Synecdoche ain’t okey dokey

Someone once said that history is just one damned¹ thing after another. It’s not: it’s just one damned thing happening everywhere all at once. However that’s rather a lot to take in, so we simplify. I did it just now. The original quote is “Some historians hold that history is just one damned thing after another” which is quite a different thing and seems to be a mutation of “life is one damn thing after another²” with a raised eyebrow suggesting that’s not quite how it is and anyone who thinks so is, well…

One simplification is to break history down into events, those atomic chunks of history restricted in geographical scope and with a definite start and end. But, like any living thing, history cannot long survive its dissection. A bladder is indubitably a bladder, but it does not serve its purpose when separated from the rest of the beast. Indeed, absent the rest its purpose though vital would be inscrutable.

We tend to do this with our weather, speaking of extreme events, storms, El Nino, and other discrete things that we further imbue with the power of agency. While it can seem intuitive, intuition can often lead us adrift. A rain storm, for example, might start and stop and, by careful attention to where each raindrop fell and when, we could precisely delineate its limits in time and space. Too much rain within those limits might lead to flooding. Put that way, one might be tempted to say that the rainstorm caused the flood and that the rain itself was caused by a particular instability in the atmosphere or some unusual combination of wind direction and local topography. In a sense the rain does cause the flood – no rain, no flood – but flooding also depends on other factors such as drainage, the condition of the soil³, specifics of the locale, timing, and, if we take a step further into the impacts of the flooding, includes things like flood defences, early warning systems, insurance, building codes, emergency management, and so on. Obviously many of these factors extend well beyond the proximate event, blurring it out in time and space till it melds imperceptibly with the continuum.

The temptation to think of the weather as chunky leads us to discount effects that aren’t chunky, overly focus on things that are, and therefore attribute too much to them. We also overlook events that don’t happen. When, for example, we think of the impacts of temperature change and variation, we tend to think of heatwaves as chunky things. They start, they stop⁴, they have a limited geographic range and they kill people. On the other hand, seasonal cold weather is background: it tends to get far less attention unless there is an unusual cold wave. A lot of effort goes into trying to understand the causes of these “events” and their impacts. A lot less goes into understanding events that didn’t happen, perhaps for obvious reasons, but when we’re talking counterfactuals, as we often are with attribution, then they are just as important.

The relationship between temperature and mortality varies from place to place but typically sees high mortality at low temperature (for the location) and high mortality at high temperature (for the location), with a minimum somewhere between the two. An analysis of heat-related mortality (deaths associated with temperatures above the local minimum of mortality) and cold-related mortality (deaths associated with temperatures below the local min) found that, for Europe during the study period, cold-related mortality per year (290,104 [213,745, 359,636]) was much larger than heat-related mortality per year (39,434 [30,782, 47,084]). An earlier study covering a much broader area had similar conclusions. Other research cited in IPCC AR6 WG2 showed that in some areas heat-related deaths exceeded cold-related deaths, particularly warmer low-income countries, so its not clear cut. The newer study shows among other things that the estimated mortality depends on the analysed timescale and not in an intuitive way. Data aggregated at daily, weekly, monthly timescales lead to different results with daily data presumably giving the most accurate results.

It’s worth noting that these figures include all deaths associated with temperatures above or below the minimum in mortality, the so-called “optimal” temperature⁵. There will be deaths of these kinds in every year. They’re different from “excess” deaths which indicate how many more people died than would be expected for the time of year⁶. Excess deaths are more commonly calculated and quoted for heatwaves and add an additional layer of complication⁷.

Over the past twenty years, reductions in cold-related deaths outnumber increases in heat-related deaths. In general though, the hot side of the temperature-mortality curve is much steeper than the cold side so at some point with continued warming, the balance will flip and even potentially butt up against human physiological limits. There’s also the issue that with heat-related extremes, we’re starting to see conditions that haven’t previously been observed in a region, at least not within living/ancestral memory. It’s also important to remember that increased mortality is only one aspect of increasing temperatures: climate affects everything. With cold extremes becoming less frequent and less extreme, we’re not moving outside of historical bounds of temperature, although we may well be for particular seasons⁸. That’s not to say that it isn’t an issue, just that the issues are different.

But back to events. By focusing on events rather than the more general relationship between climate, particularly temperature in this case, and mortality, we miss an important part of the overall picture and by focusing on the more extreme events for attribution, we can start to build a completely misleading one. Climate change affects the whole range of temperatures throughout the year, raising winter temperatures as well as summer temperatures. If we look at this by way of events, we see lots more heatwave events, but increasingly few coldwave events. Even if every heat and cold wave was reported consistently, this would still provide a misleading picture.

It is unlikely that reporting is consistent given gaps in observation networks and the relative attention that is paid to different geographical areas. We don’t have regular, systematic global monitoring of many types of extreme events⁹. Disaster databases typically focus on larger events. Further filtering comes from the interest in record breaking events which happen now more frequently at high temperatures than at low ones. We must also contend with the way that events are defined in different situations which can lead to further confusion. For example, an attribution study may look at a particular peak in extreme heat, but the impact analysis might assess mortality for the whole year. Mixing and matching these results can lead to rather alarming combinations. The article mixes up 70,000 deaths from non-optimal heat for a whole year with deaths during the “2022 European heatwave” and places this next to attribution studies of individual “events” in 2023 in which a heatwave in Southern Europe was deemed to have been virtually impossible without climate change. Munging all this together might lead an incautious reader to conclude that climate change killed 70,000 people in one heatwave in 2023. Building a globally representative picture from what comes out of this process is difficult, if not impossible even if one is paying attention the nuances.

An event-centric way of thinking can also sideline some of the worst climate-related disasters. Slow onset events, such as extended periods of drought do not lend themselves to analysis as single events. Megadroughts can conceivably last longer than the periods typically used to assess what “normal” climate is. `Recent drought in the Horn of Africa which lasted for five seasons across three years was followed by heavy rains and flooding in places. In places the dry soils from the drought increased runoff and may have contributed to flooding. One event or two or more or none? The interaction of the dry and wet conditions with everything else that is happening in the region and globally led to great suffering, food insecurity and human displacement. EM-DAT which is the most widely used repository of disaster information doesn’t list any deaths from the drought in Somalia, Kenya, or Ethiopia (the three countries I checked) although millions were “affected” and there is at least one study that suggests significant mortality in Somalia in 2022 alone.

EM-DAT also excludes small-scale disasters. Data gathered under the Sendai framework includes smaller disasters and suggests their aggregate effect is large, but the reported data does not allow disaggregation of climate-related disasters from others, so cannot easily be used to understand climate impacts.

It is important to understand the most unusual events of course, but we need to understand them within the wider context and not separated from it. Systematic monitoring (of extremes and everything in between) and reporting would help. The individual impacts of extreme events are often large and consequential, but to understand the overall impact of climate change we need also to analyse and present the far more numerous situations where impacts are individually smaller but collectively significant. Data for meteorological and impact variables at an appropriate resolution, with sufficient information to analyse are essential but not currently widely available. This could be supplemented by attribution analyses that look at impacts across the full range of conditions, not just individual “events”.

-fin-

1. The History Boys put it rather more earthily. ↩︎
2. Itself variously attributed and transcribed. ↩︎
3. Very dry and very wet soils are generally less ready to absorb further rainfall quickly. ↩︎
4. Or maybe not. Once one sits down and starts to think really hard about how to define an event, it rapidly gets difficult. Heatwaves don’t just come and go, they don’t have well defined outlines. They can shift around, individual extremely extreme extreme events can be embedded in a broader less extreme extreme event. Different definitions will lead to different groupings. It’s really a nightmare. ↩︎
5. Optimality for other impacts – crop yield, energy generation, etc – might differ. Global climate optimisation is tricky and non-local. ↩︎
6. Although confusingly, some papers refer to temperature related deaths as excess deaths indicating an excess above what would be expected at the optimal temperature (I assume, it’s a bit of a nightmare really). ↩︎
7. By comparing the target year to an average for recent years, excess deaths (in)effectively detrend the data, partly neutralising changes in mortality associated with secular changes due to anthropogenic effects, say, or adaptation. ↩︎
8. Recent very mild winters – with very few frosts – have meant that pests have been able to overwinter in our garden ready to eat what we plant the next year. The temperatures haven’t been exceptional considering the full annual cycle, but they have been unusually high for winter temperatures. ↩︎
9. We could perhaps use reanalyses, but for rainfall extremes they’re not quite ready for the primetime. ↩︎