How Temperature Changes Across a City
Temperature can differ across a city, but a useful comparison starts with comparable observations—not with a map of hot and cold neighbourhoods. Choose two documented sites, match the time, instrument, height,

Temperature can differ across a city, but a useful comparison starts with comparable observations—not with a map of hot and cold neighbourhoods. Choose two documented sites, match the time, instrument, height, exposure and unit as closely as possible, then describe what the comparison can and cannot show.

Choose a controlled two-site comparison if you can match those conditions. Choose “not comparable yet” if one reading is from a sunny wall and the other is from an open grass site, or if the observations were made at different times.
Define the two sites
Name each site by its exposure, not just its district. “Grass courtyard, 1.5 m above ground, open to north” and “shaded balcony, 1.5 m above concrete” are meaningful labels. “Site A” and “Site B” are not.
The Met Office notes that buildings can warm temperature observations and that standard observing sites have exposure requirements. This does not make household readings invalid; it tells you what their number represents.
Match the comparison conditions
Use the same temperature unit and, ideally, the same model of instrument. If instruments differ, record that fact rather than treating a difference as caused by the city. Take readings close together in time. A fast-moving cloud shadow or a change in sunlight can make a fifteen-minute gap more important than the distance between sites.
Keep height and surface context in the note. Temperature near paving, a wall, a parked car, grass or a roof is not automatically wrong. It is a measurement of air at that exposure.
Make a worked comparison
At 18:00, Site A is an open grass strip and Site B is a shaded paved courtyard. Both sensors are 1.5 m high, use degrees Celsius, and were read within two minutes. Site A reads 18.1 °C; Site B reads 19.4 °C.
The responsible conclusion is: “Under these documented exposures at 18:00, Site B measured 1.3 °C warmer than Site A.” It is not: “The whole courtyard district is 1.3 °C warmer.” The first statement preserves the comparison; the second extends it beyond its evidence.
Repeat before explaining
Repeat the comparison on several days and retain conditions: cloud cover, recent rain, sunlight, wind, time, sensor placement and changes to the site. A repeated difference may justify a question about surfaces, shelter or local airflow. It does not establish a city climate, prove a cause, or replace an official observing network.
If readings reverse, keep the reversal. It may reflect changing sun, wind, cloud, timing or an instrument issue. Do not remove inconvenient readings to create a neat story.
Check the instruments before comparing places
Place both instruments together in the same sheltered, representative position for a short comparison before beginning the two-site exercise. This does not calibrate them to a laboratory standard, but it can reveal an obvious persistent offset. Record the result: “Sensors co-located for 20 minutes in shade; readings differed by 0.2 °C.” If the difference is larger or unstable, do not quietly correct one value; label the comparison as needing investigation.
Check batteries, unit settings, display resolution, and whether a sensor has enough time to adjust after moving. A sensor carried from an air-conditioned room or from direct sunlight may need time before it represents the new environment. The useful question is not “which one is right?” but “what conditions and instrument behaviour could make this comparison unfair?”
Build a small comparison table
Use a table with columns for date, local time, Site A, Site B, time gap, cloud cover, wind, surface, exposure, instrument and remarks. A filled row is more valuable than a blank template:
| Date and time | Site A | Site B | Context | Result |
|---|---|---|---|---|
| 16 May, 18:00 | 18.1 °C, grass/open | 19.4 °C, shaded paving | light wind; readings two minutes apart; same model | B measured 1.3 °C warmer; cause not determined |
The phrase “cause not determined” protects the record from a common leap. Paving, shelter, stored heat, instrument response and exact exposure may all contribute. A two-site comparison can show a difference under stated conditions; it does not isolate one physical cause unless the design and evidence are much stronger.
Compare like with like over time
Try morning, afternoon and evening observations on several days, but do not mix them into one average without retaining the time. A shaded courtyard may be cooler than grass at one hour and warmer later. Cloud cover and wind can change the balance again. The value of repetition is not that it guarantees a simple answer; it shows when the relationship is stable, variable or unresolved.
If you make a graph, label every site and unit, keep missing observations visible, and do not draw a smooth trend through gaps. A graph can summarize your own documented points. It cannot turn two sensors into a citywide temperature network.
Use the result carefully
This method is for learning how exposure affects a personal record. It is not a neighbourhood forecast, heat-risk service, property assessment or public warning. For health and safety decisions during heat or cold, use local official information.
Your completed journal line can be short: “18:00, Site A grass/open 18.1 °C; Site B shaded paving 19.4 °C; same sensor model, readings two minutes apart; light wind; cause not determined.” That is a successful city comparison because it reports conditions without pretending to explain more than the observations support.
Decide when a comparison is not usable
Do not force a result from mismatched records. A comparison is not ready when one sensor was in direct sun and the other was shaded without that difference being the subject of the exercise; when readings are separated by a changing weather period; when the unit is unknown; or when one site changed without documentation. Mark the row “not comparable” and explain why.
That outcome is productive. It tells you what to improve next time: synchronize clocks, leave the instruments in place, choose a safer common observation time, or document the exposure more carefully. It is better than subtracting two unrelated numbers and assigning the difference to an entire urban area.
Review the evidence at the end of the month
At the end of a month, group only observations that share a comparable time and setup. Ask: how many paired readings exist; how often was Site B warmer; how large was the observed difference; and which context notes accompany the largest gaps? Report counts before impressions. For example: “Eight comparable evening pairs were collected; Site B was warmer in six; the two largest differences occurred on clear, light-wind evenings.”
This is still a description of eight documented pairs, not a general statement about every clear evening or every paved courtyard. If you alter an instrument position, start a new comparison series. Keeping that break visible protects the journal from false continuity.
The finished comparison gives you a defensible local observation exercise: two sites, stated conditions, a measured difference, and a clear limit. It does not provide a city temperature forecast, a neighbourhood service, or an official heat assessment for public decisions. Its value is disciplined context, not a bigger claim about urban climate or public risk.
That distinction is the result readers should carry into every later comparison and every new site record.