 Hello everyone, welcome to the NPTEL course on Remote Sensing and GIS for Rural Development. This is week 10, lecture 1. The past weeks, we have been looking at various issues in rural development, especially data issues to augment and increase the understanding of the rural problem. And we also looked into certain aspects of new development. For example, increasing acreage under a particular crop and crop yield. We also looked at that most of the data are outdated and or from the British era. These need immediate updation, however, lack of capacity, time and cost limitations are available and therefore, there is less data that has been procured for rural development. On this note, instead of doing the business as usual scenario, we have looked at multiple data sources that can either serve as proxy data, create new data through data mining approaches and or become augmented to the observed data. So in that case, we saw remote sensing data as a very powerful tool that can do all these three things that I mentioned, augment observed data, when data gaps are there, data limitations are there, provide new data as secondary data or proxy data and or can be used for data mining activities. So where you can create new data. On this note, in the last week, week 9, we stressed for the need to understand the land that is available for rural development, not only agriculture, but for example, if you want to start a small scale industry and or a chicken poultry farm aquaculture, you need land and resources. For that, we have used remote sensing tool to select areas for these data issues and for new development activities. So let us do a recap of week 9, what we went through in terms of understanding the data needs and how remote sensing can help. In particularly, we looked at remote sensing and GIS for crops, because rural development is still mostly focused on agriculture, most of the population works on agriculture. The water that has been given to rural regions around 89% is used for agriculture, I would say more because industries is less. So 90 to 93% of the groundwater in rural area is used for agriculture, if not 100. So in week 9, we looked at types of LULC classification, which is basically the land use land cover and we also did some land use land cover change, so that we could look at how the land changes and where we can implement new development activities. Development activities can also include increasing acreage of farms, green farming, agroforestry and subsistence farming. We also did one hands-on tutorial of LULC using data stored in the US GIS website and we were able to quickly do within the 30 minute time frame. However, there were some errors and fine tuning could have been done. As I said, these exercises were initially master's thesis some 10 years ago. Now we can do it in 30 minutes. Thanks to the development of ease to access of data and readily available remote sensing data through multiple platforms. Then we looked at issues in water availability for crop irrigation. We discussed crop irrigation as a particular entity because during the monsoon season, yes, if you have land rainfall comes, you have agriculture. But during the non-monsoon season, only those farmers who have access to water can do irrigation which is application of water for crops. So farmer sometimes have only one crop because when the monsoon comes, they do the farming, growing the crops and then harvesting and when there is no monsoon, they harvest it and then wait for the next monsoon. These are farmers who are limited with economic resources and water resources whereas farmers with access to pumps, wells and some budgets can involve in non-monsoon crops and these are divided as rubby and xyth crops. And in some locations in India, I will show through the case study. You could also see the same land used four times. So one during the monsoon, twice during the rubby and one during the xyth season. Mostly the summer monsoon may take more crops. So we discussed how remote sensing can help in identifying land and water resources for irrigation. Monsoon crops is one, but non-monsoon crops is what we looked at. And also we discussed the fact that irrigation doesn't mean that it's only non-monsoon. Even during the monsoon when the rainfall is less, because always not the same rainfall occurs. When the rainfall becomes less, we can supply excess water through groundwater or canal water. Then we looked at groundwater as a tool for irrigation and short case that India is the highest groundwater extractor in the world with approximately to 45 kW cube water extraction. I say approximately because the data doesn't fully document all the groundwater use and groundwater access. And that is where we've found that central groundwater board data which is collected once every three months may not be sufficient to capture spatially, temporally and deep and shallow aquifer connectivity. So all these three things cannot be done just by using CGW data wherein new data should be augmented and grays data was found. I explained very clearly how the grays data works and showcase some regions in the Ganges and Rajasthan-Haryana Punjab region where groundwater depletion is tremendously high. This is reflected in the central groundwater board report also. However, a longer time series analysis is absent. In the grays website just by a click of button, we were able to get around 20 years of data groundwater depletion data for the Punjab region. I say groundwater but it's actually total water storage. If we assume that the soil moisture is the same, it just cancels out in the anomaly and what you get is net groundwater change. The slope should be the same and you would see that soil moisture comes in. So the sinusoidal curve was there. Then we found out that after groundwater we also need crop type and crop acreage estimates. We discussed who are the key stakeholders that require this data and use it for rural development. And now moving on in week 10, we will see what exactly the remote sensing tools are available for crop statistics, which include crop type identification, crop acreage estimation, etc. And there has been many indicators developed using remote sensing data. These indicators serve as a matrix or a baseline from which assessments are done for crop estimation. So this is very, very important case to understand the remote sensing indicators, which indicators are suitable for Indian locations and look at crop growth, acreage, health, etc. One such indicator is the NDVI indicator, which we will focus a lot in this class. It's very easy to estimate and more importantly, now you have readymade products for NDVI. So when I was in school, I would normally download the data, do the subtractions and estimate it, but now almost three, four platforms have readymade NDVI data, we just have to download it and use it for your reports. We'll showcase what it means in the coming slides. So that is where I said, as I said, some platforms, boven NDVI, we will check and need directions. Where is the new crop type, crop acreage, crop yield estimates going to come. All these we will discuss in the week 10 lectures. So we discussed the methods of area and crop assessment and how in the past they have been using human centric approaches, wherein people would use a tape and a scale to measure a particular area for a particular crop. And then slowly there has been some updates in this method. So even now, people use the physical methods. You would see people going down during the flood season to estimate crop damage in an area, not using remote sensing data, which is quick and approximately correct. And then we have the DGPS method, where we have differential GPS measurements taken around, but the most quickest, less costly, or even free open source, I would say, is the remote sensing based estimates. They're very fast and have high precision and accuracy compared to the other methods. There is no bias error, human error. These are all neglected. Human error can only induce if the person doing the GIS work doesn't do it properly. But nowadays, as I said, there are platforms that can give this data quick and free of cost, so you don't have to rely on errors that may happen. Then the drones, we discussed about some drones, like for example, an average agricultural mapping drone would cost around 8 to 10 lakhs. Not all farmers will have that money. Even if you say a cooperative of farmers can own a drone, let's say a village owns a drone. But who's going to fly it? Who's going to get trained in flying? There's a certificate for pilot certification. And who's going to get data to analyze? So all these are there, and still GIS is needed, and still remote sensing principles are used. So why not use the satellites? One thing I want to stress here is, of all the technologies, satellite technology is growing every single year. So the data which I used when I was in college and school was 90-meter resolution. Now you get it at 30-meter and 10-meter resolutions, free. So you can see how the big jump is. And also you get new hyperspectral and multispectral images, which was not readily available in those days. So I'm trying to educate all of you through this NPTEL lecture to spend a lot of time on processing satellite data, even though drone survey data is also remote sensing data. I'm pushing for satellite data because the satellite data technology is coming very fast, is growing very fast. Indian government has relaxed a lot of rules for using satellite data for research and stuff, which means like giving it for free of cost. So I would recommend you to focus more on satellites. Similar rules and regulations may occur for drones and the technologies and the steps that you do are the same because it's the same GIS remote sensing data that you can see. So moving on, let's look into some examples of the satellite-derived products. Satellite-derived tools, indicators are very, very key. There's a lot of indicators that are used for research, for especially mapping and acreage of crop growth, vegetation, and the health of vegetation. So this NDVI ranks really, really high. A lot of studies would be done if you just Google NDVI in Google Scholar. You'll see like a lot of papers available. And you'll also see a lot of papers available for the Indian subcontinent. So I would like to first, foremost, have this discussion on NDVI and then we will take care of other indicators also. So the normalized difference vegetation index NDVI, it estimates a greenness of earth as viewed from space. So basically if a near-infrared or a green right is shined on a particular plant and a plant is fully grown, very healthy, it has green leaves, then a lot of green is reflected. The other colors are absorbed. Same, the near-infrared is given by plants and it's reflected by plants because when it's healthy, it has more near-infrared. So basically a healthy plant has different reflectance compared to a normal or a non-healthy plant or a plant which is ready to harvest. Let's take this example. So you have two plants on the screen, bring my pointer. So you have two plants. So one is the healthy plant and one is a plant which is about to lose the leaves. It's getting brown and stuff. So let's look at these two colors, near or wavelengths in the spectrum, the near-infrared, which is not visible to human eye. So that wavelength when it comes, almost 50% is observed in the plant and 50% is reflected in a healthy plant, in a non or a not that healthy plant or a plant with brown leaves, a plant which is entering fall season, autumn season, or which is dying, wilting because of no water, no fertilizer. So that actually gets 40% reflected, 60% is observed by the plant. And you could see that that 10% difference is big to tell the difference between a growing plant and a nascent plant. A growing healthy plant will have a really good green cover and the green cover of the leaves will reflect the near-infrared high. In the visible light, almost all of it is observed because in visible you have multiple spectrums, multiple wavelengths, most of that is observed, but the green is reflected. So the 8% is reflected in a healthy vegetation, whereas in the visible, 30% is reflected because it is a mixture of colors. Not only green is there, you can see green, red, brown, orange, yellow, all these colors are there because some leaves will have still some greenery, whereas most of the leaves are turning from green to red, brown, and then slowly falling down. So that will give you 30% of reflectance. Now this difference, this difference in the reflectance will be used as a function for estimating the plant health. So look at this, so 50% is reflected. So this is how the pixel will give you, right? The pixel will give you a value of near-infrared. If 50% is reflected, the 50% is stored in the pixel. So that pixel value is taken here, 0.5 minus 0.08, 0.5 plus 0.08. So the equation is very simple. NDVI is near-infrared minus red and near-infrared plus red. Since this plant has a lot of red color, red color is reflected. Since it has less red color, red color is observed in the visible red. And then you have the near-infrared, which is a particular wavelength, and that is being highly reflected in the healthy plant compared to the non-healthy plant. So this is the equation, near-infrared minus red by infrared, near-infrared plus red. So 0.5 minus 0.08 by 0.5 plus 0.08 is 0.72. It's the same thing, 0.4 minus 0.3 by 0.4 plus 0.3, that is 0.14. So now we have ranges. And this range is given by a lot of researchers based on the NDVI's function. If the NDVI is negative or below zero, any value below zero, it can be barren soil or water because there's no life. If there's no life, there is no infrared, near-infrared. So basically the top will be zero. So this part will be zero, then this negative part is there. So even whatever small part, so let's say minus 2, minus 2 by 0 plus 2, so it is 0.5, 0.05. So all these negative values would indicate that the near-infrared is totally observed and not reflected. So if it is not reflected, the satellite will not see it. So what data we are collecting is the satellite, the sun's energies are coming and it gets reflected back. Once it gets reflected back, the satellite captures it and gives you as an image. Now if 50% is reflected, the other 50% is observed, when will it be negative as per this equation? If NIR is zero, if NIR is zero, the zero percentage of NIR coming, all of it is observed, then it is either barren or water. So water observes all the colors and then it splits because water is colorless, white, etc. Barren soil is a lot of, there is nothing there to grow. So there is no green, there is no infrared reflecting agents. So you won't see much infrared. Very low is there. So these are the classes for the plant growth. So very low, very less plant growth is zero to 0.2. Low growth, moderately low growth is 0.2 to 0.44 to 0.6. High growth and really good cover, green cover is, vegetation cover is 0.8 to 1. So the max is 1 and the minimum would say, people say minus 1. So the studies have actually shown that NDVI ranges from minus 1 to plus 1 because the max you can have red is minus 1, if NIR is zero, 100 percent is affected from red. So you get minus 1. So this range which is being created for NDVI is used widely. So all the data is, they will get the reflectance and then multiple bands are there. They take the NIR band minus it by the red divided by NIR plus red and whatever the fraction comes. The fraction is between this range, each range has a particular value. So from this you can see 0.72 is almost near to moderately high and high. So which means the plant growth is really good and 0.14 is very, very low. So you could see that also the plant growth is not growing well. Let us look at some examples. NDVI has a tool using remote sensing. So this data from USGS shows the recent, recent Landsat, Landsat 8, Landsat 9 is also recent. So this band is 6, 5 and 4. So what you do is, when you collect, this is the difference between the surface reflectance of a normal Landsat and a normalized difference vegetation index, NDVI. So you have taken 6, 5, 4, normally 6 minus 5 by 6 plus 5, the band number and the band 6 is near infrared and 5 would be red or 4 would be red depending on the satellite. So here if you see, if you merge all the colors in one composite, you have this colorful image and wherever green is there, you think, okay, there is good green growth, whereas this pink and brown is there, you think, oh, there is nothing growing. But if you do the NDVI color, which is non-visible to the eye, so 6 is non-visible, the infrared is not, a near infrared is not visible. If you subtract and then do the coloring, then you will see that the range is minus 1 to 1. As indicated, the water bodies are minus 1, okay, so blue bodies are there. You can easily determine if it is a water body or not by zooming into the location. If you zoom into the locations, normally the water bodies have a shape, round shape or kind of a irregular kind of a shape. But you can see that the drainage is there, there are some pathways into the water bodies. You do not see suddenly there is blue as a patch. But here you could see these lands are low, there is no vegetation growing, whereas in the northern side, there is a lot of good vegetation growing in the Sacramento region in the US. So this is a clear difference between a normal image taken by the same satellite, same sensors. If you use a composite image, Landsat 8, all the bands that you see is this. But if you say, no, I do not want all the bands, I am going to take out 6 and 5, suppose 6 is near infrared and 5 is red, I am going to take 6 minus 5 and 6 plus 5. And there you are, you get the NDVI indicated map. And in the indicator map, you can see that clearly a lot of areas are green and there are some nascent growing also happening which is not shown well in this image. So using this, many researchers have studied the NDVI as a tool for remote sensing. So let us look at the temporal profile. Here you have days, the NDVI profile of a particular location 2001 to 2002 for irrigated and non-irrigated areas. So this is the CMS which is the complete season, growing season and then there is a first season, intermediate season and second season. So this includes the irrigated and non-irrigated. So basically there are three seasons and we are going to see for the same location how it differs between a irrigated and a non-irrigated. So if you have a land and it is the NDVI ranges from minus 1 to plus 1, we have just normalized it from 0.1 because there is no negative values, all of the time it is growing something. So let us say that there is no negative values. So what is happening here is in this paper, the Julian Day is taken. So some researchers write it as calendar day or Julian Day. In Julian Day, 97. So Jan 1 is 1 and then you just keep on adding. So Jan 31 is 31, Feb 1 is 32. So if you do that calculation, 97th day in the Julian calendar or Julian Day is 7th of April, 365th day is 31st of December, if it is a normal year. In a leap year, we will have another. So here what is happening is in the irrigated area, you always see a higher NDVI, a slightly higher NDVI, correct? You can see that 0.3 to 0.4 and then it grows, the peak is different, the peak does not happen on the same day. So in the non-irrigated, the peak happens just after the rainfall or day or two after the rainfall, the plant is happy, it turns a lot of green. Because in the irrigated, what happens is there is good supply of water and so what happens is there is a higher peak. Just look at how much difference this is, this is pretty significant. This is significant in terms of a plant being declared good or high. So this is almost high, very high or high growth of the plant and this is normal growth of the plant. This is because higher water is given irrigated. So irrigation actually helps a lot of times to increase the plant growth, need it or not, that is a different story. In this scenario, it is helping and you can see that you can differentiate based on just the NDVI data for a particular location if it is irrigated or non-irrigated. So you can see that 2, 3 seasons are there. In the IS season, which is the intermediate season, nothing has grown much. So almost barren land, almost zero. But in the 2 cycles when the water is applied and 2 monsoons, because some areas have 2 monsoons like Tamil Nadu, etc., they have 2 monsoons, North, East and South West. So what will happen is you will have these 2 monsoons that provide some relief and some water for the crops and that is why you see a bigger peak in here. You have the second peak higher. So the second monsoon is the highest. The first season is not that big. So you can see that 2 of these same location, the NDVI estimate gives you the difference between a healthy plant and a non-healthy plant. And the healthy plant can also be because of irrigation. Here there is the same paper, there is also multiple images done for entire India where A to C is given as irrigated green color, non-irrigated as brown. So they have taken the satellite data and converted it into irrigated versus non-irrigated. The irrigated is green and non-irrigated is just when there is rainfall, it happens and then there is crops. No-crop land is also given. And then in the, from 2000, 2012, 2015. So you could see that in the 3 time periods, 2000, 2012 and 2015, considerable increase in the irrigated area. The barren land, no-crop land and the non-irrigated brown land is almost the same. See the central parts. But there is considerable increase in the green area. That is because of access to ground water, most importantly. Then the D to F, it is percentage, thaluk-based irrigated area, percentage estimated by agrigated to 50 millimeter area based on modest NDVI for 2000, 2012, 2015. So just aggregating the 250 meters, what color is coming? Is it 0 to 100 percent irrigated? So 100 percent irrigated would be almost the NDVI, very, very high values. And you could see that along the ganges, there is high irrigation happening because of canal irrigation and ground water irrigation. So now if you compare this to the data set we already have, which is the CCWB data, you could clearly see that this increase from 2000 to 2015 and this data is also 2015. You could see that there is multiple blocks that are converting into red in the areas where there has been an increase, considerable increase in irrigation, which definitely depicts that it is converting from a non-irrigated area to an irrigated area like these areas for example, now it is turning green, whereas these areas are now turning red in terms of ground water. So ground water has a definite relationship to the irrigated and non-irrigated status and remote sensing is the only tool that could capture at India, pan-India scale the changes in vegetation due to access to irrigation. So now there are different satellite methods, I have already explained the Lonare et al. paper which is authored from my group, Chinasami's group. What you could see is there is different tools, different payloads, sensors and different AI ML techniques that can be used and we came to a conclusion that the Sentinel-2 data with higher resolution was better for the Maharashtra location. Similarly, this study has also done a comparison of irrigated areas based on 250 meters which is modest. But look at modest, which is very, very coarse resolution, Sentinel is 10 meters to 30 meters resolution and then you have North, East, Central and South of India, irrigated area, how it changes. Then you compare that with IMI's International Water Management Institutes, Irrigation Maps. This was the NGO where I also worked for three years when I was in Nepal. And you could see that there is a considerable difference between the models. It's the same year, but same location, East, Central and South. But there is the green color which says irrigated is different between the methods. Then the Landsat ground condition is different. You can see that the Landsat regional views depicted by Landsat ETM data and AWIFS Landsat land use land cover. Each column from left to right is North, East, Central and South region of India. You could see that, yes, if you use just the Landsat ground condition, you could see multiple images of irrigated and non-irrigated areas with water bodies and then the land use land cover based on the AWIFS data showcases that the Karif and Rabi season, the Karif and Rabi are along the South and Central regions when compared to the North and East regions. So basically what I wanted to tell you here is the accuracy of your NDVI will also depend on the satellite, the location you use and the methods. Some people would you have another formula for NDVI, which is called a developed NDVI. We will look at it in the following lectures. So here is the question when all these different colors are given. So the C is not a NDVI, it is just a normal image, green, red and blue, green, near infrared and shortwave infrared. So you can see that all these composite images somewhere give the picture, but it doesn't tell you the difference between irrigated and non-irrigated, whereas the land use land cover can give it based on the crop type. So what this tells us is there is a need for augmenting remote sensing data with observed data and other data to get at a particular understanding, which we will cover in next class, but I will stop here in terms of what is synergized mapping. It is already that we discussed in very brief terms, but I'll be happy to discuss this more in the following lecture where we will look at what are the different tools that can be used to bring data together into remote sensing platforms and then use it for rural development. So this is a trademark name, synergized mapping. It is trademarked to IIT Bombay through Chinasamy's group again, but anyone can use it as just a concept. Why we trademarked this, we wanted to show that this can happen and when you trademark it, then there is the framework is trademarked. So anyone who is using it in their publications can be acknowledged by us. So with this, I will stop today's lecture. I will see you in the lecture 2, week 10. Thank you.