A simple cultivar suitability index for low‐pH agricultural soils

Planting wheat (Triticum aestivum L.) cultivars that carry the aluminum (Al) resistance gene TaAlmt1 is a potentially lower cost alternative to lime applications in acidic agricultural soils. However, the relative importance of TaAlmt1 expression and adaptedness (to regional environmental conditions) for preserving grain yield in acidic soils is not well understood. Adaptation trials were established in low‐pH soils to compare lime‐amended (YL) and unamended grain yield (YU) among regionally adapted spring wheat cultivars with and without TaAlmt1. Averaged across YL and YU (YAVG), yield of adapted TaAlmt1 carriers was similar to adapted noncarriers (p = .939) but greater than nonadapted noncarriers (p = .024). Soil pH‐driven spatial heterogeneity appears to inflate yield coefficients of variation and the probability of committing type II errors in cultivar yield comparisons. Our results support the use of YAVG as a suitability index and decision support tool for cultivar selection in acid‐affected soils.


INTRODUCTION
Lime applications are effective for remediating low-pH surface soils but can be cost-prohibitive for farmers, especially in dryland wheat (Triticum aestivum L.) systems with low economic margins. Thus, planting wheat cultivars that carry the Al resistance gene TaAlmt1 is more common than liming. However, there is a knowledge gap regarding the relative importance of TaAlmt1 expression and adaptedness for preserving grain yield in acidic soils (Gillespie, Marburger, Abbreviations: Al KCl , KCl-extractable aluminum; Al sat , aluminum saturation; CVs, coefficients of variation; Y AVG , grain yield averaged across lime-amended and unamended low-pH soils; Y L , grain yield in lime-amended soils; Y U , grain yield in unamended, low-pH soils. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Carver & Zhang, 2020; Scott, Fisher, & Cullis, 2001). Here, we define adaptedness as the ability of a cultivar to achieve high grain yield across multiple locations with disparate weather and soil conditions. A field study was conducted to determine if adapted spring wheat cultivars carrying TaAlmt1 outyielded adapted noncarriers in lime-amended and unamended low-pH soils. Interestingly, a high degree of within-replicate and within-plot variability in plant vigor was detected. A literature search failed to corroborate this observation, although spatial heterogeneity of soil pH is well documented (Burrough, 1983;Campbell, 1978;McBratney & Webster, 1981;Wong, Asseng, Robertson, & Oliver, 2008).
We tested the hypotheses (a) that adaptedness is important for preserving grain yield in low pH soils and (b) that soil pH-driven spatial heterogeneity confounds grain yield results in low-pH field trials. Our goal was to develop a simple, generalizable cultivar suitability index robust to complications arising from field-scale heterogeneity in the edaphic environment.

Site description and study design
Field trials were established at two dryland farms in Chouteau County, MT, near the towns of Highwood and Geraldine in 2018 and 2019. The dominant soil series were Bearpaw clay loam (fine, smectitic, frigid Vertic Argiustolls) mixed with Vida clay loam (fine-loamy, mixed, superactive, frigid Typic Argiustolls) at Highwood and Scobey clay loam (fine, smectitic, frigid Aridic Argiustolls) mixed with Kevin clay loam (fine-loamy, superactive, frigid Aridic Argiustolls) at Geraldine (Soil Survey Staff, 2020). Trials were planted in a complete factorial design consisting of two lime rates (0 and 11 Mg ha −1 ) and nine cultivars with four replicates. Each plot (1.5 m by 6 m) was harvested with a small-plot combine to determine grain yield.

Lime amendments and soil analyses
Aglime (Montana Limestone Company) with 99% passing through a 0.149-mm sieve was applied on 12 Oct. 2017 at Geraldine and 31 Oct. 2017 at Highwood. The lime was incorporated with tillage to 12 cm at Geraldine and 15 cm at Highwood. One (in 2018) and four (in 2019) soil cores (0-10 cm) per lime treatment per replicate per site were collected prior to seeding for a total of 80 samples. Soil pH, base cations, and KCl-extractable Al (Al KCl ) were determined for each sample. Aluminum saturation (Al sat ) was calculated as Al KCl (cmol c kg −1 ) expressed as a percentage of the effective cation exchange capacity (Kariuki et al., 2007;Sumner & Miller, 1996).

Cultivar classification
Nine spring wheat cultivars were grouped into adapted and nonadapted classes based on yield results from Montana State University spring wheat adaptation trials established in 13 dryland environments in 2019 (MSU, 2019b). One exception was made: 'Dagmar' (PI 690450) showed promise in limited testing but was not present in all statewide trials and was excluded from the MSU (2019b) summary. 'Dagmar' was classified as adapted in the current study. For the remaining eight spring wheat lines, the median yield of the dataset, 3.6 Mg ha −1 , was used to differentiate between nonadapted cul-

Core Ideas
• Spring wheat grain yield was assessed in limeamended and unamended low-pH soils. • Regionally adapted TaAlmt1 carriers did not outyield adapted noncarriers. • Cultivar selection based on yield averages across low-and neutral-pH soils is advised. • Soil pH-driven spatial heterogeneity elevates yield CVs of low-pH field trials.

Statistical analysis
Linear mixed modeling was performed using R v. 4.0.0 (R Core Team, 2020) with the lmer function of the lme4 package v. 1.1-23 (Bates et al., 2015). Model parameters were estimated by restricted maximum likelihood. In the full model, site-year, cultivar, lime, and a lime × cultivar interaction term were designated as fixed effects, with lime included as a random effects term nested in replicate and site-year. Effects of site-year were nonsignificant (p = .511) and a lime × cultivar interaction was not detected (p = .204), supporting an assessment of cultivar yield differences across site-years and lime treatments. In the models for grain yield in lime-amended soils (Y L ) and in unamended, low-pH soils (Y U ), cultivar was designated as a fixed effect term and replicate as a random effects term nested in site-year, whereas lime treatment was designated as an additional random effects term nested in replicate and site-year in the model for grain yield averaged across amended and unamended soils (Y AVG ). Assumptions of normality and homogeneity of variance were satisfied according to the shapiro.test function in base R and the lev-eneTest function in the R package car v. 3.0-8 (Fox & Weisberg, 2019), as well as visually. Within-cultivar grain yields adjusted for random effects terms (i.e., estimated marginal means) and Tukey pairwise comparisons among cultivars were calculated using the emmeans function from the R package emmeans v. 1.4.7 (Lenth, 2020). Contrasts of adapted carriers, adapted noncarriers, and nonadapted noncarriers T A B L E 1 Lime-amended (Y L ) and unamended (Y U ) grain yields and Y L and Y U average yield (Y AVG ) of spring wheat TaAlmt1 carriers (+) and noncarriers (-), ± SD. Results are summarized across four site-years. Pairwise comparisons revealed no statistical differences among cultivars were performed using Scheffe adjustment for multiple comparisons in the contrast function of the emmeans package (Lenth, 2020).

Suitability index
Pairwise comparisons failed to detect cultivar differences in Y L , Y U , and Y AVG due to large within-cultivar standard deviations across site-years (Table 1). However, Y AVG for adapted TaAlmt1 carriers was greater than for nonadapted, noncarriers (p = .024). Similarly, Y AVG of adapted noncarriers was greater than nonadapted noncarriers (p = .008). There was no difference in Y AVG between adapted TaAlmt1 carriers and adapted noncarriers (p = .939). Results suggest that Y AVG is an appropriate index for cultivar suitability, defined here based on distribution characteristics of resistance-adaptedness clusters in Y L -Y U data space (Figure 1). We propose that cultivar suitability be quantified as the length of AB, the normal distance from a line, L, to a point, B, calculated as follows: where m is the slope of the line normal to L. Because slopes of the lines of best fit for both resistance-adaptedness clusters which is directly related to the arithmetic mean of Y U and Y L , providing conceptual justification for the use of Y AVG as a cultivar suitability index. A similar framework has been used in other disciplines, including remote sensing applications (Zhan, Qin, Ghulan, & Wang, 2007).
T A B L E 2 Soil pH, extractable Al (Al KCl ), and Al saturation (Al sat ) ± SD in limed (+lime) and unlimed (-lime) conditions grouped by site-year

Grain yield variability
Comparisons of wheat yield coefficients of variation (CVs) were made in a low-pH (<5.0) adaptation trial and three unlimed, neutral-or near neutral-pH (∼6.5-7.5) adaptation trials near the current study over the same period (2018-2019; data not presented). Trialwide grain yield CVs in neutral-pH environments (MSU, 2019a) averaged 9.4 ± 3.5% compared with 18.5 ± 5.6% in the low-pH environment. Similarly, when averaged across site-years, grain yield CVs of neutral-pH spring wheat adaptation trials (MSU, 2019a) were 38 and 51% of the current study's unamended and lime-amended yield CVs, respectively. This pattern extends beyond spring wheat to different crop species assessed in nearby trials. For example, yield CVs of neutral-pH adaptation trials with field pea (Pisum sativum L.) (MSU, 2018b; MSU 2019c), canola (Brassica napus L.) (MSU, 2018a;MSU, 2019a), and barley (Hordeum vulgare L.) (MSU, 2019a) were 28, 29, and 43% of unamended low-pH yield CVs and 28, 37, and 65% of lime-amended yield CVs, respectively. The relatively large CVs calculated for Y L and particularly Y U highlight the logistical challenges of low-pH cultivar comparisons and provide indirect evidence for spatial variability of the edaphic environment as a confounding factor.

Soil variability
Large standard deviations in Al KCl and Al sat within individual site-years and lime treatments were observed in the current study (Table 2), despite considerable effort to locate spatially homogeneous acid soils for trial establishment. Limed soils exhibited more variability in Al KCl and Al sat in 2019 than in 2018, suggesting sample size affected variability or variability increased with time from application. The high degree of variability in Al KCl and Al sat in limed and unlimed soils is especially interesting in light of the comparatively low variability in soil pH. When assessed at the field scale, where pH can range by up to 3 pH units (R. Engel, personal communication, 2020), cultivars with strong yield potential in both low-pH and neutral-pH soils will likely outperform those with strong yield potential in low-pH soils only. Basing cultivar recommendations on Y AVG could minimize impacts of potentially confounding factors (e.g., edaphic spatial heterogeneity) and increase the likelihood that crop consultants and other agricultural professionals will endorse the best-adapted cultivar for a given environment, farm, or field. Unavailability of Y L data may warrant approximation of Y AVG as mean grain yield in low-and nearby neutral-pH trials, although additional research is needed.

CONCLUSION
Visualizations in Y L -Y U data space as well as contrasts of Y AVG among well-adapted TaAlmt1 carriers, adapted noncarriers, and nonadapted noncarriers suggest (a) adaptedness is important for achieving high grain yield under low-pH field conditions and (b) Y AVG is an appropriate index for cultivar suitability. This study provides evidence supporting the hypothesis that low-pH adaptation trials may be confounded by pH-driven spatial heterogeneity in the edaphic environment. We argue that the spatial complexity of pH, Al KCl , and Al sat within agricultural fields, along with the high yield CVs observed in this and other low-pH adaptation trials, provides ample justification for cultivar recommendations based on Y U and Y L averages. Because Y L data are often unavailable, future work should investigate the integrity of cultivar recommendations based on grain yield averages across low-and nearby neutral-pH field trials. In future low-pH field trials, alternative experimental designs and/or spatially explicit yield corrections are advised.

A C K N O W L E D G M E N T S
The authors are grateful to USDA's Western Sustainable Research and Education program and the Montana Fertilizer Advisory Committee for funding this research. Special thanks to Dr. Rick Engel, whose contributions significantly improved early versions of this manuscript.

C O N F L I C T O F I N T E R E S T
The authors declare no conflict of interest.