 Welcome to remote sensing and GIS for rural development NPTEL course. This is week 5 lecture 3. In this week, we have been looking at NPTEL courses which focus on using raster data for rural development. The definition of rural development has been given earlier, however in this case we will be looking at just the data types. Week 4 we have looked at vector and week 5 we are looking at raster. We have defined the raster in the previous lectures in week 5 and showcase some examples of where this data can come handy. One important thing in week 4 and week 5 is we need to understand the data before we use it. So, in that note while we discuss the raster data, let us discuss what issues can come up in remote sensing data plus vector data for GIS while using it for rural development. Initially vector data we will discuss. In vector data as we mentioned it can either be a point or polyline or line or a polygon. So mostly these are observation data collected from ground or some data entry on tables and then converted to geospatial data. So as the name sounds observation data and user enter data etc for example sensors is done through surveys and then converted to GIS. So what happens here is there is lot of data collection issues. I will just label some of them there are detailed issues discussed in your classes which are available in NPTEL I have given you the links you could definitely look into what are the other classes that can happen. So most importantly we will have data collection issues as bias errors. Bias is what the data collector or the data itself will have an affinity or a bias towards one side. So let us say I am collecting the incidence of COVID if too much COVID cases are reported let us say in a tourist country like Thailand or Singapore then people will be very less coming in right tourism industry will collapse. So there might be some errors while data collection happens and these we call it as bias error. So the error is while entering the data sometimes not the whole picture is captured only some part of the data is collected and that is bias errors. There is also representation errors wherein the data is collected non-representated locations. We will go through this when we do a case study example but for now understand that both the data types will have errors we will focus on the raster data with some examples. But it is also necessary to understand the vector data issues. As I said there is representation errors for example this is your land and it has five parts right. If you correct in the center maybe it is representative of all the five parts five lands five channels whatever it is. But if you correct only here the data how is this data representative of the whole hand. So this is called representative errors or the data is non-representative. In a direct rural development problem I can tell that if a district is consuming lot of groundwater you should be measuring the groundwater in a farm well which is being used. You cannot take water from a location which is not being used let us say school. So that the whole village has a lot of farmland and then there is a school. If you go to the school and take the groundwater level it would be better compared to the farm because in the farm they extract the water right. So these errors are called representative errors. Then date and time lot of errors do not have a date and time or they are outdated. So these are issues that should be properly managed before converting to vectors. You find these errors a lot there is multiple literature on data errors for each aspect I am speaking about rainfall, groundwater, surface water, agricultural pesticide availability and then data on storage structures etc. The final part is instrumentation errors. The instrument that you hold to measure all so has errors. Let us say you are calculating the crop yield in a field. How do you calculate crop yield? You harvest, you put it in a sack and then you weigh the sack 1 ton per hectare, 2 tons per hectare etc. So now if your balance the weighing machine itself has errors then it is wrong correct. And this lot of people have experienced personally. In a gym you will see a different weight, at home you will see a different weight on the railway station where you can put a coin and I used to do that in Chennai. You have a coin machine you put and you get a weight. So all these are different weights. How is that possible? If the weight is more or less equal, for example final grams change, that means okay maybe you drank too much water, you had lunch or big lunch biryani or something and so you can compensate for the excess. But if the difference is like 4-5 kilos from morning to afternoon or within 2 days that cannot be explained. So how is that explained? By deeming them as instrumentation errors. That is why you see very very accurate weighing scales in a jewelry shop and in hospitals where for the babies and all they have a very sophisticated weighing scale. All these data are taken and converted to vector, these are observation data. So we will go through again on this slide, when we discuss the case studies for vectors, I will pinpoint how the errors can creep in and it is our duty to represent correct data while doing neural development work. Now we come to raster. Raster is very very complicated in terms of errors, however some of these errors are taken care of. Let us say for example bias errors. This error is not there because it is unbiased data. For example, the satellite can take images, it need not take a different image for China and India, it is the same image, whereas when you ask people to collect data, let us say you want to study the Ganges River which gives water to almost 1 billion of the population across countries, Tibet, China, Nepal and India and also Bangladesh. What happens here is you will have to look at it in a different angle because you cannot collect data in India and apply it to China. China has to give the data for the Ganges water how it is moving on Nepal. So this is where bias comes, they will give different data quality and based on their benefits etc. Whereas if you take a satellite image, it is the same. So now you could see how some errors are removed just by using satellite data. So there is no bias errors, there is no data time issues because while the image is captured automatically the data has a time and date stamp. You will see this when we download data because every data that we download at the end of the name will have the data name, the version of the data and the date and time of that image which has been populated. Instrumentation errors could be there, for example a satellite is launched in 2022, it starts to collect data 2023 and there is a lifetime, maybe 5 years, 10 years. Then after that that satellite is let go into space, it is not correcting data anymore. These are instrumentation errors but you have time whereas if you go to, for example as I said a hospital to collect hospital record data for children, height, weight etc. There is no replacement of the balance, it is just there, only if it breaks totally every place. There is no every 5 years you throw it and then bring it when you want it. Whereas this one, satellites have a lifetime but still there could be some errors. These are mechanical errors, we call drift errors. Any moving part will have some drift after sometime. It has to be calibrated again and if you do not calibrate the data you will get very, very varying information which is not correct. So by using raster data already a lot of issues have been taken care of. Now we will look at raster data issues, other issues which are very important and we should know. One is quality of the image, aerial photographs, satellite image etc. Seasonal considerations have to be given. So if you are using drones and unmanned aerial vehicles or small planes to take data, you have to make sure that you take it in a correct time and season. For example you have here one and two, one is a leaf off season and a leaf on season. You can clearly see that the leaf off season has issues in terms of the elevations, the elevations can change rapidly. So these elevation changes can be addressed if we take correct seasonal measurements. These are two same location, same area but time is different. In one you could see lot of leaves and then in number the second one you do not see leaves because it has fallen down. So you will have to make sure the date and time you select is correct. These are the reflection issues. What you could see here is a farm pond which I took from the plane. You could see that anyone who is going on a plane has put a window seat and you can just take an image from your phone. It is as good as an aerial image. You are not going to use it for data analysis because commercial plane taking from a window is different and data plane that goes exactly and takes perpendicularly is different. So I am just showing this for quality perspective. You could see that the first one there is some fog, some on the screen. On the screen there is some fog or condensation happening and that can affect the image. As much as you clean it, you still will get the fog. So those who take a flight can see that in the window pane you keep on cleaning but still the fog will come up. So it is like a car out. So you can see in the car like you always have to keep the ventilation heating on. So these errors are bad. The data is not usable because for example you want to see what is in the center part and the center part is totally fogged. So this data is unusable for that region. There are filters that can be used in remote sensing data. However, these are mathematical filters. These are not exact filters because it will just remove. But what is underlying it we do not know. The second one here is reflection from the body itself. So here most of you would not guess it in the first sense. It is a farm pond which is just full of water. And this water part of it is reflecting because there is a wind blowing and the wind creates ripples. So if you go to a beach on a beach you could see that when the wind comes the front part of the water is white in color. But the water is blue or white in the center. So if you take water it is blue in the center and then when it comes and then forms a ripples it forms up and creates a white color for visually I would say. So visually these errors can happen. However you should be careful by just looking at how do you differentiate it. We can differentiate by saying there is a boundary. So this boundary is constant. Can you see? And so this is what you need for a water body. You are mapping the water body. Government of India's very important missions of Dadi Shakti and mapping water bodies will use these kind of techniques where they say okay there is a boundary and then there is a water body. For some part of it's water this should also be water. It's not concrete because it is not built you can take another image. The best is to take another image. Then we have other types of errors which are the top is reflection error. Again this is pure reflection. The previous one was there was a wind and then ripples happened the reflection was white. Here you could see it looks like a glass pane glass window but it's not. It is actually watered blocks. So you see the canal area and then water is being applied to the fields and only the fields which are full of water is reflecting sunlight. So the sun is on the top and so it is reflecting the light and it looks blurry. So here if you want to take this image you could have taken early in the morning or late in the afternoon when the sun is not perpendicularly up and shining on the land. And there are filters for it again as I said filters help only for a particular wavelength and it becomes expensive. But this image is still very valuable because you could see like what percentage of land is water. So in rural development one important aspect is availability of water for all. So you could see that these lands have water whereas this land these lands do not have water or maybe the crop they are using is different so they don't have to water it for this particular period. But what the sharing is important and you could see that along the channel the plots along the borders get water whereas the ones away from the canal area does not get water. So this is one of the benefits of raster images, satellite images. However as I said there are issues with these kind of reflections. This is a drone image and if the drone is tilting and the angle is there then you can see shadows and elongated image. So this is taken at an angle but most importantly in the first image the sun was on the top, the second image the sun was on the side. So you see an image with reflection, I'm sorry, with shadows. So this shadow part, reflection is this one, this one is shadow. So shadow part what you could see is the trees are here but the shadows are elongated. You could see many, many of these trees have elongated shadows and that is because of the angle of the sun and the time of the image. So these can actually be reduced by using correct time images and most importantly flying the drones and UAVs in a particular angle, not a time. So make sure it is stable and horizontally flying perpendicularly to the land. If it is like this or like this then issues will happen in the image. And this is purely because you cannot control the wind. See the plane is moving as per a particular wind speed but suddenly the wind speed can come up and once the wind speed comes up, it will tilt. The plane tilts or it goes like this and this can actually affect the camera power but nowadays there are a lot of sensors to say that okay you don't have the motion but it is our important aspect to make sure that the data has all these errors removed. Then we have contrast and brightness issues. The first one you have a lot of contrast between, it's a black and white image. This is just a photograph that you can use but the same principle applies to drones and UAVs also for captured investor data. You could see that the contrast issue is there wherein you have black and white and the white overtaking the black. So higher contrast ratios are given. So there's some corrections you can retrieve the information. The bottom one has a playing field. It also has a contrast issue and a brightness issue on the borders. You can see that these parts, the image is not as good whereas in the central region it is good. So it is very important to make sure that you have taken an image, with good contrast and brightness. This is the condensation part where you see foggy and misty conditions in one image whereas the other image is clear. Both are the same location, same college, my university. But what has happened is in one it is clear whereas the other it is foggy. And with the foggy you cannot take exact locations and measurements of the objects. So planning all these is very important for your future aspects of data collection. Make sure that you have these data checked. Don't see that, why is this color like this? And if it is wrong and you cannot clean it, don't use it. So with this I will get into the raster dose. We have covered the basics, most important part of the raster errors and issues. Most of the raster errors and issues in satellites are taken care of by giving a correction. But it is our duty to read through the manual and apply the correction. They will not apply the correction in some satellites. For example, Grace, Grace has leakage error issues. This is an algorithm based error which has gravity leakages from the borders of land because sea also has gravity. So these can be taken care of by applying an algorithm and a multiplying the raster. However, Grace team will not give you the full corrected version. They will give you the version to be corrected. Then the raw data, you will have to multiply it to get the corrected version. And this is important because some people do not want to use this correction. They want their own correction terms and they can use it. For that the raw data is important and that is given. So multiple satellites are there and every satellite has errors because it is a mechanical machine with some moving parts. And it is an algorithm that gives you the data digitalization. So you can actually get more input from the metadata of the satellite, which we will cover. And these metadata are always attached to the satellite's download page. You will have to download and read them before you use the data. Because a lot of times the data might be sounding good. However, the errors might be too much. Let me show you a small example that if your error bar is 5 meters in a data set, but you are only measuring 3 meters. Your object is only 3 meters. Let's say you have a land and on each side you have 30 meters. But the error is 50 meters. Let's say 50 meters is the error. However, the smallest size you want to measure is 30 meters. So how can you measure this with a scale which has 50 meters error? So this is what you need to apply. The error bar is the difference between the errors. Min and the max is 50 meters. So maybe you're doing this. This is the measurement. It says 20. But how do you write it? 20 plus or minus 50. Is this a good way of using the data? No. Why? Because your measurement is much smaller than the error. So always look at the error before you download the data. If it is not, you cannot use it. Another example I would give is a scale. If a scale you're using to measure this pen. So the scale is 15 centimeters. But I want to measure this nib. This nib, the front part, is only millimeters thick or millimeters height. So can you use a centimeter scale to measure this? You cannot. Can you use a meter scale to measure this? You cannot. You need millimeter graduations in your scale. So that is what you'll be careful in writing and measuring. So you will need a data set that actually caters to your objective and problem. Moving on, what we'll discuss is we will look at the raster tools. Where do you find the raster tools? I'm going to open the QGIS and then show you. And then this is how the layout looks like. But we will show you in a minute. And the next class we will discuss on one or two of these raster tools that you will be using. Not all will be used in the basic scale. And also for rural development, not all are used in priority. Only two or three are very, very important. We will go through them. The most important case is, again, these tools have been extensively explained in a GIS course for a remote sensing course. I'm taking only one week off the entire course to explain these raster tools. It is our duty if you want to have more information, please learn from that in detail. We will be using the learned message to apply to rural development. Rural development is the core. But how do you use it and neglect these errors we will look at? So moving on, we will look into the QGIS interface. Let me share the QGIS part, whereas we have here. And in this QGIS template, once you open it, you open a blank window. And you have this toolbars and panels, et cetera. In the toolbars, you will be clicking the tools where you had vector initially, in which five you'll have the rasters. So each toolbar has its own help session. So these are very important to look at the tools and then read them. Because these tools get updated now and then. For example, let me open one tool. We will go through this tool later. But I'm just going to show you how it looks like in the interface. So you will have the input data here. So once you input the data, which is in the layers, that will come here as raster bands. And then you have, do you want to create on the fly raster instead of writing layer to a disk? What does this mean is it's a temporary storage. All the tools will have a temporary storage where it will get deleted after some time. Because you'll be doing these calculations again and again using the tools. You don't want to keep storing on your folders and then you don't know which one to use. So the best is to create on the fly. And then once you like it, then you can right click on it and save it. We will show you an example of how it's done. And then you have the output layer where you want to store the file location, those kind of things. You can add and delete based on it. Then you have your GeoTIF, what type of raster do you want to store? In the class, I have only said about GeoTIF, JPEG, raster, LRAS image, all those things are raster. But just look at how many different types of rasters are there. This is getting updated now and then. As I said, while the supplies are getting more powerful, sometimes they prefer a particular type of raster extension. And that they choose based on their needs. There is a NASA planetary data system, USGS data system, et cetera. The most common ones are GeoTIF, raster, HDF4D data set, NASA planetary data set, and it does image, et cetera, et cetera. So you can pick and choose which output format you want. GeoTIF is the default. Sometimes if you don't know, just keep it to the default. It is always easy to interchange. But if you keep it in the default, it is good for use. And then if you want to use an extent for the layer, x, y, z, the boundaries, you can use it. But again, default is fine. Resolution, what type of resolutions you want to get. And the coordinate system. The coordinate system is given here in this map. If you want to change it to a different coordinate system, you can apply that on here. See, this coordinate system is different from this coordinate system. And you want to add the result to the project, which is this your project, or you want it differently. Some toolboxes will have operators like this and an expression window, wherein you type an expression. It is something like programming. You write a code. But understand here that you will not be writing the full code and running, because that's why you have the GeoI, the graphical user interface. You will only be using code syntax. And all tools, as I said, have a help window. You could click the help window like this. I'm going to click. It will open a window page, which will open right now. And that is being stored. Whatever website you're using, it will open up. So let me share that window, which has just opened. And here you go. So you have the raster calculator. What did I switch there? I went to raster calculator, clicked help. And this manual in QGIS opens up. So this warrants that we need to look into this. But you can see here that the raster calculator is being explained, the figure and how it is. So as I said, two data have been put in. The output layer, they wanted to be right on a disk. So they have clicked output folder and then put it in. GeoDiff is being used. The layer extent has been used. Columns, resolutions, Alaska example has been taken. And more importantly, operations. So if you look at these operations, what is the operation saying that land cover one has been taken and only the ones lesser than 50 is multiplied by one plus land cover one greater than equal to 50 is multiplied by two. So those cells which have land cover less than 50 are multiplied by one. And land cover above 50, above and equal to 50 are multiplied by two. So now you have an updated or upscaled raster. And that raster is stored here. The names is hidden, but you can store it in a different name. And then it writes it to the layer. So all these are there. These are examples of the expression. We will go through this in there. But more importantly, this is how you learn by yourself by using the help window. So the objective of this exercise is to show that each tool has a help window. And you'll be using the help window to maneuver between your work. Because that is very important to have in terms of using a tool. So with this, I would like to conclude today's session. And we have looked into multiple formats of data, raster and vector. And in particular, we have looked into a toolbox and where to find help to read and understand the tool. Why this is important is for you to learn is the tools get updated frequently. And you should not be left back because you didn't know how to use the help box. With this, I conclude today's session. I will see you in the next session. Thank you.