As 2023 Atlantic and East Pacific hurricane seasons have come to a close, it’s an opportunity to take a look back at the season and how it unfolded. It has been a year to remember with multiple record-breaking events –Hurricane Otis standing out as the most unexpected of them all.
On 23rd of October in the Eastern Pacific basin Hurricane Otis took everyone by surprise. Overnight and without any warning from weather forecasting models, what was forecasted as a mild hurricane turned into a nightmare scenario. I went to bed expecting a “wind and some rain” in Acapulco Mexico to waking up to “Oh my god what is going on” the next morning as I was choking on my morning coffee checking daily National Oceanic and Atmospheric Administration (NOAA) updates.
This was a prime example of tropical rapid intensification – when a storm grows explosively and is likely to reach Category 3 -5 strength. These events are not unusual, and we tend to get a couple of them a year. What was so surprising about Otis was the speed at which it happened and that it was entirely missed by all forecasts. Let’s unpick…
What is rapid intensification (RI)?
Rapid intensification is defined as wind strengthening of at least 30 knots in 24 hours, the equivalent strengthening of a Category 1 cyclone (CAT) into CAT3 in under 24 hours. Otis’s winds for example, increased by 95 knots in 24 hours – that is more than three times faster than the threshold of rapid intensification, making it the second strongest rate after Hurricane Patricia in 2015. Notably in the Atlantic, we have witnessed Hurricane Lee – a monster Category 5 hurricane as winds intensified by 75 knots in 24 hours but luckily managed to stay clear of US coast.
These rapid intensification events have a higher potential to cause large losses for the most part because they are more likely to be CAT3+ by the time they make landfall. Also damages tend to creep higher if no preparation is done prior a hurricane event, Hurricane Otis gave no chance for that.
Climate change and processes
The mechanism by which a hurricane starts going through the process of RI is complicated, but in broader (and over-simplified) terms the process is very much dependent on underlying sea surface temperatures as well as moisture availability. Fastest tropical cyclone intensification rates often occur in areas of unusually warm upper ocean as it serves as a heat engine that can strengthen the storm quickly. With climate change, ocean temperatures are already significantly warmer than during pre-industrial era and 2023 ocean temperatures already off the charts, there is expectation for RI events to occur more regularly. The reality is much more complicated and RI is not that well studied by the scientific community.
It is worth noting that there is an early indication that the rates at which RI occurs are increasing – i.e. there is more explosiveness in the process. But any of these changes are very difficult to unpick because hurricane data is noisy, and the number of RI events is not sufficiently dense to enable robust statistics of trends, or in-depth understanding of the process itself.
And of course, no climate change should ever be mentioned without highlighting the role of natural variability of the climate. These are natural “swings” in the climate that can mask a climate change trend or disguise as a trend when there is none. Tropical storm activity is known to be heavily modulated by these swings (i.e. Atlantic Multidecadal Oscillation (AMO), Pacific Decadal Oscillation (PDO), El Nino-Southern Oscillation (ENSO) etc.) that introduce rhythm to temperature patterns of the oceans and therefore tropical cyclone activity. Scientists are aware of these, but it doesn’t make the task of detangling climate change signal from “natural swings” any easier.
Trends and forecasts
Research on this topic is gaining momentum. Most of the scientific studies look at RI events in context or overall activity, rather than landfalling events. Trends whilst robust, need to be aggregated across basins for reliable statistics to emerge. Below is a figure visualising occurrence at different RI thresholds aggregated across North Atlantic and East Pacific basins (the West coast of the US):
The trend is clear even the shorter period of where the technology has better capability of accurately recording hurricane statistics and is supported by existing scientific literature. It should be noted that basin-wide trends do not translate to trends in that of landfalls or losses.
More surprisingly strong hurricanes in store
It is rare for all forecasts to miss a hurricane like Otis. Having said that, forecasts have always struggled with accurately predicting the aspect of RI. Guidance models and National Hurricane centre (NHC) have exhibited more skill in the last 5 years, but intensity forecast errors for RI events are still approximately 2-3 times larger than non-RI events.
This is due to forecasting models being “statistical-dynamical” in nature – meaning that despite having a backbone of a proper numerical weather model, some weight is given to statistics and parametrizations (i.e. internal model approximations). It’s done because numerical model is not sufficient to forecast RI on its own and needs to rely on this additional information.
In the case of Hurricane Otis, all forecasting agencies had missed to forecast the RI. There was a warm patch of ocean right along the path of the hurricane, which while this is required for RI to occur, it is not always an indicator that is happening. This is something that forecasting models take into account but even with this information were still pointing at dissipation of the storm.
There is one crucial factor that serves as an indication whether rapid intensification is going to occur – the inner “anatomy” of the storm. When thunderstorms are perfectly symmetrical around the eye it causes for heat to concentrate and pressure in the eye of the storm to drop pushing the storm to go through RI. This information is often obscured due to thick clouds that prevent satellites from observing fully the structure – forecasting models didn’t take this aspect into consideration.
Hurricane Otis and it’s RI came as a surprise because this was only “discovered” when aircraft were able to fly inside the storm and take direct observations (known as reconnaissance data) – I learned about this from a blog on Science magazine. This seems to be the reason why all models missed to forecast this event.
It has been demonstrated that more frequent reconnaissance data gathering usually leads to better forecast skill, yet the frequency of how this data is gathered varies hugely across basins. The rules are set by the monitoring agency – lead by NOAA in the Atlantic and Eastern Pacific with help of US military. As I was going down the rabbit hole, it didn’t take long to find out that operational instructions from NOAA lay out explicitly the frequency of reconnaissance missions in the Atlantic which get more frequent closer it gets to the US, but much lesser detail is given for the Eastern Pacific. In existing published papers on reconnaissance, most of them focus on the Atlantic basin because the data is very sparce in Eastern Pacific. This raises questions on whether Hurricane Otis would have been detected had it originated in the Atlantic basin. In West Pacific and Central Pacific basins is even more complicated as typhoons are monitored by local agencies and governments, i.e., JMA (Japanese Meteorological Agency), PEGASA (same but for Philippines) etc, all with different rules around reconnaissance but generally less frequent.
Lesson learnt from Otis is whilst the availability of data and forecast skill is improving, we can still observe very large errors in the forecasts. Unfortunately, there errors are larger for storms that undergo RI – those are also the storms that are more likely to become destructive.
