At the beginning of June we released reask’s inaugural tropical cyclone genesis forecast for the northern hemisphere. Of course we have spent a lot of time back-testing our model, but nothing beats real world live testing and this is our first attempt at it. The complete forecast for all northern hemisphere basins is available via our web app here, however for this post we will focus on the North Atlantic only.
Let’s have a look at the big picture first:
The filled circles are an indication of the average activity in the basin since 1970 while the box and violin plots indicate confidence levels from our model. The view is for the 2018 season to experience an above average Cat0 (named storms, left of the plot) activity while the trend for the stronger storms (Cat1 / hurricanes and Cat3 / major hurricanes) is closer to the long-term mean (middle and right of the plot).
A more detailed picture is provided by the following figure where the probability of every outcome is explicitly quantified:
Here again we see some confidence from the model that 2018 could be an above average season in terms of Cat0 activity (70% chance of seeing 13 or more named storms) and a 67% probability of seeing 11-18 Cat0 storms forming (high confidence interval). The view for more intense Cat1 and Cat3 storms is much closer to the long-term average risk with a 65% chance of 4-8 hurricanes (Cat1) and a 59% chance of 2-4 major hurricanes (Cat3).
How we compare to our peers
The numbers above were submitted to the seasonalhurricanepredictions.org project organized by Phil Klotzbach from Colorado State University (CSU), Louis-Philippe Caron from the Barcelona Supercomputing Center (BSC) and XL Catlin. These were also quoted by RMS as part of their annual outlook.
Out of the three levels of intensity reported it is on the Cat0 forecast that reask seems to differ the most from its peers:
Although we are not the only organization with a view towards a season with above average Cat0 activity this is certainly not the most supported opinion among the 24 forecasts issued to date. The main reason for the disagreement has been pretty well documented in recent weeks and it comes down to some very abnormally cool May sea surface temperatures (SST) in a portion of the North East Atlantic called the Main Development Region (MDR, see white box below). Since the MDR is a hotbed for hurricane formation during the peak of the season (August-October) it is widely accepted that hurricane activity is strongly linked to how warm the SSTs are in the region.
The figure above shows the standardized SST anomaly (i.e. mean-centered and normalized by the standard deviation over the 1970-2017 period) in May with the white box highlighting the MDR. With most of the MDR in the blue end of the scale the expectation is that this season will lack the necessary fuel to sustain a very high hurricane activity. It is also worth noting that this trend towards MDR SSTs being cooler than the long-term average has carried on through the month of June (after our forecast was issued).
In the Figure below, we have plotted the number of Cat0 storms as a function of the average standardized May anomaly in the MDR. Both the values in May 2017 and May 2018 are shown (vertical bars) to highlight how much cooler the situation is this year.
So why would our model suggest an above average Cat0 activity? The short answer is: because it looks at more than just the mean May SST anomaly in the MDR, and most of our other predictors do not exhibit a strong signal towards a below average season. In an earlier post we discussed one of these predictors (“Feature A”) which was based on precipitation anomaly patterns in the March-May period. As a refresher Feature A was shown to be positively correlated to June-November Cat0 activity as well as to positive anomalies of SSTs, negative anomalies of zonal shear and negative anomalies of sea level pressure in August – October in the North Atlantic (all factors favorable to hurricane activity). Unlike the MDR May SST anomalies, Feature A is still showing similar levels as it did in 2017 and hints at a strong season in terms of Cat0 activity (see figure below).
Along with Feature A our model accounts for a wide range of other similar predictors characterizing the state of the climate at the end of May in terms of soil moisture, zonal and meridional winds at lower and upper levels as well as SSTs. The top 9 Features influencing the North Atlantic forecast are shown below, with the white vertical lines showing the 2018 levels and the dark grey ones the 2017 levels.
Beyond the model
Even if we forget for a moment the features driving our model, this 2018 season is shaping out to be a very interesting one from a forecasting point of view. As already mentioned current conditions at the end of May (and throughout June) are indicative of a cold MDR which in turns suggests lower levels of hurricane activity – but a few big question marks exist as to what the next couple of months might bring:
Will the MDR warm up in time for the peak August-September season?
As highlighted by RMS in their outlook “NOAA expects temperatures in the main development region to recover to near-average as the summer progresses”. Yet even if they do not, and the MDR remains at current levels, the biggest impact is likely to be a lack of long lived storms: the ones forming to the west of Africa and maturing over warm waters into very powerful systems. But what of Cat0 storms? There would certainly still be some potential for short lived weaker systems in the MDR and perhaps for the activity closer to the US coast to compensate a potential inactive MDR.
Will an el Niño event form during the peak of the season?
Most forecasts hint at the potential for the formation of an el Niño event during the winter but the chance of an earlier formation is very real. Should the formation occur during the peak hurricane season the increased level of shear in the basin would contribute to a lower activity… but then again we all remember what happened with ENSO forecasts last year.
Are we seeing the beginning of a new cold phase for the Atlantic Multi-decadal Oscillation (AMO)?
Expert opinions differ on the topic and given the lack of records about the climate conditions during a phase shift (the AMO is understood to have ~30 years period) there is probably no way to answer this at the moment. However there is convincing evidence that once we do enter the next cold AMO phase we should see a strong reduction in hurricane activity in the north Atlantic… the only question is when.
Based on our global TC genesis model simulations we expect a 2018 North Atlantic season with above average Cat0 (named storms) activity but only average levels of Cat1 (hurricanes) and Cat3 (major hurricanes) formation. This view accounts for a range of climate predictors as observed at the end of May and although the current cold water anomalies in the basin are taken into account these are balanced by other predictors still indicative of higher levels of hurricane genesis.
Find out what our predictions for the other 3 northern hemisphere basins look like and how the model performed in back testing in the Tools section of our website.