1 00:00:00,980 --> 00:00:01,950 [Amanda] Welcome to week five 2 00:00:01,950 --> 00:00:03,870 of Remote Sensing Foundations. 3 00:00:03,870 --> 00:00:05,160 This week we will learn all about 4 00:00:05,160 --> 00:00:06,640 image series interpretation, 5 00:00:06,640 --> 00:00:07,890 including image collections, 6 00:00:07,890 --> 00:00:10,680 image compositing and change detection. 7 00:00:10,680 --> 00:00:12,010 Upon completion of this module, 8 00:00:12,010 --> 00:00:14,197 I hope you'll be able to define and understand 9 00:00:14,197 --> 00:00:16,280 and calculate spectral indices, 10 00:00:16,280 --> 00:00:20,364 including NDVI, NDWI, and EVI, 11 00:00:20,364 --> 00:00:23,452 use Google Earth Engine to build and apply a mask, 12 00:00:23,452 --> 00:00:27,510 classify an image using both supervised classification 13 00:00:27,510 --> 00:00:31,060 and unsupervised classification and compare the methods. 14 00:00:31,060 --> 00:00:33,150 We also have a few associated readings 15 00:00:33,150 --> 00:00:35,820 and a three minute video that gives a clear perspective 16 00:00:35,820 --> 00:00:39,280 on supervised versus unsupervised classification. 17 00:00:39,280 --> 00:00:40,180 Let's get started. 18 00:00:41,470 --> 00:00:44,080 Last week we talked all about image processing. 19 00:00:44,080 --> 00:00:47,580 In fact, many of the concepts that we went over last week 20 00:00:47,580 --> 00:00:50,390 and defined during the lecture 21 00:00:50,390 --> 00:00:52,590 are the exact things that we will be doing 22 00:00:52,590 --> 00:00:54,660 and making in Google Earth Engine 23 00:00:54,660 --> 00:00:56,610 using the tutorials that are embedded 24 00:00:56,610 --> 00:00:58,760 in the readings for this week. 25 00:00:58,760 --> 00:01:02,310 So digital image processing are the steps that we take 26 00:01:02,310 --> 00:01:04,313 to turn satellite data into images. 27 00:01:05,320 --> 00:01:07,210 There are four basic divisions or categories 28 00:01:07,210 --> 00:01:08,840 of processing that we talked about, 29 00:01:08,840 --> 00:01:10,560 including restoration, enhancement, 30 00:01:10,560 --> 00:01:13,120 transformation and classification. 31 00:01:13,120 --> 00:01:16,210 Image restoration is often called pre-processing, 32 00:01:16,210 --> 00:01:19,030 because it is mostly concerned with methods 33 00:01:19,030 --> 00:01:21,440 that geometrically and radiometrically correct 34 00:01:21,440 --> 00:01:24,830 the satellite sensor data before it's made into an image. 35 00:01:24,830 --> 00:01:27,320 Broadly speaking, image restoration is concerned 36 00:01:27,320 --> 00:01:29,260 with the correction and calibration of images 37 00:01:29,260 --> 00:01:31,700 in order to achieve as faithful a representation 38 00:01:31,700 --> 00:01:33,553 of the Earth's surface as possible. 39 00:01:34,420 --> 00:01:37,015 Radiometric restoration deals with 40 00:01:37,015 --> 00:01:39,540 the correcting of distortions 41 00:01:39,540 --> 00:01:41,540 in the degree of electromagnetic energy 42 00:01:41,540 --> 00:01:43,970 registered by each detector. 43 00:01:43,970 --> 00:01:46,600 These can be due to atmospheric corrections 44 00:01:46,600 --> 00:01:48,550 or sensor malfunction. 45 00:01:48,550 --> 00:01:50,890 Geometric restoration, on the other hand, 46 00:01:50,890 --> 00:01:53,100 entails correcting the distortions in an image 47 00:01:53,100 --> 00:01:55,820 brought about by sensor and Earth positioning. 48 00:01:55,820 --> 00:01:59,360 Some reasons why we use correction processes include 49 00:01:59,360 --> 00:02:01,800 variations in the altitude, attitude 50 00:02:01,800 --> 00:02:04,450 and velocity of the sensor platform, 51 00:02:04,450 --> 00:02:08,050 Earth's curvature, Earth's eastward rotation, 52 00:02:08,050 --> 00:02:11,003 atmospheric refraction and relief displacement. 53 00:02:12,230 --> 00:02:14,130 We went over correction processes, 54 00:02:14,130 --> 00:02:16,870 including using image transformation equations 55 00:02:16,870 --> 00:02:19,740 in conjunction with ground control points. 56 00:02:19,740 --> 00:02:22,940 Three types of re-sampling procedures that we discussed 57 00:02:22,940 --> 00:02:25,530 are nearest neighbor, by linear interpolation 58 00:02:26,680 --> 00:02:28,170 and by cubic interpolation, 59 00:02:28,170 --> 00:02:30,123 which is also called cubic convolution. 60 00:02:31,400 --> 00:02:34,090 Image enhancement is predominantly concerned 61 00:02:34,090 --> 00:02:36,250 with a modification of images 62 00:02:36,250 --> 00:02:39,300 to optimize their appearance to the visual system. 63 00:02:39,300 --> 00:02:41,500 Visual analyses, as we went over last week, 64 00:02:41,500 --> 00:02:44,260 is a key element even in digital image processing, 65 00:02:44,260 --> 00:02:46,950 and the effects of these techniques can be dramatic. 66 00:02:46,950 --> 00:02:49,560 Specifically we talked about contrast stretching 67 00:02:49,560 --> 00:02:53,670 in two types of compositing, both spectral and temporal. 68 00:02:53,670 --> 00:02:55,780 We employed compositing techniques, 69 00:02:55,780 --> 00:02:59,675 or we usually employ compositing techniques 70 00:02:59,675 --> 00:03:01,290 as image correction, 71 00:03:01,290 --> 00:03:04,890 due to sensor issues and atmospheric interference. 72 00:03:04,890 --> 00:03:06,500 Temporal composites can be a way 73 00:03:06,500 --> 00:03:08,330 to backfill corrupted pixels, 74 00:03:08,330 --> 00:03:09,880 but they are also commonly used 75 00:03:09,880 --> 00:03:11,720 to make an image representation 76 00:03:11,720 --> 00:03:13,710 or representative, sorry, of a time period 77 00:03:13,710 --> 00:03:16,460 that is longer than a single snapshot in time. 78 00:03:16,460 --> 00:03:19,920 For instance, if we are comparing annual changes 79 00:03:19,920 --> 00:03:22,540 in the growing season length between two regions, 80 00:03:22,540 --> 00:03:25,290 we would compare temporal image composites 81 00:03:25,290 --> 00:03:27,020 to increase our confidence in the rate 82 00:03:27,020 --> 00:03:28,963 and timing of change between regions. 83 00:03:30,450 --> 00:03:33,000 Image transformation refers to the derivation 84 00:03:33,000 --> 00:03:36,360 of new imagery as a result of some mathematical treatment 85 00:03:36,360 --> 00:03:38,380 of the raw image bands. 86 00:03:38,380 --> 00:03:40,540 So while enhancement and restoration 87 00:03:40,540 --> 00:03:42,700 have methods that involve correcting sensor data 88 00:03:42,700 --> 00:03:44,360 to make it a better representation 89 00:03:44,360 --> 00:03:46,240 of features on the Earth's surface, 90 00:03:46,240 --> 00:03:49,130 image transformation and classification are employed 91 00:03:49,130 --> 00:03:52,160 as methodologies towards interpreting images 92 00:03:52,160 --> 00:03:54,643 using mathematic and scientific methodologies. 93 00:03:55,860 --> 00:03:58,450 Two examples of transformation that we talked about 94 00:03:58,450 --> 00:04:01,510 were vegetation indices and principle components analysis, 95 00:04:01,510 --> 00:04:02,980 but there are many others, 96 00:04:02,980 --> 00:04:06,070 and we'll go over in more detail a little bit today 97 00:04:06,070 --> 00:04:07,773 about vegetation indices. 98 00:04:08,910 --> 00:04:11,330 We also talked about image classification, 99 00:04:11,330 --> 00:04:12,163 which refers to 100 00:04:12,163 --> 00:04:14,970 the computer-assisted interpretation of images. 101 00:04:14,970 --> 00:04:16,080 We compared supervised 102 00:04:16,080 --> 00:04:19,000 versus unsupervised classification methods, 103 00:04:19,000 --> 00:04:20,410 where a major difference is that 104 00:04:20,410 --> 00:04:22,760 the supervised classification uses training sites 105 00:04:22,760 --> 00:04:24,640 to distinguish between feature classes 106 00:04:24,640 --> 00:04:29,330 and unsupervised classification uses clustering methods 107 00:04:29,330 --> 00:04:32,680 to group pixels with similar spectral signatures together 108 00:04:32,680 --> 00:04:34,820 into information classes. 109 00:04:34,820 --> 00:04:36,960 As analysts, we have a role to play 110 00:04:36,960 --> 00:04:38,730 in both classification methods. 111 00:04:38,730 --> 00:04:42,090 For supervised classification, this role is a priority 112 00:04:42,090 --> 00:04:43,860 or done before the classification 113 00:04:43,860 --> 00:04:46,180 and involves the choosing of enough training sites 114 00:04:46,180 --> 00:04:48,860 and knowledge of what is on the ground. 115 00:04:48,860 --> 00:04:50,630 In unsupervised classification, 116 00:04:50,630 --> 00:04:53,330 our only role is to group the information classes 117 00:04:53,330 --> 00:04:56,670 into feature classes after the classification has been done, 118 00:04:56,670 --> 00:04:59,670 and again, with knowledge about what is on the ground. 119 00:04:59,670 --> 00:05:03,020 And finally, we spoke briefly about the masking of water, 120 00:05:03,020 --> 00:05:05,800 whether it be solid, liquid or gas. 121 00:05:05,800 --> 00:05:07,650 This week, you will learn how to build 122 00:05:07,650 --> 00:05:10,053 and apply a mask using Google Earth Engine. 123 00:05:12,730 --> 00:05:15,010 Okay, so hopefully now from our readings, 124 00:05:15,010 --> 00:05:16,540 you already have a sense 125 00:05:16,540 --> 00:05:19,660 of what image collections are in Google Earth Engine. 126 00:05:19,660 --> 00:05:21,590 In many cases, creating an image collection 127 00:05:21,590 --> 00:05:23,870 is the first step in any workflow process 128 00:05:23,870 --> 00:05:26,400 for creating or interpreting images. 129 00:05:26,400 --> 00:05:28,640 One of the most fundamental parts of Google Earth Engine 130 00:05:28,640 --> 00:05:31,770 is selecting the images you need and displaying them. 131 00:05:31,770 --> 00:05:34,010 Here I've just put up a really colorful example 132 00:05:34,010 --> 00:05:35,560 of a workflow where someone developed 133 00:05:35,560 --> 00:05:37,810 a data product of global temperature 134 00:05:37,810 --> 00:05:41,010 and then created a image collection 135 00:05:41,010 --> 00:05:43,760 of the global temperature product, 136 00:05:43,760 --> 00:05:46,430 and finally, used that to create a GIF 137 00:05:46,430 --> 00:05:48,610 to sort of animate through time 138 00:05:48,610 --> 00:05:50,660 what's happening with global temperature. 139 00:05:54,350 --> 00:05:56,450 So at this point, I expect that you all 140 00:05:56,450 --> 00:05:59,650 understand what an image collection is 141 00:05:59,650 --> 00:06:00,760 in Google Earth Engine 142 00:06:00,760 --> 00:06:05,760 and how we start with a satellite repository essentially 143 00:06:06,200 --> 00:06:08,920 and we import that into our script, 144 00:06:08,920 --> 00:06:12,600 and then we build an image collection from that repository. 145 00:06:12,600 --> 00:06:14,070 There are numerous ways to reduce 146 00:06:14,070 --> 00:06:17,600 an entire satellite repository into an image collection. 147 00:06:17,600 --> 00:06:21,340 We can filter over space, we can filter over bands, 148 00:06:21,340 --> 00:06:23,090 we can also filter over time 149 00:06:23,980 --> 00:06:27,500 and we can filter over combinations of all of these. 150 00:06:27,500 --> 00:06:30,170 So we use filtering and mathematical reduction methods 151 00:06:30,170 --> 00:06:32,110 in Google Earth Engine to create an image 152 00:06:32,110 --> 00:06:33,360 from an image collection. 153 00:06:34,370 --> 00:06:35,360 We can also use them 154 00:06:35,360 --> 00:06:38,680 to create an image collection from a repository, 155 00:06:38,680 --> 00:06:41,140 many different combinations of both. 156 00:06:41,140 --> 00:06:43,163 In the figure here, I'm showing- 157 00:06:44,430 --> 00:06:47,150 The words reduce are used 158 00:06:47,150 --> 00:06:51,420 and reduction methods are mathematical in nature, 159 00:06:51,420 --> 00:06:54,820 and typically what they entail are writing functions 160 00:06:54,820 --> 00:06:59,820 that resample the data set for a given statistical equation. 161 00:07:02,330 --> 00:07:04,640 So mean, median, minimum, maximum 162 00:07:04,640 --> 00:07:06,160 standard deviation or percentile. 163 00:07:06,160 --> 00:07:09,060 That's also another way to go from 164 00:07:09,060 --> 00:07:10,710 image collections to images. 165 00:07:10,710 --> 00:07:12,530 That's where we take that stack of images 166 00:07:12,530 --> 00:07:15,660 and then we take, say, the mean pixel, 167 00:07:15,660 --> 00:07:18,910 the mean per pixel of all of the different images 168 00:07:18,910 --> 00:07:22,003 that are in that stack in order to create our output image. 169 00:07:24,420 --> 00:07:27,620 Last week, we talked a bit about image composites 170 00:07:27,620 --> 00:07:31,250 and in the context of both spatial 171 00:07:31,250 --> 00:07:33,630 and temporal compositing. 172 00:07:33,630 --> 00:07:34,860 So with spatial compositing, 173 00:07:34,860 --> 00:07:36,890 a lot of times we are correcting images 174 00:07:36,890 --> 00:07:39,710 and we're backfilling pixels that might have 175 00:07:40,930 --> 00:07:44,310 sensor issues with them or perhaps atmospheric ones. 176 00:07:44,310 --> 00:07:46,030 With temporal composites, 177 00:07:46,030 --> 00:07:48,550 sometimes that's the method by which we backfill, 178 00:07:48,550 --> 00:07:52,010 but we also can use them to gain a better understanding 179 00:07:52,010 --> 00:07:55,420 of what's going on for our given site through time. 180 00:07:55,420 --> 00:07:58,340 So here I'm showing an example workflow schematic 181 00:07:58,340 --> 00:08:01,530 for starting with Landsat 2 and Sentinel collections 182 00:08:01,530 --> 00:08:04,570 and creating an image composite using both. 183 00:08:04,570 --> 00:08:05,970 So we can also make fusions 184 00:08:06,941 --> 00:08:10,350 of two different satellite projects, products, I'm sorry, 185 00:08:10,350 --> 00:08:13,600 and the way in which we do this in Google Earth Engine 186 00:08:13,600 --> 00:08:16,670 is by starting with these different collections, 187 00:08:16,670 --> 00:08:19,890 you know, applying filters to kind of gain 188 00:08:19,890 --> 00:08:21,400 the date range that we're interested in, 189 00:08:21,400 --> 00:08:23,600 an area of interest, the band range. 190 00:08:23,600 --> 00:08:26,710 Then we might apply a cloud mask or maybe a water mask. 191 00:08:26,710 --> 00:08:28,960 We then would use a reducing function 192 00:08:28,960 --> 00:08:33,960 to get from all of those different data sets 193 00:08:34,800 --> 00:08:36,750 all the way down to one image composite 194 00:08:36,750 --> 00:08:39,480 that's representative of the area 195 00:08:39,480 --> 00:08:40,990 that we're interested in looking at 196 00:08:40,990 --> 00:08:44,500 and during the time period and including the bands 197 00:08:44,500 --> 00:08:46,400 that we're interested in working with. 198 00:08:49,160 --> 00:08:50,570 All right, so we'll spend a few minutes now 199 00:08:50,570 --> 00:08:52,130 talking about spectral indices. 200 00:08:52,130 --> 00:08:54,200 And as we talked about last week, 201 00:08:54,200 --> 00:08:56,170 there are a whole suite of spectral indices 202 00:08:56,170 --> 00:09:00,140 that we have to use at our fingertips. 203 00:09:00,140 --> 00:09:02,450 And, you know, basically what we can do 204 00:09:02,450 --> 00:09:04,920 or what we do when we are calculating spectral indices 205 00:09:04,920 --> 00:09:07,280 is we're taking advantage of the differences 206 00:09:07,280 --> 00:09:10,970 in percent reflectance across the different wavelengths 207 00:09:11,820 --> 00:09:15,240 that different features of the landscape have. 208 00:09:15,240 --> 00:09:17,900 And so, as we talked about a bunch last week, 209 00:09:17,900 --> 00:09:21,220 normalized difference vegetation index, or NDVI, 210 00:09:21,220 --> 00:09:25,260 is the most widely used vegetation index in research. 211 00:09:25,260 --> 00:09:26,930 It's mainly used to monitor and assess 212 00:09:26,930 --> 00:09:29,550 the greenery from regional to global scales, 213 00:09:29,550 --> 00:09:31,540 although I've also seen it applied 214 00:09:31,540 --> 00:09:33,463 at sort of finer scales as well. 215 00:09:34,512 --> 00:09:36,080 And it quantifies vegetation 216 00:09:36,080 --> 00:09:37,150 by computing the difference 217 00:09:37,150 --> 00:09:39,650 between the near infrared and the red light. 218 00:09:39,650 --> 00:09:42,900 So the equation again, as showed at the bottom left, 219 00:09:42,900 --> 00:09:45,960 and as we discussed last night or last week, sorry, 220 00:09:45,960 --> 00:09:48,593 chlorophyll strongly absorbs visible light. 221 00:09:49,462 --> 00:09:50,900 And the cellular structure of the leaves 222 00:09:50,900 --> 00:09:53,190 strongly reflects near infrared light, 223 00:09:53,190 --> 00:09:55,110 which is invisible to the human eyes. 224 00:09:55,110 --> 00:09:57,120 So unhealthy vegetation absorbs 225 00:09:57,120 --> 00:10:00,320 more of the near infrared light rather than reflecting it. 226 00:10:00,320 --> 00:10:02,640 So in that way, too, you know, 227 00:10:02,640 --> 00:10:05,350 we can look at correlations of near infrared changes 228 00:10:05,350 --> 00:10:08,863 and compare them to the red light and measure plant health. 229 00:10:09,980 --> 00:10:14,800 So the value of NDVI ranges between negative one and one. 230 00:10:14,800 --> 00:10:19,363 The NDVI values near positive one, between 0.6 and 0.9, 231 00:10:20,490 --> 00:10:23,150 usually indicate very dense vegetation. 232 00:10:23,150 --> 00:10:26,860 The negative values, so negative one to near zero 233 00:10:26,860 --> 00:10:29,950 is usually representing clouds, water and snow, 234 00:10:29,950 --> 00:10:32,650 and the values near zero but positive 235 00:10:32,650 --> 00:10:35,343 are usually indicating things like barren rock, sand, 236 00:10:36,309 --> 00:10:38,190 sometimes snow or dirty snow. 237 00:10:38,190 --> 00:10:42,210 So moderate NDVI values are often indicative, 238 00:10:42,210 --> 00:10:43,610 so like between 0.2 and 0.5, 239 00:10:45,136 --> 00:10:47,400 they're usually indicative of sparse vegetation 240 00:10:47,400 --> 00:10:49,630 or shrubs and grasslands. 241 00:10:49,630 --> 00:10:53,313 Often too, we can also use this to look at crop health. 242 00:10:54,246 --> 00:10:57,450 So in that same value range, 243 00:10:57,450 --> 00:11:00,623 you can also find, you know, crops. 244 00:11:02,740 --> 00:11:06,190 So, as I said before, the application of NDVI 245 00:11:06,190 --> 00:11:08,360 is pretty widely diversified. 246 00:11:08,360 --> 00:11:10,420 It's one of the most used indexes 247 00:11:10,420 --> 00:11:13,710 and it's used in precision agriculture farming 248 00:11:13,710 --> 00:11:16,100 for monitoring vegetation health. 249 00:11:16,100 --> 00:11:18,210 So if you're using a time series of NDVI, 250 00:11:18,210 --> 00:11:21,480 you can look at crop health. 251 00:11:21,480 --> 00:11:24,823 Same thing with forest monitoring, also drought monitoring. 252 00:11:25,815 --> 00:11:28,030 And it's also used in the analysis of land use 253 00:11:28,030 --> 00:11:29,423 and land cover changes, 254 00:11:30,350 --> 00:11:33,120 or in assessments of land degradation 255 00:11:34,060 --> 00:11:36,883 and finally and also in urban sprawl as well. 256 00:11:38,030 --> 00:11:41,830 So secondly, a slightly more complicated vegetation index 257 00:11:41,830 --> 00:11:44,353 is the enhanced vegetation index or EVI. 258 00:11:45,197 --> 00:11:47,980 EVI is calculated similarly to NDVI, 259 00:11:47,980 --> 00:11:51,130 but with some additional atmosphere corrections 260 00:11:51,130 --> 00:11:54,150 to meet with the inaccuracies of NDVI. 261 00:11:54,150 --> 00:11:55,950 So EVI is capable of eliminating 262 00:11:55,950 --> 00:11:58,350 the background and atmospheric noise 263 00:11:58,350 --> 00:12:01,310 as well as non-saturation. 264 00:12:01,310 --> 00:12:06,110 So when you have very, very, very green vegetation, 265 00:12:06,110 --> 00:12:10,090 NDVI can basically tend to saturate, 266 00:12:10,090 --> 00:12:13,650 which means it gets so, so close to plus one 267 00:12:13,650 --> 00:12:18,403 that it kind of makes the image not as usable 268 00:12:19,240 --> 00:12:24,230 in terms of assessing change or health through time. 269 00:12:24,230 --> 00:12:25,313 It's not as accurate. 270 00:12:26,630 --> 00:12:29,470 So EVI is capable of eliminating 271 00:12:29,470 --> 00:12:32,153 the background atmospheric noise 272 00:12:32,153 --> 00:12:36,870 to try to account for some of this. 273 00:12:36,870 --> 00:12:38,530 It enhances the vegetation signal 274 00:12:38,530 --> 00:12:41,360 with better sensitivity in high biomass regions. 275 00:12:41,360 --> 00:12:43,900 So it's often used in studies 276 00:12:43,900 --> 00:12:45,973 of tropical rainforest, for instance. 277 00:12:46,870 --> 00:12:49,810 So again, it's another vegetation index 278 00:12:49,810 --> 00:12:52,950 that's used to quantify the greenness of vegetation, 279 00:12:52,950 --> 00:12:57,727 and I've put a more complicated version 280 00:12:59,880 --> 00:13:02,690 of the equation in the middle. 281 00:13:02,690 --> 00:13:06,330 This week, we'll be asking you to calculate EVI 282 00:13:06,330 --> 00:13:10,890 and we'll have you use a more simplified version of that. 283 00:13:10,890 --> 00:13:14,030 So there's a couple of different equations that we use, 284 00:13:14,030 --> 00:13:15,230 but I wanted to put this here 285 00:13:15,230 --> 00:13:19,420 just to kind of differentiate as well, too, 286 00:13:19,420 --> 00:13:23,480 that we can in fact calculate these indices 287 00:13:25,770 --> 00:13:27,410 in a couple of different ways. 288 00:13:27,410 --> 00:13:32,070 And here we're showing there's an L in the equation, 289 00:13:32,070 --> 00:13:34,720 which is an adjustment for canopy background 290 00:13:35,930 --> 00:13:40,750 to address like the saturation of vegetation greenness. 291 00:13:40,750 --> 00:13:43,720 There's also coefficients of aerosol resistance 292 00:13:43,720 --> 00:13:48,140 to kind of mitigate aerosol reflectance. 293 00:13:48,140 --> 00:13:53,140 But essentially another way to calculate this index 294 00:13:53,548 --> 00:13:58,548 is to pit the blue wavelength, 295 00:14:02,060 --> 00:14:04,550 basically subtract it from the red wavelength 296 00:14:04,550 --> 00:14:07,173 in the denominator of the equation. 297 00:14:10,320 --> 00:14:11,260 Finally- 298 00:14:11,260 --> 00:14:13,552 Oh, and I guess I should also add that 299 00:14:13,552 --> 00:14:16,910 in terms of the application of EVI monitoring, 300 00:14:16,910 --> 00:14:21,400 vegetation is again typically how it's used, 301 00:14:21,400 --> 00:14:23,900 but it's got higher precision for agriculture, 302 00:14:23,900 --> 00:14:25,710 and like I said, dense forest land, 303 00:14:25,710 --> 00:14:27,010 especially in the tropics. 304 00:14:28,071 --> 00:14:30,180 And so sometimes I've also seen it used 305 00:14:30,180 --> 00:14:32,253 to calculate crop yield. 306 00:14:33,480 --> 00:14:35,609 So again, you know, you will be doing this 307 00:14:35,609 --> 00:14:37,740 this week in Google Earth Engine, 308 00:14:37,740 --> 00:14:40,380 but you won't be calculating it using the coefficients 309 00:14:40,380 --> 00:14:43,893 or the L either. 310 00:14:44,970 --> 00:14:47,470 So finally, I also wanted to talk 311 00:14:47,470 --> 00:14:49,840 a little bit about the normalized difference water index, 312 00:14:49,840 --> 00:14:53,560 which the simplified equation for, again, 313 00:14:53,560 --> 00:14:55,920 is shown down at the bottom right here. 314 00:14:55,920 --> 00:14:57,320 The normalized difference water index 315 00:14:57,320 --> 00:14:58,830 is remote sensing derived index, 316 00:14:58,830 --> 00:15:03,300 which measures the change in water content of leaves 317 00:15:03,300 --> 00:15:06,370 using the near infrared and the short wave infrared, 318 00:15:06,370 --> 00:15:08,620 so SWIR wavelengths. 319 00:15:08,620 --> 00:15:11,870 This was a method that was developed by a researcher, 320 00:15:11,870 --> 00:15:14,970 his last name is Gao, back in 1996. 321 00:15:14,970 --> 00:15:19,030 So NDWI is used to analyze vegetation's water contents 322 00:15:19,030 --> 00:15:20,940 because of its sensitivity to the water content 323 00:15:20,940 --> 00:15:22,653 of vegetation and water bodies. 324 00:15:23,780 --> 00:15:28,780 So again, the value of NDWI varies between zero, 325 00:15:29,200 --> 00:15:32,830 or sorry, negative 0.1 to 1.0, 326 00:15:32,830 --> 00:15:35,210 and it depends on the hardwood content 327 00:15:35,210 --> 00:15:37,910 and the nature of the vegetation cover. 328 00:15:37,910 --> 00:15:40,100 The higher value of NDWI represents 329 00:15:40,100 --> 00:15:42,060 the higher water content in vegetation 330 00:15:42,060 --> 00:15:44,960 and a higher fraction of vegetation cover. 331 00:15:44,960 --> 00:15:49,633 And the low NDWI indicates low vegetation water content 332 00:15:49,633 --> 00:15:53,700 and also low vegetation fraction cover. 333 00:15:53,700 --> 00:15:57,200 But when we subtract these, we can also look at stress. 334 00:15:57,200 --> 00:16:00,830 So when water is the stress factor in an image, 335 00:16:00,830 --> 00:16:05,490 the value of NDWI will decrease. 336 00:16:05,490 --> 00:16:08,070 It's also used to determine the moisture content 337 00:16:08,070 --> 00:16:11,610 in vegetation cover during an assessment of fire risk. 338 00:16:11,610 --> 00:16:15,980 So the Forest Service actually can use NDWI values 339 00:16:15,980 --> 00:16:17,440 to indicate whether there are 340 00:16:17,440 --> 00:16:20,070 sufficient moisture in fuel loads 341 00:16:20,070 --> 00:16:22,279 and use that to calculate 342 00:16:22,279 --> 00:16:26,320 how an area might be increasing in drought stress 343 00:16:26,320 --> 00:16:29,733 and increasing in fire risk and vice versa. 344 00:16:31,030 --> 00:16:33,020 So, yeah, as I said, 345 00:16:33,020 --> 00:16:35,710 the applications of NDWI are drought monitoring, 346 00:16:35,710 --> 00:16:38,870 fire risk assessment, water stress analysis. 347 00:16:38,870 --> 00:16:42,043 It's also used in reservoir mapping and discharging, 348 00:16:42,890 --> 00:16:45,170 and then it can also be used in agriculture 349 00:16:45,170 --> 00:16:47,593 in the assessment of yield reductions, 350 00:16:48,620 --> 00:16:53,393 or to determine whether ground water levels are lowering. 351 00:16:56,010 --> 00:17:01,010 So how do we detect changes between remote sensing images? 352 00:17:01,150 --> 00:17:04,090 Due to the geospatial format of raster imagery 353 00:17:04,090 --> 00:17:06,270 and repeat measurement by satellites, 354 00:17:06,270 --> 00:17:09,310 remote sensing is a very well suited scientific tool 355 00:17:09,310 --> 00:17:11,950 with which to detect changes across the landscape 356 00:17:11,950 --> 00:17:13,970 on a pixel by pixel basis. 357 00:17:13,970 --> 00:17:16,550 So as with other processes in remote sensing, 358 00:17:16,550 --> 00:17:19,290 there are numerous ways to detect change. 359 00:17:19,290 --> 00:17:21,460 The methods are typically specific 360 00:17:21,460 --> 00:17:24,260 to the question that the user is asking. 361 00:17:24,260 --> 00:17:27,400 So here at the bottom, I am showing a figure, 362 00:17:27,400 --> 00:17:28,850 or the figure here is showing 363 00:17:29,747 --> 00:17:32,110 four of the most commonly used change detection methods 364 00:17:32,110 --> 00:17:34,880 as applied to a registered normalized 365 00:17:34,880 --> 00:17:37,100 multi temporal Landsat image. 366 00:17:37,100 --> 00:17:39,560 So the methods shown from left to right 367 00:17:39,560 --> 00:17:41,003 that are mapped are, 368 00:17:42,206 --> 00:17:45,227 on the left, we've shown a post classification detection. 369 00:17:47,410 --> 00:17:50,730 The second one is using image differencing, 370 00:17:50,730 --> 00:17:53,710 so we just subtract one image from the other. 371 00:17:53,710 --> 00:17:56,610 The third is using image ratioing. 372 00:17:56,610 --> 00:17:59,380 So we can look at on a pixel by pixel basis 373 00:17:59,380 --> 00:18:01,883 the ratios of changes in certain areas. 374 00:18:02,830 --> 00:18:04,160 And finally, the fourth one 375 00:18:04,160 --> 00:18:06,550 is looking at a principle components analysis, 376 00:18:06,550 --> 00:18:09,120 so using a principle components analysis of the image 377 00:18:09,120 --> 00:18:12,810 to basically gain an understanding 378 00:18:12,810 --> 00:18:17,253 of where there is a high level of variance 379 00:18:17,253 --> 00:18:20,103 in the features on the landscape. 380 00:18:21,020 --> 00:18:22,350 So in Google Earth Engine, 381 00:18:22,350 --> 00:18:26,094 you can accomplish each of these change detection methods, 382 00:18:26,094 --> 00:18:27,910 and you'll be going into this a little bit more 383 00:18:27,910 --> 00:18:31,170 I think in the future of this course. 384 00:18:31,170 --> 00:18:32,280 We might touch on it briefly 385 00:18:32,280 --> 00:18:34,230 in the tutorials this week as well. 386 00:18:34,230 --> 00:18:37,820 But just know that these change detection methods 387 00:18:37,820 --> 00:18:41,570 can and are applied both to individual images, 388 00:18:41,570 --> 00:18:44,360 and also they're actually even more classically 389 00:18:44,360 --> 00:18:47,060 often applied to image composites. 390 00:18:47,060 --> 00:18:50,560 So I think I said sort of in a previous example, 391 00:18:50,560 --> 00:18:53,520 you know, if you're interested in looking at 392 00:18:53,520 --> 00:18:55,901 changes to vegetation, you know, 393 00:18:55,901 --> 00:19:00,901 you would set out to first gather your image collection 394 00:19:01,280 --> 00:19:03,290 and then filter it by dates. 395 00:19:03,290 --> 00:19:05,600 If it were me, I'd be filtering out dates 396 00:19:05,600 --> 00:19:07,660 that are not included in the growing season 397 00:19:07,660 --> 00:19:10,980 so as not to add more sort of noise to my data set. 398 00:19:10,980 --> 00:19:12,160 So I'd only take images 399 00:19:12,160 --> 00:19:15,300 that were during a given growing season. 400 00:19:15,300 --> 00:19:19,124 I might stack those images up and then use maybe a mean 401 00:19:19,124 --> 00:19:21,650 to sort of summarize 402 00:19:21,650 --> 00:19:24,110 where there's vegetation during the growing season. 403 00:19:24,110 --> 00:19:26,900 And then if we do this for a bunch of years, 404 00:19:26,900 --> 00:19:28,530 you can then use band math 405 00:19:28,530 --> 00:19:31,540 and other tools in Google Earth Engine 406 00:19:31,540 --> 00:19:34,850 to subtract out and create difference images. 407 00:19:34,850 --> 00:19:37,000 So there's the whole different way of, 408 00:19:37,000 --> 00:19:39,330 there's a bunch of different ways to do this, 409 00:19:39,330 --> 00:19:42,820 and I think we're definitely gonna touch on it 410 00:19:42,820 --> 00:19:45,400 a bit in this course and it's something to think about 411 00:19:45,400 --> 00:19:48,860 as we start to consider what we might wanna be doing 412 00:19:48,860 --> 00:19:51,400 for our end of semester projects. 413 00:19:51,400 --> 00:19:55,550 So please let me know if you have any questions this week 414 00:19:55,550 --> 00:19:58,030 as you go through the tutorials 415 00:19:58,030 --> 00:20:01,810 that are embedded in the readings that were assigned 416 00:20:01,810 --> 00:20:05,540 and I look forward to seeing you all on Yellowdig 417 00:20:05,540 --> 00:20:10,420 and also hopefully in the live class 418 00:20:12,231 --> 00:20:14,410 that is coming up here pretty soon too. 419 00:20:14,410 --> 00:20:16,010 All right, thank you, take care.