 Well, welcome to our webinar, Remote Sensing Applications. Thank you to our both speakers, Dr. Wilma Zlanca and Dr. Betzabe de la Barre de la Autista. I'm just going to introduce briefly our two speakers. Will is a remote sensing scientist, recently working as the remote sensing specialist for the land-wise project with the University of Reading. His research focus on estimation and analysis of relative surface or moisture observations using Sivan satellite radar bascatter data for natural fluid management research. And also he uses the optical Sentinel-2 data for crop identification and verification purposes. His PhD focus on the observation and later modeling of passive microwave radiation from the natural snowpack at multiple frequencies. Observation took place during an eight-month field campaign and it was expected over two periods, two winter periods. And this was conducted at the Finnish Meteorological Institute Arctic Research Center in Finland. Our second speaker, Betzabe, is an Earth observation scientist using GIS and statistics for understanding the environment. Her main research interests are four. First one, to improve understanding of how vegetation responds to climatological events by exploring remote sensing data. The second one is monitoring these changes through time. The third one is to understand how ecosystem productivity is linked to environmental conditions and finally assessing vegetation conditions and land cover change. She's currently working as a research fellow in the Integrate Climate Resilience Understanding the LISP project, analyzing the impact of climate change into natural vegetation, agriculture, and livelihoods in this country. Betzabe is also involved in different projects, monitoring environmental impacts and changing changes with Earth observation data, including permafrost degradation. And finally, I'm just going to present our early career representative of the GI division, Dr. Mesjin Rasol. He is a post-doctoral research fellow at the Gustav Eiffel University. He's going to be a moderator of this webinar and myself, Veronica Escobar-Riz, which I work as a research support scientist at the University of Reading. Well, thank you so much for joining us and now I will allow Will to share his screen. I am Will Maslanka. I'm going to be presenting some work that I've done with the University of Reading on the land-wise project and specifically I'm going to be talking about my work on calculating relative surface soil moisture across the Thames Valley using Sentinel-1. And I'm going to be going through from data acquisition up to verification of the relative surface soil moisture product. So, outline of the presentation, I'm going to give a very brief introduction to the land-wise project as a whole. It's project motivations and where I sat within the project. I'm then going to talk about the method in which to calculate relative surface soil moisture. So, this is whenever I mention RSSM, I talk about relative surface soil moisture. And I'm going to be talking through the processing workflow that I've got to get to that point. And then I'm going to follow on by talking about some verification of the product. So, I'm actually going to be using talking about two separate independent data sources that I've used to verify the RSSM product. So, the land-wise project, as with every kind of academic product or academic project, they often have long acronyms with interesting names that they've managed to throw all the letters in together. So, land-wise stands for the land management in lowland catchments for integrated flood risk production. And basically, the project is aiming to evaluate and assess the effectiveness of realistic and scalable land-based natural flood management measures or NFM measures, aiming, looking at groundwater-fed lowland catchments. So, in our instance, we're looking for the river Thames in the UK. The reason that we're looking for this is that within the measurements of and within the research studies of NFM, a lot of work has gone into looking at leaky barriers and leaky dams and a lot of in-channel techniques and in-channel measures, but not a lot of work has actually gone in towards land-based measures. So, this project is having a lot to try and assess and get evidence behind the land-based measures so that we can begin to have them as evidence and we can include them in flood mitigation schemes. As with many big, with many academic projects, lots of different work packages and work pillars. Land-wise is made up of five. We've got one that's generating local knowledge. So, this is talking about talking to the farmers, talking to land owners, talking to policymakers, and those that really know the land and the areas as they're the experts in their field, if you'll excuse the pun. We then got another work package on field observations. So, this is with the UK Centre of Ecology and Hydrology where they have been taking soil samples from hundreds of fields across the Thames catchment and getting loads of soil properties. I sit within the third work pillar, which is the remote sensing and I've been looking at, as I'm going to be talking about, getting data from satellites for remote sensing purposes. All of that data then gets fed to our modellers who have been then doing a lot of the work, taking our institute data, taking the remote sensing data and local knowledge, and modelling different types of NFM measures on different catchments. That data and those results can get fed to our wet app development team who have been then basically giving new ways to show the results so that we're able to feed those back to the land owners and the farmers and policymakers and show them the results from our research to hope that we're able to actually give them information that they would actually use. So, the study area and the data that we have been looking at, so our study area is the Thames Valley in the UK and for those that are unfamiliar with the catchment, excuse me, it's the red polygon that's on the map on the right hand side and it can be summarised as the non-tidal part of the river Thames. Very basically and broadly, it's more rural to the west, much more urban to the east and you have a lot more rolling hills over in the kind of rural west area and the Thames itself flows west to east from those rural areas through then into the increasingly urban parts of the likes of Reading and Windsor and London. The data, myself that I've used, has been the central one constellation and I've been using level one infreferometric-wide Grammar Engine Tech's high resolution data at the vertical vertical polarization and I've been collecting data and I've been using data between October 2015 and September 2021, so there's a nice six-year time period of six water years that I've been able to actually have a look at the hydrological cycle across the catchment and I've been looking at the two ascending orbits, so orbits 30 and orbits 132. You can see from the map on the right hand side that the catchment sits quite nicely in between those two catchments so I'm able to use both of those ascending orbits to have a good catchment and a good idea of getting a good reliable data source. I've only been looking at the ascending orbits primarily due to data management constraints and to data storage constraints and for processing time because if we're using both ascending and descending orbits while it does double our number of available orbits and available scans, it does put our data processing up to and our data storage up to about three and a half terabytes which would take a long time for us to be processing so this is why I've been only focusing on the ascending orbits so bringing it down to about 0.7 terabytes. But a little bit going into the nitty-gritty of how I've kind of gone into this work. Hopefully you may have come across these before if you've been looking and trying to get with satellites remote sensing data but there are multiple different ways of actually accessing and downloading the data. The two that I looked at for this work primarily has been the open access hub of Copanicus which is the image on the top and that's the main hub for using Sentinel data and that's one of their main storage and downloading sensors and APIs for the Sentinel data but you can also use the Alaska build and that's an Alaska satellite facility which is the image on the bottom. Both of them have got very good user interfaces as you can see from the images and have API capabilities. Personally I've used the Alaska satellite facility because I found the downloading and acquisition a little bit more user-friendly but the Copanicus one is perfectly fine. Alternatively if you didn't want to be downloading terabytes of data to go and actually be using it and you just want to have a look at actually what the data what the actual satellite imagery is looking like and then you can use that within Google Earth Engine. That is a really good piece of kit. Personally I needed to actually bring the data down and have a look and play with it so I was using the Alaska satellite facility to actually download the data and make the data itself. Now unfortunately because I'm downloading the data in its level one format you can't actually just use the format of the data as it is especially for calculating relative surface soil moisture. It needs to be processed and you need to actually remove some of the artifacts and some areas that are in within the data just from observation techniques. So this needs to be going through a processing pipeline and so this kind of pipeline workflow I'm going to split just into three separate aspects just to kind of keep it simple. I've got one that I've called pre-processing which is going to involve a lot of the nitty-gritty orbits of the orbit corrections, calibrations, those sorts of things. I've got a part that I've called manipulation and normalization and this is where I'm actually sub-setting and cropping the data to the area that I'm looking for and then doing a little bit of instant angle normalization which I'll get into in a bit more detail and then once I've done that I'm actually able to get into this RSM calculation and actually get a relative surface soil moisture time series. Now, pre-processing there's lots of different steps to it, however very hopefully there's been other people that are actually already done this and have gone through and have got a published work close so I've been following a layout that's been proposed by the phony in 2019. There's a citation at the bottom that kind of goes through it in more detail but the very basic steps and work for processes and the kind of order is shown in the flow chart on the right hand side. This removes a lot of the geometric distortion so you can get rid of any slight variations in orbit, you can get rid of a lot of issues that come up with the border noise, you can get along with some of the stuff that is involved with the terrain correction that is involved when you're looking especially if you're looking at quite topography complex areas, luckily the Thames isn't that complex in its terrain, however it's able to remove some of the artifacts from that. In order to do that there's a couple of different things that you can use however I would highly recommend using the ESA Sentinel application platform or SNAP program as this is something that has been purpose built to use with Sentinel data. Again it has a really good user interface so it's really easy, it almost works like a GIS platform so almost like QGIS but it actually has a lot of the stuff in the background. It also because it's been built it also has a lot of the pre-processing steps already inbuilt as purpose built tools so here I've got the very easy calibration thing where you're able just to read in some data, run it through a calibration command and then write the data out but because it's able to have them as individual steps you can build them into work and create workflow pipelines so rather than having to do this all manually you're actually able to write it up as a bash script so this is kind of where bash scripting and using high-power computing is actually really important as it takes a fair bit of time to chug through a lot of data but it also allows you to set something off and let it go overnight for a couple of days for a week and that way it allows you to get on with other work while it is working in the background. Now the reason why it's really good as in so you can get on with the background is because say in my instance where I have six years worth of data you're looking at two orbits there's a lot of data that's there so as I mentioned before I've got 100 1164 raw tiles which would be the data that comes across that is any data that has been observed over the catchment area from those two orbits in some instances the catchment sits across two tiles so they need to be stitched together and then cropped. For some instances it sits in the middle of one tile so it doesn't need to be stitched so they require a little bit of data manipulation just to get the line up everything correctly. I've done that through MATLAB you could do that through many other different things Python is a perfectly good we'd be perfectly fine to do this I personally have used MATLAB just because that's my area of expertise but you're able to stitch these their tiles together and then do a bit of a visual check just to make sure that everything has worked properly and then remove any erroneous tiles or scenes because there has been some data that has been corrupted but hasn't been picked up in the bash scripting so in my instance I had one that as I said I had this part of 1100 raw tiles stitched together gave me about 640 scenes and then going through and doing a visual check there's this part of about 30 scenes that had erroneous data so I've removed those out and it gave me a total of 608 valid scenes so these would be for a given day the orbit has given me some data so a little bit now in the specifics about radar backscatter in the instance of the saw mush observation it's kind of backscatter is reliant and dependent on three separate properties you have saw moisture your surface roughness and then you have the interval of the look instance angle now the saw moisture is the variant of the dependent that we're after the one that we actually want to know the surface roughness we can assume to be constant throughout the year because of the scale that we're looking at across an entire catchment the actual any slight variations in surface roughness actually gets worked out and can be removed by spatial averaging so we can ignore that whereas the we can neglect that sorry whereas the look angle the look instance angle actually needs to be removed and we need to counter that and this can be done through normalization so I have the kind of normalization equation here where very briefly we've got a normalized backscatter as a given reference angle so this would be an angle that we would want to normalize the angle of the the backscatter to you have the actual backscatter so this is the one that has gone through the processing and the manipulation and the cropping and then you have the normalization parameter which is how we're able to actually adapt the backscatter and for this study we've used a reference angle of 40 degrees that's quite common in literature and it would seem to be quite easy just to stick it to 40 degrees and work with it from there now within backscatter normalization in literature there are two ways to calculate annual backscatter or annual beta you have a simple way which is just a direct linear relationship between the instance angle and backscatter so this is called beta d or the direct normalization and you have a complex way and this uses a multiple regression relationship so this is where it has beta r or the regression normalization now part of my work is I've actually investigated and investigated an extension to this and rather than just looking at annual normalization factors I've had a look to see if there are monthly variations and you can see from the image on the the graph on the right hand side from the pale red and pale blue curves actually there is an annual a slight annual curve which we believed to be to be part of a kind of some impacts from vegetation and when we have implemented this rather than implementing the annual backscatter we've seen there has been some improvements to rssm calculations when we've compared them to in-situ data sets and I had a paper that was published earlier this year that kind of goes into this into a lot more detail so if you want to have a read of that please feel free and then you'd be happy to hear final equation we can actually get into the calculation of rssm so this is where we've taken our normalized backscatter and we compare any given pixel to the rest of the pixels in that time series and can look at them at the largest and smallest backscatter values assuming they correspond to the wettest and driest so much of values which we've seen in this is a pretty valid assumption so it's a well documented model we actually don't use the largest and smallest values but we use the statistical model to remove outliers the citation at the bottom goes through this in a lot more detail and there's a really good paper into looking at how this rssm calculation actually works so I would highly recommend giving that a read as well but basically the equation we have our rssm value on the left hand side and it's just an index between normalized backscatter and these wet and dry thresholds or these largest and smallest backscatter values so that is all of my equations I can actually now go into map actually go and talk about what that data actually look like so here I'm showing normalized backscatter for the 11th of September in 2018 and you can see that there's in this there kind of central middle part of the catchment you've got this nice high or higher backscatter value return compared to lower backscatter values to the north and the south of it you'll also see that there's some kind of almost lodges of data that has been masked out and these are urban areas that I have removed and freshwater areas that I've removed from the analysis just because it doesn't actually hold any soil moisture value soil moisture data within that signal so I've just masked it for for ease and for necessity so we would take this backscatter data we then compare it to the dry threshold as you can see here this is a dry threshold of different calculators across the entire six years time series and a wet threshold and when we run it through the equation it gives us image like this this is then showing the kind of central swath where we had lower backscatter or higher backscatter sorry showing it as we've actually got quite a large but quite a high rssm value swath in the middle compared to two drier areas so once I've gone through this and done this calculation and done this created this time series verification it's quite key to see if this is valid this is true we start with the spatial verification so again the image on the back we've got is the rssm values for that 11th of september in 2018 and superimposed on top is two hourly animation of the two hourly data from the uk met office and it's their rainfall rate is their precipitation intensity data for the two hours preceding the orbit and we can see that it actually it spatially lines up fantastically that this kind of central swath really highlights it always cause because there's I mean it's pretty obvious when you look at it but it is cause because there's rainfall over the previous two years two hours so it means that because there's been rainfall over two hours that top couple of centimeters of soil is really wet so that shows us why the soil moisture has really increased whereas the areas where it hasn't rained is really dry because there is no precipitation so I've done this study for a number of different orbits a number of different observations and it the spatial parameterization spatial distribution of the rainfall and the rssm values lines up really nicely so looking at it spatially we're now going to look at it temporally we're going to have a look and actually see full given points have a look and see how that varies with time so for this I've used the uk center of an ecology and hydrology cosmic ray network so it's the cosmos uk network and they've got sensors up and down the country but they have a number of them across within the Thames catchment so I'm looking here at chimney meadows is one of the one of those in-situ sensors and that measures volumetric water content over the top 15 centimeters so of top the top 15 centimeters of the soil so what I've done for this is I've normalized that so that I'm able to compare a relative observation with relative observation I've applied a excuse me applied a moving average over that time series so I can remove some of the noise and then competitive a temporal analysis so you've got the on the left you've got the comparison between the cosmic race sensor which is the black trace that's the vwci and then you've got the central one data which is the rssm is the red trace and I've also included the a precipitation totals for a given day at the at the chimney meadows site using a pluvia rain gauge and you can see that actually generally there's quite good agreement and you can see that as well with the scatter plot on the right hand side that there is quite good agreement usually or the surface soil moisture and the volumetric water content in our agreement that it goes up over in the winter goes down in the spring and then comes back up again in the awesome there is an overestimation in the late summer and this is quite clear you can see in july 2018 and that's july august time where the volumetric water content decreases whereas the rssm value increases for those that are unfamiliar with the weather at the time in the uk in the summer of 2018 this was during our period of time where we actually had no rainfall over the best part of about 35 days so this is where we've actually having a look at some of the data looking at some videos of the sites and photos of the site and having a look at NDVI and Sentinel 2 data to have a look at optical data we actually think that this is due to vegetation growth and this is actually where the backscatter is becoming a contribution having an additional contribution from the vegetation and from the surface soil moisture and this is something that i'm have a going into and i've got a paper that i think i submitted last week that's looking into this in a bit more detail the cosmic ray network if you're interested in that i have a citation at the bottom that is really good to have kind of a brief overview of what the data what the network is and how it can be used for scientific data and scientific research so it's again another one that i'd hardly recommend reading so in summary i've gone through kind of a brief introduction to lamb wise it's in the workings and where i sit i've gone through the calculation of this rssm dataset i've gone through the data sources the where the the study area kind of some of the pre-processing steps how i've used them how i've done them so the manipulation and normalization of the data as well as the calculation and gone through the verification steps with the in-situ datasets from the Met Office and from UKCH so with that this is my last slide so i would like to thank everybody for listening and i think i'm going to open up for questions i've seen that there's some might be some on the chat yeah so if there's any if you have any questions please do um yeah ask them now i think thank you will for for this nice presentation and very interesting work actually i will ask audience and panelists if you have any question please go ahead before we uh listen uh BitSabi for her presentation or BitSabi if you have a question it will be an open discussion we can really go ahead and discuss ideas with Will because i have already a question well if no one have any i have i'm sorry go ahead BitSabi go ahead i just wanted to check like in your in your slide of the verification that you over post the precipitation data the animation one oh the animation one certainly yeah so just to understand that is rainfall yes so this is uh from the same time kind yes yes okay yeah so the orbits um for the ascending orbits for this area they come across at about uh 18 UTC um so what i've done is i've taken the um rainfall radar data across the UK and across the Thames catchment i've taken it for the 24 hours previous so i've gone from um 18 UTC on the 10th of September across the 18 UTC on the 11th and i've done some kind of spatial analysis of looking at when the uh how much rainfall falls in the area compared to this kind of spatial patterns and if they match so this kind of central swath uh really is highlighted over rainfall that happens over the previous two hours um you can see that there's a to the northwest of the plot of this image you can see that there's kind of like a higher um not quite a strong return but there is still quite a higher rssm value area compared to the areas to the south so i'm trying to think if i can roughly in this area to the um northwest um that area has actually had a uh rainfall that's passed over in the three to four hours that's happened and that rainfall has then since passed out of the catchment so this is why there is rainfall there is kind of still high rssm values there but not quite as high as where it has just rained yeah it's had a bit of time to sink in a little bit of evaporation a little bit of infiltration um and it's removed a little bit of the um that's kind of surface almost from the time and have you measured just for curiosity like what happened after the rain as in like how long does the sign now so this is yeah so we yeah so i'm able with this data i'm able to have a look and see what it would be like tomorrow so that i can see what you put one day after so it's had a whole day to go through um one thing that i've not yet managed to have a look to see exactly how much how would the rainfall of time as it would progress um that was something that as part of the project we had hoped that we would be able to go out and do but due to technical problems and with COVID restrictions we weren't actually able to get out into the field to do a bunch of measurements i know that Veronica has lost hours on building kit for um some stuff for this project that hasn't actually um that we weren't able to use because of restrictions um but we're aiming to yeah there's something that we would like to do and something that i think would be a fantastic piece of future work to actually have a look and see how the infiltration rate would work in relation to the soil moisture observation so we could see if it's rained six hours ago can we determine okay well that means then that we're going to have lost 20 percent of our relative surface soil moisture or something like that we can get a direct relationship between the two yeah we could say it would be really nice to see exactly how long does you have the signal for yeah that's really cool thank you so much it's really nice okay well thank you so much will um can I ask you just a question in the next slide I think it was the next one yes did you think that error that you were mentioning in July is related to your um normalized parameter in the RSSM or so we see this increase regardless of the normalization parameter that we use so I've done this analysis so you can see this one is for it says up in the title of the um the kind of busy line graph um that is too many memories regression monthly so this is where I've been using the beta R the monthly beta R I've done this analysis with uh annual monthly normalization factor annual uh sorry the simple annual the simple monthly the regression annual and the regression monthly so I've done it with as a kind of a combination um and this is where we see that there's this this kind of increase over the summer is prevalent in all of the data in all of regardless of the normalization factor there's actually in the literature and in with other people that have done this work at a more coarse resolution so this this data is done at a 100 meter spatial spatial averaging I've done this also at 100 meter 250 meter 500 meter and a kilometer grid resolution which is the traditional one that's used for this method um you see this this overestimation over the summer in all of that data and you see that it's it's a documented thing that is something it's kind of like a um cutting edge a bit of trying to remove this um vegetation index and it's a it is definitely because of vegetation growth over this time um and in some of the work that I've done um recently I've got different been looking at the soil moisture values for all the soil moisture return and the backscatter return for different crop types um across the Thames Valley and there are a number of crop types that actually due to the geometry of the fruit and the pods give a massive backscatter return because it massively increases the scattering um in the kind of canopy layer um so it's definitely this this increase is definitely due to vegetation rather than a um parameterization um there is a slight discrepancy because the the cosmos the volumetric water content measures over the first say 15 centimeters and the surface soil moisture measures over the top two centimeters so there is a slight lag but we've kind of removed that in part with the um moving average that I've applied across this the data so we can remove some of the noise and remove some of that discrepancy but yeah we definitely think that the this kind of increase over the summer is due to vegetation because it's something you can see annually uh you see it in you can see the roughness of the the vegetation roughness uh it's more it's not so much it's the roughness it's more that the vegetation itself has an additional contribution so it's then whereas we have over the winter when the vegetation isn't there the contribution is is is solely from the ground over the summer it is now a combination of the ground and the vegetation and it's that extra vegetation contribution that is in causing that increase okay okay well thank you um thank you a lot for this presentation very interesting that's my pleasure I have I just put in the chart uh your most recently paper if someone wants to check it which is related to this uh presentation okay well thank you thank you a lot I have a question very unique before you uh move with the betzabi well thank you well for the for the work very interesting actually what what interests me it's the part that you're doing the the two different seasons like for example dry and let's say rainy but what is relevant what I did before in term of seasonal chains of the water content which is poly channels uh I also used two different sessions seasons like for example during the summer and winter but not exactly during the rainy season after the rain as betzabi was discussing with you uh so the saturation is more uh as you you can know that and now I can prove that that the situation in Barcelona that time and what I see in your graph for the chimney midway so for the regression uh during the january to april you have uh the the peak period I guess it's all the years yes which is interesting because I had very similar but I was working with a ground penetrating rider uh with around 30 meters with low frequency what here I have a question regarding uh for example you were talking about box scattering yes so what uh you already answered very unique in term of the combination of two reflections that happen the vegetation and the soil uh characteristic let's say but in in in my case I had the active and an active uh underground streams let's say poly channels in this work you were only working on soil moisture or you were working on the water content also or you are planning to do something related to the water content to see the active uh poly channels or something like this with this uh what's so this work we're aiming to have so because I'm looking at the surface I'm only looking at the first the top two centimeters of the soil um so anything below your your ground penetrating radar is going to see a lot further into the ground than Sentinel will or Sentinel one will um so we're really looking at kind of that first that top couple of centimeters so that we can have a look and see how much we're aiming to have this data so we have passed this on to our modelers so they have a um quite a long a long time series of what the soil this that kind of top centimeters of soil moisture so we can have a look to see how that infiltrates because we're aiming to have a look to see if we can through changing farming practices um whether that be through changing crop cycles or changing um crop management schemes so if this we need going to be increased um plowing if increased um traffic on soils increasing or decreasing the compaction if we can alter the infiltration and alter the uh the actual storage of water in the soils on the on these farms um so hopefully we'll be able to then from that and we if looking at this is some stuff the models are currently doing enough in the process of running through their hectic models um when we have these this um when we have this rainfall that's passing through we can then try and keep the water in the soil rather than keeping it from either sinking down into the kind of active channels as you were saying or running off into um kind of the river network and the river courses the water courses themselves um because this then if we have a lot of if we were to have somewhere that has got quite a lot of compaction and there's not a lot of infiltration that's going in what that's going to do is just have a lot of water runoff and that means we're going to have a lot of soil erosion there's going to be a lot of pollutants that go any fertilizers that have been put onto the soil if it falls onto compacted land that water and that uh overland runoff is just going to wash that pollutants away into the water courses so we're going to increase a lot of pollution so that's kind of what we're ultimately hoping to do and i'm this this kind of surface centimeter first couple of centimeters of soil moisture data hopefully will then help to feed them feed the modelers with their anecdotal evidence and their anecdotal data so they can spin the models up so we can see okay we know what it has been let's see what happens in the future let's see what will happen okay thank you very much thank you very much um well should we now continue with Betza there but anyway uh well first thank you for the invitation to the seminar and uh thank you for a great presentation and i'm gonna talk so i'm Betza from La Barrera uh i'm working with all these people in in a project in the swedish arctic and also uh it's part of another project that is working with Peroni Kanki and the artists as well so the title is the mapping of permafrost in the in avisco in the swedish arctic with earth of the observation data so we know that the arctic is warming rapidly even like faster than any other part of the on the globe it is warming two or four times faster which this is bringing lots of consequences um so one of this is the throwing of permafrost which permafrost i'm gonna tell you a bit more about this but it's a permanent frozen ground and that creates subsidence in the ground and also the formation of thaw legs uh that also like with this uh subsidence you have an increase of the greenhouse gas emissions and also it impacts the the infrastructure of the of the that of everything that is above the permafrost and this increase in in greenhouse gas emissions has an a strong positive climate feedback resulting in greater rates of warming for the globe so it is a quite important topic and um and this is just for you to know what is permafrost and what is the active layer that is that i'm gonna talk about this all the time so the active layer is that this layer that i show you here that so like it goes as in it goes each summer and it has a lot of biological activity and the permafrost which is let's say this like this this layer which is the permanently frozen layer of the ground and it contains a high carbon so this permafrost is what we are trying to understand and what is happening and what research is showing is that with the ground temperatures uh increasing then you have like a thaw like you like sorry the thaw is great greater during summer so the active layer actually gets bigger and in some areas the permafrost is completely lost uh on the other hand the satellite data shows that the that there's an increasing on thaw lakes because you can measure lakes in the Arctic and that indicates area that that areas where permafrost has been lost or decreased and also you can see a lot of increases in moisture in areas where permafrost is losing and there's a lot of like they are really scary uh projections that the 30 to 70 percent of permafrost will be lost by 2100 how do you say 2100 um so when it's so the idea is to to measure this through remote sensing right but like normally like you cannot like the the traditional methods are field work methods and they cannot like cover the whole the whole Arctic and and you cannot quantify the rates and the extent of perma there of permafrost thaw like so what we are doing is understanding this thaw by subsidence using different methodologies and I'm gonna show you in a bit and this these technologies are a combination of optical uh SAR and drone data so I'm telling you this because well SAR is incredible useful in these areas because you don't have like the problems with clouds but also it is really important the optical part because you need to find and understand what is happening with the vegetation there and sometimes like in this case we are using drones to have a really high spatial resolution so the objective is just as I said is measuring the subsidence which is a vertical movement of the ground and uh and use vegetation changes as a prox as a proxy of this term of permafrost thaw and also we are trying to monitor methane emissions as the consequences this so this is just to tell you about um something really important to understand why the changes in vegetations are quite important for this area so this is like more or less how it works in the swedish Arctic and so you have palsa which is like a raised wetland and what it happens is when when you have permafrost below and it's intact you have a type of vegetation that normally is drier when they when the palsa collapse is because the permafrost is losing so you have a uh you have a decrease and it's this this kind of thing that you see here so you will have more water on the surface and that will create a lot of like biological biological processes that what we'll do is to start like changing the vegetation type but also the moisture levels so what you will have at the end is a much wetter vegetation and this picture shows like how more or less is the these transitions and this is really important because this type of vegetation emits as we have found that it emits lots of methane so just to show you where are we working is in the swedish Arctic in three areas in in the tourist station store flakket and store dalin which are like wetlands in this area and this picture is how it looks in reality and i haven't gone because of covid but like the swedish and actually veronica went there um so what we did again we used drones optical images historical orthophotos and sentinel one in sar and around the field data we used active layer depth uh emissions methane emissions and land cover so with that we created land cover maps de ends and we tried to understand these changes in the pulsars and understand it as well there subsidence so i'm going to talk a bit more on the i'm i'm dividing this into the different methodologies so that you can see and so first the orthophotos so we managed to get ortho historical orthophotos for the areas these look like that so you have an orthophoto here i just put them like the main ones there uh 1960 2008 and 2018 and this is just the zoom in of an area of a pulsar so you can see that this area is a rice area and this area the one that it has collapsed and this is just an example that how is retracting on time and uh this is another option and this is how it looks in reality a pulsar that it hasn't been well like how the lateral collapsing is looking and this is how it looks after collapsing uh then we used drones and we got uh flights that were planned in the area we had uh flights in 2017 and 2020 and 21 we had onboard the drone a multispectral camera and rgb as well and uh and then with px4d software which is a software that helps to kind of do all the like ortho rectifications and create mosaics of all the images that you take in the drones we use that to create then uh our mosaics that were the ones that we used to do vegetation maps with that we use a supervised classification in this case was super vector machine and then we also create digital elevation models that to see the difference between the 2020 and 27 and 2021 we didn't because um we just have for uh 21 um some of the data and so then we used in star which in star um i think you all know what is in star so i'm gonna uh but what it what it does is like it allows you to to measure changes in the land surface altitude so what you do is you have bunch of insert data that you layer all together you stack and then you create something called interferograms that it will tell you how the ground is the form from time one to time two and uh this is this is the way that we are doing we are doing a technique that it was that it was developed by terra motion which is a company and a collaborator of us where they use apsis technique that also i can give you the reference for good papers of that if you want later on and then we did the the in the field we measured the active layer so this is um this is literally you take us like a like stick to do like how do you say like you make hole in the ground until you uh until you get the permafrost and then you see how much is the active layer and how deep is the permafrost so we did that well i didn't like my colleague did and uh it was during the summer and the idea to do this it was across different vegetation types so that we can see if there was any difference of different vegetation um and now i'm going to show you really fast what we need to have then i'm just conscious of at the time but like um so this is the orthophotos so as i can uh as i've told you we digitized the the palsa mire like the edges of the palsa it was digitized by three different people so that we can reduce the error because it's a completely visual interpretation and in those areas where the where the difference between the three observers were uh not bigger than 50 centimeters i think uh we took that those areas and then we measured how we're changing on time so in these lines you can see uh 1960 2002 2008 and 2018 and you can see how uh the red line is 2018 and the black line is to uh 1960 you can see how much the palsa has been shrinked and what we did then is to have this a to have these rates of rates of shrinkage so what we did is just to calculate the rate of the of this collapsing and then we have found that actually from 2002 like the collapse is much faster which this is in this we found this paper so from these two guys in 2016 that they found that the the temperatures in ground and ground temperatures were increasing faster from 2000 2012 so sorry for 2004 it started like growing faster so then that that coincide exactly in that moment that we that we saw as well like like a rapid shrinkage of the palsas and now what we got with the orthophotos is a like the the land cover classifications and as i've told you these are the results and sorry they might just need to some scale and north and other elements of the maps but so i just wanted to show how they looked the maps and it was just really interesting to see for example like all the permafrost lost which are in these three types of vegetation and you can see that such wetlands are these like like really bright green and these are all the areas where the palsas actually collapsed and if you have if you take this image into your again keep it on your mind because and you see all these areas of uh of the bright green that they are the areas where the permafrost has been lost and you can see in another image but i will i will tell you that so what we found is like if we put these vegetation types and then we combine it with the active layer that i that i've told you that we measured we we found that the raised pile palsa vegetation types which are the ones that are more intact have a shallow active layer which means that the permafrost is still higher however in the fen vegetation types which are those areas that i've told you that are in the in the surroundings of the palsa you have a really uh deep active layer which means the permafrost is higher than 140 centimeters so again this is another another evidence that the permafrost in that areas are are like sinking and now these are the results from the from the drone data but the part of the digital elevation model i'm just showing you how it looks the digital elevation model from the processing drone images and you can see exactly where are the areas where are higher in the palsas and lower in the palsas and then what we did is the difference between 2020 and 2017 and we found and again as i've told you to remember these areas are areas where they are shrinking and they as in the the palsas are collapsing and these are an indication of permafrost though in this area and finally like the results with the insert data so we wanted to see like how actually the insert data will do it because we have like a really high spatial resolution and this is with insert which is as in in the drone data you are talking about centimeters and the centinal one you are talking about 20 meters by 20 meters pixel so we were just checking what were the differences and this is just to show you how it looks the the insert data and again even the insert data is showing how the palsas are subsided in the in this area we found difference between the two methodologies but it's obviously like this the spatial scale that we are looking at are quite different and in this case we are this is the average surface surface motion in three years and again we found that the the the raised palsas are the ones that are subsiding faster so that indicates that the insert data and this data is telling you what is happening now like a rate like it's showing like a like a rapid change of permafrost in the area and I promise this is the final one we just did the the methane emissions and this was measured in June 2021 and we just found that the areas where where that where the permafrost was lost we are seeing more methane methane concentration so yeah this is like the combination of all these methods is a great opportunity to to monitor permafrost you can have a panartic approach you can detect rapid changes and because this is really important to measure millimeters of this because permafrost is losing all over the place in different ways and different rates and all this data gives you the opportunity to combine and understand better the subsidence and yeah this is really important because yeah it is affecting people livelihoods global warming and we need just to take action and to understand better and to create better policies to conserve these areas and I think that's all for me thank you so much thank you so much that's a very interesting um that's how that presentation was related to her recent paper and I just put it in the chat if someone wants to check this paper is very interesting um if someone has any questions um please do it I do have a question if you compare in terms of the extension of the degradation of the permafrost between the inside and the drone did you see that the inside actually uh overestimate this permafrost degradation or uh over or underestimated yeah that's a good question because we found that like inside was uh yeah I would say overestimating uh some areas however and this is something that we need to understand we don't know if it's actually overestimating or the drone is underestimating it uh because what we what we found is like in all that areas that are like in the mires that are let me just show you a how do I go back sorry I just wanted to show you the um the picture of the where I so in these areas like the inside is detecting a lot of um subsidence and the drone doesn't but these areas like uh like Sophie went there and and she says that you can see like like that it was it is collapsing and it is collapsing and it is like uh quite recent collapsing so what we are thinking is that maybe also like uh inside is telling us like maybe a rapid change that you cannot that you cannot actually detect with drone data and also because you have all the time serious at the end and also with the drone you use a different technology lighter it's true no then probably no we didn't use lighter which that's a problem because we just choose the DMs from the flight which brings a lot of error as well because obviously you have all the artifacts from like yeah wind and all the external kind of things that you have to deal with and obviously you try to do as equal as possible one flight to the other one but there I think that's something also really interesting that you have I think it shows those differences but the error is bigger in the drone data than inside data okay that's very interesting it's very good to know yeah and it would be really sorry it would be really interesting actually to use lighter and to see that but the predaginates like lighter in different times as well and that's really difficult that's another project yeah anyone has another question two bits of it no well I think we are a little bit over time then I am just going to end the webinar thank you so much for both speakers presentation and I'm just going to say to record the first part and I hope it does goes well but just free feel free to go if you have any plans or you can stay for just five minutes of the recording but yeah thank you so much for participating in this webinar because you have something to say actually Veronica maybe you can do the first part for recording that will be good well I thank you all of you first of all actually for organization of this webinar Veronica it was a big if for thank you very much thank you Will thank you with Sabir for very interesting presentations and work you are doing I wish very good luck with your work and I hope you can find some collaboration work all of us to do some stuff together in future and thank you again