Jay Armstrong and Eli Flory stand behind a seed drill at Armstrong’s farm in Muscotah, Kansas, on the last day of winter wheat seeding.
(Photo by Matthew D. LaPlante.)
By Matthew D. LaPlante and Corryn La Rue The Pitch | Dec. 12, 2023
Eli Flory snuffed out a cigarette in the dried mud, twisting the sole of his boot over the burning nub. He scanned the field ahead of him and climbed back into his tractor.
It was nearing noon on a Thursday in late September, the last day for wheat seeding at the farm where Flory works as a heavy equipment operator in Muscotah, Kansas. He had been up since before the sunrise. He would be at it for the rest of the day.
Inside the tractor, Flory consulted the bright digital screen to his right. Broad colored lines indicated the parts of the field that had been covered by the enormous machine he was pulling behind his rig. He set the tractor into gear. It rumbled forward. Flory watched the screen to make sure he wasn’t leaving any space between one line and the next, covering the field as if he were painting a fence. Behind him, the seed drill did its work. A line of bladed disks cut parallel trenches into the soil, a few inches deep. A blower pushed the seeds into the channels. A row of bulky toothed closing wheels collapsed the troughs, burying the kernels.
The sowing of any one seed happens in a split second, but it takes many days, each year, to cover the 400 acres of wheat fields on this farm, and the decisions that are made in these moments have a significant impact on the harvest that comes nine months later. The cutting discs, for example, can be adjusted up and down by fractions of inches and, on a farm like this, with gently rolling hilltops and shallow bottoms, with nearby streambeds and woodlands, and with soil that can change from silt to loam to clay in the span of a minute’s stroll, the depths at which seeds are sown are not negligible matters. And even before any of this happens, farmers have to decide not just when to plant, as the first frost approaches, but also what to plant. Soft winter wheat, like that which Flory was sowing, offers a higher yield but costs more to get to harvest than hard winter wheat does. That decision, though, is dwarfed in complexity by the choices over seed variants. With 16 billion base pairs, the wheat genome is five times larger than the human genetic code, and researchers have tinkered with tens of thousands of the plant’s genes in search of seeds that are better at one thing, often at the expense of another. For a premium, farmers can custom tailor their seed selection to their fields. Later in the growing season, there will be questions about nutrients, herbicides and pesticides, about which mixtures to use, about when to apply them, about whether to apply them, about how to apply them.
Consider the complexity. A chess game of just ten moves — five processions from white, five reactions from black — has 70 trillion possible trajectories. Wheat farming is a game that lasts thousands upon thousands of turns each season. The permutations are endless.
But there’s one variable that looms over all of the others. “We can’t do anything about the weather,” Flory said. “I’ve been here for years that had lots of drought. I’ve been here for years that had too much rain. Really good harvests, really bad ones, you can’t do much about that. It’s out of your hands.”
That’s precisely why the owner of this farm, Jay Armstrong, has a simple philosophy, one he preaches as a member of the Kansas Wheat Commission to growers hoping to emulate his decades of success, measured by far more good years than bad ones.
“You’ve got to farm with optimism,” he said. “I always start with ‘this year is going to be a good year.’ And then you sit in your tractor and you dream of big yields. That’s how I think. And so every year I’m going into it the right way, the best way, and I know I’m not going to get it every year. But I’ll keep doing it that way. And then you improvise from there — you control what you can actually control.”
Farmers cannot control the weather. They cannot control the interannual and decadal ebbs and flows of weather patterns that comprise climate. And they certainly cannot do anything about the way climates are changing as the world warms. Indeed, some are still uncertain whether that’s actually happening at all.
But in the wake of a massive drought and the worst statewide wheat harvest in decades, many Kansas wheat growers are looking to make a few better moves. And one way to do that may be to stop doing things the same way year after year, instead letting long-term atmospheric signals inform decisions at every step in the process, even long before the seeds go into the ground.
To do that, farmers cannot simply look to the skies above them, or even to the weather that is on its way, a few days or weeks out. That’s because the signs that point to meager and bountiful harvests may have already come and gone, a year or more ahead of time — from a place on the globe many thousands of miles away.
Something fantastic happened in 2016. Across the state of Kansas, everything seemed to line up perfectly for a good wheat harvest. Just about everyone was playing a perfect game of chess.
That’s easier, of course, when the board is set up well—when the variable that matters most comes aplenty. And in that year, it did.
Western Kansas, in particular, is characterized by a dry climate with abundant sunshine. Wheat can do well there, but it still needs some help. Nothing grows without any water at all, after all. And in the girth of the growing season of that harvest year—October of 2015 through April of 2016—Kansas did well with rain and snow, receiving about 15 to 20 percent more than the average from the past quarter century, with western Kansas in particular getting a good helping of precipitation. The result was an all-time state record of 57 bushels of wheat per acre.
That stood in stark contrast to what had happened just two years earlier, when rainfall across the state was hard to come by. In that year, the southern plains were anomalously dry, and Kansas registered an average of just 28 bushels per acre, a sort of yield that would have been low even back in the 1970s, before many of the technological and genetic innovations that have led to much greater yields in more recent years.
If wheat growers in Kansas could have known ahead of time that the boards would be set up this way, it would likely have impacted many of the moves they made throughout the process, from seed selection to planting decisions to nutrient supplementation to harvesting windows. But modern weather predictions are only accurate for about a week to 10 days; forecasts further out than that are generally considered to be no better than a coin flip.
To gain a better understanding of what an entire season of weather might look like, researchers often turn to the seas, where small changes in temperature can have a warming or cooling effect on the air above the ocean, which can have big impacts on atmospheric pressure, which in turn can influence the large-scale wind patterns that carry moisture west to east across North America. For decades, the most common place to look for the starting point of this atmospheric Rube Goldberg machine has been the tropical Pacific, in the regions associated with warmer El Niño and cooler La Niña trends. The former tends to bring warmer and drier weather to the northern United States states and wetter weather in the southern parts of the country. The latter often brings more precipitation to the Pacific Northwest and the Great Lakes region, with colder weather in the north and drier weather in the south.
In both cases, though, Kansas is largely missed as part of a band of states stretching from northern inland California to western Missouri, where the impacts of El Niño and La Niña are less predictable. This is one of the reasons why Jay Armstrong’s every-year-could-be-a-good-year farming philosophy makes plenty of sense. Simply put, the long-term record of wheat harvests substantiates that strategy.
The first known wheat crop was grown in this part of the country in 1839, and wheat exploded in the decades that followed. The vast wheat fields of Kansas were so awestriking to Katherine Bates during a cross-country train trip in 1893 that the college professor from Massachusetts was inspired to describe these “amber waves of grain” in a poem now famously known as “America the Beautiful.” When people sing that song, they’re singing about Kansas, a place where the average bushel per acre yield has grown steadily over nearly two centuries of cultivation. Across dozens of cycles of El Niño and La Niña during that time, Kansas hasn’t just remained a good place for growing wheat; innovation has made it an even better place for growing wheat.
The second reason why Armstrong’s optimism is warranted is that even if the influences were simple — in an El-Niño-is-good-and-La-Niña-is-bad sort of way — that information would be most helpful if growers could know with some degree of confidence that such phenomena were on their way a year or more ahead of time, long before their seeds were ordered from suppliers. And that’s simply not information that climate scientists have been able to offer to farmers.
Not yet, at least.
Long before he made the ultimate decision to abandon nearly two-thirds of his wheat fields, John Jenkinson knew it was going to be a bad season. Jenkinson, who managed his first dryland wheat crop — from sowing to sales — when he was just 13 years old, had weathered bad seasons many times before, but the winter of 2022-2023 was extreme by just about any measure.
As Jenkinson and other growers were putting seeds in the ground, about 18 percent of the landmass in Western Kansas was in a state of exceptional drought — the most extreme classification of dryness against historical averages, characterized by widespread crop and pasture losses, exceptional fire risk, and vast shortages of water in reservoirs, streams and wells. By April, when farmers generally expect to be getting a boost of spring rains that will push their crops over the finish line, exceptional drought had overtaken about 60 percent of the region, and there was not a single place in the entire state where rain and snow totals were anywhere close to historical normals.
Southwest Kansas, Jenkinson noted, generally gets about 16 inches of rain each year. In the last harvest year, he said, “we had only had about four inches of rain for the whole year. And so much of the wheat that got planted never did emerge. It never did sprout. It never did come up because it was so dry. It wouldn't grow at all.”
In the middle of it all, Jenkinson recalled, the winter winds kicked up, sending clouds of dust into the air. “It actually looked like the ‘dirty thirties’” he said, recalling the dustbowl conditions he’s read about in histories of the Great Depression.
Although Kansas exists in a liminal space between the areas of North America where the effects of El Niño and La Niña are most historically pronounced, a particularly strong phase of one or the other can push the impacts into the void. And that appears to have been the case in the most recent harvest year, which came at the end of a rare “triple dip” La Niña — three years of cooler waters in the eastern and central tropical Pacific, which tends to result in a more variable and northerly jet stream, pushing North America’s greatest source of precipitable water away from the central plains. And all the while the impacts mounted, especially for farmers in the western part of the state.
“The problem is, in western Kansas, we do what is called a summer fallow practice. We don't get enough rain that we can grow year after year after year of a crop,” Jenkinson said. “Let's say, for instance, we have 160 acres of ground… you split that up into 80 acres, and in 2022 on that 80 acres you plant wheat, but the other 80 acres you have to let sit and absorb all the moisture it can until the following year. That’s the only way we can raise crops in southwest Kansas without irrigation. And so, if you're in a crop rotation where you have to summer fallow, if you're not real cautious, if your balance is thrown off, you could have a year when you really don't have much of a crop at all, and you have to wait until the following year to have enough moisture to grow.”
But it’s hard to be cautious if you don’t know what’s coming. “Eighteen months of advanced notice would be a huge help if we could rely on it,” Jenkinson said. “You know, that would be very, very helpful if that was accurate information we could really lean on, and pre-plan for.”
That sort of information could soon be on the way.
The year that Katherine Bates came through Kansas by train, 1893, was a bad one even by turn-of-the-century standards. United States Department of Agriculture records suggest that farmers averaged just nine bushels per acre that year, but even in a good year, back then, those amber waves would have brought in less than 20 bushels per acre, on average. By way of contrast, over the most recent decade, the worst years hovered around 30 bushels and the best around 60.
To make long-term records of wheat harvests useful, then, it can be helpful to treat the data the way climate scientists often treat records of rainfall, temperature, sea-level pressure, or another climate variable, comparing any annual number to an average of the 30-years that preceded it to control for the ways in which climatological conditions naturally ebb and flow over years and decades. This also smooths out the impact of industrial advances in agriculture over time. The resulting time series is a better reflection of bountiful and meager years within the context of the era in which they occurred. It also happens to strongly correlate with total rainfall in the central plains. That should surprise no one; even given all of the many variables that impact the complex chess game that is wheat farming, more rain generally results in better yields and less rain generally results in worse yields.
And, as it turns out, over the past 120 years, the Kansas wheat harvest index has been well correlated to various indices that record the cycles and magnitudes of El Niño and La Niña across large sections of the central Pacific. Seasonally, though, these associations tend to peak one season before harvest. They weaken by two seasons before the harvest. They weaken again at three. By four seasons before the harvest — the sort of “advanced notice” that Jenkinson suggested could be helpful, if it were reliable — the correlation has all but dissipated.
In recent years, though, climate researchers have recognized that El Niño and La Niña are not binary states of being. The chess board can be set up in all sorts of ways. The neutral state between warm and cold has wide ranging impacts on global weather patterns. The speed of transition matters, too. Not surprisingly, the severity of temperatures is a big deal. Length of these phases is important, as well, hence the impact of the recent, rare “triple dip” La Niña on the typically liminal Kansas. And finally, the specific places where El Niño and La Niña are centered have a tremendous impact on the teleconnections — causal associations between meteorological phenomena that happen in distant places — that are instigated by warmer or cooler ocean waters.
So, while a large-scale assessment of the central tropical Pacific might not offer actionably predictive signals that precede wheat seeding in Kansas, some smaller areas of the ocean may offer substantial clues about the weather patterns that will be most influential to crop growth a year or more down the road.
And as it happens, one of these areas is a small region where El Niño is generally thought to have been first noticed, by fishermen working just off the coasts of Ecuador and Peru. In this region, sea surface temperatures are strongly correlated to eventual wheat harvests in Kansas one season before the harvest, strengthen two seasons before, become even stronger three seasons before, and retain a high level of significance four seasons before the harvest. At five seasons out, the relationship is weakened, but retains its statistical significance. The signal doesn’t dissipate until six seasons ahead — about half a year before seeds even go into the ground.
Correlation is not causation, and the global atmosphere is a big and noisy space, where a lot of variables come into play. Even a grandmaster chess player with a good read on their opponent’s tactics can be left without a clear strategy toward the endgame by a surprising movement of a single pawn. In like fashion, there have been years in which a warm spike in the far eastern Pacific, just off the coasts of Ecuador and Peru, didn’t result in a bountiful crop a year or more later. And there have been times in which a cold period in this region didn’t crash Kansas wheat harvests down the line.
But when the sea surface temperatures that preceded the 10 best and 10 worst harvests of the past few decades are combined into a composite picture, the pattern is starkly clear.
Bountiful wheat yields in Kansas follow El Niño events that appear to materialize in the far eastern Pacific up to four seasons ahead of harvest. Meager wheat yields follow La Niña events that begin even further ahead of time.
The insurance adjusters were already making their way around western Kansas in the early spring.
“They saw that the wheat was too thin, or too short, to be harvested and that it would have such a small yield that it wasn't worth the time and effort,” said Aaron Harries, the vice president for research and operations for the Kansas Wheat Commission. “Those fields would then be classified as abandoned, meaning they were never harvested, and abandoned acres were close to 20 percent this year in Kansas.”
The value of those abandonments will be measured in hundreds of millions of dollars. Part of that will be covered by federal taxpayers under the Federal Crop Insurance Program. Part will be covered by farmers’ deductibles. And the part will be covered by the insurers, who will correspondingly raise rates on wheat growers for this season and future seasons, just as insurers do after historic earthquakes, floods, wildfires and hurricanes.
“Crop insurance is the best risk management tool that we have in place,” said Rep. Tracey Mann, a Republican member of the House Committee on Agriculture, whose district — one of the largest in the country by square mileage — covers most of Western Kansas. “I would say crop insurance is the most cost effective way that we have in our country, and really throughout the world I believe, to have a robust, stable, steady food supply. That's really the purpose of it. And we certainly see it widely used in Kansas.”
But what if it didn’t have to be so widely used? What if farmers, insurers, and federal farm subsidy planners could know with greater certainty whether the year ahead would be meager or bountiful — not just for wheat in Kansas, but for other crops and in other states? What if the most sensical path forward wasn’t always planning for a good year — and doling out insurance payments when those years don’t come — but paying farmers proactively to fallow fields in places where climate signals suggest a high likelihood of a meager harvest, and doubling down in places where the signals suggest the coming yields will be bountiful?
To be clear: It’s unlikely many farmers would be ready for such a transition. Armstrong, Jenkinson, Harries, and Mann all expressed significant hesitation about how such signals would be interpreted, whether those signals would be reliable, and what the past can actually tell wheat growers about the future.
“The one thing farmers are not going to do, particularly in this economic environment,” Jenkinson said, “is put too much emphasis on forecasts that could be wrong.”
To get good wheat, after all, “you've gotta put some inputs into it. Fertilizer, nitrogen, those kinds of things,” he said. “Farmers in this part of the world are not afraid to do that, but they're very cautious about doing it because, too many times — more times than not, I should say — there's been a forecast that, hey, you know, it could be a wet spring. So you go out and you put the fertilizer on, but that wet spring doesn't really happen, and then you have a lot of expense into the wheat crop that you don't get back if the wheat fails.”
To be even more clear: The science isn’t there yet, anyway. Kansas wheat yields correlate to rain, but are even more significantly associated with very strong El Niño and La Niñas events in that very small region of the very far eastern Pacific — an indication that there are other meteorological variables at play. These correlations may have even slightly strengthened in recent decades, a period in which levels of greenhouse gasses have substantially risen in the atmosphere, but it’s not yet clear how this relationship will play out under continued expected warming.
What is clear is that the ability to answer these questions is increasing, and quickly.
Time and technology are why. Precipitation records in Kansas were sparse one hundred years ago. Sea surface temperature observations in the far eastern Tropical Pacific were even more sparse. Observational equipment was poorly standardized. Government records were often products of estimation. A lot of suppositions have been made in order to build vast datasets that allow researchers to compare an index of an agricultural product to an index of ocean temperature anomalies.
Starting in the late 1970s and early 1980s, though, space satellites permitted—with quickly increasing accuracy—the monitoring of weather variables across the globe, and those space-observed measurements were affirmed by observations on the ground. Now, advances in artificial intelligence are permitting vastly complex analyses — allowing not just for comparisons of temperatures in one part of the ocean to precipitation in one part of the world, but unveiling how the intricate inner-pieces of that teleconnection can impact the relationship.
In May, a team of researchers from the United States and China — two of the largest wheat producing nations in the world — demonstrated how an AI-enabled forecasting model could forecast sea surface temperature changes up to 22 months ahead of a transition into El Niño or La Niña in the central Pacific. The same month, a team from Peru showed they could use AI to make similar predictions for an oceanic region just off the coast of their nation — the same area that appears to be such a good long-term predictor of wheat harvests in the U.S. Midwest.
Michelle L’Heureux, a scientist at the national Climate Prediction Center who leads a team that updates the nation’s official El Niño-La Niña forecast, isn’t convinced AI will always be able to find an effectively predictive signal in the noise of climate data — and not because there’s too much noise, but because there’s simply not enough.
“One challenge for AI — and this is not a judgment on it at all — is that we just don’t have a lot of data,” she said, noting that most reliable observational sea surface temperature records go back to the 1950s and the even more robust data that came from weather satellites has only been around since the late 1970s and early 1980s. Cycles of El Niño and La Niña can be quite irregular, occurring every two to seven years, and each phase presents itself differently, so there simply might not be any well-established patterns for the algorithms to pick out, even if those patterns might theoretically be identifiable if the machines had hundreds or thousands of years of reliable data to work with. “You’re not dealing with a lot of historical data, which is very challenging for AI, because it’s a very data-hungry process,” she said.
What’s more, she said, there is still a lot of debate as to whether rising ocean temperatures in general — and the dynamic effects that stem from those changes throughout the atmosphere — will impact longstanding connections.
For now, at least, L’Heureux’s group has become quite proficient at predicting sea surface temperature shifts and strengths in the Tropical Pacific. The team correctly predicted a shift out of the three-year-long La Niña earlier this year. Its analysis suggests a 99 percent chance of continued moderate El Niño conditions through the winter and an 80 percent chance of it lasting through the spring. In general the team’s predictions don’t offer enough lead time to help farmers make pre-planting decisions for winter wheat, but farmers like Jenkinson said they’re willing to incorporate probabilistic assessments into their decision-making throughout the process if those forecasts can be shown to be effective.
L’Heureux is eager to have climate prediction models stand trial. Her team publishes its predictions online so that farmers and others can assess the accuracy themselves. She’d like to see AI developers do the same thing. “The proof is in the real-time,” she said. “Go ahead and set up a model. Put it on a webpage somewhere and update it and let’s see how it does with real-time data.”
However they are created, the predictions don’t have to be unfailingly accurate, or accurate at year-or-more lead times to be valuable. Every percentage of increased accuracy and every week of advanced warning can potentially help farmers put a chess piece into play a little better than they could before.
This year, just as he does every year — for so many years that he’s lost count — Scooter Scapanski made a spring and summer pilgrimage from Texas to Montana, following the historic pattern of wheat harvests, which generally come earliest the south and later further north.
The contract combine operator hits the same places year after year as part of a crew that keeps contracts with the same farmers one harvest after another. But as he drove a giant harvester across a field in eastern Kansas, Scapanski said that there were plenty of places that he and his crew didn’t even bother stopping through, especially across the Southern Plains.
“This year there was just no point,” he said. “They didn’t have anything to cut.”
Even still, Scapanski noted, there were plenty of places where the wheat came in just fine, just as it has, more often than not, for more than a century. “We still had plenty of work to do,” he said.
The devastation in western Kansas, where many farmers had to give up on their crops and others were settling for 20 to 30 bushels per acre, wasn’t mirrored in South Dakota, where farmers were flirting with 50 bushels per acre; or in Arkansas, where yields were near 60 bushels; or in Missouri, which hit a record 70; or in Illinois, which darn-near struck 90. Insurance payouts in those places will be minimal this year.
As for next year? That’s anyone’s guess. But at the moment, at least, things are looking up in Kansas, with enough summer precipitation to temper the years-long drought, and enough breaks from the rain, this fall, to get good seeds into the ground.
And, for whatever it might be worth — and that is still unclear — four seasons before the 2024 harvest, a warm patch in the far eastern tropical Pacific began to form. And from there it grew, and grew some more, just as it has in most of the years that have culminated with a bountiful wheat crop for the Wheat State.
As the cutting discs split the dirt, and the kernels fell into the trench, and the closing wheels covered the seeds with soil, the board was set.
Matthew D. LaPlante is a journalist and climate scientist. He has reported from more than a dozen nations on subjects ranging from ritual infanticide in Africa to insurgent warfare in the Middle East to gang violence in Central America. He teaches in the Department of Journalism and Communication at Utah State University. Corryn La Rue is an agriculture reporter and sixth generation California farmer. She is a former White House correspondent and the host of California Ag Today. She is an alumnus of Utah State University with degrees in agricultural communication and journalism.
LaPlante, Luthiene Alves Dalanhese, Liping Deng, and Shih-Yu Simon Wang are the authors of a recent study on the ways in which East Pacific sea surface temperatures may offer early predictive signals for harvest yields in the U.S. Midwest. The study is under peer review. None of the individuals quoted in this article were involved in the study.
