Plant Ecology

, Volume 214, Issue 10, pp 1211–1222

Ecosystem responses to woody plant encroachment in a semiarid savanna rangeland

Authors

    • Department of Ecology and Natural Resource ManagementNorwegian University of Life Sciences
    • Hawassa University
  • Ørjan Totland
    • Department of Ecology and Natural Resource ManagementNorwegian University of Life Sciences
  • Stein R. Moe
    • Department of Ecology and Natural Resource ManagementNorwegian University of Life Sciences
Article

DOI: 10.1007/s11258-013-0245-3

Cite this article as:
Belay, T.A., Totland, Ø. & Moe, S.R. Plant Ecol (2013) 214: 1211. doi:10.1007/s11258-013-0245-3

Abstract

Woody plant encroachment alters the structure and function of rangeland ecosystems. The objective of this study was to explore the association between woody plant encroachment and various ecosystem properties (i.e. vascular plant species diversity, richness, evenness, soil organic matter, herbaceous biomass, leaf litter and bare ground cover) in a semiarid savanna rangeland, and also to test whether the relationships were influenced by woody species composition, elevation and site. We carried out a vegetation survey in four rangeland sites in the lower Omo region of southwestern Ethiopia, and regressed each one of the ecosystem properties, separately, against woody plant density, elevation and site using multiple linear regressions. We found that vascular plant species diversity, richness and evenness increased with woody plant density, most likely due to increased spatial heterogeneity and soil microclimate improvement. Bare ground cover increased significantly, whereas herbaceous biomass and soil organic matter did not respond to woody encroachment. In a subsequent investigation, we used a redundancy analysis to assess whether ecosystem properties were influenced by the identity of encroaching woody plant species. Species diversity and richness responded positively to Lannea triphylla, whereas leaf litter responded positively to Grewia tenax and G. villosa. Our findings suggest that woody plant encroachment in a semiarid rangeland does alter ecosystem properties. However, its impact is highly variable, influenced by a set of factors including the level of encroachment and identity of encroaching woody species.

Keywords

Bush encroachmentEcosystem propertiesCommunity structure and functionAnimal productionBiodiversityLower Omo region

Introduction

Rapid invasion of woody plants into arid and semiarid savanna and grassland areas has been reported from different parts of the world (Hudak 1999; Moleele et al. 2002; Cabral et al. 2003; Ward 2005; Knapp et al. 2008; Lunt et al. 2010). This phenomenon, called woody encroachment (Ratajczak et al. 2012), bush encroachment (Ward 2005) or shrub encroachment (Eldridge et al. 2011) has been widely perceived as a threat to animal production and wildlife conservation due to its negative impact on herbaceous production (Hudak 1999; Dalle et al. 2006). Furthermore, woody encroachment in arid and semiarid areas is considered as one of the several factors accelerating desertification, because of its negative effect on the bio-physical properties of the soil (Schlesinger et al. 1990; Parizek et al. 2002). The reduction of grass cover due to woody encroachment may increase water runoff, and further accelerate soil erosion (Parizek et al. 2002; Huxman et al. 2005). Woody plants also absorb and transpire more water than grasses and, hence, may reduce stream flow and ground water recharge (Archer 2010).

There is also a concern that woody encroachment negatively affects animal and plant diversity (Ratajczak et al. 2012). The alteration of plant community structure, such as the displacement of C4 herbaceous plants by C3 woody plants following woody encroachment, may affect the density and diversity of native plant and animal species. Several studies (such as, Scholes and Archer 1997; Angassa 2005; Price and Morgan 2008) have reported a reduction in the diversity of understory vegetation due to woody encroachment.

A review of literature indicates that woody encroachment may have variable effects on soil organic matter. For example, comparative studies (i.e., between encroached and non-encroached sites) conducted in African (e.g., Gill and Burke 1999; Hudak et al. 2003) and North American rangelands (e.g., Springsteen et al. 2010) have shown that woody encroachment may increase the organic matter content of the soil. In general, tree- and shrub-dominated ecosystems store more biomass than open grasslands (Hibbard et al. 2001; Knapp et al. 2008), adding more leaf and root litter to the soil (McCulley and Jackson 2012). Moreover, improved micro-climate condition under tree and shrub canopies, as opposed to higher heat stress in open grassland areas, facilitates the activity of litter decomposing microorganisms (Springsteen et al. 2010; Alberti et al. 2011). Nonetheless, other studies have reported either a decline (Wessman et al. 2004) or no net change (Smith and Johnson 2003) in soil organic matter in response to woody encroachment.

The mixed reports regarding the impact of woody encroachment on different ecosystem properties might be resulted due to temporal and spatial variations on species composition, and variations on the level of disturbances (i.e., herbivory and fire) and other environmental factors, such as rainfall (Belsky et al. 1989; Ratajczak et al. 2012). Furthermore, within a given rainfall and disturbance regime, the ecosystem response may vary depending on the identity of encroaching woody species (Eldridge et al. 2011), the level of interactions among co-existing species (Bertness and Callaway 1994; Callaway and Walker 1997) and the level of encroachment.

Even though several studies in the past attempted to explore the possible impact of woody encroachment on herbaceous production, which is the ultimate concern of rangeland managers because of its direct implication on animal production, we have very limited understanding of how the phenomenon affects other ecosystem properties, such as plant species diversity in arid and semiarid areas. The specific objectives of this study are, therefore, to: (1) explore the association between woody encroachment and various ecosystem properties (i.e., vascular plant species diversity, vascular species richness, vascular species evenness, soil organic matter, herbaceous biomass, leaf litter cover and bare ground cover) in a semiarid savanna, and (2) assess if the relationship is influenced by species composition (i.e., identity of encroaching woody species).

Based on previous studies we proposed the following hypotheses: First, proliferation of woody plants into water- and nutrient-constrained arid and semiarid rangeland systems increases the level of resource stress, and therefore competition with non-woody plants may lead to the decline of species diversity and herbaceous production (Lett and Knapp 2005; Ratajczak et al. 2012). Second, the shading effect of trees and shrubs in heat-stressed semiarid areas may facilitate litter decomposition processes in the soil. Therefore, we predict soil organic matter to increase with the density of trees and shrubs (Springsteen et al. 2010). Third, because the impact of woody encroachment on different ecosystem properties is a density dependent process, we expect a linear association between different ecosystem properties and densities of trees and shrubs, regardless of the identity of encroaching woody plant species.

Materials and methods

Study area

The study was conducted at the western part of the Hamer district (4.92°–5.33°N, and 36.17°–36.37°E) in the lower Omo region of Southwestern Ethiopia. It is located south of Mago national park, at the eastern side of Omo River (Fig. 1). The elevation ranges from around 400 m, in the northwest, to 550 m above mean sea level, in the southeast. A large portion of the lowland plains is dominated by savanna vegetation, whereas areas far in the southwest are dominated by tall grasses and dwarf shrubs. The area has a semi-arid climate, characterized by moderate (mean annual rainfall = 581 mm), but highly variable (CV = 33.7 %) rainfall. The primary rainy season is between March and May followed by occasional rain between October and December. The area has been designated as one of the several wildlife conservation sites in Ethiopia since mid-1970s, and it is administered by the Murule controlled hunting area (CHA). Nevertheless, the area is at the same time considered as a communal grazing land for the Hamer pastoral communities, who practice mixed herding (cattle, sheep and goats). Although herbivory and fire have been, historically, the main disturbance factors in the area, they have been reduced considerably due to diminution of the herbaceous layer from prolonged drought (Belay et al. 2013). A detail description of the study area is given by Belay and Moe (2012).
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-013-0245-3/MediaObjects/11258_2013_245_Fig1_HTML.gif
Fig. 1

Location of the study area: the position of Hamer district in the lower Omo region of southwestern Ethiopia (left), and the location of the four study sites in the western part of Hamer district (right). The broken lines indicate track trails

Site selection and sampling

We conducted a vegetation survey in 2009 where we established four sites that have relatively similar elevation, slope, soil properties and rainfall distribution. Site-1 was located next to the Murule lodge, whereas Site-2 (Gudre), Site-3 (Zewgella) and Site-4 (Lochuba) were located 15 km south, 20 km north, and 30 km southeast of the Murule lodge, respectively (Fig. 1). Based on the interpretation of temporal satellite images (Belay et al. 2013) and also witnessed by the local residents, all the sites were open grazing lands before 1980s, but recently they have experienced various degrees of woody encroachment. In each site, we marked two parallel line-transects that run along the elevation gradient, and to identify waypoints we placed visible flags approximately every 200 m. Following the way points in each line-transect, we systematically (i.e., every 400 m) placed 20 × 20 m plots (hereafter named main plot). Furthermore, we placed two 5 × 5 m plots (hereafter called subplot) within the 20 × 20 m main plot, and another 1 × 1 m plot (hereafter called inner plot) within the 5 × 5 m sub plot to sample different ecosystem attributes (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-013-0245-3/MediaObjects/11258_2013_245_Fig2_HTML.gif
Fig. 2

Multi-scale plot design technique. The inner plot (1 × 1 m) was placed inside the subplot (5 × 5 m), and the subplot within the main plot (20 × 20 m) to record various ecosystem attributes

Woody vegetation sampling

In each of the 20 × 20 m main plot (i.e., n = 24, 29, 35 and 26 for Site-1, Site-2, Site-3 and Site-4, respectively), we recorded the number of individual woody plants by species type and age category (i.e., seedling, sapling and adult). Adults include woody plants with >2.5 cm DBH (diameter at breast height), whereas saplings include young woody plants with <2.5 cm DBH and >100 cm height. All woody plants <100 cm height were defined as seedlings (Dalle et al. 2006). From each species and age category, we selected the median-size tree or shrub, and measured its height to the nearest centimeter using a calibrated pole. We also recorded the elevation and geographic references from the center of the main plot.

Floristic composition

Within the 5 × 5 subplots, we visually estimated the percentage cover of all vascular plant species following Mueller-Dombois and Ellenberg (1974). To avoid bias during visual estimation, two people were consistently involved in the entire sampling, and the average of their independent estimation was recorded. Similarly, we recorded the percent surface leaf litter cover and bare ground cover at each plot. Voucher specimens of all plant species were collected and later identified at the National Herbarium in Addis Ababa University (ET).

Herbaceous biomass and soil sampling and analysis

Within the 1 × 1 m inner plot, we sampled the soil and herbaceous biomass. All the aboveground herbaceous biomass was clipped, oven-dried at 65 °C for 24 h, and weighed with a digital balance. The two samples were averaged to estimate the dry herbaceous biomass production at the main plot level. Similarly, we extracted a composite of soil sample by mixing surface soils (0–10 cm depth) collected from the two inner-plots. We analyzed the percent soil organic matter (OM) using the loss on ignition (LOI) technique as described by Schulte and Hopkins (1996).

Statistical analyses

We produced a site based data matrix consisting of seven ecosystem property indices (i.e., vascular plant species diversity, plant species richness, plant species evenness, soil organic matter, herbaceous biomass, percent leaf litter and bare ground cover) computed at main plot level (n = 114) (Table 1). Species diversity (H′) was calculated using Shannon–Wiener index given by ΣPi × ln Pi, where Pi is the relative cover of each species (Magurran 1988), whereas species evenness (E) was given by H′/log (S), where S is the total number of species in the community. Because the woody plant distribution was not uniform across the entire study area in terms of species composition and age classes, we standardized each individual woody plant into tree equivalent (TE) units. The tree equivalent, which is given by a woody plant with 1.5 m height, is widely used to compare woody encroachment between sites (Richter et al. 2001; Roques et al. 2001). Woody density was also transformed into log scale to normalize the data.
Table 1

Mean comparison test (Tukey’s) for explanatory and response variables (mean ± SE) between the four study sites in the lower Omo region of southwestern Ethiopia

Variables

Site1

Site2

Site3

Site4

F

p

Response variables

 Vascular sp. evenness

1.5 ± 0.1b

2.4 ± 0.1a

2.3 ± 0.1.0a

2.4 ± 0.1a

26.6

<0.01

 Vascular sp. richness

13.0 ± 1.0b

33.5 ± 1.1a

32.9 ± 1.0a

32.2 ± 1.4a

69.0

<0.01

 Vascular sp. diversity

1.6 ± 0.6b

3.6 ± 0.8a

3.4 ± 0.7a

3.6 ± 0.8a

44.6

<0.01

 Bare ground cover (%)

42.8 ± 3.8ab

30.5 ± 3.7b

47.5 ± 4.0a

43.8 ± 4.5ab

3.6

0.02

 Herb biomass (kg ha−1)

2625 ± 300a

2020 ± 170b

468 ± 44d

1176 ± 110c

34.4

<0.01

 Leaf litter (%)

18.6 ± 4.3a

0 ± 0.0b

1.0 ± 0.9b

0.0 ± 0.0b

20.6

<0.01

 Soil OM (%)

1.6 ± 0.1a

1.6 ± 0.1a

1.6 ± 0.1a

1.4 ± 0.1a

0.4

0.75

Explanatory variables

 Woody density (TE ha−1)

583.2 ± 79.3b

801.3 ± 89.8b

2475 ± 205a

2827 ± 276a

36.0

<0.01

 Elevation (m.a.s.l.)

398.0 ± 0.0c

453.0 ± 6.1b

478.7 ± 12.2b

552.1 ± 9.1a

43.6

<0.01

The presence of significant variation between sites is indicated by different superscript letters. Sites with similar letters have no significant difference for the indicated ecosystem property

To understand the response of those ecosystem properties to wood encroachment, one needs to collect a long-term time series data and compare before and after encroachment. However, because such temporal studies usually require extensive years of observation, we used a “space-for-time substitution” (sensu Pickett 1989) approach. In our case, the temporal variation on woody plant density and ecosystem responses were deduced from their spatial variation on the landscape. Such space-for-time proxies have been commonly used in several ecological studies (e.g., Wiegand et al. 2005, 2006).

To address the first research objective; i.e. to explore the association between woody encroachment and ecosystem properties, we regressed each one of the seven ecosystem indices (response variables), separately, against density of woody plants, elevation, site and the interaction terms using multiple linear regression. The significance of each explanatory variable was evaluated based on p values from the F-test. We started running the full models by including all the explanatory variables; and those variables with the highest p values were sequentially removed from the models until they were significant at 0.05 level. To assess the adequacy of the models, we also checked for normality of residuals.

To address the second research objective; i.e. to examine if variations in each one of the ecosystem responses are influenced by the identity of encroaching woody plant species, we used constrained ordination analysis. Because the gradient of each axis in a DCA analysis was short (i.e., <3), we used RDA. Shorter gradient length in the DCA implies that species responded linearly to the explanatory variables and, therefore, linear models, such as RDA, are suggested as an appropriate tool (ter Braak and Smilauer 2002). RDA is a direct ordination technique commonly used to explain the linear relationship between one set of multiple independent variables against another set of multiple dependent variables (ter Braak and Smilauer 2002). For simplicity of understanding, we included the density of eight most abundant woody plant species (account >85 % of the entire woody cover) and site as explanatory variables, whereas the seven ecosystems properties were used as response variables in the analysis. To test the significance of variables, we used Monte Carlo test using 499 unrestricted permutations. All variables with p < 0.2 level of significance were considered in the subsequent analyses. We also performed variance partitioning in order to assess the joint and independent effect of each subset of predictors (i.e., woody species and site) and to test whether they are redundant to each other in explaining the observed variation in ecosystem properties (Borcard et al. 1992; Anderson and Gribble 1998). RDA was performed using the CANOCO software (ter Braak and Smilauer 2002), whereas the remaining statistical analyses were performed using R-software (R Development Core Team 2010).

Results

We found distinct variation between the four sites (ANOVA, F = 36, p < 0.01) in terms of woody plant density (Table 1). Site-3 and Site-4 had significantly higher woody density than Site-1 and Site-2 (F = 36.0, p < 0.01). Among the encroaching woody species Maerua crassifolia, Grewia tenax, Acacia nilotica and Ormocarpum trichocarpum were the most abundant species in Site-1, Site-2, Site-3 and Site-4, respectively (Fig. 4). At main plot level (i.e., 400 m2), the mean vascular plant species richness and diversity were 28.5 (±0.97 SE) and 3.12 (±0.10 SE), respectively. The mean dry herbaceous biomass was 1,490 (±110 SE) kg ha−1, with significant variation between the four sites (F = 34.4, p < 0.01), whereas the mean leaf litter cover (%) and soil organic matter (%) were 4.26 (±1.17 SE) and 1.57 (±0.06 SE), respectively (Table 1).

The multiple linear regression analysis showed that vascular plant species diversity (p < 0.01), richness (p < 0.01) and evenness (p = 0.02) were positively associated with woody plant density (Table 2; Fig. 3). Similarly, the bare ground cover increased with woody plant density (p = 0.04). However, herbaceous biomass and soil organic matter did not show a significant linear association with woody plant density (Table 3). Moreover, plant species evenness, herbaceous biomass, leaf litter cover and soil organic matter were influenced by the interaction of density and site factors (Tables 2, 3).
Table 2

Multiple linear regressions showing the association between woody density and indices of plant community structure, in the lower Omo region of southwestern Ethiopia

Ecosystem response

Variable

Estimate

SE

t Value

p

Vascular species richness (R2 = 0.68, F = 54.37, p ≤ 0.01)

(Intercept)

0.39

6.34

0.06

0.95

Density

2.05

1.01

2.03

0.04

Site2

19.58

1.75

11.20

<0.05

Site3

16.76

2.19

7.64

<0.05

Site4

15.79

2.36

6.70

<0.05

Vascular sp. diversity (R2 = 0.58, F = 35.49, p ≤ 0.01)

(Intercept)

0.06

0.77

0.08

0.94

Density

0.25

0.12

2.03

0.04

Site2

1.92

0.21

9.01

<0.05

Site3

1.46

0.27

5.47

<0.05

Site4

1.57

0.29

5.46

<0.05

Vascular sp. evenness (R2 = 0.49, F = 13.76, p ≤ 0.01)

(Intercept)

−0.40

0.80

−0.50

0.62

Density

0.30

0.13

2.32

0.02

Site2

0.75

1.21

0.62

0.54

Site3

3.46

1.43

2.43

0.02

Site4

3.24

1.54

2.10

0.04

Density:Site2

0.01

0.19

0.05

0.96

Density:Site3

−0.40

0.20

−2.01

0.05

Density:Site4

−0.36

0.21

−1.68

0.10

Elevation was not selected in any of the models

https://static-content.springer.com/image/art%3A10.1007%2Fs11258-013-0245-3/MediaObjects/11258_2013_245_Fig3_HTML.gif
Fig. 3

Association between woody encroachment and different ecosystem properties: A vascular plant species richness, B species evenness, C species diversity, D herbaceous biomass, E bare ground cover, and F soil organic matter. Observations for each one of the four sites are represented by different shapes (open triangle for Site-1, open diamond for Site-2, open circle for Site-3, and open square for Site-4)

Table 3

Multiple linear regressions showing the association between woody density and indices of ecosystem function in the lower Omo region of southwestern Ethiopia

Ecosystem response

Variable

Estimate

SE

t Value

p

Bare ground cover (%) (R2 = 0.24, F = 7.86, p = < 0.01)

(Intercept)

−0.08

0.18

−0.46

0.64

Density

0.06

0.03

2.12

0.04

Site2

0.13

0.05

2.72

0.01

Site3

0.10

0.06

1.62

0.11

Site4

0.12

0.07

1.82

0.07

Herbaceous biomass (kg ha−1) (R2 = 0.64, F = 64.47, p = < 0.01)

(Intercept)

1.44

0.23

6.17

<0.05

Density

−0.04

0.04

−1.15

0.25

Site2

−0.29

0.30

−0.99

0.33

Site3

−1.05

0.36

−2.94

<0.05

Site4

−0.46

0.38

−1.20

0.23

Density:Site2

0.04

0.05

0.82

0.42

Density:Site3

0.10

0.05

2.01

0.05

Density:Site4

0.05

0.05

0.86

0.39

Leaf litter cover (%) (R2 = 0.41, F = 10.35, p = < 0.01)

(Intercept)

69.69

18.83

3.70

<0.05

Density

−8.28

3.03

−2.73

0.01

Site2

−69.49

26.72

−2.60

0.01

Site3

−92.60

33.47

−2.77

0.01

Site4

−69.69

36.11

−1.93

0.06

Density:Site2

8.25

4.20

1.96

0.05

Density:Site3

11.38

4.70

2.42

0.02

Density:Site4

8.28

4.96

1.67

0.10

Soil OM (%) (R2 = 0.12, F = 1.26, p = 0.28)

(Intercept)

1.04

0.27

3.80

<0.05

Density

−0.04

0.05

−0.91

0.37

Site2

−0.17

0.39

−0.43

0.67

Site3

−1.08

0.47

−2.33

0.02

Site4

0.42

1.17

0.36

0.72

Density:Site2

0.03

0.06

0.53

0.60

Density:Site3

0.15

0.07

2.33

0.02

Density:Site4

−0.03

0.15

−0.21

0.83

Elevation was not selected in any of the models

Furthermore, the RDA analysis revealed a positive association between ecosystem properties and density of individual woody species. As shown in Fig. 4, vascular plant species diversity, richness and evenness were associated with the density of O. trichocarpum, Lannea triphylla and A. nilotica, whereas herbaceous biomass and leaf litter cover were associated with the density of Grewia villosa and M. crassifolia. Similarly, soil organic matter was associated with the density of nitrogen fixing woody plants including A. nilotica and A. brevispica. The total effect of the explanatory variables (i.e., woody species density and Site) on ecosystem properties was 37.8 %, of which Site alone explains much of the variation (i.e., about 31.7 %). The variance partitioning also shows that the two sets of variables appear to be redundant in explaining the observed variation in ecosystem properties (Table 4). The eigenvalues of the first four axes in the RDA were 0.266, 0.052, 0.034 and 0.027, respectively. The first two axes accounted for 84.0 % of the variation in woody plant species—ecosystem properties relationship (Fig. 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-013-0245-3/MediaObjects/11258_2013_245_Fig4_HTML.gif
Fig. 4

RDA bi-plot showing the association between ecosystem properties (response variable) and density of individual woody plant species (A. brevis = Acacia brevispica, A. niloti = A. nilotica, G. tenax = Grewia tenax, G. villos = G. villosa, L. triphy = Lannea triphylla, M. crassi = Maerua crassifolia, M. oblong = M. oblongifolia, O. tricho = Ormocarpum trichocarpum) and sites in the lower Omo region of southwestern Ethiopia. The ecosystem properties (VascRich for vascular species richness, VascEven for vascular species evenness, VascDivr for vascular species diversity, HerbBiom for herbaceous biomass, Leaf_lit for leaf litter, BareCovr for bare ground cover, Soil_OM for soil organic matter) are shown by dashed arrows. The eigenvalues of the first four exes are 0.266, 0.052, 0.034 and 0.017, respectively. The first two axes account for 84.0 % of the variation in ecosystem properties–woody species density relationships

Table 4

RDA variance partitioning to assess the relative influence of woody density on various ecosystem properties in the lower Omo region of southwestern Ethiopia

Main variable

Co-variable

Sum of canonical eigenvalues

Explained variation (%)

F-ratio

p Value

Total effect

 Woody species + sites

0.378

37.8

5.635

0.002

Partial effects

 Woody species

Site

0.061

6.1

1.243

0.174

 Site

Woody species

0.168

16.8

9.185

0.002

 Woody species

0.210

21.0

3.488

0.002

 Sites

0.317

31.7

17.045

0.002

Joint effectsa

 Woody species and sites

 

0.139

13.9

  

aSince the variation explained by the joint effects (i.e. 0.139) is greater than the variation partially explained by Woody species (0.061), the two sets of variables (i.e. woody species and site) appear to be redundant in explaining ecosystem properties

Discussion

Woody encroachment has been considered a major threat to animal production and biodiversity conservation in different rangelands. Ratajczak et al. (2012), for example, reviewed results of 29 studies in 13 communities in North America, and concluded that plant species richness declined by 45 % due to woody encroachment. Lett and Knapp (2005) similarly indicated that displacement of the dominant herbaceous vegetation by woody plants has a dramatic negative effect on plant species richness and diversity. However, in the lower Omo region of southwestern Ethiopia, we found an increase of plant species diversity, evenness and richness with woody plant encroachment.

Our findings are consistent with the habitat heterogeneity hypothesis which suggests that species diversity increases in structurally complex habitats because of an increase in available niches and environmental resources (Macarthur and Macarthur 1961; Cramer and Willig 2005). The increased structural complexity of vertical layering in the ecosystem, following woody encroachment, increases species diversity by providing more space for different species (Eldridge et al. 2011). Large Acacia trees, for example, are known to facilitate the growth of several shade-loving shrubs, such as Grewia species under their canopies (Schleicher et al. 2011). Woody encroachment can also increase plants species diversity through aboveground and belowground facilitation. Facilitative interaction is common especially in arid and semiarid savannas (Callaway and Walker 1997; Schleicher et al. 2011), by which trees and shrubs ameliorate the growth of understory vegetation either by improving harsh environmental conditions (such as heat stress), by altering substrate characteristics or by increasing resource availability (Gomez-Aparicio et al. 2005; Maestre et al. 2009). Some C3 plants, for example, can grow better under the canopy of trees by benefiting from improved microhabitat, mainly associated with increased soil moisture and fertility (Belsky et al. 1989). Moreover, woody plants create islands of resources (also called fertility islands) that may serve as a refuge for some plant and animal species that are constrained by the poor nutrient conditions (Maestre et al. 2009; Pugnaire et al. 2011). Fertility islands are fertile soil patches beneath tree and shrub canopies that may result from an accumulation of nutrients transported from fertile areas by different agents, such as water and animals (Schlesinger et al. 1990).

We found a positive association between bare ground cover and woody plant density, pointing to the potential negative consequences of woody encroachment in accelerating soil erosion. Woody encroachment in arid and semiarid areas has previously been associated with rangeland degradation (Parizek et al. 2002) due to its negative effect on the hydrology and physico-chemical properties of the soil. Though increasing woody plant cover in moist areas can decrease soil evaporation by reducing the near-ground solar radiation (Martens et al. 2000), the loss of the understory herbaceous vegetation and the associated increase of bare ground cover may increase the potential for soil erosion (Parizek et al. 2002). Moreover, woody plants transpire more water than the herbaceous vegetation (Archer 2010), which may potentially deplete soil moisture and make the soil more prone to wind erosion (Parizek et al. 2002; Huxman et al. 2005).

Contrary to our expectation, herbaceous biomass was not related to woody plant density. Based on the previous studies (e.g., Richter et al. 2001; Angassa 2005; Dalle et al. 2006), we had expected a negative association between woody plant density and herbaceous biomass. Reduction in herbaceous biomass is largely related to interception of rainwater and solar radiation by encroaching woody plants (Belsky et al. 1989). Some woody plants, for example, Acacia mellifera have thick canopies that can intercept up to 50 % of rain water with their leaves and branches in order to maximize infiltration around the stem (Donaldson 1969). Nevertheless, in our study the competitive impact of woody plants on herbaceous biomass was not clearly evident. This is likely due to the multi-year drought that limits plant production before competition between woody and herbaceous vegetation becomes apparent (Pugnaire and Luque 2001). The level of competition, according to the stress-gradient hypothesis (Grime 1979; Pugnaire and Luque 2001), declines with increasing abiotic stress.

Like herbaceous cover, the soil organic matter also did not respond to woody plant density. Even though several previous studies (e.g., Hibbard et al. 2001; Brantley and Young 2010) have reported an improvement of soil organic matter following woody plant encroachment, we did not see any linear association between the two. This could be similarly associated with intensification of moisture stress in the area that is resulted from the prolonged drought recorded during our study period. Moisture limits the amount of soil organic matter incorporated into the soil by limiting plant production and the rate of litter decomposition (Hudak et al. 2003). Moreover, the amount of soil organic matter on the landscape is largely influenced due to spatial variation on plant species composition (Finzi and Schlesinger 2002).

We found that woody species composition (Eldridge et al. 2011) and the level of encroachment (Bertness and Callaway 1994; Callaway and Walker 1997) are important factors that influence the outcome of ecosystem responses to woody plant encroachment. As shown in the RDA analysis (Fig. 4), leaf litter, for example, was associated with the density of deciduous shrubs, such as G. tenax and G. villosa that generally contribute more leaf litter to the soil compared to other evergreen woody plants (Mooney 1972; Cornwell et al. 2008). Similarly, soil organic matter responded positively to nitrogen fixing woody plants, such as A. nilotica and A. brevispica. As previous studies indicated (e.g., Liao and Boutton 2008), the presence of nitrogen fixing woody plants may boost the ecosystem net primary production, thereby improving soil organic matter.

Even though we had expected no major difference among the four sites due to their similarity in rainfall, soil organic matter and elevation, the site factor significantly explained the variation in ecosystem properties (Tables 2, 3, 4). This could be partly due to possible correlation and redundancy in the two sets of variables (Woody plant density and Site) in explaining ecosystem properties (Table 4), and partly due to the clumped nature of species distribution. For example, site 3 was dominated by A. nilotica, whereas Site 4 was dominated by Oromocarpus trichocarpum (Fig. 4), and these two species may have a different effect on ecosystem properties. Furthermore, the variations in species evenness, leaf litter cover and soil organic matter were more significantly explained by woody density in site 3 (highly encroached site) than in site 1 (less encroached site). This is consistent with the notion that the impact of woody encroachment follows threshold dynamics (Petersen et al. 2009), by which its impact on ecosystem properties becomes apparent beyond a certain threshold level. Richter et al. (2001) and Roques et al. (2001), for example, used 30 % woody cover (or 1,400 tree equivalent per ha) as a threshold for woody encroachment, above which many ecosystem properties in the rangeland start to respond significantly.

Conclusion

Though woody encroachment potentially accelerates soil erosion, it increases plant species diversity and richness due to increasing complexity of the spatial structure of the ecosystem, including the formation of multiple vegetation layers. On the other hand, herbaceous biomass and soil organic matter did not respond to woody encroachment, most likely, due to the prolonged drought during the study period that limited plant production and litter decomposition. We conclude that like many other biological invasion phenomena, proliferation of woody plants into savanna rangelands can have ecosystem-wide effects. However, its impact is highly variable depending on the level of encroachment and the local species composition.

Acknowledgments

The fieldwork was financed by Hawassa University and Norwegian Agency for Development Cooperation (NORAD). We thank Hamer district administration and the staff of Murule lodge for their friendly support and cooperation during the fieldwork.

Copyright information

© Springer Science+Business Media Dordrecht 2013