The three months between July and September are among the warmest the city-state has experienced over the last 36 years, with 2019 poised to break temperature records.

Singapore has not been spared the recent spate of record heat waves around the world.
If anything else, an examination of the city-state’s weather data from July to September 2019 suggests that Singapore baked under a long heat spell that was noticeably more intense than the long term average over the same period (ie, Q3 2019 data compared with Q3 data in previous years).
In 84 of the 92 days in Q3 2019, or 9 out of 10 days, the mean daily temperature exceeded the long term average of 27.92°C for July-September – the highest since 1983, by this benchmark.
Q3 2019 also had 81 days where the maximum daily temperature exceeded the long term average of 31.39°C for this period, matching the current record set in 1997 when Singapore and the world felt the brunt of the El Nino effects.
The Meteorological Service had earlier said that August 2019 was likely the driest and warmest August since records started in 1929, and that 2019 was poised to set new temperature records.
It is unclear if the Weather in Singapore has shifted to a new normal. But the trend lines certainly point in a worrying direction.
Data And Benchmarks
Comparisons of weather records are always tricky, even for a small city-state like Singapore. Much depends on the data you are using, as well as the benchmarks and timeframe for comparison.
For this post, I’ll be using a 36-year weather dataset that I had previously compiled from the Singapore Met Service’s website. So references here to "long term" refer to the period between 1983 to 2019 (September). The data was pre-processed in an earlier Jupyter notebook, which I have updated with July-Sept 2019 data.
To make sure that I compare like for like, I’ll primarily be comparing Q3 2019 data with with Q3 data from previous years. Results will differ if you compare across all quarters, which include the cooler months like December.
Data Exploration And Temperature Records
Running through the data, it quickly became obvious that Q3 2019 didn’t break any single-day temperature records.
For instance, the highest maximum daily temperature in the last 36 years is 36°C, recorded on a sweltering March 26 1998. But the highest maximum daily temperature recorded during Q3 2019 was "just" 34.3°C (on September 28), well shy of the record in this dataset.
Two days in May 1998 – May 21 and May 28 – jointly hold the record for the highest mean daily temperature of 30.9°C , or about 3.23°C above the average daily temperature for the period between 1983 and 2019. This long term record was hardly threatened by the highest average daily temperature for Q3 2019— 29.8 °C (recorded on August 12 and August 16).
While the data for Q3 2019 did not break any single-day records, a closer look showed that the mean temperatures for this period were all higher than the long term average.
For instance, the average maximum daily temperature for July-September 2019 was 32.59°C, 1.07°C higher than the long term average of 31.52°C between 1983 and 2019.
Likewise, the average daily temperatures and average minimum daily temperatures in Q3 2019 were both higher than the long term averages in these two categories by 1.39°C and 1.97°C respectively.
The trend becomes clearer when you compare apples with apples, that is, compare third-quarter temperature records with each other instead of the overall long term averages, which contain cooler months.
The average maximum daily temperature for Q3 2019 (32.59°C) was 1.2°C higher than the long term Q3 average of the same category – 31.29°C – between 1983 and 2019.
Likewise, the mean daily and average minimum daily temperatures for Q3 2019 were all higher than the long term averages Q3 records between 1983 and 2018.


This is best illustrated via a few charts.
Q3 2019 Vs Long Term Q3 Weather Data
The comparison is striking, to say the least. The average temperatures for Q3 2019 are markedly higher across all three categories, including the minimum daily temperature, meaning we are getting hotter days as well as warmer nights.

The Q3 2019 records also buck the long term trends, as represented by the grey-colored lines. In general, temperatures tend to trend lower as we move from July to September.
But strikingly, the maximum daily temperature in Q3 2019 spiked as the months progressed, showing a consistent rising trend on average from July through to September. Meanwhile, the average daily temperature at the end of September 2019 was almost on par with that at the beginning of July, a notable contrast with the long term trend showing a slight cooling by the end of the third quarter.
The minimum daily temperatures for Q3 2019 followed a shape that was more consistent with the long term trend, but it was on an order significantly higher than what we’ve experienced between 1983 and 2018, as the third subplot of the chart above shows.
Q3 2019 Vs Q3 2016 & Q3 1997
We can also do single-year comparisons. For the chart below, I had picked the data for Q3s of 2016 and 1997, as both years had long heat spells.
Singapore’s Met Service had said 2016 was the hottest year on record for the city-state. In 1997, Singapore, like much of the world then, suffered long periods of intensely hot and dry weather as a result of the El Nino.

At a glance, it is clear that temperatures in Q3 2019 – represented by the solid red line – were more intense than similar periods in 2016 and 1997.
2019 certainly looks to be on track to break annual temperature records in Singapore, as media reports have suggested earlier this year.
Joyplots, Or Stacked Density Plots
The popular joyplots offer another way to visualize how intense Q3 2019 had been compared to the same period over the last 36 years. I’ll plot just one version to compare mean daily temperatures.


The differences are again obvious from a glance: The distribution of average daily temperatures for Q3 2019 (chart on the right) shows a distinct rightward shift compared to the long term trends (chart on the left), signaling a period of higher temperatures.
Comparing Days Where Mean & Maximum Daily Temperatures Were Above Long Term Average
Another way to illustrate how hot Q3 2019 had been vis-a-vis Q3s in other years is to compare the number of days where the mercury had soared above the long term mean for Q3s in the dataset.
The two benchmarks I’ll use are:
- 31.39°C: The average maximum daily temperature recorded between July and September in the years between 1983 and 2018.
- 27.92°C: The average daily temperature recorded between July and September in the years between 1983 and 2018.

In Q3 2019, there were 81 days where the maximum daily temperature exceeded the long term average of 31.39°C for this period, matching the current record set in 1997.

In 9 out of 10 days in Q3 2019 (84 of 92 days to be exact), the mean daily temperature exceeded the long term average of 27.92°C for July-September. This is highest we’ve seen since 1997, by this benchmark. Results will clearly differ if you use a different benchmark for this particular comparison.
Is Q3 2019 The Warmest Q3, or Quarter Even, On Record?
Without the full set of data – Singapore’s written temperature records go back to 1929 – it is hard to say definitively. Much also depends on the benchmarks and timeframe you adopt for comparison.
I am not a meteorologist by training or an expert on Singapore’s weather, so I can’t say if the benchmarks and assumptions I have used here reflect the acceptable norm in the field.
But Q3 2019 is certainly one of the warmest three-month period we’ve seen in the last 36 years in Singapore, and potentially an indication of the more uncomfortable times to come.
The Jupyter notebook for this post is available here, alongside the processed dataset.
My earlier attempt at visualizing and analyzing Singapore’s changing weather patterns can be found [here](https://towardsdatascience.com/detecting-abnormal-weather-patterns-with-data-science-tools-5fae48586469) and here.
As always, if you spot any errors, ping me @
Twitter: @chinhon
LinkedIn: www.linkedin.com/in/chuachinhon