RUS Webinar: Mapping waterbodies from space - HYDR01

okay so good afternoon everybody and welcome to the seventh webinar in the rich series my name is Bertha and I will be guiding you through this webinar so this is a special edition as it is given as a part of the second international electronic conference on remote sensing and moreover the topic was choosing to match with the conference on having water bodies from space it has been held last week in Ezrin Prescott Italy and so as you probably all know since you're attending the topic of this webinar will be how to map water bodies using optical and SAR data and how much service can use it can help you when you when working with Sentinel data so this webinar is being recorded and you will be able to repeat the exercise by yourself I will explain the details a little bit later so let's start one more thing I would ask you to post your questions into the greatest question staff immediately when you have them there's quite a lot of us today so in order to not have such a bag of questions for the Q&A session that we'll be following this webinar I would ask you to post them immediately and me and my critics will try to answer you as soon as we can okay so first let me give you a quick overview of what we are going to be doing today so first I will tell you a little bit about the study area then I will tell you a little bit about the data that we will be using then I will give you a short introduction into the route service then we will move on to the exercise itself and then we will have approximately 20 minutes to the Katori Q&A session so this was the total duration of development of this webinar I would expect one and a half hours okay so first the study area so as this topic is the mapping water breeze from space we have chosen as our study area has the missourian Lake District in northeast of Holland and this district is computers of more than it's shaped by mostly by glaciers during the Pleistocene Ice Age most of the lakes are are quite small and shallow and some of some of them have fun coastal areas covered with vegetation as you can see on the pictures okay so for this we will be using Sentinel one and sent to the two day does so first few words about 1701 the mission comprises of two polar orbiting satellites in the same orbit face at 180 degrees to each other their main Sentinel 1a and Sentinel one be they both carry the identical active sensor and Betsy and they have very short drive is it - time approximately maximum 4 day 4 days at the equator however for the acquisitions to be done in the same geometry meaning the same orbit it's six days with the two satellites and 12 days with only one satellite they provide all weather day and night acquisitions and four different imaging modes with different swath and resolution today we will be using dead interferometric wide swath which has a resolution of 5 by 20 meters but resembles - pixel size of 10 meters and swath width of 250 kilometres the second satellite that we will be 6 second source of lethality will be used today it's sent in YouTube The Sentinel 2 is an optical instrument or carries its optical instrument that acquires Dateline 13:30 month again it's a mission it has it comprises of satellite satellite Sentinel 1/2 a and Sentinel to P again they are in polar orbit face 180 degrees to each other as I said before the data are wiping 13 spectral bands with three different resolutions so the basic vents are optical bands so blue green red and infrared are acquired at 10 meter resolution then we have the red it's shortwave infrared vans it's 20 meter resolution and then the atmospheric that's very oriented to atmosphere composition at 60 meters approximately a visit time is 5 days at the equator ok so now if you are worried about the route service so route stands for research and user support for sentinel core products and it is an initiative funded by the european commission and managed by european space agency but the object is to promote the adaptive Copernicus's sentinel data and support R&D activities the service provides free and open scalable platform in a mmm when the powerful virtual machines that hosts the suit of open-source tool boxes that are customized to the users requirements so these here you can see some of the toolboxes that the virtual machines by default include which could be snapped QGIS are or Python and many others the virtual machines are usually accessed through a browser which I will show you a little bit later ok and they should be ready to use of course for the users and it can be customized to learn as I said before its post source of open source tool boxes that are pre-installed on the virtual machines but also we provide a specialized remote sensing helpdesk to answer any questions you might have about processing of the latest selection of data for different applications and so on we also provide webinars as you know and face to face events and various conferences or organized by ourselves to teach users how to use Sentinel a Copernicus data ok so I will just show you two and websites that are related to this project so first you probably already know it's the restraining the page where you have registered for this webinar okay so when you access this website you can you see that you have two options you can go to training or eLearning let's have a quick look to the training you can see upcoming and past trainings so for upcoming we have at the moment some face-to-face event that we are that are being organized such as next week 11 each you in Vienna or next event where we are participating in the trans-atlantic course organized in zagreb so you can also find more information about these events and how to register for them and so on also at the moment we do not have any webinar here and in the list but if you go to the past events you can see information about our past webinars including already this one and only access the page you can see the Ricoh you can find the recording of the webinar and given a document that was where all the answers to questions that we thought were interesting or important during the Q&A sessions will be listed and also the questions that we didn't manage to answer during the short Q&A sessions that we have so you can find all of that here and you can replay the webinar and in addition you can repeat you can find information about how to repeat the exercise you can also access the e-learning page that we have which provides courses at the moment only for SAR and so it's an introduction series of courses for introduction to SAR remote sensing with quizzes and lectures so I would invite you to to try to participate in our learning course the next page that is also very important is the rich Copernicus dot-eu and that is it a page where you can request the computer can you register and request for the free virtual machine that I have described before this is the layout of the page you can find out what verse is who is behind it and how does it work you can also find information about the computing environments and trainings and so on so if we go to computing environments you can see a full list of tools available on the platform and you can also see the information about the ICT resources we provide so we generally divide work environments to three different levels at level a B and C and they live in number of processors disk space and duration for which they are provided however it always depends as I said the environments are customized to the users needs so it always depends on what your project is about and what you require to do your research project or just repeat so on so if you do not have hours account yet you need to click here and log in register and go to register so account but you are ready to you to the ISA CD ssso single sign-on registration where you have to fill in your details and then of course click register and following that you will receive three separate emails so only after these three emails you are fully registered and even access to the page so remember it's three emails not just one unfortunately is your registration needs to be checked by a human operator it might take up to 24 hours of 48 hours so once you get your activation email so the third one you can go to the login and register again and login into your new account so I already have an account so I did not register but I'll just look and once you log in you can see that you have a new tab here that shows you your profile your training and also your dashboard which is the most important part that you will be located and here you have a list of virtual machines that is that have been created and are assigned for you I already have two of course you as a new user will not have any and you can request a new user service using this button so I'll just walk you quickly through that how the request works and so we will have three sets of questions for you it's quite quick and so the first question would be how many years of experience in more remote sensing do you have not just to choose some date I think might not remain empty then have you already narrowed in copernicus data via copernicus open access how yes have you already handled Copernicus data yes and this is an important part so you wish to practice within Russia virtual environment a tutorial exercise exercise shown in rusev inner or an aversive training material so at the end of this webinar and for those who have already depended different are others you know that at the end of this dinner you will receive code that you can input here such as for example for our last the venerability code was non zero for and so you can input this code to let our helpdesk know which data or which women are you wish to repeat and we will upload the data related to this weapon on including a step-by-step guide to your virtual machine so then you have to choose if you only want to repeat the tutorial exercise and you don't need any other datasets or tools or if in addition to the exercise you also plan to use your virtual machine for another project research or R&D activities and you require additional copernicus data please so for us let's say we require additional copernicus data and that's click Next here you have to define a little bit the project that you wish to perform so for us Hitler's fear what kind of operations you wish to perform unrests you can you have several levels of difficulty of tasks performed for us we will say basic processing únicos data using worse tools then you have to select your preferences for debate about downloading process so we have two options either you can use salt downloading you can download the paper yourself with our support or you can use the route downloading surface services sorry so if you prefer this move on the roof operator well handled download the data on your behalf and preload them on new virtual machine so for us let's choose self downloading and then you need to describe your foreseen activities and support needs on the virtual machine so you can have for example mentioned the p.m. size you prefer around that you prefer specific tools and software that you would prefer to have on the virtual machine and then I'll help this process it and let you know if this is zero then you just enter the name of your project so I don't know if I'm doing it in here but since the field cannot remain empty I will just select this so this is the last page in the request for the virtual machine here you need to specify the earth observation data that you need first what type of Earth Observation data you need so for us it will be system 1 and system 2 you may or may not if further details of what kind of data you and I want to use then you need to define your region of interest so as I said we will be looking at northeastern Poland and then you can also import this rigid region of interest as a shape file with maximum size of 5 megabytes and on on your shape file so it has to be a zip file containing a single shape file containing a single polygon so it's a bit complicated but actually not so much so and do we want to perform the multi temporal analysis yes or no so for us no and any additional specifications that you might have about your projects or the data that you require for it ok so now we click Submit in review so here there is a summary of your entire request so when you are submitting this request you should check this request very well then you need to agree to our terms and conditions which are generally simple I would invite you to I would invite you to have a quick read through them and then you click on submit request I already have virtual machines as you have seen so I will not submit here's this request actually so I will now go back to my dashboard and once your virtual machine is issued you will see as I do here the table where you will have one virtual machine and you can ask for support for your processing tasks if you have any specific questions about which they have to use how to process them and so on and so on and you can close your surface of course and you can of course access your virtual machine so for this turbine we will be using this one I will go through okay so once your virtual machine is issued you will receive the link and the login details to your email okay now when you access when you login to your virtual machine you can see in your browser basically a remote desktop and with all the tools listed launchers and on your on your desktop and then we can start our exercise so first step we need to do and the exercise is to of course download the data and so let's go ahead so we will download the data using them or carnivals open access huh so if you click on the browser and oh here you're automatically redirected to the Copernicus open access hell and again you have to download so Sentinel data is free however to download them you have to have an account on the Copernicus open taxes huh so I would for those who do not have it I would advise to to sign up few simple details the registration in this case is immediate again you have to receive the activation email but that shouldn't take long and once are you registered you can go to login and we can start to search for the for our data so as I had mentioned before we will be using witches and large do a little bit so we will be using sent in one and central to data so we will be using data from two different satellites we will be using one single product from The Sentinel to and why if products percent for cents in a while so let's start with looking for for our data so we can zoom in to the approximate region of interest which is right here to move the map you can click on the pan vault and then you can use the books to define your area of interest once you've done that you can go to the search settings here where you can set the same thing period so for us first we will look for a dissent in a month later and those will be for the two summer month so from the first book July to the last to the 31st of August so of 2017 sorry so let's set these days so July to August 31st 2017 ok then now we have to select which mission we are interested in so here you know you can find it up for three different missions of cetology two it's three and let's checks into the water for this moment so we know in our case that we want to let me look at they look for data from Sentinel 1:8 satellite so I have of course pre-selected this data previous to this exercise so I know what they type I am looking for the product type but we the ground range detective sensor mode will be the parametric rights law and I also know which related orbit I'm looking for you can find it a relative orbit I will show you how to find it in a second and this ensures that all the data that you get the results will be in the scene geometry okay so let's search for our data and here we have the results so it all looks like it's a one result but actually we have five different data sets we need to download all five of them so we can add them to our product card or we can just click download here I will also show you how to have a look at the properties and where to find the relative orbits information so in this sort of product overview you can see the location you can see a quick view and its slopes and you can also see product transform information and so on and so on and usually the ability for bit information is right here and this ensures that they have accurate uploading our from at the same geometry so generally when you're downloading from compounding who's open access up you can only download two products at the same time and you can you have to therefore always go and click one by one you can also add your products in the cart and use downloading utilities such as coal or area to download this data I will not be explaining this during this dinner however there is a very useful tool for downloading the data and the use of this tool is explained in the step-by-step guide that it's provided when you want to repeat this exercise so you can find all the details there however we have a limited time so I will not show how to how to do this so generally you would click car and then you can go and search for the Sentinel to data so now is search for our sensor data and % - no - I mean as I said we will be using on a while single image and this image has been acquired on the 16th of August 2017 so we need we only if you know if we always specified that one single day so August 16 also here okay then I need to unselect Sentinel model and select center - mission and I need to select a level 2 product here and the level 2 product has already been radiometrically and atmospherically corrected so the level 2 products are only available for as of now they are only available for Europe as of April 2017 operationally however if you need to perform atmospheric correction on product that is Gosai or if you need to process a productive outside Bureau or previous to this date you can achieve the same level that using the central core processor which is freely available in step again this will not be a topic of this webinar but there you can find the steps in these step by step guide ok so we want justice sorry this we already have some power selected and we can click for search so in our search here rather a lot of results we will however be interested only in this one so the setting of two is provided in tiles and the tiles are cut out of the original swath width so new tiles have one hundred by hundred kilometres size and the one that we're interested in right here so if I click on it it will be highlighted here in the in the results section and again I can add it to the cart okay in this piece I'm not interested in all the other ones and then when I want to see what I have in my cart I can go to the profile click on cart and I can see that I have six products in my card so that's correct and then I can either as I would do now download them one by one or I can download the cart which will create products meet the for a file which is used as an input to the tools that I have described before so at the moment I will not download the data because it takes quite a while there I love rather large as you can see and it would take a while to download and we do not have the time during the webinar so I will now go back to my training folder so usually if you're repeating the exercise you will know that the training folder is placed in shared so just the name of the folder is shared training and whatever the topic and name of the tutorial is and I can see that I have all my data already pre downloaded so it's five sent in one images and one since no two and now we can start with our processing so we will do our processing in snap we in the rich platform generally we use the latest version which is 36 so let's open snap and first little process a second or two data so the processing will be done separately on centinall two and no mana of course because there are very different types of data so first we will start with the optical data center - so let's open our first data set which is original here so and let's first have a quick look and the structure of snap so and the top you have a product Explorer window the bottom you have navigation window which is very useful which is whole screen view so you have a better view of what I'm doing and let's have a look at the products structure so as I said before Sentinel to product consists of 13 bands you can see that here is only number 12 but be aware that there is a - a - a - a - a which are both near-infrared man sounds like a different wavelength and then you can also see that it cool or it contains masks for us the most important part is SCL masks that are created during the atmospheric correction so during the processing to the level 2 days left and they already you can see include auto mask as well as a cloud masks so low medium and high probability of clouds and thin cirrus masks and many others we will have a closer look at them in a second so first let's have a look at our data so first let's have an open let's open an RGB image window so you right-click on the product and go to open RGB image window normally we could use the natural colors however in this case whatever these are better differentiate it using false coloring correct view so let's use and profile let's change to false color infrared and you can see the depends are now are no near-infrared red and green okay you still take a few seconds because it's not the product is very large and there we go so we can see that the tile is rather large so it's hundred by hundred kilometers our area of interest of course is mostly in this area of the district and you can see that there's many many lakes with some very small narrow under bit larger imaginary they are quite shallow Lakes and no let's have a look at the masks that are provided with the product so let's go to the mask folder here in the products is through that Explorer and you de sel masks and let's have a look if I did a good selection and if I selected the product well because the goal of course was to choose a product that's contains no clouds so let's have a look at the clone high probability and medium probability and paint series so you can see that there is some pixels that are classified as cloud in the cloud masks however you can also see that they mostly correspond to the lake coast or two roads and so on and they do not really look like clouds so they are mostly of course it's always on a closer examination but in this case I'm confident that all the pixels that are here classified as clouds are actually misclassified very bright objects such as each's wrote tarmac surfaces and so on and so on so there is no feature none here I would assume would be a cloud also in the high enough ability there's much fewer pixels and if we compare with the RGB view we can see the thing of the crest want to cities and roads and beaches and not to and any cloud features also we cannot find anything Cyrus series features so it seems that I have done a good election and we indeed have a cloudless imaged so now let's have a look how does the default included what you must perform so we open this ok so here we can see that while the big water bodies that are captured very well we can also see a lot of misclassification especially in cities so build up area like such as this one seems to be quite highly misclassified so maybe let's let's investigate if we can if we can do better in it without masks ok so let's now proceed with the actual processing of the data so what do we need to do in order to derive the cloud map let's recover water masks from a sentinel to imagery so first as I have said before a descent into two data are not all all depend not all the bands have the same resolution so we have three different resolution resolutions and unfortunately the snap tools generally require all the bands of the data set to be with identical resolution so first step that we need to do is we sample the product two identical resolution so you can do this by going to raster sorry first you need to select or highlights the product in the product explode then you can go to raster geometric operations and resampling okay so we can read the name here if you do not select here say that the product is now not going to be physically safe but it's only going to be a saving into virtual memory or virtual product and it's going to be deleted once you close snap so in our case that's that's okay so let's leave it like that and then go to the resulting parameters so we have three options we can resemble my reference values from resource product or we can resemble by my target width and height so by number of pixels to which the image supposed to be divided or my picks resolution so such as 60 meters 10 meters or 20 meters or any other value 15 meters for example in our case we will just use reference band and we will use number 1 liquor spend 1 if you remember from the table that I have been showing in the beginning has a 60 meter resolution but actually for us we will be using only data that are originally tending to resolution so we will resemble to the native resolution of that to which it is 10 meters okay and we can click well and then you know product will appear in our product Explorer so now it has an index number - agreed close this deal and now I am a product here which has this to fix for samples so that's my resemble product in which all the batteries have been identical resolution if I visualize it it's going to look exactly the same but for other purposes let's just quickly open a view so open RGB window and then now actually if I click on profile you can see I have many more options and this is due to the fact that now all my bets have the same resolution so I have much many more combinations that I can choose from before only some planets were available or had exactly the same resolution and I can only create an RGB image if all my bands or input have been and have the same resolution so now I will choose this one which is length Walter it uses the near-infrared shortwave infrared and red bands and gives us better super ability of land and water ok so now we have the image and the next thing is that we need to do is a subset so obviously we would not want to in this case process the entire image but we would like to subsidies to the area of interest and we can do this by again highlighting the product here and get it going to roster subset and here we can choose from three different ways so we can do a special subset band subset or metadata subsets for us at this point we belong the new better spatial subset and we can use pixel coordinates since we are subsetting one single product pixel coordinates are okay however if we were sex except setting multiple products in different geometries we would certainly need to use geo coordinates to subset the product in a lock lock okay so first let's select some values here and you can see that once I click here the blue line here defining the subset is going to be moved so it's showing me where my subsets my current subsets as I set it so this is my final subset I can also see the estimated size and I can click OK again this stay done now is not physically safe on your disk is only your virtual product and if you want to do any further operations with it you do need to save it physically so let's if you would need to do that so let's do that and right-click on the product and click save product ok so here I have warning that this might take a while and again it will let me choose a specific folder training and we are processing center tube so I was citizens not you and I won't say the product here again as you've seen in the warnings can take a couple minutes maybe or minute so I will not actually do that here because that might be quite boring to you and I will just cancel here and actually open the product that I have pre processed and saved for this purpose okay so now here we have the product as if saved this one so I can have a look still has my 13 bands masks including clown masks everything as we had before and now let's calculate let's create more masks so in this case we will be using the normalized difference water index and the ndwi is widely used then ratio that has been developed there any open water features and enhance their presence in remote systems with optical imagery the NDVI makes use of reflected infrared radiation and a visible green light to enhance the presence of such features open water bodies while eliminating the presence of soil and there are three of vegetation features so let me just very quickly show you a slide that I have to better illustrate this go so here you can see on this graph you can see the general reflectance of or the reflectance in a spectrum of their soil vegetation reflectance profile of soil amid the green vegetation and water so you can see that the highest reflectance of water is generally in the green part of the spectrum while in the near-infrared majority of the radiation is absorbed and therefore it is almost a zero reflected radiation of course all is dependent on the water constants so for example if you have vegetation in the water or if you have high chlorophyll moment I have high turbidity and so on and so on so the normalized difference water index uses the green part of the spectrum in this case and three or intentional 2d that corresponds to band 3 and the near-infrared which corresponds to vent 8 and generally if you are familiar with the normalized difference vegetation index it looks very much similar alright the whole concept is similar and then you can classify using this index so if the value of the resulting image is or you'll be resulting pixel it's lower than zero then it generally we can assume it crystals to land and if it's higher than zero we can assume so let's go back to our processing and let's actually perform this calculation so first let's just quickly open the bands that we will be using so it's been three it can open the view by double clicking so we will also see our subset and then feet and I will close the previous view okay so this is our subset and you can see that the reflectance of the water surfaces is very rapidly different independently and so in memory we can see that actually some certain water surfaces are the reflectance is quite high and they appear quite light on the pictures compared to for example forested areas here and invent eight in near-infrared all our water bodies have very very low reflectance and up here almost like so let's calculate the index and you can do this by right-clicking on the probe at number four here and going to mass but mass and let's set a name for the new band that we will be calculating which is and DWI let's unclick the virtual because we want to actually physically safe to spend click Edit expression and we can build our expression using the tools that we have so many three minus 8 divided by the sum of both which is been three plus eight okay so let's click ok we have the green signer there is no errors in our formula so all the bats are comfortable and you can click ok and now we can see that a new Bank has been created here in our band list and we can see that the water bodies now appear very very bright so if I go to my history round which is available I call it color activation tap here we can see that there is majority of the surfaces here are negative values while the water bodies appear as positive values here so we have search by mortal Instagram we are only interested in the two cheeks okay so what we can do next is to create a water mask so how do we do this we can again use the bank mouth operator to create threshold and then binarize our ndwi index into water services send on water surfaces so let's call it n VW i man wat mask and click virtual again go to expression and now let's define the expression so it's going to be a conditional statement which in snap you can find here so if you go to operators then you can see that here you have a preset structure of the conditional safety statement and at science you replace with value so okay so we have to say that our condition will be find all your balance here about that let me maybe write it from the beginning I would just to show you the structure of the kimono so if ndwi balance so here you can see the four before bank corresponds to the index number of the band here is larger than zero then we want to assign one also we want to assign 0 so now this command will classify our image into a binary image where all values that are in original and DWI pan higher than 0 will be assigned 1 and all other values will be assigned 0 ok so here we have our water body mask the first one I created for by a very simple method this method of course is not completely foolproof it has advantages and disadvantages we will have look at them a little bit later when we compare both methods so now what we would like to do is actually exported and in this mask which are his name I can see which we will export into a Geo tape so we can later compare both our masks derived from z2 and 701 and KPIX so let's go here to file export and goj and here we can click on subset and I forgot to do one more thing so now I created new bands in my product and I cannot export them unless I save them and physically so while I chose that known to be your true bands but to be physically safe and I actually have to first save the productivist here so the bands are written into into the product and then I can export them into theater and then we can export so let's go here to exporters buta subset and here we do not actually want to do a spatial subset but you will want to do a balanced upset so I'm not actually interested in exporting any other balance then the mask and so I will select here none and then I will go below and only select the NDVI mask okay and I can remember rename my final product as as to water mask and click export product so this is a binary mask so the expert is very quick and then so I'm done with the processing of the Sentinel to image the mask is exported and we will have a closer look at in a minute and now we will move to processing of census of one data so let's close snap because snap generally has problems with releasing the memory so if I just close the the products here in the memory will not be released and I might run out so I will just I do not want to save any of them intermediate product and I will close now and reopen it again you and there we go now actually let's move to the Sentinel - so I can import my five Sentinel two products from orange again where I have downloaded them I can select all five anymore so now you see we have five products and processing them one by one would be bit time-consuming so and we will use the batch processing tools that are available in snap let's just first open a view of one of them so here we have to view an image it's much larger than our Sentinel one isn't the two times table sorry and any conceded our area of interest is approximately here and you can also see that it's the image appears to be upside down so at the moment djenka are in Raider territory and they are shown on the screen the pixels are ordered as they were acquired so they are not actually projected into any coordinate system so this was acquired during an ascending path so the pixel first pixel in the south was acquired it was this one and it was acquired first therefore it's shown as the pixel I would say in coordinates 1 1 on this image so that's why it appears upside down because here we have the coastal area so let's see how we can do the batch processing so by processing in snap can be done using the batch processing tool here but first what we need to do is design a step by step processing graph so let's do that so Chris let's click on the graph builder and at the moment we only have two operators present here so what we can do now is to ask other operators that we want to other operations that we want to that we wish to perform on all the five products and then save the graph and the process all five products at once okay so for sentinel one gram range detected products such as we have here generally we apply in several steps so first we would like to subset the product to only have the area of interest and then not to process the whole images that would be very very time on to me so to ask this subset of territory right click to the white space and we go to raster geometric and subset and this operator will appear I can just position it grabbing it by my mouse the next step will be to apply an orbit file and that we can do in radar apply orbit file and why do we do this so the orbit state vectors provided in the metadata of the star products are generally not very accurate and can be refined the precise image files which are available days two weeks after regeneration of the product and the orbit file provides accurate satellite position and velocity information improving the analysis that requires satellite information to be as precise as possible so this is an important step that we need to do in most applications bigger then the next test is that we need to perform is thermal noise removal so again radar radiometric thermal noise removal so the thermal noise in sorry mentary is the background energy that is generated by the receiver itself and if excused the radar reflectivity towards higher values and it hampers the precision of the radar reflectivity estimates so the leather one products provide a noise lookup table for each measurement dataset and provide they are provided in the linear power which can be used to remote for you to noise from the product so again we will implement we important temperature and the next steps will be calibration again go to add radar radiometric calibration and the last step that we will perform the metering Corrections so just a few words to the calibration as well I mean the objective of it of celebration is to provide imagery in which the pixel values can be directly related to the radar texture of the scene and they unclear edits our imagery it's sufficient for qualitative use but calibrated site images are essential if we want to put the data into qualitative use so in this case actually the values of our pixels matter to us because we will be doing a train holding on them so we do need to calibrate our data go forward let the next step is to refraction last is the Train fraction and again you can go to add radar geometric turning correction turn correction so now we have the line of our steps we need to connect them so to do this you can either go to the right of each operator and drag it to whichever operator it's the next in line or you can right-click and select connect graph which will automatically connect them not in logical line but not if you have one operator here and one here then connect automatic connection to will not know how to connect so it has if it's in a line like this we can connect them automatically and here you see that you have tabs collects pointing to each of these traders and we can set the parameters there however if we are using this in a batch processing mode we do not actually want to change any parameter here because then what happens is that usually this graph and doesn't work in the batch processing mode so the only thing that we need to do now is to save the draft so we click click Save here and say yes my graphic for example click save into our s1 processing folder close the whole window and then you can go to tools touch processing and the batch processing window first asks us what are our input and output parameters so we can use this tool here to load all the currently open products in the product Explorer so let's click here let's refresh so we can get all the information from the acquisition the type of the eight of the orbit and so on and so on and here we can also see where is the output directory so where our processed products will be saved and we can also select if to keep the source product name so well in our case we can keep it and that means that our input / output track will have the same name as our input product so if you're saving them in the same folder this cannot be done because then your infant products will be rewritten or overwritten but in our case we will be saving them to the s1 processing folder and our original products are in their original folder so we don't need to worry about that then we click on load graph here and we go to the s-1 persistent folder and we're back where I've saved the graph and you can click on my graph XML and all the tools that are all the operators that I have selected in my step-by-step process are now loaded here so now we can go and actually set the parameters so first step is to set the subset so in this case we will only use we can also here we can do in one and and subset and the spatial subset so here we will do penance upset also because we will only be using the H polarization as it offers a better the probability for water surfaces so let's just select you can select two bands by holding shift we can select these two bands and those will be the only ones going further to the other processes then here we want to actually use geographic coordinates as I said we have multiple products so it's safer to use geographic coordinates in this case and these have to be given as a polygon in the well-known text format and which I will open yes so I have a little file here called subset in hope and and the polygon that I have created to it corresponds to the exact outline of a sentence to its obsessively used and I can pass it in here and I can click update see that now is a yellow point appear here and we can zoom in and we can see the orange will extend of the five products that we have as well as the subset size grape so our subset is now defined and we can go to apply orbit file here we do not need to change any details we can just leave the default values their own their own ice removal as well on the beach so eh polarization and we make sure that here for removing thermal noise is checked instead of I reintroduced hermanos let me go to conversion and again we can leave all the default values we won't have that we wanted to have an output Sigma naught X values and then it last that is Tyrion correction so interim correction use an elevation model to correct from Sergio metric to to project it to assign coordinate system and project and correct for effects such and such as SAR over lands one so here we need to select a map projection we will select here the same protection as it's used for the central to tile which is a UTM zone we can do this by going here in the custom CRS and scrolling down and it's not have this nice feature that it allows you to for the zone to be automatically selected so we select this UTM /ps 84 automatic which automatically checks the position of your image and selects which song it should be assigned to so its own 34 and that's all we need to change here we are using the SRT m3 second DM to correct our data and now we can select running just one thing we need to check here if we are saving this to the appropriate folder and we can click you can click run again this will take some minutes approximately 10 minutes to process these five files is they're quite large so we will not do this now I will load instead of files that I have pre processed before keep so I will not click random when I purchase click close and hold my mouse from the okay so here and all the files that I processed our sake in in a snap magnet format which is called snap Emma and they contain a folder and then file which is sort of a header file for it product so I only want to select the header files from the processed files so I'll do this and then I'll just click here and I can actually close the original file while today processed there we go okay so let's now see so now my products are in containing one single ban which is Sigma not underscore pH and if I open them in view okay so what we can see here is that it's not quite oriented the same way or it can seem that it's not oriented the same way as my sentiment to image with versa which was a perfect rectangle this is due to the fact that I have used the same subset and they have applied the subset before doing the terrain correction with projection of the edges so that all the full extent of the subset that I applied is actually here so it would be like this but I have also some other areas that have been added to make it square at the point where I was doing this upset okay so first what we need to do is now why do we actually process five images I don't think I have mentioned these people but Sentinel one or SAR imagery in general contains a lot of stucco so if we zoom in you can see that surface is sort of salt and pepper while you can see very well the lakes still the details are a little bit blurred by this effect so what we want to do in this case as we do not want to actually apply any speaker filtering because it will cause a degradation of the resolution we want to create a mean out of these two summer month acquisitions which will give us much better detail for our for our trash holding so first thing we need to do is actually to add hollow whisper products into a single registered stack so to do this we have to go to radar current iteration and just go just click registration and this is the greatest racing processor it looks a lot like the batch processing tool that we've seen before you can again click to import all products refresh and here I would advise to check the order of your products here and we order them so they correspond to the timeline so I need to reorder them just a little bit fortunately always have to click on the product again when you want to reorder so now now it should make sense from the lemons of July to 28th of August and let's go to create stack here you can choose this button which helps you to find the optimal master for the choristers stack so it's quick here so you see that the master has changed to the 8th of August and here we can choose nearest neighbor as a result prototype cross current correlation year we will not change any parameters while he will also change any parameters and variety where we can select where to save item on our final product and the name and the format again this is actually very highly personal so again I'm not doing because it will take some minutes so I have it prepared yet again and I go close to DLL now and open a pre processed product here and that would be here and let's open it you can see now in the pants I don't have just a single banner but I have actually five different bands named based on the date when they were acquired so let's just have a quick look on something interesting I can right click and I can open RGB now because they're for a district to the same grid and so let's choose RGB so the red band will be July the green band will be for August and the blue ballons 28 okay so this is very interesting image because you can actually see it doesn't tell us much related to the water bodies but it does tell us which surfaces have changed between the acquisitions that we chose to chose as a ballast of our composite facts of the RGB most of these changes will actually correspond to an agricultural changes such as harvesting of a field or so on these sort of grayish salt-and-pepper areas mostly correspond to forests as there is no really that much change during this summer summer period and as you can see the legs remain with very low back scatter values in general so this is just our interest it doesn't really add anything to our water mass creation not just out of interest what you can do with these gorgeous stacks can detect changes between different types of acquisitions okay so we have this so the next step that we would do would be to approach this project stack so let's go to courage istration stack tools and stack a bridge and this tool allows us to create one single bond which is going to be the mean average of all the pixels so let's and it will create a new band in our couragous to the stack I'm sorry so it will create a new product it took 30 seconds and this new product will only contain one single balanced with the averaged values and so now what I would like to do actually in order to my for the masks to be comparable I would like to subset to the same any subset to the same extent as my sensible to the minute but the first option we can understand is something that we have done to minimize the size of the data in order to have faster processing but first let's have a look and if I open use of online some of my fans here and I also have the Sigma average open here I can go to windows and tile easily then I can move my view and I can zoom in maybe and you can see that the echo filter is much reduced in the average product or average band so this is the average product these are the single acquisition bands and you can see that the coastline for example is much better defined at the forest area you get much more detail in the or recognizable detail in the product when you compute the everything now we have this product I will close all of these again and we can actually use this finally the stack to apply some sort of fresh holding generally we would do the same as we have done with sentiment to so apply sort of traditionally conditional statement as we have applied to sentiment to and ewi and so how do we actually find out what is a good threshold here because here is the lowest around really see much and treasure holding is best applied to a bimodal histogram it has two weeks and we can select a good separation point between those two these two peaks so the big enhance is view and histogram we can use you can convert our band from linear to decibel or to a logarithmic scale which will give us much better superb ility so let's click it yes create a new band let's open this new band you can see also visually it's the pocket ponies are much much more enhanced in our view and you can also see that our histogram sort of now has seems to have two peaks now when converted to the logarithmic scale so we have this little bit which corresponds to what the body is and then you have this higher peak which corresponds to the land areas so now it should be quite better to select the separation point so at the moment we have some here in the gray it's approximately 24 25 minus 25 when I put like this so this could be a good place to to select our threshold we could also do and we could also do is draw polygons using this tool and then use them to calculate statistics however we're running out of time a bit so I will not go through that however it is described in liddie a step by step guide and then it can give you further insight on how to select your threshold okay so let's apply this threshold and let's go to bad-mouth together and let's say that's one mask you select virtual and an expression and that said conditional statement and now we have to be careful to use the logarithmic scale so Sigma not underscore pH underscore DB is smaller smaller than minus 25 zero then assign one I'll assign 0 thankful so this is very useful indicator here because if you have an error in the expression economic warns you so you can check before you apply it so now we seem to have no more errors and let's click OK here with everything you have everything set ok so now we can see our water mass pay you can see that these parts have also been classified as as water but that's not actually of interest to us because we will now do a further subset to be to have exactly the same subset that we had for sentinel - ok and we will do this by here then go into a roster subset again and we will use to your coordinates and i have them defined here that correspond to the same subset essential to okay there we go you can now see this is our final subset and you click OK this window and we have a new time here and I can just quickly open it our s1 mask so now we can see we got rid of all this all these science and we only have our cloud mask and we can have a bit better look at it so it doesn't seem as nice and smooth that's the one from Sentinel - this is due to the fact that as I said something the one has a bit worse resolution so the resolution is 5 by 20 meters although there were some ones - the same resolutions of 10 meters but the original resolution of the product is 5 by 20 so it gives us a little bit less detail but it picks up very nicely also the small water bodies and you will have a look at how they compare so what do we need you to know we need to save product and we need to export demand so I'm already exploited the men as geotube so we can go directly to QGIS but again you would do it by selecting the product going to file export geotube set your subset here and export let's minimize this window and let's go to QGIS there we go so to visualize the differences in QGIS let's import both of our products then we have exported in QGIS or earth or in the object so we have s koala minutes to close first let's take this s1 pain and now let's just very quickly set the parameters so we have the best that's you so let's go to properties select at the moment it's in group and gray so let's select a single panel saw the episode of color and let's click here on add about you let's change this value to one change the color to blue it is very quick because we are quite late so I'm just going to very quickly show you how to visualize this let me go to transparency we can set it to 50% 48 50 and we will set 0 to no data value so this is important in order to get rid of all of these black friendlies here black no delete them around and only keep the values that correspond to water pixels so okay okay and now I'll let's do the same folder for dress - bottle mask top of cheese let's go to single pipe so the color sets the red transparency we will leave full and serum no data okay so now because we set the s1 water mask to be transparent we will have our our combined area so when mask is detected or when our processing has detected water body in s1 and s2 will appears purple and for s1 only it will appear as blue and for us - it will appear as red so and last tell what we can do is go to open layers plugin and open some sort of background laughs such as the google satellite okay so you can also drag it below so we have both the five layers below and now we can see actually the differences so we can see that there's some water bodies that have been detected in as one but has not have not been picked up by s2 or for example [Music] there is coastal areas around the water bodies that have generally because the s2 has generally better resolution the there's less mix pixels or not less but we pick up more of the water area then with sensible one so we sent one generally what can happen is that we get less less area covered by the water body than with s2 and okay so this is just the last view let's go now to to compare a little bit in the presentation so I have some I have a slide that just Highline advantages and disadvantages of each of the processors you can see them when you have a closer look at the details here so let's go to the presentation so yeah so basically in most times it would be better option to select an optical data to do your to your do your detection or water body detection however for example when you take and when you compare the methods that we have been processing now someone and DWI and saw recently to an DW is threshold method and Sentinel sorry too generally better resolution it is better in detection of water bodies or in coastal areas however it affected by clouds it can only acquire data in daytime and many times as we have seen on our on our comparison it can happen that the reflection of the lake bottom are the water body bottom actually causes mix cut miss classification also and the mvw by is known for having for a suffering or be affected by a buildup noise so some buildup features have the same difference or ratio as water bodies although if you only look at them in one single band they will appear very bright so not dark but very bright however the ratio of the three and been eight for these built-up areas or tarmac and so on will be similar to water body so this can cause problems this can be mitigated for example by using the shortwave infrared in the near infrared or it can be mitigated by mitigated by introducing additional thresholds to your section so for example using any WI class a one single band single band threshold as threshold in the engineer infrared and so on sentinel-2a sensor Sentinel one in the other hand and the treasure holding method we can use in all day or day and night acquisitions so you can only have two acquisitions per day and we can also use it in any weather and however the legibility is slightly lower it is affected more by the mixed pixels along the coast it is it can also cost increase classification when there is high winds so if you have waves that are pearl and to the movement of the of the platform sorry platform and they can cause higher back scatter and cosmic ossification and also it has vegetation that is submerged in the water it can cause very bright returns so overall double bombs and it causes very great return and again it can cost with us miss classification so you can see that both methods have advantages and disadvantages however also these methods that we have shown here are very basic methods they are not there's many many more much more advanced and sophisticated methods but for many applications is implemented spice however we need to be aware of the advantages and disadvantages of each of those both both windows can also be used for for example flat mapping and so on and so on okay so this is it for any exercise I hope you enjoy it I'm sorry for some technical problems that we had so thank you very much for attending the webinar remember the the code is h i hydr zero one you can use this to request the virtual machine with pre-loaded data and does the by step guide for this for this webinar so thank you very much for attending I hope you enjoyed this webinar I'm sorry for the technical problems and see you hopefully in the next month webinar the topic will be announced later this this month thank you very much and bye my

Channel: RUS Copernicus Training

Views: 7,114

Keywords:

Id: UJcS24thVvc

Length: 79min 17sec (4757 seconds)

Published: Thu Apr 05 2018