LST, Urban Heat Island Effect, and UTFVI Analysis using Google Earth Engine and Landsat dataset

hello everyone my name is valid and in this video I will show you that how you can use Google Earth engine which is a cloud computing platform to calculate land surface temperature from landsat and then using that lines with temperature how we can quantify Urban heat island index so before starting our course session I highly recommend that you read some articles about how to calculate land surface temperature from remote sensing data sets in this regard I highly suggest you read these two papers the first one is this which uses bullet engine to calculate land surface temperature and latitude land cover at provincial scale and then this paper which use which again uses Google Earth's engine to calculate LSD and a product land use land cover up to 2050. so for this tutorial uh this I will follow the approach used in this paper uh by the way if you want to download this paper you can easily get the PDF copy from my website just navigate to the publication and yeah here click on the click here to see my full publication and here you can find the PDFs and links to all of my papers so for This research for this tutorial we will focus on this paper so you can download the PDF right here okay once you have downloaded the PDF you can navigate to this section 2.5 retrieving line surface temperature and now let's jump into the code the first we will filter an image collection of landsat here um I am planning to use landsately some satellite image and filter it to get a single image so let's quickly search the data set for our image so here as you can see we have already the code which has the scaling factors and all the things so I will just copy the code paste it right into our code editor and modify it so um as we are working with the land surface temperature and I want to quantify Urban heat island effect so it is a better approach if we use any city which has high Urban heat issues so let's quickly search we should a city has um the highest so um here we can see that the Madrid the city in Spain has the highest Urban heat island effect so just as an example we will search the metric Zoom it to the city and here we will just draw RF polygon covering the city by the way this polygon will be used to just filter our image collection and here I will turn this on and now I will modify our image collection properties um for this I will choose the 2022 collection now we have filtered we have just modified the filter date and additional command filter bound which is limit the images just for our this study area and before moving on to the next section it is the vector approach to first check and visualize the image because the JavaScript based Google Earth engine just provide us the map which is which is very helpful in visualizing the satellite image in to see the situation um because in most of the cases we will get an image which has a high cloud cover which will affect our calculated line surface temperature so um for this just uh use I will just use the median to convert the image collection into a single image and here I will just visualize the image okay let's run it as you can see right here we got the image but it has um cloud cover in in the northern region here you can now see so in order to mask the cloud we will use the mask Cloud function I will just copy and paste the codes of the cloud mask so uh here is the code that uses Cloud properties in landsat specifically for landsat 8 surface surface reflection and then mask all the cloud coverage so here I will quickly add the command and one thing to notice here is that when we are applying a function the function can only be applied to the image collection not to the image because um last time when we ran the code we have the median uh function which converts the image collection into a single image so if I write the function here it will give us an error because it received a single image so we have to apply the cloud mask before the median so I will quickly copy the function name and paste it here now as you can see after applying the cloud mask we have removed all the clouds from our image and overall now I'll show you what are the steps in calculating line surface temperature in general so for length surface temperature first we filter an image uh once our image is filtered we evaluate normalized difference vegetation index which will be used to evaluate fraction of agitation then we will calculate the emissivity and at the last we will calculate the last surface temperature so let's quickly calculate the normalized different vegetation index also known as ndvi so to calculate ndbi we will use the near infrared and red band of the landsat which are surface reflectance B5 and B4 so we will give a list and inside the list we will input the band B5 and before and I will just rename it to ndbi um let's just visualize the ndbi quickly here we get the ndvi of our study area now once we have calculated the ndbi we will calculate the statistics of the ndvi specifically we need the minimum and the maximum values of ndvi to calculate the fraction um because if you if you see right here to calculate the proportion of vegetation or the friction we need the ndvi minus minimum value and maximum minus minimum and it's clear so we will I will quickly use um the reducers to calculate the statistics foreign tricks and for the statistics the first statistics is the minimum I [Music] and one important thing to note here is that we are working on the server side uh you need to understand the difference between the server side and the client side because if we just get the number the number will not be able we will not be able to incorporate the number into the later equation because all the analyzes all the equation that we will use for the land surface temperature will be on the server side so we need to convert every number to the server side in order to process it into the equation so for that we I'm using ee DOT number and inside the number I will basically write the reducer reduce region command to calculate the minimum of the maximum statistics so here for the ndbi minimum I will use the image which is ndpi and I will use reduce region to calculate the statistics in that particular region then in reduced region packets and in reduced region I will first input the reducer Dot dot minimum and then I need to specify the geometry so for the geometry is our area of Interest which I import earlier and then I need to enter the scale for this work we are working with the Lancet which has a scale of 30 meter so I will enter the scale of 30 and lastly um we need to enter the maximum pixels value um so maximum we pixels is equal to 1 e 9. next I need to enter the values command uh which will convert the server side into a client side and then get the first value from the dictionary which is our statistics um the zero and similarly we will copy and paste this for the maximum of the ndvi and just change the status reducer to the Max and all other things will be same once we have the minimum and the maximum producer the next step is to calculate the fraction of vegetation which is the fraction of vegetation and now we will um input the same equation um here so first the ndvi and then we have to subtract the minimum [Music] and we then we have to divide maximum minus minimum divide ndvi makes dot subtract ndvi Dot slash and then we need to um enter a square of this whole equation uh so since this is on the server side I will use the how um just for the Simplicity you can always search your command in the doc so for so you can easily search the power command and in the power we need to enter R2 again here if we just enter the two it will not work because 2 is a client-side object so we need to convert it into the surface side so e number and two and this is our whole equation for calculating the fraction of validation and next so once you have calculated by the way I will just add the rename command so rename it to fraction of vegetation yes so once you have calculated the fraction of agitation the next step is to calculate the emissivity and if you go to the paper here is the equation to calculate the emissive video basically we have to multiply this sector into our fraction of agitation and then we have to add this fraction into the total so in in Earth engine uh we just specify um as the image CVT variable so um to multiply I will use the fraction of vegetation dot multiply use this value 0.004 and then I can add the next fraction which was 0.986 people again ee DOT number because it is the server side and nine nine six and six nine eight six okay so once we have this equation our emissivity is uh created uh I will just rename it to uh emissivity and once we have our fraction of agitation our emissivity the last step is to use um this equation to calculate the land surface temperature so in Earth engine we can use either this approach to enter the equation or we can use the Expression um to use expression you can search expression and here is how we can use expression so basically we enter expression as a string and then we supply all of its inputs um like if we are using some variable as a name in the expression we can mention the full uh form of variable so I will just show you how to calculate so for LST LST I will use our image basically I forgot to create a single image so in order to calculate the LST we have to define a thermal image in here in our collection it has all the bands but if in the final as St calculation we only need to process the thermal band so if I go to the properties uh and search for the thermal Band Here is our thermal pen here I will write thermal and I'll just filter out this band I mean dot select and thermal band name which is s t B10 P10 and I will also rename it or thermal so this will be our single image containing a single band which will be used for calculating the LST in later on so here we will use this thermal image so in thermal image we want to apply the expression and in the expression we have to give the string as the expression and then our dictionary okay so um just quickly copy and paste all the things I will just quickly copy or paste all the things but if you want uh to understand the equation in the details you can read these five for example here I'm using the thermal band and adding one into the thermal band and the p is a constant and then the log of the emissivity which we earlier created using the fraction of agitation so um for to calculate the lens surface temperature here is the overall equation that I will use We have basically all the variables the TB is the thermal band and this is again the fraction of the thermal band P is uh this constant value and this is log of e is the emissivity which we just created using the fraction of agitation and lastly to convert the Kelvin into the Centigrade we will use this vector so I will just quickly write this down into the expression form in our code editor okay so now I have created the overall expression and I will simplify it to you so as as we can see that in the equation we have the thermal band and then divided by 1 plus expression multiplied by log of emissivity so similar to that I have write the 30 DB so basically DB is the variable that we later Define in the dictionary of the expression uh right here and then one plus uh this scale factor to the thermal band which is again DB and here I have just simplified this value which was 1.438 multiplied and raised to minus 2 and I have just simplify it into into a number and then I have used the log of the emissivity em again is a variable and at the last uh the factor to converting the Kelvin into the Centigrades so after our expression in the dictionary I have the TB which is the thermal band and then the Em which is the emissivity and lastly I have renamed all this to the LST and now I will just create some visualization parameters and then I will like our newly created LST so okay okay now I have added the visualization parameters I have just copy and paste the palette for visualize the LSD and until last I have added the map dot add clear command to I will just add the map.strander object as well assistant object which is area of interest and 10 which is the Zone level and now I will run it to see if okay if we got the error at line 50 okay in DOT number oh okay okay add the bracket here instead because we wanted to enter the number inside of the ee DOT number and the bracket was closed before getting the values to the client side so similarly here and now let's check if the error is solved yes see so see okay if we have our LST visualization just quickly stretch it so as you can see this is my color palette which I input the code manually [Music] okay so we have like uh the values between 22 to 52 degrees Centigrade which is a really high temperature in Madrid I will apply it and here you can see that we can easily observe the high temperature in the city core areas and the low temperature in the nearby for our rest allergies enable the satellite image to view it yeah so where there is a vegetation the temperature is much lesser than the urban core areas where there is high temperature let's check how much is the temperature so as you can see in the city center the temperature is like 51 degree centigrade and um outside it is so outside it is like 28 degree Centigrade so this is how we can use the Earth engine to calculate the land surface temperature this was the first part of the video now uh once we have the land surface temperature we will continue evaluating the urban heat island effect so Urban heat island effect is basically the phenomenon in which the temperature in the series are much higher than that of the surrounding so as you can see from this visualization that in the city center the temperature is quite High we can see the red spots clearly and compared to the surrounding the the vegetation or the Forest Area it has like a bluish very low temperature so in urban heat island effect we use this phenomena and based on that we quantify the Aerials that are much higher um in in the urban so uh to calculate the urban heat island effect we use multiple methodologies one methodology is to use the proxy which is the urban thermal variance index so um to calculate the urban heat island effect you can go to again my publication section and you can check my this publication here specifically you can find the equation to calculate the urban heat island effect so similarly to our recent approach of calculating LST here we have this so basically the this is the normalized method to calculate the urban heat island effect and to quantify it in terms of its proxy we can use the UT FBI which is TS minus TM divided by TS so t s is the LST which we just calculated T and MD is the mean LST of the study area and TS is a standard deviation value of the study area for calculating the normalized uhi we will need the LST minus mean and divided by the standard deviation let me quickly write the equation so basically we will use the similar method like the reduce region to calculate the statistics I'll just copy and paste it over here so here you can see that we have now calculated the normalized uhi so basically what I did I just followed this equation which was the using the LST we subtract the mean or average value of LST and then divide the standard deviation so um just following the previous reduce region command I calculated the mean using reduce ee reducer dot me and here is the mean value of LSD in this study area which is 40 degree centigrade and then I calculated the LST standard deviation using e reducer dot standard deviation and the standard deviation value is 3.9 something and basically I write the equation which is the LST minus um mean LST and divided by standard deviation So based on that we get the normalized qhi and I just give the parameter so for the parameters specific for the color palette I just copied the color palette for the this the item set which is the urban heat island it's a very good dial side but it's of course 300 meters so if you want some high resolution you can follow this approach using the land side thermal band that yes the urban areas are high Urban heat are under high Urban heat stress than the surrounding which is the first area so you can always customize the range and based on that you can check which areas are are having high heat stress than the other areas so um now the last step for this Workshop is to evaluate the urban thermal field various index which is basically the same as the uhi except instead of using the standard deviation we'll be dividing the value with the LST itself so um let's quickly do this so I have already calculated the statistics of like mean and standard deviation so I don't need to rewrite again um I just need to enter the formula for udfvi which is the LST um subtract the mean and we need to divide the this whole value with the LST itself so let me check if divide yeah divide ourselves foreign so as you can see we now have our UT fvi evaluated for our study area uh what I did in the case of utfei I followed the same approach I um subtract the mean LST and divide the LST with it and just rename it to the new name UT FEI and basically provide the parameter so one thing that is different from uhi is that we did not use the standard deviation and we divided the LST by itself and the second thing is the range so you can always go to the settings and then take the range to the hundred percent stretch to get the minimum and maximum value which from which it will be very easy to visualize your image so um as as you can see from both the UT FEI this is a UT FEI for the study area of Madrid and this one is the uhi so um as you can clearly see that by using the udfyi we can um clearly differentiate the very high hotspots of uh the or the heat stress region and even the uhi is able to differentiate between high heat stress or awkward treatment in the study area and from LST as you can see it's the overall representation of the diversity in the temperature and yeah this this was all for this Workshop um one for this step is if you want to quantify the utfvi into the classes such as they used in this paper it's like none weak to strong as you can Define the range and remap the values but um it's better to just realize in the range because you get very extra details in the in in the data so here if I zoom in so I can easily observe the the changes in the temperature the areas where there is higher vegetation for example the forest has very low UT FBI compared to the urban or bear land which shows very high um Urban thermal field variance index and even if I search for the uhi it has the same patterns as of utfi so in this way um you can easily calculate the LST and you can calculate the uhi and UT FCI using this approach and you can easily scale this up to a much larger region given that you change all the values so basically if you want to apply this to your study area all you need to do is to change the area of interest to your area and then you need to specify the start date and the end date and basically it will run and if you want to use landsat salmon or landsat 5 or even the lens at 9 you can replace um this with the image Collections and of course in case of landsat 9 or 5 or 7 you need to adjust the coefficients used here you can easily find the coefficients from the papers or from a USGS website so that was all for today's Workshop if you like this video um share it with your connection and I will see you in the next tutorial see you

Channel: Mirza Waleed

Views: 8,860

Keywords:

Id: 5W84zme9QmE

Length: 34min 4sec (2044 seconds)

Published: Sat Sep 16 2023