Land surface temperature, or LST, measures how hot the ground itself is, which is different from the air temperature reported by weather stations. It reveals urban heat islands, where concrete and asphalt run several degrees hotter than nearby parks, and it helps farmers spot water-stressed crops, firefighters assess fire risk, and planners decide where tree planting will help most. Landsat 8 and 9 carry a thermal infrared sensor that records this information at 100-metre resolution, resampled to 30 metres, every 16 days, and the archive is free.

The raw thermal band stores digital numbers, not temperatures, so the work is a chain of conversions: digital numbers to top-of-atmosphere radiance, radiance to brightness temperature using Planck's law constants from the metadata, and brightness temperature to land surface temperature by correcting for emissivity, which varies between vegetation, soil, water, and concrete. Each step uses values from the scene's metadata file, so keep that file beside your imagery throughout the workflow.

Start by downloading a Landsat Collection 2 Level-2 scene, which already includes a surface-temperature product for many users, or a Level-1 scene if you want to practise the full conversion. In QGIS, use the Raster Calculator for each conversion step, entering the radiance and reflectance rescaling factors from the metadata, and the thermal constants K1 and K2 for the brightness-temperature formula. Then estimate emissivity from NDVI, using the standard method where dense vegetation is assigned high emissivity and bare surfaces lower values, and apply the emissivity correction to reach land surface temperature in Celsius.

Style the result with a colour ramp from cool blues through yellows to hot reds, and compare it with a land-cover map of the same area. The pattern usually tells its own story: rivers and forests cool, industrial estates and dense housing hot, and new housing developments visibly warmer than the fields they replaced. Export the map as a print layout with a scale bar, north arrow, and legend, because an LST map without context is easy to misread.

Validate your temperatures against common sense and, where possible, ground data: water bodies should sit near the day's air temperature, dense vegetation a few degrees above it, and sun-baked bare soil well above it. If your whole scene reads impossibly hot or cold, the usual culprit is a missed unit conversion, such as forgetting to subtract 273.15 to convert Kelvin to Celsius, or using the wrong thermal band. Clouds are another trap, since cloud tops read extremely cold and will distort any statistics unless masked out first.

LST maps are most persuasive when they compare two dates or two neighbourhoods rather than standing alone, so pair your map with a second scene from a different season or a before-and-after comparison of a development site. That comparison turns a technical exercise into evidence a council, NGO, or community group can actually use, which is the point of learning the method in the first place.

One extension worth trying is to pair your temperature map with a vegetation map of the same date and plot temperature against NDVI for a sample of points. The relationship is usually strongly negative, with greener areas cooler, and the scatter of points around that trend highlights places where vegetation is stressed or where surfaces such as dark roofs break the pattern. That simple analysis moves your work from description to explanation, and it gives planners a quantitative argument for tree planting: each increment of greenery corresponds to a measurable cooling effect. Present the finding with a map and a chart together, since decision-makers respond to the combination of a visual pattern and a number far more than to either alone.