Revisiting IPCC Tier 1 coefficients for soil organic and biomass carbon storage in agroforestry systems

We conducted a literature review of available studies that estimated SOC stocks (or the information necessary to calculate them, i.e. SOC content and bulk density), and/or biomass C stocks. The research process included use of several research engines and knowledge platforms, namely ISI—Web of Knowledge, Google Scholar, and Scopus. A key-word search was performed using the following keys: biomass OR soil AND ('carbon stock*' OR 'carbon pool*' OR 'carbon sequestration' OR 'carbon concentration') AND (agroforest* OR parkland* OR homegarden OR multistrata OR hedgerow OR windbreak OR shelterbelt OR 'live fence' OR 'tree intercrop*' OR silvo*arable OR silvo*pasture OR 'rotation* wood*' OR tree*fallow* OR (tree* AND 'improve* fallow*') OR (tree* AND relay*crop*) OR (tree* AND alley*crop*)).

Several parameters were deemed necessary for the data to be contained in the database, specifically: the previous land use, the depth of measurement and the time frame over which the LUC had occurred should be reported. In addition, studies also had to provide the location or climate data of the study sites in order to classify the data by climate regimes according to the IPCC guidelines (IPCC 2007). If the required data were not present in the text or in tables, data were extracted from graphs using the WebPlotDigitizer software. In the case that a study included multiple sites, they were treated as separate data points. In general, only data from primary sources were collected in the database. If data from a review of primary sources were included in the database, the data were only reported once.

As a result of this systematic literature review, we collected data from a collection of 50 peer reviewed studies that reported biomass C storage and 72 studies that reported SOC stock changes in agroforestry systems. In total, 324 and 218 observations were obtained for biomass and SOC storage, respectively. The full database is available in appendix 1 and 2, available online at stacks.iop.org/ERL/13/124020/mmedia.

In general, the majority of selected publications reported SOC stock and only a few used the equivalent soil mass approach. In the case that the SOC concentration and bulk density were reported, the SOC stock was calculated. Only six publications did not report bulk densities, which were estimated using the mean bulk density per soil type as proposed by Batjes (1996).

In the case that both a synchronic (SOC stocks measured in the agroforestry and in an adjacent control plot) and diachronic (SOC stocks measured in the same plot before and after LUC) approaches were presented for the same site, the priority was given to the diachronic approach as the one most widely considered among the scientific literature as reliable (Costa Junior et al 2013). In the case that SOC stocks were presented at different periods of time, only one date was identified in order to avoid dependency of observations and priority was given to the most recent measurements as SOC stock changes are usually not detected during the first several years due to measurement uncertainties (Smith 2004). In the case that SOC stocks were measured at different depths, only the value measured at 0–30 cm was taken into account. If stocks were only measured at 0–20 and 0–40 cm, the deepest depth (0–40 cm) was considered because it included the ploughed horizon and most tree and crop roots. If only one depth was available but did not correspond to 0–30 cm, the observations was still considered.

Zoom In Zoom Out Reset image size

Two different methods were applied to calculate SOC storage or loss rates (figure 1). The first one (equation (1)), used by the IPCC (2006) for the Tier 1 method, is a stock change factor (F_LU, dimensionless). It indicates a relative change in SOC stocks in the new land use compared to the previous one. It is also called a response ratio (Ogle et al 2004, 2005), or management factor (Maia et al 2013). F_LU higher than 1 correspond to a SOC storage, while F_LU lower than 1 correspond to a SOC loss.

$$\begin{eqnarray}&&{F}_{LU}=\displaystyle \frac{Agroforestry\,SOC\,stock}{Control\,SOC\,stock}.\end{eqnarray} \tag{ 1 }$$

The second method (equation (2)) was used to estimate absolute changes in SOC stocks. SOC storage/loss rates (t C ha⁻¹ yr⁻¹) were then calculated using the following formula (figure 2):

$$\begin{eqnarray}&&\begin{array}{l}SOC\,storage/loss\,rate\\ =\,\displaystyle \frac{Agroforestry\,SOC\,stock-Control\,SOC\,stock}{Age},\end{array}\end{eqnarray} \tag{ 2 }$$

where SOC stocks are expressed in t C ha⁻¹, and the age corresponded to the agroforestry age in years.

Zoom In Zoom Out Reset image size

If only biomass was reported in a study, the carbon fraction (CF) of dry matter of 0.47 was applied to convert it into a carbon stock (IPCC 2006).

The root:shoot ratio (R) is defined as the ratio of belowground biomass to aboveground biomass and is the primary method used by nations to estimate below ground biomass and C stocks for NGHGIs (Mokany et al 2006). In the case that the root:shoot ratio was not empirically-derived, default estimates for perennial woody vegetation recommended by IPCC (2006) were used for each climatic zone: 0.27 in Boreal, 0.26 in Temperate, and 0.24 in Tropical (Cairns et al 1997).

The different types of agroforestry systems considered in this manuscript are presented in table 1. This classification is adapted from Nair et al (2009) where the same term 'alley cropping' was used for very different agroforestry systems in tropical and temperate regions. In this study, we make a distinction between tropical alley cropping, which involve dense alleys of fast-growing, usually leguminous woody species, and temperate silvoarable systems, which contain a low numbers of trees, usually for timber, in rows spaced widely enough to allow for mechanization (table 1).

Four main land conversions were studied for SOC storage/loss, cropland to agroforestry, forest to agroforestry, grassland to agroforestry, and plantation to agroforestry. Four publications also reported conversion from abandoned cropland to agroforestry (Diels et al 2004, Swamy and Puri 2005, Baumert et al 2016, Bright et al 2017). These abandoned croplands are defined as former long-term croplands uncropped (natural fallow) a couple of years before the establishment of the agroforestry system. Due to insufficient data, they were included in the category 'cropland to agroforestry'. The effect of the previous land use on annual biomass increment rates was considered negligible (IPCC 2006).

All the graphs and statistical analyses were performed using R software version 3.1.1 (R Development Core Team 2013), at a significance level of <0.05.

In general, conversion from croplands to agroforestry systems resulted in increased SOC stock but with large variation. A few studies however reported a SOC loss (Baumert et al 2016), but with no clear explanation as to the driver. It could be due to soil disturbance during tree planting, followed by an erosive event. The SOC storage rate depends on various agroforestry system characteristics, such as tree density, age and species but also on management factors (Kim et al 2016), including pruning, soil tillage and fertilization (Feliciano et al 2018). Initial conditions such as SOC stock in the previous cropland, and local pedoclimatic factors (e.g. rainfall, soil texture) are also important drivers of SOC storage (Corbeels et al 2018, Feliciano et al 2018). However, the amount of C input to the soil is probably one of the main factor explaining increased SOC stocks in croplands converted to agroforestry (Cardinael et al 2018, Fujisaki et al 2018).

Conversion from forests to agroforestry systems generally induced SOC loss. This result confirms recent findings (Chatterjee et al 2018, de Stefano and Jacobson 2018, Feliciano et al 2018, Shi et al 2018). However, the loss is usually less than if forests were converted to croplands (Schmitt-Harsh et al 2012, Norgrove and Hauser 2013). Globally, conversion from grasslands to agroforestry systems did not improve SOC stocks. This result is in accordance with Poeplau et al (2011) who found no difference in SOC stocks of afforested grasslands in Europe, and with Fujisaki et al (2015) who found slightly higher SOC stocks in grasslands than in forests. However, we did not explore here the effect of the grassland management on SOC storage. Converting degraded grasslands to silvopastures could increase SOC stocks (Mangalassery et al 2014).

As pointed out by Nair (2012), there is a significant lack of rigorous data on C sequestration in agroforestry systems. To assess changes in SOC stock and storage/loss following a LUC, some basic data are required: a description of the previous land use, SOC stocks or SOC content and bulk densities in both the previous and new land use, soil depth considered, and time span since conversion. Unfortunately, the literature review showed that these basic data were not always present. The concerned papers were therefore not used to estimate the new emission factors. Many authors have, for instance, measured SOC in agroforestry but not in a reference system, preventing assessment of whether soil had lost or gained organic C (Pandey et al 2000, Roshetko et al 2002, Isaac et al 2005, Mungai et al 2006, Muñoz et al 2007, Singh and Sharma 2007, Smiley and Kroschel 2008, Saha et al 2009, Labata et al 2012, Seddaiu et al 2013, Simón et al 2013, Guimarães et al 2014, Sitzia et al 2014, Nath et al 2015, Ramos et al 2018, Sun et al 2018). Some publications compared SOC stocks in agroforestry and in a « reference » system, but this reference system was different from the land use present before conversion to agroforestry (Drechsel et al 1991, Materechera and Mkhabela 2001, Isaac et al 2003, Nyamadzawo et al 2008, Takimoto et al 2008, Howlett et al 2011b, Cardinael et al 2012, Gelaw et al 2014, Jacobi et al 2014, Ehrenbergerová et al 2016, Rajab et al 2016). This could be very problematic, such as when a forest is converted to either a shaded-perennial crop system or a monocrop plantation. Comparing SOC stocks in the agroforestry and in the plantation could result in an apparent SOC storage, while in reality the conversion of the forest to the agroforestry system usually leads to a loss in SOC.

The age of the agroforestry system or the time span since conversion from a previous land use is also often missing in most studies, making it impossible to estimate the rate of change in SOC stock (Kater et al 1992, Kessler 1992, Walter et al 2003, Wade et al 2010, Labata et al 2012, Alvarado et al 2013, Simón et al 2013, Frazão et al 2014, Goswami et al 2014, Rocha et al 2014, Sitzia et al 2014, Baah-Acheamfour et al 2015, Jadan et al 2015, Asase and Tetteh 2016, Tumwebaze and Byakagaba 2016). Several articles only reported SOC content, and bulk densities were missing (Drechsel et al 1991, Mazzarino et al 1993, Chander et al 1998, Kang et al 1999, Kaur et al 2000, Pandey et al 2000, Singh and Sharma 2007, Kumar et al 2010, Mao et al 2012, Cardinali et al 2014, Sitzia et al 2014, Dubiez et al 2018, Nijmeijer et al 2018, Sun et al 2018).

The loss of SOC following a LUC can be very quick while it usually takes much longer to gain SOC (Smith 2004), especially with low tree densities. The studies using a diachronic approach to quantify SOC changes in agroforestry are rare (Beer et al 1990, Mazzarino et al 1993, Maikhuri et al 2000, Oelbermann et al 2004, Sierra and Nygren 2005, Swamy and Puri 2005, Raddad et al 2006, Lenka et al 2012, Singh and Gill 2014, Wang et al 2015), most of them have been performed using a synchronic or chronosequence approach. The diachronic approach has been recognized to be a more accurate method to assess SOC changes than other methods (Costa Junior et al 2013). More agroforestry trials using this approach should be established. LUC is often associated with a modification of soil bulk density, and a calculation of SOC stocks on an equivalent soil mass basis instead of on a fixed depth is recommended (Ellert and Bettany 1995, Ellert et al 2002, Wendt and Hauser 2013). Surprisingly, very few studies on SOC storage in agroforestry followed this recommendation (Bambrick et al 2010, Cardinael et al 2015a, Upson et al 2016, Cardinael et al 2017, Wiesmeier et al 2018).

The main difficulty to properly assess SOC changes in agroforestry systems compared to other land uses is the spatial heterogeneity. Scattered trees induce a gradient in organic inputs to the soil (Cardinael et al 2018), and an large number of soil samples have to be taken to explore this heterogeneity (Cardinael et al 2015a, Upson et al 2016). Developing technologies, such as visible and near-infrared spectroscopy could be used to reduce the cost and the time to monitor SOC changes (Cambou et al 2016, Viscarra Rossel et al 2016).

Roots of agroforestry trees can grow very deep in the soil due to competition with the associated crops and to soil tillage (Cardinael et al 2015b). Great uncertainties exist on the fate of fresh organic inputs in deep soil layers and on their interaction with older soil organic matter, such as priming effect (Fontaine et al 2007). However, since a large amount of root inputs can be incorporated into these systems (Germon et al 2016, Cardinael et al 2018), it would probably be very valuable to expand research on this topic. Cardinael et al (2018) indeed found that SOC profiles (2 m depth) in a long-term agroforestry systems where only well described if priming effect was included in the model. Only few studies have measured SOC storage in deep soil layers of agroforestry systems (Haile et al 2008, Howlett et al 2011a, Upson and Burgess 2013, Cardinael et al 2015a), more studies are required. Moreover, very few studies have quantified both above and belowground biomass of agroforestry systems, and the use of the root:shoot ratio determined on forest ecosystems could be problematic. Due to their low density, agroforestry trees usually grow faster than forest trees (Balandier and Dupraz 1998), and they also benefit from crop inputs (fertilization), and their root systems compete with annual crops. Moreover, agroforestry trees are often pruned. Carbon allocation between aerial and belowground parts of the trees might therefore be modified. The use of non-destructive and repeatable methods, such as ground penetrating radar could be a good way to acquire more data on agroforestry root systems (Borden et al 2014).

Several tools use tier coefficients to provide an estimation of the C-balance associated with the adoption of improved land management options, as compared with a 'business as usual' scenario. This is for instance the case of the EX-ACT (EX-Ante Carbon-balance Tool) tool developed by the Food and Agriculture Organization of the United Nations (FAO), providing ex-ante measurements of the mitigation impact of agriculture and forestry development projects, estimating net C balance from GHG emissions and C sequestration (Bernoux et al 2010). EX-ACT has been developed using primarily the IPCC (2006) Guidelines for NGHGIs (IPCC 2006), complemented by other existing methodologies and reviews of default coefficients. Default estimates for mitigation options in the agriculture sector are mostly from the 4th Assessment Report of IPCC (2007) (Smith et al 2007).

Our newly derived factors can be used to improve these tools to estimate C sequestration in any country with a minimal amount of data by using the IPCC method. These factors may have systematic biases when not representative of management effects, climate, or soils in a particular region. Consequently, nations with available resources should consider synthesizing data from local field experiments to generate country-specific factors. However, these new estimates certainly represent a significant improvement to better account for the diversity of agroforestry systems. They were all previously included into the 'perennial crops' category, together with vineyards and orchards, and are now split into eight main types of agroforestry systems per climate and region. The presentation of emission factors for the biomass and soil using the same classification of agroforestry systems represents another important added value of this study.