The DSSAT is a suite of more than 42 crop models. It can simulate crop growth and development processes as defined by the soil-plant-atmosphere dynamics, using various tools that manage databases on crop, soil, and weather and several applications that perform graphical display, seasonal analysis, rotational analysis, and genotype coefficient estimation [41] . The DSSAT suite has been used for various purposes such as precision crop management and studying agroecosystem sustainability, climate change impacts, and greenhouse gas emission [41] . Simulations are performed on a daily basis by integrating crop, soil, weather, and management data with crop models and application programs.

For rapid simulation, inspection, and analysis of results of long-term cropping sequences, DSSAT contains the Sequence Analysis tool [42] . As multiple cropping seasons are involved in a sequence analysis, this tool allows for the carryover of soil water and nutrients from the preceding crop to the following crop [43] [44] .

Llano Estacado is a semi-arid region in the southern US and comprises parts of eastern New Mexico and northwestern Texas [45] [46] . The economy of this region is predominantly agricultural, with farming of various crops prevalent. The overuse in the past of the Ogallala Aquifer, the main freshwater source for the region, has persuaded some farmers to return to dryland crops. The Texas A&M AgriLife Research & Extension Center at Amarillo is situated in this region. At the Amarillo Center, numerous experiments have been conducted to investigate, discover, develop, evaluate, and apply technology to sustain livestock and crop production in the Texas Panhandle region and beyond. In this study, Amarillo (35.19˚N, 102.06˚W), Texas was used as a representative site for the Llano Estacado region [47] .

To explore the effect of interannual climate variability on cowpea and winter wheat grain yields in the Llano Estacado region, a long-term weather dataset spanning 81 years (1942-2022) was used. Historical daily data on precipitation, temperature, and windspeed at Amarillo were obtained from the website of National Centers for Environmental Information [48] ; whereas those on solar radiation were generated using a reliable irradiation model described by [49] .

Pullman clay loam (Torrertic paleustolls) is a primary soil used for agricultural purposes in Llano Estacado. Thus, this soil was used as a representative soil for the study region [50] . The soil data (Table 1) were obtained from the Gridded Soil Survey Geographic (GSSURGO) database of the USDA NRCS [51] [52] . The run-off curve number and the drainage coefficient of the soil were 81 and 0.60, respectively.

The grain yields of cowpea and winter wheat in the three cropping systems, namely fallow-cowpea, fallow-wheat, and cowpea-wheat under double cropping, were simulated using the DSSAT Sequence Analysis tool. The cowpea or wheat crop that followed fallow, known as the mono-cropped cowpea or mono-cropped wheat, hereafter, will be referred to as ^mcowpea and ^mwheat, respectively, and the wheat crop that followed cowpea, known as the double-cropped wheat, as ^dwheat. A total of 94 scenarios were simulated that comprised four planting dates for cowpea (June 1, June 15, July 1, and July 15), six planting dates for wheat (September 15, September 30, October 15, October 30, November 15, and November 30), and three N application rates to wheat (0, 50, and 100 kg N ha⁻¹) (Table 2).

For simulations, Pullman clay loam was used as soil, and “Newton” and “Cal #5 MG4” were used as cultivars for winter wheat and cowpea, respectively. For Newton, the genetic coefficients already estimated for the study region by [53] were used. For Cal #5 MG4, the default genetic coefficients provided in the standard DSSAT release [36] that correspond to the coefficients upon which the

^†TN: Total N; OC: Organic Carbon; FC: Field Capacity; WP; Wilting Point; WH: Water Holding capacity.

cowpea model was adapted were used. For each scenario, simulation started on April 1, two months before the earliest cowpea planting date of June 1 in 1942, and terminated on the harvest date associated with the latest planting date of wheat in 2022. For simulations, 30 plants m⁻² for cowpea and 323 plants m⁻² for wheat were assumed. Dry seeds were planted in rows at 3 cm depth using the conventional tillage.

Only wheat crops received N fertilizer. Of the total quantity of N set for application, one half was applied at planting and the other half on February 15 of the following year. To let the nutrients in stover residues transfer from the preceding crop to the following crop in cycles, the residues of each crop were assumed to be automatically incorporated into the soil on the harvest day of the crop. The “Century” method in the DSSAT system was assumed for organic matter estimation, with “Cultivated, good management, initial default SOM” as the five years’ field history [54] .

For ENSO analyses, the grain yields of ^mcowpea, ^mwheat, and ^dwheat each that were simulated for each of 81 seasons (1942-2022) were assigned to a specific ENSO phase as categorized by the Japan Meteorological Agency (JMA) index [55] [56] [57] [58] . The JMA index is a 5-month running average of the sea surface temperature anomalies over the tropical Pacific (4˚S - 4˚N, 150˚W - 90˚W). An ENSO year, which starts from October through the following September, is categorized as El Niño, La Niña, or Neutral if the index values are ≥0.5˚C, ≤−0.5˚C, or between −0.5˚C and 0.5˚C, respectively, for 6 consecutive months, including October, November, and December [56] [58] . The JMA index was chosen for ENSO characterization as it selects the known ENSO events better than other similar indices [58] . According to this index, the total number of years under El Niño, La Niña, and Neutral phases during the 1942-2022 period were 18, 21, and 42, respectively.

Statistical significance tests were performed to examine yield differences across ENSO phases as influenced by cowpea planting date for ^mcowpea, as influenced by wheat planting date × N application rate interactions for ^mwheat, and as influenced by cowpea planting date × wheat planting date × N application rate interactions for ^dwheat. The tests were carried out using the pairwise Wilcoxon rank sum test [59] , a nonparametric alternative to the two-sample t-test, as the assumption of normality was not met for each Analysis of Variance (ANOVA) test. For statistical analyses, the R software environment (R version 4.1.1) was used (https://www.r-project.org/).

To assess the status of crop water stress during the cowpea and wheat growing seasons, Agricultural Reference Index for Drought (ARID), an agricultural drought index that is simple and sound [60] , widely applicable [61] , able to predict yield loss from drought for several field crops [62] , and applicable to drought forecasting [63] was used. Using data on soil and weather, daily values of ARID during cowpea and wheat growing seasons were computed and used to associate drought with yields.

The simulated, detailed response of ^mcowpea yields to ENSO phases as influenced by planting date is presented in Figure 1. The results showed that the impact of ENSO on ^mcowpea yields in the Llano Estacado region was significant only for the planting date of July 15 (Table 3). For this planting date, the ^mcowpea yields under the El Niño phase were significantly greater than those under the La Niña phase. At all other planting dates, however, the ^mcowpea yields were about the same across all ENSO phases. These results were likely because the status of crop water stress during early June through the first week of October, the cowpea growing seasons associated with most planting dates, was about the same across all ENSO phases (Figure 2). This indicated that the amounts of water taken up by the crops associated with these planting dates were not significantly different across ENSO phases. The ^mcowpea crops associated with the July 15 planting date, on the other hand, received significantly more precipitation and thus had less water stress under El Niño than under La Niña especially during the first week of October through the end of this month (Figure 2).

Although the ENSO effect on ^mcowpea yields was significant only for the July 15 planting date, the El Niño years tended to produce more cowpea yields than did the La Niña years, especially for the crops planted in July (Table 3, Figure 3). The ^mcowpea yields associated with the crops planted in July were about 10% greater in the El Niño years compared with the normal yield, an average yield of the entire 81 seasons (Figure 3(a)). These yields in the La Niña years, on the other hand, were about 20% less than the normal yield. Under the El Niño phase, the yields of cowpea crops planted in July were about 30% greater than those produced under the La Niña phase (Figure 3(b)).

These results were in agreement with those found for Piney Woods, a humid vegetational region in the southern US [32] . In that region, irrespective of soil type and N application rate to wheat, the cowpea yields under El Niño years were significantly greater than those under La Niña. However, a slight discrepancy between those results and the results from this study was that in the Piney

^†Means followed by the same letter across ENSO phases (horizontally) within a planting date are not significantly different at α = 0.1.

Woods case the greater yields under El Niño relative to La Niña occurred even with the crops planted in June; whereas in the case of Llano Estacado, the greater yields under El Niño were associated only with the crops planted in July, not June. This difference indicated that the ENSO signal during the cowpea growing season in the Llano Estacado region was weaker than that in the Piney Woods region. In fact, the ENSO signal in the southeastern US is strongest in the southernmost part of the region and gradually weakens toward the north [28] [63] . Moreover, the northern parts of the region, especially those along the I-40 corridor, have been found to have very weak or no ENSO effects during the summer [24] [25] .

Figure 4 shows the simulated, detailed response of ^mwheat yields in the Llano Estacado region to ENSO as influenced by planting date × N application rate.

The results showed that at all planting dates the impact of ENSO on ^mwheat yields in this region was significant for the N rates of 50 and 100 kg·ha⁻¹ (Table 4). For these N rates, the ^mwheat yields under the El Niño phase were significantly greater than those under the La Niña phase. At the zero N rate, however, the ^mwheat yields at any planting date were not significantly different across the three ENSO phases.

The greater yields under the El Niño phase at the N rates of 50 and 100 kg·ha⁻¹ were likely because the amount of precipitation during October through May in the following year, the winter wheat growing season associated with the planting dates studied, was greater under El Niño than under La Niña (Figure 5(a)). That is, the crop water stress during this period was less under the El Niño phase (Figure 5(b)). The amount of precipitation during the growing season associated with each planting date was greater under the El Niño phase (Figure 6(a)). Accordingly, the crop water stress during the growing season for each planting date was lower during this phase (Figure 6(b)), indicating that the amount of water taken up by the crops at each planting date was different across ENSO phases.

Even with significantly different amounts of precipitation across ENSO phases, similar yields in all ENSO phases at zero N rate were due to low inherent fertility level of the soil. As Table 1 shows, the total N and organic C contents of the soil were 0.07% and 0.67%, respectively. Since the plant production condition was N-limited under no N fertilizer application, the difference in precipitation across the ENSO phases could not lead to significantly different yields. The results indicated that under N-limited conditions the water use efficiency would be low, and thus the ENSO effect would not be evident. Indeed, when soil N level is low, it is the supply of N, not of water, that determines the grain yields; and, conversely, when soil N level is not low, it is the supply of water that determines the grain yields [64] . A higher N rate under high soil moisture condition increases biomass and grain yield by enhancing leaf area index and photosynthetic rate; that is, N fertilization enhances the amount of water extracted by a crop, and thus the water use efficiency [65] . Water and N are two vital factors influencing N uptake and utilization. While the former increases yields mainly through N

^†WPD: Wheat Planting Date; E: El Niño; L: La Niña; N: Neutral. ^‡Means followed by the same letter across ENSO phases (horizontally) within an N rate-planting date combination are not significantly different at α = 0.1.

productivity, the latter enhances yields through water productivity [66] . There is a significant synergistic relationship between crop water productivity and N use efficiency [67] . At the N rate of 50 kg·ha⁻¹ or higher, however, N was not limiting. Thus, water use efficiency at these N rates was not restricted. The El Niño phase received more precipitation during the winter wheat growing season compared with the La Niña phase (Figure 5, Figure 6), thus leading to a higher soil water content, which in turn led to a higher N use efficiency. A higher soil N content, on the other hand, led to a higher water use efficiency. Because of these higher efficiencies, wheat yields under the El Niño phase at 50 kg N ha⁻¹ or

higher were significantly greater than those under the La Niña phase.

Although the ENSO effect on ^mwheat yields was significant only for the N application rates of 50 and 100 kg·ha⁻¹, the El Niño years tended to produce higher yields relative to the La Niña years for all systems comprising all the six planting dates x three N rates (Table 4, Figure 7). Irrespective of the planting date, the ^mwheat yields associated with the N rates of 0, 50, and 100 kg·ha⁻¹ were approximately 5%, 20%, and 27% greater, respectively, in the El Niño years compared with the normal yield, an average yield of the entire 81 seasons (Figures 7(a)-(c)).

In the La Niña years, on the other hand, the ^mwheat yields at 0, 50, and 100 kg N ha⁻¹ were about 5%, 15%, and 20% less than the normal yield, respectively, regardless of the planting date. Compared with the La Niña phase, the ^mwheat grain yields under the El Niño phase were greater by about 8%, 40%, and 60% at the N rates of 0, 50, and 100 kg·ha⁻¹, respectively, irrespective of the planting date (Figure 7(d)). These results demonstrated that the ENSO impact on ^mwheat yields in the Llano Estacado region would be greater with an increase in N application rate from 0 to 100 kg·ha⁻¹ due to an increase in water use efficiency.

Statistically, the ENSO difference across planting dates in terms of yield departure from the normal yield was not significant (Figure 7). However, the difference tended to be greatest for November 15 planting date, especially at 50 kg N ha⁻¹ or less. This was probably because the growing season associated with this planting date fell in peak winter when ENSO signal in the southern US is the strongest of all seasons.

The results regarding the greater wheat yields during the El Niño phase relative to La Niña in the Llano Estacado region were in agreement with those observed by reference [33] . They [33] evaluated ENSO effects on wheat yields in the Texas High Plains, which lies in the Llano Estacado region, using about 19

years’ field-observed wheat yield data and found that the yields under the El Niño phase were about 55% greater than those under the La Niña phase. The greater yields under El Niño, relative to La Niña, that reference [33] as well as this study found in the Llano Estacado region were mainly due to precipitation and temperature, key weather variables determining growth, development, and yields of wheat. The precipitation difference during wheat seasons across ENSO phases and its effects on yields have been previously discussed (Figure 5, Figure 6). Regarding the temperature effects, due to lower temperatures during wheat growing seasons, the El Niño phase years generally resulted in greater yields. The La Niña phase years, in contrast, had lower yields due to higher temperatures during the growing seasons. Higher temperatures led to a shorter growing season, a shorter grain-filling period, lack of vernalization, and increased leaf senescence [26] .

Our finding of the greater yields under El Niño for the semi-arid region of Llano Estacado, however, did not agree with those found for the humid region of Piney Woods [32] . In the Piney Woods case, the grain yields of winter wheat were greatest under the La Niña phase. The likely reasons for this inconsistency were as follows. Llano Estacado is a semi-arid region that received an 81-year average precipitation of about 29 mm, roughly one-fourth of that received by the humid region of Piney Woods (116 mm), during the winter wheat growing season of October through May. As a dryland, therefore, Llano Estacado had water-limited growing conditions. Accordingly, wheat crops in this region were more sensitive to water than those in Piney Woods. Thus, an El Niño year in Llano Estacado, relative to a La Niña year, produced proportionately more yields than did an El Niño year in the Piney Woods region. Moreover, the ENSO signal in the northern part of the southern US is weaker than that in the southern part [28] [63] . Wheat shows a strong yield enhancement during weak-to-moderate El Niño events as it normally benefits from enhanced winter precipitation during El Niño; however, strong El Niño events, which bring excessive winter precipitation, suppress wheat yields [34] . Thus, the lesser wheat yields during El Niño, relative to La Niña, in the Piney Woods region caused by proportionately more or excessive winter precipitation were likely due to the following conditions. The relatively excessive precipitation during El Niño led to more losses of N through leaching. The excessively cooler conditions caused by excessive precipitation under El Nino possibly provided fewer wheat tillers per unit area [68] and more freeze injury to wheat crops, especially during jointing to flowering [69] .

The simulated, average responses of ^dwheat yields in the Llano Estacado region to ENSO phases as influenced by cowpea planting date, wheat planting date, and N application rate are presented in Table 5. As the results showed, the impact of ENSO on ^dwheat yields in this region was significant especially for the scenarios comprising the wheat crops that were planted on October 15 or later following the cowpea crops planted in June and the wheat crops that were planted on October 30 or later following the cowpea crops planted in July. For these scenarios, the ^dwheat yields under the El Niño phase were significantly greater than those under the La Niña phase. The greater yields under El Niño were likely because the amounts of precipitation and crop water uptake during the wheat growing seasons associated with the significant scenarios were greater under this phase (Figure 5, Figure 6).

For the scenarios comprising wheat planted before October 15 or October 30 following cowpea planted in June or July, respectively, either the ENSO comparison was not possible because of no feasibility of cowpea-wheat doubling cropping in the study region [70] , or, in the cases of feasible scenarios, the ^dwheat yields were not significantly different across ENSO phases. With a long-term simulation study conducted for the Llano Estacado region using 80 years’ (1942-2021) weather data, reference [70] found that the number of feasible years

^†CPD: Cowpea Planting Date; WPD: Wheat Planting Date; E: El Niño; L: La Niña; N: Neutral. ^§NA: Not Available. For this scenario, cowpea-wheat double cropping system was not possible. ^‡Means followed by the same letter across ENSO phases (horizontally) within an N rate-planting date combination are not significantly different at α = 0.1.

for cowpea-wheat double-cropping in the Llano Estacado region ranged from 0 to 52, depending on the double-cropping scenario comprising the planting dates of cowpea and wheat (Figure 8). As they demonstrated, the feasibility was highest, about 65%, with July 15 and November 30 as the planting dates of cowpea and wheat, respectively. They further observed that the feasibility of double-cropping

was primarily determined by the number of days available for and required by the preceding cowpea crop and the total number of days needed by the double crops of cowpea and wheat, which decreased with delays in cowpea and wheat planting.

Even with a higher soil N level, the generally lesser yields of ^dwheat (Table 5), relative to ^mwheat (Table 4), were because of the number of feasible years for double-cropping (Figure 8). As the number of feasible years for any double-cropping scenario was not 80 (out of total 80 years available), the ^dwheat yields for the unfeasible years (80 minus feasible years) were assumed to be zero. The averaging of ^dwheat yields associated with both feasible and unfeasible years, therefore, led to the smaller values compared with the yields of ^mwheat that were associated with the years that were all feasible.

Unlike ^mwheat yields, which were impacted by ENSO only at the N rate of 50 kg·ha⁻¹ or higher (Table 4), the ^dwheat yields were impacted by ENSO also at the zero N rate (Table 5). As explained above, the insignificant impact of ENSO on ^mwheat yields at the zero N rate was due to low inherent fertility level of the soil. With the inclusion of cowpea, a legume crop, just before wheat crops, a significant amount of N (about 100 kg ha⁻¹ season⁻¹) was incorporated into the soil through symbiotic N fixation and cowpea stover residues application. Thus, under double-cropping systems containing cowpea under dryland conditions, wheat production was not N-limited but water-limited. As soil N level was not low even at the zero N rate, the ^dwheat grain yields were determined by the water supply [64] that was associated with ENSO conditions and the water use efficiency that was enhanced by high N level [65] [66] [67] . Thus, the ^dwheat yields under El Niño were significantly greater than those under La Niña even when no N fertilizer was applied.