Commentary by Prescient Weather co-founder Richard James, Ph.D
After a winter of notable extremes in the Northern Hemisphere, a period of renewed volatility has brought both exceptional warmth and unusual cold to mid-latitude areas in early April. Following a very warm March in the eastern U.S., a spring freeze has occurred in the Southeast, and western Europe is also seeing a dramatic change to colder conditions.
Severe spring freezes can wreak havoc on agricultural operations that involve early-flowering crops such as fruit trees, blueberries, and even winter wheat. For example, Georgia’s peach and blueberry crops suffered large losses in 2017, when a freeze followed exceptional late winter warmth that caused the plants to come out of dormancy much earlier than usual. To illustrate the problem, here’s a chart of Atlanta’s temperatures in early 2017: temperatures were persistently much above normal from mid-January to early March, but then a hard freeze occurred in mid-March. The freeze itself was not climatologically unusual, but the previous warmth was, and the combination of the two was devastating.
Spring Freeze Trends
The 2017 spring freeze was one of the most damaging in many decades for the U.S. Southeast, based on the accumulation of growing degree days prior to the freeze. It’s interesting to consider, then, whether this kind of thing is happening more frequently around the world. From one standpoint we might expect damaging freezes to become less common as the globe warms, because freeze itself becomes less likely in the critical early spring window. However, if increasing winter warmth is causing plants to break dormancy earlier, then subsequent freeze could become an increasing problem; here’s a paper that discussed this idea in the wake of another damaging U.S. freeze in 2007. Moreover, a popular recent hypothesis is that “weather whiplash” is on the rise, and if this is correct, then it suggests that damaging late freeze events will become more widespread.
But what does the data show? If we just look at Atlanta, there’s no immediately obvious trend. The chart below shows the total growing degree days (sum of daily average temperature excess above 10°C) that occurred prior to the last hard freeze each year (defined as a minimum temperature of -2.5°C or lower). If it wasn’t for 2017, we might say there’s a hint of a downward trend, but 2017 was the most extreme event in at least 90 years by this metric. Note that this data comes from Atlanta’s international airport, and localized urban warming could have a significant effect on the results here.
A nearby site in Georgia with much less urban development is Athens, home of the University of Georgia, and also home of this author. The 2007 freeze stands out in the data from Athens, and both 2017 and 2018 saw damaging freezes; but there’s no statistically significant long-term trend.
A more comprehensive view can be obtained from the ERA5 reanalysis. Based on hourly data back to 1958, here’s a map of the long-term average of total GDDs prior to the last -2.5°C freeze at each grid cell. Interestingly the problem of the spring freeze is heavily concentrated in the southern and eastern U.S., where the climate involves the necessary spring volatility; it seems that nowhere else in the hemisphere regularly sees the same combination of sustained warmth prior to a hard freeze in spring.
That’s not to say that a damaging spring freeze can’t occur elsewhere: a map of the 1958-2020 maximum shows a few areas of concern, notably in the Middle East.
How about ERA5 trends over time? Here’s a chart of the Northern Hemisphere land area percentage that reaches freezing (not necessarily a hard freeze) after at least 200 GDDs have been accumulated. There’s perhaps a very slight hint of a rising trend, but certainly nothing statistically significant.
If we use the -2.5°C threshold, there’s no suggestion of a trend at all.
In the results above, I’ve used a 10°C base for the GDD calculations, which is standard for many warm-season crops, but 5°C or even 0°C is often preferred for other crops such as wheat. The two charts below show results for these base temperatures, with the threshold GDDs (400 and 800 respectively) chosen to reflect levels that are rarely attained and would probably involve significant crop damage. Again, there’s no clear trend.
Interestingly, then, there’s no clear sign of hemispheric-wide trends in accumulated warmth prior to the last freeze of spring. However, a map view suggests that some changes have occurred regionally, as we might expect. A simple difference map between 1958-1989 and 1990-2020 shows an increase in pre-freeze GDDs over much of the central and eastern U.S., and also in parts of interior Asia, but a decrease in freeze damage is indicated along the U.S. Gulf Coast.
Let’s take a closer look at the results over the contiguous U.S.:
The pattern in the U.S. appears to be consistent with a northwards migration of the spring freeze zone. In the deep south, it seems that warming may have significantly reduced the frequency of hard freezes in early spring, and so freeze damage has decreased. But in contrast, the risk has risen in the central and eastern U.S., because spring freezes can still readily occur, but GDDs are now accumulating more rapidly than they used to prior to that last freeze.
If we take a look at historical data from Wichita (southern Kansas) and College Station (southeastern Texas), the contrasting trends support the idea that spring freeze risk has migrated north in the past 70 years – see below. Kansas is seeing more warmth prior to the last hard freeze (although perhaps not enough to be a big problem yet), but southeast Texas is seeing fewer hard freezes (although they are certainly not yet rare).
Similar global patterns of change are seen in the Japanese JRA-55 reanalysis:
There’s a lot more work that could be done – and is being done – to investigate these trends and to quantify the changing risk of freeze damage across a range of agricultural sensitivities. This paper found that freeze risks in Europe “vary largely depending on the species and geographical locations”, which is no surprise. For anyone that’s interested, I’ll be happy to share the ERA5 derived data I used in this post.
To round off the discussion, it’s also worth noting that the reanalysis data – and especially JRA-55 – show increasing spring freeze risks in the Southern Hemisphere.
In terms of area affected, southern South America appears to be the focus, but this should be confirmed with ground-truth climate data, as reanalysis is more prone to error in the data-sparse south.