Year : 2005 | Volume
| Issue : 1 | Page : 150-173
Quantifying Changes in Vegetation in Shrinking Grazing Areas in Africa
Randall B Boone
Natural Resource Ecology Laboratory, Colorado State University, Fort Collins, Colorado, USA
Randall B Boone
1499 Campus Delivery - B234 NESB, Natural Resource Ecology Laboratory, Colorado State University, Fort Collins, Colorado, 80523-1499
Source of Support: None, Conflict of Interest: None
|Date of Web Publication||11-Jul-2009|
| Abstract|| |
Pastoralists around the globe are being sedentarised and livestock mobility is declining. Animals once able to move about landscapes to access ephemeral green forage are being confined to small areas with fewer forage choices. The ecosystem model SAVANNA was used to quantify the effects of land subdivision and sedentarisation on vegetation traits in South Africa and Kenya. In South Africa, significant declines in high palatability green leaf biomass, annual net primary productivity, and root biomass were recorded as a 300 km 2 block of land was subdivided into parcels of 10 km 2 . In contrast, low palatability biomass measures generally increased. Woody plant populations and slow decomposing soil organic matter increased significantly, whereas surface litter declined. In southern Kajiado District, Kenya, group ranches in which livestock populations declined under subdivision showed increases in herbaceous biomass, whereas the ranch where livestock populations did not change under subdivision had less herbaceous biomass. Livestock within small parcels were food stressed in the dry season and their populations declined so that vegetation increased beyond what could be eaten in the wet season. The vegetation changes modelled led to, or reflected, significant declines in livestock. The results suggest that stakeholders should retain open access to subdivided lands to reduce loss of vegetation productivity.
Keywords: landscape heterogeneity; SAVANNA ecosystem model, livestock movement.
|How to cite this article:|
Boone RB. Quantifying Changes in Vegetation in Shrinking Grazing Areas in Africa. Conservat Soc 2005;3:150-73
| Introduction|| |
INTACT GRAZING AREAS may be viewed as a series of interconnected landscape patches that produce forage ephemerally. Patchiness can be due to effects such as differences in slope and aspect, soil nutrients and depth, water availability and absorption, and plant dispersal and colonisation events. Grazing itself can promote landscape patchiness as patches are used differentially (e.g., Adler et al. 2001; Augustine and Frank 2001). Nomadic and transhumant pastoralists and their livestock use mobility to access patches as the vegetation pulses in its usefulness to herbivores (Pickup and Stafford Smith 1993). Livestock in a patch that may have unsuitable grass may move to patches that are more suitable at the time. If many patches are unsuitable, livestock may be moved to key resource areas (Scoones 1991, 1993), such as wetlands or high elevation grasslands, which produce forage when other sites do not.
The mobility of pastoralists and their livestock is being reduced around the globe. Traditional long-range movements made in response to extreme climatic effects are no longer options for many pastoralists. Privatisation of lands, gazetting of conservation lands, and increasing transport expenses are limiting long-range movements (FAO 2001a). More generally, sedentarisation has been a specific goal of policy reforms, a secondary outcome of governmental administration or neglect, or a philanthropic goal of non-governmental organisations to ease the provisions of services (Niamir-Fuller and Turner 1999). For example, Kajiado District, Kenya, one of the study areas, once included large sections of unfragmented land (there were 8 averaging 2731 km 2 each; Ole Katampoi et al. 1990) that were used by Maasai who made regular, medium range movements between seasons to access green forage. Lands were used communally, and movements were subject to complex use rights (Galaty 1980). In the late 1960s and 1970s, the Kenyan government cooperated with the World Bank to divide the Maasai sections into what are now 52 group ranches (Galaty 1980; Kimani and Pickard 1998). Group ranches were formed to secure communal land title for Maasai, ease the provision of government services, and instil self-interest in intensifying livestock production. Most of the goals of group ranch formation have not been achieved, which has been written about extensively (e.g., Rutten 1992; Galaty 1994; Heath 2000). Early in the history of group ranch formation, some ranches included parcels owned by individual Maasai, a process that was sanctioned by the government in 1983 (Kristjanson et al. 2002). Subdivision continues today, with ranches nearest to Nairobi likely to be divided, and those most arid least likely to be divided. There have been many social and institutional effects of subdivision in Kajiado (e.g., Galaty 1980; Bekure et al. 1991; Rutten 1992; Galaty 1994; Kristjanson et al. 2002). Here I focus upon changes in ecosystem services, principally changes in vegetation production and traits.
Theory suggests (Ritchie and Olff 1999) and modelling demonstrates (Boone and Hobbs 2004; Boone et al. in press) that declines in livestock occur as parcels available for grazing are reduced in area, even though the total area available for grazing remains the same. In intact landscapes, livestock in an unsuitable patch may move elsewhere to access forage. In subdivided landscapes, where parcels are used exclusively by owners, livestock in an unsuitable patch must make do. Theorised declines agree with the experiences of pastoralists, who recognise that parcels that are used exclusively may become too small to support a viable livestock herd. Straightforward assessments of the effects of parcel isolation on vegetation responses will be complicated because of concurrent changes in livestock populations. Would changes be due to differences in animal mobility under sedentarised conditions, or differences in stocking rate due to livestock population declines? It may not be possible to isolate the effects completely, but I have taken steps to focus upon the effects that changes in mobility of animals have on vegetation. A generalised model application from South Africa was used with livestock populations held constant. That simplified application was somewhat removed from reality, but allowed for carefully paired simulations that isolated effects of landscape fragmentation (i.e., paired simulations with a parcel part of the intact landscape, and as an isolated parcel). The second application was for Kajiado District, Kenya. That application sacrificed some rigour in pairing simulations for a more realistic application. In this paper, I quantify the changes in vegetation associated with patch isolation, exploring effects of reduced mobility of livestock and the sources of livestock declines, plus the effects of more constant grazing on landscape patches heretofore adapted to resting periods as livestock move elsewhere.
| Methods|| |
Kajiado [Figure 1] is a semi-arid district of south-western Kenya (36° 0' E to 37° 55' E, 1° 1'S to 3° 3' S). The area modelled was the southern half of the district, which is 10,746 km 2 , and includes Amboseli National Park. Elevation ranges from 790 m along the south-eastern border of the study area, to 2159 m within the Chyulu Hills [Figure 1]. Ash and pumice soils dominate the area, with brown calcareous clay loams also important. Rainfall ranges from 400 to 800 mm annually, is bimodal, and is variable through time (e.g., coefficient of variation of 27% from 1969 to 1998). Plant communities include grasslands dominated by red oat grass (Themeda triandra) and wooded grasslands and bushlands dominated by acacia shrubs and trees, such as Acacia drepanolobium. Wildlife congregate within Amboseli National Park during the dry season, and disperse into surrounding group ranches to graze during the wet season. Large herbivores are stocked at about 4.7 ha / large herbivore unit (250 kg body mass, using Coe et al. 1976 for masses), with livestock comprising 80% of large herbivore biomass. For both wildlife and livestock, swamps and other places with permanent water act as key resources in the dry season. In 2002, about 52,000 Maasai inhabited the study area, with high human population growth and in-migration by non-Maasai (reviewed in Thornton et al. in press). Resident Maasai employ a combination of activities as livelihoods, but the foundation of the economy remains livestock raising, with about 440,000 livestock in the study area. These analyses are based on three group ranches within the district, Eselenkei, Osilalei, and Olgulului/Lolarashi, hereafter referred to as Olgulului [Figure 1]. Osilalei Group Ranch was subdivided in 1981, whereas the other two group ranches are as yet not subdivided.
The Vryburg area of North West Province in South Africa has an annual rainfall averaging 500 mm in the area modelled, but rainfall is variable across space and time (e.g., the temporal coefficient of variation was 30% from 1900 to 1995) and declines in a gradient toward the northwest and the Kalahari Desert. The elevation is ca. 1200 m, and is classed as the Kalahari thornveld and shrub brushveld (Acocks 1975; NBI 1996). The area studied contains parcels where livestock are produced commercially, with a recommended stocking rate of 7 ha per large stock unit (Department of Agriculture 1999). Boone et al. (2004) and Thornton et al. (2004) used an ecosystem model application to assess the utility of climate forecasts to livestock producers in this region. The parameters from that model were used to create a simplified, generic model for the region. This generic model allowed me to use custom subdivision maps, limit model complexity to make interpretation straightforward, and model effects of fragmentation in a southern African ecosystem, without being bound by regional ownership or subdivision patterns.
SAVANNA Modelling System
Initial development of the SAVANNA Modelling System began in the early 1980s in the Turkana District of northern Kenya (Coughenour 1985), and since then, there have been numerous applications and improvements around the globe (e.g., Coughenour 1992; Buckley et a]. 1993; Boone et al. 2002; Boone et al. 2004; reviewed in Ellis and Coughenour 1998). SAVANNA is a model of semi-arid and arid lands used to model landscapes, as opposed to points or small areas. Also, some ecosystem models are rule-based, where, for example, a drought of a given severity will lead to a given proportion of plant death. SAVANNA is process-based, where a drought of a given severity will decrease soil moisture, increase competition for water between plants, and lead to a quantity of plants dying from water stress. The outcome from rule-based and process-based models may be similar, but the opportunity for novel results is greater in process-based SAVANNA.
SAVANNA is a series of interconnected FORTRAN computer programmes that model primary ecosystem interactions, simulating plant and animal functional groups, such as palatable grasses, acacia shrubs, or cattle. SAVANNA represents landscapes by dividing them into a grid of square cells. Maps containing information about these cells are processed by SAVANNA, and files that describe weather history, soils, and plant and animal traits are used to model ecosystem function. SAVANNA predicts water, nitrogen, light, and space availability for plants within each cell. Based upon competition, quantities of carbohydrates are photosynthesised. Photosynthetic products are allocated to leaves, stems, and roots using plant allometrics and priorities unique to herbaceous and woody functional groups. These allocations yield estimates such as primary production, seed production, and wood accumulation. Changes in plant populations are calculated based on primary production, and are related to factors such as seed reproduction and germination, vegetative reproduction rates, and sources of mortality, including herbivory. At each weekly time-step, for example, plants may produce seeds that become established, grow into older age classes, out-compete other plant functional groups and disperse, or die. Animals feed upon the available vegetation, depending upon defined dietary preferences and consumption rates. The energy they gain from herbivory is reduced by energy expenditures, including basal metabolism, but also gestation, lactation, and travel costs. Net energy remaining causes weight changes, with weights compared to typical body masses and deviations reflected in condition indices. A habitat suitability index is calculated for each of the cells in the landscape at every weekly time-step and for every animal functional group. Habitat suitability is based upon forage quantity and quality, slope, elevation, the presence of thickets and shade, distance to water, the density of green and dead herbaceous material, and the density of herbivores. Individuals in the populations are distributed on the landscape based upon these habitat suitability indices. SAVANNA produces monthly charts and maps that summarise ecosystem states, over simulations spanning about 10 to 100 years. For more detail about SAVANNA, see Ellis and Coughenour (1998) and Boone (2000).
Confining livestock to parcels may yield two main effects, a decline in the number of livestock for the region, and an increase in grazing pressure or duration within the parcel. Isolating these responses is problematic and complete isolation may not be possible. I used two applications to help isolate the effects: one, a simplified application from South Africa where livestock densities were kept constant, and the second, a more realistic application from Kajiado District, Kenya, where livestock densities were allowed to vary.
A Generic Adaptation of SAVANNA to South Africa
In the original SAVANNA (using version 4L) application to the Vryburg area (Boone et al. 2004; Thornton et al. 2004) and in the research here, seven plant functional groups were used: 1-3) high, moderate, and low palatability grasses, 4) annual grasses, 5) acacia shrubs (e.g., Acacia mellifera), 6) camphorbush shrubs (Tarchonanthus camphoratus), and 7) acacia trees (e.g., A. tortilis). Palatability was assigned to grass functional groups based upon their use by livestock. Five animal groups were defined in the original model: cattle, goats, sheep, horses, and donkeys. Animals in SAVANNA, as in reality, can exhibit complex compensatory changes in populations. To simplify interpretation, I removed four of the species, so that only the dominant livestock species, cattle, remained.
Spatial data used for the 20 x 15 km (300 km 2 ) block included elevation, slope, and aspect, produced at 1 km 2 resolution by the US Geological Survey and available from the African Data Dissemination Service (ADDS 2001). The 300 km 2 block was located within a single vegetation type (Low and Robelo 1996), the Kalahari Plains Thorn Bushveld. Soil types were provided by the Department of Agriculture. Distance to water maps are typically used in SAVANNA as part of measures of habitat suitability for herbivores. However, water maps were not used in the generic application so that whether or not a parcel contained a water source would not complicate interpretation of results. Spatial data were generalised to 1 x 1 km grids. Weather data from 1990 to 1995 was provided by the South African Weather Service for 166 weather stations (see Boone et al. 2004). From these weather data, 12 unique weather patterns were generated by selecting rainfall and temperatures randomly, by year (see Boone and Hobbs 2004 for more detail).
Parameters were set in the model based on an extensive literature review, previous SAVANNA applications (e.g., Coughenour 1992; Kiker 1998; Boone et al. 2002), field work conducted specifically for the purpose, and expert opinion. Listing the numerous parameters within SAVANNA is beyond the scope of this paper, but they may be grouped into ecological or sociological processes. Digital Appendix 1 gives example sources from the literature used to set parameters within these groups. Taking a large stock unit to be equal to one animal, the 300 km 2 used in analyses [Figure 2] would support 4286 cattle according to the recommendations of the Department of Agriculture (1999). Parameters were adjusted in SAVANNA until the modelled population averaged about 4286 cattle over a 30-year simulation. The population varied by more than 50%, however, as seen in the region during droughts and wet years.
The generic application is not intended to represent a specific place, but rather a simplified ecosystem with the flexibility to accommodate a variety of experiments. However, the structure and suitability of SAVANNA has been demonstrated many times (e.g., Coughenour 1992; Kiker 1998; Boone et al. 2002; Boone et al. 2004), and the usefulness of its component algorithms was confirmed by the original authors of the algorithms. Also, the suitability of the SAVANNA application from which the generic application was derived was assessed by comparing livestock population changes during drought to those reported by commercial and communal livestock raisers in the study area (Hudson 2002). The total green leaf biomass modelled was compared to observed greenness measured from satellites (see Boone et al. 2004), with agreement (Spearman's p = 0.69, df = 156 months) the same as that between measured rainfall and observed greenness (p = 0.69).
An Adaptation of SAVANNA to Southern Kajiado District, Kenya
In the southern Kajiado District application, SAVANNA (version 4Lc) included seven plant functional groups: palatable grasses, palatable herbs, unpalatable herbaceous vegetation, swamp vegetation, palatable shrubs, unpalatable shrubs, and deciduous woodlands. Palatable grasses include many species but are typified by red oat grass (Themeda triandra), palatable shrubs include Acacia and Commiphora spp., and deciduous woodlands are mostly acacias. Nine animal functional groups included six wildlife groups: wildebeest (Connochaetes taurinus), zebra (Equus burchellii), African buffalo (Syncerus caffer), grazing antelope, browsing antelope, and elephants (Loxodonta africana), and three livestock groups: cattle, goats, and sheep. Grazing antelope included Grant's gazelle (Gazella granti), Thomson's gazelle (Gazella thomsoni), impala (Aepyceros melampus), kongoni (Alcelaphus bucelaphus), oryx (Oryx gazella), and waterbuck (Kobus ellipsiprymnus). Browsing antelope included eland (Taurotragus oryx), greater kudu (Tragelaphus strepsiceros), lesser kudu (Tragelaphus imberbis), and gerenuk (Litocranius walleri).
Like in the generic application, elevation, slope, and aspect were derived from 1 km 2 data from the ADDS (2001) for Kajiado. A land cover map was modified from the vector-based AfriCover map by FAO (2001b), which contained 18 land cover types. In SAVANNA, the land cover map was joined with woody plant cover and height maps generated from aerial surveys (see De Leeuw et al. 1998 for details) and other parameters to initialise vegetation attributes throughout the study area. Soils were from the Kenya Soil Survey (1995) KenSOTER database, augmented by more detailed maps from the same source. Water sources were edited versions of point data provided by the Ministry of Agriculture. Edits included removing boreholes known to not function, and the addition of pipelines, permanent rivers, and new dams and boreholes. Distinct distance to water maps were created for wet, transitional, and dry seasons for livestock and wildlife groups. Maps, called force maps, are used in SAVANNA to limit the distributions of animals in ways that are not associated with ecological relationships. Force maps were created to limit the distribution of elephants to be within ca. 35 km of Amboseli National Park (Ole Katampoi et al. 1990). Other limitations were minor, making livestock less likely to enter fenced swamps outside the park and unable to graze within the park, and preventing elephants from entering fenced forested patches within the park. The geographic data used in the analyses reported here were generalised to 1 x 1 km in resolution. Weather data came from 47 stations within Kajiado, from 1969 to 1998, but the data were sparse. Weather data were augmented using rainfall estimates derived from satellites for 1998 to 2002 (Xie and Arkin 1997) available from ADDS (2001) at 8 x 8 km resolution. Surface maps showing average rainfall from the satellite estimates were used as covariates in smoothing algorithms within SAVANNA, in rainfall interpolation. Temperature data were not available beyond 1985 for the stations, so data from 1969 to 1985 were used for later years. The variation in temperature in the region is exceedingly small (i.e., standard deviation of monthly minimum temperature, 0.15° C, maximum temperature, 0.18° C), so this substitution should have minimal impact on results.
As for the generic SAVANNA application, parameters were set in the model using existing literature, especially Bekure et al. (1991), previous applications, field work (Galvin et al. 2000; Mworia and Kinyamario 2000; BurnSilver et al. 2003), and expert opinion, as well as another ecosystem model (Toxopeus et al. 1994). Digital Appendix 1 provides examples of literature sources used to set parameters. Wildlife and livestock populations were set using 12 aerial surveys conducted by the Kenya Department of Resource Surveys and Remote Sensing between 1980 and 2000 (see De Leeuw et al. 1998 for details). Elephant populations came from the Savanna Elephant Vocalisation Project (SEVP 2002). See Boone et al. (in press) for more detail on sources for setting animal populations.
SAVANNA parameters were calibrated until responses were similar to those in the literature (e.g., Bekure et al. 1991) and available data (e.g., Kinyamario 1996; Mworia and Kinyamario 2000; Department of Resource Surveys and Remote Sensing aerial survey data). Similar to the application to South Africa, an assessment of biomass production and plant phenology was made by comparing modelled total leaf greenness and actual greenness observed from space as shown in normalised difference vegetation indices. Similar to the South African model, the agreement between modelled greenness and observed greenness (Spearman's p = 0.65, df = 180) was greater than the agreement between measured rainfall and observed greenness (p = 0.61).
Comparisons must be made in the light of changes in animal populations across time, which are related to weather patterns that were determined randomly. A series of comparisons were therefore made, exploring effects on vegetation while keeping livestock populations comparable. [Figure 2] shows a series of parcel maps for a 300 km 2 block of land that were used in analyses. Changes in vegetation were quantified as parcel areas declined from 300 km 2 to 10 km 2 , an area sufficiently small to be of interest, but that contained enough 1 km 2 modelling cells (10) for an ample sample. Two steps were used to conduct simulations using the generic SAVANNA application. Thirty-year simulations were conducted, one for each parcel area for each of 12 randomised weather patterns [Figure 2], e.g., 1 simulation at 300 km 2 , a simulation for both 150 km 2 parcels, etc., summing to 76 simulations, and with each repeated 12 times). In each simulation, cattle could move throughout the 300 kmz block, and the number of cattle that inhabited each parcel each month was calculated. These populations reflected specific differences in the suitability of parcels and weather patterns as animals moved about seeking forage. A second series of simulations was conducted using these parcel-and-weather specific population trends throughout each 30-year run, but with animals confined to parcels. Summarised changes in vegetation during the last 15 years of simulations are presented, comparing responses with equal livestock populations but with livestock able to move throughout the 300 kmz block or confined to a parcel.
Results from simulations using the generic application quantified changes in vegetation as livestock were confined to parcels, but their populations were dictated to remain what they were when the animals were fully mobile (Boone and Hobbs 2004; Boone et al. in press). In simulations in southern Kajiado District, realismrather than flexibility-was emphasised. For example, livestock populations were allowed to change as mobility declined, and the observed weather pattern was used; the only difference between simulations was the area available to herbivores. Simulations spanned 24 years, with results from the last 15 years summarised, allowing nine years for herbivores to reach some balance with forage availability. A single simulation was done for intact Eselenkei, Olgulului, and Osilalei Group Ranches. Herbivore populations were then set for different sized parcels (10, 5, 3, and 1 kmz) based on area, so that for a given ranch the average starting densities remained the same across simulations. These subdivisions represented incrementally smaller subdivided parcels, down to the resolution of a modelling cell, which was 1 x 1 km. Under subdivision, Maasai may expect to receive somewhat less than 1 kmz of land (100 ha) for their use, with parcels being between 24 and 40 ha (60 to 100 ac). Twenty simulations were conducted for each parcel area, with the locations of parcels placed randomly within the ranches. In all simulations, wildlife populations were kept constant, or varying in a cyclical manner for Olgulului, where animals will move into the ranch from Amboseli National Park in each wet season. This simplified interpretation of model results. For reference, based on biomass (using Coe et al. 1976), livestock were 92% of large herbivores in Osilalei Group Ranch, 86% in Eselenkei Group Ranch, and 67% in Olgulului.
Analysis and Display
Analyses using the generic application were paired well, meaning that the only difference between control and simulation experiments was the mobility of cattle confined to parcels, or allowed to move about the 300 kmz block. The differences in ecosystem metrics were typically small at any given period, and changes in the paired simulation experiments yielded increased values in some months and decreased values in other months. The method used to summarise changes is shown in [Figure 3]. Because of the close pairing of simulations, for a given variable, I subtracted the response (such as standing biomass) under fragmentation from the response when the parcel was intact. The difference in biomass was integrated (i.e., monthly values summed) across the simulation period, yielding in the example a decline across the entire simulation period of 807 g m -2 , the result plotted. Whether or not results for parcels were different from the entire 300 km 2 block were tested using one-sampled Student's t-tests.
| Results|| |
Declines occurred in herbaceous biomass in the South African application, as the 300 km 2 block was subdivided into smaller and smaller parcels. Above-ground green leaf and standing dead herbaceous biomass [Figure 4a] declined significantly compared to the biomass when the block was intact. Herbaceous stem biomass showed a similar pattern (not shown), and root biomass responses were mixed. Annual net primary productivity (NPP), both above ground and below ground, progressively declined [Figure 4b] as parcel areas shrank and animals were forced to graze within patches year-round.
Changes in herbaceous above-ground biomass were not uniform across plant functional groups. High palatability grasses declined most rapidly, with small declines in moderate palatability grasses, and a small increasing trend in low palatability grasses [Figure 5a]. The decline in herbaceous biomass allowed woody plant biomass to increase, reflected in a modest increase in acacia shrub leaf biomass [Figure 5a]. Herbaceous root biomass exhibited a similar pattern to leaf biomass, with roots of high palatability grasses declining, and those of moderate and low palatability grasses increasing [Figure 5b]. Tree and shrub populations increased as herbaceous biomass declined and germination of woody seeds improved [Figure 6a]. Nitrogen content showed complex responses, confounded by changes in nutrient content such as those related to plant phenology and changes in plant density. High palatability grasses declined in nitrogen mass [Figure 6b], but increased somewhat at the smallest parcel areas, associated with an increasing ratio of nitrogen-rich seeds and young plants versus mature plants. Surface litter declined dramatically as parcel areas declined [Figure 6c], as cattle confined to the parcel throughout the year ate more standing dead herbaceous biomass in the dry season, before it could fall as surface litter. As a contrasting example, the pool of slow-to-decompose soil organic matter increased slightly, associated with the increase in woody plants and greater contributions of wood to the soil. Overall, soil organic matter declined slightly.
In simulations for southern Kajiado District group ranches, the balance of palatable grasses, palatable herbs, and unpalatable herbaceous vegetation was idiosyncratic. For larger parcels (e.g., 10 km 2 ), the proportions of palatable grass, herbs, and unpalatable herbaceous vegetation closely mirrored the proportions for the entire group ranch. But as expected, some small parcels (1 km 2 ) came to be dominated by one of any of the three herbaceous groups. Changes in vegetation within Kajiado were less dramatic than in the generic application because livestock were allowed to decline in response to subdivision, rather than being held artificially high. Also, in the Kajiado analyses, parcels were randomly distributed throughout the ranches and were compared to overall group ranch means, rather than closely paired as in generic application. The standard errors around the mean estimates are therefore broader than for the generic application, and I emphasise trends in mean responses rather than statistically significant changes in modelled results.
In Eselenkei and Olgulului Group Ranches, livestock populations declined by 25% and 20% respectively in 1 km 2 parcels, relative to the intact group ranch (Boone et al. in press). Herbaceous vegetation biomass increased in these group ranches [Figure 7a], but decreased in Osilalei Group Ranch, which did not show a decline in livestock as parcel area declined (Boone et al. in press). The pattern of increases for Eselenkei and Olgulului Group Ranches and a decline for Osilalei Group Ranch occurred for other biomass measures, such as herbaceous green leaf biomass [Figure 7b]. However, nitrogen mass increased for all group ranches [Figure 7c], reflecting a slight shift in plant phenologies to younger plants, associated with increased grazing pressure.
When the land was subdivided into 1 km 2 parcels and livestock were much less mobile, herbaceous green leaf biomass declined in the dry season to such a degree that some livestock could not survive [igure 8]. With fewer livestock surviving the dry season, leaf biomass in the wet season was high, relative to when animals could move about the ranch under pre-subdivided conditions. In contrast, herbaceous green leaf biomass in Osilalei remained relatively high in the dry season [Figure 8], and livestock survived to forage in the wet season and decrease standing biomass, as reflected in [Figure 7a]. The decline in biomass in Eselenkei Group Ranch, for example, was not solely due to normal seasonal drying, but because livestock were confined to 1 km 2 parcels and eating dried vegetation they otherwise would not. Herbaceous green leaf biomass declined in the dry season beyond what was seen when Eselenkei Group Ranch was intact [Figure 9]. [Figure 10] illustrates the livestock on a single 1 km 2 parcel, with the population more finely tuned to the conditions on the parcel when the ranch is intact (solid line), as animals move into and out of the parcel. In contrast, livestock confined within the parcel (dotted line) cannot access other sources of forage, resulting in lower animal condition and higher death rates.
| Discussion|| |
Sedentarisation of nomadic people is "the greatest single transformation of pastoralism as both a production system and a way of life." (Lane and Moorehead 1995:123) But although sedentarisation was once a laudable goal and remains a frequent outcome of policy, it is now discouraged. The importance of livestock mobility is one of the few hard-earned lessons of pastoral anthropology (Sandford 1995). Dramatic examples of improved survival in herds that moved relative to sedentary herds have been reported (e.g., Scoones 1992; Kavoori 1999). Large declines in livestock populations under sedentarisation can occur (Boone and Hobbs 2004; Boone et al. in press), and this paper documents changes in vegetation associated with reduced mobility that bode poorly for pastoral food security, especially in landscapes with insufficient productivity to support livestock in the dry season.
It is counter-intuitive for average herbaceous biomass to increase in areas where livestock decline under subdivision - livestock populations should increase to make use of the extra forage. However, the decline is explained when primary productivity and temporal variation are considered. Eselenkei and Olgulului Group Ranches are generally less productive than Osilalei Group Ranch. In intact Eselenkei and Olgulului, livestock that inhabited landscape patches that were unsuitable moved elsewhere within the ranch to locate forage. Grazing intensification related to subdivision can decrease standing herbaceous biomass (live and dead) in the dry season [Figure 9]. But the resulting decline in livestock populations allows herbaceous vegetation to rebound to a greater density than when the animals were able to move about the intact group ranches. Grazing areas may therefore appear to be in better condition under subdivision at some times of the year, without quantitative comparisons. The condition of grazing lands must be assessed over time, and past and current stocking rates across the region compared to identify long-term changes in vegetation traits.
This paper focuses upon changes in vegetation production and other traits associated with subdividing lands used by livestock, but reduced access to land will impact wild herbivores as well (Western and Gichohi 1993). Also, the generic and southern Kajiado study areas include livestock populations that appear to influence, and be influenced by, vegetation (see Vetter 2004, and citations therein). Areas exhibiting non-equilibrium dynamics and with poor links between grazing and vegetation typically have coefficients of variation in excess of 30% and less rainfall than occurs in these sites (Ellis 1995). Analyses are planned to explore primary and secondary productivity for sites with less rainfall and higher coefficients of variation. These analyses focused upon comparisons between simulations where the only variable changed between runs was livestock mobility. The applications were parameterised to agree with observations to the degree possible, but the approach is not predicated on absolute predictive ability, but rather on comparisons between simulations parameterised identically. Parcels were used exclusively by owners in these analyses. This approximates what is seen in reality in group ranches long-subdivided, such as in the Kitengela region of Kajiado District, which has many fenced parcels (Kristjanson et al. 2002). Further, in subdivided Osilalei Group Ranch, herders moved their animals shorter distances than in group ranches that were not subdivided (BurnSilver et al. 2003), although they were able to move to Imbirikani Group Ranch during the drought of 2000.
Maasai call themselves 'people of the cattle,' and changes in vegetation within Kajiado District will be felt by that society through changes in livestock populations. A 25% decline in livestock populations under subdivision, as seen in modelling for Eselenkei Group Ranch (Boone et al. in press) would cause great stress for Maasai herders. In addition, in these analyses and in Boone et al. (in press), we used a 1 km 2 (100 ha) parcel as the smallest subdivision, but Maasai in southern Kajiado ranches where subdivision is being considered may expect to receive parcels only 24 to 40 ha (60 to 100 acres), so population declines and vegetation changes may be more extreme. Related analyses used a pastoral household decision-making model tightly linked to SAVANNA (Thornton et al. in press). In those analyses, even modest subdivision and declines in livestock production led to food deficits that caused families to sell some animals to purchase grains, which in turn decreased livestock production, which led to more animal sales, etc. Such downward spirals in Maasai economies have been documented (Rutten 1992). Lastly, human population growth, both endogenous and in-migration, is greater than 3% per year in Kajiado (reviewed in Thornton et al. in press). Stable or declining primary and secondary production but a dramatically increasing human population often leads to increased food insecurity.
Subdivision may be an inevitable outcome of existing policy priorities, land use intensification, and the socio-economic pressures pastoralists face (Kimani and Pickard 1998; BurnSilver et al. 2003), but governmental and non-governmental organisations may need to subsidise sedentarised pastoral communities to offset losses in ecosystem services, such as primary and secondary production as demonstrated (Boone and Hobbs 2004). Subdivision can impart some benefits upon community members, such as title to parcels of land that may be used as collateral for loans (Kimani and Pickard 1998), but care must be used in subdivision planning. In general, modelling results suggest that land owners and policy makers should maintain open or flexible access to individually owned parcels or share access to parcels with other herders, so that benefits of subdivision are enjoyed and ecosystem services are maintained over time.
Michael Coughenour developed the SAVANNA Ecosystem Model, which made these analyses possible. Joyce Acen provided digitised maps of aerial survey data for Kajiado District, Kenya. I thank the pastoralists of Kajiado and commercial livestock producers of the North-West Province, South Africa for their participation in workshops and the information they provided. Reviews by Shauna BurnSilver, Richard Conant, two anonymous reviewers, and the Editor allowed me to improve the manuscript. This work was supported by the US National Science Foundation under Grant No. DEB-0119618.
| References|| |
|1.||Acocks, J.P.H. 1975.Veld types of South Africa. Memoirs of the Botanical Survey of South Africa 40:1-128. |
|2.||ADDS. 2001. African Data Dissemination Service. US Geological Survey/EROS Data Center, Sioux Falls, South Dakota, USA. URL: adds/ |
|3.||Adler, P.B., D.A. Raff, and W.K. Lauenroth. 2001. The effect of grazing on the spatial heterogeneity of vegetation. Oecologia 128:465-479. |
|4.||Augustine, D.J. and D.A. Frank. 2001. Effects on migratory grazers on spatial heterogeneity of soil nitrogen properties in a grassland ecosystem. Ecology 82:3149-3162. |
|5.||Bekure, S., P.N. de Leeuw, B.E. Grandin, and P.J.H. Neate (eds.). 1991. Maasai Herding: Analysis of the Livestock Production System of Maasai Pastoralists in Eastern Kajiado District, Kenya. ILCA Systems Study 4, ILCA (International Livestock Centre for Africa), Addis Ababa, Ethiopia. |
|6.||Boone, R.B. 2000. Integrated Management and Assessment System: Balancing Food Security, Conservation, and Ecosystem Integrity. A Training Manual. Global Livestock Collaborative Research Support Program, University of California, Davis, California. URL: prods/rboone/Workshop.html |
|7.||Boone, R.B. and N.T. Hobbs. 2004. Lines around fragments: Effects of fragmentation on large herbivores. African Journal of Range and Forage Science 21:79-90. |
|8.||Boone, R.B., S.B. BurnSilver, P.K. Thornton, J.S. Worden, and K.A. Galvin. (in press). Quantifying declines in livestock due to land subdivision in Kajiado District, Kenya. Rangeland Ecology and Management. |
|9.|| Boone, R.B., M.B. Coughenour, K.A. Galvin, and J.A. Ellis. 2002. Addressing management questions for Ngorongoro Conservation Area, Tanzania, using the SAVANNA modelling system. African Journal of Ecology 40:138-50. |
|10.||Boone, R.B., K.A. Galvin, M.B. Coughenour, J.W. Hudson, P.J. Weisberg, C.H. Vogel, and J.E. Ellis. 2004. Ecosystem modelling adds value to a South African climate forecast. Climatic Change 64:317-40. |
|11.||Buckley, D.J., M.B. Coughenour, C.B. Blyth, D.J. O'Leary and J.A. Bentz. 1993. Ecosystem management model - Elk Island National Park: A case study of integrating environmental models with GIS. Proceedings of the Second International Conference on Integrating GIS and Environmental Modelling, National Center for Geographic Information Analysis, Santa Barbara, California. |
|12.||BurnSilver, S., R.B. Boone, and K.A. Galvin. 2003. Linking Pastoralists to a Heterogeneous Landscape: The Case of Four Maasai Group Ranches in Kajiado District, Kenya. In: Linking Household and Remotely Sensed Data: Methodological and Practical Problems (eds. J. Fox, V. Mishra, R. Rindfuss, and S. Walsh), pp.173-199. Kluwer Academic Publishing, Boston. |
|13.||Coe, M.J., D.H. Cumming, and J. Phillipson. 1976. Biomass and production of large African herbivores in relation to rainfall and primary production. Oecologia 22:341-354. |
|14.||Coughenour, M.B.1985. Graminoid responses to grazing by large herbivores: Adaptations, exaptations, and interacting processes. Annals of the Missouri Botanical Gardens 72:852-863. |
|15.||Coughenour, M.B. 1992. Spatial Modelling and Landscape Characterisation of an African Pastoral Ecosystem: A Prototype Model and its Potential Use for Monitoring Drought. In: Ecological Indicators (eds. D.H. McKenzie, D.E. Hyatt, and V .J. McDonald), pp. 787-8 10. ElsevierApplied Science, New York. |
|16.||De Leeuw, J., H.H.T. Prins, E.C. Njuguna, M.Y. Said, and R.A. de By. 1998. Interpretation of DRSRS Animal Counts (1977-1997) in the Rangeland Districts of Kenya. International Institute for Aerospace Survey and Earth Sciences, Enschede, The Netherlands and The Department of Resource Surveys and Remote Sensing, Nairobi, Kenya. |
|17.||Department of Agriculture. 1999. Atlas Products of the North West Province. Technical Supportive Services, GIS Section, Department of Agriculture, North West Province, Potchefstroom, Republic of South Africa. |
|18.||Ellis, J.E. 1995. Climate Variability and Complex Ecosystem Dynamics: Implications for Pastoral Development. In: Living with Uncertainty: New Directions in Pastoral Development in Africa (ed. I. Scoones), pp. 37-46. Intermediate Technology Publications, London. |
|19.||Ellis, J.E. and M.B. Coughenour. 1998. The SAVANNAIntegrated Modelling System: An Integrated Remote Sensing, GIS and Spatial Simulation Modelling Approach. In: Drylands: Sustainable Use of Rangelands into the Twentyfirst Century (eds. V.R. Squires and A.E. Sidahmed), pp. 97-106. Technical Reports, WAD Series, Rome. |
|20.||FAO. 2001a. Pastoralism in the New Millennium. Food and Agricultural Organisation, Rome. |
|21.||FAO. 2001b. The AfriCover Initiative. AfriCover East Africa Module, Nairobi, Kenya. URL: |
|22.||Galaty, J. 1980. The Maasai Group-Ranch: Politics and Development in an African Pastoral Society. In: When Nomads Settle: Processes of Sedentarization as Adaptation and Response (ed. P.C. Salzman), pp. 157-172. Praeger Special Studies, J.F. Bergin Publishers, New York. |
|23.||Galaty, J.G 1994. Ha(l)ving land in common: The subdivision of Maasai group ranches in Kenya. Nomadic Peoples 34/35:109-122. |
|24.||Galvin, K., P. Thornton, and S. Mbogoh. 2000. Integrated Modelling andAssessment for Balancing Food Security, Conservation, and Ecosystem Integrity in East Africa. Final Report: Socio-economic Modelling Component, 1997-2000. Global Livestock Collaborative Research Support Program, University of California, Davis, California. |
|25.||Heath, B. 2000. Ranching: An Economic Yardstick. In: Wildlife Conservation by Sustainable Use (eds. H.H.T. Prins, J.G. Grootenhuis andT.T. Dolan), pp. 2133. KluwerAcademic Publishers, Boston. |
|26.||Hudson, J.W. 2002. Response to Climate Variability in the Livestock Sector in the North-West Province, South Africa. M.A. thesis. Colorado State University, Fort Collins, Colorado. |
|27.||Kavoori, P.S. 1999. Pastoralism in Expansion: The Transhuming Herders of Western Rajasthan. Oxford University Press, Oxford |
|28.||Kenya Soil Survey. 1995. 1:1 M Scale Soil and Terrain Database of Kenya (KenSOTER). Ministry of Agriculture, Nairobi, Kenya. |
|29.||Kiker, GA. 1998. Development and Comparison of Savanna Ecosystem Models to Explore the Concept of Carrying Capacity. Ph.D. thesis. Cornell University, Ithaca, New York. |
|30.||Kimani K. and J. Pickard. 1998. Recent trends and implications of group ranch subdivision and fragmentation in Kajiado District, Kenya. Geography Journal 164:202-213. |
|31.||Kinyamario, J.I. 1996. NPP Grassland: Nairoi, Kenya, 1984-1994. Oak Ridge National Laboratory, Distributed Active Archive Center, Oak Ridge, Tennessee. URL: |
|32.||Kristjanson, P., M. Radeny, D. Nkedianye, R. Kruska, R. Reid, H. Gichohi, F. Atieno, and R. Sanford. 2002. Valuing Alternative Land-use Options in the Kitengela Wildlife Dispersal Area of Kenya. International Livestock Research Institute (ILRI) Impact Assessment Series 10. A joint ILRI/African Conservation Centre Report, Nairobi, Kenya. |
|33.||Lane C. and R. Moorehead. 1995. New Directions in Rangeland Resource Tenure and Policy. In: Living with Uncertainty: New Directions in Pastoral Development in Africa (ed. I. Scoones), pp. 96-133. Intermediate Technology Publications, London. |
|34.||Low, A.B. and A.G. Robelo (eds.). 1996. Vegetation of South Africa, Lesotho and Swaziland. Department of Environmental Affairs and Tourism, Pretoria, South Africa. |
|35.||Mworia, J. and J.I. Kinyamario. 2000. The Impacts of Land Use Changes on the Vegetation, Soils and Water Balance in Kajiado, Kenya. Report to the Global Livestock Collaborative Research Support Program. Davis, California. |
|36.||NBI. 1996. Vegetation Map of South Africa, Lesotho, and Swaziland. Department of Environmental Affairs and Tourism. National Botanical Institute, Deon Marais, Pretoria, South Africa. |
|37.||Niamir-Fuller M. and M.D. Turner. 1999. A Review of Recent Literature on Pastoralism and Transhumance in Africa. In: Managing Mobility in African Rangelands: the Legitimization of Transhumance (ed. M. Niamir-Fuller), pp. 18-46. Intermediate Technology Publications, Ltd., London. |
|38.||Ole Katampoi, K., G.O. Genga, M. Mwangi, J. Kipkan, J. Ole Seitah, M.K. van Klinken, and M.S. Mwangi. 1990. Kajiado District Atlas. ASAL Programme Kajiado and the Ministry of Reclamation and Development of Arid, Semi-Arid Areas and Wastelands, Kajiado, Kenya. |
|39.||Pickup, G. and D.M. Stafford Smith. 1993. Problems, prospects and procedures for assessing the sustainability of pastoral land management in arid Australia. Journal of Biogeography 20: 471-487. |
|40.||Ritchie, M. and H. Olff. 1999. Spatial scaling laws yield a synthetic theory of biodiversity. Nature 400: 557-560. |
|41.||Rutten M.M. 1992. Selling Wealth to Buy Poverty: The Process of the Individualisation Among the Maasai Pastoralists of Kajiado District, Kenya: 1890-1990. Verlag Breltenbach Publishers, Fort Lauderdale, Florida. |
|42.||Sandford S. 1995. Improving the Efficiency of Opportunism: New Directions for Pastoral Development. In: Living with Uncertainty: New Directions in Pastoral Development in Africa (ed. I. Scoones), pp. 174-182. Intermediate Technology Publications, London. |
|43.||Scoones, I.E. 1991. Wetlands in drylands: Key resources for agricultural and pastoral production in Africa. Ambio 20:366-371. |
|44.||Scoones, I.E. 1992. Coping with drought: Reponses of herders and livestock in contrasting Savanna environments in southern Zimbabwe. Human Ecology 20:293-314. |
|45.||Scoones, I.E. 1993. Why are There so Many Animals? Cattle Population Dynamics in the Communal Areas of Zimbabwe. In: Range Ecology at Disequilibrium: New Models of Natural Variability and Pastoral Adaptation in African Savannas (eds. R.H. Behnke, I.E. Scoones, and C. Kerven), pp. 62-76. Overseas Development Institute, International Institute for Environment and Development, and Commonwealth Secretariat, London. |
|46.||SEVP. 2002. SEVP's Home Ground. Savanna Elephant Vocalization Project, Nairobi, Kenya. URL: |
|47.||Thornton, P.K., S.B. BurnSilver, R.B. Boone, and K.A. Galvin. (in press). Modelling the impacts of group ranch subdivision on agro-pastoral households in Kajiado, Kenya. Agricultural Systems. |
|48.|| Thornton, P.K., R.H. Fawcett, K.A. Galvin, R.B. Boone, J.W. Hudson, and C.H. Vogel. 2004. Evaluating management options that use climate forecasts: modelling livestock production systems in the semi-arid zone of South Africa. Climate Research 26: 33-42. |
|49.||Toxopeus, A.G., X. Bakker, and A. Kariuki. 1994. An interactive spatial modelling (ISM) system for the management of the Amboseli Biosphere Reserve (southeast Kenya). ITC Journal 4: 392-400. |
|50.||Vetter, S. (ed.). 2004. Rangelands at Equilibrium and Non-equilibrium: Recent Developments in the Debate Around Rangeland Ecology and Management. Programme for Land and Agrarian Studies, University of the Western Cape in association with the Rhodes University Botany Department and the Leslie Hill Institute for Plant Conservation, Cape Town, South Africa. |
|51.||Western, D. and H. Gichohi. 1993. Segregation effects and the impoverishment of savanna parks: the case for ecosystem viability analysis. African Journal of Ecology 31:269-281. |
|52.||Xie, P. and P.A. Arkin. 1997. A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bulletin of the American Meteorological Society 78:2539-2558. |
[Figure 1], [Figure 2], [Figure 3], [Figure 4a], [Figure 4b], [Figure 5a], [Figure 5b], [Figure 6a], [Figure 6b], [Figure 6c], [Figure 7a], [Figure 7b], [Figure 7c], [Figure 8], [Figure 9], [Figure 10]