Recently, we received enquiries regarding differences in maximum and minimum temperatures between our weather data and the monthly reports published by Bureau of Meteorology (BoM). In some cases, the differences reached up to 2–3°C for both maximum and minimum temperatures.
The main reasons for these differences are related to:
- Different definitions of a “day”
- Different temperature measurement and processing methods
BoM defines a climatological day differently from most building simulation software. Historically, weather observations in Australia were manually recorded at 9:00 am, particularly for max-min temperature and precipitation measurements, and this convention continues today.
According to BoM, the daily maximum temperature is determined over the 24-hour period from 9:00 am to 9:00 am the following day, while the daily minimum temperature is determined over the 24 hours leading up to 9:00 am on the day indicated.
However, simulation software such as EnergyPlus, WUFI, and IES VE uses calendar-day data (typically 1:00–24:00) when calculating daily maximum and minimum temperatures and they simply select the highest and lowest hourly instantaneous temperatures. As a result, the same “day” can contain different temperature extremes depending on the time window used.
The chart below demonstrates how minimum temperature values can vary depending on the definition of the daily time window.
For example, for 2 January 2026, the BoM hourly minimum temperature is derived from observations recorded near 10:00 am on 1 January (17.4°C), whereas the simulation-based daily minimum temperature comes from 11:00 pm on 2 January (17.8°C). The minimum temperature in the BoM monthly report for 2 January is also 17.3°C which probably happened near 10:00 am on 1 January. This is an uncommon example because the minimum temperature usually occurs around dawn (1 January had the lowest maximum temperature on the month).
Historically, the BoM had no record of when the max and min temperatures occurred but they do now because of the way they are recorded in an Automatic Weather Station (AWS). Because we only use the max-min data to QA (and occasional gap filling) of the 24 hourly values, we have not accessed this time data because it is not useful to us.
For maximum temperature (Chart below zoomed at 2 January 12:00), both BoM daily maximum temperature and the simulation-based daily maximum are 25.2°C. However, the maximum temperature in the BoM monthly report for 2 January is 26.0°C which probably happened near 12:00 pm on 2 January.
Another source of difference relates to how temperatures are measured and processed. According to the Bureau of Meteorology (BoM) documentation on air temperature measurements, maximum and minimum temperatures at AWS sites are derived from 1-second observations recorded by the Automatic Weather Station (AWS). Therefore, BoM daily maximum and minimum temperatures are based on high-frequency point measurements rather than hourly observation data.
In contrast, our EPW and ACDB weather files are based on hourly observation temperatures. Therefore, short-duration temperature peaks or dips captured by AWS point measurements may not appear in hourly observation datasets. This means:
- Daily maximum temperatures in BoM reports can be slightly higher than hourly observation simulation data
- Daily minimum temperatures in BoM reports can be slightly lower than hourly observation simulation data
This difference is especially noticeable during rapidly changing weather conditions, where temperature extremes may occur briefly between hourly timestamps.
Although our hourly maximum and minimum temperatures do not always align with the BoM monthly report due to differing time windows, we use the BoM daily extremes as reference values in our QA process before 2024. When estimating missing values, our algorithm refers to these BoM-derived extremes to guide the estimation and ensure that the estimated temperatures remain realistic and within reasonable bounds. However, from 2025 onward, datasets from BoM contain very few missing observations and remaining gaps are supplemented using our internal quality-control and estimation algorithms.