Influence of seasonal climatic water deficit and crop prices on rainfed crop grain harvest, repurposing, and abandonment in the western U.S.A.

Increasing climate aridity and drought, exacerbated by global warming, are increasing risks for western United States of America (U.S.A.) rainfed farming, and challenging producers’ capacity to maintain production and profitability. With agricultural water demand in the region exceeding limited supplies and fewer opportunities to develop new water sources, rainfed agriculture is under increasing pressure to meet the nation’s growing food demands. This study examines three major western U.S.A. rainfed crops: barley, spring wheat, and winter wheat. We analyzed the relationship between crop repurposing (the ratio of acres harvested for grain to the total planted acres) to seasonal climatic water deficit (CWD). To isolate the climate signal from economic factors, our analysis accounted for the influence of crop prices on grain harvest. We used historical climate and agricultural data between 1958 and 2020 to model crop repurposing (e.g. forage) across the observed CWD record using a fixed effect model. Our methodology is applicable for any region and incorporates regional differences in farming and economic drivers. Our results indicate that farmers are less likely to harvest barley and spring wheat for grain when the spring CWD is above average. Of the major winter wheat growing regions, only the Northern High Plains in Texas showed a trend of decreasing grain harvest during high CWD. For the majority of major crop growing regions, grain prices increased with lower levels of grain harvest. Interestingly, winter wheat repurposing is significantly higher in the southern Great Plains (∼50% harvested for grain) compared to the rest of the West (∼90%). Our results highlight that the major barley and spring wheat regions’ grain harvests are vulnerable to high spring CWD and low summer CWD, while winter wheat grain harvest is unaffected by variable CWD in most of the West.


Introduction
Western United States of America (U.S.A.) rainfed crop farmers have a long history of adapting to variations in local climatic conditions in order to maximize revenue while minimizing risk [1][2][3][4][5]; however, ongoing and notable changes in the climate increases the risk to dryland farming production [6,7]. By 2050, food production will need to increase to feed the increasing global population [8,9], and despite U.S. farmland acreages declining roughly 6% since 1997 [10], food production has increased by using fertilizers, better crop varieties and agronomic practices, and increasing irrigation [9,[11][12][13]. Irrigated crops are less sensitive to climate-induced water stress [13,14], but water demands in the West are already exceeding available water supplies and there are few opportunities to develop new water sources; therefore, rainfed agriculture is important in meeting food demands.
Despite the increase in irrigated land during the past century, rainfed agriculture continues to dominate planted acreage in the western U.S.A. According to the U.S. Department of Agriculture National Agricultural Statistics Survey (USDA NASS), three annual rainfed crop types, barley, spring wheat and winter wheat, comprise approximately 50% of the cropland in the West and are critical to the U.S. agricultural economy. Western U.S.A. barley sales exceeded $500 million in 2017, and $1 billion in 2012 [10]. Barley is mainly grown in Idaho, Montana, and North Dakota. Wheat is more economically important, as western U.S.A. sales exceeded $5.5 billion in 2017, and $11 billion in 2012 [10]. Wheat is one of the most widely cultivated grain crops in the world and contributes approximately 20% of the world's total calories [15]. The U.S. ranks fifth for wheat production and third for wheat exports [16]. Roughly 80% of U.S. wheat production occurs west of the Mississippi River, and approximately 80% of planted wheat is winter wheat [10]. Spring wheat is grown in the Pacific Northwest (PNW) and northern Great Plains (NGP), while winter wheat is grown throughout the West, with higher acreages in the PNW and the central and southern Great Plains (CGP and SGP, respectively).
Meeting food demands involves harvesting a crop for grain, but crops can be repurposed as livestock feed. During the growing season, farmers have to decide their end goals to maximize revenue, i.e. managing for and harvesting for grain, or repurpose the crop by either cutting and baling hay, using the field for livestock grazing, and/or abandoning the crop unharvested. Their choice influences management factors, such as agronomic inputs and coordinating harvesting plans, all of which affect their net revenue. Farmers have a relative idea of how a crop will yield based on prior experience with the land, weather, and inputs, but farmers also need to consider the grain market, their ability to market and sell hay, and how their decision will impact their fields' soil health for subsequent crops [17]. Farmers adapt their decisions to interannual variability in weather and markets, but due to climate change, increasingly hot and dry conditions are impacting crop production [7,18,19]. There remain questions about what water deficit levels or change in market prices may trigger farmers to abandon their grain harvest and repurpose planted rainfed crops. Moreover, an increasing frequency or severity of crop water deficits may be indicative of a regional climate threshold towards more persistent crop grain failure, forcing farmer adaptations.
While many studies have addressed the influence of precipitation and temperature on crop productivity and yields, less attention has been paid to the climatic water deficit's (CWD's) impact on crop repurposing or abandonment [6,7,13,14,[19][20][21]. We use CWD, the difference between potential evapotranspiration (ET) and actual ET [22], to measure crop climate stress. We used CWD because it represents the evaporative demand that is not met by available water, it is often more representative of ecologically-significant changes than temperature or precipitation [18], and because it is a conceptually useful, plant-centric metric to define anomalously humid or arid conditions [22]. Understanding how farmers respond to changes in CWD can highlight regions where rainfed crops could be bolstered or threatened by a changing climate. Non-climatic factors, such as grain and hay prices, can also influence farmer decisions to harvest a cereal crop for grain, and confound farmer sensitivity to climate thresholds. For this reason, climatic and market factors need to be considered simultaneously to understand and predict climate thresholds that may trigger abandoning crop grain harvest in favor of a more economical purpose.
In this study, we pose three hypotheses. (1) Because farmers have a history of adapting to variations in local climate conditions, we hypothesize that rainfed crop farmers have historically optimized planted area to normal CWD conditions in their region, instead of optimizing planting to the conditions of more favorable years, in order to maximize revenues while minimizing risk. (2) Because research has shown the importance of crop seedling health to crop production [21,23,24], we hypothesize that historically anomalous CWD during the early growing season had relatively more influence on crops harvested for grain than anomalous CWD during the late growing season. (3) Given that rainfed winter wheat is produced more widely across the western U.S.A.
[10], it will be less sensitive to market forces than barley or spring wheat. Using county-level USDA surveys of rainfed planted and harvested acreages, crop prices, and downscaled historical climate datasets, we tested our hypotheses using two methods: (1) plotting the ratio of acres harvested to acres planted, i.e. the harvested to planted ratio (HPR), as a measure of crop repurposing over the available historical record (1958-2020) and observed CWD anomaly range; and (2) a multiple linear fixed effect regression (MLR) model to analyze how seasonal CWD and changes in crop prices between planting and harvest influence the HPR. While a farmer decision to harvest, repurpose, or abandon a crop involves numerous factors besides seasonal CWD and market prices, such as participation in the Federal Crop Insurance Program, extreme weather events (e.g. spring frosts, hail, high winds, floods), diseases and insects, personal finances, soil health, and many more, our goal was to identify if seasonal CWD and/or change in crop prices influence grain harvests of barley and wheat. Analyzing the regression coefficients associated with the CWD and grain market prices can shed light on the sensitivity of crop repurposing across the western U.S.A. due to climate variations.

Methods
We used downscaled historical observed data to quantify the impact of climate and market prices on crop abandonment over the western U.S.A. The USDA NASS [16] provides annual, county-level rainfed crop planted and harvested acreages and monthly, state-level crop prices received for the three rainfed crops analyzed: barley, spring wheat, and winter wheat. Daily CWD was obtained from TerraClimate, a 4 × 4 km gridded monthly surface climate and water balance dataset [20]. The TerraClimate dataset extends from 1958 to the current date. Thus, the analysis in this study is confined to 1958-2020.
The USDA NASS agricultural data are recorded at the county-level; however, the analysis between CWD and HPR was performed at the USDA Agricultural Statistics District (ASD) spatial scale (figure 1, and table A1, appendix A in supplementary material) to increase statistical power since some counties do not have sufficient data to allow robust inference. For the analysis, county-level data was pooled in a single ASD level dataset, and the gridded CWD data was aggregated by spatially averaging the gridded climate data [25] within each ASD area using a threemonth, seasonal time scale; therefore, the daily CWD values were aggregated for the spring season (April-June), summer season (July-September), and fall season (October-December). The wide seasonal windows used in this analysis capture the peak planting and harvesting times in all ASD regions, i.e. the course temporal resolution is insensitive to the differences in the growing season period between regions.
The CWD and agricultural data time series from each ASD were detrended prior to analysis. CWD data was subsequently transformed into standardized anomalies (z-score). HPR data, which are bound between 0 and 1, were transformed using the logit transformation. By using county-level acreage data and then aggregating to the ASD level, we increase the number of datapoints for a more robust statistical analysis, but reduce biases resulting from variations in soil type, soil quality, and landscape topography [14]. Using these transformed datasets we plotted the change in HPR over the seasonal CWD anomaly range using the LOWESS method [26,27].
We calculated the percent change in crop prices between the earliest usual planting month and the latest usual harvesting month. The USDA NASS provides the usual field crop planting and harvesting dates for each crop and each state. For barley and spring wheat, March is the earliest planting month and September is the latest harvesting month. For winter wheat, September is the earliest planting month and July of the following calendar year is the latest harvesting month [28]. Percent change in prices received were transformed into standardized z-score anomalies.
The percent change in crop prices received from planting to harvesting was used in combination with CWD using a MLR model (equation (1)) where HPR is the response variable for crop i, season j, and ASD k. α is the fixed effect intercept for crop i, season j, and ASD k; β1 is the coefficient of CWD for season j, and ASD k; and β2 is the coefficient of the percent change in price received for crop i, and ASD k. The coefficients represent the relative weight each factor has on the HPR. A negative β1 indicates decreased HPR (i.e. higher crop repurposing) with increased CWD (i.e. drier conditions). A negative β2 indicates decreased HPR when the associated crop price increased between planting and harvesting. As presented in the Results, the α values were untransformed and indicate the baseline HPR if there were no CWD or price change anomalies. Given that the inputs are normalized and that CWD and prices received are uncorrelated (figure A1, appendix A in supplementary material), the MLR coefficients can be interpreted as the fraction of the HPR variance explained by each factor, and therefore permit to analyze in which region each factor is significant in affecting farmer decisions to repurpose crops. The relationship between seasonal CWD and crop prices received to annual HPR was evaluated for all ASDs that have any USDA NASS data for the selected crop types between 1958 and 2020. In the Results section below, we include ASDs that currently devote large acreages to each crop type [29]. The CWD versus HPR plots for all crops and all ASDs are presented in the supplementary material.

Impacts of anomalous seasonal CWD on HPR using LOWESS
Currently, Idaho east (ID-e), Montana north central (MT-nc), and North Dakota northeast (ND-ne) ASDs (figure 1) account for over half of the total acres dedicated to barley in the western U.S.A [29]. As to be expected, considering drought negatively impacts crop production, barley repurposing increased in these regions during positive CWD anomalies (drier years). This was observed during both spring and summer, as shown by the negative slope of the LOWESS curves on the positive CWD anomaly scale (figure 2), although spring was more impactful. The  The LOWESS curve highlights the nonlinear behavior of HPR over the CWD anomaly range. The inflection points of the spring LOWESS curves occur within one standard deviation (SD) of normal for the majority of these barley and spring wheat regions (figures 2(a), (c), (e), and 3(a), (c), (e)). The inflection points of the summer LOWESS curves are less definitive; however, for regions where crop repurposing occurred, either during high or low CWD anomalies, the inflection point is within ±1 SD.
Winter wheat is currently grown across the western U.S.A. and in markedly different climatic regions; the PNW, California's Central Valley, the CGP and SGP, and the Northern Rocky Mountains [29]. Twelve of the most heavily planted regions, spanning the different climatic regions are presented in figure 4. Unlike barley or spring wheat, winter wheat HPR was more stable across the CWD anomaly range for all regions except the regions in the SGP (Texas northern high plains (TX-nhp), Texas northern low plains (TX-nlp) (figures 4(a) and (b), and Oklahoma southwest (OK-sw) (figures 4(c) and (d)). In these regions, HPR decreased with positive CWD anomalies during both fall and spring seasons, with the Texas regions more so than OK-sw. Appendix D in the supplementary material contains the LOWESS plots for any ASDs with winter wheat data from 1958-2020.
Winter wheat harvest most commonly occurs in mid-June in Texas, but occurs in early August in Montana [28]. Given our results, and this spatiotemporal variability in usual harvesting dates across the West, we analyzed if June CWD impacted the LOWESS curve in TX-nhp by preforming the LOWESS regression using average March, April, and May CWD data, and also average April and May CWD data. Omitting June, nor adding March CWD data significantly changed the LOWESS curve (appendix D, figure D1).

Effect of anomalous seasonal CWD and change of crop prices received on HPR using MLR
The resulting MLR coefficients (β1 and β2) for the explanatory variables (CWD and percent change in price received, respectively) are mapped in figure  Regarding winter wheat, negative fall β1 values dominated most of the study region ( figure 7(a)). Of the seven regions currently growing the most winter wheat, MT-nc, CO-ec, KS-sc, and TX-nhp showed a significant relationship between HPR and CWD (figures 7(a) and (d)). The PNW and eastern portions of the CGP and SGP regions showed positive β1 values (figures 7(a) and (d)), indicating that too much moisture results in more repurposing.
Conversely to CWD, barley and spring wheat were more sensitive to summer percent change in price (figures 5(b), (e), and 6(b), (e), respectively). The negative β2 values indicate the price increased between planting and harvesting when HPR decreased. This apparent contradiction can be explained by the effects of supply and demand imbalances. Barley and spring wheat are not as widely grown across the West, therefore grain loss can significantly influence market prices. Winter wheat, however, is more widely cultivated, buffering changes in market prices from increased grain loss in certain regions. This is represented in our results shown by the pale colors in figures 7(b) and (d), indicating that commodity price change had little influence on HPR across the West.
The HPR baseline maps (figures 5(c), (f), 6(c), (f), and 7(c), (f)) indicate that farmers harvested the vast majority of planted acres for grain when there were no CWD or crop price anomalies. The barley and spring     wheat HPR baseline values were greater than 90% for both seasons across the majority of the study region (figures 5(c), (f), and 6(c), (f)). However, the most dramatic MLR results were the low winter wheat HPR baseline values in New Mexico, Texas, and southern Oklahoma. For example, TX-nhp had a baseline HPR of 57%.

Discussion and conclusions
Given our results, above average CWD conditions trigger a threshold on farmer decisions to repurpose planted rainfed barley, spring wheat, and winter wheat; however, winter wheat farmers outside of the SGP were largely unaffected by variable CWD. Shown by the LOWESS curves (figures 2 and 3), western U.S.A. farmers have adapted to normal CWD conditions within their respective regions, planting acreages to maximize grain returns under their regional climate normals; therefore, allowing them to maximize revenue while minimizing risk. That being said, farmers consider many factors beyond CWD; including but not limited to, personal finances, traditional knowledge and instincts, recent weather and soil conditions, social and community influence, market conditions, and crop insurance participation [20,30,31]. In addition, the use of agronomic inputs has reduced crop failure risks since 1958 [32], and agricultural scientists have been developing more drought resistant crop varieties [33]. With all these factors it is extremely difficult to identify a purely climatic effect on observed data [34]. We found that the spring season had more influence on barley and spring wheat repurposing, while the fall season had relatively more influence on winter wheat, but only in the SGP. Among the major barley and spring wheat growing regions, only ID-e experienced an increasing (drying) trend in spring CWD between 1958-2020 (+0.28 mm yr −1 (figure 8)). Planted rainfed barley acreages have been declining in this region since the late 1980s which coincides with spring CWD more often surpassing normal (figure 8). Strictly considering our results, if these trends continue then barley grain harvest would decrease in ID-e.
Spring, for barley and spring wheat, and fall, for winter wheat, are the seasons of planting and emergence of the crop. Inadequate moisture can retard seed germination or dry out newly emerged seedlings [21,23,24], heightening the risk of crop failure or poor grain filling during unfavorable CWD conditions, which increases the chance that farmers abandon or repurpose the crop. For barley and spring wheat, summer is the most arid season in the West; therefore, we did not expect a strong influence on crop repurposing. This was supported by our results (figures 2(b), (d), (f), and 3(b), (d), (f)). By July in the northern Rockies and NGP, barley and spring wheat are mature with full leaves to shade the soil, preventing evaporation, and they have fully developed root systems to access soil moisture. A counter point is that a crop with more biomass has a greater demand for transpiration [35]; however, under drought conditions stomatal conductance decreases and transpiration efficiency increases [36].
There are numerous unfavorable conditions that can negatively affect plant health and trigger crop abandonment or repurposing, including excessive moisture and water logged soils [37]. Our LOWESS results show decreasing HPR at negative CWD anomalies in eastern North Dakota (figures 2(f) and 3(f), and Appendices B and C in supplementary material). Soils in eastern North Dakota are highly susceptible to waterlogging [38]. Other issues outside the scope of this research include excessive cool temperatures (low growing degree days) hindering plant growth [39], and/or damaging spring frosts [40]. Both of these factors may have heightened the impact of the CWD on HPR.
While farmers have adapted to some degree to the effects of climate change and increasing climate extremes represented within our multi-decadal study record , the magnitude of historical climate variability may be less extreme than projected, end-of-century climate conditions [41]. Under projected warmer winter temperatures, farmers may transition from spring wheat to winter wheat to exploit more favorable conditions for growth [42].
Our results indicate that winter wheat is more stable across the CWD anomaly range throughout the western U.S.A., except in the SGP (figures 4(a)-(d), and 7(a), (d)). The Texas High Plains region is key for U.S. winter wheat production and is primarily rainfed [43]. A marked result in our study is the dramatically different winter wheat baseline HPR in the SGP compared to the rest of the West (figures 7(c) and (f)). One reason for this result is the common use of winter wheat for cattle grazing [44]. In much of the SGP, favorable conditions, including warmer winter temperatures and few snow-covered days, allow cattle to forage on winter wheat when other forages are limited [39,45]. Weather related damage, such as from wind and hail, are also common in the SGP [46]. Other issues that have affected winter wheat in the SGP include vernalization and leaf rust. Vernalization issues can be caused by mild winters that reduce the number of chilling hours required (<45 • F) by winter wheat to switch from vegetative to reproductive growth [47]. Leaf rust is a disease and survives during mild winters [47]. Vernalization issues and leaf rust may become more persistent in the SGP and areas affected may expand north with climate change as winter temperatures increase [48].
From an economic perspective, the MLR percent change in price coefficient results were as anticipated for the majority of the regions: when HPR decreases, i.e. an increase in crop repurposing, the price received for the commodity increases by the time of harvesting. However, for winter wheat in MT-nc and CO-ec the MLR percent change in price coefficient is positive, implying that grain harvesting increased when prices increased by the time of harvesting. This may be a result of the spatiotemporal nature of harvesting times, i.e. the earlier harvest time in the SGP impacts Colorado and Montana farmers' decisions on crop repurposing. Since 1958 neither of these regions have had a significant change in the area devoted to rainfed winter wheat; however, OK-nc and TXnhp rainfed winter wheat planted acreage have been declining since the 1970s [16]. The change in price coefficient for barley in ID-e is also positive, but since the late 1980s the area devoted to rainfed barley in this region has been declining [16]. The area devoted to irrigated barley in ID-e has remained constant since the mid-1980s; therefore, our results could indicate that the land currently used for rainfed barley in IDe is the most ideal for producing high yields. Many factors can influence crop prices and crop abandonment, which go beyond the analysis presented here. However, given our results, winter wheat grain harvesting north of Oklahoma appears to be more resilient to not only CWD, but also market forces. Addressing our hypotheses, (1) the barley and spring wheat LOWESS curves provide evidence that farmers have optimized planted area to their regional normal CWD conditions. Across Montana and North Dakota, more arid spring seasons had significant impact on crop repurposing, and more humid summer seasons decreased HPR in central and eastern North Dakota. In some regions, normal spring and summer CWD may have slightly decreased grain harvest, but the risk associated with greater than one SD away from normal CWD was significant. For barley and spring wheat, positive (arid) spring CWD anomalies were more impactful than arid summer CWD anomalies; however negative (humid) summer CWD anomalies in central and eastern North Dakota reduced HPR. (2) Seedlings and newly emerged crops are more sensitive to climatic anomalies than mature crops. While arid spring conditions did impact winter wheat in the SGP, arid fall seasons were more detrimental to grain harvest. Therefore, CWD during the planting season for all three crops was more significant to overall HPR. Finally, (3) winter wheat was less sensitive to market forces than barley or spring wheat. It may be a result of winter wheat being grown throughout the West, but our results also showed that winter wheat HPR was less sensitive to CWD anomalies; therefore, the winter wheat grain market is elastic, especially compared to barley or spring wheat.

Data availability statement
No new data were created or analysed in this study.