QUANTITATIVE REMOTE SENSING ANALYSIS OF THERMAL ENVIRONMENT CHANGES IN THE MAIN URBAN AREA OF Guilin BASED ON GEE

The dynamic change of urban thermal environment caused by the change of land use type has become one of the important problems of urban ecological environment protection. In Guilin city as research area, based on the Google Earth Engine (GEE), the random forest algorithm was used to classify the land use classification of Landsat remote sensing images in 2010, 2014 and 2018, and the mono-window algorithm was used to calculate the surface temperature. The surface vegetation was solved according to the NDVI pixel binary model. Coverage, and finally dynamic statistics and comparative analysis of land use, vegetation cover and surface temperature. The main results as follows. (1) From 2010 to 2018, the average temperature in the main urban area of Guilin is on the rise (increased by 1.29 °C), and the temperature zones in each class are converted from low temperature zone, lower temperature zone and medium temperature zone to higher temperature zone and high temperature zone. (2) Lower temperature zone and the low temperature zone is mainly distributed in vegetation and water body coverage areas, while the medium temperature zone, higher temperature zone and the high temperature zone are mainly distributed in construction land and unused land cover area. (3) High vegetation cover area in 2014-2018 (reduced by 31.34%) The main reason for the sharp decline is the substantial increase in the area of construction land (expansion 30.19%). (4) GEE-based random forest algorithm Land use classification had higher classification accuracy (more than 80% in all three periods). The results can provide scientific basis for improving urban thermal environment and scientific reference for the development strategy of Guilin city. * Corresponding author


INTRODUCTION
The urban thermal environment is an important parameter of the urban ecological environment (Xu, 2015;Fu et al, 2018;Meng et al, 2018). Satellite remote sensing can provide earth observation data in a short period of time, which is an important means to obtain surface temperature (Hou et al, 2018). The use of satellite remote sensing for dynamic analysis of urban thermal environment and its influencing factors is of great significance.
The current land use change caused by land use has brought about a series of ecological and environmental problems, such as the large reduction of vegetation coverage area (Zhang et al, 2018), the large number of rivers and lakes disappeared (Maimaitijiang et al, 2018), and the heat island effect continued to increase (Xie et al, 2017). In response to these problems, domestic and foreign scholars have carried out research on the impact of land use change on urban thermal environment The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLII-3/W10, 2020 International Conference on Geomatics in the Big Data Era (ICGBD), 15-17 November 2019, Guilin, Guangxi, China changes in recent years. MODIS and AVHRR surface temperature products are mostly used for surface temperature inversion, but their spatial resolution is low, which is only suitable for large-area research (Hou et al, 2010;Weng et al, 2015 (Hu et al, 2017).

GEE-based random forest land use classification
The random forest algorithm is a flexible and easy-to-use machine learning algorithm. Even if parameter tuning is not performed, better classification accuracy can be obtained (Breiman, 2001). The basic principle is: (1)  Each sampling process has 1/3 of the data not drawn, called extra-bag data, which can be used for internal error estimation and OOB error.
In the GEE, the random forest classifier is called, and the land use type is used as the target variable. The number of decision trees (ntree=500) is set by using the SRTM DEM data and NDVI and NDWI as the characteristic variables input during the training process, the default number of variables (mtry) is the square root of the total number of characteristic variables (Ghosh, 2014), and then the land use classification of the main urban area of Guilin.

GEE
In the formula, NDVI Soil is the NDVI value of the bare soil or no vegetation coverage area, and NDVI Veg is the NDVI value of the pixel completely covered by the vegetation, that is, the NDVI value of the pure vegetation pixel. The empirical value is selected as NDVI Soil =0.05, and NDVI Veg is 0.70 (the maximum pixel value when less than 0.70). When a pixel NDVI value is greater than 0.70, the PV value is 1; when the NDVI is less than 0.05, the PV value is 0.

Surface temperature
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLII-3/W10, 2020 International Conference on Geomatics in the Big Data Era (ICGBD), 15-17 November 2019, Guilin, Guangxi, China In this paper, the Mono-window algorithm improved by the traditional radiation equation is used to invert the Land Surface Temperature (LST) in the main city of Guilin (Wang et al, 2015;Zhou et al, 2010). The USGS points out There is calibration instability in the TIRS 11 band (Qin et al, 2001a), so the Landsat TM/ETM+ band 6 used in the original algorithm is improved to the 10th band of the Landsat 8 image, and the parameters a and b in the formula 2 are corrected. The calculation formula (Qin et al, 2001b) is as follows: In the formula, T is the actual surface temperature in K. a and b are -62.735657 and 0.434036, respectively (Qin et al, 2001b). C is atmospheric transmittance, T 10 is the pixel brightness temperature of the thermal infrared band 10 of Landsat 8, in K.
In winter (November to January), the atmospheric mean temperature (T a ) of the location of the main urban area of Guilin      (4) The random forest algorithm based on GEE cloud platform not only has a fast calculation and has better classification accuracy. The verification accuracy in 2010, 2014 and 2018 is higher than 80%, which is an effective land use classification. technology.
In summary, reduce the unused land area, increase vegetation coverage through afforestation, and thus affect Surface runoff to conserve water sources can effectively improve the gradual deterioration of the thermal environment in Guilin.