Drivers of forage provision and erosion control in West African savannas—A macroecological perspective

Agriculture, Ecosystems and Environment 251 (2018) 257–267

Contents lists available at ScienceDirect

Agriculture, Ecosystems and Environment journal homepage: www.elsevier.com/locate/agee

Drivers of forage provision and erosion control in West African savannas—A macroecological perspective

MARK

⁎

Reginald T. Guuroha,b,c, , Jan C. Ruppertb,e,f, Jessica Fernera,b, Kristijan Čanakb,g, Sebastian Schmidtleina,d, Anja Linstädterb a

University of Bonn, Geographical Institute, Centre for Remote Sensing of Land Surfaces, Walter-Flex-Straße 3, D-53113 Bonn, Germany University of Cologne, Botanical Institute, ZülpicherStraße 47b, D‐50674 Cologne, Germany, c CSIR-Forestry Research Institute of Ghana, P. O. Box 63, KNUST, Kumasi, Ghana d Karlsruhe Institute of Technology, Institute of Geography and Geoecology, Kaiserstraße 12, D‐76131 Karlsruhe, Germany e Risk and Vulnerability Science Centre, University of Limpopo, Private Bag X 1106, Sovenga 0727, South Africa f University of Tübingen, Institute of Evolution and Ecology, Auf der Morgenstelle 5, D‐72076 Tübingen, Germany, g Center for Development Research (ZEF), University of Bonn, Walter-Flex-Straße 3, D-53113 Bonn, Germany b

A R T I C L E I N F O

A B S T R A C T

Keywords: Ecosystem service Rangeland Forage provision Vegetation properties West African savanna Environmental drivers

Rangelands’ ability to provide ecosystem services (ESs) depends on ecosystem properties and functions, which are interactively driven by biophysical and land-use drivers. In West Africa’s savanna rangelands, the relative importance of these drivers for ES supply is still poorly understood, hampering the identification of appropriate management strategies. In this context, trade-offs between the ES of forage provision and the regulating ES of erosion control are of particular importance. Taking a macroecological perspective, we aimed at detecting consistent patterns in ES drivers and identifying good predictors. The study area comprises a steep gradient of climatic aridity across West Africa’s Sudanian savannas from northern Ghana to central Burkina Faso, in combination with local gradients of land-use intensity and topo-edaphic conditions. We used aboveground biomass, metabolisable energy and metabolisable energy yield as proxies for forage provision, and the cover of perennials in the grass layer as a proxy for erosion control. Linear mixed-effect models and model selection were used to test relationships between multiple environmental variables and ES proxies. We found differential responses of ES proxies to environmental drivers. Vegetation properties were important for all ESs. Antecedent rainfall was the most important predictor of aboveground biomass, while plants’ phenology and land-use were most important for metabolisable energy. Environmental variables (such as aridity, soil properties and grazing intensity) mediated via vegetation properties were the most important predictors of erosion control followed by the direct effect of climatic aridity. Our finding that antecedent rainfall was more important for forage provision than climatic aridity implies that the effects of long-term climatic aridity may in a given year be overridden by current season’s precipitation particularly in case of a good rain year. The observed importance of land-use and vegetation properties implies that well-conceived adaptation strategies could mitigate potential negative effects of climate change.

1. Introduction Ecosystem services (ESs) are the benefits that society derives from nature (MEA, 2005). Ecosystems used as rangelands deliver multiple ESs (Sala and Paruelo, 1997; Sala et al., 2017). Among them, forage supply is the most prominent provisioning ES; it supports approximately 50% of global livestock production (MEA, 2005). Rangeland ecosystems also deliver various supporting, regulating and cultural ESs, with erosion control being of major importance (Orwin et al., 2015;

Sala et al., 2017). Accelerated soil erosion is accompanied by the loss of soil-mediated ESs such as nutrient and greenhouse gas regulation (Delgado-Baquerizo et al., 2013; Orwin et al., 2015). Rangelands’ ability to provide essential ESs is mostly determined by biophysical factors such as climate and topo-edaphic conditions, and by land-use (Sala et al., 2017). These factors are also the major determinants of vegetation structure in savanna and grassland rangelands (McNaughton, 1983; Oesterheld et al., 1999; Augustine, 2003). Hence, rangelands’ ability to provide essential ESs is intimately linked to

⁎

Corresponding author at: CSIR-Forestry Research Institute of Ghana, P.O. Box UP 63, KNUST, Kumasi, Ghana. E-mail addresses: [email protected] (R.T. Guuroh), [email protected] (J.C. Ruppert), [email protected] (J. Ferner), [email protected] (K. Čanak), [email protected] (S. Schmidtlein), [email protected] (A. Linstädter). http://dx.doi.org/10.1016/j.agee.2017.09.017 Received 10 August 2016; Received in revised form 31 July 2017; Accepted 19 September 2017 0167-8809/ © 2017 Elsevier B.V. All rights reserved.

Agriculture, Ecosystems and Environment 251 (2018) 257–267

R.T. Guuroh et al.

environmental drivers on vegetation-mediated ESs mostly have a local scale (e.g. Nacoulma et al., 2011; Schmidt et al., 2011), which hampers detecting consistent patterns of ES drivers at the regional level. The third challenge relates to the fact that biophysical and land-use drivers do not only have direct effects on ES supply, but also exert indirect effects via their imprint on ecosystem structure and function (de Bello et al., 2010; Gaitán et al., 2014). Although these indirect effects are often subtler than direct effects (Cardinale et al., 2012), they remain useful in predicting effects of ecosystem properties on ES supply (Díaz et al., 2007a; Kandziora et al., 2013). To incorporate indirect effects, studies from dryland rangelands often rely on key vegetation properties such as plant diversity (Gaitán et al., 2014) and/or relative abundances of plant functional types (PFTs; see Linstädter et al., 2014; Ruppert et al., 2015). Although the use of PFTs has widely been accepted in plant ecology, the challenge remains to select trait sets that capture plant responses to environmental drivers of interest (Fry et al., 2013). Due to convergent effects of grazing and climatic aridity on vegetation (Gaitán et al., 2017), this task is particularly challenging in dryland environments (Linstädter et al., 2014). Here, plant traits related to life history, growth form and plant height have been found to be responsive (Díaz et al., 2007b). In West Africa’s Sudanian savannas, a steep regional gradient of climatic aridity shapes vegetation patterns (White, 1983). ES studies along such gradients may allow extrapolating climate change effects via a space-for-time substitution, if spatial trends reflect projected temporal trends (Dunne et al., 2004). Moreover, steep local gradients of land-use intensity are observable (Ouédraogo et al., 2015). West Africa’s savannas are thus an ideal study area for improving our understanding of ES delivery from savanna rangelands under contemporary and future conditions. However, this opportunity has remained rather unharnessed due to logistical challenges. Taking a macroecological perspective, our study thus aims at (i) identifying predictors of important ESs (forage supply and erosion control) from West Africa’s savanna rangelands over a broad geographical scale, (ii) quantifying the relative importance of biophysical and land-use drivers for ES supply, with the ultimate goal to identify significant and potentially universal predictors. We use a conceptual model for drivers’ single and interactive effects on ES proxies (Fig. 1), based on current knowledge (e.g. Díaz et al., 2007a). We specifically hypothesise that climate is the major driver of ES supply over and above the effects of more local drivers such as topo-edaphic factors and landuse intensity.

vegetation structure, and ES delivery can be quantified by ecosystem properties that are commonly measured in ecological studies (Díaz et al., 2007a). However, the relative importance of these drivers for ES supply is still poorly understood, hampering the identification or design of appropriate land management strategies (van Oudenhoven et al., 2012). Unfortunately, it is not possible to manage ecosystems to simultaneously maximise all ESs; trade-offs exist (López-Ridaura et al., 2002; Polasky et al., 2008; Smith et al., 2012) and are particularly common between provisioning and regulating ESs (Raudsepp-Hearnea et al., 2010; Castro et al., 2014). In dryland rangelands, one of the most problematic trade-offs is that between forage supply, which is of immediate importance for livestock production, and erosion control, which if reduced has detrimental feedbacks on forage supply in the long-term due to rangeland degradation (Milton et al., 1994; Kimiti et al., 2016; Linstädter et al., 2016). Since environmental drivers may benefit immediate ES provision while negatively affecting ES provision in the long-term, trade-offs need to be detected and considered in management decisions. However, most studies on rangeland ESs do not address these trade-offs. Several challenges are pertinent in this context. First, the quantification of ES supply is challenging since many aspects cannot be directly measured (Burkhard et al., 2012). To overcome the methodological obstacles of a direct ES measurement, several sets of ES indicators (or ‘proxies’) have been proposed on local, regional and global scales (MEA, 2005; Kandziora et al., 2013; Albert et al., 2016). Forage quantity is seen as the main factor constraining livestock carrying capacity (Yahdjian and Sala, 2006); it is estimated as aboveground net primary production (ANPP) (Scurlock et al., 2002; Yahdjian and Sala, 2006; Swemmer et al., 2007; Ruppert and Linstädter, 2014) or as aboveground biomass (AGB) (Oomen et al., 2016). Forage quality is assessed with various indices, such as crude protein, in-vitro digestibility or a combination of both into metabolisable energy (ME) (Changwony et al., 2015). It is also desirable to integrate forage quantity and quality into a single proxy, such as metabolisable energy yield (MEY), which quantifies forage nutritive energy per area (Niemeläinen et al., 2001). Vegetation cover can serve as a proxy for erosion control (Kandziora et al., 2013). In dryland ecosystems, a high perennial plant cover (PPC) is particularly important to prevent accelerated wind and water erosion (Rietkerk et al., 2000; Munson et al., 2011). PPC is also a good indicator of an ecosystem’s capacity to capture and retain resources such as water and nutrients (Soliveres et al., 2014). However, forage services related to forage quality are rarely included in ES studies (e.g. Sawadogo et al., 2005; Palmer et al., 2010; Oomen et al., 2016) probably because measurements are labour- and cost-intensive. To overcome this challenge, suitable ES proxies are required for a fast and cost-efficient estimation of both quantitative and qualitative components of forage provision (Ferner et al., 2015). Such indirect methods are important since the high costs and logistical requirements for direct ES measurements usually constrain their assessment. These constraints are further exacerbated when studies involve large and remote study areas as in our case. Consequently, field studies on rangeland ESs are still often restricted to small study areas and/or tend to focus on ESs that are relatively easy to (directly) measure. The second challenge is to identify key ES drivers (Díaz et al., 2007a), and to quantify their relative importance for vegetation structure and hence ES supply. As stated above, the structure of grasslands and savanna rangelands is simultaneously influenced by multiple biophysical and land-use drivers. Among them, the frequency and intensity of grazing disturbances is of particular importance, as it determines the extent of biomass removal and destruction (Milchunas et al., 1989; Altesor et al., 2005). The simultaneous influence of multiple drivers makes it difficult to disentangle their independent effects, especially in regions that face high environmental variability (Archer, 1995). Consequently, empirical savanna studies assessing the effects of

2. Methods 2.1. Study area The study area (∼106 000 km2) reaches from Northern Ghana to Central Burkina Faso in the West African Sudanian savanna zone (Fig. 2). Climate is seasonal; in the southern Sudanian zone, it is humid to dry sub-humid, and in the northern Sudanian zone it is semi-arid (UNEP, 1997). Mean annual precipitation (MAP) and mean annual temperature (MAT) range from 1200 to 600 mm/a, and from 26 °C to 28 °C, respectively. The grass layer is co-dominated by annual and perennial grasses such as Brachiaria spp. and Andropogon spp., intermixed with annual forbs. The tree layer consists of species with high resprouting ability (Ouédraogo et al., 2015). Soils develop on acidic metamorphic rocks and have coarse texture (> 80% sand) with low water holding capacity (Callo-Concha et al., 2012). They are highly susceptible to erosion and compaction and, depending on the cultivation history, exhibit low levels of organic matter, nitrogen and phosphorus (Callo-Concha et al., 2012). Besides subsistence agriculture, grazing by domestic herbivores is the most widespread type of land-use in the area (Naah et al., 2017).

258

Agriculture, Ecosystems and Environment 251 (2018) 257–267

R.T. Guuroh et al.

Fig. 1. Flow chart describing the conceptual approach used in this study for showing effects of different sets of biophysical and land-use drivers on proxies of ecosystem services. Solid thick arrows indicate individual effects and single-line arrows indicate interactions.

Vegetation properties were sampled on three circular subplots (1 m2) per plot. To capture intra- and inter-seasonal ES dynamics, we sampled during two consecutive growth periods (June to October 2012 and 2013), and varied the time of sampling independently from other sources of variation. In total, we sampled forty-four sites and 300 plots.

2.2. Sampling design We stratified sampling along the climate gradient into three rainfall zones (Fig. 2). Sampling per zone aimed at capturing local gradients in topo-edaphic conditions and grazing pressure achieved by choosing sites that maximised gradient ranges. For this purpose, we sampled at least two sites in protected areas per rainfall zone. For topo-edaphic conditions, we used a geological map to select sites (with ≥3 km distance) in major geological units within each rainfall zone (see Ferner et al., 2015). Within-site sampling was stratified into slope position (upslope, midslope and lowland). We placed ≥3 plots (100 m2) per slope position and site in homogeneous vegetation (distance ≥30 m).

2.3. Data acquisition 2.3.1. Climatic variables For each site, data on MAP and MAT, mean temperature of the wettest quarter, minimum and maximum temperature of the coldest and warmest month respectively (TMin, TMax) were obtained from Fig. 2. Study area and location of 44 sampling sites. The study area covers the southern and northern Sudanian vegetation zones, following White (1983). Sampling is stratified into three zones of decreasing climatic aridity (hereafter called ‘rainfall zones’), as indicated by isohyets (low: mean annual precipitation 600–800 mm/a; intermediate: 800–1000 mm/a; high: 1000–1200 mm/a).

259

Agriculture, Ecosystems and Environment 251 (2018) 257–267

R.T. Guuroh et al.

phenological stages (Appendix B in Supplementary material). Established equations performed well in predicting AGBSpec (adjusted R2 = 0.74) and were used for AGBSpec estimation on all subplots. To capture the effects of biophysical and land-use drivers on ES delivery via their imprint on ecosystem properties, we calculated plant species richness on the plot level, and aggregated floristic composition into PFTs. Following recommendations of Díaz et al. (2007b) for regional-scale studies, we established PFTs based on life history (‘annual’ or ‘perennial’), growth form (‘forb’, ‘graminoid’, or ‘woody’), and plant vegetative height (‘small’: ≤50 cm, ‘tall’: > 50 cm; Appendix C in Supplementary material). Species’ life history and growth form was extracted from taxonomic literature (e.g. Schmidt et al., 2011). Trait combinations resulted in ten PFTs (Appendix E in Supplementary material). Relative abundances of PFTs were calculated based on AGBSpec. In analogy to community-aggregated functional traits (Vile et al., 2006), we obtained a community’s phenophase (Phen) by weighting species’ phenological stage according to their relative contribution to the biomass of the plant community.

WorldClim (averages over 1950–2000; http://www.worldclim.org/). We quantified climatic aridity based on the UNEP aridity index (AI) as 1-AI (UNEP, 1997; Delgado-Baquerizo et al., 2013). To capture effects of recent rainfall, we obtained season’s accumulated precipitation (SAP) until the month preceding field sampling from the GPCC database (Schneider et al., 2011). 2.3.2. Topo-edaphic variables Besides slope position, we recorded various soil variables on the plot level (Appendix A in Supplementary material). Although bare soil cover is an unspecific indicator that could represent edaphic and climatic aridity as well as grazing disturbances (Linstädter et al., 2014), we categorised it as a topo-edaphic variable, emphasizing that it is closely related to soil properties such as soil depth and water infiltration. Following FAO (2006), we estimated the cover of surface fragments. To quantify topsoil properties, a composite sample from five soil cores (0–4 cm depth) was collected. Samples were homogenised and air-dried (> 21 days); only fractions < 2 mm were analysed. Particle size distribution was determined by laser diffraction method, using a Laser Particle Size Analyser (Horiba LA–960). Soil acidity was determined in a 1:2.5 water suspension. Plant-available phosphorus was measured via calcium-acetate-lactate (CAL) extraction (mg kg−1), following standard protocols (VDLUFA, 2008). N and C content was analysed by dry combustion with a CN analyser (Vario EL cube). Analyses were performed at Geography Institute of the University of Bonn, Germany. Variables were categorised into those with ‘slow’ and ‘fast’ response to land-use intensity, based on findings from African dryland rangelands (Linstädter and Baumann, 2013; Sandhage-Hofmann et al., 2015).

2.3.5. Ecosystem service supply We established plot-level proxies for forage quantity (aboveground biomass, AGB; kg dry matter (DM) ha−1) and forage quality (metabolisable energy, ME; MJ kg−1 DM). We estimated AGB by summing AGBSpec data per subplot, and then averaging over subplots. As no actions were taken to prevent losses in biomass from herbivory before sampling, AGB may not serve as an estimate for ANPP (Ruppert and Linstädter, 2014). However, given that our sampling efforts have covered the full range of grazing pressure in Sudanian savannas, we are confident that it will serve as a reliable estimate of actual forage quantity within years of sampling. For ME estimation, we used a portable spectro-radiometer (FieldSpec 3Hi-Res, ASD Inc., Boulder, CO, USA) to measure plant reflectance on subplots. With the aid of a regression model calibrated in our study area (see Ferner et al., 2015), we estimated ME and averaged to plot-level (see Appendix D in Supplementary material for details). Due to difficult measurement conditions in West Africa’s Sudanian savannas (Ferner et al., 2015), spectral data were obtained for 1–3 subplots per plot and 1–9 plots per site. We calculated the product of AGB and ME as a combined proxy of forage provision (metabolisable energy yield, MEY; in GJ ha−1). For the regulating ES ‘erosion control’, we used the cover of perennial plants (PPC) as a proxy, as many studies conducted under varying environmental conditions reported a positive effect of vegetation cover on reducing soil erosion (e.g. Nunes et al., 2011; Kandziora et al., 2013; Mohammad and Adam, 2010). This includes West Africa’s savanna rangelands, where Rietkerk et al. (2000) found that the presence of perennial plants was particularly important for erosion control due to reductions in the run-off, and increases in the infiltration of rainwater. For this reason, PPC is an appropriate indicator for erosion control in our study area.

2.3.3. Land-use intensity variables In West Africa’s Sudanian savannas, an official protection status does not necessarily mean that an area is not grazed by domestic livestock (Ouédraogo et al., 2015). For this reason, we did not consider a site’s protection status as a predictor variable, but directly assessed land-use intensity on the plot level. We visually estimated surface characteristics with positive (+) or negative (−) relations to grazing intensity in African savannas (Linstädter et al., 2014), i.e. the cover of cattle and donkey dung (+), smallstock droppings (+), litter (−), and moribund material (−). Following Linstädter et al. (2014), we also recorded physical evidence of grazing pressure (GP), ranging from 0 (ungrazed) to 4 (heavily grazed). We treated ordinal data as quasi-numerical in further analyses. 2.3.4. Vegetation properties Within subplots, we assessed the cover and vegetative height of all vascular plant species. We focused on the herbaceous layer, but included seedlings and saplings (≤2 m) of woody species. Species’ phenological stage was recorded with a simplified BBCH scale; hence instead of using all ten development stages on the full BBCH scale (Hess et al., 1997), we reduced it to those relevant for our study, thus distinguishing germinating, sprouting, shooting, flowering, fruiting, and senescent. Species’ standing aboveground biomass (AGBSpec; in kg dry matter ha−1) was estimated via allometric equations. On 203 harvesting quadrats (1 m2) representing the full range of grazing pressure in Sudanian savannas, we recorded species’ vegetative height, ground cover and phenological stage. We then harvested plant biomass at stubble height (ca. 3 cm), and separated it into species, discarding moribund material. Samples were oven-dried (60 °C, 48 h) and weighed. For allometric models, we calculated species’ biovolume per quadrat as cover x height (Jauffret and Visser, 2003) and established linear regressions with species’ biomass per quadrat as response variable. Explanatory variables were – besides biovolume – species’ growth form and phenological stage to account for their modulating effects on aboveground biomass (Rigge et al., 2013). Model selection procedures rendered separate calibrations for two aggregate growth forms and four

2.4. Data analysis To assess the relative importance of drivers of ES supply, we first selected environmental variables as potential predictors, and then explored the relationship of selected potential predictors with ES proxies. 2.4.1. Selection of environmental variables We performed principal component analyses (PCAs) to select potential predictors of ES supply from six site-level variables (long-term climatic variables) and twenty-nine variables recorded on plot level. Separate PCAs were performed for five variable sets: (i) climate, (ii) topo-edaphic variables with slow response or (iii) fast response to landuse, (iv) land-use intensity, and (v) PFTs. We then identified variables highly loading (≥│0.8│) on principal components (PCs) with eigenvalues > 1 to reduce collinearity within predictor sets. In case of competing variables, we chose the variable with the highest proportion 260

Agriculture, Ecosystems and Environment 251 (2018) 257–267

R.T. Guuroh et al.

under the given model (Cohen, 1973). Moran’s I spatial correlograms were used to check for spatial autocorrelation in final models (Griffith, 2009). As a mean of validation and to estimate uncertainty in standard errors (SEs) of model parameters, we bootstrapped our final models (10 000 repetitions with replacement). We also calculated the relative bias in SE for model parameters by comparing the bootstrap estimates and our final LMM, following Thai et al. (2013):

of explained variance in single-variable models. We additionally chose site, grazing pressure (GP), slope position (SP), season’s accumulated precipitation (SAP), phenophase (Phen), species richness (SRic) and two interaction terms (GP x Phen, GP x SP). Given their ordinal or categorical nature, GP and SP could not be included in PCAs and were selected due to their prevalent importance (Augustine, 2003). SAP and Phen were selected to account for seasonal variation in precipitation and forage provision, respectively (Brüser et al., 2014). Interaction terms and SRic were selected based on expert knowledge, as we assumed important effects on ES supply. Multicollinearity of selected variables was checked using Pearson’s correlation and variance inflation factors.

RBias =

SEBoot − SELMM x100 SELMM

Where RBias is the relative bias; SEBoot = Bootstrap standard errors averaged over 10 000 runs SELMM = Final LMMs’ standard errors. Following Thai et al. (2013), we classified model predictors as unbiased (RBias < ± 5%); moderately biased ( ± 5–10%); and strongly biased (> ± 10%). Bootstrapping was performed with the bootpackage (in version 3.2.2) for R (Canty and Ripley, 2015). Statistical assumptions were explored visually as proposed by Zuur et al. (2010). To achieve normality of errors and homoscedasticity, we applied square-root transformation for PPC and logarithmic transformation for AGB and MEY. All analyses were conducted using the statistical software R in version 3.2.2 (R Core Team, 2015).

2.4.2. Exploring environmental variables’ relationship with ESs We used linear mixed-effect models (LMMs) to explore the effect of selected variables on ESs. Before modelling, all explanatory variables were standardised by scaling and centering. Initially, full LMMs – including all selected variables as fixed effects – were established for each of the four ESs; ‘site’ was included as random-intercept term (e.g. ES proxy ∼climate variables + land‐use variables + topo-edaphic variables + vegetation properties + interactions + (1|site)). Initial full models were subject to BIC-based model selection using restricted maximum likelihood estimation. LMMs were calculated using the lme4package (in version 3.2.2) for R (Bates et al., 2015). To estimate the variance explained by fixed and random effects in final models, we used the method proposed by Nakagawa and Schielzeth (2013) to obtain marginal and conditional R2 (MR2 and CR2, respectively). MR2 is the proportion of explained variance by fixed-effects, and CR2 is the proportion explained by fixed plus random effects (Ruppert et al., 2015). For each ES, variance explained by random effects was calculated as CR2 minus MR2. Final models were further explored using ANOVAs (Type III). We estimated the importance of each individual predictor by calculating classical eta-squared values as effect size metric. The eta-square value for a given predictor reflects the proportion of total explained variance in the dependent variable that is associated with this very predictor

3. Results 3.1. Selection of environmental variables From the five PCAs, we selected fourteen variables (nine environmental variables and five PFTs) for further analyses (Appendix E in Supplementary material). Soil sand content was dropped from the set of potential predictors due to its high correlation with soil N content (Appendix F in Supplementary material). In total, 21 terms (eighteen fixed-effects, two interactions, and one random effect) were selected as potential predictors of ES supply (Table 1).

Table 1 Selected variables for linear mixed-effect models. Fixed effects are grouped into five predictor sets. Two interaction terms and the random effect ‘study site’ are also considered. Effect type Fixed

Predictor set Climate

Slow responding topo-edaphic variables

Land-use

Fast responding topo-edaphic variables Vegetation properties

Interaction

Interactions

Random

Study site

Potential predictor a

Climatic aridity Min. temperature of coldest month Season’s accumulated precipitationb Slope positionc Soil acidity Bare soil cover Grazing pressured Litter cover Moribund material cover Soil content of plant-available phosphorus Soil nitrogen content Small annual forbs Tall annual forbs Small annual graminoids Small perennial graminoids Tall perennial graminoids Species richness Phenophase Grazing pressure × phenophase Grazing pressure × slope position Study site

Acronym (unit)

Mean

SD

Min.

Max.

CA (n.a.) TMin (°C) SAP (mm) SP (n.a) pH (n.a) BSC (%) GP (n.a) LC (%) MMC (%) P (mg kg −1) N (%) SAF (% AGB)e TAF (% AGB)e SAG (% AGB)e SPG (% AGB)e TPG (% AGB)e SRic (#) Phen (CWM)f GP*Phen GP*SP Site

0.464 21.71 516.4 – 5.40 19.24 – 3.00 0.502 14.20 0.090 0.078 0.115 0.049 0.047 0.283 17.86 2.55 – – –

0.093 0.431 174.04 – 0.517 9.98 – 3.95 1.53 17.10 0.054 0.082 0.094 0.085 0.103 0.240 5.53 0.616 – – –

0.307 20.81 161.4 1 4.00 5.00 0 0.00 0.00 1.00 0.031 0.00 0.00 0.00 0.00 0.00 6 2.00 – – 1

0.694 22.61 808.9 3 6.90 60.00 4 25.00 15.00 147 0.324 0.441 0.558 0.545 0.686 0.991 40 4.85 – – 44

n.a. = not applicable. a 1-AI, with AI = UNEP aridity index (mean annual precipitation/potential evapotranspiration; UNEP (1997)). b Antecedent rainfall of a rainy season until month preceding field sampling. c Categorical variable (1–3): lowland (1), midslope (2), upslope (3). d Ordinal scale (0–4): 0 (ungrazed), 1 (light GP), 2 (moderate GP), 3 (heavy GP), 4 (very heavy GP). e AGB = total aboveground biomass per plot. f Community-weighted mean of phenology; species’ phenological stage (0–5) weighted by their relative abundance (% AGB).

261

Agriculture, Ecosystems and Environment 251 (2018) 257–267

R.T. Guuroh et al.

Fig. 3. Variance explained by biophysical and land-use predictors in linear mixed-effect model. For each ecosystem service proxy, bars denote the proportion of variance explained by significant predictors, calculated as classical eta squared. Predictors are grouped into predictor sets (see Table 2). Unexplained variance is included as residuals. SAP = Season’s accumulated precipitation, GP = grazing pressure, Phen = phenophase; AGB = aboveground biomass, ME = metabolisable energy, MEY = metabolisable energy yield, PPC = perennial plant cover.

3.2. Performance of predictors across ES proxies

3.3. Relationships between predictor sets and ES proxies

The importance, bias and direction of predictor effects for the four ES proxies varied considerably (Fig. 3). Explained variance in the ME model was comparatively low (44%), but higher for other ES proxies (MEY and PPC = 67%; AGB = 77%). The importance of the random factor ‘site’ (calculated as CR2 minus MR2) also varied considerably across ES proxies. It was negligible for PPC (5% of variance), small for AGB (10%) and MEY (15%), but high for ME (26%). Following bootstrapping and the calculation of relative bias, we found that the ME model was of limited reliability, since it had a ratio of four biased to three unbiased predictors (Table 2 and Fig. 4).

3.3.1. Climate variables Among climate-related variables, SAP was the most important predictor of AGB and MEY, with high levels of explained variance (38% and 22%, respectively; Fig. 3). Climate variables were less important for ME and PPC. SAP had positive effects on AGB and MEY, while climatic aridity negatively affected AGB and PPC (Fig. 4). 3.3.2. Topo-edaphic variables Effects of topo-edaphic conditions were of secondary importance for all ES proxies: only bare soil cover was retained in final models. This predictor was an unbiased predictor of all ES proxies, with negative effects on AGB, MEY and PPC, but positive effects on ME (Fig. 4). However, it only explained a small portion of variance in ES proxies (< 5%; Fig. 3).

Table 2 Relative bias of standard errors (SEs) for all significant predictors of ecosystem services (AGB = aboveground biomass, ME = metabolisable energy, MEY = metabolisable energy yield, PPC = perennial plant cover). Levels of bias are: unbiased, with relative bias < ± 5% (given in brackets); moderately biased, with relative bias from ± 5% to ± 10%; and strongly biased, with relative bias > ± 10% (given in bold). Empty cells indicate that a potential predictor was not retained in the final model of the respective ecosystem service. Predictor set

Potential predictor

Relative bias of SEs (%) AGB

Climate variables

Topo-edaphic variables Land-use variables Vegetation properties

Interaction

3.3.3. Land-use variables Grazing, litter cover and moribund material cover were important predictors for forage services, but were not relevant for PPC (Fig. 3). Grazing pressure was the most important land-use predictor in all cases, explaining 12% variance in AGB, 14% in ME and 8% in MEY. It had negative effects on AGB and MEY but positive effects on ME.

ME

MEY

Climatic aridity Season’s accumulated precipitation Bare soil cover

−23 −23 (1)

(0)

(3)

Grazing pressure Litter cover Moribund material cover Small annual forbs Small perennial graminoids Tall perennial graminoids Species richness Phenophase Grazing pressure × phenophase

6 −5

−17 29 −5 −11 15 –

(2) (−3)

18 (4) 5 14 10

PPC −24

3.3.4. Vegetation properties Vegetation properties were important predictors for all ES proxies and had strong positive effects in all cases except phenophase which was negatively related to ME and PPC (Fig. 4). SRic, SAF, SPG, TPG, and Phen were important for forage services while only a subset (SPG, TPG, Phen) were important for PPC (Fig. 3). ME was mainly driven by phenophase (19%) and grazing pressure (14%), while PPC was mainly driven by TPG (52%). Vegetation properties contributed high levels of variance (greater than 20%) for all ES proxies.

−33

(−2)

6 8 (2) 21 8

(−4)

−10 −5 −21 –

3.3.5. Interactions Of the two interactions tested, only the interaction of grazing pressure with phenophase was important for AGB and MEY (Fig. 3,

262

Agriculture, Ecosystems and Environment 251 (2018) 257–267

R.T. Guuroh et al.

Fig. 4. Importance, direction and bias of predictor effects on ES supply. Arrows indicate that a predictor was retained in the final ES model, and visualize a predictor’s effect size class (after Cohen, 1988)1, direction of effect2 and bias3. AGB = aboveground biomass, ME = metabolisable energy, MEY = metabolisable energy yield, PPC = perennial plant cover. 1 Arrow width indicates predictors’ effect size (classical eta squared); = very small effect (< 0.1), = small effect (0.1–0.3) and = medium effect (0.3–0.5) 2 Arrow direction indicates relationship; upward/downward arrows = positive/negative relationship of predictors with response variable 3 Arrow colour indicates relative bias of predictors; green = unbiased (relative bias < ± 5%), yellow = moderately biased (relative bias from ± 5% − ± 10%) and red = strongly biased (relative bias > ± 10%).

ES supply (Kandziora et al., 2013); and that variables reflecting vegetation structure are of primary importance (Gaitán et al., 2014; Ruppert et al., 2015). In the following, we will first discuss our results from a methodological point of view, and then highlight the ecological context of our predictor sets’ performance.

Appendix G in Supplementary material) but explained < 1% of variance in both cases. 3.4. Relationships between vegetation properties and environmental predictors Plant functional types (SAF, SPG and TPG) were mainly driven by topo-edaphic and land-use variables with the exception of TPG which was additionally driven by climatic aridity (Table 3). SRic was only driven by season’s accumulated precipitation.

4.1. Methodological considerations 4.1.1. Feasibility of ES proxies In our study, we estimated ES supply via ES proxies based on calibrated models, as the size and remoteness of our study area hampered a direct ES quantification (Burkhard et al., 2012; Kandziora et al., 2013). A direct estimation of forage quantity would have implied to harvest aboveground biomass from 1350 sampling quadrats of 1 m2 along 106 000 km2 and to immediately transport them to a lab with a drying oven. Both the volume of the biomass samples (> 60 m3) and the limited availability of lab facilities made this approach logistically unfeasible. We assume that the poor research infrastructure is one of the main reasons why Africa in general and West Africa in particular is chronically under-represented in ES studies (see e.g. Seppelt et al., 2011; Maestre et al., 2012). Likewise, high costs and logistical challenges in

4. Discussion We assessed effects of various environmental variables and vegetation properties on vital ESs provided by African savannas. Our macroecological study gives valuable insights into the importance of climate, topo-edaphic conditions, land-use intensity and vegetation properties as ES drivers on a regional scale. We found that the relative importance of predictor sets differed considerably across ES proxies, and that vegetation properties always played an important role. This highlights that it is critical to consider multiple potential predictors of

Table 3 Summary of LMM results showing relationships between vegetation properties and climate, topo-edaphic and land-use variables. SAF = small annual forbs, SPG = small perennial graminoids, TPG = tall perennial graminoids, SRic = species richness, Phen = phenophase. n.s = non-significant (but retained in model), Numbers indicate model parameter coefficients. Predictor set

Potential predictor

Climate variables

Season’s accumulated precipitation Climatic aridity Min. temperature of coldest month Slope position Soil acidity Bare soil cover Soil content of plant-available phosphorus Soil nitrogen content Grazing pressure Litter cover Moribund material cover Grazing pressure × Phenophase

Topo-edaphic variables

Land-use variables

Interaction

SAF

SPG

TPG

−2.7 (**)

0.2 (***)

−0.2 (**)

0.2 (**)

0.2 (**) 0.3 (***) 0.2 (***)

2.3 (***)

* < 0.05, ** < 0.01, *** < 0.001.

263

−2.2 (**)

SRic

Phen

0.3 (**)

-0.02 (n.s)

Agriculture, Ecosystems and Environment 251 (2018) 257–267

R.T. Guuroh et al.

by a high climatic aridity (Zimmermann et al., 2008; Hou et al., 2013). Moreover, our result is consistent with the convergence model of aridity and grazing stating that aridity and grazing are convergent selective forces (Quiroga et al., 2010) which act together to result in a higher environmental harshness. Similarly, a lower relative abundance of perennial plants has been observed in South Africa’s grasslands and savannas along a gradient of additive forces of aridity and grazing pressure (Fynn and O’Connor, 2000; Linstädter et al.,2014). In our study, tall perennial graminoids only dominated near-natural vegetation in the two rainfall zones with intermediate and high MAP (results not shown). Our results underline that non-protected areas in the northern Sudanian savannas are particularly prone to soil erosion (see Rietkerk et al., 2000), which may lead to losses of other soil-mediated ESs such as carbon storage (Orwin et al., 2015).

estimating ME via laboratory-based analyses may explain why this important aspect of forage provision is often neglected in rangeland studies globally and particularly from Africa (e.g. Sawadogo et al., 2005; Palmer et al., 2010; Oomen et al., 2016). Similarly, although several methods have been proposed for soil erosion assessment (Hsieh et al., 2009), a direct measurement of soil erosion under field conditions, especially at a landscape scale, remains a major challenge (Keim and Schoenholtz, 1999; Nearing et al., 2000; Trimble and Crosson, 2000). Soil erosion measurements are usually hard to carry out, are time consuming and expensive (Benyamini, 2004) so that the large study area in our case coupled with the logistical constraints made the use of direct measurement unfeasible. Please note that even an indirect estimation of forage quality is highly challenging from a methodological point of view (Ferner et al., 2015), which might explain why it is rarely included in ES indicator sets (MEA, 2005; Kandziora et al., 2013). To circumvent these methodological challenges, our study relied on robust calibration models, both for forage quantity and quality, which were specifically established for our study area as a preparatory work for the present study (Ferner et al., 2015). We are thus confident that our indirect estimations of forage supply are reliable. With respect to erosion control, we used the cover of perennial plants as a proxy for the prevention of erosion (Kandziora et al., 2013), but did not consider other environmental constraints of an accelerated soil erosion such as surface slope or topsoil texture (Tamene and Le, 2015). Hence, soil erosion vulnerability was not in our focus.

4.2.2. Topo-edaphic variables Among the variables within this predictor set, five were selected for further analyses (Table 1), and only bare soil cover was retained in final models. Although it was an unbiased predictor of all ESs, it always had very small effects (Fig. 4). Bare soil is an unspecific indicator, reflecting not only edaphic aridity but also other aspects of environmental harshness such as climatic aridity and disturbances (Augustine, 2003; Linstädter et al., 2014). It has also been described as an indicator of low ecosystem integrity (Kandziora et al., 2013). The very small effects of topo-edaphic variables indicate that in West Africa’s Sudanian savannas they are of minor importance for ES supply from herbaceous vegetation. To ensure that this was not due to a masking effect of other potential predictors (e.g. climate and land-use), we checked for multicollinearity among them (cf. Appendix F in Supplementary material) before performing modelling. Our results are somewhat surprising, as other regional studies from semi-arid savannas in eastern and southern Africa found topo-edaphic conditions to be a major source of spatial variation in herbaceous vegetation, e.g. in floristic composition and/or ANPP (Augustine, 2003; Linstädter et al., 2014; Viljoen et al., 2014). However, variation in vegetation structure (and thus ES supply) along environmental gradients also depends on gradient length (Shipley, 2010). We assume that the broad range of land-use intensity captured in our study area has masked the comparatively small variation in topoedaphic conditions. As gradient length reflects the importance of ecological effects (Shipley, 2010), standardization of predictors does not necessarily normalise these effects. Our considerations are in congruence with a global study on environmental constraints of savannas, which found that both soil fertility and topographic complexity were of local and divergent importance (Lehmann et al., 2011).

4.1.2. Performance of predictor variables across ES proxies Standard errors of model predictors were mostly unbiased, indicating that our sampling efforts for AGB, MEY, and PPC were sufficient. An exception was climate; here, variables were always biased. This might be due to the fact that they were only available at site (and not plot) level. The fact that predictors explained a high proportion of variance in AGB, MEY and PPC (67–77%) and that site effects were comparatively small (5–15%) suggests that relevant predictor sets were used. However, our findings for ME are less robust. The seven predictors retained in the final model explained only 44% of variance in ME; four of them were also highly biased. This finding is supported by the relative high amount of variance explained by the random factor ‘site’ (26%). Hence, important drivers of ME were not taken into account. We assume that large differences in plant species’ forage quality – not captured by our PFT classification – may have driven ME differences on the plot level. More generally, our results underline that it is still a major challenge to identify key biophysical and land-use drivers for spatio-temporal patterns in ES supply (Díaz et al., 2007a; Kandziora et al., 2013). 4.2. Relationships between predictor sets and ES supply

4.2.3. Land-use variables Grazing, litter cover and moribund material cover were significant predictors for forage provision, but only had small to very small effects. In agreement with earlier findings (e.g. Schönbach et al., 2012; Hempson et al., 2015), grazing had strong negative effects on AGB and MEY, but positive effects on ME. For AGB, our results are supported by earlier findings from dryland rangelands worldwide (Milchunas and Lauenroth, 1993), including savannas and grasslands in West Africa (Savadogo et al., 2007a) and southern Africa (Fynn and O’Connor, 2000; Moreno García et al., 2014; Oomen et al., 2016). The positive effect of grazing on ME is probably due to modulatory effects of grazing on plant phenology. Grazing typically delays plant phenology (Han et al., 2015); as advanced phenological stages have lower nutritive values (Changwony et al., 2015), grazing indirectly increases forage quality (Hempson et al., 2015). More generally, our results underline that grazing has great effects on plant community structure and functioning, and subsequently on forage provision (McNaughton, 1983; Sala et al., 1986). This emphasises that it is critical for ES studies to (also) consider management effects on ecosystem functioning (de Bello et al., 2010; Sala et al., 2017).

4.2.1. Climate variables Contrary to our hypothesis, the long-term climate regime of a given site was of minor importance for forage provision as compared to landuse. Forage quality (ME) was not predicted by any climate variable, while AGB and MEY were more driven by antecedent rainfall than by climatic aridity. Our results however corroborate earlier findings from savannas and grasslands worldwide that long-term effects of climatic aridity on forage production may be largely overridden by inter-seasonal fluctuations in rainfall, particularly in the case of a good rain year (Oesterheld et al., 1999; Fynn and O’Connor, 2000; Blanco, 2008; Ruppert et al., 2012). As we designed our study to (also) capture intraseasonal variation in ES and their drivers, it is not surprising that a season’s accumulated precipitation played an important role for forage production. A similar importance of antecedent rainfall has been reported for Southern African grasslands and savannas (Swemmer et al., 2007; Brüser et al., 2014). The strong negative relationship between PPC and climatic aridity also corroborates earlier findings that perennial plants are filtered out 264

Agriculture, Ecosystems and Environment 251 (2018) 257–267

R.T. Guuroh et al.

5. Conclusions

4.2.4. Vegetation properties Species richness (SRic), phenophase (Phen), tall perennial graminoids (TPG), small perennial graminoids (SPG), and small annual forbs (SAF) were important predictors for forage services, but only Phen, SPG and TPG could predict PPC. In agreement with earlier findings, especially from experimental sites (e.g. Marquard et al., 2009), there was a significant positive relationship between SRic and AGB. Positive effects of species diversity on productivity (estimated here as AGB) can be due to a number of mechanisms (see Craven et al., 2016 for a recent review), such as species complementarity or facilitation, or the presence of key species or functional groups that have a disproportionately positive effect on community performance. It is noteworthy that the relationship between SRic and productivity is contentious and might be site-specific (Tredennick et al., 2015). Our findings highlight the critical role of SRic for maintaining functioning in rangelands, and provide additional evidence of its role in providing key ESs (Cardinale et al., 2012). As expected, we found a stronger positive relationship between AGB and tall perennial graminoids (11.7% of explained variance) than between AGB and small perennial graminoids (0.5% of explained variance). In contrast to forage quantity (AGB), we only found weak effects of PFTs on forage quality; the positive effects of two ‘small’ PFTs (SPG and SAF) were very small and strongly biased (Fig. 4). Apparently our PFT approach was successful in aggregating species with similar effects on forage quantity but not on forage quality. This is somewhat surprising, as we explicitly distinguished between ‘tall’ and ‘small’ forbs and graminoids to account for the typically higher forage quality of ‘small’ (low-stature) grasses and forbs, as found in African grazing lawns (Hempson et al., 2015). In agreement with earlier findings (Schönbach et al., 2012; Changwony et al., 2015), ME was negatively related to phenophase (Fig. 4), which is mainly due to a reduction in leaf-to-stem ratio at advanced phenological stages (Ball et al., 2001). Our result of a positive relationship between PPC and the relative abundance of perennial graminoids is expected; it implies that ecosystems dominated by perennial grasses (e.g. in protected areas) are comparatively little affected by accelerated erosion (Rietkerk et al., 2000), but should also have a good capacity to capture and retain water and nutrients (Soliveres et al., 2014). More generally, it underlines that management efforts aiming at erosion control in African grasslands and savannas should focus on the retention of perennial grasses. In support of this recommendation, a modelling study from the southwestern United States found that declines in perennial vegetation cover resulted in exponential increases in wind erosion (Munson et al., 2011). It is usually difficult or sometimes even impossible to manage ecosystems to simultaneously maximise all ES so trade-offs commonly occur (LópezRidaura et al., 2002; Polasky et al., 2008; Smith et al., 2012). For instance, it is often tricky to find the right balance between managing grasslands for forage supply and for erosion control. However, this situation can be minimised if management is carefully planned to reduce trade-offs and enhance synergies. There are examples from dryland rangelands showing how the cover of palatable (and often tall) perennial grasses could be maintained even under grazed conditions (Menke, 1992; Kemp and Culvenor, 1994), with the idea to balance the trade-off between the immediate ES of forage provision and more longterm benefits, in particular the ES of erosion control. Strategies for achieving this goal include rotational grazing and carefully adjusted resting schemes (Fynn and O’Connor, 2000; Müller et al., 2015). Finally, synergies and trade-offs of forage provision with other rangeland ESs such as biodiversity or carbon sequestration are pertinent (Sala et al., 2017), but were beyond the scope of our study. This would be an important task for future studies to shed new light on management options in the face of synergistic and antagonistic interactions among drivers of rangeland ESs.

Our study adopted a macroecological approach, including a large number of study sites and covering a large area with variable environmental conditions. In this way, we aimed at determining the relative importance of drivers of ES supply in African savanna rangelands. Our findings are useful for rangeland management and conservation within the context of ongoing climate change. Studies along steep climatic gradients may enhance our understanding of climate change effects on ES supply via a space-for-time substitution (Dunne et al., 2004). Our results indicate that climate change will indeed have an impact on the sustainability of ES supply from the region, both directly and supposedly also indirectly via its effects on vegetation properties. The higher importance of antecedent rainfall compared to climatic aridity corroborates earlier findings that antecedent rainfall, if good in a given year, could override the effects of long-term climatic conditions (Ruppert et al., 2012). Considering the importance of grazing pressure as a driver of forage supply and erosion control, we deduce that appropriate land management strategies (such as an adaptive regulation of stocking densities on a local and regional scale) can potentially mitigate negative effects of climate change on ES supply. Moreover, our result that vegetation properties − in particular the dominance of some functional groups – were generally of high importance for ecosystem service provision suggest that land managers can achieve considerable success by conserving or re-introducing key PFTs such as native perennial grasses which can enhance ES synergies by playing a dual role of contributing both to immediate (forage provision) and long-term ES supply such as erosion control (Kandziora et al., 2013; Zimmermann et al., 2015). Such a re-introduction is also a common restoration practice in many dryland rangelands worldwide (van den Berg and Kellner, 2005; Guerrant and Kaye, 2007; van Oudtshoorn et al., 2011). For this purpose, different approaches may be used, e.g. passive methods such as resting during critical time windows (e.g. post-drought years and high-rainfall years) and/or active restoration methods such as reseeding (Milton and Dean, 1995; Holmgren and Scheffer, 2001; van Oudtshoorn et al., 2011). Despite the fact that we could not directly measure ESs, we are confident that our reliance on ES proxies provides a good representation since robust calibration models were specifically developed for this study. Our approach provides important insights on the predictors of vegetation-mediated ESs and their relative importance for driving ES supply in African savannas.

Acknowledgements This study was mainly funded by the German Federal Ministry of Education and Research (BMBF; http://www.bmbf.de/en/) through WASCAL (grant 01LG1202-A). We thank the Catholic Academic Exchange Service (KAAD) for providing a scholarship for the first author. AL and JCR also acknowledge funding by BMBF via the ‘Limpopo Living Landscapes’ project (grant 01LL1304-D). We are grateful to the staff at the University of Ouagadougou and Senckenberg Museum in Frankfurt (particularly Oumarou Ouédraogo and Marco Schmidt) for assisting in plant identification. Statistical advice by Klara Dolos and Guido Lüchters is greatly appreciated. We thank John Baptist Naah for fruitful discussions.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.agee.2017.09.017.

265

Agriculture, Ecosystems and Environment 251 (2018) 257–267

R.T. Guuroh et al.

common factors, questions and approaches. Aust. J. Bot. 55, 362–370. Han, J., Chen, J., Xia, J., Li, L., 2015. Grazing and watering alter plant phenological processes in a desert steppe community. Plant Ecol. 216, 599–613. Hempson, G.P., Archibald, S., Bond, W.J., Ellis, R.P., Grant, C.C., Kruger, F.J., et al., 2015. Ecology of grazing lawns in Africa. Biol. Rev. 90, 979–994. Hess, M., Barralis, G., Bleiholder, H., Buhr, L., Eggers, T., Hack, H., Stauss, R., 1997. Use of the extended BBCH scale – general for the descriptions of the growth stages of mono- and dicotyledonous weed species. Weed Res. 37, 433–441. Holmgren, M., Scheffer, M., 2001. El Niño as a window of opportunity for the restoration of degraded arid ecosystems. Ecosystems 4, 151–159. Hou, Y., Zhou, G., Xu, Z., Liu, T., Zhang, X., 2013. Interactive effects of warming and increased precipitation on community structure and composition in an annual forb dominated desert steppe. PLoS One 8. Hsieh, Y.P., Grant, K.T., Bugna, G.C., 2009. A field method for soil erosion measurements in agricultural and natural lands. J. Soil Water Conserv. 64, 374–382. Jauffret, S., Visser, M., 2003. Assigning life-history traits to plant species to better qualify arid land degradation in Presaharian Tunisia. J. Arid Environ. 55, 1–28. Kandziora, M., Burkhard, B., Müller, F., 2013. Interactions of ecosystem properties, ecosystem integrity and ecosystem service indicators—a theoretical matrix exercise. Ecol. Indic. 28, 54–78. Keim, R.E., Schoenholtz, S.H., 1999. Functions and effectiveness of silvicultural streamside management zones in loessial bluff forests. For. Ecol. Manage. 118, 197–209. Kemp, D.R., Culvenor, R.A., 1994. Improving the grazing and drought tolerance of temperate perennial grasses. N. Z. J. Agric. Res. 37, 365–378. Kimiti, D.W., Riginos, C., Belnap, J., 2016. Low-cost grass restoration using erosion barriers in a degraded African rangeland. Restor. Ecol. 25 (3), 376–384. http://dx. doi.org/10.1111/rec.12426. López-Ridaura, S., Masera, O., Astier, M., 2002. Evaluating the sustainability of complex socio-environmental systems: the MESMIS framework. Ecol. Indic. 2, 135–148. Lehmann, C.E.R., Archibald, S.A., Hoffmann, W.A., Bond, W.J., 2011. Deciphering the distribution of the savanna biome. New Phytol. 191, 197–209. Linstädter, A., Baumann, G., 2013. Abiotic and biotic recovery pathways of arid rangelands: lessons from the High Atlas Mountains, Morocco. CATENA 103, 3–15. Linstädter, A., Schellberg, J., Bruser, K., Garcia, C.A.M., Oomen, R.J., du Preez, C.C., et al., 2014. Are there consistent grazing indicators in drylands? Testing plant functional types of various complexity in South Africa’s grassland and savanna biomes. PLoS One 9, e10467. Linstädter, A., Kuhn, A., Naumann, C., Rasch, S., Sandhage-Hofmann, A., Amelung, W., et al., 2016. Assessing the resilience of a real-world social-ecological system: lessons from a multi-disciplinary evaluation of a South African pastoral system. Ecol. Soc. 21, 35. Müller, B., Schulze, J., Kreuer, D., Linstädter, A., Frank, K., 2015. How to avoid unsustainable side effects of managing climate risk in drylands –the supplementary feeding controversy. Agric. Syst. 139, 153–165. MEA, 2005. Ecosystems and Human Well-Being: Synthesis. Island Press, Washington, DC, USA. Maestre, F.T., Quero, J.L., Gotelli, N.J., Escudero, A., Ochoa, V., Delgado-Baquerizo, M., et al., 2012. Plant species richness and ecosystem multifunctionality in global drylands. Science 335, 214–218. Marquard, E., Weigelt, A., Temperton, V.M., Roscher, C., Schumacher, J., Buchmann, N., et al., 2009. Plant species richness and functional composition drive overyielding in a six-year grassland experiment. Ecology 90, 3290–3302. McNaughton, S.J., 1983. Segrenti grassland ecology: the role of composite environmental factors and contingency in community organization. Ecol. Monogr. 53, 291–320. Menke, J.W., 1992. Grazing and fire management for native perennial grass restoration in California grasslands. Fremontia 20, 22–25. Milchunas, D.G., Lauenroth, W.K., 1993. Quantitative effects of grazing on vegetation and soils over a global range of environments. Ecol. Monogr. 63, 327–366. Milchunas, D.G., Lauenroth, W.K., Chapman, P.L., Kazempour, M.K., 1989. Effects of grazing, topography, and precipitation on the structure of a semiarid grassland. Plant Ecol. 80, 11–23. Milton, S.J., Dean, W.R.J., 1995. South Africa’s arid and semiarid rangelands: why are they changing and can they be restored? Environ. Monit. Assess. 37, 245–264. Milton, S.J., Dean, W.R.J., du Plessis, M.A., Siegfried, W.R., 1994. A conceptual model of arid rangeland degradation. Bioscience 44, 70–76. Mohammad, A.G., Adam, M.A., 2010. The impact of vegetative cover type on runoff and soil erosion under different land uses. Catena 81, 97–103. Moreno García, C.A., Schellberg, J., Ewert, F., Brüser, K., Canales-Prati, P., Linstädter, A., et al., 2014. Response of community-aggregated plant functional traits along grazing gradients: insights from African semi-arid grasslands. Rangeland Ecol. Manage. 17, 470–481. Munson, S.M., Belnap, J., Okin, G.S., 2011. Responses of wind erosion to climate-induced vegetation changes on the Colorado Plateau. Proc. Natl. Acad. Sci. 108, 3854–3859. Naah, J.-B.S.N., Guuroh, R.T., Linstädter, A., 2017. Factors influencing local ecological knowledge of forage resources: ethnobotanical evidence from West Africa’s savannas. J. Environ. Manage. 188, 297–307. Nacoulma, B., Schumann, K., Traoré, S., Bernhardt-Römermann, M.R., Hahn, K., Wittig, R., Thiombiano, A., 2011. Impacts of land-use on West African savanna vegetation: a comparison between protected and communal area in Burkina Faso. Biodivers. Conserv. 20, 3341–3362. Nakagawa, S., Schielzeth, H., 2013. A general and simplemethod for obtaining R^2 from generalized linearmixed-effects models. Methods Ecol. Evol. 4, 133–142. Nearing, M.A., Romkens, M.J.M., Norton, L.D., Stott, D.E., Rhoton, F.E., Laflen, J.M., et al., 2000. Measurements and models of soil loss rates. Science 290, 1300–1301. Niemeläinen, O., Kontturi, M., Jauhiainen, L., Nissinen, O., 2001. Estimated metabolisable energy yield of perennial and annual grass swards compared with those of

References Albert, C., Bonn, A., Burkhard, B., Daube, S., Dietrich, K., Engels, B., Frommer, J., Götzl, M., Grêt-Regamey, A., Job-Hoben, B., Koellner, T., Marzelli, S., Moning, C., Müller, F., Rabe, S.-E., Ring, I., Schwaiger, E., Schweppe-Kraft, B., Wüstemann, H., 2016. Towards a national set of ecosystem service indicators: insights from Germany. Ecol. Indic. 61, 38–48. Altesor, A., Oesterheld, M., Leoni, E., Lezama, F., Rodríguez, C., 2005. Effect of grazing on community structure and productivity of a Uruguayan grassland. Plant Ecol. 179, 83–91. Archer, S., 1995. Tree-grass dynamics in a Prosopis-thornscrub savanna parkland: reconstructing the past and predicting the future. Ecoscience 2, 83–99. Augustine, D.J., 2003. Spatial heterogeneity in the herbaceous layer of a semi-arid savanna ecosystem. Plant Ecol. 167, 319–332. Ball, D.M., Collins, G.D., Lacefield, N.P., Martin, D.A., Mertens, K.E., Olson, D.H., Putnam, D.J., Undersander, M.W., 2001. Understanding Forage Quality. American Farm Bureau Federation Publication, 1-01 Park Ridge, Illinois. Bates, D., Maechler, M., Bolker, B., Walker, S., 2015. Fitting Linear Mixed-Effects Models Using lme4. J. Stat. Software 67 (1), 1–48. http://dx.doi.org/10.18637/jss.v067.i01. Benyamini, Y., 2004. Measuring and monitoring soil erosion for soil conservation and soil protection in Israel. Soil Conservation and Protection for Europe. Blanco, L.J., 2008. Grazing effect on NDVI across an aridity gradient in Argentina. J. Arid Environ. 72, 764–776. Brüser, K., Feilhauer, H., Linstädter, A., Schellberg, J., Oomen, R.J., Ruppert, J.C., Ewert, F.A., 2014. Discrimination and characterization of management systems in semi-arid rangelands of South Africa using RapidEye time series. Int. J. Rem. Sens. 35, 1653–1673. Burkhard, B., Kroll, F., Nedkov, S., Müller, F., 2012. Mapping ecosystem service supply, demand and budgets. Ecol. Indic. 21, 17–29. Callo-Concha, D., Gaiser, T., Ewert, F., 2012. Farming and cropping systems in the West African Sudanian Savanna, WASCAL research area: Northern Ghana, Southwest Burkina Faso and Northern Benin. ZEF Working paper series, 100. Canty, A., Ripley, B., 2015. Boot: Bootstrap R (S-Plus) Functions. R Package Version 1. pp. 3–17. Cardinale, B.J., Duffy, J.E., Gonzalez, A., Hooper, D.U., Perrings, C., Venail, P., et al., 2012. Biodiversity loss and its impact on humanity. Nature 486, 59–67. Castro, A.J., Verburg, P.H., Martín-López, B., Garcia-Llorente, M., Cabello, J., Vaughn, C.C., López, E., 2014. Ecosystem service trade-offs from supply to social demand: a landscape-scale spatial analysis. Landscape Urban Plann. 132, 102–110. Changwony, K., Alvarez, M., Lanyasunya, T.P., Dold, C., Becker, M., Südekum, K.H., 2015. Biomass and quality changes of forages along land use and soil type gradients in the riparian zone of Lake Naivasha, Kenya. Ecol. Indic. 49, 169–177. Cohen, J., 1973. Eta-squared and partial eta-squared in fixed factor anova designs. Educ. Psychol. Meas. 33, 107–112. Cohen, J., 1988. Statistical power analysis for the behavioral sciences, Second ed. Lawrence Erlbaum Associates, Hillsdale, New Jersey. Craven, D., Isbell, F., Manning, P., Connolly, J., Bruelheide, H., Ebeling, A., et al., 2016. Plant diversity effects on grassland productivity are robust to both nutrient enrichment and drought. Phil. Trans. R. Soc. B : Biol. Sci. 371. Díaz, S., Lavorel, S., de Bello, F., Quétier, F., Grigulis, K., Robson, T.M., 2007a. Incorporating plant functional diversity effects in ecosystem service assessments. Proc. Natl. Acad. Sci. 104, 20684–20689. Díaz, S., Lavorel, S., McIntyre, S., Falczuk, V., Casanoves, F., Milchunas, D.G., et al., 2007b. Plant trait responses to grazing – a global synthesis. Global Change Biol. 13, 313–341. de Bello, F., Lavorel, S., Díaz, S., Harrington, R., Cornelissen, J., Bardgett, R., et al., 2010. Towards an assessment of multiple ecosystem processes and services via functional traits. Biodivers. Conserv. 19, 2873–2893. Delgado-Baquerizo, M., Maestre, F.T., Gallardo, A., Bowker, M.A., Wallenstein, M.D., Quero, J.L., et al., 2013. Decoupling of soil nutrient cycles as a function of aridity in global drylands. Nature 502, 672–676. Dunne, J.A., Saleska, S.R., Fischer, M.L., Harte, J., 2004. Integrating experimental and gradient methods in ecological climate change research. Ecology 85, 904–916. FAO, 2006. Guidelines for Soil Description. Food and Agricultural Organization of the United Nations, Rome, Italy. Ferner, J., Linstädter, A., Südekum, K.-H., Schmidtlein, S., 2015. Spectral indicators of forage quality in West Africa’s tropical savannas. Int. J. Appl. Earth Obs. Geoinf. 41, 99–106. Fry, E.L., Manning, P., Allen, D.G.P., Hurst, A., Everwand, G., Rimmler, M., Power, S.A., 2013. Plant functional group composition modifies the effects of precipitation change on grassland ecosystem function. PLoS One 8. Fynn, R.W.S., O’Connor, T.G., 2000. Effect of stocking rate and rainfall on rangeland dynamics and cattle performance in a semi-arid savanna, South Africa. J. Appl. Ecol. 37, 491–507. Gaitán, J.J., Oliva, G.E., Bran, D.E., Maestre, F.T., Aguiar, M.R., Jobbágy, E.G., et al., 2014. Vegetation structure is as important as climate for explaining ecosystem function across Patagonian rangelands. J. Ecol. 102, 1419–1428. Gaitán, J.J., Bran, D.E., Oliva, G.E., Aguiar, M.R., Buono, G.G., Ferrante, D., Nakamatsu, V., Ciari, G., Salomone, J.M., Massara, V., Martínez, G.G., Maestre, F.T., 2017. Aridity and overgrazing have convergent effects on ecosystem structure and functioning in Patagonian rangelands. Land Degrad. Dev. http://dx.doi.org/10.1002/ldr.2694. Griffith, D., 2009. Spatial autocorrelation in spatial interaction: complexity-to-simplicity in journey-to-work flows. In: Reggiani, A., Nijkamp, P. (Eds.), Complexity and Spatial Networks: In Search of Simplicity. Springer, Berlin. Guerrant Jr., E.O., Kaye, T.N., 2007. Reintroduction of rare and endangered plants:

266

Agriculture, Ecosystems and Environment 251 (2018) 257–267

R.T. Guuroh et al.

performance of sheep in continental steppe. Grass Forage Sci. 67, 535–545. http://dx. doi.org/10.1111/j.1365-2494.2012.00874.x. Schmidt, M., Thiombiano, A., Zizka, A., König, K., Brunken, U., Zizka, G., 2011. Patterns of plant functional traits in the biogeography of West African grasses (Poaceae). Afr. J. Ecol. 49 (4), 490–500. http://dx.doi.org/10.1111/j.1365-2028.2011.01283.x. Schneider, U., Becker, A., Finger, P., Meyer-Christoffer, A., Rudolf, B., Ziese, M., 2011. GPCC Full Data Reanalysis Version 6.0 at 0.5°: Monthly Land-Surface Precipitation from Rain-Gauges Built on GTS-based and Historic Data. Scurlock, J.M.O., Johnson, K., Olson, R.J., 2002. Estimating net primary productivity from grassland biomass dynamics measurements. Global Change Biol. 8, 736–753. Seppelt, R., Dormann, C.F., Eppink, F.V., Lautenbach, S., Schmidt, S., 2011. A quantitative review of ecosystem service studies: approaches, shortcomings and the road ahead. J. Appl. Ecol. 48, 630–636. Shipley, B., 2010. From plant traits to vegetation structure. Chance and Selection in the Assembly of Ecological Communities. Cambridge University Press, Cambridge, UK. Smith, M.D., van Wilgen, B.W., Burns, C.E., Govender, N., Potgieter, A.L.F., Andelman, S., et al., 2012. Long-term effects of fire frequency and season on herbaceous vegetation in savannas of the Kruger National Park, South Africa. J. Plant Ecol. 6 (1), 71–83. http://dx.doi.org/10.1093/jpe/rts014. Soliveres, S., Maestre, F.T., Eldridge, D.J., Delgado-Baquerizo, M., Quero, J.L., Bowker, M.A., Gallardo, A., 2014. Plant diversity and ecosystem multifunctionality peak at intermediate levels of woody cover in global drylands. Global Ecol. Biogeogr. 23, 1408–1416. Swemmer, A.M., Knapp, A.K., Snyman, H.A., 2007. Intra-seasonal precipitation patterns and above-ground productivity in three perennial grasslands. J. Ecol. 95, 780–788. Tamene, L., Le, Q.B., 2015. Estimating soil erosion in sub-Saharan Africa based on landscape similarity mapping and using the revised universal soil loss equation (RUSLE). Nutr. Cycl. Agroecosyst. 102, 17–31. Thai, H., Mentre, F., Holford, N., Veyrat-Follet, C., Comets, E., 2013. A comparison of bootstrap approaches for estimating uncertainty of parameters in linear mixed-effects models. Pharm. Stat. 12, 129–140. Tredennick, A.T., Adler, P.B., Grace, J.B., Harpole, W.S., Borer, E.T., Seabloom, E.W., et al., 2015. Comment on Worldwide evidence of a unimodal relationship between productivity and plant species richness. Science 351. Trimble, S.W., Crosson, P., 2000. U.S. soil erosion rates—myth and reality. Science 289, 248–250. UNEP, 1997. World Atlas of Desertification, 2nd ed. United Nations Environment Programme, London. van Oudenhoven, A.P.E., Petz, K., Alkemade, R., Hein, L., de Groot, R.S., 2012. Framework for systematic indicator selection to assess effects of land management on ecosystem services. Ecol. Indic. 21, 110–122. van Oudtshoorn, F., Brown, L., Kellner, K., 2011. The effect of reseeding methods on secondary succession during cropland restoration in the Highveld region of South Africa. Afr. J. Range Forage Sci. 28, 1–8. van den Berg, L., Kellner, K., 2005. Restoring degraded patches in a semi-arid rangeland of South Africa. J. Arid Environ. 61, 497–511. VDLUFA, 2008. Datenblatt A 6.2.1.1: Bestimmung von Phosphor und Kalium im CalciumAcetat-Lactat-Auszug VDLUFA Methodenbuch. In: Vdlufa (Ed.), Band 1. Die Untersuchung von Böden. VDLUFA Verlag, Darmstadt, Germany. Vile, D., Shipley, B., Garnier, E., 2006. Ecosystem productivity can be predicted from potential relative growth rate and species abundance. Ecol. Lett. 9, 1061–1067. Viljoen, F., Bullock, K., Panagos, M., Myburgh, W., 2014. Topographical units and soil types prove more efficient for vegetation sample site placement than land type units in semi-arid savanna, North West province, South Africa. Afr. J. Range Forage Sci. 31, 49–54. White, F., 1983. The Vegetation of Africa. A Descriptive Memoir to Accompany the UNESCO/AEFTAT/UNSO Vegetation Map. UNESCO/AETFAT/UNSO, Paris. Yahdjian, L., Sala, O.E., 2006. Vegetation structure constrains primary production response to water availability in the Patagonian steppe. Ecology 87, 952–962. Zimmermann, J., Higgins, S.I., Grimm, V., Hoffmann, J., Münkemüller, T., Linstädter, A., 2008. Recruitment filters in a perennial grassland: the interactive roles of fire, competitors, moisture and seed availability. J. Ecol. 96, 1033–1044. Zimmermann, J., Higgins, S.I., Grimm, V., Hoffmann, J., Linstädter, A., 2015. Local density effects on individual production are dynamic: insights from natural stands of a perennial savanna grass. Oecologia 178, 1125–1135. Zuur, A.F., Ieno, E.N., Elphick, C.S., 2010. A protocol for data exploration to avoid common statistical problems. Methods Ecol. Evol. 1, 3–14.

spring barley and oat. Agric. Food Sci. Finland 10, 335–346. Nunes, A.N., de Almeida, A.C., Coelho, C.O.A., 2011. Impacts of land use and cover type on runoff and soil erosion in a marginal area of Portugal. Appl. Geogr. 31, 687–699. Oesterheld, M., Loreti, J., Semmartin, M., Paruelo, J.M., 1999. Grazing, fire, and climate effects on primary productivity of grasslands and savannas. In: Walker, L. (Ed.), Ecosystems of Disturbed Ground. Elsevier Science, Oxford, pp. 287–306. Oomen, R.J., Linstädter, A., Ruppert, J.C., Brüser, K., Schellberg, J., Ewert, F.A., 2016. Effect of management on rangeland phytomass, cover and condition in two biomes in South Africa. Afr. J. Range Forage Sci. 33, 185–198. Orwin, K.H., Stevenson, B.A., Smail, S.J., Kirschbaum, M.U.F., Dickie, I.A., Clothier, B.E., et al., 2015. Effects of climate change on the delivery of soil-mediated ecosystem services within the primary sector in temperate ecosystems: a review and New Zealand case study. Global Change Biol. 21, 2844–2860. Ouédraogo, O., Bondé, L., Boussim, J.I., Linstädter, A., 2015. Caught in a human disturbance trap: responses of tropical savanna trees to increasing land-use pressure. For. Ecol. Manage. 354, 68–76. Palmer, A.R., Short, A., Yunusa, I.A., 2010. Biomass production and water use efficiency of grassland in KwaZulu-Natal, South Africa. Afr. J. Range Forage Sci. 27, 163–169. Polasky, S., Nelson, E., Camm, J., Csuti, B., Fackler, P., Lonsdorf, E., et al., 2008. Where to put things? Spatial land management to sustain biodiversity and economic returns. Biol. Conserv. 141, 1505–1524. Quiroga, R.E., Golluscio, R.A., Blanco, L.J., Fernandez, R.J., 2010. Aridity and grazing as convergent selective forces: an experiment with an Arid Chaco bunchgrass. Ecol. Appl. 20, 1876–1889. R Core Team, 2015. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria(URL http://www.R-project. org/). Raudsepp-Hearnea, C., Petersona, G.D., Bennettc, E.M., 2010. Ecosystem service bundles for analyzing tradeoffs in diverse landscapes. Proc. Natl. Acad. Sci. U. S. Am. (PNAS) 107, 5242–5247. Rietkerk, M., Ketner, P., Burger, J., Hoorens, B., Olff, H., 2000. Multiscale soil and vegetation patchiness along a gradient of herbivore impact in a semi-arid grazing system in West Africa. Plant Ecol. 148, 207–224. Rigge, M., Smart, A., Wylie, B., Gilmanov, T., Johnson, P., 2013. Linking phenology and biomass productivity in South Dakota mixed-grass prairie. Rangeland Ecol. Manage. 66, 579–587. Ruppert, J.C., Linstädter, A., 2014. Convergence between ANPP estimation methods in grasslands—a practical solution to the comparability dilemma. Ecol. Indic. 36, 524–531. Ruppert, J.C., Holm, A.M., Miehe, S., Muldavin, E., Snyman, H.A., Wesche, K., Linstädter, A., 2012. Meta-analysis of rain-use efficiency confirms indicative value for degradation and supports non-linear response along precipitation gradients in drylands. J. Veg. Sci. 23, 1035–1050. Ruppert, J.C., Harmoney, K., Henkin, Z., Snyman, H.A., Sternberg, M., Willms, W., Linstädter, A., 2015. Quantifying drylands’ drought resistance and recovery: the importance of drought intensity, dominant life history and grazing regime. Global Change Biol. 21, 258–1270. Sala, O.E., Paruelo, J.M., 1997. Ecosystem services in grasslands. In: Daily, G.C. (Ed.), Nature’s Services: Societal Dependence on Natural Ecosystems. Island Press, Washington, DC, pp. 237–252. Sala, O.E., Oesterheld, M., Le6n, R.J.C., Soriano, A., 1986. Grazing effects upon plant community structure in subhumid grasslands of Argentina. Vegetation 67, 27–32. Sala, O.E., Yahdjian, L., Havstad, K., Aguiar, M.R., 2017. Rangeland ecosystem services: nature’s supply and humans’ demand. In: Briske, D.D. (Ed.), Rangeland Systems. Springer, pp. 467–489. Sandhage-Hofmann, A., Kotze, E., van Delden, L., Dominiak, M., Fouche, H.J., van der Westhuizen, H.C., et al., 2015. Rangeland management effects on soil properties in the savanna biome, South Africa: a case study along grazing gradients in communal and commercial farms. J. Arid Environ. 120, 14–25. Savadogo, P., Sawadogo, L., Tiveau, D., 2007a. Effects of grazing intensity and prescribed fire on soil physical and hydrological properties and pasture yield in the savanna woodlands of Burkina Faso. Agric. Ecosyst. Environ. 118, 80–92. Sawadogo, L., Tiveau, D., Nygård, R., 2005. Influence of selective tree cutting, livestock and prescribed fire on herbaceous biomass in the savannah woodlands of Burkina Faso, West Africa. Agric. Ecosyst. Environ. 105, 335–345. Schönbach, P., Wan, H., Gierus, M., Loges, R., Müller, K., Lin, L., et al., 2012. Effects of grazing and precipitation on herbage production, herbage nutritive value and

267