1. Introduction
The 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.” More than 130 years later, Kansas remains the largest producer of wheat in the United States; indeed, it is often called “The Wheat State.” Today, winter wheat, which accounts for about three-quarters of the total wheat production of the United States, is grown on approximately 7.3 million acres across Kansas, with average yields of about 42 bushels per acre over the past decade and a value of more than $2 billion annually (U.S. Department of Agriculture 2022).
This cereal grain is a worldwide staple food, providing about one-fifth of the calories consumed by humans across the globe (Erenstein et al. 2022). However, despite more than 10 000 years of cultivation (Feldman and Kislev 2007) and modern advances in agriculture, wheat yields remain widely inconsistent from year to year. Over the past quarter century, statewide average yields in Kansas have ranged from 28 bushels per acre in 2014 to a historic high of 57 bushels per acre just 2 years later (U.S. Department of Agriculture 2022). At the heart of these bountiful and meager harvests are meteorological patterns that are highly variable from year to year. Records for total annual statewide precipitation have ranged from 15.3 to 40.6 in. with a significant share of precipitation falling during the summer between harvest and seeding (NOAA National Centers for Environmental Information 2022). Total precipitation during the girth of the wheat-growing season (October through April) is also highly variable, and there is evidence that the frequency of extreme precipitation events has shown a general increase (Rahmani et al. 2014), but there has been no statistically significant trend in total precipitation since the late 1800s (Fig. 1, blue line). Kansas growing season temperatures are also highly variable on an interannual basis and have trended warmer since the 1800s (Fig. 1, red line) echoing global averages.
Time series (light dotted lines) and trend lines (bold unbroken lines) for Kansas statewide ONDJFMA averages for temperature (red) and precipitation (blue) from the NOAA–CIRES–DOE Twentieth Century Reanalysis (20CR) datasets.
Citation: Journal of Applied Meteorology and Climatology 63, 8; 10.1175/JAMC-D-23-0121.1
It has long been known that remote forces from the Pacific Ocean like El Niño–Southern Oscillation (ENSO) exert a significant impact on seasonal precipitation and temperature across the globe through atmospheric teleconnection processes, with resultant agricultural impacts across the tropical and subtropical world (Iizumi et al. 2014). This includes parts of the central United States, with warmer-state El Niño conditions broadly thought to be consistent with higher precipitation over the nation’s most prosperous winter wheat-growing region, including Kansas, and colder-state La Niña conditions consistent with lower precipitation (Rajagopalan et al. 2000). Accordingly, winter wheat yields tend to increase during El Niño and decrease during La Niña, but this is an association that Mauget and Upchurch (1999), among others, have noted would be more valuable if ENSO forecast accuracy and lead times could be improved at least past the 9-month planting-to-harvest season for winter wheat (and preferably further, as the logistics of planting require many months of planning and field preparation). Across the globe, however, predictions at these lead times remain limited, and the skillfulness of interseasonal forecasts has been particularly weak in a liminal band of U.S. states including Northern California, northern Nevada, Utah, Colorado, and Kansas, which lie between the regions of the country that are most reliably and predictably impacted by ENSO phases. Even outside of this liminal band, however, seasonal precipitation has been shown to be particularly influenced by the differential impacts of ENSO’s distinct flavors, chiefly the East Pacific (EP) and central Pacific (CP) El Niño (Wiedermann et al. 2021), as well as the often-more-subtle interevent differences in the spatial patterns of La Niña (Capotondi et al. 2015). While there is not yet a widespread consensus on how to differentiate each of these flavors, with some identifying distinct EP and CP states for El Niño alone and others discerning up to nine separate states within the warm, cold, and neutral phases of ENSO in general (Johnson 2013), the flavors are most often described geographically based on the location and propagating direction of SST anomalies (An and Jin 2004). Considerable study has also been dedicated to the frequency at which different flavors have appeared over time and how models such as those developed under the Coupled Model Intercomparison Project suggest that frequency may or may not shift under various warming scenarios (Xu et al. 2017). Importantly, it has long been noted that, unlike the ternary view of ENSO (i.e., El Niño, La Niña, and the neutral phase), a flavor-based view defies description by a single index (Trenberth and Stepaniak 2001).
Connecting harvest yields to ENSO flavors may be particularly beneficial for agricultural planning and policymaking as prior research has identified several potential precursors for these various patterns of SSTs and their related oceanic and atmospheric effects. CP ENSO, for example, has been shown to exhibit high correlations with the 20-to-30-yr vacillations of the North Pacific meridional mode, the leading coupled variability mode of the subtropical Pacific (Chiang and Vimont 2004; Stuecker 2018). Amplified EP anomalies, meanwhile, have been linked to the South Pacific meridional mode, also at decadal frequencies (Zhang et al. 2014), as well as subsurface equatorial thermocline variations, which have been demonstrated to lead EP SST anomalies by at least two seasons (McPhaden et al. 2006). It has also been shown that cold wintertime SST anomalies in the western North Pacific (WNP) can strengthen westerly summer winds in the western equatorial Pacific, thus promoting the intensification of surface convergence and anomalous Ekman and geostrophic advection; depending on the relative strength of these mechanisms and the background state of ENSO, this may promote either CP-centered El Niño or EP-centered El Niño a full year after the WNP SST anomalies occur (Borhara et al. 2023).
Thus far, however, there has been little study on whether the historic variability of ENSO flavors impacts the yields of specific crops, such as wheat, in production-relevant places, such as Kansas. Addressing the ways in which these conditions have influenced crop yields in the past could substantially enhance the industry’s resilience and sustainability in a changing climate, particularly if reliable signals can be identified and used to inform agricultural decision-making in the preplanting and even preplanning phases of the crop cycle.
2. Data
For our analysis, we obtained Kansas’ extensive 1866–the present wheat harvest history from the U.S. Department of Agriculture National Agricultural Statistics Service. This record includes planted acres, harvested acres, yield per acre, and price received, providing insights into production trends over time. For instance, yields have improved from an average of 15.1 bushels per acre in the 1870s to an average of 42 bushels per acre in the 2010s, reflecting global advancements in seed selectivity, irrigation, crop management, and weather forecasting (Najafi et al. 2018). To limit the influence of industrial advances that lead to greater average yields over time and smooth the influence of climatological oscillations that can influence precipitation and other meteorological variables in the U.S. Midwest at interannual and interdecadal scales (Birka et al. 2010), thus highlighting yield variability in temporal context, we utilized the annual yield in bushels per acre divided by a 30-yr moving average for the same metric. We chose to use a running mean of 30 preceding years rather than a running mean of ±15 years, as this approach is preferential for providing near-present index correlations at the expense of presumably less accurate values from the late nineteenth century (i.e., 1895–2022 rather than 1880–2007). The outcome is a wheat yield index (WYI) spanning 128 years that retains substantial interannual variability (Fig. 2).
(a) From the U.S. Department of Agriculture National Agricultural Statistics Service, average statewide winter wheat yields in bushels per acre (dotted line), a 30-yr running mean of annual yields (thin double line), and the Kansas WYI (thick line), representing annual yield divided by the 30-yr mean yield. (b) Pearson’s one-point correlation maps for WYI and precipitation, 2-m temperature, and 250-hPa geopotential height from the ERA5 datasets.
Citation: Journal of Applied Meteorology and Climatology 63, 8; 10.1175/JAMC-D-23-0121.1
No meteorological variable is more influential to wheat production than water availability (Amir and Sinclair 1991), and even irrigation-supported cultivation is generally insufficient without natural precipitation (Zeng et al. 2021). To affirm that WYI is indeed well reflective of this variable, we used the NOAA Climate Prediction Center morphing technique (CMORPH) dataset for total precipitation (TP), which was chosen for its fine (8 km by 8 km) spatial resolution and observation-based, bias-corrected analysis from January 1998 to December 2022 to produce a one-point Pearson’s correlation map illustrating the connection between WYI and precipitation during the girth of the growing season [October–April (ONDJFMA)], demonstrating a high degree of correlation (p < 0.01) across the state of Kansas. Like many precipitation datasets, the global accuracy of the CMORPH set has been shown to be highly variable, with a warm-season overestimation bias across the United States and especially in the central plains (Tian et al. 2007), although this bias is significantly reduced in the cold season (e.g., the October–April period that we chose to assess potential precipitation associations with WYI). Moreover, comparative assessments using multiple other reanalysis datasets at same- and longer-term temporal resolution, including the fifth major global reanalysis produced by European Centre for Medium-Range Weather Forecasts (ERA5; Fig. 2b, top), at 0.5° resolution, produced correlative centers that were likewise arranged over Kansas, suggesting that, as an integrated system, the wheat yield index captures precipitation as an integral variable.
Multiple reliable ENSO-related indices provide estimates of SSTs and anomalies going back to the late 1800s, thus coinciding with the onset of reliable harvest records in Kansas, including Niño-1 + 2, Niño-3, Niño-3.4, Niño-4, the bivariate ENSO time series (BEST), and the Southern Oscillation index (SOI). (As the SOI tends to demonstrate an inverse relationship with the other ENSO-related indices, for ease of visual interpretation we have transposed it on all figures, where it is thus designated SOI*-1.) These indices were gathered from the NOAA Physical Science Laboratory’s El Niño Index Dashboard (https://psl.noaa.gov/enso/dashboard.html). Several other Niño-related indices, as well as the Pacific decadal oscillation and the Atlantic multidecadal oscillation, were tested without obvious associations.
To examine SSTs outside of the reference regions used for these indices, and to explore the reference indices in greater oceanic and atmospheric contexts, we additionally utilized the ERA5 datasets for SSTs and mean sea level pressure (MSLP) at annual and seasonal time scales. Although we include TP and MSLP in our analysis and discuss other climate variables known to be impacted by ENSO states and clearly influential to the wheat harvest, our choice to emphasize WYI as the outcome of greatest interest reflects a desire to identify simple and potentially actionable associations, permitting farmers, agricultural planners, and water managers to inform decision-making through a lens of historical climate outcomes.
3. Preliminary analysis
As the winter wheat-growing season in Kansas generally runs from the September planting to the June and July harvest, with August weather impacting the soil conditions under which seeds will be planted in September and October (Shroyer et al. 1996), our analysis commenced with an examination of August–July “crop year” averages of each of the chosen ENSO indices from 1895 to 2022. A cross correlation of all the indices across the 128-yr time series (Fig. 3) shows that each was significantly and strongly associated with the WYI (p < 0.01) in the same crop year (green bars), from SOI on the low side (r = 0.23) to Niño-1 + 2 on the high side (r = 0.32).
Correlation coefficients for the Kansas WYI and average SST anomalies from the August–July crop year immediately preceding the harvest (green), preceding the harvest by 1 year (yellow), and preceding the harvest by 2 years (orange) for six ENSO-related indices.
Citation: Journal of Applied Meteorology and Climatology 63, 8; 10.1175/JAMC-D-23-0121.1
Interestingly, while Niño-1 + 2 (0°–10°S, 90°–80°W) is generally regarded as having the greatest variability of the most common ENSO-related SST indices (National Center for Atmospheric Research 2023), a 30-yr running correlation with WYI across the 128-yr time series suggests that within a single crop year, Niño-1 + 2 also maintains a much more stable relationship with annual harvest yields [standard deviation (SD) = 0.101] than the other indices (ranging from SD = 0.146 to SD = 0.240). As further illustrated in Fig. 4, the Niño-1 + 2 and WYI relationship (thick red line) has remained significant (p < 0.1) more often than the other indices and has consistently maintained the steadiest relationship with WYI. A notable exception came during a 30-yr period preceding the mid-1970s when the connection between SSTs and WYI fell across every index, and this period of faltering correlation was reflected in a similar analysis (not shown) using NOAA’s Extended Reconstructed SST, version 5 (ERSST.v5), from 1854 to 2023. (This period of weak association between the SST indices and WYI may be owing to natural multidecadal variability, but it might also align with a period of agricultural innovation and industrialization during the mid-1900s. Dimitri et al. (2005) explained that from 1945 to 1970 farm workers fell from 16% to 4% of the total labor force even as harvest yields were substantially increased, owing to rapid improvements in mechanization and specialization. However, both speculations require further exploration.)
A running correlation between 30 years of the Kansas WYI and the same 30 years of six ENSO-related SST indices. The lines for Niño-1 + 2 (red) and Niño-4 (blue) are emphasized for comparison and complemented with a linear trend line (dotted red and blue lines, respectively).
Citation: Journal of Applied Meteorology and Climatology 63, 8; 10.1175/JAMC-D-23-0121.1
Across the time series, the Niño-1 + 2 relationship with WYI also appears to have nominally strengthened over time (as indicated by the red-dotted linear trend), while an index that demonstrated a much stronger correlative relationship with WYI in the first half of the twentieth century, Niño-4 (thick blue line), appears to have lost a significant degree of this association (as indicated by the blue-dotted linear trend). (Whereas the more westward areas of the tropical Pacific, such as the Niño-3.4 reference region, have a more direct known teleconnective relationship with precipitation in North America, this may be related to a gradual shift from dryland to irrigated farming in Kansas, as well as improvements in reservoir storage and groundwater access, thus muting the “live or die” impact of receiving enough precipitation although not eliminating the broad impact of SSTs in general to influence weather conditions that continue to play a strong role in the quality of a crop, although this is also speculative and requires greater study.)
As shown in Fig. 3, none of the indices remained significantly correlated with WYI at a 1- or 2-yr lead (yellow and orange bars, respectively), seemingly indicating that primarily intra-annual teleconnective cycles are at play in the association between tropical Pacific SSTs and wheat harvests in Kansas, but not excluding the potential for associations that precede the planting season (i.e., those that exist in the time frame of four to six seasons ahead of harvest). To further explore these possibilities, we examined the seasonal precursive relationship of each index with WYI. While June, July, and August are a more commonly used unit of months to represent the boreal summer, the Kansas wheat harvest is generally complete in July, so we have chosen to use May, June, and July (MJJ) as the first preceding “season.” This also aligns with our decision to designate August–July as the “crop year.”
In the first MJJ preceding harvest completion, only one ENSO-related index, Niño-4 (based on surface temperature anomalies from 5°N–5°S to 160°E–150°W) demonstrated a statistically significant relationship with WYI across a 128-yr record (Fig. 5 upper table), and that relationship was relatively weak (r = 0.162, p < 0.05), indicating that, to a great extent, any concurrent influences that SSTs have on the conditions that impact cultivation may be resolved by the season of harvesting.
(top) A 128-yr cross correlation of six ENSO-related indices during the six seasons preceding harvest and Kansas WYI. (bottom) As in cross correlation, but for a more recent 44-yr period. The strongest seasonal correlations are bolded, and the confidence level of the correlation is indicated by blue shading.
Citation: Journal of Applied Meteorology and Climatology 63, 8; 10.1175/JAMC-D-23-0121.1
However, at a two-season lead (FMA), all six indices were significantly associated with WYI at p < 0.05 (Niño-1 + 2) or p < 0.01 (SOI, BEST, Niño-3, Niño-3.4, and Niño-4), and these relationships all strengthened at the three-season lead (NDJ) and generally maintained significance through a four- and five-season lead [August, September, and October (ASO) and MJJ-1, respectively]. Indeed, the strongest relationship between WYI and any of the preceding seasonal SST indices in our 128-yr cross correlation was Niño-1 + 2 in the first ASO season (r = 0.364, p < 0.01), which occurs during or just before the planting season. Niño-1 + 2 is also the strongest indicator of coming WYI in MJJ-1 (r = 0.311, p < 0.01), which suggests that SSTs in this ENSO reference region (based on anomalies from 0°–10°S to 90°–80°W) may be a worthwhile consideration for water managers and agricultural planners a season ahead of sowing.
Cognizant of the apparently shifting relationship between several of the indices and WYI in more recent decades, and of the expanded and increasing inputs and thus greater reliability of the indices in general since satellite-enabled observations, we also produced a cross correlation of each of the indices since 1979. Although the sample number was much smaller (n = 44 years), the outcomes were generally similar, albeit with fewer significant correlations at longer lead times. Notably, however, the relationships between Niño-4 and WYI appear to be drastically diminished in the more recent decades, while the associations between Niño-1 + 2 and WYI remained robust and even strengthened from MJJ-1 onward (Fig. 5 lower table).
The strength of both the longer- and shorter-term cross correlations (Fig. 5) as well as the 30-yr running correlation (Fig. 4) suggests Niño-1 + 2 anomalies may offer an early window into Kansas crop yields. This is interesting, as Niño-1 + 2 is the smallest and easternmost of the commonly used ENSO SST reference regions (corresponding to the area off the Peruvian and Ecuadorian coasts where local fishermen first recognized phases of warmer waters corresponding to variable fishing conditions), and may not itself have much, or any, direct influence on the eventual meteorological conditions impacting wheat yields in the U.S. Midwest. Niño-1 + 2 does, however, have a known precursive relationship with eventual ocean temperatures further west, such as those recorded in the Niño-3.4 reference region (5°N–5°S, 170°–120°W), which are the preferred areas for coupled ocean–atmosphere interactions (Trenberth and Hoar 1997) and where teleconnective associations with North America are known to be more direct (Hanley et al. 2003). Indeed, it can be seen in the 128-yr record that while a significant relationship between Niño-1 + 2 and WYI emerges six seasons ahead of harvest, in FMA-1 (r = 0.176, p < 0.05), sooner than any of the other indices, it begins to weaken by the vital spring before harvest, whereas most of the other indices strengthen or retain correlation with WYI in these later months, suggesting a more direct impact on variables, especially precipitation, that impact wheat growth during cultivation.
Albeit exhibiting a slightly more significant and earlier association with eventual wheat yields over 128 years, and an even stronger correlation in more recent decades, Niño-1 + 2 does not have an exceptionally stronger predictive quality to WYI than the other ENSO-related indices at lead times of four and five seasons (presowing) and is particularly close to the reference region to the immediate east, Niño-3 (5°N–5°S, 150°–90°W). For instance, in the first ASO before harvest in the 128-yr cross correlation, the coefficient for Niño-1 + 2 was 0.364 (p < 0.01), but Niño-3 was only nominally less correlated at 0.320 (also p < 0.01). One season earlier, MJJ-1, Niño-1 + 2 was correlated with WYI at 0.311, while Niño-3 demonstrated a correlation of 0.276 (both p < 0.01). In the more recent cross correlation, Niño-1 + 2 and Niño-3 swap nominally greater correlative behavior with WYI in NDJ and FMA, and both exceed the reference regions that are based on SSTs further to the west and the multivariate indices that are based on a greater range of variables.
It is also possible, of course, that other regions of the Pacific may demonstrate significant associations with the Kansas WYI. To explore these potential connections across the Pacific Ocean, we generated one-point Pearson’s correlation maps for SSTs and WYI for each of the six seasons preceding harvest completion from 1979 to 2022 (Fig. 6).
Pearson’s one-point correlation for WYI with ERA5 SST anomalies in the Pacific Ocean for the six seasons preceding the harvest.
Citation: Journal of Applied Meteorology and Climatology 63, 8; 10.1175/JAMC-D-23-0121.1
In line with the cross correlation from the same time period (Fig. 5 upper table), these maps reflect the initiation of a region of strong positive correlation with WYI in the preplanting period, five seasons ahead of harvest (MJJ-1), in the eastern tropical Pacific and within the Niño-1 + 2 (green box) and Niño-3 (orange box) reference regions, a combined area that is generally associated with EP-flavored El Niño events, which feature stronger western Pacific precursors than CP events do (Fosu et al. 2020) and, therefore, may be more likely to excite cross-Pacific teleconnections toward the U.S. Midwest.
This region retains its association with the eventual wheat harvest as the spatial distribution of significant correlation (p < 1.0) spreads west into the central tropical Pacific and latitudinally deepens during the cultivation period, four through two seasons ahead of Kansas harvest (ASO, NDJ, and FMA), but it does not extend much into the western portion of the Niño-3.4 reference region, or the Niño-4 region, which are generally associated with CP-flavored (aka “Modoki”) El Niños, affirming the finding, illustrated in Fig. 4, that the long-term connection between central tropical Pacific SSTs and Kansas WYI may have weakened. Interestingly, as the harvest grows near, the region of most significant correlation is substantially south of the Niño-1 + 2 and Niño-3 reference regions, a finding that may be worth consideration for near-harvest planning.
It is important to note that Kansas would not be the nation’s largest wheat-producing state if long-term conditions were not generally favorable toward the harvest of that crop, so this progression for all years and all yields from 1979 to 2022 represents SST conditions that are, generally speaking, beneficial for wheat production in that state. Even though ENSO is most often quantified in terms of SSTs, ocean temperatures alone, no matter where they are located, do not directly impact local wheat-growing conditions in the U.S. Midwest. What could affect the meteorological picture in Kansas is an atmospheric teleconnection inclusive of the impacts of ocean temperature on MSLP; MSLP on tropical airflow via the Walker circulation; and both MSLP and the Walker circulation on resultant upper-atmospheric pressure and circulations, with implications for temperature and precipitable water, among other variables.
Yu et al. (2017) described that substantially different global patterns of MSLP occur during EP and CP events. These characteristic arrangements can be seen for EP (as indicated by SLP correlations with an end-of-year annualized Niño-1 + 2 index) and CP (as indicated by SLP correlations with an end-of-year annualized Niño-4) in Figs. 7a and 7b, respectively. In Fig. 7c, the spatial distribution of mean sea level pressure is correlated with WYI, with a dipolar pattern closely resembling the EP arrangement of high and low pressure anomalies over the Pacific Ocean, including a more southern and eastern center of anomalously low pressure, jutting from the western coastlines of Colombia, Ecuador, and Peru, and a region of high pressure anomalies northeast of Papua New Guinea. This contrasts with the east Indian Ocean–centered high pressure anomalies associated with CP conditions.
(left) Annual ERA5 MSLP correlated with the (a) Niño-1 + 2 index, (b) Niño-4 index, and (c) Kansas WYI, demonstrating the responsiveness of wheat to sea level pressure anomalies similar to EP-flavored El Niño. (right) Annual ERA5 geopotential height at 250 hPa for (d) Niño-1 + 2 index, (e) Niño-4 index, and (f) Kansas WYI.
Citation: Journal of Applied Meteorology and Climatology 63, 8; 10.1175/JAMC-D-23-0121.1
The picture that emerges higher in the troposphere is also telling, with a region of low pressure in the subtropical Pacific of particular interest as its orientation in association with Kansas WYI (Fig. 7f), which is centered in the far eastern Pacific, more closely resembles EP-associated conditions (Fig. 7d) than CP-associated conditions (Fig. 7e). These are the conditions that may have a more direct potential force upon the patterns that impact Kansas weather (i.e., differentially expressed wave trains that are excited in the extratropical atmosphere, affecting both the strength and location of jet streams that impact winter storm paths over the United States, as described by Yu and Zou 2013).
Zhang et al. (2020) noted that both observations and coupled models have been employed to demonstrate that between the years 1981 and 2020 EP El Niño events were associated with generally lower average winter temperatures and substantially higher average winter precipitation in the U.S. Midwest generally, and Kansas specifically, than during CP events. Although that report was based only on DJF temperatures and precipitation conditions from the Global Historical Climatology Network/Climate Anomaly Monitoring System (GHCN/CAMS) and the gauge-based gridded monthly dataset from the Global Precipitation Climatology Centre (GPCC), respectively, the findings generally correspond to our observation (in Fig. 2) that the strongly EP El Niño–aligned WYI is positively associated with precipitation and weakly negatively associated with temperatures in Kansas.
To gain a better view of how wheat yields are impacted by both warm and cold phases in the tropical Pacific, we further explored the relationships between the indices and “bountiful” versus “meager” WYI. As the WYI for any given year is a comparison of that year and a running mean of the previous 30 years, and agricultural advances over time have steadily led to greater yields, this index exhibits a progressive bias (i.e., a greater number of years are quantified as above “1” than below “1”). To account for this bias, and because we also desired to examine these bountiful and meager harvests in relation to a greater view of SST behavior across the entire Pacific Ocean, which is better facilitated by satellite-era reanalysis, we once again limited this part of the analysis to the years between 1979 and 2022. The 10 most bountiful harvests during that time frame, as indicated by a high WYI, were 1979, 1980, 1982, 1983, 1984, 1997, 1998, 1999, 2003, and 2016. The 10 most meager harvests, as indicated by a low WYI, were 1981, 1989, 1995, 1996, 2002, 2006, 2007, 2011, 2014, and 2022.
Nearly predominantly across the indices, bountiful harvests broadly coincided with warm anomalies and meager harvests coincided with colder anomalies across the span of ENSO reference regions, affirming the associations that have been noted by Mauget and Upchurch (1999), among others. This also raises the prospect that the strongest precursive signals for bountiful harvests may come from one part of this region at a certain lead time, while the most predictive signals for meager harvests come from another ENSO reference region at a different lead time. Thus, while Niño-1 + 2 and Niño-3 may have the strongest sustained relationship with WYI at leads of three, four, five, and six seasons ahead of harvest in all years, it may not be the strongest indicator of a coming bountiful or meager year specifically.
The progression of anomalous SSTs preceding a recent, record-high bountiful harvest (2016) and a recent near-record-low meager harvest (2014) helps illustrate this point. In 2016, 8.2 million harvested acres yielded a total production of 467 million bushels for an all-time record yield of 57 bushels per acre in Kansas (National Agricultural Statistics Service). As shown in Fig. 8a, six seasons ahead of the harvest, in FMA 2015, an El Niño pattern was beginning to form, although it was not until five seasons ahead of the harvest, in MJJ 2015, that the warm anomalies flared in the easternmost ENSO index region (Niño-1 + 2, green box; and Niño-3, orange box). It can then be observed that the center of the warm anomaly shifted gradually west in ASO 2015 and NDJ 2015–16, retaining its signature El Niño shape in FMA 2016, at which point the association between Niño-1 + 2 considerably weakened and, by MJJ 2016, disappeared. This progression broadly follows the east-initiating, west-moving pattern of associations demonstrated in the cross correlations of Fig. 5.
ERA5 SST anomalies in the Pacific Ocean for the six seasons preceding (a) the harvest of 2016, which was one of the most bountiful in recent years, and (b) the harvest of 2014, which was one of the most meager. Composite ERA5 SST anomalies in the Pacific Ocean for the six seasons preceding (c) the 10 years with the highest WYI and (d) the 10 years with the lowest WYI between 1979 and 2022.
Citation: Journal of Applied Meteorology and Climatology 63, 8; 10.1175/JAMC-D-23-0121.1
As demonstrated in Fig. 8b, however, the meager harvest of 2014 (in which 8.8 million harvested acres resulted in 246 million bushels of wheat for a yield of just 28 bushels per acre, the weakest harvest in decades) was preceded by mixed anomalies in Niño-1 + 2 at a six-season lead and the emergence of a La Niña–like pattern of colder-than-usual temperatures in the eastern tropical Pacific in MJJ 2013. While a small center of cold anomalies remained in the Niño-1 + 2 reference region in ASO 2013, the La Niña pattern across the rest of the tropical Pacific faded, with no strong pattern emerging through the rest of the growing season; thus, this progression failed to follow the east-initiating, west-moving pattern of associations demonstrated in the cross correlations of Fig. 5. The most noticeable influencing feature in this series of images may be the emergence of the so-called blob, a persistent mass of anomalously warm water in the subtropical northern Pacific that impacted weather patterns across North America starting in 2013 (Bond et al. 2015), a reminder that while ENSO-related signals may exert a significant influence on the U.S. Midwest, other pan-Pacific and transoceanic forces will always be at play.
When viewed in composite, however, Pacific SSTs provide an informative picture of a distinctive progression of oceanic temperatures preceding bountiful and meager harvest years in general. As illustrated in Fig. 8c, composites of the six seasons preceding the 10 most bountiful harvests show a weak prevailing CP condition being overtaken by a westward-progressing EP El Niño, with a warm eastern “tongue” emerging five and four seasons ahead of harvest (MJJ-1 and ASO), deepening three ahead of harvest (NDJ), and weakening as harvest approaches and commences (FMA and MJJ). In the ASO and NDJ seasons, these composites most closely resemble a “type 8” or “type 9” ENSO formation in the nine-configuration spectrum developed at the International Pacific Research Center (IPRC) (Johnson 2013) in that a robust “warm tongue” of ocean temperature anomalies is centered in the East Pacific, “connected” to the South American west coast (as opposed to the “detached” warm anomaly in “type 7”), and met in the central Pacific by moderately cool anomalies.
Composites of the six seasons preceding the 10 most meager harvests (Fig. 8d) also show a strong eastern orientation of SSTs, in this case cold anomalies, emerging six and five seasons ahead of harvest (FMA-1 and MJJ-1), strengthening four and three seasons ahead of harvest (ASO and NDJ), and beginning to weaken as harvest approaches (FMA and MJJ). In the preplanting MJJ-1 and ASO seasons (and, in fact, throughout the six seasons leading up to harvest), these composites most closely resemble an IPRC “type 4” ENSO formation in that the cold anomalies do not stretch across the Pacific and are met near the 180th meridian by a distinctive and large expanse of warm anomalies in the west Pacific (WP).
4. Summary and discussion
In most years, a combination of both local and remote forces assures the necessary conditions for the production of both dryland and irrigated winter wheat in Kansas. However, the climatology of the Wheat State is characterized by considerable variability, which has shifted toward generally higher temperatures and may become even more volatile under anthropogenic climate change, particularly when it comes to interannual patterns of precipitation (Rahmani and Harrington 2019). Increasingly accurate predictions of bountiful and meager harvests—particularly forecasts that may inform decision-making in advance of the planting season—could substantially enhance the industry’s resilience under global warming.
While statistically significant associations exist between eventual wheat yields and concurrent-to-cultivation ocean temperature anomalies in most ENSO-associated regions, these associations are mostly weak to nonexistent in the year prior. However, the presence of a relatively strong association between far eastern Pacific Ocean temperatures a year ahead of eventual wheat yields offers the enticing prospect of identifying signals that can help inform cultivation decisions before the growing season. To this end, this study demonstrates that ENSO 1 + 2 has maintained a statistically significant correlation with eventual wheat yields in Kansas at leads of up to six seasons ahead of harvest for most of the past century, and this relationship may have strengthened in more recent decades. In both individual cases and composites, these associations exist for bountiful and meager crop years.
These early associations may in part be a condition of the ways eastern-oriented warm phases tend to produce stronger overall El Niños (Abdelkader Di Carlo et al. 2023) with resultant effects on Northern Hemisphere atmospheric blocking (McKenna and Karamperidou 2023). Whereas Kansas may typically exist in a liminal region for ENSO-related predictability, in a strong El Niño the typical area of impact for a persistent and extended jet stream and amplified storm track (i.e., the southwest and southern plains) may extend into higher latitudes to additionally encompass the Midwest. Indeed, as shown in Fig. 2b, a positive Kansas WYI is correlated with reduced upper-tropospheric pressure across the canonical Niño-affected regions of the U.S. Southwest, with resultant increased precipitation and reduced temperatures.
This analysis thus provides evidence of the differing agricultural impacts of ENSO in Kansas, suggesting that EP-flavored phases may be an area of interest for Midwestern wheat farmers, water managers, insurers, and agricultural policymakers. It is important to note, however, that correlations can be scale dependent and even associations that exist robustly at a statewide scale might be diluted or nonexistent at smaller scales as a result of local climatological, geological, and cultural factors. For example, the state’s western counties tend to have drier winters and sandier soils, whereas the central and eastern regions tend to have more precipitation and soil with higher concentrations of clay (Perkins and Schrenk 1948), and optimum planting dates and seeds planted per acre vary accordingly, while insect and disease risks may also vary by region (IMP Centers 1999). As a result, it would not be appropriate to assume that the relationships identified in this study necessarily exist in all areas of the state, and future research would be needed to understand the regions of the state that are most predictably connected to the precursive relationships identified here.
These findings offer a potential avenue of exploration for other states and other crops. Future work should further explore the integrative elements of WYI, including the qualities of precipitation (i.e., the impact of extreme rainfall), and develop dynamic models of WYI and similar yield indices in the past, present, and modeled-warming future, not just as proxies for precipitation but as reflections of deeply integrated systems informed by both local climate and remote forces and resultant changes to precipitation, pressure, wind, cloud cover, temperature, and soil moisture, permitting for even more skillful and reliable models for crop prediction at actionable lead times.
Acknowledgments.
This research is supported by the U.S. Department of Energy/Office of Science under Award DE-SC0016605 and the U.S. SERDP Project RC20-3056. SYSW was also supported by the U.S. Department of Interior, Bureau of Reclamation, with Grant R19AP00149. LD is supported by the National Natural Science Foundation of China with Grant 41875071.
Data availability statement.
The USDA Kansas Wheat History report is available at https://www.nass.usda.gov/Statistics_by_State/Kansas/Publications/Cooperative_Projects/KS-wheat-history22.pdf. ENSO-related indices were retrieved from the National Center for Atmospheric Research, https://climatedataguide.ucar.edu/climate-data/nino-sst-indices-nino-12-3-34-4-oni-and-tni. CMORPH high-resolution global precipitation estimate datasets are available at https://www.ncei.noaa.gov/data/cmorph-high-resolution-global-precipitation-estimates/access/ and have been integrated into the Climate Reanalyzer Monthly Reanalysis Correlations tool from the Climate Change Institute at the University of Maine, https://climatereanalyzer.org/reanalysis/monthly_correl/, where ERA5 data for sea surface temperatures and mean sea level pressure have also been integrated for the daily temperature, SST, and sea ice mapping tool, https://climatereanalyzer.org/reanalysis/monthly_maps/. We also employed ERA5 data from the ECMWF Copernicus Climate Data Store, https://www.ecmwf.int/en/forecasts/dataset/ecmwf-reanalysis-v5.
REFERENCES
Abdelkader Di Carlo, I., P. Braconnot, M. Carré, M. Elliot, and O. Marti, 2023: Different methods in assessing El Niño flavors lead to opposite results. Geophys. Res. Lett., 50, e2023GL104558, https://doi.org/10.1029/2023GL104558.
Amir, J., and T. R. Sinclair, 1991: A model of water limitation on spring wheat growth and yield. Field Crops Res., 28, 59–69, https://doi.org/10.1016/0378-4290(91)90074-6.
An, S. I., and F.-F. Jin, 2004: Nonlinearity and asymmetry of ENSO. J. Climate, 17, 2399–2412, https://doi.org/10.1175/1520-0442(2004)017<2399:NAAOE>2.0.CO;2.
Birka, K., R. Lupo, P. Guinan, and E. Barbieri, 2010: The interannual variability of midwestern temperatures and precipitation as related to the ENSO and PDO. Atmósfera, 23, 95–128.
Bond, N. A., M. F. Cronin, H. Freeland, and N. Mantua, 2015: Causes and impacts of the 2014 warm anomaly in the NE Pacific. Geophys. Res. Lett., 42, 3414–3420, https://doi.org/10.1002/2015GL063306.
Borhara, K., B. Fosu, and S.-Y. S. Wang, 2023: The role of the western North Pacific (WNP) as an El Niño–Southern Oscillation (ENSO) precursor in a warmer future climate. Climate Dyn., 61, 3755–3773, https://doi.org/10.1007/s00382-023-06773-z.
Capotondi, A., and Coauthors, 2015: Understanding ENSO diversity. Bull. Amer. Meteor. Soc., 96, 921–938, https://doi.org/10.1175/BAMS-D-13-00117.1.
Chiang, J. C. H., and D. J. Vimont, 2004: Analogous Pacific and Atlantic meridional modes of tropical atmosphere–ocean variability. J. Climate, 17, 4143–4158, https://doi.org/10.1175/JCLI4953.1.
Dimitri, C., A. Effland, and N. Conklin, 2005: The 20th century transformation of U.S. agriculture and farm policy. Electronic Report from the Economic Research Service, USDA, 17 pp., https://www.ers.usda.gov/webdocs/publications/44197/13566_eib3_1_.pdf.
Erenstein, O., M. Jaleta, K. A. Mottaleb, K. Sonder, J. Donovan, and H.-J. Braun, 2022: Global trends in wheat production, consumption and trade. Wheat Improvement, M. P. Reynolds and H.-J. Braun, Eds., Springer, 47–66.
Feldman, M., and M. E. Kislev, 2007: Domestication of emmer wheat and evolution of free-threshing tetraploid wheat. Isr. J. Plant Sci., 55, 207–221, https://doi.org/10.1560/IJPS.55.3-4.207.
Fosu, B., J. He, and S.-Y. S. Wang, 2020: The influence of wintertime SST variability in the western North Pacific on ENSO diversity. Climate Dyn., 54, 3641–3654, https://doi.org/10.1007/s00382-020-05193-7.
Hanley, D. E., M. A. Bourassa, J. J. O’Brien, S. R. Smith, and E. R. Spade, 2003: A quantitative evaluation of ENSO indices. J. Climate, 16, 1249–1258, https://doi.org/10.1175/1520-0442(2003)16<1249:AQEOEI>2.0.CO;2.
Iizumi, T., J.-J. Luo, A. J. Challinor, G. Sakyrai, M. Yokozawa, S. Hirofumi, M. E. Brown, and T. Yamagata, 2014: Impacts of El Niño Southern Oscillation on the global yields of major crops. Nat. Commun., 5, 3712, https://doi.org/10.1038/ncomms4712.
IMP Centers, 1999: Crop profile for wheat in Kansas. 21 pp., https://ipmdata.ipmcenters.org/documents/cropprofiles/KSwheat.pdf.
Johnson, N. C., 2013: How many ENSO flavors can we distinguish? J. Climate, 26, 4816–4827, https://doi.org/10.1175/JCLI-D-12-00649.1.
Mauget, S. A., and D. R. Upchurch, 1999: El Niño and La Niña related climate and agricultural impacts over the Great Plains and Midwest. J. Prod. Agric., 12, 203–215, https://doi.org/10.2134/jpa1999.0203.
McKenna, M., and C. Karamperidou, 2023: The impacts of El Niño diversity on Northern Hemisphere atmospheric blocking. Geophys. Res. Lett., 50, e2023GL104284, https://doi.org/10.1029/2023GL104284.
McPhaden, M. J., S. E. Zebiak, and M. H. Glantz, 2006: ENSO as an integrating concept in Earth science. Science, 314, 1740–1745, https://doi.org/10.1126/science.1132588.
Najafi, E., N. Devineni, R. M. Khanbilvardi, and F. Kogan, 2018: Understanding the changes in global crop yields through changes in climate and technology. Earth’s Future, 6, 410–427, https://doi.org/10.1002/2017EF000690.
National Center for Atmospheric Research, 2023: Nino SST indices (Nino 1 + 2, 3, 3.4, 4; ONI and TNI). Accessed 1 March 2024, https://climatedataguide.ucar.edu/climate-data/nino-sst-indices-nino-12-3-34-4-oni-and-tni.
NOAA National Centers for Environmental Information, 2022: State climate summaries: Kansas. https://statesummaries.ncics.org/chapter/ks/.
Perkins, A. T., and W. G. Schrenk, 1948: The analysis and characteristics of selected Kansas soils. Trans. Kansas Acad. Sci., 51, 215–223, https://doi.org/10.2307/3626294.
Rahmani, V., and J. Harrington Jr., 2019: Assessment of climate change for extreme precipitation indices: A case study from the central United States. Int. J. Climatol., 39, 1013–1025, https://doi.org/10.1002/joc.5858.
Rahmani, V., S. L. Hutchinson, J. M. Hutchinson, and A. Anandhi, 2014: Extreme daily rainfall event distribution patterns in Kansas. J. Hydrol. Eng., 19, 707–716, https://doi.org/10.1061/(ASCE)HE.1943-5584.0000839.
Rajagopalan, B., E. Cook, U. Lall, and B. K. Ray, 2000: Spatiotemporal variability of ENSO and SST teleconnections to summer drought over the United States during the twentieth century. J. Climate, 13, 4244–4255, https://doi.org/10.1175/1520-0442(2000)013<4244:SVOEAS>2.0.CO;2.
Shroyer, J. P., C. Thompson, R. Brown, P. D. Ohlenbusch, D. L. Fjell, S. Staggenborg, S. Duncan, and G. L. Kilgore, 1996: Kansas crop planting guide. Kansas State University, 8 pp., https://bookstore.ksre.ksu.edu/download/kansas-crop-planting-guide_L818.
Stuecker, M. F., 2018: Revisiting the Pacific meridional mode. Sci. Rep., 8, 3216, https://doi.org/10.1038/s41598-018-21537-0.
Tian, Y., C. D. Peters-Lidard, B. J. Chaudhury, and M. Garcia, 2007: Multitemporal analysis of TRMM-based satellite precipitation products for land data assimilation applications. J. Hydrometeor., 8, 1165–1183, https://doi.org/10.1175/2007JHM859.1.
Trenberth, K. E., and T. J. Hoar, 1997: El Niño and climate change. Geophys. Res. Lett., 24, 3057–3060, https://doi.org/10.1029/97GL03092.
Trenberth, K. E., and D. P. Stepaniak, 2001: Indices of El Niño evolution. J. Climate, 14, 1697–1701, https://doi.org/10.1175/1520-0442(2001)014<1697:LIOENO>2.0.CO;2.
U.S. Department of Agriculture, 2022: Kansas wheat history. 42 pp., https://www.nass.usda.gov/Statistics_by_State/Kansas/Publications/Cooperative_Projects/KS-wheat-history22.pdf.
Wiedermann, M., J. F. Siegmund, J. F. Donges, and R. V. Donner, 2021: Differential imprints of distinct ENSO flavors in global patterns of very low and high seasonal precipitation. Front. Climate, 3, 618548, https://doi.org/10.3389/fclim.2021.618548.
Xu, K., C.-Y. Tam, C. Zhu, B. Liu, and W. Wang, 2017: CMIP5 projections of two types of El Niño and their related tropical precipitation in the 21st century. J. Climate, 30, 849–864, https://doi.org/10.1175/JCLI-D-16-0413.1.
Yu, J.-Y., and Y. Zou, 2013: The enhanced drying effect of central-Pacific El Niño on US winter. Environ. Res. Lett., 8, 014019, https://doi.org/10.1088/1748-9326/8/1/014019.
Yu, J., X. Wang, S. Yang, H. Paek, and M. Chen, 2017: The changing El Niño–Southern Oscillation and associated climate extremes. Climate Extremes: Patterns and Mechanisms, Geophys. Monogr., Vol. 226, Amer. Geophys. Union, 3–30.
Zeng, R., F. Yao, S. Zhang, S. Yang, Y. Bai, J. Zhang, J. Wang, and X. Wang, 2021: Assessing the effects of precipitation and irrigation on winter wheat yield and water productivity in North China Plain. Agric. Water Manage., 256, 107063, https://doi.org/10.1016/j.agwat.2021.107063.
Zhang, H., A. Clement, and P. Di Nezio, 2014: The South Pacific meridional mode: A mechanism for ENSO-like variability. J. Climate, 27, 769–783, https://doi.org/10.1175/JCLI-D-13-00082.1.
Zhang, T., M. P. Hoerling, A. Hoell, J. Perlwitz, and J. Eischeid, 2020: Confirmation for and predictability of distinct U.S. impacts of El Niño flavors. J. Climate, 33, 5971–5991, https://doi.org/10.1175/JCLI-D-19-0802.1.
