Environmental and management drivers of soybean yield in a subtropical region with recurrent drought and heat stress

Agricultural expansion in South America’s Semiarid and Subhumid Chaco has accelerated in recent decades, predominantly driven by the growth of soybean farming. Our objectives were to determine the main climatic and management constraints that explain soybean yields variability and to identify strategies to cope with them. Given the lack of locally validated best management practices, our approach was to analyze on-farm data from 2840 production paddocks with mixed models, attempting to capture the environmental and management heterogeneity of the area. The high variability in soybean yields in the study area was greatly determined by climatic variables. The predictors that explained the largest proportion of yield variance were the number of years with soybean as preceding crop, the effective rainfall between October (i.e. before planting) and April, the number of days with maximum temperatures above 35 °C between day 50 and 100 from planting and the mean maximum temperature between December and April. Fields that avoided soybean in recent rotations achieved higher water use efficiency, an effect further conditioned by rainfall. Based on our findings, the following agronomic strategies and research areas should be prioritized: i) increasing crop diversification and avoiding soybean monoculture; ii) implementing agricultural practices aimed to increase pre-plant water availability (i.e. maintaining high soil coverage); iii) sowing soybean at different dates within the same farm to expose the crops to varied climatic conditions

1 Introduction

Agricultural frontiers are still expanding in several areas of the world, especially on less favored agricultural lands since the most suitable ones are already being cultivated or used for other purposes. Areas recently incorporated into agriculture are therefore overly sensitive to environmental degradation and need specific management practices, likely different from those with longer agricultural history. Among these new areas, the Semiarid and Subhumid Chaco in South America has been a hotspot of agricultural expansion in the last decades (Vallejos et al., 2015; Volante et al., 2016), favored by the increase in annual rainfall (Ricard et al., 2015), the adoption of no-till farming which has resulted in reduced cultivation costs and increased water use efficiency, and the changes in the land tenure regime (Piquer-Rodríguez et al., 2015). In addition, the favorable international prices, the acceptable yields and the introduction of adapted varieties has led soybean to constitute the main crop in the area (Dominguez and Rubio, 2019; Goldfarb and Van Der Haar, 2017). The climate of the region is characterised by having the maximum absolute temperatures of the continent while precipitations present a monsoon regime with great interannual variability (De Ruyver and Di Bella, 2019; Ricard et al., 2015). Despite the frequent limitations exerted by either hydric or thermic stress (Giménez et al., 2015), recent reports confirm that soybean crops are economically viable in the area (Madias et al., 2021; Casali et al., 2021). Furthermore, the Chaco region has shown the largest yield gaps for soybean in Argentina, suggesting that there is great potential to increase yields (Aramburu-Merlos et al., 2015). However, the Semiarid and Subhumid Chaco lacks, like any other new agricultural area, a sufficient agricultural history to provide a sound guidance on the best management practices adapted to the local conditions.

The present challenge in the study area is to identify the key factors limiting soybean yields and to define the most effective management practices considering soil, climate, and genotype interactions (Andrade et al., 2017). Given the lack of locally validated best management practices, one approach is to rely on farm data attempting to capture the environmental and management heterogeneity of the area. Agricultural research embedded in real-world farm management allows addressing complexity and uncertainty that it is often not evident from experiments under controlled or semi-controlled conditions (Lacoste et al., 2022; Madias et al., 2025). In addition, when farmers are directly involved, the results gain credibility and can be more easily shared with the local community. Using this approach, a recent work covering a wide area of the Argentinean Gran Chaco region identified the number of days with maximum temperatures above 35 °C from R1 to R7, the amount of rainfall from 30 days before sowing to R7, the years after land conversion to agriculture, the evapotranspiration from sowing to R7, the sowing date, the genotype selection and the soil phosphorus availability as the main predictors of soybean yield (Madias et al., 2021). Given the broad surface area and environmental contrasts of the Gran Chaco, this type of analysis should be extended to specific subregions, as the Semiarid and Subhumid Chaco.On-farm research usually involves large databases with spatial and temporal dimensions including correlated variables (Lacoste et al., 2022). In such sense, mixed models provide a framework to capture the complex correlation structures of multidimensional data through random effects and their associated components (Gambin et al., 2016; Coyos et al., 2018; Madias et al., 2025). This statistical approach also has the advantage that can be used with unbalanced data and missing observations (Smith et al., 2005).

The evidence gathered so far in the Chaco Region suggests that management practices aimed at mitigating water deficits and finding less restrictive environments during critical growth periods would be essential for obtaining stable yields (Aramburu-Merlos et al., 2015; Casali et al., 2021, 2022; Kettler et al., 2025; Madias et al., 2021). For example, studies using mechanistic models comparing different local climatic scenarios indicated that, in contrast to maize, delaying soybean sowing dates lead to significant decreases in crop yield, even though the duration of the crop cycle is not affected (Casali et al., 2021).

In the present work, an extensive data set was collected from on-farm production paddocks in the Semiarid and Subhumid Chaco to assess the following objectives: i) to determine the main climatic and management constraints that explain soybean yields, and (ii) to identify management strategies with the potential to alleviate the effects of climatic constraints.

2 Materials and methods

2.1 Study area

The Semiarid and Subhumid Chaco areas are included in the Great American Chaco Region (Baumann et al., 2016) and occupy a large part of Northern Argentina, mainly in the Santiago del Estero and Chaco provinces (Figure 1). Annual precipitation ranges from 600 to 800 and from 800 to 1000 mm (1901-2020) in the Semiarid and Sub-humid Chaco, respectively Barnatán et al. (2025), but with a high interannual variability and higher rainfall in summer (December to March) than in winter (June to September). For much of the year, especially during the summer, the environmental demand exceeds the soil water supply, thus episodes of water deficit for crops are common (Maddonni, 2012). Mean annual temperatures range between 19 - 22 °C in a gradient from South to North, with monthly averages in the summer exceeding 27 °C in the Northern most areas of the region.

Figure 1

Figure 1. Geographical distribution of the 2840 agricultural paddocks in the Semiarid and Subhumid Chaco (range 5-443 paddocks per yellow point).

2.2 Field data

Soybean field data were obtained from production paddocks (Guayacán and Gancedo-La Paloma CREA groups) (63% of the data) and from genotype field evaluation networks (Aapresid; INTA-EEA Las Breñas; INTA-EEA Famaillá) (37% of the data). The consolidated dataset included 2840 production paddocks in eight cropping seasons (2006 to 2015) and 29 locations (21 from the Subhumid Chaco and 8 from the Semiarid Chaco). A total of 168 genotypes were tested in those paddocks. The dataset included crop yields and variables related to agronomic management (sowing date, density at harvest, row spacing, genotype and fertilization), agricultural history (years that the paddock has been cropped, years under no-till, and crop rotation), water availability (monthly rainfall between October and April and soil water profile characterization) and soil features (organic matter, available nitrogen, phosphorus, and sulfur and pH). Meteorological data, including maximum and minimum temperature, solar radiation and rainfall, was obtained from six INTA (Argentinean National Agricultural Technology Institute) stations (Las Breñas, Bandera, Sachayoj, Añatuya, Castelli and Los Frentones – https://siga.inta.gob.ar/#/), and one from the Sociedad Rural Quimili (i.e. a local association of farm owners). A series of environmental variables were calculated from the collected data set, including mean maximum temperatures, mean temperatures, accumulated rainfall, accumulated effective rainfall according to a method recommended by INIA from Uruguay (http://www.inia.uy/gras/Monitoreo-Ambiental/Balance-H%C3%ADdrico/Calculo-Precipitacion-Efectiva), potential evapotranspiration (Hargreaves and Samani, 1985) and heat stress indices.

2.3 Predictor variables

The potential yield predictors were classified into five groups of thematically related variables and each of them were identified as quantitative or categorical:

1. Agronomic management and agricultural history variables. The quantitative variables of this group were: sowing date, plant density, distance between rows, genotype maturity group, years after land conversion to agriculture, years of no-till, and preceding crops in the last five years. The latter variable was expressed as number of years with each crop (gramineous crops were counted in the same group). The categorical variables of this group were: fertilization (yes/no), and growth habit of the genotype (determinate/semi-determinate/indeterminate). Nitrogen was the only nutrient added as fertilizer, but only 8% of the plots were fertilized.

2. Water availability variables. The quantitative variables of this group were: precipitation (mm), effective precipitation (mm), potential evapotranspiration (mm) according to Hargreaves and precipitation/evapotranspiration ratio (mm). All these variables were calculated monthly and in different periods between October and April. We included in this group a categorical variable about available water at sowing along the soil profile classifies as poor, fair, or good by the farmer.

3. Temperature variables. The quantitative variables of this group were average maximum and average temperatures recorded monthly and in different periods between December and April (°C).

4. Heat stress variables. The quantitative variables of this group were: number of days and cumulative degree days (°C) in which the maximum temperature was above six thresholds (20, 25, 30, 35, 40, 45 °C) in nine different periods during the crop cycle (0-50, 50-100, 50-150, 100-150, and 0-150, days after sowing).

5. Soil variables. The quantitative variables of this group were organic matter (%), total nitrogen (%), extractable phosphorus (mg kg ^-1; Bray), extractable sulfur (mg kg ^-1) and pH, all at 0-20 cm depth.

2.4 Statistical analysis and model selection

We were interested in identifying and ranking the most relevant environmental and management variables determining soybean yields in the Semi-arid and Sub-humid Chaco. We began by evaluating single-constraint models for each of the five groups of predictor variables described above. The Site×Year combination was considered as the environment because some observations came from nearby sites, and from the same site in different years. This factor was considered as random and the predictor variable within each group as the fixed factor. These models were of the type:

$${\text{Y}}_{\text{ij}}={\beta }_{0\text{j}}+{\beta }_1*{\text{X}}_{\text{i}}+{\in }_{\text{ij}}$$

where Y_i is the grain yield in farm’s paddock i at environment j, β_0j is the random effect of environment j, β₁*X _i is the fixed effect of predictor X and ∈_i is the residual. Once that the best predictor had been selected within each group of variables, we considered their potential use for a multiple constraint mixed models. For the model selection procedure, we started with a full model that included all predictors (i.e. the best predictor identified within each group of similar variables) and all the possible interactions. Seven simplified models (Table 1) with less predictors and/or no interactions (i.e. additive effects) were compared with the full and null models. The structure of these models was the same as the single-predictor models, with identical random component, but considering more than one predictor. The observations of the candidate fixed-effects predictors were standardized by z-scores to address the fact that the predictors varied at very different scales. The z-scores have the advantage that do not change the functional relationship between the predictor variables and the response variable.

Table 1

Table 1. General structure of multiple mixed-effects models for predicting soybean yield in the Semiarid and Subhumid Chaco, shown with some examples.

The lme4 package (Bates et al., 2015) of the R statistical software (R Core Team, 2022) was used to perform the mixed effects analysis. For both univariate and multivariate mixed models, an established protocol (Zuur et al., 2010) was followed to assess i) outliers, ii) homogeneity of variances, iii) normal distribution, and iv) independence. The existence and type of relationship between predictor variables and soybean yields were evaluated through scatter plots and Pearson’s correlation coefficient analysis. To verify homoscedasticity, standardized residual plots were visually inspected against the predicted residuals for the overall model and for each of the predictor variables. Normality was inspected through the comparison of the distribution of the residuals of the model (assumed to be normal) with the theoretical normal distribution. The temporal correlation between residuals was assessed with the auto-correlation function (ACF). Model coefficients were estimated using ML (maximum likelihood), as the random component was always the same, and only the predictors were modified. Multicollinearity between predictors in the multivariate models was assessed through the VIF (Variance Inflation Factor) values, using the car package (Fox and Weisberg, 2018), and following the criterion that VIF values less than 10 indicate that there are no multicollinearity problems. The selection of models followed the multi-model inference approach based on information theory (Burnham et al., 2011).

Our multi-model inference approach did not rely on the assumption that there was a single “true model”,but rather examined simultaneously several competing hypotheses to identify the best set of models capable of summarizing which “effects” (represented by predictors) can be supported by the data. The selection of model predictors was based on the Akaike Information Criterion (AIC) (Burnham et al., 2011). For the selection of predictors for the final multivariate model, the top-down model selection strategy was followed. The goodness-of-fit of the mixed models was assessed, in addition to their AIC value and relative weight, with the R² of the fitted models, following Nakagawa and Schielzeth (2013). The marginal R² represents the variance explained by the fixed factors, while the conditional R² represents the variance explained by the whole model (fixed and random effects).

3 Results

3.1 General database results

Data obtained from 2840 paddocks that included a wide variability of management and environmental conditions and soybean yields ranging from 104 kg ha^-1 to 6004 kg ha^-1, provided a comprehensive overview of the soybean production systems in the Semiarid and Subhumid Chaco. Sowing dates varied from December 2^nd to February 20^th, plant density at harvest from 11.8 to 42.7 pl m^-2 and inter-row spacing from 35 to 70 cm. The 168 genotypes belonged to maturity groups ranging from 4.4 to 9. The number of years from land conversion to agriculture varied from 3 to 18 years. In the last five years, the most common preceding crop was soybean, followed by maize and to a lesser extent wheat, sorghum, sunflower, cotton, oats, cover crops, the perennial grass Gatton Panic (Megathyrsus maximus) and perennial natural pastures. The plots were almost exclusively cultivated under no-tillage and only 7% of them were double cropped with either sunflower/soybean, cover crop/soybean, maize/soybean, or wheat/soybean. Accumulated precipitation between December and March varied widely, ranging from 89 to 857 mm, while the average maximum temperature ranged from 29.6 to 36.9°C for the same period.

3.2 Single-constraint models

Single-constraint models identified the predictors that explained the largest proportion of yield variance for each of the five groups of thematically related variables. They were the number of years with soybean as preceding crop (PC) (Figure 2), the effective rainfall between October and April (POA) (Figure 3), the number of days with maximum temperatures above 35°C between day 50 and 100 from planting (D35_50-100) and the mean maximum temperature between December and April (TDA) (Figures 4, 5), for the agronomic management, water availability, heat stress and temperature groups, respectively. None of the predictors of the soil group showed a significant association with soybean yield.

Figure 2

Figure 2. Relationship between soybean grain yield (kg ha^-1) and years with soybeans in the last 5 years. The line is a linear regression fitted to the data, with the shaded area indicating the 95% confidence interval.

Figure 3

Figure 3. Relationship between soybean grain yield (kg ha^-1) and effective rainfall between October and April (mm) as affected by the number of years with soybeans in the five years prior to planting. The lines represent the linear regressions of the data, with the shaded area indicating the 95% confidence interval.

Figure 4

Figure 4. Relationship between soybean grain yield (kg ha^-1) and days with maximum temperature above 35°C between 50 and 100 days after planting. The line represents the linear regression of the data, with the shaded area indicating the 95% confidence interval.

Figure 5

Figure 5. Relationship between soybean grain yield (kg ha^-1) and average maximum temperature between December and April (°C) as affected by days with maximum temperatures greater than 35°C between the 50th and 100th day after planting (D35). The lines represent the linear regressions of the data, with the shaded area indicating the 95% confidence interval.

3.3 Multiple linear mixed-effects models

After identifying the best yield predictor variables within each group of thematically related variables, we evaluated multiple mixed-effects models to identify the most influential factors affecting soybean yields and the main interactions. The model that best explained soybean yield variability (model 1 in Table 1) included the four predictor variables selected in the previous phase: number of years with soybean in the last five years (PC), the effective precipitation between October and April (POA), the number of days with maximum temperatures above 35°C during the period 50 and 100 days after sowing (D35_50-100), and the average maximum temperature during the growth cycle (TDA), and two interaction terms: POA * PC and D35_{5 0-100} * TDA (Table 2).

Table 2

Table 2. Estimates ± standard error (SE) of fixed effects for the final model (Model 1 in Table 1), shown for unstandardized data.

4 Discussion

Soybean clearly appeared as the most unfavorable preceding crop in the agricultural rotation (Figure 2). The lowest grain yields were observed under soybean monoculture (i.e. five consecutive years of soybean), whereas the highest grain yields were observed in plots with only one year with soybean out of the last five. Therefore, reducing soybean monoculture and promoting crop diversification were essential for increasing resilience and improving crop performance, as previously observed in the Chaco Region (Casali et al., 2022) and elsewhere (e.g. Crookston et al., 1991; Wilhelm and Wortmann, 2004; Yamoah et al., 1998). We also observed that the effective precipitation affected the role played by soybean in the crop sequence: when soybean appeared less frequently in the previous five years, the crop produced more grain per millimeter of rainfall. Yields increased by roughly 4.7 kg ha^-1; per mm of rainfall under such conditions, compared to 4 kg ha^-1; per mm when soybean was grown in more than two of the past five years (Figure 3). Wright et al. (2012), in a review of agrosystems from subtropical regions, concluded that water productivity in crop-livestock mixed systems was higher than in systems based exclusively on agricultural crops. Although this practice is not common in the area, some initial results are promising (Peri et al., 2016). This sensitivity to the proportion of soybean cultivation in previous years identified in this study underscores the importance of diversifying crop rotations by incorporating crops that generate a greater amount of residue (Casali et al., 2022). Soybean monocultures are associated with yield penalties due to low soil coverage, high soil evaporation, accelerated soil carbon loss, poor soil structure, as well as greater difficulties in the control of weeds, insects and diseases (e.g. Karlen et al., 2006; Villarino et al., 2017).

Among variables related to water availability, the effective precipitation between October and April was the best predictor of crop yield. Since this parameter refers to precipitation that infiltrates into the soil, its preponderance over absolute precipitation suggests that a significant proportion of local rainfall is lost through runoff and evaporation. This finding highlights the importance of rainwater harvesting strategies, as early suggested by Magliano et al. (2015) for the Chaco region. Among the different periods evaluated for the calculation of effective rainfall, the period between two months before sowing and harvest (October to April) outperformed other periods such as the entire soybean growth cycle (December to April) as the best predictor of soybean yield (Figure 3). Obtained results suggest that management practices enhancing pre-planting soil water storage can help maximize crop yield. Examples of these practices include anticipating the harvest of the preceding crops, increasing soil cover through cover crops or the inclusion in the rotation of cash crops that leave a high amount of residues production (e.g. maize)) (Blanco-Canqui, 2013) and no-tillage, which is already widely adopted by local farmers (Villarino et al., 2017).

Among variables related to temperature, the best predictor of soybean yields was, unexpectedly, the mean maximum temperature during the entire growth cycle (December to April) (Figure 5), outperforming other variables such as mean or minimum temperatures or maximum temperatures during other periods of the growth cycle. This result highlights the role of maximum temperatures during the entire crop cycle, rather than in any specific phase. Other reports found that mean temperatures showed a greater association with soybean yield than maximum values. For example, in Argentina’s main soybean-growing region (i.e. the Pampean region, located south of Chaco), Andrade and Satorre (2015) found that the mean temperature in the reproductive stages of soybean was the most important determinant of yields. Soybean crops grown in the Pampean region face considerably lower mean temperatures than in Chaco and are rarely exposed to extreme temperatures and therefore to heat stress events, which could explain why the effect of temperature would be limited to the reproductive stages. Obtained results confirm previous modelling studies with DSSAT in the Semiarid/Subhumid Chaco which also found that the daily maximum temperature averaged from sowing to physiological maturity (in this case from January to April) was the best predictor of soybean yield within variables related to temperature (Casali et al., 2022).

The number of days with maximum temperatures above 35°C during the period 50-100 days after sowing (D35_50-100) (Pearson’s correlation coefficient, r= 0.47), which coincided with the phenological phases R4 -R6, emerged as the best predictor of soybean yields among the variables related to heat stress. This confirms previous findings obtained in the broader Gran Chaco Region (Madias et al., 2021) and modelling results using local climate records (Casali et al., 2022), who also identified this variable as a key predictor of local soybean yields. Other periods, such as 0-50 and 100-150 days after sowing, presented null or lower associations with yield. The agreement between the current on-farm results and previous DSSAT model simulations (Casali et al., 2022) for heat stress variables (D3550-100 in both cases) strongly supports the reliability of modeling approaches to test hypotheses and simulate scenarios in the region The identification of critical maximum temperatures is relevant not only to identify current soybean yield limiting factors but also for modeling the effects of climate change. Predicted future climate scenarios in our study site indicated an increase in daily maximum temperatures (Casali et al., 2021). Despite soybean maturity group not emerging as a significant variable in yield determination in our database, Madias et al. (2021) underscored the importance of selecting maturity groups based on the sowing date in this region. In early sowings, maturity groups VII and VIII surpassed the yields of maturity groups V and VI, while in late sowings, the latter exhibited better performance. These results contrast with Yang et al. (2007) who proposed using genotypes with a longer growing season to counterbalance the shortening of the growing season caused by high temperatures.

None of the soil variables were associated with soybean yield, likely due to which could be associated with the relative homogeneity of the soils of the area under study. For example, soil available phosphorus consistently showed values well above the critical threshold determined for soybean (14.3 mg kg⁻¹ P Bray, Sucunza et al., 2018) which rules out its role as a yield limiting factor. Another variable that also failed to appear as a predictor of soybean yield as previously expected was the sowing date. Almost 70% of the paddocks were sown between December 1^st and January 3er, that is the optimal sowing window for the study region (Madias et al., 2021), which may reflect the fact that most local farmers are already aware of the benefits of concentrating sowing in this period to avoid water and heat stresses in the critical periods of grain yield determination.

Relative to the best model (model 1), removing temperature (model 3) or heat stress (model 4) resulted in higher AIC values, but the increases were relatively small (3.9 and 23.8 units; Table 1). In contrast, the exclusion of both variables simultaneously (model 5, Table 1)) increased the AIC value considerably (3146.2 units), which is consistent with the inclusion of the interaction term D35_50-100 * TDA in the model with the best fit.The negative coefficients associated with TDA (-158 kg ha^-1; per °C) and D3550-100 (-309 kg ha^-1; per day) indicate strong sensitivity of soybean yields to heat stress. The relevance of this interaction is related to the more pronounced negative effect on soybean yields of higher average maximum temperatures (> to 32°C) at D35_50-100 greater than 14 days (Figure 5). This means that more heat stress events resulted in lower yields in years of high maximum temperatures. The analysis of the marginal R² values indicated that model 3 (Table 1), which did not include the mean maximum temperature (TDA), had a value of 0.83, while model 4, which did not include heat stress (i.e., D35) had a lower value (0.63). Considering that the marginal R² explains the variance of the fixed factors and that it was higher in the model with heat stress than in the model with temperature, this result implies that heat stress had greater explanatory power for soybean yield than maximum temperature alone, as confirmed by the coefficients of the obtained model.

5 Conclusions

The high variability in soybean yields in the study area was greatly determined by climatic variables. Although these variables cannot be controlled, understanding their impact on crop yields and their association to management variables can assist in devising strategies to reduce risks and enhance the resilience of the crop production. In such sense, this study is the first to identify the main drivers and their interactions determining soybean yield in an area characterized by frequent hydric and heat stress. We identified a maximum temperature threshold that causes heat stress events (35°C) and a tolerance threshold period for these heat stress events (14 days) after which soybean yields are strongly affected. The amount of rainfall was a decisive factor in defining yields, even a month before planting. The mixed model analysis quantified the marked negative effect of soybean as preceding crop (i.e. monoculture) which indicated the need to diversify the crops included in the agricultural rotation. Interestingly, our findings suggest that rainfall and crop rotation history work together to influence water use efficiency. Fields where soybean was avoided or planted just once in the previous five years showed greater water use efficiency.

Management strategies and research needs that emerge from the results of our on-farm research include: i) increasing crop diversification, avoiding soybean monoculture; ii) implementing agricultural practices aimed to increase pre-plant water availability (i.e. maintaining high soil coverage); iii) sowing soybean at different dates in the paddocks of the same farm so that soybean crops are exposed to different weather conditions, helping to mitigate production risks; iv) developing genotypes with higher water use efficiency and tolerance to heat stress.

Data availability statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Author contributions

LC: Formal Analysis, Validation, Methodology, Data curation, Writing – original draft, Writing – review & editing, Investigation. JH: Conceptualization, Funding acquisition, Writing – review & editing, Formal Analysis, Software, Writing – original draft, Methodology. AM: Writing – review & editing, Formal Analysis, Validation. GR: Investigation, Funding acquisition, Writing – review & editing, Methodology, Writing – original draft, Supervision, Conceptualization, Project administration.

Funding

The author(s) declare financial support was received for the research and/or publication of this article. Financial support was provided by CONICET, UBA, ANPCYT (Argentina) and Agroscope (Switzerland). We are grateful to the agronomists and farmers who contributed to building the dataset for this research, and to the members of AAPRESID and INTA for their valuable support. Open access funding by Agroscope.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declare that no Generative AI was used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

References

Andrade J. F., Poggio S. L., Ermácora M., and Satorre E. H. (2017). Land use intensification in the Rolling Pampa, Argentina: diversifying crop sequences to increase yields and resource use. Eur. J. Agron. 82, 1–10. doi: 10.1016/j.eja.2016.09.013

Crossref Full Text | Google Scholar

Andrade J. F. and Satorre E. H. (2015). Single and double crop systems in the Argentine Pampas: Environmental determinants of annual grain yield. Field Crops Res. 177, 137–147. doi: 10.1016/j.fcr.2015.03.008

Crossref Full Text | Google Scholar

Aramburu-Merlos F., Monzon J. P., Mercau J. L., Taboada M., Andrade F. H., Hall A. J., et al. (2015). Potential for crop production increase in Argentina through closure of existing yield gaps. Field Crops Res. 184, 145–154. doi: 10.1016/j.fcr.2015.10.001

Crossref Full Text | Google Scholar

Barnatán I., Díaz M., Garay N., Petino E., Stella J. L., and Veiga H. (2025). Atlas climático de Argentina (Ciudad Autónoma de Buenos Aires: Servicio Meteorológico Nacional). Available online at: http://repositorio.smn.gob.ar/handle/20.500.12160/2981 (Accessed August 25, 2025).

Google Scholar

Bates D., Maechler M., Bolker B., and Walker S. (2015). Fitting linear mixed-effects models using lme4. J. Stat. Softw 67, 1–48. doi: 10.18637/jss.v067.i01

Crossref Full Text | Google Scholar

Baumann M., Piquer-Rodríguez M., Fehlenberg V., Gavier Pizarro G., and Kuemmerle T. (2016). “Land-use competition in the South American Chaco,” in Land Use Competition: Ecological, Economic and Social Perspectives. Ed. Niewöhner J., et al (Springer, Cham), 215–229. doi: 10.1007/978-3-319-33628-2_13

Crossref Full Text | Google Scholar

Blanco-Canqui H. (2013). Crop residue removal for bioenergy reduces soil carbon pools: How can we offset carbon losses? Bioenergy Res. 6, 358–371. doi: 10.1007/s12155-012-9221-3

Crossref Full Text | Google Scholar

Burnham K. P., Anderson D. R., and Huyvaert K. P. (2011). AIC model selection and multimodel inference in behavioral ecology. Behav. Ecol. Sociobiol 65, 23–35. doi: 10.1007/s00265-010-1029-6

Crossref Full Text | Google Scholar

Casali L., Herrera J. M., and Rubio G. (2021). Modelling maize and soybean responses to climatic change and soil degradation in a region of South America. Agron. J. 113, 1–13. doi: 10.1002/agj2.20585

Crossref Full Text | Google Scholar

Casali L., Herrera J. M., and Rubio G. (2022). Resilient soybean and maize production under a varying climate in the semi-arid and sub-humid Chaco. Eur. J. Agron. 135, 126463. doi: 10.1016/j.eja.2022.126463

Crossref Full Text | Google Scholar

Coyos T., Borrás L., and Gambin B. (2018). Site-specific covariates affecting yield response to nitrogen of late-sown maize in Central Argentina. Agron. J. 110, 1544–1553. doi: 10.2134/agronj2017.09.0520

Crossref Full Text | Google Scholar

Crookston R. K., Kurle J. E., Copeland P. J., Ford J. H., and Lueschen W. E. (1991). Rotational cropping sequence affects yield of corn and soybean. Agron. J. 83, 108–113. doi: 10.2134/agronj1991.00021962008300010026x

Crossref Full Text | Google Scholar

De Ruyver R. and Di Bella C. (2019). “Climate,” in The Soils of Argentina. Ed. Rubio G., et al (Springer Nature, Cham), 27–48. doi: 10.1007/978-3-319-76853-3_3

Crossref Full Text | Google Scholar

Dominguez J. and Rubio G. (2019). “Agriculture,” in The Soils of Argentina. Ed. Rubio G., et al (Springer, Cham), 209–238. doi: 10.1007/978-3-319-76853-3_16

Crossref Full Text | Google Scholar

Fox J. and Weisberg S. (2018). “An R Companion to Applied Regression,” in Thousand Oaks (Sage Publications, CA).

Google Scholar

Gambin B. L., Coyos T. A., Di Mauro G., Borrás L., and Garibaldi L. A. (2016). Exploring genotype, management, and environmental variables influencing grain yield of late-sown maize in central Argentina. Agric. Syst. 146, 11–19. doi: 10.1016/j.agsy.2016.03.01

Crossref Full Text | Google Scholar

Giménez R., Mercau J. L., Houspanossian J., and Jobbágy E. G. (2015). Balancing agricultural and hydrologic risk in farming systems of the Chaco plains. J. Arid Environ. 123, 81–92. doi: 10.1016/j.jaridenv.2014.09.004

Crossref Full Text | Google Scholar

Goldfarb L. and Van Der Haar G. (2017). “The moving frontiers of genetically modified soy production: shifts in land control in the Argentinian Chaco,” in Soy, Globalization, and Environmental Politics in South America. Eds. Oliveira G. L. T. and Hecht S. B. (Routledge, London), 312–332.

Google Scholar

Hargreaves G. H. and Samani Z. A. (1985). Reference crop evapotranspiration from temperature. Appl. Eng. Agric. 8, 96–99. doi: 10.13031/2013.26773

Crossref Full Text | Google Scholar

Karlen D. L., Hurley E. G., Andrews S. S., Cambardella C. A., Meek D. W., Duffy M. D., et al. (2006). Crop rotation effects on soil quality at three northern corn/soybean belt locations. Agron. J. 98, 484–495. doi: 10.2134/agronj2005.0098

Crossref Full Text | Google Scholar

Kettler B. A., Nalli Sonzogni F. D., Neiff N., Boscarino S., and Maddonni G. (2025). Climatic constraints for rainfed maize production in the Northeastern subtropical region of Argentina. Theor. Appl. Climatol 156, 1–17. doi: 10.1007/s00704-025-05530-y

Crossref Full Text | Google Scholar

Lacoste M., Cook S., McNee M., Gale D., Ingram J., Bellon-Maurel V., et al. (2022). On-farm experimentation to transform global agriculture. Nat. Food 3, 11–18. doi: 10.1038/s43016-021-00424-4

PubMed Abstract | Crossref Full Text | Google Scholar

Maddonni G. A. (2012). Analysis of the climatic constraints to maize production in the current agricultural region of Argentina—a probabilistic approach. Theor. Appl. Climatol 107, 325–345. doi: 10.1007/s00704-011-0478-9

Crossref Full Text | Google Scholar

Madias A., Di Mauro G., Vitantonio-Mazzini L. N., Gambin B. L., and Borrás L. (2021). Environment quality, sowing date, and genotype determine soybean yields in the Argentinean Gran Chaco. Eur. J. Agron. 123, 126217. doi: 10.1016/j.eja.2020.126217

Crossref Full Text | Google Scholar

Madias A., Simón C. G., Stahringer N. I., Borrás L., Rubio G., and Gambin B. L. (2025). On-farm insights in the South American Gran Chaco reveal the importance of soil organic matter and crop management decisions for boosting maize yields. Eur. J. Agron. 168, 127612. doi: 10.1016/j.eja.2025.127612

Crossref Full Text | Google Scholar

Magliano P. N., Murray F., Baldi G., Aurand S., Páez R. A., Harder W., et al. (2015). Rainwater harvesting in Dry Chaco: Regional distribution and local water balance. J. Arid Environ. 123, 93–102. doi: 10.1016/j.jaridenv.2015.03.012

Crossref Full Text | Google Scholar

Nakagawa S. and Schielzeth H. (2013). A general and simple method for obtaining R2 from generalized linear mixed-effects models. Methods Ecol. Evol. 4, 133–142. doi: 10.1111/j.2041-210x.2012.00261.x

Crossref Full Text | Google Scholar

Peri P. L., Dube F., and Varella A. (2016). Silvopastoral Systems in Southern South America (Cham: Springer). doi: 10.1007/978-3-319-24107-4

Crossref Full Text | Google Scholar

Piquer-Rodríguez M., Torella S., Gavier-Pizarro G., Volante J., Somma D., Ginzburg R., et al. (2015). Effects of past and future land conversions on forest connectivity in the Argentine Chaco. Landsc. Ecol. 30, 817–833. doi: 10.1007/s10980-014-0147-3

Crossref Full Text | Google Scholar

R Core Team (2022). R: A language and environment for statistical computing (Vienna, Austria: R Foundation for Statistical Computing). Available online at: https://www.R-project.org/ (Accessed December 20, 2022).

Google Scholar

Ricard M. F., Viglizzo E. F., and Podestá G. (2015). Comparison of adaptative strategies to climate variability in rural areas of Argentine Chaco and US Southern Plains during the last century. J. Arid Environ. 123, 68–80. doi: 10.1016/j.jaridenv.2014.10.009

Crossref Full Text | Google Scholar

Smith A. B., Cullis B. R., and Thompson R. (2005). The analysis of crop cultivar breeding and evaluation trials: an overview of current mixed model approaches. J. Agric. Sci. 143, 449–462. doi: 10.1017/S0021859605005587

Crossref Full Text | Google Scholar

Sucunza F. A., Boem F. H. G., Garcia F. O., Boxler M., and Rubio G. (2018). Long-term phosphorus fertilization of wheat, soybean and maize on Mollisols: Soil test trends, critical levels and balances. Eur. J. Agron. 96, 87–95. doi: 10.1016/j.eja.2018.03.004

Crossref Full Text | Google Scholar

Vallejos M., Volante J. N., Mosciaro M. J., Vale L. M., Bustamante M. L., and Paruelo J. M. (2015). Transformation dynamics of the natural cover in the Dry Chaco ecoregion: A plot level geo-database from 1976 to 2012. J. Arid Environ. 123, 3–11. doi: 10.1016/j.jaridenv.2014.11.009

Crossref Full Text | Google Scholar

Villarino S. H., Studdert G. A., Baldassini P., Cendoya M. G., Ciuffoli L., Mastrángelo M., et al. (2017). Deforestation impacts on soil organic carbon stocks in the Semiarid Chaco Region, Argentina. Sci. Total Environ. 575, 1056–1065. doi: 10.1016/j.scitotenv.2016.09.175

PubMed Abstract | Crossref Full Text | Google Scholar

Volante J. N., Mosciaro M. J., Gavier-Pizarro G. I., and Paruelo J. M. (2016). Agricultural expansion in the semiarid Chaco: poorly selective contagious advance. Land Use Policy 55, 154–165. doi: 10.1016/j.landusepol.2016.03.025

Crossref Full Text | Google Scholar

Wilhelm W. W. and Wortmann C. S. (2004). Tillage and rotation interactions for corn and soybean grain yield as affected by precipitation and air temperature. Agron. J. 96, 425–432. doi: 10.2134/agronj2004.4250

Crossref Full Text | Google Scholar

Wright I. A., Tarawali S., Blümmel M., Gerard B., Teufel N., and Herrero M. (2012). Integrating crops and livestock in subtropical agricultural systems. J. Sci. Food Agric. 92, 1010–1015. doi: 10.1002/jsfa.4556

PubMed Abstract | Crossref Full Text | Google Scholar

Yamoah C. F., Varvel G. E., Francis C. A., and Waltman W. J. (1998). Weather and management impact on crop yield in rotations. J. Prod. Agric. 11, 219–225. doi: 10.2134/jpa1998.0219

Crossref Full Text | Google Scholar

Yang X., Lin E., Ma S., Ju H., Guo L., Xiong W., et al. (2007). Adaptation of agriculture to warming in Northeast China. Clim. Change 84, 45–58. doi: 10.1007/s10584-007-9265-0

Crossref Full Text | Google Scholar

Zuur A. F., Ieno E. N., and Elphick C. S. (2010). A protocol for data exploration to avoid common statistical problems. Methods Ecol. Evol. 1, 3–14. doi: 10.1111/j.2041-210X.2009.00001.x

Crossref Full Text | Google Scholar