https://www.youtube.com/embed/oAbJVq5wFS0
If you can choose the others, how much you get a less complicated access. >> > > Ankit: Yeah. Let me promptly initially try to share the screen. And after that I bring them. It'' s like, you recognize, initial I have to share screen. Present your. Yes, I can do the entire display. And also you see the slides. > >> Greg Bronevetsky: Yep, that'' s excellent. >> That was remarkably simple. >> > > Ankit: Great. > > Greg Bronevetsky:'Okay. Yes, if you if you ' re the
>> one driving, that works completely. > > Ankit: Yeah, I believe BA will certainly need to change to his very own computer system due to the fact that some of the slides obtained corrupted on >> > > Greg Bronevetsky: It ' s okay >>. > > Ankit: Google and also we already recognize the concern. So he stated he will just switch. Yeah. >> > > Greg Bronevetsky: Okay, that works. Let me see >> > > Ankit: Kind here. >> > > pierre hiernaux: Hi, Ankit.How are you doing? > > Greg Bronevetsky: How ' s it going? > > pierre hiernaux: Penalty, thank you >>. Greetings, Greg. > > Greg Bronevetsky: oh, > > pierre hiernaux >>: I believe it'' s early morning in your area. >> > > Greg Bronevetsky: good night, almost consuming. Looks like there'' s some light still. Oh, you desire to attempt presenting your slides, I believe you have a few >> > > pierre hiernaux: I have to enhance the noise, sorry. Please, can you speak? >> > > Greg Bronevetsky: Hey there, can you hear me? >> > > pierre hiernaux: Yes, but it'' s still Devote. Okay, this will be much better. >> > > Greg Bronevetsky: Remain to pointers for the stocks now. >> > > pierre hiernaux: So, might remove this. We ought to put right here. My treatment is the last one, however you desire you want me to attempt and also see if I >> > > Greg Bronevetsky: It ' s confirm.So if you ' ve made use of Google meet in the past, it needs to function penalty. > > pierre hiernaux: allow me see >>. where it is below. > > Greg Bronevetsky >>: let me simply email her to make certain he is able to make it. > > pierre hiernaux: Have not> accessed to, where is it? Sharing. > > Greg Bronevetsky: So there ' s in> the bottom of the home window. There ' s a switch that has an arrowhead up.'annotation screen. > > pierre hiernaux: But I don ' t see it.This task. No. > > Greg Bronevetsky: See. I am, I ' m gon na share my >> screen. Or two, on my screen. > > pierre hiernaux: Yeah. >> > Greg Bronevetsky: Switch right here, this blue one. That ' s I already pushed but you click it. >'> pierre hiernaux: Okay. >> > > Greg Bronevetsky: It provides you the rest. > > pierre hiernaux: alright, I ' ll try >> So I ' m attempting this set integrated with a with confidence needed to go right into the recall And then this set. Birthday celebration. Can you see it? > > Greg Bronevetsky: Yes. Fantastic. Okay. So I didn ' t understand to Tucker yet you understand, just how a lot longer it will be? Um, so we'can produce like a couple mins and afterwards start > > pierre hiernaux: So I ' ll quit it now.A little bitage. Okay? > > Greg Bronevetsky: Okay >>. Allow ' s just inspect,'see what we ' ll obtain is. I will >>, you recognize. Oh fine. we ' ll have if anyone joined this point live, they can ask concerns concerning or else I obtain, you understand, questions from the outside thing is you understand ask my very own concerns or ask inquiries that people ' s part is injecting you between, fine. > > Ankit: And also fine with me as well as others. > > Greg Bronevetsky: Okay. >> All. Yeah. Therefore should we await >> football, or > > Ankit: yeah, because I think Jim is the one who is beginning to discussion >>, so it ' s be great to await, but > > pierre hiernaux: What is away? That ' s why. > > Greg Bronevetsky: Away. It ' s come >> to not Tucker. Yeah. Come yeah. That makes sense. He ' s away. >'> pierre hiernaux>: I put on'' t recognize if I am far away.You ' re in Australia, aren'' t you? > > Greg Bronevetsky: No, I'' m us West Coastline. Someplace to remain location so that'' s> > > pierre hiernaux: Okay >>. > > Greg Bronevetsky: not from you. >> > > pierre hiernaux: so >>, > > Greg Bronevetsky: Australia would certainly be >> > > pierre hiernaux: Okay. I maintain have including ads that is available in. I don'' t recognize how to do away with these.> > > Greg Bronevetsky: No. Okay. Look he ' s Had the ability to make it or not. >> > > pierre hiernaux: Shall we send a message to Jim? >> > > Greg Bronevetsky: I yeah, if you people can currently have >> > > pierre hiernaux: Or you have currently, okay? >> > > Greg Bronevetsky: But if you have like a conversation point with him, I just have the email. >> > > pierre hiernaux:'I ' ve only the email. >
>> > Greg Bronevetsky: Okay. >> > > Ankit: I can give him a telephone call on his very own. >> > > Greg Bronevetsky: Yeah, that would be fantastic. >> > > pierre hiernaux: yeah, simply get the mail you sent to >> > > Greg Bronevetsky: Yes, there ' s always that delay for email. You can'' t rather trust fund it. >> > > pierre hiernaux: Yeah, I think it takes a sometimes to reach. First. >> > > Greg Bronevetsky: Yeah, he was never rather developed for fast communication. >> > > Ankit:'I ' m sorry, yet'Jim didn ' t pick up which is quite, extremely unlike Jim.But shocked.
> > Greg Bronevetsky: Yeah, what ' s good suggest? Well', what ' s a great means of I mean, is it better to wait or to start? >> > > pierre hiernaux: Eventually we can begin by the end. Which means? >> > > Greg Bronevetsky: you understand, inevitably >> > > pierre hiernaux:'I ' m not mystery the primary. Issue. >> > > Greg Bronevetsky: yeah. Most likely we can provide one more like, you know, 5 mins and also then choose. >> > > pierre hiernaux: Allow ' s see. Okay, 5 mins greater than to make a choice due to the fact that Well, there'' s nothing wrong in examining by the end. >> > > Ankit: I suggest >>. > > Greg Bronevetsky: Maybe, yes, rational flow. It'' s great. >> > pierre hiernaux: Do you concur? Ankit. >> > > Ankit: I absolutely concur. >> > > pierre hiernaux: you know, it'' s much like if'you ' re coding a program, you can can begin by the end. >> > > Greg Bronevetsky: Oh oftentimes, you do, there'' s an entire design of you start with an examinations specify what it is that you wish to have actually written as well as then you create the thing.
>> > > pierre hiernaux: and yes, well in the paper, you never ever begin by the intro. You recognize? you never begin by the you always thought by somewhere. In the outcomes. and after that you attempt and develop up. >> > > Greg Bronevetsky: As Well As you'' ve started intro, compose the remainder of it, after that alter the direction. Having actually recognized what it is that I indicated to have actually created. >> > > pierre hiernaux: Yeah, if you begin by the introduction without a doubt, you'' ll have it to change totally at the end. >> > > Greg Bronevetsky: Yeah, right
>>. > > pierre hiernaux: Okay >>. > > Greg Bronevetsky: yet it'' s motivation. Like >> > > pierre hiernaux: Yeah, in some cases it makes It aids to have this. To red line. Then you transform it. >> > > Greg Bronevetsky: Yeah, at this paper that. I imply, I'' ve not been documents in lengthy time, however I have this paper, I began in 11. It'' s still > >> > pierre hiernaux: Hmm. >> > > Greg Bronevetsky: Superior. Oh, there is and yeah, that direction has actually been has actually been composed lot of times as we found You understand. What individuals did not comprehend the very first a number of times. >> > > pierre hiernaux: okay, here is >> > > Greg Bronevetsky: Perfect. Hey, just how'' s it going? > >> Tucker: Switch on my microphone. >> > > Greg Bronevetsky: Yep, that'' s perfect. You wish to begin? Share your slides as the various other thing that can >> > > pierre hiernaux: hydrant. >> > > Tucker: Hi, everyone starting to be late. >> > > Greg Bronevetsky: That ' s alright. As well as we can run a couple of minutes late. That'' s fine. >> > > Tucker: Okay. That should reveal the slides? >> > > Ankit: Jim, I can do it for my part in your component.
>> > > Tucker: Okay, excellent and also excellent. Why put on'' t you? Hello there Pierre. >> > > pierre hiernaux: Hi. > > Tucker: Hi end Set. I Greg > > Greg Bronevetsky: Hey, how ' s it going? >> > > Tucker: A little bit spread. >> > > Greg Bronevetsky: No concerns. Yet yeah, we'' ll handle. I guess, Jim, you'' re speaking. Okay. >> > > pierre hiernaux: Okay >>. > > Tucker: Yes. I ' ll be presenting Ankit as well as Pierre that have the hardest part of the discussion. >> > > Greg Bronevetsky: All. Well, then allow me begin the recording and also once it dents I can offer everybody brief control as well as yeah. >> > > Tucker: Okay, we ' re prepared to go. >> > > Greg Bronevetsky: Second. >> > > Ankit: Can you? >> > > Tucker: Greetings. I just begin. >> > > Greg Bronevetsky: Yes. Oh hey. So hey everybody, I'' d love to introduce video clip. Jim Tucker below currently and obtain the Aparia from. I guess when you composed the paper was >> >
>> > Tucker: Yes. > > Greg Bronevetsky: NASA Goddard, best as well as obtained discussing well taking a look at Forest from room, you understand. Welcome. >> > > Tucker: Many thanks. I'' ll be undergoing some slides from a recent paper of ours in nature or from very early last month as well as The 4. primary authors, or the first four authors on our paper. Myself Martin Brant that regrettably can not join us Pierre. No and ankit.And we suggest
in the paper that the 4 people must be considered the very first writer on the paper due to each person'' s one-of-a-kind payments to what became a rather wide paper. in addition to a relatively deep paper. So what we did was we made use of a huge quantity of satellite data, I can have the following slide in Set. To convert tree crowns over a location as big as the location of the continental. USA and also Alaska regarding 10 million square kilometers from the Atlantic Sea to the Red Sea north of the Equator as well as south of the Sahara. We utilize a 3rd of a million maxar images to Make use of maker finding out the finding out component of our research study, was 89,000, training day of specific trees.We burned
a great deal of hb'' s of high performance computing time on the data. We arranged. For a matrix of 2 times. 10 to the 13th components. and and after 60 million core hrs on blue waters. before it was decommissioned, we mapped in fact 13 billion trees over the whole area and also 10 billion trees with tree. crowns, better than three square meters, in the semi-arid portion of our research study, which we exchanged carbon amounting to factor. Eight petabytes of carbon Disaggregated, right into fallen leave wood. and also root carbon at the private tree level. Anytime you do a research study like this, you put on your own in a box. Exactly how do people use your information? We resolve that and we have an audience connected with our paper that are since which allows you to check out our whole information set of the 10 billion trees place as well as carbon at the treat down to the tree level next off, slide. Our study took component on the south side of the Sahara as well as we used Maxar satellite data.In the early dry period, the top picture is the wet season where every little thing is environment-friendly. The lower image coincides area. You see the same trees, you see all the floral greenery is senescent, And also we concentrate on the tree grounds following. Slide, please. Below ' s the very same picture with an NDVI'profile of the of this location and also Senegal which you see between where the guideline is currently. As well as what we have is a solitary wet season. It is really pronounced. You have the dry season and you have a rainy season of 3 or four months, five months, and some locations, relying on latitude.And then you go
into the dry season, all of our data, which we make use of are from October to March throughout the completely dry period next slide. >> > > Greg Bronevetsky: An inquiry, Why? Our tree premises more pronounced because season. >> > > Tucker: Yes. so we wish to work the well, they'' re two factors. One is cloud. cover is considerably less and also secondly, we have an obvious distinction between trees and every little thing else.As you can see
, in this photo in In the rainy season. The floral plant life. Is. is eco-friendly or actively expanding as well as the completely dry season in the bottom image. They'' re basis plants is senessa. So we selected months of maximum contrast between tree crowns and also their background. Allow'' s go to the following slide. Ankit. thanks. This is how points look for our training information. This is and likewise what we mapped the arrows on the best indicate the same tree, which is on the left.And we use
the pancromatic band to determine the shadow and our meaning of a tree is it has to have an environment-friendly fallen leave cover. as well as that'' s, which ' s suggested in the base, photo with the normalized Difference, Plant life index for this particular tree and an associate in a connected shadow combined with the eco-friendly tree canopy. This is what every one of our training information are based upon under a large range of problems viewing angles, aerosol optical densities different cloud cover situations and you see you. you can see how the environment-friendly tree crown is divided from the darkness but adjoining to it. Next slide. Below'' s one example of one tree. on the left hand side of the picture and the same tree as we have actually processed. Our data with a convoluted neural internet, you net architecture. and identified that tree. and also we have actually not identified as trees. any one of the bushes, and also and grassy locations either due to the fact that they are inactive or due to the fact that they if they'' re green, they do not cast a darkness. So next slide, please. From our research study, we recognized 9.9 billion tree crowns better than three square meters of location for the area.We program right here.
For our, which goes Forgive me, which has a rainfall variety from 0 millimeters each year in the far north of the picture to the bottom of the picture, which has a thousand millimeters of rainfall per year. and after that convert it, each of those 10 billion trees into carbon at the tree degree next slide. This isn'' t, this is a photo of me taking God forbid, 42 years earlier. In Senegal I started functioning in sub-Saharan Africa in the very early 80s making use of satellite data.This is where we had NASA established Time series remote noticing utilizing information from meteorological satellites. This is where I fulfilled Pierre or no, who will certainly adhere to Ankit and the program and just a photo of me in the same area throughout the stormy season. This is a location where where Pierre and I have functioned. Pierre has extensive experience. I sanctuary'' t episodic series of experiences there next slide. Now, when you develop a data set of 10 billion trees and also transform them right into carbon, you can look at different distributions of what you have. you definitely have a huge example dimension. If people ask you, what'' s your sample size in our situation, we can claim in equates to 9.9 billion trees as well as on the left.You see a distribution in megagrams of carbon per hectare? with respect to rainfall, if you increase these worths times a thousand you transform the worths right into kilos of carbon On the right-hand man side. you see data summed up by specific tree as well as you see, once more that if you multiply the megagrams of carbon times a thousand, you obtain the The kgs of carbon from timber root as well as leaf from each tree in question. As well as you see the circulation of these with various percentiles, these are figures from our paper and also nature and also it ' s crucial to check out the mean, as well as circulations as you see here.Next line, end Set. > > Greg Bronevetsky: An inquiry on this. The plot shows, For the various environment
zones you, It appears like you ' re overlooking the tree varieties due to the fact that I ' m presuming you can ' t truly figure that out. And also you ' re saying that, whenever it ' s, it ' s Rainier these'trees because environment build up even more mass. Is that? > > Tucker: Well,'Pierre or no will enter into the botanical circulation of trees thoroughly, yet yes, that ' s> what we ' re saying that we as a result of our place. You have a restricted varieties variety as well as we are able'to draw to'merge our lama, tree relative to varieties. If we didn ' t have the very same distribution of trees are the exact same distribution of tree types from the Atlantic Ocean to the Red Sea.You couldn ' t do that, but we do. and we have and also these are the results. > > Greg Bronevetsky: Okay. And after that you were getting roots yet this you understand, you ' re getting anything regarding the dirt past the tree origins,> > > Tucker: No, we don ' t care regarding the dirt. We have sufficient issues to fret about in terms of roots wood and leaves. > > Greg Bronevetsky: There. yeah. Okay, trendy. Thanks. > > Tucker: You ' re welcome. Okay. Now this is a plot of the of >> the carbon chance distribution on the left hand side and also a >> distribution of tree crown area and also a cumulative style on the right hand side, I ' ll attract your attention on the right-hand man side to the arrow indicating a 50% factor at the 15 square meter tree crown area. Fifty percent of our trees are tiny and also have a tree cover from three square Meters to 15 square meters.And then 50 %have a lot more. So, this is why it ' s critical to use 50 centimeter information. Or else you miss a big number
of trees, which have a whole lot of carbon. On the,'on the left hand side, I draw your attention to the top to sections, which display in the red numbers. the the portion of carbon in the hyper-arid dry semi-arid as well as subhumid areas specifically and below them listed below the arrows defining. the array for Hyper-arid as well as all of the 4 zones with respect to annual precipitation, you see the location, which they make up. So, the factor below, is 90 %of our carbon comes from. only 15% or from 30%of our location. And also as we most likely to greater rainfall amounts, 15 %of our carbon comes our brothers, 64% of our carbon originates from. 15 %of the location, which has from 600 to a thousand millimeters per year. You see the sample dimension. You see, the location? From which these information come.And you see our total carbon worth. When you sum everything is 0.8 pentagrams, following line. We were really interested to contrast our outcomes scaling up from the tree degree to what various other investigators
had reported in terms of huge grams of carbon for above ground trees. this would certainly be the timber and also the fallen leave portion of what we determined, as well as omitting roots. On the left hand side. we see contrasts to Six previous researches. making use of relatively program resolution satellite data to 30 meters by 30 meters from the Landsat satellites. And also you observe our information are the black dashboard lines and the gray area around the black dashed lines of our information stands for plus or minus one conventional inconsistency on the right hand side.You just see a simple bar chart summing up the various research studies with the amount of tobal of complete carbon, which they estimate we discovered just one previous research was close to what we had actually estimated which was Nile Hannon and also Purple as well as it followed our story at least up on the
left hand side, 800 millimeters each year and afterwards and after that diverged from our data there. So, these comparisons, reflect the > > pierre hiernaux: Okay. > > Tucker: truth that you require 50 centimeter information to gauge distinct and also scattered trees. If you utilize rugged resolution satellite data, you ' re mosting likely to miss out on a great deal of the trees. If you utilize planet data at three to 4 meter, spatial resolution, you ' re mosting likely to miss out on a huge number of trees concerning fifty percent >> of the overall trees in the location because they are small Following slide. now, this is a various plot on the right hand side, but the exact same plot. We saw previously on the left hand side and also this is, this is what occurs when you have a significant information set You can plot things and also you see in the very first You have you see the overestimation of of those four? Private investigators. In fact, the initial five shown overestimation by the 4, various other by the 5, various other investigators and also only Nile Hannan is a little on the bottom of what we have and also because these sample sizes are so big, it makes it very challenging when you when you make use of simple statistics like an R worth or our square value to reveal. Well, these data agree because we ' re dealing with the research location over 10 million square kilometers next. Slide, We additionally made contrasts to numerical simulation models and also not surprising lee mathematical. Simulation designs operate at an extremely core spatial resolution. There are few which operate at a spatial scale of kilometers. Many are at tens of kilometers range. As well as if you wear ' t represent the tree cover as it is, you ' re going to have very different arise from what we have.And so that is shown in C, on the ideal hand side and also the The standard inconsistency of all of the trendy designs represented in the right-hand man plot show up in gray and as well as the factor for this broad variant is that previous studies have not used or previous modeling researches of not made use of acres exact, Depictions of the tree cover.'slide > > Greg Bronevetsky: On that one. What kind of designs are these? I suggest, what version mean. I know it ' s usual change, however what sort of dynamics do they capture? > > Tucker: Well, they are. They imitate carbon. leaving from from thought distributions of plants and after that they follow this through time. So below we ' re looking at a recap of this to see how well they concur in accumulation with what we have actually done with regard to rainfall. > > Greg Bronevetsky: fine, so what I ' m truly going for now is you have a research that has a specific price quote in a specific factor in time as well as currently we can do this from.You recognize that we have satellites that suffice from now on. >> Exactly how would you criteria? those versions to benefit from your data? > > Tucker: That would certainly be done really quickly. > > Ankit: Yeah. > > Tucker: Thanks. > > Greg Bronevetsky: An inquiry, Why? > > Tucker: Yes. > > Tucker: That would certainly be done extremely quickly.> > Tucker: You ' re welcome following slide. > > Ankit: So, this is the weight made use of below in the loss function.So generally, throughout training, we pass the weight to the loss function for each spot, the the loss feature gets, what ' s the weight at each pixel and also it ' s one everywhere other than in in these in this boundary locations which combine in? > > Greg Bronevetsky: ah, > > Ankit: So, where you have > > Greg Bronevetsky: So, you compute these these weights using > > Ankit:. > > Tucker: You ' re welcome following slide. > > Tucker: You ' re welcome.Next slide. > > Ankit: So, this is the weight used below in the loss function.So essentially, throughout training, we pass the weight to the loss feature for each spot, the the loss function gets, what ' s the weight at each pixel as well as it ' s one almost everywhere except in in these in this border areas which merge in? > > Greg Bronevetsky: ah, > > Ankit: So, where you have > > Greg Bronevetsky: So, you compute these these weights using > > Ankit:. > > Ankit: it ' s not found out.> > Tucker: We have no trouble with shrubs. > > Greg Bronevetsky: If you especially concentrate on off angle pictures, your shadows are going to be a whole lot more apparent and Height ' s are gon na be more obvious.Is that usable > > Tucker: Yes. We ended up not remarkably with > > Greg Bronevetsky: Yes. > > pierre hiernaux: Okay, but then it differs with the the Eyesights. > > Greg Bronevetsky: That makes sense. > > Tucker: We have no trouble with bushes. I mean, this in a little stand > > pierre hiernaux: What? > > Greg Bronevetsky: If you specifically concentrate on off angle photos, your darkness are going to be a whole lot a lot more obvious as well as Elevation ' s are gon na be more obvious.Is that usable > > Tucker: Yes.
No comments:
Post a Comment