- Email: [email protected]
Roads increase woody cover under varying geological, rainfall and fire regimes in African savanna
Journal of Arid Environments 80 (2012) 74e80
Contents lists available at SciVerse ScienceDirect
Journal of Arid Environments journal homepage: www.elsevier.com/locate/jaridenv
Roads increase woody cover under varying geological, rainfall and fire regimes in African savanna I.P.J. Smit a, *, G.P. Asner b a b
Scientific Services, South African National Parks, Private Bag X402, Skukuza 1350, South Africa Department of Global Ecology, Carnegie Institution, 260 Panama Street, Stanford, CA 94305, USA
a r t i c l e i n f o
a b s t r a c t
Article history: Received 1 June 2011 Received in revised form 21 October 2011 Accepted 29 November 2011 Available online 20 January 2012
Roads extend throughout savannas, yet few studies have quantified their effects on adjoining woody vegetation structure. Airborne LiDAR imagery collected over 168 experimental fire plots in the Kruger National Park, all bounded by graded firebreak roads, provided an opportunity to study if, and to what extent, roads influence woody vegetation structure under different rainfall, geologic and fire conditions. In 91.7% of the plots, woody canopy cover was higher on the edges of roads compared to areas farther away. The increase was most pronounced within 5 m of the road edge, but was detectable 10e15 m from the edge. On average, the area within 15 m from the road had approximately 6% and 2% higher woody vegetation cover than areas further than 15 m from the edge on wetter granitic and drier basaltic savanna landscapes, respectively. Increased edge effects on woody cover were observed even in fire exclusion plots, suggesting that non-fire processes, likely altered hydrological regimes, may be the underlying reason for woody encroachment. This study illustrates that roads cause selective woody plant thickening in savannas, even in areas without road edge management, and therefore careful consideration should be paid on how road edges are managed and when roads are planned. Ó 2011 Elsevier Ltd. All rights reserved.
Keywords: Bush thickening Hydrology Kruger National Park LiDAR Road edges Woody encroachment
1. Introduction Roads, and their associated edges, have numerous impacts on biodiversity patterns and processes. In a review, Trombulak and Frissell (2000) identified seven general negative effects of roads on biodiversity: mortality from road construction, mortality from collision with vehicles, modification of animal behaviour, alteration of the physical environment, alteration of the chemical environment, spread of exotics, and increased use of areas by humans. However, other papers have shown that road edges can also have several positive effects, especially in highly transformed landscapes, where they act as refugia and/or movement corridors (Eversham and Telfer, 1994). Foreman and Alexander (1998) proposed that road impacts operate at two primary levels, namely (i) the individual, species and population level (e.g. road kills, disturbance, etc.), and (ii) the ecosystem process and landscape level (e.g. hydrology, erosion, sedimentation, etc.). In 2003 it was estimated that there were approximately 8000 km of roads in the Kruger National Park (KNP), South Africa
* Corresponding author. Tel.: þ27 13 735 4257; fax þ27 13 735 4055. E-mail addresses: [email protected] (I.P.J. Smit), [email protected] (G.P. Asner). 0140-1963/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.jaridenv.2011.11.026
(Freitag-Ronaldson and Foxcroft, 2003). Considering the extent of roads and their potential impacts, surprisingly few “road ecology” studies have been conducted to understand potential impacts of roads in the KNP. The only published study we could find for KNP was a qualitative description of road impacts based on personal observations of a resident scientist (Pienaar, 1968). Beyond KNP, we found that studies focussing on the effects of roads on adjoining woody vegetation structure in savannas seem to be lacking from the literature. This study aims to address that gap by determining if, and to what degree, woody vegetation structure is influenced on the edges of low-maintenance graded firebreak roads under different rainfall, geological and fire management regimes (in the absence of road-edge management like scraping or mowing). Graded firebreak roads refer to linear strips that are created when bulldozers and/or scrapers denude an area of vegetation (woody and herbaceous) to leave a rudimentary road (Fig. 1). These graded areas act as firebreaks and serve as management roads for very low volume traffic, such as game ranger patrols and research vehicles. Here we addressed the following questions: Is there a difference in woody vegetation structure in areas closer to edges of roads compared to areas further from the edges? If so, how consistent is the edge effect, and how far from the edge of the road is the effect measureable? How does the edge effect differ between different geologically- and rainfall-defined landscapes? Do different fire
I.P.J. Smit, G.P. Asner / Journal of Arid Environments 80 (2012) 74e80
75
Table 1 Description of the four studied landscapes (based on van Wilgen et al. (2007)). Landscape
Vegetation type
Dominant tree species
Geology
Mean annual precipitation (mm)
Pretoriuskop
Sourveld
Granite
705
Skukuza
Combretum
Granite
572
Satara
KnobthornMarula Mopane
Terminalia sericea, Dichrostachys cinerea Combretum collinum, Combretum zeyheri Acacia nigrescens, Sclerocarya birrea Colophospermum mopane
Basalt
507
Basalt
451
Mopane
Fig. 1. Graded firebreak roads of approximately 6 m wide delineate Experimental Burn Plots (EBPs) in the Kruger National Park. Different fire management regimes are practiced on each side of the road.
management regimes lead to different edge effects on woody vegetation structure? 2. Materials and methods 2.1. Study site KNP covers an area of almost 20,000 km2 in the north-eastern corner of South Africa, bordering Zimbabwe to the north and Mozambique to the east. The Park contains a north-south rainfall gradient with rainfall increasing from the north (w350e420 mm yr1) to the south (w680e750 mm yr1). The western part of the Park is mainly underlain by granite-derived soils and the east by soils of basaltic origin, with some other geologies covering smaller sections of the Park. This abiotic template gives rise to 35 major vegetation types which are dominated by knobthorn (Acacia nigrescens), marula (Sclerocarya birrea), leadwood (Combretum imberbe), red-bush willow (Combretum apiculatum), silver cluster leaf (Terminalia sericea) and mopane (Colophospermum mopane) (Gertenbach, 1983). 2.2. Experimental Burn Plots (EBPs) The aim of the KNP Experimental Burn Plots (EBPs) is to study the effects of fire (frequency and season) on the vegetation under the grazing pressure of indigenous herbivores (Van der Schijff, 1958). The EBP experiment consisted of the application of fires at varying return intervals and seasons, and protection from fire, on a series of w7 ha plots in four of the major vegetation landscapes of the Park (Table 1). The treatments were replicated four times in each of these landscapes (replicates are called “strings”). The treatments originally included annual winter fires in August, and biennial and triennial fires in August, October, December, February and April. In 1976, further treatments to examine the effects of fires every 4 and 6 years in October were added to selected landscapes (Satara & Mopane) by subdividing the February treatment plots. This resulted in 48 plots (4 replicate strings 12 treatments) in the Skukuza and Pretoriuskop landscapes and 56 plots (4 replicate strings 14 treatments) in the Satara and Mopane landscapes. One string in each of the Skukuza, Satara and Mopane landscapes was not included in this study because local soil differences made them unrepresentative of the particular landscapes (Venter, 2004), leaving 48 plots in the Pretoriuskop landscape, 36 in the Skukuza landscape and 42 in the
Satara and Mopane landscapes. Full details and history of the experimental design and application of treatments are available from Biggs et al. (2003). 2.3. LiDAR derived top of canopy dataset Airborne Light Detection and Ranging (LiDAR) data were acquired in April 2008 using the Carnegie Airborne Observatory (CAO) (Asner et al., 2007). The CAO was mounted in an aircraft which flew over the 168 EBPs at an altitude averaging 2000 m above ground level. The CAO LiDAR was operated at 50 kHz pulse repetition frequency and a 34 scanning configuration cross-track of the aircraft direction. Global positioning system (GPS) and Inertial Motion Unit (IMU) data were combined to determine the 3-D location of the laser returns. From the laser “point cloud” data, a physical model was used to estimate top-of-canopy and ground surfaces (digital elevation models; DEM) using the Terrascan/Terramatch (Terrasolid Ltd., Jyväskylä, Finland) software package. Vegetation height was subsequently estimated by subtracting the ground DEM from the top-of-canopy layer. The CAO generated a laser spot spacing of 1.12 m. Since the beam diameter at ground level was also 1.12 m, this laser spot spacing configuration resulted in a 50% overlap between LiDAR observations, decreasing the likelihood of missing vegetation canopy. The resulting canopy DEM had a spatial resolution of 1.12 1.12 m, such that a surface was generated where each 1.24 m2 contained the estimated vegetation height in that particular pixel. The absolute vertical resolution of the CAO LiDAR is 15 cm, but following digitization and application to porous tree canopies, the effective vertical resolution is 0.2e0.5 m. All further analyses were based on these top of canopy surfaces created for each EBP, which is an estimate of woody vegetation cover. Woody vegetation cover is a widely used variable to quantify how horizontal woody structure varies across space and time (e.g. Fensham and Fairfax, 2003). Other variables like tree density can also be used to characterize woody structure, but these parameters were not directly available from the LiDAR dataset employed here. Due to the lack of late summer rains in 2008 (http://www. sanparks.org/parks/kruger/conservation/scientific/weather/) (accessed 30 May 2011), some woody vegetation already started dropping leafs in April when the LiDAR survey was conducted. This resulted in decreased leaf area, especially on the drier northern basaltic landscapes (Satara and Mopane). Lower leaf cover, together with the spatial resolution employed, resulted in some woody vegetation, especially smaller and sparse individuals, not being detected. However, as we do not compare sites over time but over space, the effect of underestimation on the emerging structural patterns was small. That is, since we are comparing edges to core areas of each EBP, the systematic underestimation would be consistent across and between plots, and therefore would not affect the overall patterns and trends (Smit et al., 2010).
76
I.P.J. Smit, G.P. Asner / Journal of Arid Environments 80 (2012) 74e80
An ANOVA followed by Sidak pairwise comparisons (Sidak, 1967) was conducted to test whether landscape (four levels; Pretoriuskop, Skukuza, Satara, Mopane) and fire treatment (14 levels; August annual, February tri-annual, fire exclusion, etc.) significantly influenced the average difference in percent woody cover between the edge and the core area of the EBPs. Finally, we tested how percent woody cover differs for different vegetation height classes between the edges and cores of the EBPs to determine how the edge effects influence the 3-D woody vegetation structure. 3. Results
Fig. 2. The Experimental Burn Plot (EBP) is depicted in light grey, the graded firebreak road surrounding the EBP is depicted in dark grey, and the 5 m distance bands from the graded firebreak road towards the centre of the EBP is depicted by the dotted lines (not to scale). Percentage woody cover was determined for each of these 5 m incremental distance bands.
2.4. Data analyses Percentage woody cover was determined in 5 m incremental distance bands from the outside of an EBP towards the centre, making use of the CAO LiDAR derived “top-of-canopy” layer described above. For example, woody cover was determined for the area 0e5 m from each graded firebreak road, 5e10 m from the road, 10e15 m from the road, etc., systematically moving away from the edge of the plot towards the centre (Fig. 2). This was repeated on four fire exclusion EBPs (plots on which fire has been excluded since 1954) in the Pretoriuskop landscape and on three fire exclusion EBPs in the three other landscapes (Skukuza, Satara and Mopane) in order to assess the extent of the edge effect in the absence of fire. Based on the extent of the edge effects demonstrated from the analysis above, each EBP was subsequently subdivided into two zones, namely (i) an edge area (area inside EBP within 15 m from firebreak road), and (ii) a core area (area inside EBP further than 15 m from firebreak road) (Fig. 3).
Areas closer to the graded firebreak roads were more densely wooded than corresponding areas closer to the centre of the EBP, with areas directly bordering the road (<5 m) being the most densely wooded (Fig. 4). This was consistently observed up to about 10 m from the firebreak, with the % woody cover levelling out at a background level between 10 and 15 m. In the Pretoriuskop landscape, the edge effect occurred up to about 15 m before levelling off between 15 and 20 m. In 154 of the 168 EBPs studied (91.7%), percent woody cover was higher in the edge area than in the core area. On the Pretoriuskop and Skukuza EBPs, the edge maintains on average 5.5e6.0% more woody cover than the paired core area of the same EBPs (Fig. 5) (Paired T-test {edge minus core}: Pretoriuskop t ¼ 6.99, df ¼ 47, p < 0.001; Skukuza t ¼ 9.60, df ¼ 35, p < 0.001). For the Satara and Mopane EBPs, the edge contains about 2% more woody cover than the paired core area (Fig. 5) (Paired T-test {edge minus core}: Satara t ¼ 7.26, df ¼ 41, p < 0.001; Mopane t ¼ 8.77, df ¼ 41, p < 0.001). An ANOVA with “edge effect” (difference in % woody cover between edge and core area) as the dependent variable, and landscape and fire treatment as independent variables (including the interaction term), indicates that the landscape is a significant explanatory variable (F ¼ 13.88; df ¼ 3; p < 0.0001), but not the fire treatment (F ¼ 0.62; df ¼ 13; p ¼ 0.62) nor the interaction term (F ¼ 0.63; df ¼ 35; p ¼ 0.63). In other words, the fire treatment does not play a statistically significant role in determining the magnitude of the edge effect. Sidak pairwise comparisons (Sidak, 1967) show that the average edge effect is significantly more pronounced in the wetter granitic Pretoriuskop and Skukuza landscapes than on the drier basaltic Satara and Mopane landscapes. Note that the average edge effect is not statistically significantly different 70 60
% woody cover
50 40 30 20 10 0 0-5
5-10
10-15
15-20
20-25
25-30
30-35
35-40
40-45
45-50
Distance from firebreak (m) Fig. 3. The striped area depicts an Experimental Burn Plot (EBP) whereas the solid area depicts the graded firebreak road surrounding the EBP. The area inside each EBP closer than 15 m from the road was termed the “edge” (light striped), whilst the remainder of the plot (i.e. the area further than 15 m from the road) was termed the “core” (dark striped). For each plot the “edge effect” was defined as the difference between the percent woody cover in the edge and the percent woody cover in the core of the same plot.
Pretoriuskop
Skukuza
Satara
Mopane
Fig. 4. Average percentage woody cover (and standard error) for fire exclusion EBPs at incremental 5 m distance bands from the graded firebreak road (n ¼ 4 for Pretoriuskop and n ¼ 3 for Skukuza, Satara and Mopane). The dotted line indicates where background levels of percent woody cover were reached (i.e. where the edge effect is no longer detectable).
I.P.J. Smit, G.P. Asner / Journal of Arid Environments 80 (2012) 74e80
(Fig. 7). For the Satara and Mopane landscapes, the edge effect is more pronounced in the lower height classes, whereas in Skukuza and Pretoriuskop, canopy cover is more enhanced in the intermediate height classes between 3.5 and 6 m. Note that this is “top-ofcanopy” cover, and hence smaller individuals growing under higher individuals will not be reflected in the analysis.
8 7 6 5 4
4. Discussion
3
Edge effects of roads on woody vegetation structure have been poorly studied in savanna systems. This study is, to our knowledge, the first to quantify how roads in different geological, rainfall and fire contexts influence woody vegetation structure in a natural semi-arid savanna system. The results presented here illustrate that graded firebreak roads can increase woody vegetation cover. Since the effects of roads on woody vegetation structure were observed even in the absence of five decades of fire (i.e. on the plots where fire has been excluded since 1954), we conclude that increased woody cover along road edges is not primarily due to changes in fire characteristics in close proximity to roads. In fact, results presented in Fig. 6 suggest that the effect of increased woody cover on road edges is often more pronounced in the absence of fire. We therefore hypothesize that these patterns of increased woody cover are primarily caused by changes in hydrological patterns and processes, and are not due to differences in fire characteristics and fire behaviour along roads (e.g. cooler fires next to roads). Road-induced hydrological changes may cause a gradient of decreasing wetness away from the roads due to:
2 1
Pretoriuskop
Skukuza
Satara
Mopane
Fig. 5. Average (and standard error) difference in % woody cover between the edge of the EBPs and the core area (n ¼ 48, 36, 42 and 42 for Pretoriuskop, Skukuza, Satara and Mopane landscapes respectively).
between Pretoriuskop and Skukuza and between the Satara and Mopane landscapes respectively. The ANOVA and pairwise comparison results are in agreement with Fig. 5. The data provide some suggestion that the edge effect may be more pronounced (although not statistically significantly e see ANOVA above) on the fire exclusion sites as compared to the fire treatment sites. For 10 of the 13 replicate strings, the fire treatments (all treatments combined) harbored weaker edge effects than observed in the fire exclusion plots (Fig. 6). We also found that percent woody cover between the edges and cores of the EBPs varies among different vegetation height classes
(% woody cover edge) minus (% woody cover core)
18 16 14 12 10 8 6 4 2
se
e
nd e
e
-T
-M
se
op
m zo
M
op
an
an op M
op M
an
bo
a -D an
e
ta r
a
-N
-S
wa
at
ne
ar
ts
i
ya he ra ta
ra ta Sa
-N
Sk
uk
uz
a
uk Sk
Sa
it s wa
sw
-S uz
a
uz
ar
ha
ku
ka
za
i ku
-N a
ha
uk Sk
-S op sk iu
or et Pr
ap
ni be
bi um -N
op sk iu or et
iu or et Pr
Pr
sk
or
op
iu
sk
-K
op
am
-F
be
ay
ni
i
0
et Pr
increased run-off from roads (i.e. similar to rainfall-harvesting and run-off farming) (Wani et al., 2009),
Sa
0
-M
(% woody cover edge) minus (% woody cover core)
77
Landscape - Replicate String Fire treatment
Fire exclusion
Fig. 6. Average (and standard error) difference in % woody cover between the edges of the EBPs and the paired core areas. The results are reported separately for the fire treatments (grey shading; n ¼ 11 for Pretoriuskop and Skukuza and n ¼ 13 for Satara and Mopane) and fire exclusion plots (black shading; n ¼ 1) respectively. Note that for 10 of the 13 replicate strings the edge effect was more pronounced for the fire exclusion compared to the fire treatment plots.
78
I.P.J. Smit, G.P. Asner / Journal of Arid Environments 80 (2012) 74e80
0.8
0.6
0.4
0.2
0
0. 51 11. 5 1. 52 22. 5 2. 53 33. 5 3. 54 44. 5 4. 55 55. 5 5. 56 66. 5 6. 57 77. 5 7. 58 88. 5 8. 59 99. 9. 5 51 0 10 -1 1 0 0.5 .5 11 11 -1 1 1 1.5 .5 12 12 -1 1 2 2.5 .5 13 13 -1 1 3 3.5 .5 14 14 -1 1 4 4.5 .5 15 15 -1 1 5 5.5 .5 16 16 -1 1 6 6.5 .5 -1 7
(% woody cover edge) m inus (% woody cover core)
1
-0.2
Height class (m) Pretoriuskop
Skukuza
Satara
Mopane
Fig. 7. Average (and standard error) difference in % woody cover between the edge area and the core area of the EBPs at 0.5 m height increments of woody vegetation cover.
roads acting as barriers for water movement (i.e. water obstructed and dammed by roads due to poor cross-drainage) (Stoeckeler, 1965), and positive feedback between plant density, water infiltration and soil moisture through spatial redistribution of run-off (Hillerislambers et al., 2001; Rietkerk et al., 2000). If roads are altering local spatial patterns in hydrological inputs, woody vegetation closer to the edges of roads will have access to elevated soil moisture for longer periods, resulting in longer growing seasons (Fig. 8). This possibility is supported by our observations of roadside woody canopy height distributions along the regional precipitation gradient (Fig. 7): Relative differences in canopy height change from taller canopies in wetter zones (Pretoriuskop, Skukuza) to smaller canopies in the drier regions of Satara and Mopane. This likely occurs because woody plants are able to take advantage of the additional moisture provided by roadside run-off in the wetter areas, whereas the difference is suppressed in the drier zones. Overall, the edge effect is most pronounced closer to roads (<5 m), but is detectable up to about 5e10 m from roads in the Skukuza, Satara and Mopane landscapes and 10e15 m in the Pretoriuskop landscape (Fig. 4). Similar to the relative effects of precipitation on height distributions, here too we postulate that the edge effect extends further from road edges in the Pretoriuskop landscape due to higher rainfall conditions, resulting in higher runoff over a larger distance. The distances reported here agree with studies that have shown that run-off can affect soils up to 25 m from the road (Garcia et al., 1996; Reinirkens, 1996). Edge effects were detected both in the relatively open as well as in well-wooded landscapes of KNP (ranging from the Satara and Mopane landscapes with <10% woody cover to the denser cover of >50% observed in the Skukuza and Pretoriuskop landscapes) (Fig. 4). This covers a wide range of woody cover estimates which typically occurs in African savannas (most of the 854 African savanna sites explored by Sankaran et al. (2005) had woody cover ranging between 0 and 60%).
The data suggest that the edge effect may be more pronounced on the fire exclusion sites than on the fire treatment sites (Fig. 6). This may be due to the fact that the fire exclusion sites, with higher vegetative biomass (Smit et al., 2010), have greater hydraulic roughness, which increases infiltration and storage of water in soils (Foreman and Alexander, 1998). The higher hydraulic roughness (i.e. more vegetation) of the fire exclusion sites reduces the lateral movement of the run-off when it moves across the surface adjoining the road, thus facilitating vertical recharge (Hillerislambers et al., 2001).
Fig. 8. Mopane shrubland in northern Kruger National Park. Compare the denser and taller mopanes (and greener e see online colour image) on the edge of the tar road with individuals further away from the road (area directly bordering the road was scraped). This photo was taken November 2010 after the first spring rains e these rains were not enough to break winter dormancy for the trees further away from the road which have no leaves, whereas the increased run-off from the road caused the trees in close proximity to flush. This demonstrates how trees on road edges are favoured by (i) having access to more water and (ii) having a longer growing season. Although the edge effect of roads is obvious in this example, it is subtle and hard to measure and detect in most of the Park where woody vegetation structure is complex and multilayered. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
I.P.J. Smit, G.P. Asner / Journal of Arid Environments 80 (2012) 74e80
Trees growing relatively close to roads but not directly in the “wet” edge, may also be advantaged because savanna trees can access water and nutrients over wider areas than just directly below their main stems. For example, Kulmatiski et al. (2010) found that trees in southern KNP can access water up to 8 m from their main stems. Moreover, studies have found that trees are more productive when competition with grass is reduced (e.g. Riginos, 2009; Van Der Waal et al., 2009). Therefore, woody vegetation on road edges may have roots underlying the (grassless) graded area thereby avoiding competition with grasses for water. These hypotheses can be tested by installing data loggers that monitor soil moisture gradients and dynamics in the vadose zone on the firebreak road and at increasing distances from the road. This will provide insights into how roads influence the soil moisture regime through space and time as a result of altered hydrological patterns. If the firebreak is often used as a road and if intense storms compact the soil, the surface may increasingly seal, reducing infiltration on the road itself and increasing run-off. Increased grazing pressure may also contribute to selective bush thickening along road edges. Many herbivores are attracted to the short green grasses which grow along the roads (Pienaar, 1968). This is particularly evident during the drier seasons when most of the grass in the Park is senesced, and greenery appears along the roads as a result from run-off from small rainfall events. Since increased grazing is one of the potential contributing mechanisms responsible for woody encroachment/thickening (Archer et al., 1995), we postulate that increased grazing along road verges may also contribute, together with the hydrological changes described above, towards increased woody vegetation along the road edges over the long-term. 5. Management implications and future studies Roads (including graded firebreaks, management roads and airstrips) covered 7926 km in the KNP in 2003 (Freitag-Ronaldson and Foxcroft, 2003). Assuming that other types of roads may have similar effects to the graded firebreaks examined in this study, the edge effect of roads on woody vegetation may affect significant areas of the Park. A woody cover edge effect of 15 m on either side of roads equates to approximately 12,000 ha throughout the Park, which is roughly 1.2% of the total Park area (this is over and above the surface area of the road itself). This suggests that roads may potentially contribute to selective bush thickening of 12,000 ha of the park through subtle edge effects. This effect needs to be considered in addition to the impacts normally considered when new roads are planned. It must also be stressed that we considered edge effects on the EBPs, where no road edge management was conducted for over five decades and where the firebreaks were created e and are maintained e with minimal disturbance. The EBP firebreaks can therefore be considered the “least disturbance scenario”, with build-up roads with hard and sealed surfaces changing the hydrological regime even more, which may result in more pronounced edge effects through even larger changes in the localized hydrology. Tourism is one of the most important sources of revenue for conservation areas like the KNP (Ferreira and Harmse, 1999; Lindberg, 1991). A significant part of this tourism product is based on game viewing from private or open-top safari vehicles that utilize the extensive road network. Our results suggest that game visibility may be negatively affected by bush thickening along the road edges (Watson, 1995). Considering that roads have an artificial effect on woody cover, selective thinning of woody vegetation along the roads may therefore be justified both from an ecological and a tourism perspective. The Experimental Burn Plots (EBPs) in Kruger National Park are considered one of the largest and longest-running fire experiments
79
in Africa, and as such have become an international research area for studies of savanna fire ecology (van Wilgen et al., 2007). Our results have significant implications for ongoing and future studies on these important experimental plots. We suggest that monitoring and sampling must be conducted at least 15 m from the side of the firebreak if it is to exclude the confounding effect of changed hydrological regimes and altered woody vegetation structure. This recommended buffer distance can possibly also be extended to research and monitoring studies in other comparable semi-arid savanna systems. Finally, this study shows that the roadside edge effect is not primarily due to fire feedbacks, as was originally expected (i.e. due to back-burns and reduced herbaceous biomass resulting in cooler fires on edges of the plots), but rather it results from other effects like changed hydrological regimes. Acknowledgements We thank South African National Parks (SANParks) for permission to conduct this study within the Kruger National Park. The team that maintained the experiment for many decades deserves special mention, especially Andre Potgieter. We thank Ty KennedyBowdoin, James Jacobson, David Knapp, and the CAO team for data collection, processing and analysis. We thank Kate Parr for sparking initial interest in this topic. Shaun Levick, Renaud Mathieu, Konrad Wessels, Jan van Aardt, Barend Erasmus and Moses Cho are thanked for support and advice during the CAO campaign. Eddie Riddell, Navashni Govender, Rina Grant and two anonymous reviewers are thanked for comments and ideas that have improved the manuscript. IPJS thanks the Andrew Mellon Foundation for funding a trip to the Department of Global Ecology that contributed towards this manuscript. Airborne mapping and analysis of the KNP was funded by the Andrew Mellon Foundation. The Carnegie Airborne Observatory is made possible by the W.M. Keck Foundation and William Hearst III. References Archer, S., Schimel, D.S., Holland, E.A., 1995. Mechanisms of shrubland expansion e land-use, climate or CO2. Climatic Change 29, 91e99. Asner, G.P., Knapp, D.E., Kennedy-Bowdoin, K., Jones, M.O., Martin, R.E., 2007. Carnegie airborne observatory: inflight fusion of hyperspectral imaging and waveform light detection and ranging (wLiDAR) for three-dimensional studies of ecosystems. Journal of Applied Remote Sensing 1, 1e21. Biggs, R., Biggs, H.C., Dunne, T.T., Govender, N., Potgieter, A.L.F., 2003. Experimental burn plot trial in the Kruger National Park: history, experimental design and suggestions for data analysis. Koedoe 46, 1e15. Eversham, B.C., Telfer, M.G., 1994. Conservation value of roadside verges for stenotopic heathland Carabidae: corridors or refugia? Biodiversity and Conservation 3, 538e545. Fensham, R.J., Fairfax, R.J., 2003. Assessing woody vegetation cover change in northwest Australian savanna using aerial photography. International Journal of Wildland Fire 12, 359e367. Ferreira, S.L.A., Harmse, A.C., 1999. The social carrying capacity of the Kruger National Park, South Africa: policy and practice. Tourism Geographies 1, 325e342. Foreman, R.T.T., Alexander, L.E., 1998. Roads and their major ecological effects. Annual Review of Ecological Systems 29, 207e231. Freitag-Ronaldson, S., Foxcroft, L.C., 2003. Anthopogenic influences at the ecosystem level. In: du Toit, J.T., Rogers, K.H., Biggs, H.C. (Eds.), The Kruger Experience: Ecology and Management of Savanna Heterogeneity. Island Press, Washington, pp. 391e421. Garcia, R., Millan, E., Maiz, I., 1996. Heavy metal contamination analysis of roadsoils and grasses from Guipuzkoa (Spain). Environmental Technology 17, 763e770. Gertenbach, W.P.D., 1983. Landscapes of the Kruger National Park. Koedoe 26, 9e121. Hillerislambers, R., Rietkerk, M., Van Den Bosch, F., Prins, H.H.T., De Kroons, H., 2001. Vegetation pattern formation in semi-arid grazing systems. Ecology 82, 50e61. Kulmatiski, A., Beard, K.H., Verwij, R.J.T., February, E.C., 2010. A depth-controlled tracer technique measures vertical, horizontal and temporal patterns of water use by trees and grasses in a subtropical savanna. New Phytologist 188, 199e209. Lindberg, K., 1991. Policies for Maximizing Nature Tourism’s Ecological and Economic Benefits. World Resources Institute, Washington.
80
I.P.J. Smit, G.P. Asner / Journal of Arid Environments 80 (2012) 74e80
Pienaar, U.de V., 1968. The ecological significance of roads in a national park. Koedoe 11, 169e174. Reinirkens, P., 1996. Analysis of emissions through traffic volume in roadside soils and their effects on seepage water. The Science of the Total Environment 189/ 190, 361e369. 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 Ecology 148, 207e224. Riginos, C., 2009. Grass competition suppresses savanna tree growth across multiple demographic stages. Ecology 90, 335e340. Sankaran, M., Hanan, N.P., Scholes, R.J., Ratnam, J., Augustine, D.J., Cade, B.S., Gignoux, J., Higgins, S.I., Le Roux, X., Ludwig, F., Ardo, J., Banyikwa, F., Bronn, A., Bucini, G., Caylor, K.K., Coughenour, M.B., Diouf, A., Ekaya, W., Feral, C.J., February, E.C., Frost, P.G.H., Hiernaux, P., Hrabar, H., Metzger, K.L., Prins, H.H.T., Ringrose, S., Sea, W., Tews, J., Worden, J., Zambatis, N., 2005. Determinants of woody cover in African savannas. Nature 438, 846e849. Sidak, Z., 1967. Rectangular confidence regions for the means of multivariate normal distributions. Journal of the American Statistical Association 62, 626e633. Smit, I.P.J., Asner, G.P., Govender, N., Kennedy-Bowdoin, T., Knapp, D.E., Jacobson, J., 2010. Effects of fire on woody vegetation structure in African savanna. Ecological Applications 20, 1865e1875.
Stoeckeler, J.H., 1965. Drainage along swamp forest roads lessons from northern Europe. Journal of Forestry 63, 772e776. Trombulak, S.C., Frissell, C.A., 2000. Review of ecological effects of roads on terrestrial and aquatic communities. Conservation Biology 14, 18e30. Van der Schijff, H.P., 1958. Inleidende verslag oor veldbrandnavorsing in die Nasionale Krugerwildtuin. Koedoe 1, 60e93. Van Der Waal, C., de Kroon, H., de Boer, W.F., Heitkonig, I.M.A., Skidmore, A.K., de Knegt, H.J., van Langevelde, F., van Wieren, S.E., Grant, R.C., Page, B.R., Slotow, R., Kohi, E.M., Mwakiwa, E., Prins, H.H.T., 2009. Water and nutrients alter herbaceous competitive effects on tree seedlings in a semi-arid savanna. Journal of Ecology 97, 430e439. van Wilgen, B.W., Govender, N., Biggs, H.C., 2007. The contribution of fire research to fire management: a critical review of a long-term experiment in the Kruger National Park, South Africa. International Journal of Wildland Fire 16, 519e530. Venter, F.J., 2004. An identification and evaluation of outliers in the experimental burning plots, Kruger National Park: an objective benchmarking. Internal report. South African National Parks, Skukuza, South Africa. Wani, S.P., Rockstrom, J., Owels, T., 2009. Rainfed Agriculture: Unlocking the Potential. CABI, Wallingford. Watson, H.K., 1995. Management implications of vegetation changes in the Hluhluwe-Umfolozi Park. South African Geographical Journal 77, 77e83.
Contents lists available at SciVerse ScienceDirect
Journal of Arid Environments journal homepage: www.elsevier.com/locate/jaridenv
Roads increase woody cover under varying geological, rainfall and fire regimes in African savanna I.P.J. Smit a, *, G.P. Asner b a b
Scientific Services, South African National Parks, Private Bag X402, Skukuza 1350, South Africa Department of Global Ecology, Carnegie Institution, 260 Panama Street, Stanford, CA 94305, USA
a r t i c l e i n f o
a b s t r a c t
Article history: Received 1 June 2011 Received in revised form 21 October 2011 Accepted 29 November 2011 Available online 20 January 2012
Roads extend throughout savannas, yet few studies have quantified their effects on adjoining woody vegetation structure. Airborne LiDAR imagery collected over 168 experimental fire plots in the Kruger National Park, all bounded by graded firebreak roads, provided an opportunity to study if, and to what extent, roads influence woody vegetation structure under different rainfall, geologic and fire conditions. In 91.7% of the plots, woody canopy cover was higher on the edges of roads compared to areas farther away. The increase was most pronounced within 5 m of the road edge, but was detectable 10e15 m from the edge. On average, the area within 15 m from the road had approximately 6% and 2% higher woody vegetation cover than areas further than 15 m from the edge on wetter granitic and drier basaltic savanna landscapes, respectively. Increased edge effects on woody cover were observed even in fire exclusion plots, suggesting that non-fire processes, likely altered hydrological regimes, may be the underlying reason for woody encroachment. This study illustrates that roads cause selective woody plant thickening in savannas, even in areas without road edge management, and therefore careful consideration should be paid on how road edges are managed and when roads are planned. Ó 2011 Elsevier Ltd. All rights reserved.
Keywords: Bush thickening Hydrology Kruger National Park LiDAR Road edges Woody encroachment
1. Introduction Roads, and their associated edges, have numerous impacts on biodiversity patterns and processes. In a review, Trombulak and Frissell (2000) identified seven general negative effects of roads on biodiversity: mortality from road construction, mortality from collision with vehicles, modification of animal behaviour, alteration of the physical environment, alteration of the chemical environment, spread of exotics, and increased use of areas by humans. However, other papers have shown that road edges can also have several positive effects, especially in highly transformed landscapes, where they act as refugia and/or movement corridors (Eversham and Telfer, 1994). Foreman and Alexander (1998) proposed that road impacts operate at two primary levels, namely (i) the individual, species and population level (e.g. road kills, disturbance, etc.), and (ii) the ecosystem process and landscape level (e.g. hydrology, erosion, sedimentation, etc.). In 2003 it was estimated that there were approximately 8000 km of roads in the Kruger National Park (KNP), South Africa
* Corresponding author. Tel.: þ27 13 735 4257; fax þ27 13 735 4055. E-mail addresses: [email protected] (I.P.J. Smit), [email protected] (G.P. Asner). 0140-1963/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.jaridenv.2011.11.026
(Freitag-Ronaldson and Foxcroft, 2003). Considering the extent of roads and their potential impacts, surprisingly few “road ecology” studies have been conducted to understand potential impacts of roads in the KNP. The only published study we could find for KNP was a qualitative description of road impacts based on personal observations of a resident scientist (Pienaar, 1968). Beyond KNP, we found that studies focussing on the effects of roads on adjoining woody vegetation structure in savannas seem to be lacking from the literature. This study aims to address that gap by determining if, and to what degree, woody vegetation structure is influenced on the edges of low-maintenance graded firebreak roads under different rainfall, geological and fire management regimes (in the absence of road-edge management like scraping or mowing). Graded firebreak roads refer to linear strips that are created when bulldozers and/or scrapers denude an area of vegetation (woody and herbaceous) to leave a rudimentary road (Fig. 1). These graded areas act as firebreaks and serve as management roads for very low volume traffic, such as game ranger patrols and research vehicles. Here we addressed the following questions: Is there a difference in woody vegetation structure in areas closer to edges of roads compared to areas further from the edges? If so, how consistent is the edge effect, and how far from the edge of the road is the effect measureable? How does the edge effect differ between different geologically- and rainfall-defined landscapes? Do different fire
I.P.J. Smit, G.P. Asner / Journal of Arid Environments 80 (2012) 74e80
75
Table 1 Description of the four studied landscapes (based on van Wilgen et al. (2007)). Landscape
Vegetation type
Dominant tree species
Geology
Mean annual precipitation (mm)
Pretoriuskop
Sourveld
Granite
705
Skukuza
Combretum
Granite
572
Satara
KnobthornMarula Mopane
Terminalia sericea, Dichrostachys cinerea Combretum collinum, Combretum zeyheri Acacia nigrescens, Sclerocarya birrea Colophospermum mopane
Basalt
507
Basalt
451
Mopane
Fig. 1. Graded firebreak roads of approximately 6 m wide delineate Experimental Burn Plots (EBPs) in the Kruger National Park. Different fire management regimes are practiced on each side of the road.
management regimes lead to different edge effects on woody vegetation structure? 2. Materials and methods 2.1. Study site KNP covers an area of almost 20,000 km2 in the north-eastern corner of South Africa, bordering Zimbabwe to the north and Mozambique to the east. The Park contains a north-south rainfall gradient with rainfall increasing from the north (w350e420 mm yr1) to the south (w680e750 mm yr1). The western part of the Park is mainly underlain by granite-derived soils and the east by soils of basaltic origin, with some other geologies covering smaller sections of the Park. This abiotic template gives rise to 35 major vegetation types which are dominated by knobthorn (Acacia nigrescens), marula (Sclerocarya birrea), leadwood (Combretum imberbe), red-bush willow (Combretum apiculatum), silver cluster leaf (Terminalia sericea) and mopane (Colophospermum mopane) (Gertenbach, 1983). 2.2. Experimental Burn Plots (EBPs) The aim of the KNP Experimental Burn Plots (EBPs) is to study the effects of fire (frequency and season) on the vegetation under the grazing pressure of indigenous herbivores (Van der Schijff, 1958). The EBP experiment consisted of the application of fires at varying return intervals and seasons, and protection from fire, on a series of w7 ha plots in four of the major vegetation landscapes of the Park (Table 1). The treatments were replicated four times in each of these landscapes (replicates are called “strings”). The treatments originally included annual winter fires in August, and biennial and triennial fires in August, October, December, February and April. In 1976, further treatments to examine the effects of fires every 4 and 6 years in October were added to selected landscapes (Satara & Mopane) by subdividing the February treatment plots. This resulted in 48 plots (4 replicate strings 12 treatments) in the Skukuza and Pretoriuskop landscapes and 56 plots (4 replicate strings 14 treatments) in the Satara and Mopane landscapes. One string in each of the Skukuza, Satara and Mopane landscapes was not included in this study because local soil differences made them unrepresentative of the particular landscapes (Venter, 2004), leaving 48 plots in the Pretoriuskop landscape, 36 in the Skukuza landscape and 42 in the
Satara and Mopane landscapes. Full details and history of the experimental design and application of treatments are available from Biggs et al. (2003). 2.3. LiDAR derived top of canopy dataset Airborne Light Detection and Ranging (LiDAR) data were acquired in April 2008 using the Carnegie Airborne Observatory (CAO) (Asner et al., 2007). The CAO was mounted in an aircraft which flew over the 168 EBPs at an altitude averaging 2000 m above ground level. The CAO LiDAR was operated at 50 kHz pulse repetition frequency and a 34 scanning configuration cross-track of the aircraft direction. Global positioning system (GPS) and Inertial Motion Unit (IMU) data were combined to determine the 3-D location of the laser returns. From the laser “point cloud” data, a physical model was used to estimate top-of-canopy and ground surfaces (digital elevation models; DEM) using the Terrascan/Terramatch (Terrasolid Ltd., Jyväskylä, Finland) software package. Vegetation height was subsequently estimated by subtracting the ground DEM from the top-of-canopy layer. The CAO generated a laser spot spacing of 1.12 m. Since the beam diameter at ground level was also 1.12 m, this laser spot spacing configuration resulted in a 50% overlap between LiDAR observations, decreasing the likelihood of missing vegetation canopy. The resulting canopy DEM had a spatial resolution of 1.12 1.12 m, such that a surface was generated where each 1.24 m2 contained the estimated vegetation height in that particular pixel. The absolute vertical resolution of the CAO LiDAR is 15 cm, but following digitization and application to porous tree canopies, the effective vertical resolution is 0.2e0.5 m. All further analyses were based on these top of canopy surfaces created for each EBP, which is an estimate of woody vegetation cover. Woody vegetation cover is a widely used variable to quantify how horizontal woody structure varies across space and time (e.g. Fensham and Fairfax, 2003). Other variables like tree density can also be used to characterize woody structure, but these parameters were not directly available from the LiDAR dataset employed here. Due to the lack of late summer rains in 2008 (http://www. sanparks.org/parks/kruger/conservation/scientific/weather/) (accessed 30 May 2011), some woody vegetation already started dropping leafs in April when the LiDAR survey was conducted. This resulted in decreased leaf area, especially on the drier northern basaltic landscapes (Satara and Mopane). Lower leaf cover, together with the spatial resolution employed, resulted in some woody vegetation, especially smaller and sparse individuals, not being detected. However, as we do not compare sites over time but over space, the effect of underestimation on the emerging structural patterns was small. That is, since we are comparing edges to core areas of each EBP, the systematic underestimation would be consistent across and between plots, and therefore would not affect the overall patterns and trends (Smit et al., 2010).
76
I.P.J. Smit, G.P. Asner / Journal of Arid Environments 80 (2012) 74e80
An ANOVA followed by Sidak pairwise comparisons (Sidak, 1967) was conducted to test whether landscape (four levels; Pretoriuskop, Skukuza, Satara, Mopane) and fire treatment (14 levels; August annual, February tri-annual, fire exclusion, etc.) significantly influenced the average difference in percent woody cover between the edge and the core area of the EBPs. Finally, we tested how percent woody cover differs for different vegetation height classes between the edges and cores of the EBPs to determine how the edge effects influence the 3-D woody vegetation structure. 3. Results
Fig. 2. The Experimental Burn Plot (EBP) is depicted in light grey, the graded firebreak road surrounding the EBP is depicted in dark grey, and the 5 m distance bands from the graded firebreak road towards the centre of the EBP is depicted by the dotted lines (not to scale). Percentage woody cover was determined for each of these 5 m incremental distance bands.
2.4. Data analyses Percentage woody cover was determined in 5 m incremental distance bands from the outside of an EBP towards the centre, making use of the CAO LiDAR derived “top-of-canopy” layer described above. For example, woody cover was determined for the area 0e5 m from each graded firebreak road, 5e10 m from the road, 10e15 m from the road, etc., systematically moving away from the edge of the plot towards the centre (Fig. 2). This was repeated on four fire exclusion EBPs (plots on which fire has been excluded since 1954) in the Pretoriuskop landscape and on three fire exclusion EBPs in the three other landscapes (Skukuza, Satara and Mopane) in order to assess the extent of the edge effect in the absence of fire. Based on the extent of the edge effects demonstrated from the analysis above, each EBP was subsequently subdivided into two zones, namely (i) an edge area (area inside EBP within 15 m from firebreak road), and (ii) a core area (area inside EBP further than 15 m from firebreak road) (Fig. 3).
Areas closer to the graded firebreak roads were more densely wooded than corresponding areas closer to the centre of the EBP, with areas directly bordering the road (<5 m) being the most densely wooded (Fig. 4). This was consistently observed up to about 10 m from the firebreak, with the % woody cover levelling out at a background level between 10 and 15 m. In the Pretoriuskop landscape, the edge effect occurred up to about 15 m before levelling off between 15 and 20 m. In 154 of the 168 EBPs studied (91.7%), percent woody cover was higher in the edge area than in the core area. On the Pretoriuskop and Skukuza EBPs, the edge maintains on average 5.5e6.0% more woody cover than the paired core area of the same EBPs (Fig. 5) (Paired T-test {edge minus core}: Pretoriuskop t ¼ 6.99, df ¼ 47, p < 0.001; Skukuza t ¼ 9.60, df ¼ 35, p < 0.001). For the Satara and Mopane EBPs, the edge contains about 2% more woody cover than the paired core area (Fig. 5) (Paired T-test {edge minus core}: Satara t ¼ 7.26, df ¼ 41, p < 0.001; Mopane t ¼ 8.77, df ¼ 41, p < 0.001). An ANOVA with “edge effect” (difference in % woody cover between edge and core area) as the dependent variable, and landscape and fire treatment as independent variables (including the interaction term), indicates that the landscape is a significant explanatory variable (F ¼ 13.88; df ¼ 3; p < 0.0001), but not the fire treatment (F ¼ 0.62; df ¼ 13; p ¼ 0.62) nor the interaction term (F ¼ 0.63; df ¼ 35; p ¼ 0.63). In other words, the fire treatment does not play a statistically significant role in determining the magnitude of the edge effect. Sidak pairwise comparisons (Sidak, 1967) show that the average edge effect is significantly more pronounced in the wetter granitic Pretoriuskop and Skukuza landscapes than on the drier basaltic Satara and Mopane landscapes. Note that the average edge effect is not statistically significantly different 70 60
% woody cover
50 40 30 20 10 0 0-5
5-10
10-15
15-20
20-25
25-30
30-35
35-40
40-45
45-50
Distance from firebreak (m) Fig. 3. The striped area depicts an Experimental Burn Plot (EBP) whereas the solid area depicts the graded firebreak road surrounding the EBP. The area inside each EBP closer than 15 m from the road was termed the “edge” (light striped), whilst the remainder of the plot (i.e. the area further than 15 m from the road) was termed the “core” (dark striped). For each plot the “edge effect” was defined as the difference between the percent woody cover in the edge and the percent woody cover in the core of the same plot.
Pretoriuskop
Skukuza
Satara
Mopane
Fig. 4. Average percentage woody cover (and standard error) for fire exclusion EBPs at incremental 5 m distance bands from the graded firebreak road (n ¼ 4 for Pretoriuskop and n ¼ 3 for Skukuza, Satara and Mopane). The dotted line indicates where background levels of percent woody cover were reached (i.e. where the edge effect is no longer detectable).
I.P.J. Smit, G.P. Asner / Journal of Arid Environments 80 (2012) 74e80
(Fig. 7). For the Satara and Mopane landscapes, the edge effect is more pronounced in the lower height classes, whereas in Skukuza and Pretoriuskop, canopy cover is more enhanced in the intermediate height classes between 3.5 and 6 m. Note that this is “top-ofcanopy” cover, and hence smaller individuals growing under higher individuals will not be reflected in the analysis.
8 7 6 5 4
4. Discussion
3
Edge effects of roads on woody vegetation structure have been poorly studied in savanna systems. This study is, to our knowledge, the first to quantify how roads in different geological, rainfall and fire contexts influence woody vegetation structure in a natural semi-arid savanna system. The results presented here illustrate that graded firebreak roads can increase woody vegetation cover. Since the effects of roads on woody vegetation structure were observed even in the absence of five decades of fire (i.e. on the plots where fire has been excluded since 1954), we conclude that increased woody cover along road edges is not primarily due to changes in fire characteristics in close proximity to roads. In fact, results presented in Fig. 6 suggest that the effect of increased woody cover on road edges is often more pronounced in the absence of fire. We therefore hypothesize that these patterns of increased woody cover are primarily caused by changes in hydrological patterns and processes, and are not due to differences in fire characteristics and fire behaviour along roads (e.g. cooler fires next to roads). Road-induced hydrological changes may cause a gradient of decreasing wetness away from the roads due to:
2 1
Pretoriuskop
Skukuza
Satara
Mopane
Fig. 5. Average (and standard error) difference in % woody cover between the edge of the EBPs and the core area (n ¼ 48, 36, 42 and 42 for Pretoriuskop, Skukuza, Satara and Mopane landscapes respectively).
between Pretoriuskop and Skukuza and between the Satara and Mopane landscapes respectively. The ANOVA and pairwise comparison results are in agreement with Fig. 5. The data provide some suggestion that the edge effect may be more pronounced (although not statistically significantly e see ANOVA above) on the fire exclusion sites as compared to the fire treatment sites. For 10 of the 13 replicate strings, the fire treatments (all treatments combined) harbored weaker edge effects than observed in the fire exclusion plots (Fig. 6). We also found that percent woody cover between the edges and cores of the EBPs varies among different vegetation height classes
(% woody cover edge) minus (% woody cover core)
18 16 14 12 10 8 6 4 2
se
e
nd e
e
-T
-M
se
op
m zo
M
op
an
an op M
op M
an
bo
a -D an
e
ta r
a
-N
-S
wa
at
ne
ar
ts
i
ya he ra ta
ra ta Sa
-N
Sk
uk
uz
a
uk Sk
Sa
it s wa
sw
-S uz
a
uz
ar
ha
ku
ka
za
i ku
-N a
ha
uk Sk
-S op sk iu
or et Pr
ap
ni be
bi um -N
op sk iu or et
iu or et Pr
Pr
sk
or
op
iu
sk
-K
op
am
-F
be
ay
ni
i
0
et Pr
increased run-off from roads (i.e. similar to rainfall-harvesting and run-off farming) (Wani et al., 2009),
Sa
0
-M
(% woody cover edge) minus (% woody cover core)
77
Landscape - Replicate String Fire treatment
Fire exclusion
Fig. 6. Average (and standard error) difference in % woody cover between the edges of the EBPs and the paired core areas. The results are reported separately for the fire treatments (grey shading; n ¼ 11 for Pretoriuskop and Skukuza and n ¼ 13 for Satara and Mopane) and fire exclusion plots (black shading; n ¼ 1) respectively. Note that for 10 of the 13 replicate strings the edge effect was more pronounced for the fire exclusion compared to the fire treatment plots.
78
I.P.J. Smit, G.P. Asner / Journal of Arid Environments 80 (2012) 74e80
0.8
0.6
0.4
0.2
0
0. 51 11. 5 1. 52 22. 5 2. 53 33. 5 3. 54 44. 5 4. 55 55. 5 5. 56 66. 5 6. 57 77. 5 7. 58 88. 5 8. 59 99. 9. 5 51 0 10 -1 1 0 0.5 .5 11 11 -1 1 1 1.5 .5 12 12 -1 1 2 2.5 .5 13 13 -1 1 3 3.5 .5 14 14 -1 1 4 4.5 .5 15 15 -1 1 5 5.5 .5 16 16 -1 1 6 6.5 .5 -1 7
(% woody cover edge) m inus (% woody cover core)
1
-0.2
Height class (m) Pretoriuskop
Skukuza
Satara
Mopane
Fig. 7. Average (and standard error) difference in % woody cover between the edge area and the core area of the EBPs at 0.5 m height increments of woody vegetation cover.
roads acting as barriers for water movement (i.e. water obstructed and dammed by roads due to poor cross-drainage) (Stoeckeler, 1965), and positive feedback between plant density, water infiltration and soil moisture through spatial redistribution of run-off (Hillerislambers et al., 2001; Rietkerk et al., 2000). If roads are altering local spatial patterns in hydrological inputs, woody vegetation closer to the edges of roads will have access to elevated soil moisture for longer periods, resulting in longer growing seasons (Fig. 8). This possibility is supported by our observations of roadside woody canopy height distributions along the regional precipitation gradient (Fig. 7): Relative differences in canopy height change from taller canopies in wetter zones (Pretoriuskop, Skukuza) to smaller canopies in the drier regions of Satara and Mopane. This likely occurs because woody plants are able to take advantage of the additional moisture provided by roadside run-off in the wetter areas, whereas the difference is suppressed in the drier zones. Overall, the edge effect is most pronounced closer to roads (<5 m), but is detectable up to about 5e10 m from roads in the Skukuza, Satara and Mopane landscapes and 10e15 m in the Pretoriuskop landscape (Fig. 4). Similar to the relative effects of precipitation on height distributions, here too we postulate that the edge effect extends further from road edges in the Pretoriuskop landscape due to higher rainfall conditions, resulting in higher runoff over a larger distance. The distances reported here agree with studies that have shown that run-off can affect soils up to 25 m from the road (Garcia et al., 1996; Reinirkens, 1996). Edge effects were detected both in the relatively open as well as in well-wooded landscapes of KNP (ranging from the Satara and Mopane landscapes with <10% woody cover to the denser cover of >50% observed in the Skukuza and Pretoriuskop landscapes) (Fig. 4). This covers a wide range of woody cover estimates which typically occurs in African savannas (most of the 854 African savanna sites explored by Sankaran et al. (2005) had woody cover ranging between 0 and 60%).
The data suggest that the edge effect may be more pronounced on the fire exclusion sites than on the fire treatment sites (Fig. 6). This may be due to the fact that the fire exclusion sites, with higher vegetative biomass (Smit et al., 2010), have greater hydraulic roughness, which increases infiltration and storage of water in soils (Foreman and Alexander, 1998). The higher hydraulic roughness (i.e. more vegetation) of the fire exclusion sites reduces the lateral movement of the run-off when it moves across the surface adjoining the road, thus facilitating vertical recharge (Hillerislambers et al., 2001).
Fig. 8. Mopane shrubland in northern Kruger National Park. Compare the denser and taller mopanes (and greener e see online colour image) on the edge of the tar road with individuals further away from the road (area directly bordering the road was scraped). This photo was taken November 2010 after the first spring rains e these rains were not enough to break winter dormancy for the trees further away from the road which have no leaves, whereas the increased run-off from the road caused the trees in close proximity to flush. This demonstrates how trees on road edges are favoured by (i) having access to more water and (ii) having a longer growing season. Although the edge effect of roads is obvious in this example, it is subtle and hard to measure and detect in most of the Park where woody vegetation structure is complex and multilayered. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
I.P.J. Smit, G.P. Asner / Journal of Arid Environments 80 (2012) 74e80
Trees growing relatively close to roads but not directly in the “wet” edge, may also be advantaged because savanna trees can access water and nutrients over wider areas than just directly below their main stems. For example, Kulmatiski et al. (2010) found that trees in southern KNP can access water up to 8 m from their main stems. Moreover, studies have found that trees are more productive when competition with grass is reduced (e.g. Riginos, 2009; Van Der Waal et al., 2009). Therefore, woody vegetation on road edges may have roots underlying the (grassless) graded area thereby avoiding competition with grasses for water. These hypotheses can be tested by installing data loggers that monitor soil moisture gradients and dynamics in the vadose zone on the firebreak road and at increasing distances from the road. This will provide insights into how roads influence the soil moisture regime through space and time as a result of altered hydrological patterns. If the firebreak is often used as a road and if intense storms compact the soil, the surface may increasingly seal, reducing infiltration on the road itself and increasing run-off. Increased grazing pressure may also contribute to selective bush thickening along road edges. Many herbivores are attracted to the short green grasses which grow along the roads (Pienaar, 1968). This is particularly evident during the drier seasons when most of the grass in the Park is senesced, and greenery appears along the roads as a result from run-off from small rainfall events. Since increased grazing is one of the potential contributing mechanisms responsible for woody encroachment/thickening (Archer et al., 1995), we postulate that increased grazing along road verges may also contribute, together with the hydrological changes described above, towards increased woody vegetation along the road edges over the long-term. 5. Management implications and future studies Roads (including graded firebreaks, management roads and airstrips) covered 7926 km in the KNP in 2003 (Freitag-Ronaldson and Foxcroft, 2003). Assuming that other types of roads may have similar effects to the graded firebreaks examined in this study, the edge effect of roads on woody vegetation may affect significant areas of the Park. A woody cover edge effect of 15 m on either side of roads equates to approximately 12,000 ha throughout the Park, which is roughly 1.2% of the total Park area (this is over and above the surface area of the road itself). This suggests that roads may potentially contribute to selective bush thickening of 12,000 ha of the park through subtle edge effects. This effect needs to be considered in addition to the impacts normally considered when new roads are planned. It must also be stressed that we considered edge effects on the EBPs, where no road edge management was conducted for over five decades and where the firebreaks were created e and are maintained e with minimal disturbance. The EBP firebreaks can therefore be considered the “least disturbance scenario”, with build-up roads with hard and sealed surfaces changing the hydrological regime even more, which may result in more pronounced edge effects through even larger changes in the localized hydrology. Tourism is one of the most important sources of revenue for conservation areas like the KNP (Ferreira and Harmse, 1999; Lindberg, 1991). A significant part of this tourism product is based on game viewing from private or open-top safari vehicles that utilize the extensive road network. Our results suggest that game visibility may be negatively affected by bush thickening along the road edges (Watson, 1995). Considering that roads have an artificial effect on woody cover, selective thinning of woody vegetation along the roads may therefore be justified both from an ecological and a tourism perspective. The Experimental Burn Plots (EBPs) in Kruger National Park are considered one of the largest and longest-running fire experiments
79
in Africa, and as such have become an international research area for studies of savanna fire ecology (van Wilgen et al., 2007). Our results have significant implications for ongoing and future studies on these important experimental plots. We suggest that monitoring and sampling must be conducted at least 15 m from the side of the firebreak if it is to exclude the confounding effect of changed hydrological regimes and altered woody vegetation structure. This recommended buffer distance can possibly also be extended to research and monitoring studies in other comparable semi-arid savanna systems. Finally, this study shows that the roadside edge effect is not primarily due to fire feedbacks, as was originally expected (i.e. due to back-burns and reduced herbaceous biomass resulting in cooler fires on edges of the plots), but rather it results from other effects like changed hydrological regimes. Acknowledgements We thank South African National Parks (SANParks) for permission to conduct this study within the Kruger National Park. The team that maintained the experiment for many decades deserves special mention, especially Andre Potgieter. We thank Ty KennedyBowdoin, James Jacobson, David Knapp, and the CAO team for data collection, processing and analysis. We thank Kate Parr for sparking initial interest in this topic. Shaun Levick, Renaud Mathieu, Konrad Wessels, Jan van Aardt, Barend Erasmus and Moses Cho are thanked for support and advice during the CAO campaign. Eddie Riddell, Navashni Govender, Rina Grant and two anonymous reviewers are thanked for comments and ideas that have improved the manuscript. IPJS thanks the Andrew Mellon Foundation for funding a trip to the Department of Global Ecology that contributed towards this manuscript. Airborne mapping and analysis of the KNP was funded by the Andrew Mellon Foundation. The Carnegie Airborne Observatory is made possible by the W.M. Keck Foundation and William Hearst III. References Archer, S., Schimel, D.S., Holland, E.A., 1995. Mechanisms of shrubland expansion e land-use, climate or CO2. Climatic Change 29, 91e99. Asner, G.P., Knapp, D.E., Kennedy-Bowdoin, K., Jones, M.O., Martin, R.E., 2007. Carnegie airborne observatory: inflight fusion of hyperspectral imaging and waveform light detection and ranging (wLiDAR) for three-dimensional studies of ecosystems. Journal of Applied Remote Sensing 1, 1e21. Biggs, R., Biggs, H.C., Dunne, T.T., Govender, N., Potgieter, A.L.F., 2003. Experimental burn plot trial in the Kruger National Park: history, experimental design and suggestions for data analysis. Koedoe 46, 1e15. Eversham, B.C., Telfer, M.G., 1994. Conservation value of roadside verges for stenotopic heathland Carabidae: corridors or refugia? Biodiversity and Conservation 3, 538e545. Fensham, R.J., Fairfax, R.J., 2003. Assessing woody vegetation cover change in northwest Australian savanna using aerial photography. International Journal of Wildland Fire 12, 359e367. Ferreira, S.L.A., Harmse, A.C., 1999. The social carrying capacity of the Kruger National Park, South Africa: policy and practice. Tourism Geographies 1, 325e342. Foreman, R.T.T., Alexander, L.E., 1998. Roads and their major ecological effects. Annual Review of Ecological Systems 29, 207e231. Freitag-Ronaldson, S., Foxcroft, L.C., 2003. Anthopogenic influences at the ecosystem level. In: du Toit, J.T., Rogers, K.H., Biggs, H.C. (Eds.), The Kruger Experience: Ecology and Management of Savanna Heterogeneity. Island Press, Washington, pp. 391e421. Garcia, R., Millan, E., Maiz, I., 1996. Heavy metal contamination analysis of roadsoils and grasses from Guipuzkoa (Spain). Environmental Technology 17, 763e770. Gertenbach, W.P.D., 1983. Landscapes of the Kruger National Park. Koedoe 26, 9e121. Hillerislambers, R., Rietkerk, M., Van Den Bosch, F., Prins, H.H.T., De Kroons, H., 2001. Vegetation pattern formation in semi-arid grazing systems. Ecology 82, 50e61. Kulmatiski, A., Beard, K.H., Verwij, R.J.T., February, E.C., 2010. A depth-controlled tracer technique measures vertical, horizontal and temporal patterns of water use by trees and grasses in a subtropical savanna. New Phytologist 188, 199e209. Lindberg, K., 1991. Policies for Maximizing Nature Tourism’s Ecological and Economic Benefits. World Resources Institute, Washington.
80
I.P.J. Smit, G.P. Asner / Journal of Arid Environments 80 (2012) 74e80
Pienaar, U.de V., 1968. The ecological significance of roads in a national park. Koedoe 11, 169e174. Reinirkens, P., 1996. Analysis of emissions through traffic volume in roadside soils and their effects on seepage water. The Science of the Total Environment 189/ 190, 361e369. 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 Ecology 148, 207e224. Riginos, C., 2009. Grass competition suppresses savanna tree growth across multiple demographic stages. Ecology 90, 335e340. Sankaran, M., Hanan, N.P., Scholes, R.J., Ratnam, J., Augustine, D.J., Cade, B.S., Gignoux, J., Higgins, S.I., Le Roux, X., Ludwig, F., Ardo, J., Banyikwa, F., Bronn, A., Bucini, G., Caylor, K.K., Coughenour, M.B., Diouf, A., Ekaya, W., Feral, C.J., February, E.C., Frost, P.G.H., Hiernaux, P., Hrabar, H., Metzger, K.L., Prins, H.H.T., Ringrose, S., Sea, W., Tews, J., Worden, J., Zambatis, N., 2005. Determinants of woody cover in African savannas. Nature 438, 846e849. Sidak, Z., 1967. Rectangular confidence regions for the means of multivariate normal distributions. Journal of the American Statistical Association 62, 626e633. Smit, I.P.J., Asner, G.P., Govender, N., Kennedy-Bowdoin, T., Knapp, D.E., Jacobson, J., 2010. Effects of fire on woody vegetation structure in African savanna. Ecological Applications 20, 1865e1875.
Stoeckeler, J.H., 1965. Drainage along swamp forest roads lessons from northern Europe. Journal of Forestry 63, 772e776. Trombulak, S.C., Frissell, C.A., 2000. Review of ecological effects of roads on terrestrial and aquatic communities. Conservation Biology 14, 18e30. Van der Schijff, H.P., 1958. Inleidende verslag oor veldbrandnavorsing in die Nasionale Krugerwildtuin. Koedoe 1, 60e93. Van Der Waal, C., de Kroon, H., de Boer, W.F., Heitkonig, I.M.A., Skidmore, A.K., de Knegt, H.J., van Langevelde, F., van Wieren, S.E., Grant, R.C., Page, B.R., Slotow, R., Kohi, E.M., Mwakiwa, E., Prins, H.H.T., 2009. Water and nutrients alter herbaceous competitive effects on tree seedlings in a semi-arid savanna. Journal of Ecology 97, 430e439. van Wilgen, B.W., Govender, N., Biggs, H.C., 2007. The contribution of fire research to fire management: a critical review of a long-term experiment in the Kruger National Park, South Africa. International Journal of Wildland Fire 16, 519e530. Venter, F.J., 2004. An identification and evaluation of outliers in the experimental burning plots, Kruger National Park: an objective benchmarking. Internal report. South African National Parks, Skukuza, South Africa. Wani, S.P., Rockstrom, J., Owels, T., 2009. Rainfed Agriculture: Unlocking the Potential. CABI, Wallingford. Watson, H.K., 1995. Management implications of vegetation changes in the Hluhluwe-Umfolozi Park. South African Geographical Journal 77, 77e83.