Did the HK Observatory play with its Typhoon signaling? A quick look
Typhoon season was a big part of my upbringing in subtropical Asia. They are quite dangerous and occasionally deadly. However, as a kid I was quite excited whenever they came. If a particularly Typhoon is strong enough, the Observatory would put out signals 8, 9 or 10 (higher signals mean stronger Typhoons - out of 10), canceling school (and work for most industries). I would then stay home all day and play video games. Since I lived close to the sea, it was also just fascinating to stare at the choppy waves.
There are all sorts of urban myths about Typhoons. A popular one in Hong Kong is the âLiâs force fieldâ; named after Li Ka Shing, the wealthiest tycoon in Hong Kong. Li supposedly has the power to swat Typhoons away when the strong ones get too close, hence robbing everyone of a free day off and preserving the cityâs productivity.
The other myth is also about productivity, but just much darker. The rumor is that the Hong Kong Observatory deliberately refrains from busting out stronger signals during working hours so that everyone would go to work. This is quite a serious allegation, since it accuses the Observatory (hence government) of putting productivity ahead of the safety of citizens. Before looking at the data, I thought that this is probably bullshit. I wholeheartedly believe that the Observatory is a professional organization, and that the officials there do their best to be conservative, especially given how easily one could get hurt in a typhoon.
Using the Observatoryâs own data, however, a fairly shocking picture emerged. (Note that the Hong Kong Typhoon warning system comprises of signals 1, 3, 8, 9 and 10. School and work are usually cancelled for anything at or above signal 8. Signals 4, 5, 6, 7 are not used). First off, the incidence of signals 8 or above is dramatically lower during working hours in the last 7 years â see chart 1a. Effectively, we were 1.7x more likely to have signals 8 or above during non-working hours vs. working hours (normalized for number of hours).
Warning: for all you consultants, I did this on a Mac MS Office, which is rubbish, so please donât complain about my graph formatting
Perhaps we have been getting weaker typhoons compared to average? Figures 2 tells us that this is likely not the case. As a percentage of total time under typhoon, we are spending about the same amount of time under signals 8 or above, ~5-10%. If anything we are spending a bit more time in the stronger signals compared to the previous decade. Figure 3 also shows us that the distribution of typhoon strength is about average in the 2010s, with about 26% of typhoons eventually converting into signals 8 or above.
Perhaps the distribution of typhoon signals over the course of a day has always been this way? That also is not true. While there is some evidence that there was a similar trend in 2000s (less pronounced), this was clearly not the case in the 4 decades before this â see chart 4a. The ratio of incidence â chart 4b (same as chart 1b, just plotted over time) shows that the last 7 years were indeed quite unusual in how unlikely it was to get signals 8 or above during the work day.
I am no expert in typhoons, but I find the idea that Typhoons are more likely to come at night a bit bizarre. In fact, if you just look at the data for signals 1 and 3 (see figures 5a and 5b), you see that the same analysis above produces a fairly uniform distribution.
Very suspicious huh? I still donât believe that there is conclusive evidence that the Observatory has been manipulating typhoon signals for economic gains. However, there is certainly reason to look deeper into this.
The definitive proof would be to get historical weather data, apply rules for each signal (which are public), and backtest that against the actual signals hoisted by the observatory. That weather data seems pretty hard to get, so I didnât do this here. Once the data is in place, this analysis is actually probably really straightforward. Hong Kong Observatory â if you are somehow reading this, this would be a quick way to clear your name.
There are also other alternative explanations. For example, weather systems are notoriously hard to predict â so depending on how much the signaling process is based upon âexpectationsâ, models might introduce some noise into the ultimate signaling decision. However, I find it extremely hard to believe that the models somehow have a time bias and that we systematically expect less severe typhoons from 9-5pm. Itâs just too strange to believe.
The other explanation could be that the Observatory is not aware of its bias. Given that signaling is ultimately a bit of a judgment call, worries about false positives (hoisting signal 8, but then nothing happens -> Observatory looking like a fool) might have creeped into the decision makersâ subconscious and lead them to be more conservative. This is definitely plausible, depending on how the decision making is actually done.
The other explanation is that we are just getting loads of weird Typhoons that approach HK at night. But again, I would think that charts 5a and 5b disproves this.
The data only covers 6 years out of the possible 10 for the 2010s, letâs see how this plays out in the next few years.
Finally, random canto song about the wind â not really related at all but itâs awesome:Â
Raw data here for anyone interested:Â http://www.hko.gov.hk/wxinfo/climat/warndb/warndb1_e.shtml










