Potential for soil organic carbon sequestration in grasslands in East African countries: A review

Grasslands are areas where the vegetation is dominated by grasses (Poaceae family). They are an intrinsic part of both rangelands and pasturelands and constitute about half of the global land area. Grassland provides different functions, such as livestock feed, environmental regulation, sequestration of soil carbon, biological diversity and maintenance of soil health (Carlier, ROTAR, Vlahova, & Vidican, 2009; Hönigová et al., 2012; Ribeiro, Fernandes, Dalila Espırito-Santo, 2014). Soil organic carbon (SOC) content is a key indicator of soil fertility and grassland productivity, and SOC sequestration is considered a means to mitigate climate change through Received: 17 April 2018 | Revised: 14 August 2019 | Accepted: 1 October 2019 DOI: 10.1111/grs.12267

Globally, grasslands contain more than one-third of aboveground and belowground stocks of organic carbon (Allen-Diaz et al., 1996;Mengistu, 2006). Fisher et al. (1994) and Mishra, Ranade, Joshi, and Sharma (1997) highlight the potential for SOC sequestration by deep-rooting African grasses, such as Andropogon gayanus and Brachiaria humidicola. Guo and Gifford (2002) estimated that the conversion of cropland to grassland increases SOC by about 19%, while recent meta-analysis shows an increase of 3%-5% (Conant, Paustian, & Elliott, 2001;Conant et al., 2017. The higher SOC stocks in grasslands are due to their perennial nature which results in constant carbon inputs from aboveground vegetation and the large quantities of carbon to the subsoil via root exudates and decomposing deep roots (Zimmermann, Dauber, & Jones, 2012). However, grassland management practices have substantial effects on the turnover rates of soil organic matter, carbon inputs and soil nutrients (Blair, Lefroy, & Lisle, 1995), leading to varying effects on SOC.
For instance, Dlamini, Chivenge, Manson, and Chaplot (2014) reported a 90% decline in SOC in the heavily grazed highlands of KwaZuluNatal in South Africa. Globally, the increased grassland degradation caused by overgrazing has been shown to substantially reduce SOC, mainly in the dry climates, with between 1.2% and 4.2% of the grassland SOC stocks likely to be lost . Grassland rehabilitation through adopting improved management has been suggested as one way to reverse this loss. Chaplot, Dlamini, and Chivenge (2016) show that, compared to traditional free grazing, livestock enclosures where the livestock intensity is increased in one area for short duration increases SOC in Southern Africa.
About 75% of Eastern Africa is dominated by managed grassland systems. However, at the regional level little is known about the types of grassland managements and their impacts on SOC, with the existing studies focussing on small localized areas. Given the large land size of grasslands in East Africa, there is a yet untapped potential for C storage in the grasslands as an option to mitigate climate change. However, there is a need to review studies undertaken in the past to better understand the current and future potential effects of grassland management practices. Therefore, the aim of this study was review observation-based studies on grass-

| Key terminology
Terminology referring to grasslands varies based on context. For instance, in some cases rangeland and grassland are used interchangeably, while in others grasslands are considered part of different land use classes, e.g., rangelands and pasturelands. The terminology used in this review paper was as follows:

Rangelands
Land on which the native vegetation is predominantly grasses, grasslike plants, forbs or shrubs. This includes land revegetated with such species naturally or artificially by routine management, mainly through manipulation of grazing.

Pasturelands
Enclosed tracts used for grazing distinguished by periodic cultivation to maintain introduced (non-native) forage species and use of agronomic inputs such as irrigation and fertilization (Holechek, Pieper,& Herbel, 2001).

Grasslands
Areas where the vegetation is dominated by grasses (plants of the Poaceae family) and which are an intrinsic part of rangelands and pasturelands.

| Study area
The present study covered six countries (Burundi, Ethiopia, Kenya, Rwanda, Tanzania and Uganda) in East Africa (Figure 1). The total grassland coverage in this area is an estimated 80 Mha (ESA, 2017

| Literature search and selection
A systematic literature search was undertaken using Google Scholar and key terms such as: "SOC" and "grasslands," combined with country names. The results were compiled in a database for further analysis. From the total number of hits in each search, subsets were derived in three steps by applying criteria on the title (subset 1), applying criteria on full-text availability (subset 2) and applying criteria on the full text itself (subset 3). Subset 1 was derived by discarding papers where the title clearly indicated that the study covered another geographical region or another land use. The results of the literature search are presented in Table 1. In total, 23 publications that quantified SOC concentration and/or SOC sequestration potential, or for which we could derive those values, were identified. No research work was found for Burundi, and extractable data were only available for Ethiopia, Kenya, Uganda and one regional study for East Africa as a whole. Of the total number of studies, seven reported data from before-after intervention cases, so sequestration rates were computed.
The fact that only 23 relevant publications were located indicates a scarcity of hard data on the SOC sequestration potential of grasslands in the region. About half of the studies were carried out in Kenya (8 studies) and Ethiopia (6 studies) ( Table 2), which indicates a geographical bias in empirical evidence on SOC sequestration potential. Moreover, the selected countries differ in size and are therefore not directly comparable. In addition, a language bias could have excluded studies produced in French-and/or non-English-speaking countries (e.g., Burundi and Rwanda). In the retrieved publications, eight different management practices were examined: enclosure; improved management (using rotations and by adding different inputs such as manuring, fertilizer, etc); free grazing; light grazing; heavy grazing; fencing; restoration measures; and conversion from natural forest to grazing.

| Variables and formulae used to calculate SOC stocks and sequestration rates
In the analysis, direct SOC stock and C sequestration rate values were used if these were available in the studies reviewed. In the absence of values on stocks and sequestration rates, SOC concentration (SOC%) and bulk density (BD) were used to calculate SOC stock for the soil layer specified. Total SOC stocks (Mg C/ha) were then calculated as a product of SOC%, BD and sampling depth (Equation 1). The sequestration rate of SOC was calculated by dividing the SOC stock by the number of years since establishment of each specific management practice/intervention (Equation 2). In the absence of bulk density data, an average BD value for the respective countries (~1.4 g/cm 3 on average for all countries) was taken from the global soil mapping database SoilGrids.org (ISRIC, Wageningen, the Netherlands).

| SOC sequestration potential of grasslands in East Africa
The SOC sequestration potential reported in the 23 publications is summarized in Table 3. There was great variation in both the reported amount of SOC in the soil and the SOC sequestration rate under different management options. This may be due to the small number of studies available, supporting the need for additional experiments and/or observational studies for accurate assessments of

SOC sequestration potential in grassland soils in countries in East
Africa.
The SOC amounts and the rate at which SOC is sequestered in grasslands under different management interventions in the selected East African countries are presented in Table 3. Soil sampling depths behind the reported data were not uniform in the papers reviewed, and therefore, it was difficult to draw comparisons between countries and management intervention measures. Three different soil depth categories (0-0.1 m, 0-0.3 m and 0-1.0 m) were covered specifically by the soil samples analyzed in the 23 papers reviewed.
However, no soil depth values were mentioned for the SOC values reported in three papers.
From the reported data, we inferred that the SOC stocks at 0-0.3 m soil depth in grasslands range from 3 Mg C/ha (Ethiopia) to 93 Mg C/ha (Kenya) ( Table 3). The reported impact of management interventions on SOC sequestration rate varied from 0.1 to 3.1 Mg C ha -1 year -1 for 0-1.0 m soil depth in Kenya, while lower sequestration rates (0.10 Mg C ha -1 year -1 ) were reported in general for some management practices in Ethiopia. The highest rate was reported for fencing (3.10 Mg C ha -1 year -1 for 7-10 year fenced grasslands in Kenya), followed by conversion from natural forest to grazing (2.4 ± 2.1 Mg C ha -1 year -1 in Uganda) and enclosure (1.8 ± 0.2 Mg C ha -1 year -1 ). The extent to which grazing is practiced was also reported to affect the sequestration rate, e.g., light grazing (0.77 ± 0.4 Mg C ha -1 year -1 ), giving a higher sequestration rate than heavy grazing (0.9 ± 0.5 Mg C ha -1 year -1 ) for 0-0.1 m soil depth in Ethiopia.

| Geographical location, soil and climate
Two factors are important for increasing the amount of carbon sequestered in grasslands: 1) Carbon input to the soil by net primary production, and 2) the rate of organic matter decomposition (US Department of Energy, 1999). More than two-thirds of the carbon stored in grasslands is located below ground (Burke et al., 1989;Parton, Schimel, Cole, & Ojima, 1987). Grassland SOC stocks Age (

| Grass productivity
The production and decomposition of plant biomass can influence a number of ecosystem processes (Windham, 2001). Hence, plant biomass production and decomposition can determine carbon inputs to the soil profile. Moreover, plant allocation between aboveground and belowground parts, and between shallow and deep roots, can leave distinct imprints on the distribution of soil carbon with depth (Jobbágy & Jackson, 2000). The high carbon input derived from high plant root biomass production in grassland systems can provide the potential to increase soil organic matter content, which is a key factor for enhanced SOC storage in grasslands (Farage et al., 2004). Therefore, maximizing productivity and root inputs in grassland systems is crucial in increasing their SOC sequestration (Trumbmore, Davidson, Barbosa de Camargo, Nepstad, & Martinelli, 1995). However, some studies report that a large root biomass supports substantial soil microorganism populations and their metabolic processes, thus contributing significantly to soil organic matter decomposition and carbon turnover (Kuzyakov, 2002).
Accumulation of SOC in grasslands is also a function of the length of time for which the land remains under grasses (Neill et al., 1997).
Therefore, regardless of technologies or mechanisms, grassland age must be considered when assessing longer-term carbon storage potential.

| Grazing management
Managing grasslands is crucial, since grasses are the main feed source for livestock production systems (Conant et al., 2001;Ni, 2002). Grazing management aims at optimal use of grasslands for animals, which can be achieved by manipulation of one or more of three variables: species/type(s) of animals, number of animals and distribution of animals.
Grassland grazing systems are classified as continuous, seasonal or rotational (Pratt & Gwynne, 1977). Rotational grazing is the typical grazing management system for grasslands in East Africa and gives the land time to recover from past overgrazing (Pratt & Gwynne, 1977).
Poor management of grasslands, such as excessive free grazing, has a major negative influence on grassland C cycling, affecting not only transfers between vegetation and soil compartments, but also ecosystem input and output flows (Franzluebbers et al., 2012;Holechek et al., 2001;Scurlock & Hall, 1998). In the long run, these alterations may have important consequences for the capacity of managed grasslands to store SOC (Holechek et al., 2001). In contrast, improved grazing management is reported to increase SOC storage in grasslands (Conant et al., 2017;Schuman, Janzen, & Herrick, 2002).
In Africa, increased human and livestock populations on a shrinking land area, in combination with drought, which has always been a part of Africa's climate, have intensified ecological degradation of terrestrial ecosystems, resulting in food shortages and famines (Holechek et al., 2001;SIDA, 2010). East African countries, particularly Ethiopia, Kenya and Tanzania, have some of the highest population growth rates in the world (Solomon et al., 2015), which could potentially force more intensive land use and increasingly shift farming onto grasslands (Holechek et al., 2001;SIDA, 2010).
Degradation of remaining grasslands as a consequence of land conversion has become a major concern, as grasslands support nearly a billion domestic animals worldwide (Mengistu, 2006). In sub-Saharan Africa, grassland degradation is much more obvious around watering points because of overstocking (SIDA, 2010). Hence, maintaining and improving the health of grassland ecosystems is a major challenge (Holechek et al., 2001).
Grasslands in dryland areas of East African countries are generally degraded (Farage et al., 2004). Conversion of grasslands to marginal agriculture causes low productivity, which leads to SOC losses by erosion and decomposition (Franzluebbers et al., 2012;Scurlock & Hall, 1998). According to FAO (2004), as a consequence of cropland conversion to grasslands in dryland areas, SOC sequestration rate declines

| SOC sequestration potential of grasslands in East Africa
For many years, crop breeders and growers have been selecting specifically for tropical grasses with deep, large root systems that can exploit nutrients and water in deeper soil layers (Fisher et al., 2007). Deep root penetration into the subsoil and deposition of root biomass are believed to be the primary vehicle for SOC sequestration (Kell, 2011;Kuzyakov, 2002;McKenzie & Mason, 2010;Nguyen, 2003 McKenzie & Mason, 2010;Schwenke et al., 2014). Even grasslands subjected to controlled grazing have higher SOC levels than croplands, primarily due to the lower losses of SOC that occur when the soil is not tilled (Farage et al., 2004). However, estimates of carbon inputs from biomass in grassland savannah vary significantly, from 0 Mg C ha -1 year -1 in dry periods to between 5 and 15 Mg C ha -1 year -1 in the rainy season (Vågen, Lal, & Singh, 2005).

Grasslands can play a significant role in carbon sequestration
if grazing management is optimized. Alternatively, if grasslands are exposed to prolonged overgrazing and soil degradation, the consequence is high SOC losses (Conant, 2012;Farage et al., 2004).
Improved grazing management has been estimated to increase SOC storage in US rangelands by 0.1-0.3 Mg C ha -1 year -1 , while newly established grasslands have been shown to sequester as much as 0.6 Mg C ha -1 year -1 (Conant et al., 2017;Schuman et al., 2002).
There is much less empirical evidence available for East Africa and elsewhere in Africa, but the data at hand indicate that pasture improvements, including fertilization, liming, irrigation and sowing of more productive grass varieties, generally result in relative gains of 0.1-0.3 Mg C ha -1 year -1 (Sanderman et al., 2009).
Grasses also have the potential to sequester carbon in previously degraded soil. For example, in a study where switchgrass (Panicum virgatum L.) was grown on degraded land, SOC sequestration rate increased by an estimated 12% over 10 years (Garten & Wullschleger, 2000). Miscanthus and found that the rate increased with increasing mean annual temperature. A study conducted by International Center for Tropical Agriculture (CIAT) scientists on gamba grass (Andropogon gayanus), which has a root system penetrating to 1.0 m depth in tropical soils, revealed outstandingly high SOC sequestration rates of up to 195 Mg C ha -1 year -1 (Vietmeyer,1997 ).
Grasslands are the basis of livelihood for East African communities, which depend heavily on livestock (Mengistu, 2006). Animal numbers are expanding in the region to meet the needs of the growing human population and the animals are depending on a shrinking rangeland area, which is putting rangelands under pressure (Holechek et al., 2001). In addition, the high demand for different kinds of products from rangeland causes frequent modification of land use, which results in changes in the relative area of different rangeland uses (Mengistu, 2006). In East Africa, a great number of wild animals depend on rangelands that they must share with humans and their herds and flocks ( Figure 2). Grasslands of East Africa therefore need careful management, because they have great potential for feed production and SOC storage (Mengistu, 2006).

| Opportunities for governance of grasslands in East Africa
African soils in general have great potential for storing SOC, provided that appropriate land management practices and carbon input measures are undertaken (Solomon et al., 2015). In Ethiopia, agricultural production and land productivity remain low, due to ever-increasing human population pressure on land resources, in combination with unfavorable land policies (small land holdings and land ownership issues). Shiferaw, Hurni, and Zeleke (2013) report some success for the Ethiopian Sustainable Land Management Program (SLMP), which aims at improving soil quality and productivity by restoring degraded lands, including grasslands, using best land management practices and climate change mitigation and adaptation measures (https :// www.giz.de/en/world wide/18912.html). The program reflects the goals of the growth and transformation plan for Ethiopia, one of which is to enhance SOC stocks in Ethiopian soils, including grassland soils (SIDA, 2010;Solomon et al., 2015).
In the studies reviewed in the present analysis, it is widely suggested that rangeland/grassland regeneration and protection practices could be a sustainable option for SOC sequestration, contributing to climate change mitigation at both national and regional level. In addition to the climate change benefit, storage of carbon in grassland soils through improving the productivity of grass biomass offers economic returns (Singh, Guleria, Rao, & Goswami, 2011). However, grasslands are complex ecosystems. For instances, restoring and sequestering carbon may come at the cost of reduced feed availability when free grazing is restricted, negatively affecting pastoralists in the short-term. Similarly, while conversion of grasslands into cropland may benefit smallholder farmers, this land use conversion can entail further losses of SOC.

| Data availability
Field studies on the SOC sequestration potential of rangeland systems in East Africa form only a small part of research on the subject globally (Tubiello, Soussana, & Howden, 2007 (Fisher et al., 1994) or even down to 6.0 m (Carvalheiro & Nepstad, 1996;Sommer, Denich, & Vlek, 2000). In addition, variations in the vertical and horizontal distribution of SOC in tropical, subtropical and temperate soils need to be assessed.

| SOC sequestration rates under planted cut and carry forages
Another knowledge gap concerns SOC sequestration rates under planted cut and carry forages on cropland, which was not part of this review. Little is known about the SOC dynamics in these systems, and measuring C sequestration is a challenge, as the areas planted are patchier and integrated with other (food) crops. Hence, computer modeling might play a role. This could be linked to the interventions/ technologies CIAT is working on in Africa, which focus on cut and carry because rangelands are often communally managed and there is no individual incentive to introduce improvement measures.

| CON CLUS IONS
Grasslands have great potential to sequester SOC due to their fast establishment, growth and perennial biomass production, and improved management of grasslands can increase this potential, particularly on degraded soils. However, this review revealed a lack of published studies on the SOC sequestration potential of grasslands in East Africa, so it is difficult to draw any firm general conclusions.
Overall, it was found that: • More studies are needed on the SOC sequestration potential of East African grasslands when promising management practices are adopted.
• There are tremendous variations in reported SOC sequestration rates. In combination with the low number of individual studies, this is likely to lead to biased and uncertain estimates of SOC sequestration potential.
• The effects of grass type and management practices on SOC storage potential need to be fully quantified, including extending soil sampling to greater depths and under deep-rooted perennial grasses.
• Ways of stabilizing carbon stored in soils so that it is not released by management-induced disturbances need to be identified.
Given the vast areas under grasslands worldwide and the tendency for grassland systems to store more carbon in soils, their carbon sequestration potential should be compared with that of other land use systems. This could help improve decision making on prioritizing SOCenhancing measures and policies.

ACK N OWLED G EM ENTS
We thank the AgriFoSe2030 program and Sida for the financial support provided. Salaries were partly funded by Formas/SIDA (contract: 220-1975-2013)