In my view that will not be feasible only with Landsat data, as the spatial resolution (100 m) will be a limiting factor here. However if there is a relatively bigger parcel of land where cool/reflective roofs have been installed, then Landsat data analysis may give reliable results.

Dear DR Richa Sharma

Thank you for your answer.

I derived LST of Tabriz city (northewest of iran) using Landsat 5 and 8 thermal bands. and I saw temperature of impervious surfaces area, specially in High density part of city, during 17 years significantly reduced. we think reflective roof has a great impact on this,because many of roofs materials, specially in High density part of city,from Asphalt or something like that to materials have relatively higher albedo. I attached LST and classification maps to this message.

I agree with Dr. Richa's comments.

In addition to that, I think one needs to be careful while interpreting the LST results derived from Landsat 8 thermal bands. It is still a question for me whether the results of Landsat 8 based LST are reliable enough for UHI applications. At the time of this writing, the 'Stray Light' issue that has plagued Landsat 8 Band 10 & 11 since 2013 has still not been completely eliminated. Therefore, I think it is critical to first implement corrections to L8 LST before comparing/contrasting it with LST derived from Landsat 5 TM or 7 ETM+

Amir, do you have any in-situ ground surface temperature measurements or any other sources to verify the reliability of your Landsat 8 LST results?

https://landsat.usgs.gov/how-are-thermal-infrared-sensor-tirs-thermal-bands-aboard-landsat-8-used

Dear James Varghese

thank you for your attention. i used Landsat 8 band 10 and Jimenez-munoz single channel method to derive LST. I think NASA suggests to use band 10 Landsat 8, because band 11 has some problems.

In addition to crucial issues flagged by James Varghese, another simple yet important consideration is that LST values will be greatly influenced by weather conditions prevailing on the day data is captured. One way could be to normalise the LST values and then compare them across the years (just a preliminary analysis). However, as pointed out in previous responses ground validation should be done to support results.