Globally, agricultural-use land occupies about 40-50% of the land surface and generates about 10-12% of the total global anthropogenic emissions, i.e., 5,1-6,1G t CO₂-eq per year (IPCC, 2006). Land uses in the study area (Villavicencio) are characterized mainly by conversion from grassland to continuous croplands, which causes a large degradation of soil organic matter (SOM) (Silva, 2018; Silva and Orozco, 2018), mainly due to conventional tillage (García et al., 2018). Greenhouse gas emissions are influenced by the type of land used, especially by the types of crops and/ or pastures in diverse environments (Mangalassery et al., 2014; Chambers et al., 2016; Lal, 2018). The effects of land use on the emission of CO₂ are dominated by the type of tillage and/or pasture management (IPCC, 2006; Haddaway et al., 2017; Behnke et al., 2018; FAO, 2018). Land use may have direct and indirect effects on soil carbon stocks, and these changes may be conditioned to meet the social needs of farmers, such as the production of food, energy and fossil fuel, water supply, and crop residues management, in order to achieve higher productions in the short term (Popp et al., 2017; Nyambo et al., 2020). Several studies applying IPCC models to different extents corroborate that different soil management practices and types of tillage increase soil CO₂ emissions (IPCC, 2006; Cardinael et al., 2018; Lal, 2018; Parra et al., 2019). As an option for reducing soil CO₂ emissions, conservation practices have increased in many parts of the world, aiming to also increase production and sustainable development (Ogle et al., 2019). Agroforestry systems with trees and crops in interactive and simultaneous cultivation have been regarded as a key cropping practice for improving the productivity of agroecosystems and reducing soil CO₂ emissions (Nair, 2012; Feliciano et al., 2018). This can be seen in Figure 1.

Figure 1 Source: Authors

In general, the transition from cropland to an agroforestry system is beneficial to soil organic carbon (SOC) (Cardinael et al., 2018). The accumulation of SOC due to the sequestration of carbon in the soil is certainly one of the major benefits of agricultural systems, as it is effective in helping to mitigate the increase in atmospheric CO₂ concentrations (Lal, 2018). For example, in the southeastern USA, conservation tillage, combined with intensive crop rotations that include cover crops, can sequester an average of 1 700 lb of CO₂ each year, i.e., 464 lb of C per acre (Franzluebbers, 2015). There are various methods to estimate soil CO₂ emissions from agriculture, ranging from simple Tier 1 methods (IPCC, 2006; Ogle, 2004; Parra et al., 2019) to complex process-based models that simulate the changes in soil carbon with some detail (Tiers 2 and 3) (FAO, 2018), although this relation is even more complex under the influence of climate change in tropical zones. Cardinael et al., (2018), applying a Tier 1 IPCC methodology, validated that the mean SOC storage rate (± confidence intervals) for croplands converted to agroforestry systems was 0,75±0,19 t C ha^-1 yr^-1, while the mean SOC loss rate for forests converted to agroforestry systems was -1,15±1,02 t C ha^-1 yr^-1, in all regions, climates, and agroforestry systems taken together. The mean SOC change rates for the conversion from grasslands to agroforestry systems were not significantly different from zero (0,23±0,25 t C ha^-1 yr^-1). In this sense, by employing soil management practices, agricultural lands can both sequester soil carbon and reduce GHG emissions (IPCC, 2006). Nemo et al., (2017) showed a variant of Tier 1 testing with the 2 RothC model (IPCC, 2006), which is used to simulate the interaction between GHG emissions, growth, and grazing in managed grasslands, where the C-input was adjusted so the equilibrium C matched the measured total SOC at the end of the spin-up period. This variant is useful for grassland systems in which the plant-derived carbon input is the most uncertain parameter, as well as the one to which the model is most sensitive (Poeplau, 2011). These types of approaches have been used to estimate C sequestration potentials in grasslands, as well as the potential effects of pasture management on SOC and stock changes on global, national, and regional scales (Henderson et al., 2015; Chambers et al., 2016; Novaes et al., 2017; FAO, 2018; Parra et al., 2019). Our specific objectives were the following: (1) to test the performance of the IPCC Tier 1 ensemble approach to simulate soil CO₂ emissions; (2) to quantify differences in soil CO₂ emissions between agricultural systems and zones across the Villavicencio area by modeling and measuring SOC according to the effect of soil management practices as a first approximation to national inventories; and (3) to assess the performance of the IPCC Tier 1 approach with parameters such as the RMSE, the Nash-Sutcliffe efficiency (NSE), and R².

To evaluate the accuracy of our model parameterization, the model performance was evaluated with four widely used quantitative methods, i.e., the R² (squared correlation coefficient), which is described in Equation (2); the RMSE (root mean squared error) (Equation (3)) and RMSE/n (Moriasi et al., 2007), a measurement of accuracy calculated as the differences between model-predicted and measured SOC values; and the (E) (model performance efficiency) (Nash-Sutcliffe, 1970), which evaluates the degree of closeness between modeled and observed data (Ludwig et al., 2011) (Equation (4)). These indicators were analyzed with P < 0,005. The statistical analysis was performed using Infostat v. 17.0 for Windows.

where: SSres= sum (Oi - Pi)² and SStot = sum (Oi - Ornean)²; Oi = observed values (known results); Pi = expected values or unknown results.

The RMSE ranges from 0 to 100. At an ideal fit, the RMSE is equal to zero. A lower RMSE is better.

The Nash-Sutcliffe model efficiency coefficient E was calculated as follows:

where Ō is the observation mean. A higher E is better, and it can be expressed as a percentage when multiplied by 100 (Smith and Smith, 2007). A linear regression of the simulated SOC (yPi) and observed SOC (xOi) time-series data was performed (Equation (5)):

which resulted in a slope sPi, an intercept IPi, and the coefficient of determination R². In this sense, only the agricultural systems with the best fit in the model were graphed.

After validation, the Tier 1 IPCC model was used to simulate the current SOC initial stock changes in order to assess the effect of soil management factors on SOC final stocks (Equation (6)).

where SOC₀₁ is the mean of the final soil C stocks (t ha-1) over the next 20 years, (SOC₀) = SOCinitial Stock (t ha^-1); and FLU , FMG, and FI are the default values for soil management (IPCC, 2006). SOC stock changes in the top soil (0-30 cm) over a period of 20 years were calculated as follows (Equation (7)):

where: ΔSOC = losses and/or gains in SOC rates, T = default time for transition between equilibrium SOC values (20 years). ΔSOC can be converted to atmospheric CO₂ stored in or emitted from the soil by multiplying the tons of C by 44/12 (the ratio of molecular weight for CO₂ and C) (IPCC, 2006).

Rates of gains/losses of soil C and soil CO₂ emis sions for each agricultural system

A significant limitation for model validation was that the change in SOC stock was equated with CO₂ emissions, as only input variables had an assigned probability distribution within each default value, depending on the soil management factor identified, without considering other input factors. The observed SOC storage rates regarding the conversion from pasture for coffee-based agroforestry (AFCf_7) in zone 1 were higher than those of the conversion of rice/corn crop rotation for improved banana fallow (IFB_2) of zone 3 (Table 4) -levels: ΔSOC = AFCf_7 = 2,96 t C ha^-1 year^-1 and IFB_2 = 0,08 t C ha^-1 yr^-1, accounting for a neutralization of -10,83t CO₂eq ha^-1 year^-1 in AFCf_7, which was due to default values FLU, FMG, and FI in the simulations that demonstrated potential mitigation.

Table 4

Legend: AS: Agrisilvicultural system; SS: Shade system; SP: Silvopastoral system; NDP: Non-degraded pasture; IF: Improved fallow; HG: Homegarden system; MDP: Moderately degraded system; CL: Cropland; CR: Crop rotation; Cf: Coffee; Cc: Cocoa; B: Banana; Ci: Citrus; P: Papaya; R: Rice; S: Soybean; C: Corn; F_LU: land use factor; F_MG: pasture management and/or tillage regime in cropland (full, reduced, no tillage); F_I:carbon input level (low, medium, high without manure, high with manure).

The mean stock change factor (default value, ± confidence intervals) was 1,19±0,10 for grassland converted to agroforestry (Cardinael et al., 2018). Poeplau et al., (2011) demonstrated that conversion from grasslands to agroforestry systems did not improve SOC stocks. However, this work agrees that the returns of organic material (leaf litter) in coffee-based agroforestry systems is higher than in the monoculture of coffee as, demonstrated by Zaro et al., (2020). In general, the transition from cropland to an agroforestry system is beneficial to SOC (Lu et al., 2015), as demonstrated in the simulated IFB_2. The results in Table 4 also show that the effect of pasture management on gains/losses of soil C rates varied substantially between NDP_15 compared to the MDP116 of zones 3 and 4 (Table 4) -levels: ΔSOC = NDP_15 = 0,12 t C ha^-1 year^-1 and MDP1_16 = -0,02 t C ha^-1 year^-1. These values are in line with those reported by Silva and Orozco (2018) in degraded pastures of Ariari, Meta, Colombia. Typically, in the model-integrated soil management factors, NDP demonstrated better pasture management than MDP. Several previous studies also showed that many pasture techniques have been applied to mitigate GHG emissions from agriculture (Jadan et al., 2015; Parra et al., 2019). A plausible adoption rate of 30% for improved deep-rooted legumes associated to Brachiaria pastures in Cerrado, Brazil, represented a mitigation potential of -29,8 t CO₂-eq yr^-1 to the atmosphere (Thornton and Herrero, 2010). However, belowground C-inputs from exudation and root sloughing from C4 grasses are high in Brachiaria pastures, forming the base for soil organic matter buildup in these systems (Anderson-Teixeira et al., 2016). Moreover, elements from agricultural systems (the impact of pasture management and grazing on growth and yield) need to be included in the modeling in order to allow predicting future food security (Van et al., 2018). The responses of our predictions showed that the adoption of silvopastoral systems SP_1 and SP_8 stemming from pastures by farmers of zone 3 resulted in a significant absorption of soil CO₂ emissions (Table 4) -levels: -3,37 and -3,39 t CO₂ eq ha^-1 year^-1, respectively. In the tropical zone of Colombia, Parra et al. (2019), using the Tier 1 IPCC approach (2006), showed that a silvopastoral system had the highest potential for offset GHG emissions (-4,8 t CO₂eq ha^-1 year^-1) due to soil C accumulation plus biomass C fixation in Acacia trees. Soil C sequestration by the world's grasslands could offset up to 4% of global GHG emissions (IPCC, 2006). This research showed that 34% of the agricultural systems evaluated (7 and 10 out of the 23 and 21 agricultural systems of zones 3 and 4) turned out to be CO₂eq emitters, mainly due to CL and MDP (Table 4). Default emission factors used in the Tier 1 IPCC model can be further sources of uncertainty, as they may not be representative of SOC changes and, in many cases, they can under- or overestimate soil CO₂ emissions and/or absorptions (FAO, 2018; Silva and Orozco, 2018). Soil C losses rates differ considerably between continuous the crop rotation of soybean/corn CRS/C25 including FT of zone 3 and the MDP B. decumbens MDP116 with MT of zone 4 (Table 4) -levels: ΔSOC = CRS/C 25 = -0,70 and MDP116 = -0,02 t C ha^-1 yr^-1. In fact, easily decomposable materials are fully or partially depleted, and the microbial population and decomposition rate of litter materials decline due to full tillage (Lal, 2018). Reduced tillage and 'no till', residue incorporation, improving soil biodiversity, and mulching enhance the sequestration of carbon in the soil. NRCS conservation practices can be expected to sequester approximately 0,07 to 0,96 t C ha^-1 yr^-1 due to improved soil management in croplands (Chamber et al., 2016).

Representativeness of soil CO₂ emissions by each category of land use and each zone across the Villavicencio area

The simulated total of contributions regarding emission and/ or neutralization of soil CO₂ across the Villavicencio area, zone 1 (comprising AFS), account for -5 416 t CO₂eq yr^-1 (Figure 4, Table 5) (-4,86 t CO₂eq ha^-1 yr^-1 * 1 115 ha), which is possibly due to a high SOC initial state (Table 5) and a higher default value used in the simulation. This is shown in Table 4.

Figure 4 Source: Authors

Silva (2018) found that the SOC initial state has the greatest impact on emissions dynamics. Numerous studies across the globe demonstrate that the use of integrated practices such as AFS can increase soil C gains by 10-60%, thus reducing the carbon emissions associated with the residue inputs by 20-50% (Nair, 2012). For example, a cocoa AFS has the capacity to sequester about 3 t C ha^-1 year^-1, with a reduction of 11 t CO₂eq ha^-1 year^-1 (Jadan et al., 2015). On the other hand, zone 3 showed the highest soil CO₂ emissions across the Villavicencio area, followed by zone 4, due to the contributions of CL (Figure 4), i.e., 7 223 t CO₂eq yr^-1, and 3 809 t CO₂eq yr^-1, respectively. As shown in Table 5, the lowest initial SOC stocks mostly appear in the continuous CL of zones 3 and 4, where there is no land use transition between rice crops and crop rotation and soils are frequently disturbed by full tillage. Soil conservation practices such as crop rotation can fix a large amount of soil organic C and achieve a balance in C storage as long as full tillage is not performed (IPCC, 2006). Efforts for the mitigation of soil CO₂ should focus mainly on zones 3 and 4, adopting more sustainable soil management practices. In a study by Silva (2018), conversion from rice monoculture or crop rotation would increase the soil C stock by about of 12,3 t C ha^-1 in 20 years, equivalent to 0,61 t C ha^-1 yr^-1, which would reduce emissions by 2,27 t CO₂eq ha^-1 year^-1. In zone 3, the contribution of NDP to the mitigation of emissions (-12 252 t CO₂eq yr^-1) (Figure 4) is more influenced by the area planted (ha) than by the rates of soil C gains per year (Table 5). However, the time reference of the NDP does not usually pose a challenge. This change is usually fast -soon after the introduction of the new practice- and eventually stabilizes when a new equilibrium is close by (FAO, 2018). The silvopastoral systems of zone 3 and 4 stemming from the conversion of NDP showed an apparent SOC storage (Table 4) and neutralizations of -9 794 and -462 t CO₂ eq yr^-1, respectively (Figure 4). MDP contributed with 7,40% of the total emissions of zone 3 (Figure 4). Converting degraded grassland to silvopastures could increase SOC stocks ((Mangalassery et al., 2014). Silvopastoral systems are agricultural strategies that can act positively for the potential mitigation of soil CO₂ emissions (Parra et al., 2019). As pointed out by the FAO (2018), there is a significant lack of rigorous data on C sequestration in silvopastoral systems, since a large amount of the root inputs (FI) of trees can be incorporated into these systems.

As climate change research becomes more and more relevant, agroforestry system (AFS) models can play a major role in understanding the interplay between environmental change, SOC, and the functioning of these systems. In this sense, GHG simulations across Villavicencio zones showed that the highest removals took place in the AFS of zone 1 (-5 416 t CO₂ eq yr^-1). However, the better matches for AFS (between observations and simulations) were obtained in zones 3 and 4 in comparison with zone 1 (RMSE/n = 0,05, 0,17, and 1,61 t C ha^-1). In zone 1, there may be an overestimation of the modeled SOC in AFS. Our simulation analyses clearly indicate that a pathway for the reduction of soil CO₂ emissions is possible through a wide-scale adoption of different types of AFS that can optimize soil management factors for increased SOC. Silvopastoral systems (SP) have gained large attention during the last decades due to their SOC accumulation and should be considered to improve Moderately Degraded Pastures (MDP). In zones 3 and 4, the potentially significant negative impacts on soil CO₂ emissions (7 223 and 3 809 t CO₂ eq yr^-1, respectively) are due to SOC losses in Intensive Cropland (CL), which account for 30 017 and 6 925 t CO₂ eq yr^-1. In this sense, in CL, the reduction of soil CO₂ emissions can generate a large portion of the needed mitigation through the adoption of cropping rotation and soil management practices such as minimal tillage and higher above-crop residues. AFS are important Factor Land Use (FLU) to incentive low-carbon footprint agriculture, as a plan by the Colombian Government to reach its GHG emission reduction targets.