Multispectral vs. thermal-only imagery: what actually tells you which block needs shade

If you've asked a satellite vendor for "heat imagery" and gotten back a single thermal band, you've got half of what you need. The other half is what's underneath that heat signal, and that's where multispectral imagery earns its keep.
What thermal-only imagery actually shows
A thermal band measures land surface temperature at the moment the satellite passed overhead. It's a reading of the surface itself: how hot the asphalt, the roof, or the bare soil got right then. It carries no information about what produced it.
A hot pixel could be an empty asphalt lot, a flat commercial roof, or a block of three-story walk-ups where forty renters have no air conditioning. The thermal band reads the same regardless. You get a temperature grid, and turning that grid into a reason to act on a particular square of it takes another dataset.
That gap matters more at block resolution than it does at a city scale. A citywide heat map is useful for a press release. A cooling-centre siting decision or a tree-planting route needs to know which specific blocks are both hot and full of people who can't escape the heat, and a single band can't answer the second half of that question.
Why the extra bands matter for block-level targeting
Multispectral imagery adds visible and near-infrared bands to the thermal one. Those bands let you classify land cover: impervious surface, tree canopy, bare soil, building footprint. Cross that classification against the thermal read and you can start separating "hot because it's an industrial lot with no residents" from "hot because it's dense multifamily housing with zero canopy." Add demographic density, age 65+, enrollment in assistance programs, population per block, and the thermal grid turns into a ranked list instead of a color gradient.
Resolution plays into this too. At 10-30 m ground sample distance, a single pixel can straddle a parking lot, a sidewalk, and part of a roofline. Land-cover classification sorts that pixel to a surface type, and the surface type tells you whether tree cover or a cool roof would do anything.
Seasonal cadence adds another layer. A summer pass shows the heat island at full strength. Comparing it to a spring pass over the same blocks shows you where canopy is thin enough that a block heats up fast once the season turns, a signal that's useful before the hottest months rather than after a heat advisory has already come and gone.
What this looks like for a planning department
In practice, working from a thermal-only dataset means a second round of work: pulling building footprints, parcel records, and census tables separately, then joining them to the thermal grid by hand in GIS. It's doable, and plenty of departments have done it for years. It's also slow, and it's easy to miss blocks where the heat signal and the vulnerability data don't line up the way you'd expect from either dataset alone.
Overlaying surface temperature, building type, and demographic density in a single pass skips that join step and hands you a block list instead of three spreadsheets to reconcile. That's the whole premise behind Heat Exposure Mapping: point it at a city and get back a ranked list of the blocks that run hottest and house the most heat-vulnerable residents, block by block rather than as a single citywide number.
If your next budget cycle needs a shortlist of blocks instead of a heat map to frame on the wall, that's worth a look.