Increasing Demand for Agricultural Commodities

The increasing demands of a growing world population and a change in the quality of diets have made it necessary to increase agricultural production (Andrade, 2020). The Argentinian Pampas are among the most naturally fertile regions in the world (Calviño & Monzón, 2009), and Argentina has a long tradition in extensive agriculture. Therefore, Argentina is a key player in the quest for global food security.

At the beginning of the 20th century, Argentinian producers already planted 12 million hectares of maize, wheat, and flax (Hora, 2012). The increasing demand for animal protein and biofuels during the 20th century and early 21st century further expanded the market for Argentine grains (Lamers et al., 2008). In combination with other factors, this triggered a process of expansion of grain and oilseed production.

As of 2021, Argentinian producers planted 35 million hectares, predominantly with crops of maize, soybean, and wheat (Satorre & Andrade, 2021). Commodity prices act as the main driver of Argentinian farmers’ decision-making, because the agriculture sector in Argentina is not subsidized, and therefore farmers fully take on the risks inherent to commodity markets.

Technological Advances

Agricultural systems in the Argentinian Pampas underwent significant changes during the last decades of the 20th century and the first decades of the 21st century. This was partly driven by the increased demand of most agricultural commodities produced in Argentina, but also by the evolving local economic and political context, fluctuations in climate and weather, and technological innovations (Bert et al., 2011).

The intertwined effects of these drivers induced significant changes in land use and land tenure patterns in Argentinian agricultural systems (Baldi & Paruelo, 2008). The most remarkable land use change was the expansion—westward and northward—of the area devoted to annual crops, replacing and displacing pastures and native grasslands to marginal areas (Magrin et al., 2005; Viglizzo et al., 2011).

Most of this expansion was tied to the advance of soybean cultivation, which quickly became the dominant crop in most agricultural regions. After its introduction to the country in the early 1970s, soybean production in Argentina reached 11 million tons in 1990, with 5.1 million hectares planted.

The introduction of no-tillage planting and genetically modified herbicide-tolerant soybean varieties in the early 1990s resulted in large cost reductions due to improved weed control, lower energy costs, and simplified agronomic management (Qaim & Traxler, 2005; Trigo & Cap, 2003).

These technological advances resulted in a sustained rapid expansion of soybean cultivation, and as of 2015, Argentinian producers planted 20 million hectares of the crop, yielding 55 million tons.

Organizational Innovations

An additional driver of changes in land use in Argentina has been the rapid expansion of land leasing. As of 2021, approximately half of the area cultivated in the Argentinian Pampas is managed by tenant farmers (Instituto Nacional de Estadística y Censos, 2021). Favorable economic conditions for agricultural production, together with the advent of simplified agronomic systems based on new technologies, have provided incentives for farmers to increase their production scale through land rent and the outsourcing of labor and services. However, advances in technology have also left some farmers out.

The existing evidence indicates that leased agricultural land is managed differently from owned land (Carolan, 2005). Studies of agricultural production in the Argentinian Pampas (e.g., Arora et al., 2016; Senesi et al., 2017) suggest that landowners are more likely to implement agricultural practices that can have a positive medium-term impact. Tenant farmers, on the other hand, are more likely to focus on agronomic practices that allow them to maximize immediate profits, particularly when their land leases are short term.

Given the expansion of land tenancy in Argentina’s agriculture sector, the divergence in incentives between landowners and tenants has resulted in changes in overall agronomic practices.

Favorable Climate Fluctuations

During the 20th century, central-eastern Argentina experienced some of the most significant changes in precipitation trends (Giorgi, 2002). Due to the high cost of irrigation, more than 80% of the agriculture in the Argentinian Pampas is rainfed, and the performance of crops is strongly tied to climate variations (Penalba et al., 2007; Verón et al., 2015). Accordingly, the long-frequency climate variation has contributed to large changes in land use patterns.

A steady increase in both annual and extreme precipitation has been observed in the Pampas since the 1960s (Boulanger et al., 2005; Minetti et al., 2003; Rusticucci & Penalba, 2000). Rainfall variation has been distributed unevenly through the seasonal cycle: Increased precipitation has been concentrated in late spring to summer, whereas the winter season has seen little or no change in precipitation. Furthermore, the increase in rainfall has been particularly marked near the western margin of the Argentinian Pampas, displacing westward the transition to semi-arid regions that represented the traditional boundary of rainfed agriculture and allowing for areas previously considered marginal to become arable (Berbery et al., 2006).

Diminished Soil Fertility

Agricultural crop rotation systems, such as those predominant in Argentina, show a lower capacity to sustain soil organic matter balances compared to rotation sequences that mix agricultural crops with pastures (Caride et al., 2012). In turn, within agricultural crop rotation systems, the maintenance of soil fertility (measured in terms of soil organic matter balance and nitrogen stock) is more likely to be achieved with no-tillage systems with rotations that combine different agricultural crops compared to monoculture of a single agricultural crop (Alvarez et al., 2014; Novelli et al., 2011).

Therefore, agricultural systems focused on soybean cultivation, like those used widely in Argentina, are more likely to maintain soil fertility when periodically incorporating alternative crops, such as grasses (maize and wheat) or, ideally, pasture.

However, it is worth noting that the focus of local agricultural rotation systems used in Argentina and other countries is to provide enough nutrients to maximize yields, regardless of whether those are sufficient to compensate for crop uptake (García et al., 2018). In contrast, in worldwide terms, the focus is on limiting the agricultural (and industrial) fixation of nitrogen and the inflow of phosphorus to oceans, so as to avoid transgressing planetary boundaries and causing environmental damage (Rockström et al., 2009).

Reduction and Increased Variability in Crop Yields

The evidence from long-term experimentation indicates that the monoculture of grasses (e.g., maize or wheat) or legumes (e.g., soybean) results in a reduction in crop yields compared with grass–legume crop rotations (Farmaha et al., 2016; Kelley et al., 2003). For example, in Argentina’s agriculture sector, it is considered a well-established fact that the rotation of grasses and legumes is more likely to achieve and sustain high yield levels compared to monoculture (Andrade et al., 2017).

However, the impact of crop rotation systems on yields varies across regions, and their relative profitability depends on exogenous variables such as input costs and taxation regimes. In addition, any medium-term positive impacts may not be considered relevant by producers with a short-term horizon, such as tenant farmers leasing land for a single season.

Ecological Imbalance

The introduction in the 1990s of genetically modified crops with resistance to herbicides resulted in very significant reductions in agricultural costs and greatly simplified agronomic management systems throughout the world.

However, the emergence of herbicide-resistant weeds, together with the mounting evidence regarding the negative impact of herbicide use on ecosystems, makes it necessary to develop alternative weed control strategies such as integrated weed management (IWM) (GROW, n.d.; Stokstad, 2019). Although these systems generally still feature chemical eradication solutions, they emphasize proper herbicide management practices and limit their use, favoring prevention strategies and non-chemical weed control tactics. These include mechanical weed management (e.g., mowing, harvest weed seed control, or robotic weeding), biological solutions (e.g., the introduction of bacteria, fungi, or insects that have a preference for certain weed species), and improved cultural practices (e.g., nutrient management, reduced row spacing, cover crops, or crop rotation) (Scursoni et al, 2019).

In addition, the stress on water resources induced by the monoculture or quasi-monoculture of agricultural crops can result both in increased water scarcity and in excessive soil saturation and more frequent flooding. In particular, monoculture affects, among other things, the physical fertility of soils. For this reason, water infiltration and soil storage capacity tend to be reduced. On the other hand, annual crops, and especially soybean, tend to consume less water. Studies in Argentina show an increase of groundwater levels as a consequence of this and, accordingly, a higher risk of flooding (Jobbagy et al., 2021; Nosetto et al., 2012).

Increased GHG Emissions and Vulnerability to Climate Change

Land use, including agricultural land use, is a significant net source of GHG emissions, contributing more than 20% of anthropogenic emissions of carbon dioxide (CO₂), methane, and nitrous oxide combined as CO₂ equivalents in 2007–2016. Therefore, climate change mitigation response options related to land use are a key element of most modeled viable climate scenarios (Intergovernmental Panel on Climate Change [IPCC], 2019).

In the specific context of Argentina’s agriculture soybean and maize are the main summer crops. According to Arrieta et al. (2018), the average yields per hectare for the 2012–2013 season were 2.62 tons of soybean and 6.02 tons of maize (Pampas region). In terms of energy efficiency and GHG emissions, on average 0.887 tons of soybean and 0.740 tons of maize were produced per gigajoule of energy, and 6.06 tons of soybean and 5.01 tons of maize were produced per ton of CO₂ equivalent (CO₂e) emitted (all regions). However, these figures only consider the emissions from cultivation and fertilization, without taking into account carbon changes in soil organic matter.

In overall terms, there are large differences in crop yields, GHG emission, and energy efficiency across regions in Argentina. These are likely driven by distinct climatic patterns, and in particular by differences in mean annual precipitation across regions (Calviño & Monzón, 2009). This is because agricultural production in Argentina is especially sensitive to climate and weather fluctuations. Therefore, agronomic adaptation strategies will play an essential role in increasing climate resilience. These include changing land and cropping practices, the development of improved crop varieties, and changing food consumption patterns (IPCC, 2019).

Impact of Short-Term Crop Rotation on Agricultural Yields

The impact of short-term crop rotation on agricultural productivity has been studied in many contexts using producer- and plot-level data, generally finding positive impacts. For example, Farmaha et al. (2016) find a positive impact of crop rotation on yields and resource efficiency in the U.S. corn belt when implementing maize–soybean rotations in high-yield irrigation applications.

An analogous analysis can be carried out based on the plot-level data compiled in the 2018 AACREA–IADB Integrated Crop Rotation Database (1998–2016) to estimate the impact of crop rotations in Argentina, where soybean has been the most prevalent monoculture or quasi-monoculture during the past few decades. By looking at the summer planting sequence by plot, it is possible to compare soybean yields when soybean was also cultivated during the previous planting season versus when a summer crop other than soybean was rotated in the plot or it was left fallow during the summer.

For simplicity purposes, only summer planting periods in the season are considered in the analysis (i.e., no distinction is made between a plot planted with a winter crop or left fallow during the winter). Also, for comparison robustness purposes, only the yield of early summer soybean crops (i.e., soybean crops planted early in the summer, after the plot was left fallow during the winter) are included in the analysis. This is the most common type of soybean crop planting in Argentina because late summer soybean crops (i.e., soybean crops planted later in the summer, typically following a winter crop such as wheat or barley) usually result in lower yields.

Figure 6 shows the distribution of early summer soybean yields for individual plots included in the 2018 AACREA–IADB Integrated Crop Rotation Database, relative to the average early summer soybean yield during the reference planting season in the local CREA group where the plot is located. A plot with an early summer soybean yield corresponding to the average early summer soybean yield in its local CREA group during the planting season of reference is considered to have a relative yield of 100. Comparing plot-level soybean yields with the average yield during the reference planting season in the relevant CREA local group avoids potential biases caused by season- and group-specific factors (e.g., average local soil characteristics and weather patterns, or differences in technology use and agronomic practices across seasons and CREA local groups). However, this type of comparison is not robust with respect to potential bias arising from plot- or producer-specific factors, such as those that may have affected producers’ choice to plant soybean this or the previous season (e.g., if more skilled producers are more likely to have rotated crops the previous season, or if crop rotation is more likely to be used in plots with poorer soil). Although this may be of no particular concern in the case of the 2018 AACREA–IADB Integrated Crop Rotation Database, given that the technological level and managerial skills of CREA members are likely to be very similar due to cooperation and information sharing. it may be an important consideration in other contexts.

Open in new tab

Figure 6. Distribution of early summer soybean relative yields by short-term rotation intensity in 2018 AACREA–IADB Integrated Crop Rotation Database (1998–2016).

Note: OS: other summer crop/summer fallow in Year 0 followed by early summer soybean in Year 1 (short-term rotation); SS: early or late summer soybean crop in Year 0 followed by early summer soybean in Year 1 (no short-term rotation). Distributional summary statistics: ${\overline{x}}_{\text{OS}}=101.85;{\overline{x}}_{\text{SS}}=96.29$.

Source: Authors’ compilation based on data from Asociación Argentina de Consorcios Regionales de Experimentación Agrícola (2018).

Two distributions are graphed in Figure 6: other–soybean (OS), the relative yield of early summer soybean crops in Year 1 for plots left fallow during the summer or where a summer crop different than soybean was planted in Year 0; and soybean–soybean (SS), the relative yield of early summer soybean crops in Year 1 for plots where a soybean (either early or late summer) crop was planted during the previous season. That is, OS and SS are the distributions of early summer soybean relative yields for plots with and without short-term soybean crop rotation, respectively. Both distributions feature comparable dispersions, but the distribution of early summer soybean relative yields with short-term crop rotation (OS) features a higher mean than the distribution without short-term crop rotation (SS). Although simple, this comparison already suggests a positive impact of short-term crop rotation on soybean yields.

The insights provided by the graphical comparison of relative yield distributions can be confirmed by fitting a linear regression model to the plot-level data in the 2018 AACREA–IADB Integrated Crop Rotation Database. Table 3 shows the estimated coefficients for the ordinary least squares (OLS) regression of early summer soybean relative yields y on a constant term c and an indicator variable $d_{\text{ss}}$ that equals 1 if soybean was also planted the previous season (no short-term crop rotation), and 0 otherwise (short-term crop rotation).

The estimated constant term corresponds to the mean early summer soybean relative yield with crop rotation, 101.85. This indicates that, ceteris paribus, plots with short-term crop rotation indeed have mean early summer soybean yields that are 1.85 percentage points higher than the average early summer soybean yield for their respective local group during the reference planting season.

The estimated regression coefficient for $d_{\text{ss}}$, –5.56, is significant with 99% confidence. This indicates that, ceteris paribus, plots without short-term crop rotation have mean early summer soybean relative yields 5.56 percentage points lower than plots with short-term crop rotation.

Although this analysis continues to be very simple (e.g., ignoring any potential omitted variable bias due to plot-level characteristics), it suggests that the positive impact of short-term crop rotation on soybean yields can be quite large in agronomic terms.

Impact of Medium-Term Crop Rotation on Agricultural Yields

If sufficient historical data are available, the analysis of the impact of crop rotation on agricultural productivity can be expanded to estimate the impact of medium-term crop rotation over several planting seasons.

Figure 7 shows the distribution of early summer soybean relative yields in the reference (last in the sequence) planting season for plots in the 2018 AACREA–IADB Integrated Crop Rotation Database (1998–2016) for which data are available over periods of four or more consecutive planting seasons, ending in a season during which early summer soybean was planted. A plot with an early summer soybean yield corresponding to the average early summer soybean yield in its local CREA group during the last planting season (in the last consecutive period of four or more seasons) is considered to have an early summer soybean relative yield of 100.

Open in new tab

Figure 7. Distribution of early summer soybean relative yields by medium-term rotation intensity in 2018 AACREA–IADB Integrated Crop Rotation Database (1998–2016).

Note: Distribution of early summer soybean relative yields for [0, 20] = plots producing a summer crop other than soybean or left fallow during the summer in up to 20% of the last consecutive planting seasons; (20, 40] = plots producing a summer crop other than soybean or left fallow during the summer in more than 20% and up to 40% of the last consecutive planting seasons; (40, 60] = plots producing a summer crop other than soybean or left fallow during the summer in more than 40% and up to 60% of the last consecutive planting seasons; (60, 100] = plots producing a summer crop other than soybean or left fallow during the summer in more than 60% of the last consecutive planting seasons. Distributional summary statistics: ${\overline{x}}_{\left[0,20\right]}=93.97;{\overline{x}}_{(20,40]}=97.69;{\overline{x}}_{(40,60]}=104.59;{\overline{x}}_{(60,100]}=106.15$.

Source: Authors’ compilation based on data from Asociación Argentina de Consorcios Regionales de Experimentación Agrícola (2018).

Given that soybean was the most prevalent crop in Argentina during the time period covered in the 2018 AACREA–IADB Integrated Crop Rotation Database (1998–2016), a medium-term crop rotation intensity variable can be constructed based on the percentage of the last consecutive seasons during which a summer crop other than soybean was planted in a plot or it was left fallow during the summer. This intensity variable can then be converted into a multinomial categorical variable or into a set of binomial indicator variables corresponding to each medium-term crop rotation intensity interval.

For example, the indicator variable [0, 20] takes a value of 1 if a summer crop other than soybean was planted in the corresponding plot or it was left fallow during the summer in up to 20% of the last consecutive planting seasons, and a value of 0 otherwise (i.e., if early or late summer soybean was planted during more than 20% of the last consecutive planting seasons, such as in sequence of five planting seasons during which maize was planted instead of soybean during one season, or during which early or late summer soybean was planted in all five seasons). Analogously, the indicator variable (20, 40] takes a value of 1 if a summer crop other than soybean was planted in the corresponding plot or it was left fallow during the summer in more than 20% of the last consecutive planting seasons up to 40%, and a value of 0 otherwise (i.e., if early or late summer soybean was planted during more than 40% of the last consecutive planting seasons, such as in sequence of five planting seasons during which maize was planted instead of soybean during two seasons). The length of the intervals could be chosen arbitrarily, but narrow intervals may cause data sparsity problems, whereas wide intervals may be uninformative. In the case of Argentina, the interval [0, 20] captures plots devoted to soybean cultivation with no or limited long-term crop rotation. The intervals (20, 40] and (40, 60] capture the two crop rotation intensities that are generally considered feasible by soybean producers in Argentina. Finally, the interval (60, 100] captures plots generally devoted to the cultivation of other crops, with limited rotation of soybean cultivation.

As in the analysis of short-term rotations, for simplicity purposes, only summer planting periods in the season are considered in the analysis (i.e., no distinction is made between a plot planted with a winter crop or left fallow during the winter). Similarly, for comparison robustness purposes, only the yield of early summer soybean crops (i.e., soybean crops planted early in the summer, after the plot was left fallow during the winter) are included in the analysis—that is, all the sequences of four or more consecutive seasons included in the analysis end in a season during which early summer soybean was planted.

Four distributions are graphed in Figure 7, corresponding to plots in each of the four defined medium-term crop rotation intensity intervals: [0, 20], (20, 40], (40, 60], and (60, 100]. As in the case of short-term crop rotation intensity, all four early summer soybean relative yield distributions feature comparable dispersions, but mean early summer soybean productivity during the season of reference increases with the intensity of medium-term crop rotation during the previous seasons.

Analogous to the case of short-term crop rotations, the insights provided by the graphical comparison of relative soybean yield distributions can be confirmed by fitting a linear regression model to the plot-level data in the 2018 AACREA–IADB Integrated Crop Rotation Database (1998–2016).

Table 4 shows the estimated coefficients for the OLS regression of early summer soybean relative yields y on a constant term c and the crop rotation intensity variable r (i.e., the percentage of the last four or more consecutive seasons for which data are available during which a summer crop other than soybean was planted in the plot or it was left to fallow during the summer). The estimated regression coefficient for r, 0.22, is significantly different from 0 with 99% confidence. This indicates that, ceteris paribus, an additional season of crop rotation in four seasons is estimated to result in early summer soybean yields that are 5.5 percentage points higher than the average early summer soybean yield for the respective local group during the reference planting season.

This result is very similar to the estimated impact of short-term crop rotation previously obtained by comparing only two seasons. However, the estimated impact of crop rotation on early summer soybean relative yields is even larger when the possibility to rotate during several seasons in the medium term is taken into account: Table 4 also shows the estimated coefficients for the OLS regression of early summer soybean relative yields y on a constant term c and binomial indicator variables $d_{\left(20,40\right]},d_{(40,60]}$, and $d_{\left(60,100\right]}$, which equal 1 if a plot belongs to the corresponding medium-term intensity crop rotation interval, and 0 otherwise. The results indicate that, ceteris paribus, the mean early summer soybean relative yields for plots in which a summer crop other than soybean was planted or which were left fallow during the summer in more than 20% and up to 40% of the last consecutive seasons (i.e., 60% or more, but less than 80% soybean cultivation, moderate crop rotation) are 3.72 percentage points higher than the mean early summer soybean relative yields for plots in which a summer crop other than soybean was planted or which were left fallow during the summer in less than 20% of the last consecutive seasons (i.e., 80% or more soybean cultivation, low or no crop rotation). But, ceteris paribus, the mean early summer soybean relative yields for plots in which a summer crop other than soybean was planted or which were left fallow during the summer in more than 40% and up to 60% of the last consecutive seasons (i.e., 40% or more, but less than 60% soybean cultivation, balanced crop rotation) are estimated to be 10.62 percentage points higher, whereas the mean early summer soybean relative yields for plots in which a summer crop other than soybean was planted or which were left fallow during the summer in more than 60% of the last consecutive seasons (i.e., less than 40% soybean cultivation, predominance of other crops) are estimated to be 12.18 percentage points higher.

This analysis continues to rely on a very simple identification strategy, but it suggests that the positive productivity impact of balanced crop rotation over many seasons in the medium term can be even larger than that of short-term crop rotation over just two seasons.

Panel Data Analysis

So far, the analysis of the impact of crop rotation on soybean productivity has only attempted to avoid potential biases caused by season- and group-specific factors, by examining (early summer) soybean yields relative to the average productivity of plots in the relevant local group during the reference planting season. However, the panel structure of the 2018 AACREA–IADB Integrated Crop Rotation database allows the use of more sophisticated analysis strategies to address potential omitted variable bias arising from non-observable time-invariant plot characteristics (e.g., non-observable soil quality or non-observable producer ability).

Figure 8 summarizes the results of estimating the impact of medium-term crop rotations on relative soybean yields using three panel data models, which leverage the fact that the data set includes observations corresponding to medium-term planting sequences of the same plots at different points in time: (a) A random effects model assumes that time-invariant plot-specific characteristics (i.e., effects) are random variables (orthogonal to any time-varying explanatory variables included in the analysis); (b) a within model assumes that time-invariant plot-specific characteristics are fixed (i.e., non-random) quantities, and it controls for this by examining the deviations from the overall plot-level mean values over time; and (c) a first-difference model also assumes that time-invariant specific characteristics are fixed, but it controls for this by instead examining the evolution of plot-level values from one time period (i.e., planting sequence) to the next. In addition, a pooling model is also considered: This simply pools all observations together in the estimation—even observations that correspond to the same plot at different points in time.

Open in new tab

Figure 8. Panel data estimates of impact of medium-term rotation intensity on early summer soybean relative yields in 2018 AACREA–IADB Integrated Crop Rotation Database (1998–2016).

Note: The y-axis represents the estimated increase in mean early summer soybean relative yields during the last season in a medium-term planting sequence (four or more consecutive planting seasons). The pooling model ignores the panel structure of the data. The random effects model assumes that time-invariant plot-specific effects are random variables. The within model is the standard fixed effects model, where time-invariant plot-specific characteristics are assumed to be fixed quantities, and unbiasedness is achieved by controlling for deviations from the overall plot-level mean values over time. Finally, the first-difference (fd) model also assumes fixed effects, but unbiasedness is achieved by instead controlling for the evolution of plot-level values from one planting sequence to the next.

Source: Authors’ analysis based on data from Asociación Argentina de Consorcios Regionales de Experimentación Agrícola (2018).

Given that the pooling model ignores the panel structure of the data, its results are comparable to those obtained so far. The results of estimating a random effects model are also comparable, but this estimation will only be consistent if time-invariant plot characteristics are indeed orthogonal to any time-varying explanatory variables. Otherwise, the results of the within or first-difference fixed effects models suggest that the positive impact of medium-term crop rotation on productivity may be even larger than those estimated so far, with balanced (40–60%) medium-term rotation resulting in an estimated 15–20 percentage point increase in mean (early summer) soybean relative yields compared to no or low medium-term crop rotation intensity (0–20%).

For a detailed discussion of econometric panel data models, see Wooldridge (2010).

Impact of Short-Term Crop Rotation on Overall Agricultural Resilience

Figure 9 graphs early summer soybean absolute yields with and without short-term crop rotation for plots included in the 2018 AACREA–IADB Integrated Crop Rotation Database (1998–2016), grouped by CREA local group. Each observation represents the average early summer soybean absolute yield (kg/ha) of plots in a CREA local group for which early summer soybean cultivation in the reference planting season was preceded by soybean (either early or late summer) or by a summer crop other than soybean or summer fallow in the planting season sequence. That is, each observation corresponds to the average early summer soybean yield during a single planting season of two sets of plots managed by producers in a CREA local group: those for which early summer soybean cultivation during the corresponding planting season was preceded by soybean cultivation (either early or late summer) during the previous planting season (x-axis, average CREA local group early summer soybean absolute yield without short-term crop rotation) and those for which soybean cultivation during the same planting season was preceded by cultivation of a summer crop other than soybean or summer fallow during the previous planting season (y-axis, average CREA local group early summer soybean absolute yield with short-term crop rotation).

Open in new tab

Figure 9. Average soybean yields with and without short-term crop rotation, by CREA local group and planting season in 2018 AACREA–IADB Integrated Crop Rotation Database (1998–2016).

Open in new tab

y = Average Early Summer Soybean Absolute Yields (kg/ha) by CREA Local Group and Planting Season	OLS Coefficient	p Value
Constant	142.69 (75.96)	0.061
1 = short-term rotation	0.90 (0.02)	0.000

Open in new tab

Note: Analysis excludes observations (unique local group-planting season combinations) comprising less than 20 plots in either the with or the without crop rotation sets. Linear regression is unweighted by observation size (i.e., the linear regression does not take into account the number of plots included in each observation). The estimated regression model is $y_{ijt}=c+{\beta }_{\text{OS}}\times d_{\text{OS}ijt}+e_{jt}$, where $y_{ijt}$ denotes the average early summer soybean absolute yield of plot set $i$ in CREA local group $j$ during planting season $t$, and $d_{\text{OS}ijt}$ is an indicator variable equal to 1 if plot set $i$ in CREA local group $j$ during planting season $t$ corresponds to a short-term rotation (other-soybean) planting sequence, and 0 otherwise. A $t$-test rejects the null hypothesis of equality of average absolute yields with and without short-term crop rotation ($p$ value = 0.000).

Source: Authors’ analysis based on data from Asociación Argentina de Consorcios Regionales de Experimentación Agrícola (2018).

Therefore, observations in Figure 9 only capture the average absolute yield of plots for which early summer soybean was cultivated during the reference season. Also, each observation represents a unique combination of planting season and CREA local group: Several observations may correspond to the same CREA local group, provided that they correspond to distinct planting seasons. In addition, the diameter of each observation in Figure 9 is proportional to the number of plots included in that observation, relative to the total number of plots included in the analysis. That is, a larger diameter means that the local CREA group corresponding to that observation cultivated soybean in more plots during the corresponding season.

The linear regression analysis of average early summer soybean absolute yields suggests that short-term crop rotation, from one season to the next, had a positive but uneven impact on soybean yields in Argentina. In particular, the estimated y-intercept of the linear regression line, 142.69, is greater than 0, but its estimated slope, 0.90, is less than 1. This indicates that although the overall short-term impact of crop rotation is positive, ceteris paribus, the average impact of short-term crop rotation is smaller for higher yield season–local group combinations (observations up and to the right in Figure 9). In other words, the average positive impact of short-term crop rotation appears to be larger in less favorable season–local group contexts where yields are lower (observations down and to the left in Figure 9): Whereas the t-test null hypothesis of equality in average yields with and without short-term crop rotation is rejected with 99% confidence, at 95% confidence the difference is not significant for season–local group combinations with average yields in excess of 4,000 kg/ha or higher (i.e., starting approximately at that point, the linear regression line’s 95% confidence area overlaps with the 1:1 yield line).

This basic result suggests that the positive impact of short-term crop rotation on soybean yields may be larger in adverse conditions so that crop rotation may contribute to increase the overall resilience of Argentina’s agricultural sector.

Impact of Short-Term Crop Rotation on Agricultural Climate Resilience

What about climate resilience in particular? In order to address this more specific question, the analysis of the overall impact of short-term crop rotation on relative soybean yields can be replicated for different climate scenarios, such as the phases of the ENSO phenomenon. In the context of Argentina, the El Niño phase of the ENSO phenomenon is generally considered to bring more favorable (summer) agricultural conditions than the neutral ENSO phase. Conversely, the La Niña ENSO phase is generally considered to bring less favorable (summer) agricultural conditions than the neutral ENSO phase. Therefore, the analysis of the impact of short-term and medium-term crop rotation during distinct ENSO phases allows the quantification of the effectiveness of crop rotation as a climate adaptation tool.

Analogous to Figure 6, Figure 10 graphs the distributions OS (the early summer relative soybean yields for plots where a summer crop different than soybean was planted or which were left fallow during the summer in the previous season—that is, with short-term crop rotation) and SS (the early summer soybean relative yields for plots where soybean was planted during the previous season—that is, without short-term crop rotation). However, Figure 10 further disaggregates the distributions by ENSO phase, graphing separately the early summer soybean relative yields for planting seasons during the neutral, La Niña, and El Niño ENSO phases.

Open in new tab

Figure 10. Distribution of early summer soybean relative yields by short-term rotation intensity and ENSO phase in 2018 AACREA–IADB Integrated Crop Rotation Database (1998–2016).

Note: OS: other summer crop/summer fallow in Year 0 followed by early summer soybean in Year 1 (short-term rotation); SS: early or late summer soybean crop in Year 0 followed by early summer soybean in Year 1 (no short-term rotation). Distributional summary statistics: ${\left[{\overline{x}}_{\text{OS}},{\overline{x}}_{\text{SS}}\right]}_{\text{neutral}}=\left[101.34,95.08\right];{\left[{\overline{x}}_{\text{OS}},{\overline{x}}_{\text{SS}}\right]}_{\text{Niña}}=\left[103.54,95.36\right];{\left[{\overline{x}}_{\text{OS}},{\overline{x}}_{\text{SS}}\right]}_{\text{Niño}}=\left[101.72,97.26\right]$.

Source: Authors’ compilation based on data from Asociación Argentina de Consorcios Regionales de Experimentación Agrícola (2018).

As it can be observed in Figure 10, the distribution of early summer relative soybean yields features more dispersion during the neutral and La Niña ENSO phases compared to the El Niño ENSO phase. Most important, short-term crop rotation appears to have a large positive distributional impact during the La Niña ENSO phase compared to the neutral and El Niño ENSO phases.

Again, the insights provided by the graphical comparison of early summer soybean relative yield distributions can be confirmed by fitting a linear regression model to the plot-level data in the 2018 AACREA–IADB Integrated Crop Rotation Database (1998–2016). Analogous to Table 3, Table 5 shows the estimated coefficients for the OLS regression of early summer soybean relative yields y on a constant term c and an indicator variable $d_{\text{ss}}$ that equals 1 if soybean was also planted the previous season (no short-term crop rotation) and 0 otherwise (short-term crop rotation). However, in Table 5, the estimation results are further disaggregated by ENSO phase.

As it can be observed, ceteris paribus, the estimated mean positive impact of short-term crop rotation on early summer soybean relative yields is almost double during the less favorable La Niña ENSO phase compared to the more favorable El Niño ENSO phase (8.18 versus 4.46 percentage points, respectively).

This still basic comparison suggests that the positive impact of short-term crop rotation on soybean yields may be considerably larger in unfavorable climate conditions, such as the drier La Niña ENSO phase in Argentina.

Impact of Medium-Term Crop Rotation on Agricultural Climate Resilience

The analysis of the overall impact of medium-term crop rotation on early summer soybean relative yields can also be replicated for the different phases of the ENSO phenomenon.

Analogous to Figure 7, Figure 11 shows the distribution of early summer soybean relative yields by medium-term crop rotation intensity intervals. However, analogous to Figure 10, Figure 11 further disaggregates the distributions, graphing separately the relative soybean yields for planting seasons during the neutral, La Niña, and El Niño ENSO phases.

Open in new tab

Figure 11. Distribution of early summer soybean relative yields by medium-term rotation intensity and ENSO phase in 2018 AACREA–IADB Integrated Crop Rotation Database (1998–2016).

Note: Distribution of early summer soybean relative yields for [0, 20] = plots producing a summer crop other than soybean or left fallow during the summer in up to 20% of the last consecutive planting seasons; (20, 40] = plots producing a summer crop other than soybean or left fallow during the summer in more than 20% and up to 40% of the last consecutive planting seasons; (40, 60] = plots producing a summer crop other than soybean or left fallow during the summer in more than 40% and up to 60% of the last consecutive planting seasons; (60, 100] = plots producing a summer crop other than soybean or left fallow during the summer in more than 60% of the last consecutive planting seasons. Distributional summary statistics: ${\left[{\overline{x}}_{[0,20]},{\overline{x}}_{(20,40]},{\overline{x}}_{(40,60]},{\overline{x}}_{(60,100]}\right]}_{\text{neutral}}=\left[90.43,96.24,105.04,105.54\right]$; ${\left[{\overline{x}}_{[0,20]},{\overline{x}}_{(20,40]},{\overline{x}}_{(40,60]},{\overline{x}}_{(60,100]}\right]}_{\text{Niña}}=\left[91.75,101.8,112.93,112.55\right]$; ${\left[{\overline{x}}_{[0,20]},{\overline{x}}_{(20,40]},{\overline{x}}_{(40,60]},{\overline{x}}_{(60,100]}\right]}_{\text{Niño}}=\left[99.1,98.68,101.33,103.09\right]$.

Source: Authors’ compilation based on data from Asociación Argentina de Consorcios Regionales de Experimentación Agrícola (2018).

As it can be observed in Figure 11, the distribution of early summer soybean relative yields for all medium-term crop rotation intervals again features more dispersion during the neutral and La Niña ENSO phases compared to the El Niño ENSO phase. Also, as in the case of short-term crop rotation, medium-term crop rotation appears to have a large positive distributional impact during the La Niña ENSO phase compared to the neutral and El Niño ENSO phases. This is particularly noticeable for moderate (20–40%) and balanced (40–60%) medium-term rotation intensities.

Analogous to Table 4, Table 6 shows the estimated coefficients for the OLS regression of early summer soybean relative yields y on a constant term c and (a) the crop rotation intensity variable r (i.e., the percentage of the last four or more consecutive seasons for which data are available in which a summer crop other than soybean was planted in the plot or it was left fallow during the summer) and (b) binomial indicator variables $d_{\left(20,40\right]},d_{(40,60]}$, and $d_{\left(60,100\right]}$, which equal 1 if a plot belongs to the corresponding medium-term intensity crop rotation interval and 0 otherwise. However, in Table 6 the estimation results are further disaggregated by ENSO phase.

The estimated regression coefficients indicate that although moderate medium-term crop rotation (20–40%) did not have a significant impact during El Niño ENSO phases, it did have a large significant impact during neutral and La Niña ENSO phases: Compared to no or low medium-term crop rotation intensity, ceteris paribus, the estimated mean early summer soybean relative yields with moderate medium-term crop rotation during neutral and La Niña ENSO phases were 5.81 and 10.05 percentage points higher, respectively. The impact is even more pronounced for balanced medium-term crop rotation (40–60%): Compared to no or low medium-term crop rotation, ceteris paribus, the estimated mean early summer soybean relative yields with balanced medium-term crop rotation during neutral and La Niña ENSO phases were 14.61 and 21.18 percentage points higher, respectively. However, the impact of balanced medium-term crop rotation during El Niño ENSO phases was estimated at just 2.23 percentage points compared to no or low medium-term crop rotation. Also, the impact of increased medium-term crop rotation intensity on mean early summer soybean relative yields appears to taper off as the cultivation of other crops becomes dominant (60–100% medium-term crop rotation).

Although with the available data precision in the estimation is low, resulting in wide confidence intervals for the coefficients, these results suggest that moderate (20–40%) and, especially, balanced (40–60%) medium-term crop rotation may be a very effective tool to increase climate resilience in Argentina’s agriculture sector, softening the impact of unfavorable climate conditions such as the drier Neutral and La Niña ENSO phases.