 Hello everyone, welcome back to the next lecture in the course. Today, we are going to start with the last part of the course that is we are going to see few applications of remote sensing. So, actually applications of remote sensing is that itself is like extremely wide to discuss because whenever we want to apply remote sensing techniques to some applications we need to have some sort of domain knowledge in that particular field so that we can effectively use remote sensing data sets. So, dealing with applications normally will involve discussing the domain knowledge plus how remote sensing will help. So, we have to discuss both the things which is like a huge task actually because there are plenty of different fields in which remote sensing is applied and it will be almost impossible to cover all applications and with the essential background knowledge required for us to apply the remote sensing data. So, what we will see in this course is just like few basic applications in different fields like we will not go much deeper into it. This will give you all a good flavor of different ways in which remote sensing can be applied. We will broadly touch different different data sets optical, active micro view, passive micro view and so on with which you will be able to learn further or there will be like lot of references I will be giving along with this particular lecture itself like within the if you note within the video itself there will be like lot of references from which the materials are taken or the references which will provide you like a good knowledge about the subject those things will be there. So, those who are interested can look at them and gain more knowledge about that particular field. So, today in the first applications part we are going to start with land use, land cover monitoring and change reduction. Actually land use, land cover falls under category of analysis remote sensing image analysis like after we acquire remote sensing data we will do certain analysis over it or certain operations over it which will help us to categorize land use, land cover and also to understand or identify changes in the land ok. In this lecture we are not going to discuss those techniques because that itself again will be very elaborate but I will just briefly tell you what land use land cover is, what all the basic base in which we can get it from remote sensing data sets and some fields or some applications in which these land use land cover data sets are directly applied even without having much domain knowledge we will be able to appreciate the way such examples and we are going to discuss in the lecture. First of all what is land use and land cover, land cover indicates the physical land type that is present on the surface say vegetation, built up area, water body, these are all physical features that are present on the land. So, identifying them using remote sensing data sets we call it as land cover mapping whereas land use means how we humans are putting that particular land puzzle to use. Say vegetation is a broad class, within that vegetation if someone is using that particular puzzle of land for cropland then that classifies as land use ok there is a piece of land containing vegetation as land cover but humans are using it as cropland. Again a built up area, so built up area is just buildings that is all but for what use that built up area is being put whether it is for residential purposes, for commercial purposes, industrial needs, plenty of different use we can put, so that is land use. So, mapping and monitoring these two like land use and land cover are essentially carried out using remote sensing data sets. In olden days people used to do ground based survey to collect such information but after the advent of remote sensing actually this was one of the earliest applications of remote sensing identifying like land use land cover what people called us like classification ok. So, normally land use and land cover will go together we would not be exactly splitting them like strictly speaking ok this is a land cover map, this is a land use map normally we would not do that both there will be like a slight mix whenever we identify something as vegetation we will always identify whether it is cropland or forest or plantation, shopland and so on. Similarly, whenever we identify a built up we will identify whether it is urban, rural, residential, commercial and so on to some extent. So, both of them go together normally and for various purposes we will be in need of such information. So, what is given there in this slide is an example for a global land cover map obtained from modus sensor ok. So, this is produced yearly. So, here you can see there are like 17 classes together say blue color parcels or water and there are like different classes of forest, grasslands, wetlands, croplands and so on. So, you can see here it is for if you talk about India there is forest near the western guards majority portion is kind of like cropland according to this land cover map and so on. So, this is just one example for land cover mapping and in this lecture we will see some users of these sort of maps. So, why LULC information is necessary? So, land use land cover data is needed both for scientific as well as administrative purposes. Say for example, someone needs to understand how the hydrological nature of region has changed because of land cover change. Say for example, there may be a kind of like area people might have built up like huge residential apartments, they might have changed the characteristic of land cover or some forest have made might have been cleared to introduce croplands all these things might have happened. When such thing happens normally there will be some change in other processes like hydrological processes, ecological processes and so on. If we want to understand them what is the effect of change in the water cycle in the region due to change in land use land cover change for that purpose we need. Okay, this was like the older map and this is the new map when the land cover changed like this how what was the change in hydrology correspondingly. So, these are like scientific investigations also for administrative purposes say someone wants to develop like a new airport let us say like our country is growing really fast and many airports are coming up and actually many different civil engineering projects are being taken up. When those things happen when you want to do like a large scale infrastructure development we should take a look at LULC map identify okay this is like a land that is empty it is not being used for anything or whether there is any other plan for this particular parcel of land. So, it is like a complex process people will not just quickly go okay this is the land I am going to start constituting an airport it is not going to happen like that. So, for identifying where to develop and if we develop something whether it will be useful to public say someone wants to develop like a new airport for Mumbai city it is already being taken let us say. So, the airport should be not very far from the city it should not be like 2 to 3 hours drive from the city it should be close to the extent possible. So, there are like plenty of different planning has to go through when such large scale infrastructure projects comes up. So, for such things you need or some if a government wants to develop like a satellite township definitely land use land cover map will be need up. So, for all these purposes LULC map is needed in addition to this due to normal human activity LULC pattern changes over time in response to our own needs and demands. So, when this happens it will have its own impact in the natural cycles like ecological processes as I told like water cycle even like climate and so on. So, understanding the effect of this change also is really necessary for sustainable development. We cannot like go on doing something which is causing lot of ecological damage. So, when this study happens the land use changed from the last 20 years like this because of this such damage has occurred to the environment. So, can we further proceed? So, such kind of decisions it is both like scientific as well as administrative. So, not only LULC map at this instant, but also the change that occurred in the last say 2 decades or 3 decades is of vital need for both scientific as well as administrative purposes. What has changed reduction? We just gone through about like LULC. So, what exactly changed reduction is very simple identifying what has changed between 2 time instances or different time instances say in the last 30 years what has happened to the city. So, we can do kind of like a multi temporal analysis say every year what has happened or every 5 years what has happened. Identifying such things is known as change reduction simply put. So, as I told earlier the changes that are happening in the land use land cover is again going to impact our surroundings, our environment and whatever natural things that we are dependent upon. So, this is also really important and say like what like some of the clues that we may get about like change reduction. Say when there is kind of like a large scale like urban expansion we may see like whenever you are seeing from the top especially from satellites or airborne you will be seeing like a sudden large new kind of patch that is growing up or when a forest is cleared for like agricultural use that will appear differently when you look from satellite. So, all these things are kind of like clues for identifying ok some changes happened. So, even with like kind of visually looking satellite images we will be able to identify such changes. How remote sensing is useful in getting this LULC information or for change reduction purposes? In olden days like in the early days of remote sensing people used to do what is known as visual interpretation like even before the advent of satellite remote sensing aerial photographs were being taken like when we discussed about different platforms I briefly told about like taking photographs from aircrafts kites or PGNs and so on. So, aircraft based photographies were in wide use in earlier days in the technology developed rapidly during the world war times and people realize the civilian applications of such photographs. So, so many developed countries had like a planned program of acquiring aerial photographs. So, people will use to look at those photographs and identify what feature is there. So, that is known as photo interpretation or visual interpretation of aerial photographs. We have to visually look at it and identify ok this particular say this is like a photo normally it will be like a square ok this is like a patch of land with the crops in it. So, this is crop land this is like a township. So, this is built up area maybe some sort of semi urban. So, here there is like a pond which is like a water body. So, identifying this visually urban water. So, whatever is present in the photograph identifying them visually is called visual image interpretation or photo interpretation. Even after the advent of satellite images people were still doing it for quiet sometime. So, normally in satellite images I told you especially in the olden eras we will be getting like each band will consist of like one image whether it is like hard copy printed maps or something ok. So, each band you will get one one data if it is computer based analysis like you will see an image in a big screen and identify it. So, it is still visual image interpretation, but identifying it on a screen that was also done like just displaying the image on the screen whatever screen like TV screen or computer delineating there using a some sort of input device or during the days in which we were receiving paper based satellite images like hard copy satellite images will be delivered in olden days. So, when that happens normally people will order for what is known as like a true color composite or a standard false color composite. So, what is true color and standard false color? True color is say satellite can take data sensors can take data in RGB, NIR bands all these things we know. So, whenever you get like a printed map on like a particular type of like a paper or something the print not a map is a satellite image. When you get like a satellite image printed on a sheet we would not be getting just one band separately. So, normally especially like for a Indian remote sensing satellites they used to give us kind of like different bands combined together say red band, green band, blue band. When you want to display color images in some sort of like display medium either on a paper or on a screen we will mix it in terms of its corresponding colors. Say in display pattern there will be again red, green, blue colors when these three colors mix you get an image color image display that is the nature how normal display systems works. So, when you display a red band image correspondingly in the red color of the display system like a bright red color will appear like bright red in the display like high Dn values in red band will appear bright red color. Low values in low Dn values in red band will appear like very light red in color like that. Similarly, green band from satellite data assigned to green color in display, blue band in satellite assigned to blue color in display. So, such corresponding mapping between the data and your display system we call this as true color composite. But since blue band undergoes lot of attenuation like atmosphere scattering normally blue may not be printed people will replace blue with NAR band okay. So, satellite data it will be red, green NAR in display system NAR will be mapped to red, red will be mapped to green, green will be mapped to blue. So, you change this see this particular pattern of colors. So, this is known as like a false color composite. So, the true colors which we see are not there. So, what is actually appearing red that is not red that is the signals in the NAR band. Similarly, what is appearing as green it is not green but what is due to the signal in red actually that is red color but we are seeing it as green. So, this kind of mix and matching like or displaying some other color for a data in some other band is known as like false color composite. So, whenever like hard copy satellite images were printed and delivered to users normally users can order either like a true color or like a false color composite. Say people will order like a false color composite they will have a look at it okay. There will be what is known as like a light table and there will be a huge tables attached with lights underneath. So, the satellite image will be spread across that particular table then transparent tracing sheets will be placed over the image and they will be like bounded properly and expert image interpreters will manually trace whatever different features are by looking at like the false color composite data and by using what are known as visual interpretation clues I write here. So, visual interpretation clues people will be able to identify what is there to like a very good extent. So, they will manually trace such features and that thing will be considered as like LULC map then it will be transformed to like a proper map. So, that was how LULC map were prepared in the olden days. So, if you want to do like a change deduction analysis okay this is like the LULC map produced using this year this is the LULC map maybe in the next 5 years compare them and what to say overlay them and identify what are the changes. So, such things were done manually using what is known as photo interpretation or image interpretation. So, normally there is like a slight difference between photograph and image if you look at like some of the textbooks they will define as if you use the detector itself as a storage medium you call it as photograph say our film what we used in our olden day camera that is the one which deducts the incoming light at the same time that will be used to store the image also right when you take the film you will develop it. So, that contains the actual information you call that as kind of like photographs in satellite image what you call it as an image it will be stored somewhere like digitally or something and then it will be transformed to the ground and you will take like separate printouts. So, the detector is different system what displays the image is kind completely like different system. So, there are like minor difference especially some textbooks will have this difference and also it is convention to call aerial photographs and satellite images we will not normally call satellite photographs nowadays most of we call them as like images. So, that is like the minor difference. So, this was done in the olden days at present especially in the last say 2 decades or so digital classification techniques have developed. So, digital means rather than using like a hard copy satellite images people moved on to digital images which can be directly displayed on a computer. So, digital classification techniques emerged or developed using by looking at an image we will be able to extract information using computer based analysis using either spectral clues or spatial clues spectral clues means say each particular land cover category may have like one spectral reflectance pattern. If you have like multi band image you can just try to extract this spectral reflectance pattern and maybe match with your ground or reference information such thing can be done or you can again or you can like manually digitize few pixels telling that this is like water body this is like built up say maybe like some 100 pixels from like a big image if we classify this and then tell the computer I have identified you few pixels then you classify the rest of the image. These are digital classification techniques which is widely used now. So, from satellite images by normal either by looking at it visually or by using digital classification techniques it will be possible to get LUL's information and also to do change reduction. So, this is like a schematic of how remote sensing is helpful in getting LUL's information. So, there are like plenty of different satellites different modes of platforms are available. So, we have different landscapes each having its own spectral and spatial characteristics. So, by we will normally acquire the image over the landscape we will have images. So, we will get the image downloaded we will develop what are known as classification schemes or decision rules. So, by combining the image and by combining the spectral and spatial characteristics of different landscapes inputting them as inputting them into classification schemes we will be able to get like a LULC map. So, we will use not only satellite images we can use digital elevation models we can use other existing maps and so on. So, these are called like ancillary data sets apart from satellite images if you use something we will call it as ancillary data sets. So, we get LULC map. So, let us say I have LULC map for say year 2000 and now I have LULC map for year 2020. If I compare them and produce like a change map that is what is depicted here. So, temporal change analysis dynamic change this can be done. Otherwise, so this is called post classification change reduction. Post classification change reduction that is we do LULC mapping identify the land use land cover features there then after producing two maps we will do change reduction and hence this is called post classification change reduction. But this need not be done always like this and some people even do not prefer this rather than this if our goal is to only to do change reduction we can directly do from here. Just by comparing multi temporal images and by applying this classification schemes to those images like without producing this we can use what are known as certain change reduction schemes and we can directly classify this. So, we can either create LULC map then perform change reduction analysis or directly by using multi temporal images and change reduction algorithms we can directly come up with a change map. So, both of them are possible. So, this is like the brief or broad schematic of how LULC information or change reduction information is obtained from remote sensing datasets. So, there are plenty of or variety of algorithms and methods available to do land use land cover classification and the different using different criteria we can classify them. So, we will just broadly see how the different algorithms for LULC mapping can be classified in what all the different ways it can be classified. We would not be going into details of seeing those algorithms but we will see what different classes of algorithms exist for our basic knowledge. So, first thing is whether we are going to give any sort of like inputs to the computer during like what is known as like a training sample to the computer or not based on that we can classify the algorithm. So, whether training samples are used or not that is like the first category of classification that is I told you like we will display an image in the computer and then we will identify few pixels for each class. Say this is water, this is built up area, this is cropland like that we will identify few pixels and then we will ask the computer algorithm to work on the rest of the image to classify it entirely. Say that can be like a thousands of pixels in the image but we will identify one maybe like a few tens of pixels for each class and rest of the pixels will be classified by the algorithm. Such classification procedure are called supervised classification approach that is as humans we are supervising, we are telling this particular pixel this group of pixels is built up. So, similar pixels should be categorized as built up by the algorithm. So, we are training or we are inputting something to the computer. Then on the other hand sometimes we may not be knowing what is there but we may want some preliminary information. So, we may ask the computer to identify spectrally similar features like I told you we will be looking at like the spectral characteristics basically we will spectral or spatial characteristics whatever say this is similar how spectrally similar these features are that can be like say 100 pixels comprising of cropland. So, most likely the spectral signature will be closer to each other they will not be exactly matching but they will be closer to each other. So, we will try to identify such pixels which have similar spectral characteristics or maybe spatial characteristics. So, identifying those pixels the computer will tell these are all the groupings I can make say these are this is one cluster of pixels this is second cluster of pixels it will tell us. So, what we can do is we can have a look at such cluster of pixels and identify and then maybe go to ground and identify okay this is like one cluster the cluster in the ground is actually like a cropland. So, this entire cluster must be cropland this cluster is actually like we can like have a corresponding look in the image as well as the ground information right doing such thing like getting the clusters as output first from the computer then assigning actual clusters to it that procedure is called unsupervised classification because the computer is actually doing the clustering thing as to us it will not classify computer will not tell this is urban or this is cropland it will just tell maybe like the basic algorithms maybe nowadays AI based algorithms with our input it can function many things. But still older classical algorithms will be telling okay these are all the clusters or group of pixels it is up to us again to further define what those group represent on the ground. So, this is called unsupervised classification. Then next major category of class if I erase whether we do one pixel one class or sub pixel classification that is let us say there is like some image is there with 100 meter by 100 meter pixel resolution. But so each pixel corresponds to 100 meter by 100 meter so there can be many different features present within that particular pixel that can be a small point that can be like a building and many so and so on. So, are we telling okay one pixel corresponds to only one class say what is their present majority if you assign that particular thing we call it as like per pixel classifier or sometimes we will be also interested to know what are all different features present within that pixel say okay this pixel contains 50% cropland 20% built up area remaining thing as like say water body if we identify such sub pixel level classification that is called sub pixel classifier there are a few algorithms even to do that. Then like the next major classification will be whether are we using only like the spectral information or spatial information or both. So, spectral information means we will just take an image derive the spectral characteristics of the data and just by looking at the spectral characteristics of data we will identify what is what so those sort of classifiers are spectral classifiers. But with the advent of very high spatial resolution data like less than 5 meter spatial resolution if you have a look at those data only by looking at the spectral reflectance curve we may not be able to get actually what is there present on the land we may also need to have a look at the spatial information present like say when you look at like 50 meter by 50 meter data 50 meter pixel data a building will be like a small part of the 50 meter pixel. But when you have a look at say 2 meter data so a building will be the same building will now be occupying several pixels maybe like 10 pixel by 10 pixel or something. So, the building can be say 20 meter by 20 meter can be right. So, when you have a look at like a medium resolution data a land cover feature will be present entirely present within the pixel but now it is now spread across multiple pixels. So, we have to look at have a look at the spatial patterns. So, such algorithms also develop they are known as contextual classifiers or spatial classifiers. So, whether are we using spatial information or not in our classification using that we can classify our algorithms as spectral classifiers or contextual classifiers or hybrid. It can use both spectral information as well as spatial information. So, these are some examples of different classification procedures that we have these are just like a very brief introduction to the topic. The topic is very wide like image classification pattern identification is very wide but we have just discussed very briefly about the different classifiers. So, with this we end this particular lecture. Thank you very much.