In forensic meteorology, we are often asked whether a given weather event was “foreseeable” – in other words, whether those impacted by or involved in the event should have or could have seen it coming.
Foreseeability generally has three components when considering weather events – climatology (what is “normal”), forecast (what is expected) and communication (whether those making decisions have access to crucial information).
In many ways, climatological expectation is about timeline. Over a long enough timeline, even extreme weather events can be expected to occur.
That may sound counterintuitive, so let’s take a familiar example. If you flip a coin 20 times, what are the odds that you’ll get a run of 15 tails in a row? It doesn’t take a statistician to tell you those odds are pretty low.
But what if you flip the coin 200 times? 200 thousand times? As the number of total coin flips grows so too do the odds of seeing a run of 15 tails in a row, for no other reason than that there are more opportunities for it to occur. While a run of 15 tails would be extremely unusual within a trial of only 20 flips, it would normal and expected within a trial of 200 thousand flips. In other words, if you flip a coin 200 thousand times, you should expect to see a run of 15 tails somewhere within that trial. An event that is unusual within a short trial is actually expected within a longer one.
The same is true of extreme weather events. Take the “100-year flood”. What a meteorologist or hydrologist means when they say a “100-year flood” is an event that has a 1-in-100 chance of occurring in any given year. Over a long period of time, we would therefore expect to see such a flood occur, on average, about once every hundred years.
What “100-year flood” doesn’t mean is that a given location is “due” for such an event if the last one occurred more than 100 years ago or that it is “safe” if the last one occurred less than 100 years ago. Every year, the odds are the same: 1-in-100.
The expectation of a given event (from a statistical, climatological perspective) does not change based on past weather. A “100-year flood” is just as likely to occur next year as it was last year.
Go back to the coin flip example: If you have a correctly balanced coin, each flip carries 1-in-2 odds of showing tails. Even if you flip 10 tails in a row, the 11th flip carries the exact same odds of coming up tails: 1-in-2. It may feel like you’re “due” for heads after 10 tails in a row, but that’s psychology speaking not statistics.
What does change after an extreme weather event is knowledge among the affected population. Once an extreme event occurs, people know 1) that it can happen, and 2) what the impacts are. So while the occurrence of one extreme event does not shift the climatological expectation for when a similar event will recur (e.g. suffering through a “100-year flood” this year doesn’t mean you’re safe next year), it can impact the foreseeability of a similar event. After all, people know that it can happen because they’ve already seen it happen.
The fact that a given weather event is within climatological expectations doesn’t necessarily make it foreseeable for the affected population. The weather-related impacts that population could reasonably have anticipated depend on a number of factors, including the weather forecast and impacts statements, as well as how that information was disseminated to the public and/or decision makers.
Weather forecasts, especially for major events, are generally quite good and have improved significantly over the past several decades. Anyone not living under a rock is unlikely to get surprised by a hurricane or a heat wave. Even the warning lead-time for tornadoes is up to an average of 13 minutes – plenty of time to take shelter if you get the warning. (We’ll return to that “if” in a moment.)
That said, sometimes meteorologists get it wrong. Small changes to atmospheric conditions can have major impacts on how a weather event evolves (that proverbial butterfly flapping its wings in Brazil). When there is significant uncertainty in a forecast, meteorologists try to express it. However, scientific uncertainty is a nuanced thing, and often the uncertainty statements are the first thing to get stripped for the headlines. So while meteorologists and decision makers with direct access to them may understand the forecast uncertainty and its implications, the general public may not.
Remember when Hurricane Irma was definitely going to wipe Miami off the map? Neither do I, because at no point during Irma’s evolution was an impact on the southeast FL coast a sure and immanent thing. But “Miami About to be Obliterated!” makes a far more clickable headline than “landfall near Miami looking possible, but storm could still swing further west or east”.
Which brings us to the final crucial piece in foreseeability: information access.
Even if a given weather event is within the climatological norms; even if the forecast is close to perfect, with uncertainty and potential impacts clearly and understandably communicated – even then a weather event can be “unforeseeable”.
If the people impacted by the weather event do not have access to or should not be reasonably expected to seek out forecast information, the event may not have been foreseeable. For example, every few years, hikers are injured or killed in the desert southwest when flash floods turn dusty arroyos into raging rivers.
That last question is arguably more difficult and is where forensic meteorology largely bows out of the discussion.
A forensic meteorologist can tell you whether a given event was climatologically unusual, whether the forecasts leading up to the event were accurate, and how that forecast information was disseminated.
But whether those who were impacted should have ultimately “seen it coming” is a more complex issue.
What if different forecasts disagreed? If a truly extreme event was forecasted, is it reasonable for people to think the forecast was exaggerated? What is reasonable ignorance or incredulity in this age of information-overload and often hyperbole? If someone has access to forecast information, should they be expected to actively seek it out? Should they be expected to act on it?
These are often the types of questions that arise in weather-related investigations and disputes. What starts out as a series of relatively straightforward meteorological questions ultimately winds into a complex web of human psychology and societal and cultural expectations.
The meteorological science is a crucial component of such investigations, but developing a rigorous and comprehensive understanding of the situation requires cross-disciplinary collaboration and communication. The weather is where the questions start, but not necessarily where they end.
If you are involved in a weather-related investigation and would like to discuss how a forensic meteorologist could support that work, please reach out to Blue Skies Meteorological Services for a free, no-obligation consultation.
Between 1963 and 2012, only 11% of tropical cyclone (hurricane and tropical storm) deaths were due directly to the wind. Yes, hurricane-force winds can be frighteningly destructive and frankly awe-inspiring, but when the numbers are tallied, water is undeniably the deadlier threat. Combined, storm surge, inland flooding due to heavy rains, and high surf are responsible for 88% of tropical cyclone deaths. The storm surge alone kills nearly half of all people who lose their lives during these events.
The reasons for this are simple but not necessarily obvious.
Especially in areas with modern building codes and construction, structures can withstand long duration strong winds. Even when structures are significantly damaged by debris or the wind itself, they continue to offer some shelter and protection from the wind. This is why people in the path of a tornado are advised not to jump in their car and drive away but rather to shelter in place in an interior room on the lowest floor. The roof can be ripped off a building and the exterior walls toppled, yet a tiny bathroom or closet often remains intact and offers a pocket of safety amidst the chaos.
Water is different. It rises inexorably, under doors and through windows. No room is safe. Entire neighborhoods are engulfed at once – there is no accessible safe haven once the water begins to rise. Storm surge water seethes with debris, from toxic waste to power lines to floating vehicles. And unlike the wind, which weakens quickly as a storm departs, storm surge and inland floodwaters can engulf a region for days to weeks. Those who don’t drown during the storm are often trapped without water or food until rescue personnel arrive, and when the impacted area is large and damage is extensive, rescue can be slow coming.
If you were watching news coverage as Hurricane Matthew marched menacingly up the east coast of Florida earlier this month, you probably noticed that meteorologists and local officials were focusing far more on the potential for unprecedented storm surge than on the storm’s maximum winds, despite Matthew being a Category 3 or 4 storm with winds up to 145 mph at times. Major Florida cities along the Atlantic coast and along the St. John’s River (which empties into the Atlantic Ocean just east of Jacksonville, FL) were threatened by a storm surge unseen for at least a century.
Matthew was a major hurricane with destructive winds, but it was the water, not the wind, that led governors, mayors, meteorologists, and emergency managers alike to issue impassioned pleas for those in the path of the storm to evacuate. Like hurricanes Sandy, Ike, and Katrina before it, Matthew testified to the fact wind tells only part of the story of a cyclone’s destructive power.
The rest of that story, and the far deadlier chapters, are told by water.
And starting this year, emergency managers, public officials, and the general public have another tool to evaluate the impact of that water on their communities. Developed by the National Weather Service and made operational for the 2016 hurricane season, Potential Storm Surge Flooding Maps are specific to each tropical cyclone and show a reasonable worst-case scenario for storm surge inundation at the neighborhood level. In other words, it shows the storm surge heights that a person should prepare for before a storm, given the uncertainties in the meteorological forecast.
The Potential Storm Surge Flooding Maps are based on the existing National Weather Service (NWS) Sea, Lake, and Overland Surges from Hurricanes (SLOSH) model. (Modeling involves a dizzying and inescapable array of acronyms.) The SLOSH model takes into account forecast uncertainty and, when run as an ensemble, provides an envelope of possible storm surge outcomes based on the current forecast.
Ensemble modeling involves running a model (or group of models) many different times, each time with small differences in the initial conditions or model assumptions. For the SLOSH model, these differences represent plausible futures for the track and intensity of the hurricane given current observations and historical forecast errors. The result is a set of possible storm surge scenarios, with each member of the set representing a slightly different hurricane track and intensity.
Once this set – or “ensemble” – is assembled, it is analyzed statistically to determine a reasonable worst-case storm surge scenario at each location, and the depth of that storm surge is displayed on the Potential Storm Surge Flooding Maps. In this instance, “reasonable worst-case” is defined as that storm surge depth which has a 1-in-10 chance of being exceeded at each location. (In other words, at any given location, there is a 90% chance that the actual storm surge will be less than or equal to what is displayed on the map, given the current forecast.)
The Potential Storm Surge Flooding Maps are a valuable visual tool to assess storm surge risk as a tropical cyclone approaches land. The SLOSH model that forms the basis of the maps takes into account the factors most critical in determining storm surge:
However, the potential storm surge flooding map does not take into account a number of factors that do not impact storm surge directly but that nonetheless greatly impact overall flooding and water damage. The maps do not account for:
Wave action can significantly increase the impact of storm surge along the immediate coast, overtopping seawalls and sandbags where the storm surge alone would not. The potential storm surge flooding maps show only the potential storm surge – they do not provide any information about expected wave height. In many cases, the height of waves riding atop the storm surge can exceed the height of the storm surge itself.
Information about expected wave height and wave action impacts can be found in hurricane-related text products from the National Weather Service, including the local area forecast discussions and hazards and impacts statements. Interests along the coast and in areas protected by levees should pay particular attention to the impacts of wave action and weigh those impacts in addition to the direct impacts of the storm surge when making hurricane preparations.
Freshwater and Inland Flooding
Flash flooding, areal flooding, and river flooding due to excessive rainfall are responsible for over one quarter of all tropical cyclone-related deaths and can impact areas far inland and outside of the storm surge risk area. Hurricanes and tropical storms can dump tremendous amounts of rain in a short period of time. Even weak tropical systems can produce devastating amounts of rain. For example, the remnants of Tropical Storm Amelia in 1978 flooded central Texas with four feet of rain. Interests in low-lying areas prone to flooding and along creeks and rivers expected to be impacted by a tropical system should pay careful attention to rainfall forecasts and expected impacts.
Forecasting the protection offered by levees during a tropical cyclone is complex and difficult, as Hurricane Katrina tragically demonstrated in 2005. Wave action, the depth and speed of the storm surge, and the strength and construction of the levees all influence the amount of protection they provide. Interests living in leveed areas need to remain especially vigilant as a tropical cyclone approaches, and consider all relevant risks, not just those posed by the storm surge.
The potential storm surge flooding maps issued during Hurricane Matthew’s approach of Florida’s east coast offered a clear and dire assessment of the risk for unprecedentedly severe storm surge inundation.
The maps for Hurricane Matthew were so dire because the worst-case scenario was squarely within the cone of uncertainty. It wasn’t just possible. It was likely that locations along the east coast and the St. John’s River would be under several feet or more of water if the hurricane continued to hug the coast or made landfall in northeast Florida or southeast Georgia.
More than just the beaches were under threat. The potential storm surge flooding maps showed areas along the St. Johns River — inland regions that don’t typically think of themselves as vulnerable to storm surge –- under up to 6 feet of water as Matthew pushed a wall of water at the coast and up the mouth of the St. Johns. Although the worst-case scenario fortunately did not materialize, reports were still received of the St. Johns River flowing backwards on the morning of Saturday, October 8th, as storm surge and hurricane-force onshore winds pushed the sea inland.
“Fortunately” is the correct word in this situation, because it was nothing more or less than luck that prevented a truly worst-case scenario from actualizing. Had Matthew drifted 20 miles farther the west, the Atlantic coast of Florida, and inland river cities like Jacksonville and Palatka would have experienced truly devastating flooding.
The National Weather Service’s new potential storm surge flooding maps provide a graphical, easy-to-understand, quantitative assessment of storm surge risk along the U.S. Gulf and Atlantic coasts. By displaying a reasonable worst-case scenario at each location, they show both local officials and the general public the storm surge depth for which they should prepare themselves, enabling well-informed preparation and evacuation decisions.
However, these maps are not intended to offer a complete picture of water-related risks associated with tropical cyclones. They do not take into account wave action (i.e. waves riding atop the storm surge), rainfall-induced flooding, or potential levee failures. These impacts must be considered separately when assessing the impact of a hurricane or tropical storm in a given area.
And, as always, the output of any model is only as good as its input.
Ultimately, the accuracy of the potential storm surge flooding maps depends on the accuracy of the hurricane track and intensity forecast. Those forecasts – and the storm surge inundation maps derived from them – are continually refined as new data from in-situ and remote sensing platforms (like hurricane hunter aircraft measurements and satellite imagery, respectively) as well as new model guidance become available.
After several relatively quiet years in the equatorial Pacific Ocean, El Niño may be on its way back.
A new research study published in the Proceedings of the National Academy of Sciences (PNAS, February 2014) utilized a novel, long-range statistical approach to El Niño forecasting and found a 75% likelihood that El Niño conditions will begin to present by the end of 2014.
El Niño is the warm phase of a larger ocean-atmosphere cycle called the El Niño Southern Oscillation (ENSO). During an El Niño event, the waters of the eastern equatorial Pacific off the coast of Central and South America become anomalously warm. During the opposite phase of the cycle, La Niña, those same waters become anomalously cold (see figure at right, credit: NASA).
This fluctuation in water temperature may seem like a relatively localized phenomenon, but because the ocean and atmosphere are coupled (interconnected) and circulate the entire globe, an increase in water temperatures off the coast of Peru is not only devastating to the local fishing industry, but is also the most important driver of natural interannual climate variability across the entire planet.
Globally, El Niño conditions result in a major shift in atmospheric circulations and, consequently, weather patterns (see figure below right, credit: NOAA), as well as an increase in globally averaged temperatures.
In the northern hemisphere, El Niño conditions typically result in
Conventional El Niño forecasting techniques rely on dynamical and statistical climate models that analyze observations of sea surface temperatures (SSTs) and wind patterns. Although this forecast method can be quite accurate when making predictions a few months out, its skill is rather limited at longer-range forecasting. Accurate long-range forecasting is critical, however, to preparing for and mitigating the economic effects of El Niño events. For example, in the agricultural sector, farmers need to be able to plan which crops to plant based on expected weather conditions (e.g. hotter than normal, wetter than normal, drier than normal, etc) to reduce the likelihood of crop failure.
The study published by Ludescher, et al. this month claims to have developed a forecasting technique that can accurately predict ENSO fluctuations up to a year in advance by relying solely on statistical correlations between air temperatures across the Pacific region and upcoming changes to equatorial Pacific SSTs (i.e. upcoming El Niño or La Niña events). Although the study’s authors tout its long-range predictive ability, it cannot currently predict the magnitude (severity) of those upcoming events.
The study’s authors say that their technique accurately predicted the absence of El Niño during 2012 and 2013, but because the forecasting methodology is so new, it has yet to be tested in a prediction of non-neutral ENSO conditions. Many atmospheric scientists not involved with the study remain skeptical of the skill of the new technique for that reason as well as for the fact that the study does not propose an explanation as to why the statistical correlation should work. In other words, the study does not advance scientists’ understanding of the physical mechanisms that drive the ENSO cycle.
If the new statistical forecasting technique proves successful, though, it may alleviate a problem that has been plaguing conventional ENSO forecasters for the last couple of years and that may now be negatively affecting the skill of seasonal climate (e.g. ENSO) forecasts.
The National Oceanic and Atmospheric Administration (NOAA) maintains a network of moored buoys in the tropical Pacific Ocean to monitor real-time ocean temperatures for input into the climate models used to forecast El Niño and La Niña. Since budget cuts in 2012 forced NOAA to reduce its maintenance schedule of the buoys, however, over half of them have failed. With less detail about ocean temperatures in this critical location being provided by the thinning buoy network, forecast models may suffer a loss of accuracy.
Only time will tell. The dynamical and statistical climate models used to provide conventional ENSO forecasts are also beginning to predict an increased likelihood of El Niño conditions beginning in late 2014. It’s still too early for significant confidence, but those readers who are involved in weather-sensitive industries should monitor the situation closely and consider planning early for possible El Niño conditions beginning in late fall 2014.