Assessing the relationship between land surface temperature with vegetation and water area change in Arsanjan county, Iran

Land cover and soil moisture changes have a significant impact on land surface temperature (LST). Therefore, LST can be used to study land cover and desertification changes. Arsanjan County, which is located in the northeast of Fars province, has a relatively good forest and rangeland. Unfortunately, excessive harvesting of the groundwater resources and also reduced precipitation in this area caused to decrease water levels and dried up many wells in this area during recent years. So the area of the farmland and Bakhtegan Lake has decreased in this region during the last decades. However, so far, the condition of the LST and its relationship with land cover changes has not been assessed in Arsanjan County. In this study, spatial-temporal changes of LST and its relationship with vegetation and the water area of Bakhtegan Lake have been studied.

The eleven images related to Level-1 data of Landsat satellite was taken from 2003 to 2018. Since the vegetation situation in the study area is in the best vegetation and water area condition in April and May, so the images related to these months were selected to check the fluctuation of vegetation cover and water level of Bakhtegan Lake. The data pre-processing was performed in three sections: geometric, radiometric and atmospheric correction by ENVI software. The FLAASH algorithm, which is one of the best methods of atmospheric correction, was applied for atmospheric correction. In this study, NDVI was used to estimate the amount of vegetation. The Planck algorithm method was applied to calculate the LST. The change detection process was done using the index differencing method. To classify the LST map and the temporal-spatial changes, the LST difference map was normalized. Then, the normalized image was categorized using the standard deviation parameter in five temperature classes.

In the present study, 11 Landsat images were examined to investigate the spatial-temporal changes in land coverage and LST and the relationship between these two parameters from 2003 to 2018. The NDVI mean value was 0.25 in 2003, which decreased to 0.18 in 2018. On the other hand, the LST mean value had an upward trend as it increased from 29℃ in 2003 to 41.7℃ in 2018. The NDVI mean value was 0.66 in the farmland in 2003, however, its value reached to 0.33 in 2018. In contrast, LST mean value increased in the farmland from 20.9℃ in 2003 to 39.5.5℃ in 2018. Also, the LST mean value in the lake area increased from 20.1℃ in 2003 to 36.5 in 2018. Based on the results, the NDVI mean value in the rangeland and farmland decreased by 0.07 and 0.33, respectively, in 2018. However, due to the positive relationship between NDVI and LST in water-covered areas, the NDVI mean value increased by 0.39 in Bakhtegan Lake area in 2018. In contrast, the LST mean value in the rangeland, farmland and Bakhtegan Lake increased by 12.7℃, 18.6℃ and 16.4℃, respectively, in 2018 compared to 2003. The results indicated a negative relationship between NDVI and LST (R2= 0.862). The LST value decreases by increasing NDVI value in the vegetated area. In contrast, there was a positive correlation between NDVI and LST in salt-marshes and barren areas. According to the results, the highest negative correlation was obtained for the farmland, which was  -0.94. The reason for this high correlation can be related to the high density of vegetation cover in agricultural areas. The low negative correlation between NDVI and LST in the rangeland indicates the low vegetation density in rangeland and forest area. In order to study the area of decrease or increase of LST in the farmland, rangeland and water classes, the LST difference map was classified to five categories including very low temperature, low temperature, medium temperature, high temperature and very high temperature. According to the result of LST classification, the highest area was related to the moderate temperature class in all land covers, so that the highest area of this temperature class was associated with the rangeland by 86733 hectares. Since the vegetation density, especially in the farmland, had a significant decrease in 2018 compared to 2003, the area of high and very high-temperature classes increased in 2018, so that their area reached to 4625 ha and 7192 ha, respectively, in the farmland. Also, since the water area of the lake decreased in 2018 compared to 2003, the area of high and very high-temperature classes in these classes reached to 1824 ha and 3919 ha, respectively.

According to the results, the NDVI mean value in 2018 decreased in the farmland and rangeland and increased in the Bakhtegan Lake area. In contrast, the LST increased in the mentioned areas. The results of the LST classification showed that the highest amount of LST change is related to the moderate temperature class. Since the vegetation density, especially in the agricultural area, had a significant decrease in 2018 compared to 2003, the results showed that the area of high and very high temperatures had a higher increase than low and very low temperatures. Also, since the lake's water level decreased in 2018 compared to 2003, the area of high and very high temperatures in these classes increased. The findings show that there is a negative correlation between vegetation and land surface temperatures.