The DSSAT crop model is a software application program comprising simulation models for many crops and tools to facilitate the effective use of the models [28]. The crop models simulate growth, development, and yield as a function of the soil-plant-atmosphere continuum. The model tools include database management programs for applications, crop data, soil, utilities, and weather. The DSSAT suite of crop models has been used for various applications at different spatial and temporal scales such as on-farm and precision management, climate variability and climate change impacts, breeding and gene-based modeling, water use, greenhouse gas emissions, and agroecosystem sustainability [28]. The inputs for the crop models include data on daily weather, soil surface and profile, and crop production-management details. The simulations are conducted primarily on a daily basis. At the end of each day, the balances of soil-plant water, nitrogen, phosphorus, and carbon, as well as development states are updated. To simulate multi-year outcomes of crop management strategies, DSSAT integrates the effects of soil, crop phenotype, weather, and management options by combining crop, soil, and weather databases with crop models and application programs.

The Sequence Analysis tool allows the user to conduct rapid inspection and analysis of results of long-term cropping sequences [29]. The model tool allows the user to calculate a series of statistics and create various graphics that examine relationships between trends and variability. The main aspect of the sequence analysis is that simulation studies are conducted across multiple cropping seasons; thus, status of soil water and nutrient is carried over from one cropping season to the subsequent one [30] [31].

The Texas A&M AgriLife Research & Extension Center at Overton (32.29˚N, 94.97˚W) is situated in the Pineywoods region of the southern United States. This region, where agriculture is a major economic activity, comprises eastern Texas, western Louisiana, and southern Arkansas. At the Overton Center, numerous soil fertility and cultivar trials and grazing experiments have been conducted by various scientists since 1967. In this study, Overton, Texas was used as a representative site for the Pineywoods region and the humid Southeastern United States [21] [22].

To explore the interannual climate variability effects on the yields of cowpea and wheat in the Pineywoods region, a long-term weather dataset spanning 80 years (1942-2021) was used. The historical daily data on temperature and precipitation were obtained from https://www.ncei.noaa.gov/access/search/data-search/daily-summaries (National Centers for Environmental Information) and the reports and publications from the Texas A&M AgriLife Research and Extension Center at Overton, whereas those on solar radiation were generated using a reliable irradiation model described by [32].

Darco (Grossarenic Paleudults) and Lilbert (Arenic Plinthic Paleudults) were used as representative soils for the study because they are some of the major soils used for agricultural purposes in the Pineywoods region [33] [34]. The soil data were obtained from the Gridded Soil Survey Geographic (gSSURGO) database of the USDA NRCS [35] [36]. These soils are distinct in various aspects, including texture and inherent fertility level. Compared with Lilbert, Darco had less clay and silt contents, greater saturated hydraulic conductivity, deeper A and E horizons (122 cm vs. 58 cm), less organic C and inorganic N contents, and smaller values of field capacity and wilting point (Table 1). However, the water holding capacities of both soils were about the same.

The DSSAT Sequence Analysis tool was used to simulate grain yields for cowpea and winter wheat in two sequences: cowpea-wheat double crop and fallow-wheat. The wheat crop that followed cowpea or fallow, hereafter, will be referred to as ^cwheat and ^fwheat, respectively. Simulations were made using two soils, Darco and Lilbert, and three N application rates only to wheat at 0, 100, and 200 kg N ha⁻¹. Thus, a total of 12 scenarios were assessed, which comprised 2 sequences × 2 soils × 3 N rates (Table 2).

For simulations, the following management and environmental inputs were assumed. For cultivars, we used “Hyfowet” for winter wheat and “Cal #5 MG4”

a. MH = master horizon, WP = wilting point, FC = field capacity, SA = saturation, WH = water holding capacity, HC = saturated hydraulic conductivity, TN = total N, OC = organic carbon.

a. Double crop.

for cowpea. The Hyfowet cultivar was the highest-yielding wheat cultivar in east Texas as identified by [21]. Because the genetic coefficients for this cultivar were already estimated by [21], there was no need to further calibrate and evaluate the wheat model for this cultivar. For the Cal #5 MG4 cultivar, the default genetic coefficients for this cowpea cultivar are those in the standard DSSAT release [27] and correspond to the cultivar coefficients upon which the cowpea model was adapted (K. J. Boote, personal communication, 11 February 2022). The simulation start date was assumed to be June 20, 1942, and simulation would terminate on the harvest date of cowpea in 2021. On the simulation start day, soil moisture was assumed to be at field capacity and soil N content 25 kg ha⁻¹. For planting dates, we assumed June 20 for cowpea and October 20 for wheat. The plant populations used were 30 plants m⁻² (about 40 kg seed ha⁻¹) for cowpea and 323 plants m⁻² (about 100 kg seed ha⁻¹) for wheat. Using the conventional tillage, dry seeds were planted on rows at 3 cm depth. Inorganic N fertilizer was given only to wheat, not cowpea.

Of the total amount of N applied to wheat, 50% was applied at the planting time and the remainder on February 15 of the following year. For organic amendments, the cowpea stover residue of 2125 kg DM ha⁻¹, with the N content of 1.5%, was incorporated into the soil on the planting date of wheat, and the wheat residue of 500 kg DM ha⁻¹, with the N content of 1%, was incorporated in the soil on the planting date of cowpea. For soil organic matter, Century was used as the method, with the five years’ field history of “Cultivated, good management, initial default SOM” [37].

For ENSO analyses, the yields of cowpea, ^cwheat, and ^fwheat that were simulated for each of the 80 years (1942-2021) were assigned to an ENSO phase – El Niño, La Niña, or Neutral – as categorized by the Japan Meteorological Agency (JMA) index [38] [39] [40] [41]. For ENSO characterization, the JMA index was chosen because it best selects the known ENSO events [41]. Accordingly, the total numbers of El Niño, La Niña, and Neutral years analyzed were 18, 20, and 42, respectively. The JMA index is a 5-month moving average of the sea surface temperature anomalies over the tropical Pacific (4˚S - 4˚N, 150˚W - 90˚W). The ENSO year of 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 [39] [41].

For cowpea, ^cwheat, and ^fwheat each significance tests were carried out to assess grain yield differences across soil types as influenced by N rate and ENSO interactions, across N rates as influenced by soil type and ENSO interactions, and across ENSO phases as influenced by soil type and N rate interactions. The tests were done using a nonparametric alternative to the two-sample t-test, known as the pairwise Wilcoxon rank sum test [42]. The reason for using the Wilcoxon test was that 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/).

The yields of ^cwheat and ^fwheat were both impacted by the N application rate, regardless of soil type or ENSO phase (Table 4). On both soils under all ENSO phases, the yields of both ^cwheat and ^fwheat increased with an increase in N rate from 0 to 100 and 200 kg ha⁻¹, respectively. However, with an increase in N rate from 100 to 200 kg ha⁻¹, ^cwheat yields did not increase. The yield responses of ^cwheat to N rate were lower than those of ^fwheat because ^cwheat yields were associated with a higher soil fertility level from cowpea residue (about 32 kg N ha⁻¹ season⁻¹) and N transfer through N fixation (about 117 kg N ha⁻¹ season⁻¹). Since the inherent fertility level of the soil associated with ^cwheat was relatively high, additional amounts of N contributed less; thus, the ^cwheat yields started plateauing at 100 kg N ha⁻¹. The ^fwheat yields, however, did not level off at this N rate due to a lower soil fertility level.

For cowpea yields, however, no N application rate effect was observed on either soil under all ENSO phases. These results were expected because cowpea is a

^†,cwheat = wheat preceded by cowpea, ^fwheat = wheat preceded by fallow, ^‡Means followed by the same letter across N rates (vertically) within a soil-ENSO-crop combination are not significantly different at α = 0.1.

legume, despite no N fertilizer, and wheat residues carry-over (5 kg N ha⁻¹ season⁻¹ from the residue of 500 kg DM) did not enhance cowpea yields. The N applications were made only for ^cwheat and ^fwheat crops. These N applications did not have any significant residual effects on the yields of cowpea which followed wheat.

The yields of cowpea were impacted by ENSO on both soils and at all N rates considered. Regardless of soil type and N rate, cowpea yields in El Niño years were greater than those in La Niña years (Table 5).

The greater yields under El Niño conditions were likely due to the greater amounts of precipitation this ENSO phase delivered during the establishment phase, the growing season, the peak summertime, and the flowering and pod formation stages (Figure 1).

Although the ENSO impacts on ^cwheat and ^fwheat yields were statistically evident only at 0 kg N ha⁻¹ and at 0 and 100 kg N ha⁻¹, respectively (Table 5), the general trend was that, regardless of soil type and N rate, yields of both ^cwheat and ^fwheat in La Niña years were greater than those in El Niño years (Figure 2). These results are in agreement with the findings of [19] [46] [47] and [48]. The likely reasons for the greater yields of both ^cwheat and ^fwheat under the La Niña phase, in general, were as follows. First, due to warmer conditions at

^{†, c}wheat = wheat preceded by cowpea, ^fwheat = wheat preceded by fallow, ^‡Means followed by the same letter across ENSO phases (vertically) within a soil-N rate-crop combination are not significantly different at α = 0.1.

the planting time under La Niña than under the other ENSO phases, wheat had better germination. For good germination and growth of wheat, 12˚C to 25˚C is the optimum temperature range [49]. Second, during germination and early vegetative stages, no significant harmful effect of drier conditions under La Niña was likely because the total amount of water required for these stages was small (<1.5 cm·d⁻¹) relative to the ones that occurred later in the season [50]. Third, due to warmer conditions under La Niña, more wheat tillers per area would be likely due to warmer temperatures [51], with the optimum temperature for tiller development being 13˚C [52]. Fourth, under the La Niña phase, more photosynthetic assimilates would be expected as this phase has been generally associated with clearer skies and more solar irradiation. Fifth, the La Niña phase had less likelihood of the occurrence of various insect pests and diseases due to drier and warmer conditions [17] [53]. Finally, La Niña was less likely to provide freeze injury to wheat crops, particularly during jointing to flowering [54].