Toward neutral carbon integrated production y silvopastoral systems: an Uruguayan study case modeled using agent-based technique

doi:10.21203/rs.3.rs-3640428/v1

Download PDF

Research Article

Toward neutral carbon integrated production y silvopastoral systems: an Uruguayan study case modeled using agent-based technique

https://doi.org/10.21203/rs.3.rs-3640428/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

The aim of this study was to assess management practices, productivity, and greenhouse gas outcomes using multi-agent modeling of a livestock system incorporating agroforestry modules. The agent-based model simulates the dynamics of animal growth, forage, and tree vegetation in several scenarios. Simulation results were evaluated over 10-year periods using a case study. Rotational grazing and forestry presence in grazing paddocks induce changes in system-wide interactions and productivity. That combined production allows for adjusting animal stocking rates and achieving a balance between forage availability and cattle demand, considering also tree-cattle interactions. Through stocking rate management, a balance between forage supply and demand can be achieved, with an improvement in net greenhouse gas reduction as livestock weight gain increases. In our study case, carbon-neutral beef production is achieved with a ratio of 380 steers at 606 ha grazing area with 13% forested surface. Estimated greenhouse gas emission intensity exhibits an inverse exponential relationship with weight gain. The agent-based modeling is a valuable tool for understanding the complexity of these systems, generating emergent properties. The study underscores the importance of implementing sustainable livestock practices and integrating agroforestry systems as a mitigation strategy.

ABM

simulation model

carbon neutral meat

Silvopastoral systems (SPS) represent a sustainable form of production that combines livestock farming with the presence of trees. These systems offer both environmental and economic benefits, including climate change mitigation through carbon sequestration (Hanisch et al. 2019). However, the potential for Carbon fixation in these systems will depend primarily on the characteristics of the herbaceous understory (metabolism, species, shade tolerance, etc.), as well as the density, spatial distribution, and age of woody species, and the animal stocking rate of the system (Jackson & Ash, 1998).

Uruguay is a traditional grass extensive livestock-based economy-based country according to Uruguay XXI national agency (Uruguay XXI 2023), where integration of forestry at farm level represents an opportunity to farmers to improve incomes and also animal welfare (Pastorini 2017). According to Scoz (2015) since the early 1990s, due to Law 15939 declaring forest resources a matter of national interest and promoting the development of the forest economy, there has been a significant increase in the planted area with fast-growing species. Today, this area covers above one million ha, with the Eucalyptus spp. genus accounting for over 70% of planted species.

Within this context Uruguay has committed to achieving the goal of net-zero greenhouse gas (GHG) emissions before 2050. The livestock sector is in the process of developing a national strategy for the promotion of climate-smart livestock production. This strategy aims to significantly reduce the greenhouse gas emissions balance in the sector (Uruguay XXI 2023).

In general traits, despite the agricultural sector being highly scrutinized due to GHG emissions from livestock (FAO 2021), there is a lack of national studies that assess the long-term evolution of productive and environmental variables when combining forestry plantations with livestock production.

Considering the combined production times from daily animal management to the completion of a timber harvest cycle, in addition to the complexity of estimating GHG modeling emerges as a fundamental tool for projecting the system in the long term and its evaluation. In this regard, agent-based modeling (ABM; Railsback & Grimm 2019) allows the simulation of complex systems where spatial and temporal management of system agents and heterogeneity among its, as well, are a center of interest. With this approach and considering SPS as an agroecosystem with its components and interactions (Sarandón & Flores 2014), the ABM approach is also recommended to emulate ecological interactions (Grimm et al. 2005).

The objective of this study was to evaluate the long-term trajectory of sustainable livestock carrying capacity, livestock emissions, woody biomass fixation, and cattle greenhouse gas emission and biomass fixation in a silvopastoral system. A case study simulates the forest component with different animal stocking rate, and the carbon assessment was estimated using the ABM platform. Through this, the temporal trajectory of management practices, such as rotational grazing and soil use, interactions with afforestation, productivity, and carbon balance, is obtained. For this purpose, an ABM model based on producer information was developed to find the equilibrium point between animal emissions and carbon sequestration by woody biomass. Different animal management alternatives (stocking rate adjustment and optimal stocking rate animal, among others) can result in carbon-neutral meat production systems. With this approach, the aim is to produce meat with very low or zero net GHG emissions, through mechanisms such as feed type or offsetting emissions with woody biomass production.

The study was carried out between 2020 and 2022, considering interviews with the farm manager and incorporating various sources of system of information (electronic spreadsheet records of productive information, georeferenced information of soil use, and others). Using this information, an ABM was implemented in the NetLogo v6.2 platform (Wilensky & Stroup 1999) to simulate the production system. The modeled farm is located in the Department of Florida (Uruguay) at the Cristaline shield region of the country (Panario 1988) and covers a total area of 2,018 hectares. However, for this study, only a grazing area of 606 hectares, predominantly consisting of native pasture (Perez-Rocha 2020), was considered, while the remaining area is dedicated to agriculture and was not included in the developed model.

The analyzed system is a cattle farm with a pastoral base that engages in a full-cycle operation (breeding-finishing system). The breeding herd is in another part of the farm, and the developed model only includes the pasture-fattening phase of steers, which are then finished in a feedlot. The producer has also established forest plantations with the primary objective of providing shade and shelter for the livestock. Additionally, the aim is to obtain wood for commercial purposes when the harvesting cycle is reached.

The study area includes Eucalyptus dunnii forest plantations at two different planting densities distributed across three paddocks. On one hand, two dense stands occupying an effective planting area of 50.4 ha, with a density of 1,100 plants/ha (called frame 1). On the other hand, there is a stand covering an area of 26.6 ha, with a planting spacing of (2 x 3) + 18 m, resulting in an initial density of 526 plants/ha (called frame 2). Figure 1 shows the satellite photo sketch and its implementation in the NetLogo platform; tree plantations in frame 2 are located in paddock number 2, while stands in paddocks number 5 and 6 are at frame 1, which are denser. All plantations were established between 2017 and 2018, except an afforested paddock (number 4) with more than 30 years old plantation, and it was assumed without growing in simulation periods.

As observed in Fig. 1, the study area encompasses a total of 31 paddocks where the fattening of steers occurs, following a defined paddock rotation. Steers are fattened on pastures, entering with a weight range of 350 to 450 kg/animal. The property currently manages a total of 420 animals. At next paragraphs, a description of the ABM model, applying an adaptation of the ODD protocol (Grimm et al. 2013) is provided.

2.1 General Description

In general terms, the ABM integrates two pre-existing local developed simulation models: i) the dynamics of pasture growth and steer fattening, based on the prey-predator model (Dieguez & Fort 2017), which simultaneously considers pasture growth and animal grass intake densodependant, estimating a daily forage balance, and ii) the dynamics of tree growth were implemented using the SAG growth projection model (Methol 2008; Hirigoyen & Rachid 2014; Rachid-Casnati et al. 2019), based on inventory data collected in the case study.

With dendrometric results, the annual biomass production of the stem was estimated using SAG model, and the estimation of branch biomass is obtained from the Hirigoyen et al. (2021) function. For the estimation of carbon in woody biomass, a parameter of 21.87% based on the density of E. dunnii (Doldán et al., 2008) and Carbon in sampled biomass of E. grandis (Schinato 2022) was applied. Annual values fixed by biomass and annual greenhouse gas emissions from livestock are converted to the common unit of CO_2e, according to the Intergovernmental Panel of Climatic Change (IPCC 2006, 2019) factors.

The model structure and dynamics (number of animals, animal and forestry management measures, reference of animal liveweight gains; LWG) were adjusted based on information provided by the system manager. Data collection involved electronic records, property sketches, and in-person and video conference interviews conducted in 2021 and 2022. Additionally, the model considers a decision subsystem representing pasture management and the entry and exit of animal lots to the system.

The ABM aims to simulate the interaction between animals, forage, and trees in an SSP system, where livestock production is combined with sustainable forest management, from the perspective of carbon-neutral production systems. The model focuses on assessing the impact of management strategy on primary (forage and forestry) and secondary (livestock) productivity, considering the dynamics of grass, animal liveweight (LW) coevolution, and tree growth, all in interaction. In addition to productive variables, GHG emissions by animals and carbon fixation by forest plantations are calculated according to IPCC (2006, 2019) standards.

2.2 Design Concepts

The model includes three types of agents: animals (steers), pasture (patches grouped in paddocks), and trees with two planting densities, one dense and the other with defined alleys, referred to as frames 1 and 2, respectively (see Fig. 1). Each agent has specific attributes: steers have a LW and LWG, the forest has stem and branch volume, and other dendrometric variables such as the diameter at breast height (DBH), basal area (BA) and projected height (H) provided by the SAG submodel. The pasture is described by its dry matter (DM) production.

2.3 Agent Behavior

The animals move through the paddocks following their numerical sequence and directly harvesting forage biomass from the patch they are in. Their LWG depends on the available biomass in the patch they stay at the forage rotation, with a Holling type III consumption, and pasture grows with a logistic function (Pastor 2008) adjusted using satellite remote sensing data (Baeza et al. 2010) for the case study. Trees grow using the equations of the SAG model (Methol 2008; Hirigoyen & Rachid 2014; Rachid-Casnati et al. 2019), calculating stem and branch volume. Both volume, growth and also annual loss of specimens are functions of planting density, according to SAG model estimation.

2.4 Interaction Between Agents

The main animal-pasture interaction is the grass intake and is defined according to the PPGL model (Dieguez & Fort 2017) for prey-predator or Lotka-Volterra dynamics (Pastor 2008). Trees cast shade, which positively affects LWG according to the SimForGan model (Varela 2019).

2.5 State Variables

State variables include the quantity of available pasture, tree density, energy consumed by animals, animal LW, as well as pasture and tree growth rates. GHG by animals and CO₂e fixation by trees are also calculated, combining Tier 2 methodologies proposed by IPCC (2006, 2019). Carbon fixation is considered only in above-ground biomass (branch and branches), with no consideration of underground biomass. Carbon fixation of pasture biomass is also not considered.

2.6 Events

Events include pasture and tree growth and the daily movement of animals. Another event is the exportation (and incorporation) of lots of 20 animals, according to established weight limits (initial weight and target weight). Sequence of movements of animals and LW thresholds were provided by the farmer.

2.7 Details

The model was implemented using the NetLogo platform (Wilensky & Stroup 1999). The Time and GIS extensions were used to adjust the time step (daily tick) to the Gregorian calendar and the distribution of patches based on georeferenced paddock location georeferenced information.

2.8 Initialization

The total number of animals managed simultaneously is 420 steers, as farmer records stated. The location and size of paddocks, as well as those including forests, are defined based on information provided by the producer (see Fig. 1).

2.9 Parameters

The following parameters were included in the model calculations: native grass pasture digestibility 50% (Miller & Albicette 2005), pasture crude protein content 9%, estimated from study samples. For GHG emission calculation, a Y_m value of 0.065 was considered. The maintenance and activity coefficients of animals (Cfi and Ca) were 0.322 and 0.17 MJ/day/kg, respectively, and the conversion factors for Carbon, CH₄, and N₂O were 3.67, 28 and 273, respectively. The parameter for animal nitrogen retention was set at 0.07 (IPCC 2006, 2019). The annual pasture productivity in the case study, based on remote sensing, was 4,558 kg DM/ha/year, with seasonal growth rates of 11, 6, 16, 17 kg DM/ha/day for autumn, winter, spring, and summer, respectively, according to satellite remote sensing data (Baeza et al. 2010). These values were used to define the K_max and K_min values (22.0 and 5.3 cm/ha, respectively) for logistic growth, required for the PPGL model (Dieguez & Fort 2017). The conversion of DM volume and grass height considered was 300 kg DM/cm (Do Carmo et al. 2015), and a metabolizable energy value of 2 Mcal/kg DM for the pasture (Mieres et al. 2004). The effect of grazing in paddocks with forest plantations increases LGW by 10% for both planting frames, based on previous studies (Varela 2019).

2.10 Model Rules

The livestock management decisions were established by the farmer, considering steers exiting the system in batches when at least 20 animals reach the target LW (450 kg/animal). Upon departure, the batch is replaced with the same number of steers but with the initial LW (350 kg/animal), and this batch begins the fattening process again. Another rule defined by the producer is that grazing is conducted sequentially in numerical paddock order (e.g. 1, 2, 3..31, 1, etc). In terms of grazing decisions, the technical details of moving animals according to DM allowance were discussed with the farmer and generally align with local technical advice on extensive grassland systems management (Do Carmo et al. 2015). However, some DM thresholds were discussed with the farmer to introduce an objective parameter into the simulation model. The daily decision was modeled such that if the next paddock in the numerical sequence to which animals will be assigned does not have an offer of 2,000 kg DM/ha, they are assigned to the paddock with the highest biomass available in the system. Animals remain in the paddock until a remaining amount of 1,500 kg DM/ha is reached, after which the numerical sequence of paddocks continues, adhering to the previously mentioned rules. Paddocks with forests are also included in the rotation."

2.11 Statistical analysis

To assess the model's coherence, its results were compared separately with the original models incorporated in the Agent-Based Model (ABM). While this comparison may seem redundant, it is essential due to the potential bias introduced by spatial and individual heterogeneity among agents. Therefore, a regression analysis between the ABM and the other reference models is warranted. Specifically, we compared the animal component of the ABM with PPGL simulations, considering parameters such as LW and LWG of animals over a 10-year simulation period. Additionally, the results related to afforestation were compared with the SAG model, focusing on the annual evolution of the number of trees, basal area, diameter at breast height, and wood volume for both plantation frames.

To perform a goodness-of-fit test, several regression-related parameters were calculated using the R statistical software (R Core Team 2021), including:

The Mean Squared Error (MSE) is a common estimate for measuring the predictive accuracy of a model (Tedeschi 2006), being the sum of the square of the differences between the ABM monthly results (Y_i) and reference model output (X_i), divided by the number n of data (Eq. 1).

\(MSE=\frac{{\sum }_{i=1}^{n}\left({Y}_{\text{i}}-{X}_{\text{i}}\right){ }^{2}}{n}\) (Eq. 1)

From this value, which is an absolute indicator, the result is relativized by the average or deviation of the observed values, leaving an indicator without magnitude. The Relative Prediction Error (RPE) is then calculated, which is expressed as a percentage from the square root of the MSE and the mean of the observed values (Eq. 2). This is an indicator of goodness-of-fit between actual and predicted values, where RPE values < 10% indicate that the model's predictions are good, values between 10–20% suggest reasonable predictions and values > 20% indicate mediocre predictions (Fuentes-Pila et al. 1996).

\(RPE=\frac{\sqrt{MS}E}{\widehat{Y}}\) (Eq. 2)

The square root of the MSE (RSE), but relative to the standard deviation of the observed values, (σ) is calculated to evaluate the error associated with the predictions of the model regarding the inherent variation in the observed values (Eq. 3). Values close to zero are considered better than positive values (< 0.5, very good prediction; 0.5–0.75, good prediction; 0.75–1, moderate prediction; > 1.0, the model needs improvements (Moriasi et al. 2007).

\(RSE=\frac{\sqrt{MS}E}{\sigma }\) (Eq. 3)

The Concordance Correlation Coefficient (CCC) was calculated as presented in (Eq. 4) (Lin 1989, 2000) using the R DescTools package (R Core Team, 2022). The CCC combines measures of both precision and accuracy to determine how far the observed data deviate from the line of perfect diagonal of concordance. This coefficient increases in value as a function of the nearness of the data's reduced major axis to the line of perfect concordance (the accuracy of the data) and of the tightness of the data about its reduced major axis (the precision of the data). Values of CCC near + 1 indicate strong concordance between x and y, values near − 1 indicate strong discordance, and values near zero indicate no concordance; there is no agreement on how to interpret other values.

\(CCC=\frac{2{\sigma }_{XY}}{{{\left(\widehat{X}- \widehat{Y}\right)}^{2} + \sigma }_{X}^{2} + {\sigma }_{Y}^{2}}\) (Eq. 4)

The Model efficiency (MEF) was calculated according to Tedeschi (2006) (Eq. 5) as an indicator of goodness-of-fit where MEF = 1 indicates perfect goodness-of-fit; MEF < 0 indicates that the values predicted by the model are worse than the observed mean.

\(MEF=\frac{{\sum }_{i=1}^{n}\left({Y}_{\text{i}}-\widehat{Y}\right){ }^{2}- {\sum }_{i=1}^{n}\left({Y}_{\text{i}}-{X}_{i}\right){ }^{2}}{{\sum }_{i=1}^{n}\left({Y}_{\text{i}}-\widehat{Y}\right){ }^{2}}\) (Eq. 5)

2.12 Scenarios

With the implemented ABM model, simulations were conducted over a 10-year period varying the total stocking rate from 300 to 700 (each 50) steers to study its impact on system carbon assessment.

The results of the ABM model implementation for the case study, incorporating the system information provided by the producer, are presented below. Subsequently, the results of simulations conducted with different carbon balance scenarios while varying the stocking rate are presented and discussed.

3.1 Model goodness-of-fit

Prior to show long term simulations results, to check coherence of the ABM model, its results were compared with original models by separated ways, as benchmarking reference. Normalized monthly LGW (kg/animal/day) and LW (kg/animal) evolution are presented in Fig. 2.

The regression between ABM and PPGL differs considering LWG or LW of steers. The LWG adjust poorly compared to LW, with more disperse result. Nevertheless, regression plot presents a reasonably good arrangement of paired plots on diagonal ABM = PPGL. Table 1 presents the goodness-of-fit between ABM and PPGL models, considering the defined statistical analysis parameters.

Table 1

Results of goodness-of-fit parameters between ABM and PPGL model results of monthly average weight gain (LWG) and weight (LW) of animals on 10-years simulations.
	LWG (kg/animal/day)	LW (kg/animal)
mean PPGL	0.378	390.196
mean ABM	0.388	396.512
standard deviation PPGL	0.107	25.032
standard deviation ABM	0.103	25.970
Fitness test parameters
R²	0.742	0.981
MSE	0.003	52.915
MSE bias (%)	3.10	75.39
MSE slope (%)	3.44	2.58
MSE random (%)	93.45	22.02
RPE (%)	14.87	1.86
RSE	0.523	0.291
CCC	0.852	0.931
CCC regression	0.857	0.960
CCC bias	0.995	0.969
MEF	0.724	0.915

Although MSE is an absolute indicator (its magnitude corresponds to the same one of variables), it often provides limited insight. However, the proportion of MSE behaves differently between the response variables of LW and LWG (Table 1). For the former, random effects primarily account for the differences between ABM and PPGL models. This phenomenon can be attributed to the heterogeneity of agents (working with a population) instead of using the mean behavior of individual animals multiplied by the total number of animals. Additionally, management practices can introduce the random noise observed between the models for LWG, as the PPGL model represents a continuously grazing system, while the ABM considers a rotational grazing system.

It is interesting to see the LW response variable behavior, where the primary differentiating factor is a bias where the ABM underestimates this variable by an average of 6.3 kg per animal.

For both response variables, the slope of the regression does not account for the differences between the models (only about a 3% impact).

The first relevant parameter is the RPE, which yields reasonable predictions for LWG and good predictions for LW. Similarly, for the RSE parameter, the predictions are good and very good for LWG and LW, respectively.

Concerning the CCC parameter, both response variables exhibit strong concordance between the models. For both variables, both components (regression and bias coefficients) result in good regressions with minimal deviation from the diagonal. Lastly, the Mean Efficiency (MEF) indicates good and excellent fitness for LWG and LW, respectively.

Considering these results, the model fits very well for a crucial variable in system decision-making, which is the weight of animals. It is important to note that the LW of animals determines their exit from the system, leading to the inclusion of a new batch of animals for fattening. In long-term simulations, this effect becomes more relevant than the average daily LWG of animals, which exhibits fair fitness between the models. Both the ABM and PPGL models show consistent dynamics in 10-year simulations, as discussed in the following section.

For main dasometric annual variables ABM was compared with SAG model results over 10-years simulations (see Table 2).

Table 2

Results of goodness-of-fit parameters between ABM and PPGL SAG model results of annual evolution of number of trees (NT), basal Area (BA), diameter at breast hight (DBH) and wood volume (WV) on 10-years simulations.
	NT (trees/ha)	BA (m²/ha)	DBH (cm)	WV (m³/ha)
mean SAG	788.85	17.52	16.98	108.88
mean ABM	788.95	19.25	5.29	120.90
standard deviation SAG	286.12	8.13	18.05	73.85
standard deviation ABM	286.22	8.26	5.786	79.12
Fitness test parameters
R²	0.999	0.952	0.969	0.991
MSE	0.10	6.09	2.29	216.57
MSE bias (%)	10.00	48.87	49.61	66.71
MSE slope (%)	10.22	1.68	13.88	13.645
MSE random (%)	79.73	49.44	36.50	19.63
RPE (%)	0.040	14.08	8.918	13.515
RSE	0.001	0.303	0.286	0.199
CCC	0.999	0.931	0.939	0.966
CCC regression	0.999	0.953	0.961	0.981
CCC bias	0.999	0.977	0.976	0.985
MEF	0.999	0.903	0.914	0.958

As depicted in Table 2, all goodness-of-fit parameters indicate a very good fit between ABM and SAG dasometric variables. Even though these model comparisons may appear self-evident, the tests provide confidence in the successful implementation of trees evolution over time in the ABM model. Regarding specific behaviors, it is notable that WV exhibits a bias in MSE, suggesting that the ABM model overestimates forestry yield by 11%. For the other parameters, random noise is the primary source of MSE differences. However, in all cases, the ABM model exhibits an overall strong performance.

3.2 Case Study Simulation

Among the key response variables, the LW of the animals exhibited 14 fattening cycles in the 10 simulated years (Fig. 3). It is worth to mention that PPGL model comparations also presents 14 fattening cycles in the same simulation period.

The total volume of wood (stem and branches) showed a logistic growth pattern, in accordance with the equations of the SAG model (Fig. 3).

Analyzing the results of the ABM model from the point of view of animal production, the annual yield expressed in kg of LW exported averaged 98.7 ± 2.6 kg/ha with an average stocking rate of 0.72 Livestock Units per hectare (LU/ha). These values fall within the range reported by the Plan Agropecuario Institute for the latest agricultural exercises for stocking rates of 0.79 LU/ha (Ghelfi & Molina 2022) and are expected for livestock production on native pasture. According to a report by UPM (2010), analyzing various studies on the incorporation of forestry into local livestock systems, LWG of 97 kg/ha are reported with stocking rates also of 0.79 LU/ha.

Considering the forestry point of view, simulations over a 10-year period revealed promising results in terms of the forested area in the SPS, which accounted for 13% of the total area (77 hectares of total forested land, with approximately two-thirds of the area planted densely for pulpwood and one-third with alleys for silvopasture). According to Cabrera et al. (2016), the results suggest that the inclusion of forestry in livestock systems does not necessarily have to be on a large scale or occupy a significant portion of the grazing land, with economic, and furthermore, environmental benefits, which will be discussed later in this work.

Even if Eucalyptus dunnii plantations in dense frames for cellulose pulp can affect negatively grass growth and composition, frames with alleys and in low proportion of livestock farm do not impact on forage yield (Boscana et al. 2019). Since the present study considered a modest inclusion of afforested areas (13% mentioned above), the performance of animals resulting from their interaction with the grass did not differ significantly from extensive livestock farming without forestry (see section 3.1 Model goodness-of-Fit). Moreover, as animals do not spend all their grazing time in silvopastoral paddocks (see Fig. 1; three of thirty-one total paddocks), the 10% improvement in LWG due to shelter effects, as indicated by Varela (2019), did not significantly affect the overall animal performance of the system.

From an environmental point of view, and regarding the GHG emissions, the average methane emissions per individual -considered the primary GHG from bovine cattle and identified as the major air impact of livestock farming, as reported by MA (2021)- was 0.279 ± 0.004 kg CH₄/animal/day, assuming a relatively low grass digestibility (50%). This value aligns with the range observed by Montagnini et al. (2013) for SPS in Colombia, which reported methane emissions between 0.196 and 0.38 kg CH₄/day for pastures with digestibility ranging from 55–75% to 45–55%, respectively. In a recent study Congio et al. (2023) in a large literature review to adjust CH₄ emissions for Latin America livestock farming systems ranged from 0.0382 from 0.364 kg/day.

In current simulations, the mean methane emission factor per steer fattening cycle was 72.7 kg CH₄/animal. Annualizing this emission factor results in 101.8 kg/animal/year. Since the emission factor is linked to LW and LWG of animals, steers in the modeled native pasture-based system exhibit relatively high emissions for finishing systems with relatively low pasture quality in terms of protein and digestibility (see 2.9 Parameters section). However, the modeled LW for these systems is 397 kg, with a LWG of 0.388 kg/steer/day (see Table 1). This aligns the emission with mature LW animals exhibiting low LWG, as reported by the United States Environmental Protection Agency (US-EPA 2023) and Ominski et al (2007).

Analyzing the relationship between greenhouse gas emission intensity and animal LWG, an inverse exponential relationship was observed, as depicted in Fig. 4

As animals gain more weight, emissions per animal decrease. This is attributed to the LGW being associated with improved feed conversion efficiency and reduced fattening time, leading to lower methane production per unit of LGW by the animals (Bilotto et al. 2019). The results obtained are comparable to those reported by Picasso et al. (2014) and Becoña et al. (2020) for local pasture-based fattening systems, where increased LWG directly reduces greenhouse gas emission intensity. Bilotto et al. (2019) modeled GHG emission intensity for different grassland-based systems in Argentina, founding the same trend and results: factors that determines lower LGW to (e.g. animals nutrition quality and it category) leads to increase GHG emission intensity. Those authors found ranges from 22 to 18 kg CO₂e/kg of LGW for those including backgrounding/finishing phases or pure cow-calf systems, respectively.

The average emission intensity values resulting from the ABM model were 24.9 kg of CO₂e/kg LWG, for an average gain of 0.388 kg/animal/day. These values are slightly higher than those reported locally by the Ministry of Environment (MA 2021) with ranges close to 20 CO₂e/kg of LWG. It is worth noting that the modeled case study involves a fattening phase on native pasture, leading the animals to relatively high weights (450 kg LW steers) which are finished in pens in another part of the farm system.

The type of animal, as well as the potential weight gain on native pasture and animal genetics, are factors influencing the calculation of GHG emission intensity, which result in a relatively low efficiency from this perspective. Congio et al. (2023) in a metanalysis work found that feed intake variables were the main predictors of CH₄ emission, primarily DM intake. Additionally, these authors found that LWG is a good predictor of emissions, with goodness-of-fit parameters RSE = 0.99, RPE = 32.3, MSE bias = 1.57, MSE slope = 1.74 and CCC = 0.11. In ABM model, CH₄ emissions and LGW where positive and linearly related (R² = 0.83) with RSE = 0.52, RPE = 14.87, MSE bias = 3.1 MSE slope = 3.45 and CCC = 0.85. Even if results are logically explained considering that CH4 calculations implies LGW (among other energy requirements of animals), it is important to note the relative high value of the coefficient of determination, explaining a high proportion of the relationship between variables.

Finally, considering the carbon balance of the modeled system, and as mentioned earlier, a positive carbon balance is achieved in the system after 10 years with only a total of 77 hectares of forested area (13% of the total area), as shown in Fig. 5.

Figure 5 illustrates that the first three years exhibit a negative carbon balance. This is since the forest biomass production in years 2 and 3 is insufficient to offset the emissions from livestock. The total emissions of GHG per hectare obtained with the ABM model averaged 2,457.7 kg CO₂e/ha/year, a value comparable to the 2,400 kg CO₂e/ha/year reported by Picasso et al. (2014) for production systems with a yield of 100 kg LW/ha. In a similar context, Silva et al. (2023) reported 2,709 kg CO₂e/ha/year for silvopastoral systems in Colombia with a stocking rate of 0.8 LU/ha.

The incorporation of forested areas into the productive system is a key factor in greenhouse gas emissions mitigation. Trees, through photosynthesis, can sequester atmospheric carbon dioxide, reducing its concentration and contributing to carbon capture (Montagnini & Nair, 2004). This carbon capture can partially offset the emissions generated by livestock, helping to balance the carbon budget of the system. While the inclusion of a forested area may be questioned by producers at both the local (Cabrera et al. 2016; Boscana et al. 2019) and international levels (Fischer & Vasseur 2002; Iñamagua-Uyaguari et al. 2023) when considering its incorporation into their production systems, it is evident that from a carbon balance perspective, the inclusion of trees offers significant benefits.

At the local level, through interviews with livestock farmers who introduced afforestation as a complementary component, Cabrera et al. (2016) observed that the primary benefit perceived by producers is the shelter provided during winter and shade in summer. This improves the well-being of animals in the face of extreme temperatures typical of southern temperate regions, and in some cases, it translates into increased LWG and higher calving rates at raring systems. Additionally, the authors found that producers reported that despite the need to reduce grazing area (primarily during the tree establishment period), this reduction did not result in significant yield losses to their production systems. On the contrary, the improvements obtained more than compensated for the temporary reductions in productive area. Other benefits, such as a lack of increased workload on the farming task, the possibility of saving for the future, and no negative impact on livestock health due to afforestation inclusion, are relevant factors for adoption. Considering economic barriers to adoption of SPS for familiar farming systems, according to Montagnini et al. (2013) a way to promoting adoption are financial support provided by specific research projects, or government programs of payments for environmental services.

While the benefits of incorporating afforestation into livestock systems seem evident, the inclusion of SPS (with wide alleys) or pastures with dense spacings for pulpwood production appears to be a matter of controversy. A recent local study (Conde et al. 2023) analyzes the integration of afforestation into extensive grazing livestock farms using linear programming solutions, primarily focusing on economic aspects in a case study. The authors do not recommend the inclusion of integrated silvopastoral grazing areas, but rather suggest that separate afforestation from livestock farming is the better choice. This type of study underscores the complexity of decision-making processes when considering only one (economical) perspective within dynamic and interactive components developing over time. In this context, for small-scale livestock farming systems (typical of most cattle owners in the country), integrated SPS with a low proportion of afforested areas may offer solutions for both economic and animal welfare concerns.

3.3 Sensitivity to stocking rate variations

The sensitivity analysis to stocking rate (ranging from 300 to 700 steers) reveals an inverse linear behavior between the increase in the total number of steers and the carbon balance (Fig. 6).

According to the obtained results, the carbon equilibrium point for the simulated system is 380 animals over a simulation period of 10 years. This value is lower than that obtained in the simulation of the current system, which is 420 animals. It is important to note that, during the first two years of the simulation, grazing in the forested paddocks is excluded, resulting in a reduction in forest biomass production, as mentioned earlier (see Fig. 5).

The results reveal a conflict between economic-productive and the environmental dimension. To achieve a neutral carbon system, it is necessary to maintain a lower number of livestock heads than the property's potential. In fact, considering the average weight of steers in the ABM model as 396.5 kg/animal (see Table 1), and taking into account the livestock equivalents established for Uruguay at 380 kg per animal, and considering the total system's 606 hectares, it results in a stocking rate of 0.72 UG/ha. This value is even slightly lower than the average carrying capacity of native fields, although it can be partially adjusted for non-grazeable areas in paddocks with forests, representing a low proportion of the property. As discussed earlier, the carrying capacity of native livestock grassland systems in Uruguay is 0.79 UG/ha (UPM 2010; Ghelfi & Molina 2022), for an average productivity of approximately 100 kg of LW/ha. A reduction in the number of animals and, consequently, a decrease in livestock stocking rate would lead to situations of undergrazing or, at least, working with a stocking rate lower than its potential.

According to Bussoni et al (2021) and Montagnini et al. (2013), this conflict could be resolved through the market if this service is recognized at a higher value. Additionally, an increase in livestock conversion efficiency is a key aspect to improve environmental parameters (Becoña et al. 2020).

The results obtained demonstrate the potential of silvopastoral systems to achieve greater carbon capture and reduce GHG emissions. By combining livestock with the presence of trees, a synergy between animal production and climate change mitigation can be achieved. Other emergent benefit of SPS is the improvement of natural resource management, enhancing the provision of environmental services like biodiversity, carbon, water, land degradation control, and the improvement of whole agroecosystem productivity (Montagnini et al. 2013).

This study, using multi-agent modeling, has demonstrated that SSPs with proper management practices, such as rotational grazing and the use of afforested paddocks, can have a significant impact on productivity and the reduction of greenhouse gas emissions. The results emphasize the importance of promoting and adopting sustainable approaches in animal and plant production, as SSP offer an integrated solution that combines economic and environmental benefits. This work models a real production system based on data provided by the producer. A more in-depth exploration of productive, economic, and environmental variables will allow for models that are closer to represent reality and feasible for implementation by livestock producers to promote long term impact decision-making tools.

Acknowledgments

This work had the financial support of the Comisión Sectorial de Investicación Científica of the Universidad de la República, by the project FAgro-CSIC 02100-500193-21

Statements and Declarations

This article is based on a previous work presented at the XII International Congress of Silvopastoral Systems held in Montevideo City, October 2023. This work had the financial support of the Comisión Sectorial de Investigación Científica of the Universidad de la República, by the project FAgro-CSIC 02100-500193-21.

Baeza S, Lezama F, Piñeiro G, Altesor A, Paruelo J (2010) Spatial variability of above-ground net primary production in Uruguayan grasslands: a remote sensing approach. Applied Veg Sci 13:72–85. http://10.1111/j.1654-109X.2009.01051.x
Becoña G, Ledgard S, Astigarraga L, Lizarralde C, Dieguez F, Morales H (2020) EMAG- National model to evaluate environmental impacts of cattle production systems in Uruguay. Agrociencia Uruguay 24:48. https://doi.org/10.31285/AGRO.24.48
Bilotto F, Recavarren P, Vibart R, Machado C (2019) Backgrounding strategy effects on farm productivity, profitability and greenhouse gas emissions of cow-calf systems in the Flooding Pampas of Argentina. Agric Sys 179:102688. https://doi.org/10.1016/j.agsy.2019.102688
Bussoni A, Cubbage F, Alvarez Giambruno J (2021) Silvopastoral systems and multi-criteria optimization for compatible economic and environmental outcomes. Agric Syst 190:103118. https://doi.org/10.1016/j.agsy.2021.103118
Cabrera G, Rachetti M, Condon R, Olivera B, Artigalás A, Martínez J, Inzaurralde S, Garrone I, Sánchez I, Bentancor L, Bussoni A, Munka C, Boscana M, Alvarez J, Fernandez E (2016) La integración de forestaciones en predios familiares: resultado de un estudio participativo. http://www.ainfo.inia.uy/digital/bitstream/item/15316/1/Integracion-de-forestaciones-en-predios-familiares-CNFR-2016.pdf. Accessed 6 November 2023
Conde R, Robledo F, López de Lacalle A (2023) Silvopastoral and agrorgoforestry systems: An integer linear programming model for investment decisions. J Dyn Games 10:304–329. https://doi.org/10.3934/jdg.2023013
Congio G, Bannink A, Mayorga O, Rodrigues J, Bougouin A, Kebreab E, Carvalho P, Berchielli T, Mercadante M, Valadares-Filho S, Borges A, Berndt A, Rodrigues P, Ku-Vera J, Molina-Botero I, Arango J, Reis R, Posada-Ochoa S, Tomich T, Castelán-Ortega O, Marcondes M, Gómez C, Ribeiro-Filho H, Gere J, Ariza-Nieto C, Giraldo L, Gonda H, Cerón-Cucchi M, Hernández O, Ricci P, Hristov A (2023) Improving the accuracy of beef cattle methane inventories in Latin America and Caribbean countries. Sci Total Env 856:159128. https://doi.org/10.1016/j.scitotenv.2022.159128
Dieguez F, Fort H (2017) Towards scientifically based management of extensive livestock farming in terms of ecological predator-prey modeling. Agric Sys 153:127–137. https://10.1016/j.agsy.2017.01.021
Do Carmo M, Soca P, Cardozo G, Balzarini M (2015). Ajuste de carga animal en base a la oferta de forraje. Certilla técnica. Proyecto Ganadería Familiar y Cambio Climático, DGDR, Ministerio de Ganadería Agricultura y Pesca, Montevideo, Uruguay, http://www.inia.uy/Documentos/Privados/INIA%20Tacuaremb%C3%B3/Dia%20de%20campo%20manejo%20de%20carga%20en%20Basalto%20oct%202015/protocolo%20estimacion%20forraje%20y%20ajuste%20de%20carga%20Laura_Maxi.pdf. Accessed 6 November 2023
Doldán J, Fariña I, Tarigo F (2008) Utilización de Eucalyptus spp, alternativas de plantaciones uruguayas para pulpa Kraft. INNOTEC 3:30–34. https://doi.org/10.26461/03.04
FAO (2021) Food and Agriculture Organization. Emissions due to agriculture. Global, regional and country trends 2000–2018.https://www.fao.org/3/cb3808en/cb3808en.pdf. Accessed 6 November 2023
Fischer A, Vasseur L (2002) Smallholder perceptions of agroforestry projects in Panama. Agroforestry Sys 54:103–113. https://doi-org.proxy.timbo.org.uy/10.1023/A:1015047404867
Fuentes-Pila J, DeLorenzo M, Beede D, Staples C, Holter J (1996) Evaluation of Equations Based on Animal Factors to Predict Intake of Lactating Holstein Cows. J Dairy Sci 79:1562–71
Ghelfi M, Molina C (2022) Síntesis de resultados del Programa de Monitoreo de Empresas Ganaderas. Instituto Plan Agropecuario. https://www.planagropecuario.org.uy/uploads/monitoreos/37_Monitoreo_Empresas_Ganaderas_21_22.pdf. Accessed 6 November 2023
Grimm V, Polhill G, Touza J (2013) Documenting Social Simulation Models: The ODD Protocol as a Standard. In: Edmonds B & Meyer R (eds) Simulating Social Complexity. Understanding Complex Systems. Springer, Berlin, Heidelberg, pp 117–133
Grimm V, Revilla E, Berger U, Jeltsch F, Mooij W, Raislback S, Thulke H, Weiner J, Wiegand T, DeAngelis D (2005) Pattern-Oriented Modeling of Agent-Based Complex Systems: Lessons from Ecology. Science 310, 987–991. https://10.1126/science.1116681
Hanisch A, Negrelle R, Araújo Bonatto R, Nimmo E, Biscaia Lacerda A (2019) Evaluating Sustainability in Traditional Silvopastoral Systems (caívas): Looking Beyond the Impact of Animals on Biodiversity. Sustainability 11, 3098. https://doi.org/10.3390/su11113098
Hirigoyen A, Resquin F, Cerrillo R, Franco J (2021) Stand biomass estimation methods for Eucalyptus grandis and Eucalyptus dunnii in Uruguay. Bosque 42. https://10.4067/S0717-92002021000100053
Hirigoyen, A., Rachid, C., 2014. Sistema de Apoyo a la Gestión Forestal. Modelos de Simulación de Crecimiento y Análisis Económico. Revista INIA 68:69–74.
Iñamagua-Uyaguari JP, Fitton N, Smith P (2023) Planting trees in livestock landscapes to protect soil and water also delivers carbon sequestration. Agroforest Syst 97:1199–1214. https://doi-org.proxy.timbo.org.uy/10.1007/s10457-023-00857-9
IPCC (2006) Guidelines for National Greenhouse Gas Inventories, Prepared by the National Greenhouse Gas Inventories Programme, Eggleston H, Buendia L, Miwa K, Ngara T and Tanabe K (eds). Vol 4. IGES, Japan. https://www.ipcc-nggip.iges.or.jp/public/2006gl/vol4.html. Accessed 6 November 2023
IPCC (2019) Refinement to the 2006 IPCC Guidelines for National Greenhouse Gas Inventories. https://www.ipcc-nggip.iges.or.jp/public/2019rf/. Accessed 6 November 2023
Jackson J, Ash A. (1998) Tree-grass relationships in open eucalypt woodlands of northern Australian: influence of trees on pasture productivity, forage quality and species distribution. Agroforestry Syst 40:159–176.
Lin L (1989) A Concordance Correlation Coeffi-cient to Evaluate Reproducibility. Biometrics 45: 255–268. https://doi.org/10.2307/2532051
Lin L (2000) A note on the concordance correlation coefficient. Biometrics 56:324–5. https://doi.org/10.2307/2533516
MA (2021) Ministerio de ambiente, Presidencia de la República Oriental del Uruguay. https://www.gub.uy/ministerio-ambiente/sites/ministerio-ambiente/files/documentos/publicaciones/Huella_Ambiental_Ganadera_Informe%2010-2022.pdf. Accessed 6 November 2023-
Methol R (2008) SAG Eucalyptus: Sistema de Apoyo a la Gestión de Plantaciones de Eucalyptus orientadas a la Producción de Celulosa en Uruguay, Instituto Nacional de Investigación Agropecuaria. Serie Técnica INIA 173, Montevideo, Uruguay.
Mieres J, Assandri L, Cúneo M (2004). Tablas de valor nutritivo de alimentos. Guía parala alimentación de rumiantes. Serie Técnica INA 142, Instituto Nacional de Investigación Agropecuaria, Montevideo, Uruguay.
Miller R, Albicette M (2005) Seminario de actualización técnica en manejo de campo natural. Serie Técnica INIA 15, Instituto Nacional de Investigación Agropecuaria, Montevideo, Uruguay.
Montagnini F, Nair P (2004) Carbon sequestration: an underexploited environmental benefit of agroforestry systems. Agroforestry Sys 61:281–295.
Montagnini F, Ibrahim M, Murgueitio Restrepo E (2013) Silvopastoral systems and climate change mitigation in Latin America. Bois et Forêts des Tropiques 316:3–11
Moriasi D, Arnold J, Van Liew M, Bingner R, Harmel R, Veith T (2007) Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Transactions of the ASABE 50:885–900. https://doi.org/10.13031/2013.23153
Ominski K, Boadi D, Wittenberg K, Fulawka D, Bassarb 1 (2007) Estimates of enteric methane emissions from cattle in Canada using the IPCC Tier-2 methodology. Can J of Anim Sci 87:459–467. https://doi.org/10.4141/CJAS06034
Panario D (1988). Geomorfología del Uruguay. Facultad de Humanidades y Ciencias, Universidad de la República. Montevideo, Uruguay
Pastor J (2008) Mathematical Ecology of Populations and Ecosystems. A John Wiley & Sons, Ltd, Sussex, United Kingdom
Pastorini V (2017) Enfoques y avances en diversificación de especies forestales y forestación a pequeña escala: acciones de la Dirección general de desarrollo rural – Ministerio de ganadería, agricultura y pesca. http://www.ainfo.inia.uy/digital/bitstream/item/9158/1/SAD-774p11-14.pdf. Accessed 6 November 2023
Pérez-Rocha J (2020). El estado del campo natural en el Uruguay. Montevideo. FAO, MVOTMA y MGAP. https://doi.org/10.4060/cb0989es
Picasso V, Modernel P, Becoña G, Salvo L, Gutiérrez L, Astigarraga, L (2014) Sustainability of meat production beyond carbon footprint: a synthesis of case studies from grazing systems in Uruguay. Meat Sci 98:346–354. https://doi.org/10.1016/j.meatsci.2014.07.005
Priesack E, Gayler S, Hartmann HP. The impact of crop growth sub-model choice on simulated water and nitrogen balances. Nutr Cycl Agroecosyst. 2006;75(1–3):1–13.
R Core Team (2021) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/. Accessed 6 November 2023
R Core Team (2022) DescTools package. Concordance Correlation Coefficient. https://search.r-project.org/CRAN/refmans/DescTools/html/CCC.html. Accessed 6 November 2023
Rachid-Casnati C, Mason E, Woollons R (2019). Using soil-based and physiographic variables to improve stand growth equations in Uruguayan forest plantations. iForest, 12:237–245. http://10.3832/ifor2926-012
Railsback F, Grimm V (2019) Agent-Based and Individual-Based Modeling. Princeton University Press
Silva A, García, D, Martínez E (2023) Silvopastoral Systems Ecological Strategy for Decreases C Footprint in Livestock Systems of Piedmont (Meta), Colombia. Braz. arch. biol. technol. 66;e23220340. https://doi.org/10.1590/1678-4324-2023220340
Sarandón S, Flores C (2014) Agroecología: bases teóricas para el diseño y manejo de Agroecosistemas sustentables. Facultad de Ciencias Agrarias y Forestales. Universidad de la Plata, Argentina
Schinato F (2022) Evaluación de Servicios Ambientales en un Sistema Silvopastoril. Tesis Magister en Ciencias Agracias, Facultad de Agronomía, UdelaR. https://www.colibri.udelar.edu.uy/jspui/bitstream/20.500.12008/35851/1/SchinatoFranco.pdf. Accessed 6 November 2023
Scoz R (2015) Sistemas silvopastoriles en Uruguay; un sistema productivo que no se adopta con solo promocionarlo. 3° Congreso Nacional de Sistemas Silvopastoriles - VIII Congreso Internacional de Sistemas Agroforestales. http://www.ainfo.inia.uy/digital/bitstream/item/4511/1/Scoz-2015.pdf.. Accessed 6 November 2023
Tedeschi L (2006) Assessment of the Adequacy of Mathematical Models. Workshop on Mathematical Model Analysis and Evaluation Sassari. http://www.cncps.cornell.edu/modeval. Accessed 6 November 2023
USEPA (2023) United States Environmental Protection Agency. Enteric Fermentation-Greenhouse Gases. https://www3.epa.gov/ttnchie1/ap42/ch14/final/c14s04.pdf. Accessed 6 November 2023
UPM (2010) La forestación y la ganadería en Uruguay. https://www.upm.uy/siteassets/documents/forestacion-ganaderia-uruguay.pdf Accessed 6 November 2023
Uruguay XXI (2023) Sector ganadero en Uruguay. https://www.uruguayxxi.gub.uy/uploads/informacion/b28305203e10f6d7bb3689eb619160dd922b2036.pdf. Accessed 6 November 2023
Varela F (2019) Modelo de simulación para sistemas de silvopastoreo en regiones centro sur y sureste de Uruguay. Tesis Magister en Ciencias Agrarias. Facultad de Agronomía, UdelaR. https://www.colibri.udelar.edu.uy/jspui/bitstream/20.500.12008/29153/1/VarelaCasadeyEduardoFabi%C3%A1n.pdf. Accessed 6 November 2023
Wilensky U, Stroup W (1999) HubNet.. Center for Connected Learning and Computer-Based Modeling, Northwestern University. Evanston, IL. http://ccl.northwestern.edu/netlogo/hubnet.html. Accessed 6 November 2023

No competing interests reported.

Download PDF

Editorial decision: Revision requested
24 Jan, 2024
Reviews received at journal
09 Jan, 2024
Reviewers agreed at journal
15 Dec, 2023
Reviewers invited by journal
15 Dec, 2023
Editor assigned by journal
21 Nov, 2023
Submission checks completed at journal
21 Nov, 2023
First submitted to journal
20 Nov, 2023

You are reading this latest preprint version

Toward neutral carbon integrated production y silvopastoral systems: an Uruguayan study case modeled using agent-based technique

Status:

Version 1

Abstract

Figures

1. Introduction

2. Materials and methods

2.1 General Description

2.2 Design Concepts

2.3 Agent Behavior

2.4 Interaction Between Agents

2.5 State Variables

2.6 Events

2.7 Details

2.8 Initialization

2.9 Parameters

2.10 Model Rules

2.11 Statistical analysis

2.12 Scenarios

3. Results and Discussion

3.1 Model goodness-of-fit

3.2 Case Study Simulation

3.3 Sensitivity to stocking rate variations

4. Conclusions

Declarations

References

Additional Declarations

Status:

Version 1