Landsat 7 Landsat 7 Land Surface Temperature (LST) Atmospheric Correction Method Retrieval and Ground Object Temperature Analysis Based on ENVI

1 Preliminary image processing and theoretical part of this article

Update : Landsat inversion of land surface temperature based on GEE can be seen in this blog , automatic batch operation, faster processing.

The operation part of this article will directly start with the calculation of vegetation coverage; for a complete land surface temperature inversion calculation process, there is still a lot of data preparation and preprocessing before the step of calculating the vegetation coverage. In order to better understand the entire experimental process, the steps we have not carried out are sorted out as follows. Among them, you can refer to the specific pre-operation methods:
click here (https://blog.csdn.net/zhebushibiaoshifu/article/details/113915477)

First, obtain the required remote sensing image data, and perform preprocessing steps including the content of the above article-data import, radiometric calibration, geometric correction, atmospheric correction, image stitching and cropping, etc.
Among them, radiation calibration needs to be divided into two steps, namely, the visible light band data (such as the 1, 2, 3, 4, 5, and 7 bands in the above article) and the thermal infrared band data (such as the 6 band in the above article) Carry out radiation calibration separately.
Second, it is necessary to calculate NDVI (ie, Normalized Difference Vegetation Index, normalized vegetation index, not vegetation coverage). NDVI refers to the sum of the difference between the reflection value of the near-infrared band and the reflection value of the red light band in a remote sensing image; it can be used to detect the growth state of vegetation, vegetation coverage, and to eliminate part of the radiation. Error etc. The specific value range of NDVI is limited to between -1 and 1. Its negative value means that the ground is covered with clouds, water, snow, etc., that is, it has high reflection of visible light; a value of 0 means that the ground is covered with rocks or bare soil, etc., so that NIR (Near Infrared, near infrared band) and R (Red, red band) are approximately equal; its positive value indicates that the ground is covered by vegetation, and the higher the vegetation coverage, the higher the value. At present, some websites (such as NASA's official website) have NDVI finished product data, which can be directly downloaded and used by us; and through the near-infrared band data and red light band data in the initial remote sensing image, we can directly use the aforementioned definition formula, namely

Calculate it. When calculating NDVI, it should be noted that the selected remote sensing image should not have excessive cloud interference.
Third, it is necessary to resample the calculated NDVI result image according to the actual situation. This is because in this article, it is necessary to use the "Band Math" tool to calculate the image data multiple times, and this tool requires the input data to be exactly the same in terms of image resolution (ie pixel size) and the number of rows and columns. At the same time, the thermal infrared data we will use (ie Landsat ETM + sixth band data) has a resolution of 60 meters; and the calculated image resolution of the NDVI data is 30 meters. Therefore, we need to resample the NDVI data image with higher resolution accuracy to make the two resolutions consistent.
The resampling function can be realized by selecting "Basic Tools"→"Resize Data (Spatial/Spectral)" in the ENVI software.
At present, the main single-band inversion algorithms for land surface temperature include atmospheric correction method (also known as Radiative Transfer Equation, RTE), single-channel algorithm and single-window algorithm. In this paper, we use the atmospheric correction method. The basic principle of the atmospheric correction method is: estimate the influence of the atmosphere on the surface thermal radiation, and then subtract this part of the atmospheric influence from the total thermal radiation received by the satellite sensor to obtain the surface thermal radiation intensity; finally, the surface thermal radiation The intensity can be converted into the corresponding surface temperature.
The thermal infrared radiance value Lλ received by the satellite sensor is composed of three parts, which are: the upward radiance of the atmosphere L↑, the true radiance of surface objects after passing through the atmosphere, reaching the satellite sensor, and the downward radiance of the atmosphere after passing through the satellite sensor. The energy reflected from the ground. Therefore, combined with the above theoretical process, the radiation transfer equation can be used to express the thermal infrared radiance value received by the satellite sensor:

where ε is the surface emissivity, T_s is the true surface temperature, and B(T_s) is the surface at T_s The thermal radiation brightness of the black body at this real temperature, τ is the transmittance of the thermal infrared band of the atmosphere. The surface specific emissivity ε is also called emissivity, and according to Kirchhoff's law, emissivity and absorptance are equal, then (1-ε) can represent reflectivity. Therefore, [ε*B(T_s )*τ] is the energy that the true radiance of the surface object reaches the satellite sensor after passing through the atmosphere, and [(1-ε) L↓ τ] means the downward radiance of the atmosphere is passing through The energy reflected from the ground.
The specific emissivity of an object is a characterization of the ability of an object to radiate electromagnetic waves outward, and refers to the ratio of the amount of radiation emitted by the surface to the amount of radiation emitted by a black body at the same temperature. It is not only dependent on the composition of the surface object, but also related to the surface state (such as surface roughness) and physical properties (such as dielectric constant, water content, etc.) of the object, and changes with the measured wavelength and observation angle. It is difficult to accurately and quantitatively measure the surface emissivity, so this article estimates the surface emissivity based on the empirical method:


Among them, F_C is the vegetation coverage. It can be seen from this that we need to calculate the surface vegetation coverage at the same time to determine the surface specific emissivity calculation formula and participate in the calculation. In this paper, we adopt the mixed pixel decomposition method to solve the vegetation coverage. It is more consistent with the above-mentioned surface specific emissivity calculation formula. We still divide the surface into three parts: water, vegetation, and buildings; among them, this category is specifically distinguished based on the NDVI value. When NDVI is less than 0, the ground feature is considered to be a water body, and the vegetation coverage is 0; when NDVI is greater than 0.7, the ground feature is considered to be vegetation, and the vegetation coverage is 1; when the NDVI value is [0,0.7], the ground feature is considered to be in the water body Between vegetation and vegetation, the vegetation coverage is calculated according to the formula. For different ground objects, calculate the vegetation coverage:

Finally, after all the above calculations are completed and B(T_s) is obtained according to the radiation transfer equation, the true surface temperature can be obtained according to the inverse function of Planck's formula. The formula is: the

real surface temperature unit is Kelvin (K), and we need to convert it to the common degree Celsius. The conversion of the temperature unit can be calculated separately after obtaining the real temperature image of the surface, or it can be calculated directly in the above formula.

2 actual operation

2.1 Calculation of vegetation coverage

Based on the above analysis, we first need to calculate the vegetation coverage of the study area with the help of the 60-meter resolution NDVI data image of the study area.
(1) Open the ENVI Classic 5.3 (64-bit) software, select "File" → "Open Image File", select the "TM-NDVI-60m.img" file in the pop-up file selection window; click "Open".

(2) Select "Basic Tools" → "Band Math", and enter the first formula of this experiment, the vegetation coverage formula, in the formula creation window that pops up. After entering the formula, click the "Add to List" button below to save the formula into the area to be selected. In order to reduce the unnecessary workload in the later stage, you can click the "Save" button after editing a formula each time to save the formula in the area to be selected.

(3) After saving the formula, click the "OK" button in the lower left corner to start the calculation of the formula. In the formula variable file selection window that pops up, select the variable "B1" of this formula as the file "TM-NDVI-60m.img" we just added. Then, configure the output file address and other information.

(4) After the configuration is completed, click the "OK" button in the lower left corner to start the operation of the formula. At the end of the operation, import the resultant image of vegetation coverage in the study area into the ENVI software, as shown below.

2.2 Calculation of surface emissivity

As mentioned earlier, the calculation of vegetation coverage is actually to prepare for the calculation of surface emissivity. The ground features are also divided into three categories: water, town, and natural surface. Based on the empirical method, using a segmented calculation formula similar to the form of vegetation coverage, different types of ground features are given different surface emissivity calculation formulas.
Considering that the variable band name "B1" is included in the formula used to calculate the vegetation coverage, in order to avoid confusion between the variables of different formulas, I used different variable names in each formula at the beginning ——For example, "B2", "B3", etc. are used in the calculation formula of surface emissivity here. However, through the later experiment process, it was discovered that in fact, even if the variable names of the formulas are consistent each time, the experiment will not be significantly affected.
(1) Select "Basic Tools" → "Band Math", and enter the second formula of this experiment in the pop-up formula creation window, namely the surface emissivity formula. After entering the formula, click the "Add to List" button below to save the formula into the area to be selected. In order to reduce the unnecessary workload in the later period, you can click the "Save" button after editing this formula, and overwrite the last saved formula file to save the two formulas in the area to be selected.

(2) After saving the formula, click the "OK" button in the lower left corner to start the calculation of the formula. In the formula variable file selection window that pops up, select the variable "B2" of this formula as the image file "TM-NDVI-60m.img" we added at the beginning, and select the variable "B3" of this formula It is the result image file of vegetation coverage obtained through the above steps. Then, configure the output file address and other information.

(3) After the configuration is completed, click the "OK" button in the lower left corner to start the operation of the formula. At the end of the run, import the resulting image of the surface emissivity of the study area into the ENVI software, which is displayed as follows.

2.3 Calculation of blackbody radiance value at the same temperature

Based on the analysis in the first part of this article, we can calculate the radiance value of the black body at the same temperature by knowing the specific emissivity of the surface of the study area and the brightness of the thermal infrared band.
(1) Select "File" → "Open Image File" in the ENVI Classic 5.3 (64-bit) software, and select the "TM6-rad-subset-jz-xiangfan.img" file in the pop-up file selection window; click " turn on".
(2) Select "Basic Tools"→"Band Math", and enter the third formula of this experiment in the formula creation window that pops up, namely the blackbody radiance value formula. After entering the formula, click the "Add to List" button below to save the formula into the area to be selected. In order to reduce the unnecessary workload in the later stage, you can click the "Save" button after editing this formula, and overwrite the last saved formula file to save the three formulas in the area to be selected.

(3) After saving the formula, click the "OK" button in the lower left corner to start the calculation of the formula. In the formula variable file selection window that pops up, select the variable "B4" of this formula as the image file "TM6-rad-subset-jz-xiangfan.img" we just added, and set the variable "B5" of this formula "Select the result image file of the surface emissivity obtained through the above steps. Then, configure the output file address and other information.

(4) After the configuration is completed, click the "OK" button in the lower left corner to start the operation of the formula. At the end of the run, the obtained image of the calculated blackbody thermal infrared radiance value at the same temperature in the study area is imported into the ENVI software, which is displayed as follows.

2.4 Calculation of the true surface temperature

From the foregoing analysis, it can be seen that the blackbody thermal infrared band radiance obtained through the above steps is not the actual surface temperature, and we still need to realize the conversion relationship between the two through the inverse Planck formula.
At the same time, the unit of the actual surface temperature obtained in this step is Kelvin (K), which is not the degree Celsius (°C) that we often use in daily life. Therefore, we also need to realize the conversion of temperature units.
The conversion from Kelvin temperature to Celsius only needs to subtract 273.15 from the original temperature, which is relatively simple and does not need to be converted separately. Therefore, I chose to directly complete the conversion of the temperature unit in this step.
(1) Select "Basic Tools" → "Band Math", and enter the fourth formula of this experiment in the formula creation window that pops up, that is, the true surface temperature formula. After entering the formula, click the "Add to List" button below to save the formula into the area to be selected. In order to reduce the unnecessary workload in the later period, you can click the "Save" button after editing this formula, and overwrite the last saved formula file to save the four formulas in the area to be selected.

(2) After saving the formula, click the "OK" button in the lower left corner to start the calculation of the formula. In the formula variable file selection window that pops up, select the variable "B6" of this formula as the image file of the blackbody thermal infrared band radiance result we just calculated, and then configure the output file address and other information.
(3) After the configuration is completed, click the "OK" button in the lower left corner to start the operation of the formula. At the end of the run, import the calculated result image of the real surface temperature of the study area into the ENVI software, which is displayed as follows.

(4) Right-click on any position of the result image and select "Cursor Location/Value" or "Quick Stats" option to view the pixel information of each pixel in the image. Here, I use "Cursor Location/Value" to view the image pixel information and get the result as shown below.

(5) It can be seen that the value of "302.902771" appears in the pixel value of the pixel (ie, the "Data" value). The result of this step is the actual surface temperature in degrees Celsius, and it is impossible to have more than 300 data. Therefore, it is likely that an error occurred in the previous operation part.
(6) Through the return inspection, it was found that when calculating the surface emissivity, there was an input error in the input formula. After correcting this error, the actual surface temperature image is regenerated and statistically inspected.

(7) It can be seen that the modified image data of the real surface temperature conforms to the actual situation, and the error can be considered to have been eliminated.

2.5 Image export

Since it is necessary to make a thematic map in the later stage of the experiment to classify the temperature into different levels and compare the temperature characteristics of different ground objects, it is necessary to save the actual surface temperature result image obtained in ENVI as a ".tif" format to facilitate the use of ArcGIS series The software further classifies, beautifies and produces pictures.
(1) First, I save the image as operation in the ENVI image window. Select "File"→"Save Image As"→"Image File", select the file format "TIFF/GeoTIFF" in the pop-up save configuration window, and configure the result image file save path, save file name, etc.

(2) After getting the saved result image file, add it to the AcrMap 10.2 software. At this point, it is found that the obtained layer file has been stretched to the range of 0-255 in pixels. Through the "recognition" function of AcrMap, it is found that all the pixels of the pixel have been stretched, that is, their original values ​​have become the above-mentioned data in the range of 0-255. Therefore, it is believed that the image should no longer restore the original temperature value through related operations. Later, attempts to adjust with tools such as "Set Raster Attributes" all ended in failure.

(3) By consulting the network resources, I saw someone pointed out that this situation was to save the "image" instead of the "image"; under this reminder, try to directly use the "File" → "Save" in the toolbar at the upper left corner of the ENVI software File As"→"TIFF/GeoTIFF" to save the image. Select the file in the pop-up save configuration window, and configure the result image file save path, save file name, etc.

(4) Import the ".tif" format image result file saved the second time into the AcrMap 10.2 software, and see that the result value is normal.

(5) After getting the normal result, think about the two different preservation results. The software used in this article is ENVI Classic 5.3 (64-bit) version, and the non-Classic version also found this problem. Recalling the process of saving pictures in the image interface, it is further discovered that such saving should directly set the image format to an 8bit plastic storage format, thus losing its original pixel data information and meaning-that is, it only saves this picture The "appearance" of the image uses data with no practical significance between 0 and 255 to represent the gray level of the image; but the original image, that is, the image we hope to obtain, its data and gray level should be the expression of temperature. If the original image with different gray levels divided by different temperatures is changed to an image with different gray levels divided by numbers between 0 and 255, it will naturally lose its original temperature meaning.

(6) After importing the normal image into the AcrMap 10.2 software, it is found that the black border of the image cannot be removed at the same time-using the "recognition" function, you can see that the black border has no data, and it is already in the "NoData" state. . In response to this phenomenon, it may be a common error that occurs when different software saves and reads data files. You don't need to pay attention to it, and it will disappear after re-grading the temperature later.
(7) The final imported image and its information are shown in the figure below. Among them, there are two layers with similar results because at the beginning of the actual surface temperature conversion process, the value I subtracted was "273" instead of "273.15".

3 Thematic map production

The image imported into the AcrMap 10.2 software has pixels ranging from 22.3997 to 47.0569. If the temperature range is not re-divided, the colors of the resulting color image will be very discrete, which is not conducive to observation and analysis.
(1) The imported image itself does not have statistical information, and it is not possible to divide the range of its various values ​​in the "Classified" module. Therefore, it is necessary to use tools to collect statistics on the data. Click "Data Management Tools.tbx" → "Grid" → "Raster Properties" → "Calculate Statistics".

(2) In the pop-up configuration window, select the layer that needs statistical data, and other items do not need to be adjusted.

(3) Click "OK" to complete the calculation of statistical data for the layer image.

(4) Right-click on the layer name, select "Properties" → "Symbolism" → "Classified", open the "Classification" button, select the classification method as "Manual", and set four "break values" as required ——The maximum value of 30, 35, 39 and image pixels is 47.0569. Click OK to save the break point setting.

(5) After returning to the layer properties window, mark the content represented by the four ranges to facilitate later drawing. At the same time, adjust the appropriate color scheme according to the characteristics of the temperature thematic map. Originally, I planned to use red as the color scheme, but I found that the overall distinguishing effect after the picture was not very good. So choose another color scheme with relatively strong contrast.

4 Comparative analysis of surface temperature of different features

In order to compare the difference in surface temperature of different features more clearly, with the help of the satellite map module in Tencent Maps, with the Hanjiang River in Xiangyang City as a control, the regional satellite map and thematic map were compared. The conclusions are as follows:

(1) The temperature of the water body is significantly lower than that of the surrounding cities and villages. This feature can be described as one of the most obvious elements in the temperature thematic map-the thematic map can clearly identify a low temperature zone entering from the west of Xiangyang City, extending from the northwest to the southeast; and forming an impact island in the center of the city to the southwest. , And then continue to travel south. In addition, in the northern part of the impact island, the low temperature zone formed by the Tangbai River is also more obvious; it is clearly distinguishable and can even be traced back to the source of the river along the low temperature zone. The same is true for the Qinzui Reservoir in the southeast of the city. It can be seen from the temperature ribbon of the water body that at the moment of imaging, its temperature is below 30°C. Because the temperature of the water body is closely related to the light, especially the above-ground rivers such as the Han River and Tangbai River; the imaging time of the image used in this experiment is about 10:30 in the morning, and the sunshine time of the day is not long, so the water temperature is low.

(2) The urban temperature is significantly higher than the surrounding area. It can be seen from the temperature thematic map that the main high temperature gathering areas are located along the Han River and Tangbai River; combined with the satellite map, it can be seen that most of these are the main urban gathering centers of Xiangyang, with dense buildings, populations, and developed economies. It can be seen from the urban temperature ribbon that at the moment of imaging, its temperature is above 39°C. On the other hand, for some suburbs and rural areas with scattered buildings, the temperature is relatively not as high as in urban centers; however, the temperature is generally within the range of 35-39°C, which is still higher than the surrounding area. Due to the large urban population, more emissions from automobiles and domestic combustion, and frequent industrial production, the temperature is relatively high.

(3) The temperature of farmland and cultivated land is relatively high. Another reason why areas above 35°C are mainly distributed along the Han River and Tangbai River is that these arable land, like towns, are mostly distributed on the banks of rivers with flat terrain, fertile soil and convenient irrigation. It can be seen from the thematic map that the temperature of most arable land is in the range of 35-39℃; of course, there are also large areas of farmland whose temperature is in the range of 30-35℃. Therefore, it can be considered that the temperature of the cultivated land is around 35°C. Its temperature is lower than that in cities and towns. It is due to the large and sparsely populated villages, and the population density is significantly smaller than that in urban areas, so it generates less heat; while its temperature is higher than that of forests and mountainous areas because of flat terrain and artificial farming to regulate soil temperature. .

(4) The temperature in forests and mountainous areas is obviously lower. In the temperature thematic map, there are three main low temperature gathering centers, namely the southwest of Xiangyang City, the north bank of the Han River in the west, and the east of the Han River in the middle. Among them, the most obvious and largest low-temperature gathering center-the low-temperature southwest of the city is a mountainous area, while the other two areas are mostly forests. It can be seen that the temperature in these low-temperature areas is below 30°C. Due to the shading and heat absorption of the trees in the forest, the surface temperature of the forest is significantly lower than that of other areas; for mountainous areas, due to the low air pressure and thin air, the atmospheric heat preservation effect is poor; at the same time, the surface of the mountains is far from the ground and cannot be Absorb the ground heat, making its surface temperature lower. At the same time, both have the characteristics of very low population density.

Welcome to pay attention to the public account: crazy learning GIS