from clicking here. But if I have a lot more granules in here, I can use this Earth Data download and it can pull everything in one download for me.

43:40 Speaker 5: This is a different software you need to put on your computer to have that work, but it's really quite simple and we are happy to help out with that. So I have one more slide to show, and that's just some resources that I've provided. This is a link to the ATL twenty four data set landing page, a way to get the Earth Data log in for NASA data in case you don't have that already. Here's a direct link to NASA's Earth Data search for all the decks the Panopolice software from NASA, and we have some help articles the Earth Data Search Guide.

44:24 Speaker 5: This will tell you how to use Harmony subsetting a little bit, but we also have Harmony in as an API as well, and there is a Python library called Harmony Pi that you can use as well if you're doing a programmatic subsetting in a Python notebook for example. And then here's also a link to slide rule Earth. But other people will be talking about that. So I hope this helped you get a good orientation to resources that we have for supporting ATL twenty four and in the User Services Office. We are here to help you, so feel free to email us at NSIDC at NSIDC dot org.

45:13 Speaker 1: Thank you, Diane. Next we'll hear from JP Swinsky, who will present on the slide rule web service for plotting and downloading ATL twenty four data. JP over to you.

45:26 Speaker 3: Okay, thanks for that introduction.

45:29 Speaker 6: I'm going to be talking about slide rule Earth, which is a collaboration between NASA, Goddard and University of Washington.

45:38 Speaker 3: So what is slide rule.

45:40 Speaker 6: Slide rule is a public web service that has rest like APIs for processing science data. The goal is to provide researchers with low latency access to one demand data products, and we provide those data products using parameters supplied at the time of the request, so it's like a real time data access system. The entire system runs in Amazon Web Services US West two and we have access very efficient access to all of NASA's Earth data products stored in US West two like ISA two, Jedi, Lansat, and a growing list of others.

46:20 Speaker 6: As a team, we support multiple clients for accessing slide Rule. The two main clients are a Python web client and our Python client and our web client which I'll be talking about today, but we also support no JS and for those power users, you can access our system with CURL. Here's on the screen you see it just a cartoon of what the system looks like at a very high level, the users typically interact with slide rule using either a Jupiter notebook or their web browser.

46:56 Speaker 3: If they're using the web client.

46:58 Speaker 6: We run a bunch of under the domain slide earth dot io, and then behind those servers we have access to things like NASA's Common Metadata Repository, which is their index for the NASA's Earth data products, as well as the Earth Data Cloud that I mentioned before. You can find us more information about us on our website, which is slide wile worth do I Owe. You can point your browser to that. You're always free to contact me. I have my contact information there and for developers, please look us up on GitHub.

47:32 Speaker 6: We're completely open source under a BSD three clause license.

47:38 Speaker 3: So why did we create slide role.

47:42 Speaker 7: So the.

47:44 Speaker 6: Idea is that we're hopefully improving the way data is accessed in the cloud. So here's just a picture of what it's like to access data in the cloud without a service. So here you have different s three buckets holding different data products like the ATL three data from I Set two, the Jedi data, the Blue Topo data, the Harmonized lands At Sentinel data, and it immediately becomes evident that if you're working from a single Jupiter Hub instance or your laptop, it is very challenging to work with these data sets at a global scale.

48:22 Speaker 6: There's just too much data to work with. You also need to know the format of all these different data sets. For instance, if you're trying to correlate ATLO six ice surfaces to the Arctic dem you need to be able to figure out how to get the acquisition date from the Arctic dem rasters, and that's going to be different than the way you're going to get an acquisition date from HLS, and different then it's going to be from Blue Topo and so forth. And then lastly, there's going to be common operations that you're going to want to do in the data that are going to be very compute intensive.

49:01 Speaker 6: For instance, if you're working with a photon cloud like we are with ATL twenty four, it is very compute intensive to take that photon cloud and resolve it to a single elevation over a certain aggregated area.

49:17 Speaker 3: So that is why.

49:18 Speaker 6: We came up with slide rule as a service for accessing this data. We provide optimized direct access to the raw data. So if you want to just read Atail twenty four or blue topo. You can use our services and we will very efficiently in a parallel way, read the data as fast as possible and give it to you directly. But we also supply efficient subsetting services. So if you want to take the Atail twenty four data and look at it just in an area of interest, you can provide that polygon to slide Rule and it will go out and it'll pull all the different grands and in a parallel way provide them to you in one single response.

50:05 Speaker 3: And then we do on demand processing right next to the data.

50:08 Speaker 6: So we are running in US West two, right next to the buckets that are stored in that same data center, and we can produce customized data products on demand using parameters supplied by your request. So that's in a nutshell what slide rule is. And now I'm going to transition to talk about to give a demonstration of our web client. So this is pretty new. It's it just came out this year, and this is one of the ways, maybe the easiest way to accessing slide Rule, which is doing so from your web browser.

50:47 Speaker 6: So this is the this is the web client. You can get to it going to client, do slide rule or do io, and you can also find links to it. You can also find links to it from our main web page.

51:08 Speaker 3: So when you.

51:08 Speaker 6: Navigate to the web plan, this is the screen that you'll see. And the first thing you're going to want to do for accessing the ATL twenty four Coastal Pathymetry data is you're going to want to go over here and select the ice AD to Coastal Bathymetry selection box. What that does is that puts a search mask up onto overlays it onto the map, showing you.

51:33 Speaker 3: Where there's data available.

51:34 Speaker 6: So everything that you want to do should be within that search mask so that you're sure to get a response back from our service. So I went and did a little poking around and I found a spot that has some good bathymetry off the coast of Cuba. So I just searched for it and zooms me in there, and I know that there's some good pathymmetry in this area here. So what I'm going to do with the I SID two Coastal Pathymetry box selected is I'm going to take the box tool and I'm going to draw just a bounding box over this area and you can see that it automatically shows you the size of the area for requests using the our public service.

52:24 Speaker 6: For a tail twenty four, we ask that it be underneath a less than one thousand square kilometers, so five hundred and fifty some square kilometers is just fine, and you can.

52:35 Speaker 3: See what it is if you hover over it.

52:37 Speaker 6: Once you have the bounding box, you can just click run side rule and it will generate that request for you. Send that request to the servers running in US West two. Those servers will subset all of the data that needs to be subseted, collect it all in a single response, and send it back to the browser that you're running.

53:03 Speaker 6: Once the data is all loaded, it'll pop up this analysis screen on your browser. The first couple times through you're gonna get a lot of helpful advice. I think it's after five requests, we figure that you've read all the helpful advice and you'll no longer see those. If you ever want to reset or get it back, you can always reset the system here and you'll go back to being like a new user. So now we have a display of all the different tracks. You can hover over it to see the information of the different photons returned.

53:43 Speaker 6: These are all the bathymetry photons in that bounding box in ATL twenty four. So you can hover over. You can see the confidence level, the rithometric height, the reference groundtrack, things like that. You can select the different tracks, and on the right you'll see a plot of the along track photon cloud, just the bothymetry photons. So let me just I'll just show a couple of different ones and see what you can see.

54:15 Speaker 3: What they look like.

54:18 Speaker 6: That's a neat one. Okay, well, let's let's choose maybe choose this one here. Let's choose this one here. It goes down to five five meter depth. So now this shows me as zoomed in view of the pathymmetry. But I want to get a better context. But I want to look at the whole photon cloud that it comes from. So if I click this show photon cloud, it actually goes out and makes another request to the server to grab the ATL O three raw photon cloud and combine it with the ATL twenty four photon classifications, and then it gives now this full view of the data.

55:03 Speaker 6: So what we're looking at now is the input data, if you will, to the ATL twenty four algorithms that produce the classifications.

55:11 Speaker 3: That what we saw on the previous plot.

55:13 Speaker 6: And so if we zoom, let's zoom in here. So I just zoomed into a section. You can see the white here is our sea surface. That's the sea surface classified photons. The purple or the unclassified photons. The ATAIL twenty four algorithm is just saying it's neither sea surface.

55:34 Speaker 3: Nor with imagery.

55:36 Speaker 6: And then the green is the photons that the algorithm identified as a bathometric return. Okay, I think one other thing I'll show here is let's go back to this is sometimes it's useful to just get a quick view of what the whole data set looks like, and so we do for that, we do provide a three D view of it.

56:06 Speaker 3: And so.

56:08 Speaker 6: This here is showing what the what the whole data set looks like. And you can see there's some sea floor here at the bottom there.

56:22 Speaker 3: I guess also.

56:26 Speaker 6: We can you have a lot of things at your disposal as far as controlling what the what the plot looks like. So here we are coloring the points based on the arithometric height. If we wanted to color them based on their confidence level. This is now the same plot, but it's showing which photons have a very high confidence in the algorithm of being a mathemetric return versus ones that have a low confidence. So you can see up here we have some really high confident ones, and then right here appropriately so it's a little bit less confident.

57:08 Speaker 6: Lastly, I will just say there is for those that are power users. You can see what the actual request to the server looked like here. Okay, so this flow has so far been just a very simple, straightforward flow of getting access to the ATAIL twenty four data. We do have in the web client some options for some more advanced access, So now go over here, go back to the request and go up here to this selection button and choose advanced. So now we see we're selecting the atail twenty four x API, and this is going to give us access to more features that the server exposes.

57:58 Speaker 6: Specifically, I'm going to go now to this ATL twenty four X fields. Before I do that, I'll draw a bounding box, and I'm gonna make this a much smaller bounding box because we're gonna get some more data. You'll see why. Okay, So I'm gonna go here and now in my request of atail twenty four instead of saying, you know, I don't just want the bathymetry, I want all of it, the unclassified and the sea service, similar to what we saw when we did the ATLO three requests and used the ATL twenty four as a classification, and then.

58:34 Speaker 3: Also there, I don't want just the main fields.

58:39 Speaker 6: I want all of the fields that are in the data product.

58:44 Speaker 3: So here I'm just gonna select.

58:47 Speaker 6: This whether or not this compact selector, and I'm gonna leave it off now because I don't want the compact version, which happens to be the default version when we do just a general request. If you wanted to, you could also do some filters here, like you could filter on the confidence threshold. You could filter on the turbidity in the in whether or not there was valid turbidity or wind speed, and whether or not it was at night and the sensor death was exceeded. You could also even pull in some ancillary fields from the data product.

59:22 Speaker 3: But we'll leave it just like this, and.

59:27 Speaker 6: We'll show the request parameters and you're going to see, now we've got some additional parameters.

59:31 Speaker 3: In our request. We'll make this request.

59:38 Speaker 6: So this is again a smaller area because My guess is it's going to pull in a lot more data because we're asking for all of the photons and not just some of the photons, not just the pothymmetry photons.

59:48 Speaker 3: Yeah, look at that.

59:50 Speaker 6: So you can see this just in this really small area. And I would we would just caution if you want all of the photons beyond just the pathymetry photons, you do want a small area not just for the load on our surface, but it can very easily overwhelm a typical client laptop running this in your browser as well. So this very small request had seven hundred and nine thousand photons returned.

1:00:17 Speaker 3: But you can see.

1:00:20 Speaker 1: This is.

1:00:22 Speaker 6: A plot here, and then we can do things like now, instead of doing that with the metric, right, let's do the photon classification. And you can see that over here we have these higher classifications, which is sea surface and pathymetry. So right away I can look at this and I can say, over here's land and over here is your sea surface and pathymetry. And if that wasn't super clear just by looking at it here, I'm gonna keep my pointer here and I want you to look over at the left side of the screen where you see the dear graphic map and you should see a moving target showing which photon you're highlighting on the map, corresponding to this elevation plot on the right, And that just confirms that, yeah, this is what you're what you're looking at here is land and what you're looking at here is the water.

1:01:20 Speaker 6: Now what you can do when you get all of the data is you can export it if you want. So we can go to our table and this table tab, and now we can run an SQL query against the data. We can get a full Now all of the raw data is being shown, well actually just the first thousands, so you can if this is more for power users, but it gives you full access to all of the data that you're looking at. I should note right now that this entire application, everything you're seeing here is running in the client.

1:01:53 Speaker 6: So this is a browser side, client side application, single page application. All of the querying, all of it is done on your computer, so you own the data. This is not being stored on the server anywhere, and so we're just giving you tools to be able to interact with the data data that is being returned by the server. So you make the request of the server using this client the server returns the data, and then you own that data and can interact with it how you please. And if you build a query that you really like, you can export the CBS to your computer.

1:02:31 Speaker 6: Or if you want to just export the raw data that came back from the server, you can go over here and choose either.

1:02:39 Speaker 3: Geopark or CSV as the format. The export.

1:02:43 Speaker 6: Coming soon is LAS that should be out probably mid December. Now we support it on the server side, and very soon we're going to tie it in on the client side as well. Okay, I think with that, I'm done.

1:03:10 Speaker 1: What a great demo on the use of slade roll web service for accessing and plotting a TAIL twenty four data. Thank you JP. The following summarized the concepts covered in Part one of the webinar series. NASA's ice AT two carries a photon counting laser altimeter, which is named Atlas. Atlas provides near contiguous a long track sampling using six individual beams of green light at five hundred and thirty two nanimeter wave length, providing high vertical resolution. Every photon detected by ATLAS has a latitude, longitude, and elevation associated with it.

1:03:46 Speaker 1: Atlas provides surface specific data products developed by experts and include products for land ice, sea ice, the atmosphere, vegetation and land oceans and inland water. ATL twenty four product provides ATLAS derived coastal bathymetry via ensemble machine learning models, where each photon classified as bathymetry will have a confidence value assigned. The National Snow and Ice Data Center or NSIDC provides access to ATL twenty four data metadata and tools. NASA's Earth Data Search is used to find and spatially subset ATL twenty four bathymetry data.

1:04:27 Speaker 1: Slide roll is a public web service with low latency access to on demand data products stored in S three and slide roll provides on demand processing next to the data for generating customized data products using parameter supplied in the user's request. Looking ahead to Part two of the webinar series, we will be covering the following topics. How to generate satellite derived bathymetry using methodologies combining ATL twenty four bathymetry with optical satellite data from Sentinel TIS, two Python workflows for filtering, visualizing, and analyzing bathymetric point clouds, Integration of ATL twenty four into existing Noah workflows for nautical charting and coastal zone management.

1:05:14 Speaker 1: An overview of Noah's satbathy tool and its capabilities for automated satellite drive bethymmetry generation using ATL twenty four. Before we transition to the question and answer session, I want to remind you there will be one homework assignment which you will be able to access from the training page on December fourth. Answers must be submitted by Google Form with the due date of December thirty first. To receive a certificate of completion, you must attend all three live webinars and complete the homework assignment by the deadline.

1:05:46 Speaker 1: You receive a certificate via email approximately two months after completion of the course. We want to thank once more doctor Lauri Magruder from the University of Texas at Austin, Doctor Diane Fritz from the National Snow and Ice Data Center, and JP Swinsky from NASA Goddard Space Flight Center for their presentations and instructive demos. Below is the contact information for Lori, Diane and Jp, along with links to the ourset website and social media. If you enjoyed today's webinar, we hope you will sign up on the arset list, serve to receive notifications of future trainings, and follow us on social media for future relevant announcements pertaining to NASA's Earth Sciences.

1:06:29 Speaker 1: Below our list of resources relevant to the material covered in part one of the training. We will now transition to the question and answer portion of today's training.

1:06:48 Speaker 8: Great, and thank you for everybody it's been submitting your questions. We've got a lot of good ones so far, so jumping right into it. Question number one, can we develop watershed models using the data from the I st to imagery what would be the minimum resolution?

1:07:07 Speaker 5: I'll go ahead and take that. I'm not one hundred percent sure about the kind of model that is wanting to be developed, but I think hopefully the presentation was good at describing the data format. So we have a long track photon returns, and those tracks, we have six beams that are coming three pairs, and the pairs are ninety meters apart from each other, with three kilometers separating each beam pair. But the along track resolution is quite good at around seventy centimeters. And so it's really the elevation data that you get from IAT too, regardless of which product that you're looking at is really good validation data at a long that track, So I hope that helps answer that question.

1:08:10 Speaker 8: Hey, Diane, thank you. Question number two, can we derive data for different time periods to do comparisons? It would be useful for cebid mobility assessments.

1:08:23 Speaker 5: This was me as well, So a TAIL twenty four as many of the other data products are derived from the main ATLO three photons and those have been gathered since twenty eighteen, and you can separate out those files with temporal queries and pick what you're looking at at different time periods. Of course, the return time of the I SAT to satellite is ninety one days, so between that and getting good data, your particular location may limit some of what you're trying to do with those comparisons, but those are the parameters for it.

1:09:09 Speaker 8: Awesome, Thank you, Diane. Question three is the ATL twenty four product already corrected for tides or is this the post processing step we have to consider, especially for locations where tidal range is considerably wide.

1:09:22 Speaker 4: This is Chris, I'm as speaker on the Thursday session, but I can answer this if nobody else wants to jump in. We actually don't need an explicit Todd correction step in the ATAIL twenty four workflow because we use a process similar to what's called allipsoidlarly reference serveang and hydrosurveying world. Basically, it means that our data, our heights are, our depths are not on the instantaneous water surface the relative to a vertical data at every step in the process. See, they're an orthometric data more an ollipsoidal datum.

1:09:55 Speaker 4: And for people that want to transform to a tidal EDWM like mean low Er La water in the US, at least we can use a tool called Noah's v d ADAM to be able to do vertical datum transformations. So basically we don't we don't actually need an explicit TIPES step, but we leave it to the user to be able to transform to different vertical datams, including title datams, if they want to do.

1:10:17 Speaker 8: So. Great, Chris, thank you so much. Question number four, that's a very clear scatter plot. Could you tell me about the steps taken to check for overfitting?

1:10:32 Speaker 7: I can answer that if no one else is. This is Jeff Perry. I'm on, UT's a TIL twenty fourteen, and I kind of wrote this assuming I understood the question correctly. I don't actually have the plot in front of me, but the individual algorithm should have wrote this first. The individual algorithms may or may not overfit depending on their native They're all different, right, they all are very different algorithms. The final model, the ensemble is we use xty boost just because it's fast, for no other reason.

1:11:09 Speaker 7: It is prone to overfitting. We can we can fix that by adjusting the hyper parameters of the model. But after all is said and done, a lot of the overfitting we correct by applying this blunder detection. That's, like, you know, really obvious errors that some of the models can make, like they can some of them can put bithymmetry on top of sea surface and obviously you can't have that. So that fixes a lot of those overfitting errors. I hope I answered that sufficiently.

1:11:43 Speaker 8: Jeff, thank you so much. I really appreciate that. Question Number five, you're welcome. How much work will it take to extend this approach to inland waters? Is it possible? And is any team working on such a product? What about in northern latitude lakes and ponds?

1:12:00 Speaker 4: This is Chris again. I can answer this one if no one else wants to we in developing ATL twenty four were very purposefully did not consider inland waters because there's already a dedicated data product for inland waters, which is ANTIL thirteen, and ATL thirteen does also include agathymmetry attribute.

1:12:22 Speaker 8: Terrifect. Thank you so much, Chris, and hopefully whoever asked that question can go and search for that ATL thirteen product. Question six, what applicability does the STATEA have for bethymetry mapping of inland and near coastal lakes?

1:12:39 Speaker 4: Is that the sorry, that one similar similar to the it's similar to the previous question right inland waters again, that's at ATL thirteen.

1:12:49 Speaker 8: Is some inland waters? Okay, great, thank you Chris. Question number seven, our positive depth biases ATL twenty four common and what can be the cause?

1:13:05 Speaker 4: I can maybe take that one too, unless Jeff for somebody else wants to go ahead. I think that's still to be determined. I think that's a great case where we'd like to hear from users of the data as they do comparisons against independent reference data if they're seeing positive depth biases and that accuracy test that was in Laurie mcgruder's presentation, we actually didn't see a bias being consistently in one direction or another. I think we saw biases of around ten centimeters that were that could either be positive or negative.

1:13:41 Speaker 4: There are some reasons why we might expect a bias. For example, one of the things that are ATL twenty fourteen has started talking about recently is what's known as forward scattering bias and mathematric ladder. But I think we just haven't. I mean, part of the importance of ATL twenty four and I I sat it to pythymmetry in general, is that there's just so many areas in the world where we don't have really good existing pathymmetry, and so that's kind of a double edged sword. It makes the data products quite valuable, but it also means that there's in many places in the world not a lot of existing data to compare to.

1:14:23 Speaker 8: Great Chris, Thank you so much. Question number eight. Does the slide Real web client apply any de noising or classification algorithms before presenting ATL twenty four bythymmetry points or are we viewing the raw ATL twenty four photon returns.

1:14:39 Speaker 3: This is JP.

1:14:42 Speaker 6: What was shown when we show it through the accessing it through the web client it's it's just the raw pathymmetry.

1:14:50 Speaker 3: In the general tab.

1:14:52 Speaker 6: We do by default filter any low confidence pthymmetry just to keep requests as fast as possible, but then allow the user to drill down and raise the threshold if they want now in a couple of days. Well, there's other APIs where you can apply different processing steps to the data being returned. Those steps aren't available in the web client, but are available through the Python client. Also, I guess, just going back a couple questions for those interested in ATL thirteen the inland bodies of water, there is also Cydebrugle also supports accessing that data from the web browser as well.

1:15:41 Speaker 6: Instead of the Coastal bothymetry tab, you just select the Inland bodies of Water tab and you'll get the ATL thirteen data product.

1:15:51 Speaker 8: Great JP, thanks so much. Question number nine, how does ice have to you distinguish between photon returns from the sea surface and photon returns from the seafloor in shallow coastal waters, especially considering issues like photon noise and surface wave variability.

1:16:11 Speaker 7: This is Jeff. Again, Matt had answered this, but he's having technical difficulties. Otherwise he would probably be speaking up.

1:16:23 Speaker 7: But as you can see in his answer there, Sorry now I'm just kind of reading it. But the question is, how does it distinguish between photon returns and shallow coastal waters. I mean, like he says, we train on a lot and a lot of data, and sometimes in really you know, high waves, we do make mistakes in the trough of a wave can be classified as seafloor. In release we've had so what we've been calling version one of the algorithms. We've had two releases of this, and the second release fixes a lot of those.

1:17:05 Speaker 7: It goes back and looks, especially for those places where this trough of the wave is misclassified as bethymetry and fixes them. So that's one way we deal with that.

1:17:22 Speaker 8: Okay, Jeff, thank you so much for chiming in, and we apologize that Matt is not able to access or unmute himself. But moving on question number ten, how does the ATAIL twenty four handle the challenge of separating true seafloor returns from water column backscatter in shallow coastal zones where turbidity is high.

1:17:44 Speaker 7: Hia, it's Jeff again, So I mean, obviously, if there's too much turbidity and the signal disappears. There's nothing you can do. You know, at some point that's going to happen, right, And if the noise is too high, well there's you know, if you lose the signal, you lose a signal. But uh, you know, again, we have an ensemble here, multiple algorithms, and so for example, the algorithm costnet. It operates on high uh you know, imagery with high vertical resolution. You can very clearly separate the two signals if there's not too much sturbidity, and so it separates those out.

1:18:23 Speaker 7: And again, as we keep saying, the ensemble has an opportunity if some algorithm misses it or confuses the two signals, or makes assumptions about the signal. Like I mentioned in the beginning, some algorithms don't don't won't even predict bathymetry if the bathymetry signal is stronger than the sea surface. But the ensemble and the blunder detection has a chance to go back and fix all that. It doesn't always fix it, but but it has that opportunity.

1:18:55 Speaker 8: Great, Jeff, thank you so much. Question number eleven is the symmetry available for large room systems like the Mississippi River or are there too many issues with turbidity or orbittle path shoreline geometry.

1:19:13 Speaker 4: I think that's again similar to the previous questions on inland waters, where that would be the A Tail thirteen product and not ATL twenty four, which is coastal focused.

1:19:25 Speaker 8: All right, Chris, thank you. Question twelve in slide role, is the data derived per line only so the spatial variability is poor?

1:19:36 Speaker 6: Yeah, So I think what you're asking is all of the data products right now that we generate or provide services for a long track, and they are. We would love to have produce image data raster data sets. We hope to in the future, but right now we're funded only to support the along track products.

1:20:06 Speaker 8: All right, thanks GP. Question thirteen does the Python library allow for larger bounding boxes?

1:20:17 Speaker 5: So this this is just limited by the retrieval of the granules from the cloud hosted data. So you can make larger bounding boxes, but your your code may stall a little bit. But I put in a reference here for being able to look up Harmony and look into the Harmony pie library and see. So I to be honest, I haven't done a very large grab myself, so I don't feel one hundred percent confidence about this answer.

1:20:56 Speaker 6: And for side rule, it's really easy, and I'm guessing the harmony it would have the same fundamental problem. It's it's really easy when you're looking at a globe to not realize how.

1:21:14 Speaker 3: Big of a request you're making.

1:21:18 Speaker 6: I don't know if you remember when I was drawn that really small box in the demo, and that had seven hundred thousand points in it. You can imagine a factor of one hundred, factor of a thousand on that easily. You know, you're easily talking a billion rows and and so with these photon photon cloud data sets, it's very easy to get huge, huge returns. So all that being said, on side rule, we let we let you do it, and and we'll do the best we can. But it's typically the client crashes before the servers do.

1:22:01 Speaker 6: That's just what we notice.

1:22:04 Speaker 5: Yeah, So, as JP is saying, these data sets are massive, so really kind of doing some pre look of you know, what is your temporal subset that you might want or your spatial subset and really restricting it to what you truly need is the best way.

1:22:23 Speaker 8: Yeah, thank you, Diane and GP. Question fourteen is turbidity the main thing affecting the confidence level? Does that come from veers?

1:22:42 Speaker 7: I'm not sure who who put this answer in? So I can answer if no one else's the I mean yes, the answer the first question is yes, turbidity and background noise is what's affecting the confidence level. But it's implicit. It's not we don't measure the turbidity and then measure the background noise and then make assumptions, you know, make predictions based upon that. It's just implicit in most of the algorithms. When there's more turbidity and more background noise, it's just going to have a harder time, uh, deciding what you know, what the label, what, what classification of photon is and and so uh.

1:23:22 Speaker 7: And then down in the answer the X you boost model. It's a tree based model, and so it uses that tree structure to determine what the what the competence level is. It's like a probability. It's a soft max if you know what that is. But uh, it's it's not strictly speaking, it's not a probability, but it looks like a probability and acts like a probably. So so that's where that comes from. Is it's it's part of the UH decision tree model. UH will will for every photon, assign probabilities or SoftMAC probabilities to each label and then choose the most probable one.

1:24:07 Speaker 8: Thank you, Jeff. Question fifteen, can we use it for detecting plate movement of tidal areas to get information about upcoming earthquakes.

1:24:26 Speaker 5: I'm just going to unmute and say that this is a really interesting application idea. Chris, if you please comment, But I think that this is beyond the production of ATL twenty four data. But it's great to have people thinking this way and thinking about applications for this data.

1:24:49 Speaker 4: I was going to say the exact same thing. It's really interesting, but beyond the scope of what we've looked at. I don't know if the special resolution and accuracy support that. Yeah, if anybody looks into that, i'd I'm very curious to hear what they find.

1:25:06 Speaker 8: Diane and Chris, thank you. Question sixteen I think has been addressed in previous questions. So for those that are interested in inland athymmetry, definitely check out the ATL thirteen product. Question seventeen. In satellite arrived bathymmetry workflows that use ATL twenty four as reference data for training models on sentinel to imagery. Are tubidity and bottom type limitations generally encountered first in the multi spectral reflectance rather than in the ATL twenty four lighter signal.

1:25:40 Speaker 4: I was just reading.

1:25:43 Speaker 8: This question.

1:25:44 Speaker 4: Oh, somebody scrolled and I'm not sure which one was this again seventeen in the Thursday sessions this week, and the presentations that both I give and gretch Nami from Noah, We're gonna We're gonna talk about using ATL twenty four in combination with satellite drive with imetry from multi spectral imagery, So hopefully those presentations will help answer this. Let me read this question again. Though our turbidity and bottom type limitations generally encountered first in the multi spectral reflectance.

1:26:23 Speaker 4: They're sort of independent workflows, but some of the constraints are the same. So obviously, in any method of satellite drive with imagery relying on spectral data and relying on having returns from the sea floor, turbidity is one of the main limiting factors, and that's the case also with with IAT with imetry. Generally the ranges in which they both work are are pretty similar, which makes ATL twenty four a pretty good choice for calibration and validation or just integration with special drift immetry.

1:27:04 Speaker 4: I hope that answers the question, and I hope that the talks that Gretchen and I give on Thursday will also help answer that.

1:27:14 Speaker 8: Made Chris thank you so much. Question eighteen. Are their tools available for accessing the ATL thirteen products such as slide rule and Diane thank you so much for adding this link so that they can access that data set. Question nineteen. Is there a gridded product available or planned for the future for the US East West? It's covered by coasts? Thank you.

1:27:45 Speaker 4: I don't think our team has explicitly talked about doing a gridded product. From the beginning, the intent was for ATIL twenty four to be in a long track product. I think that certainly they're their tools and options that will help end users create credited products in some areas. I think we're just starting to see enough coverage from all of the different i ST two overpasses that we maybe have enough data in places to do like a gridded bathmetric DM solely from i ST to data, But in other areas where people want a gridded product kind of like wall to wall coverage.

1:28:28 Speaker 4: Those are cases where the integration with specially drived withthymetry from multi spectral imagery can be very valuable.

1:28:38 Speaker 8: You thank you, Chris. Question twenty what changes are expected at the confidence levels next year? It looks like Matt might have answered this one, and unfortunately that does not have access to this WebEx but he did provide an answer which I will read. But I apologize because I'm not the one that contributed this answer. But we are actively improving the constituent models and investigating new models to add into the mix. The confidence levels are entirely dependent on capabilities of the constituent models.

1:29:19 Speaker 8: After building the ensemble and prior to publishing the next version of ATL twenty four, we will conduct a study of confidence values relative to reference with symmetry data sets to determine the updated best practice value to use. Thank you Matt for adding that and for answering that question. Question number twenty one. How critical is viers KT four ninety to the uncertainty assessment? I would think clouds and land adjacency would greatly reduce the quantity of KD four ninety retrievals and quality of these retrievables would be reduced in optically optically shallow waters unless alterations to the standard KD four ninety algorithm are used to minimize benthic contamination.

1:30:05 Speaker 4: I'm not sure to type that answer, but I agree with what's given. There's the answer. The Deer's Katie four ninety is not directly used in the UH the deeporate retrieval. It is the only place where that's currently used is in the uncertainty calculation. And as it says an answer there that that's used in data that's that goes into the uncertainty look up tables. Yeah, hopefully that that answers it.

1:30:36 Speaker 8: And Chris, I don't know if you can answer this next one as well, but why not use Katie four ninety in a green band example? Given the viewers five hundred fifty five.

1:30:46 Speaker 4: The the global Katie diffuse attenuation coefficient of downwelling radiance. That's what Katie represents. The global data products that we're aware of and work with are generally provided as Katie four ninety, meaning diffuse attenuation coefficient at four hundred and ninety nanimeters. It would be better to use one closer to five hundred and thirty two nanometers, which is the wave length at which I SAT two's ladder system operates. What we do is we actually provide a conversion from KT four ninety to KTIE five thirty two.

1:31:28 Speaker 4: And that's that's based on a drived empirical linear relationship between KTI four ninety and KT five thirty two. That said, and I'm not aware of it, but if there is a global readily available vers KDE or Vere's KD five point fifty five, the question as a good one that probably would be a better product to use.

1:31:54 Speaker 8: Thank you, Chris. Question twenty three, can I get access to the data for Nigeria?

1:32:05 Speaker 5: It is a global data set, so along the coast there, I don't know what the mask actually looks like in that space. So Chris, maybe at the top of your mind you might have a better sense. But yes, what is available from finding mathymetry from ATLO three to ATL twenty four should be in that area.

1:32:31 Speaker 8: All right, allan thank you. Question twenty four could this be used to map objects in your coasts like vessels, booies, jetties, et cetera.

1:32:43 Speaker 5: I wrote, what is there? But Chris, you will definitely have a better sense of this because depending on the size of the sunken vessel things like that, and the clarity of the water being able to see that in the seafloor. Not sure if if anybody has come across that in the tests, but possible if if the there isn't any turbidity in the water. I would think, Chris, I think you're still muted.

1:33:19 Speaker 4: I was just saying, I agree with Diana. It's it's really just a function of the of the resolution, and I think that one was addressed already. But basically, in the along track, along any individual beam track, the resolution is approximately seventy centimeters, although it also depends on the posymetry cass, it depends on water water clarity, and then it's coarser cross tracks passing. So just you know, is the size of an object is it? Is it detectable at that special resolution at least for things like jetties or you know, large you know, coastal engineering structures.

1:33:54 Speaker 4: Uh, if if you happen to get a track line over it, you will, you'll you'll most definitely say it. And I've seen cases where you can get with immetry right up to say a jetty and then and then you'll see the see the surface elevations going over the top of the jetty.

1:34:15 Speaker 8: Hey Diana, Chris, thank you. A last question twenty five. What happens in the presence of algae blooms or sea grass.

1:34:27 Speaker 4: That's a great question, and I think the answer is it just it just depends. You know, when you've got algae bloms, that's obviously impacting the serbidity, impacting the ability of that laser light to penetrate through the water column and to get a reflect a detectible reflection off the sea floor that makes it back up to the to the sensor. And then it's kind of a similar thing with with UH with sea grass, UH that can do a couple of different things that it can obscure the sea floor. It also seagrass tends to have relatively low reflectance at five hundred and thirty ten animeters, so it can make it difficult to get returns.

1:35:06 Speaker 4: But more commonly, what I've seen in areas for example, you know, we've looked at some data around Saint Croix where there might be a particular track that goes over a seagrass bed, and you know, sometimes we have seen what looks like good withthymmetry in those areas. I don't think we've done an accuracy assessment specifically over sea grass, so you know, how much does that impact the accuracy. I would say kind of unknown at this point and probably depends a lot on the density of the sea grass, depth and other factors.

1:35:47 Speaker 8: Great, Chris, thank you so much. As we wrap up, I just want to thank one of all the participants that joined wherever you joined from, Thank you so much. We do look forward to seeing you in two days on Thursday for the second and final part of this training series. We also want to thank once again Laurie McGruder, Diane Fritz, and JP Swinsky, and also for those that were able to join to help answer the questions, Jeff Perry and Chris Parrish. Thank you both. And we also want to acknowledge that Amy Neely who's also helping out and supporting this training as well.

1:36:16 Speaker 8: So thank you to all the people that presented and also helped to answer the questions, and we look forward to seeing you all in two days for the second part of the webinar series. Thank you