Rapid assessment of Restoration Outcomes: Case study of Acacia reficiens control in Samburu

Approximately 65–70% of rangelands in Sub-Saharan Africa classified as being moderately to severely degraded, with significant reductions in vegetation cover, increases in undesirable species, or both. Several initiatives have been initiated to halt or reverse degradation across the continent at various scales. However, scarce resources available to managers on degraded African Rangelands discourages experimental testing of restoration methods. Where there are restoration projects carried out, very few conduct monitoring rigorous enough to answer questions about the ‘success’ of the resulting outcomes. This unfortunately often leads to uniform and untested application of restoration methods, despite the success of these methods depending on topography, soils, climate, and degradation type.

Due to persistent overgrazing and recurrent drought, much of the Ewaso ecosystem in northern Kenya is degraded. This landscape is characterized by large patches of bare ground and the spread of undesirable native species like Acacia reficiens and exotic invasive species like Opuntia stricta. The spread of A. reficiens, especially in the Samburu portion of this ecosystem, is of great concern as it reduces both habitat for endangered wildlife species like the Grevy’s zebra, as well as available forage for pastoral communities (1).

Very little literature exists about the historical vegetation patterns of most eastern African rangelands, and even less so about the Samburu area. The few existing literary and oral indigenous sources indicate that vegetation complexes in this area were historically dominated by dryland Acacia-Commiphora savannas. Most sources identify acacia tortillis as the dominant species, with one survey in 1976 finding that this species accounted for 44% of all tree cover in the then Samburu-Isiolo game reserve. In most historical accounts, A. reficiens was not identified as a major constituent of the plant community or a particularly problematic encroaching plant, as it was confined to specific low productivity parts of the landscape. Climatic variations, alterations of human movements through reduction in effectiveness of pastoral nomadism, fluctuation in wildlife and livestock populations, and changes in land use have resulted in shifts of vegetation community composition across much of Northern Kenya, including Samburu. There has been a perceived reduction in cover and diversity of perennial grasses, as well as an increase in shrubs and other woody species, with A. reficiens especially documented as taking over a significant amount of what was previously grassland, and replacing more desirable woody vegetation like A. tortillis on these landscapes.

A. reficiens spread is associated with a lack of persistent herbaceous understory growth. The exact mechanism of this exclusion is not well understood, but studies of invasive Acacia species in other ecosystems have pinpointed microhabitat modification, soil property changes and to a lesser extent, allelopathy, or chemical deterrence. By taking over large tracts of community land, A. reficiens has in turn increased degradation in the rest of the landscape by increasing pressure on existing forage resources for livestock and wildlife as a result of overgrazing in the remaining grassland areas. The lack of persistent understory herbaceous growth is problematic from a forage availability perspective for wildlife conservation, with the decline in Hirola numbers in Garissa being partially attributed to a similar reduction in herbaceous cover (2). To compound matters, the reduction in herbaceous cover is a direct precursor to large scale soil erosion, with sheet erosion and formation of hard caps on low sloping areas making way to rill erosion and massive gulleys in slightly sloping areas. The Samburu landscape in A. reficiens dominated areas is characterized by heavy amounts of both kinds of soil erosion.

To better understand how tree cover by itself doesn’t necessarily stop or reduce soil erosion, let’s carry out a thought experiment. Imagine a regular floor in an empty room slightly sloping towards a drain on one side. If you poured water down the slope of the room, it would very quickly flow from one end of the room to the other without any obstruction and nearly all the water would flow out of the room. Now imagine if you put in several chairs into the same room and repeat the water pouring exercise. The chairs will slow down the water somewhat, but most of the water will still make it to the bottom of the room and flow out. However, if you spread carpets onto the same floor, even only covering half the area, these carpets will absorb most of the water and only a portion of the water will make it to the drain. Similarly, a landscape with only trees and no understory will still experience massive surface run-off velocities and consequently high risks of erosion. A landscape with no trees but fully covered with foliar ground cover will experience little to no water erosion, and very limited wind erosion, as is usually the case with high productivity black cotton grassland savannas. Peak degradation resistance in terms of deeper soil structure, wind erosion, and water erosion is achieved when there is significant ground cover coupled with varying vegetation structure including perennial grasses, forbs, shrubs, and trees. With this context in mind, most pastoral communities in A. reficiens covered areas have identified clearing of this species and replacement with perennial grass cover as a priority for both livestock production and wildlife conservation.

In the mid-2000s, community members in affected landscapes in Samburu county began searching for options for halting A. reficiens spread and possibly reclaiming former grazing land on affected lands. Beginning in 2009, a few community conservancies, with support from the Northern Rangelands Trust and the Grevy’s Zebra Trust (GZT), set up pilot restoration projects. These projects were aimed at reducing density and cover of A. reficiens through manual clearing with machetes and increasing forage available to livestock and wildlife in these mixed use community conservancies through reseeding with the native perennial grass Cenchrus ciliaris or African foxtail grass/ buffel grass.

While these projects have been considered ‘successful’ by community members at achieving their stated goals of increasing forage production and ground cover, there was no quantitative or even semi-quantitative data that allowed an objective evaluation. The definition of ‘success’ has always been a point of contention amongst practitioners, with most people typically measuring vegetation conditions, structure, processes, and similarity to original or idealized reference sites. One approach is to incorporate assessment of ecological process and structure compared to reference states unique to each ecological site and its associated unique site or land potential. As discussed in an earlier article, land potential of a site means that it has specific soil and physical characteristics that makes it differ from other sites in its ability to produce different types and amounts of vegetation, and differs in its ability to respond to management actions and natural disturbances.

Several rapid frameworks for restoration evaluation have been suggested, and one approach uses the establishment of control plots, or untreated comparison areas. These control areas are then used in a BACI framework, or Before-After, Control-Impact design. This simply means that the practitioner collects data on a site before treatment, and then repeats the same assessment after a relevant amount of time has passed post-treatment. This is the Before-After part of the protocol. A second comparison is made using control sites that were also affected by the degradation damage being treated, but that were left out of the treatment. This allows one to see if any changes that are visible are because of the treatment or just incidental to conditions on the site. This second step (Control-Impact) is especially important to avoid claiming success due to a treatment when the changes are due to other external factors. For example, if we reseed an area, we want to be sure that when grass grows its because of the effects of the reseeding intervention, and not because there was already seed in the soil. Having an untreated control gives us an easy way to test this difference.

However, for a control to be effective, it requires establishing paired plots where the conditions (soils, topography, climate, and current vegetation) that are likely to affect response to restoration are as similar as possible, i.e. their land potential is similar. This is especially important if the assessment is being done after restoration has already been carried out, and there is no baseline data to carry out a Before-After comparison. Treatment and control plots in this case have to have similar underlying land potential, and have a reasonable probability of having had similar degradation conditions before restoration was carried out. Since the restoration work on our area of focus in Samburu (Westgate and Kalama Community Conservancies) had been carried out before any baseline data were collected, we by necessity had to carry out a restrospective Control-Impact assessment, as described in Kimiti et al. 2020 b (3).

We selected 13 restoration sites set up in different years across the two conservancies, allowing us to select 22 restoration plots. Our experimental plots were more than our restoration sites since some of our sites were large enough that they had areas with multiple land potential within them. We used the Land-Potential Knowledge System (LandPKS) platform to assess our sites for productivity potential, using the Mobile Apps and the LandInfo module that allowed us to collect soil and site data used to calculate Plant Available Water Holding Capacity (PAWHC) using the Rosetta pedo-transfer model. Newer verions of the app carry these calculations out automatically. We combined this information with info from other global climate and crop growth databases to obtain grass growth potential and soil erosion potential estimates from the Agricultural Policy/ Environmental eXtender model (APEX). We used all this information to select the best matching control plot to each of our treatment plots, creating a matched-pairs experimental design.

For each treatment-control pair, we collected data on select vegetation measures primarily using the LandCover module of the LandPKS mobile app, which is based on the ‘stick method’ for monitoring Rangeland health (4). To ascertain overall effectiveness of the restoration project at increasing vegetation cover, we collected data on total ground cover, foliar cover, bare ground cover, perennial grass cover, shrub and sub-shrub cover, annual plant cover, and both woody and herbaceous litter cover. To assess differences between treatment and control plots in water and wind erosion risk, we collected information about the number of large gaps between plant bases and plant canopies in each plot. To find out the effectiveness of treatments at reducing the cover of our target woody species, we compared tree cover, as well as the density of A. reficiens plants between our treatment and control plots. Additionally, we collected standing biomass data; clipping, drying, and weighing herbaceous vegetation from selected areas in each plot.

Results and discussion

First of all, we carried out matched pair analyses for our land potential measures for our treatment and control plots, confirming that there were no differences in slope, potential grass productivity, potential erosion, soil texture (relative proportion of sand, silt, and clay) at four different soil depths, and finally in the PAWHC measure selected. This was a critical step in ensuring that our treatment and control plots had reasonably similar land potential at the beginning of the restoration project.

Treatment plots, when compared to control plots, had 30% higher total ground cover, 28% total vegetation (foliar) cover, and nearly 35% higher perennial grass cover. Cleared areas had 60% more standing herbaceous biomass, meaning higher forage capacity for livestock and grazing wildlife. Treated areas as expected had 85% fewer A. reficiens plants per acre, but had double the shrub and perennial forb cover, meaning there was more browse availability as well, given A. reficiens exhibits little to no utilization by both wild and domestic browsers. There was no difference in annual plant cover and herbaceous litter cover between treatment and control plots, but unsurprisingly, woody litter cover was higher in the cleared areas. Soil texture at 20-50 cm interestingly affected our percent of basal gaps, with plots having Sandy Clay Loam at that depth exhibiting 50% fewer large gaps than control areas. This is not unexpected as soil texture at different levels in the soil profile, slope shape, slope size, and local climate conditions will affect the ability of a site to capture water and nutrients and make them available to plants. Conceptually, this validates the need to collect as much site characterization data as it may help explain differences in treatment success between different locations.

As a strict comparison to Restoration objectives therefore, our results show that the clearing and reseeding project demonstrably achieved its stated goal of reducing A. reficiens cover and density and increasing herbaeous forage cover for wildlife and livestock. The image below especially works as a qualitative real world explanation of what is meant by increase in ground cover and reduction in erosion potential. Our previous discussion of water flow on the surface being obstructed and slowed down by covering vegetation is clearly visible here, further illustrating the fragility caused by A. reficiens spread in a previously highly productive expanse of land.

It is worth noting that although perennial grass cover was higher in treated than untreated areas, total perennial grass cover was still below the 50% cover target set by management on these conservancies. Treatment plots on both conservancies produced enough grass cover that the communities could harvest seeds at the end of each rainy season that was then used for further reseeding projects on the landscape. Seeding efficiency could potentially be optimized by both ensuring seeding rates are standardized, as well as accounting for small differences in surface topography when reseeding, perhaps even digging shallow berms or trenches (5). Using a higher diversity of native grass species when reseeding would also be a potential method for increasing species diversity and patch resilience. Eragrostis superba (Maasai Love Grass) and Enteropogon machrostachyus (Bush rye/ Needlegrass) are two other indenous grass species that have been used successfully for reseeding other Kenyan rangelands.


Lack of information about reference conditions and land potential affects land managers’ ability to evaluate restoration success objectively as well as predict it. We found that site characteristics affected some restoration targets at certain sites. A large body or work exists describing impacts of land potential information on potential productivity (and therefore potential response to restoration interventions) of a site. Collecting this information could help land managers better decide where to direct scarce resources for restoration.

Our results demonstrate the effectiveness of manual clearing of A. reficiens stands as a restoration method. When paired with reseeding interventions, especially those that optimize seed retention and sediment capture, it has the potential to reverse losses in forage for both wildlife and livestock in affected community owned, mixed-use rangelands. However, for these treatments to have a greater impact on landscape and ecosystem level restoration, they must be monitored over time, documented vigorously, and disseminated to a wider audience to allow upscaling and testing in different landscape and climate conditions. For example, testing of these methods in similalry problematic A. reficiens areas like Garissa in Eastern Kenya would be very informative from a replication perspective.

As encouraging as these results are, there is need to emphasize that any rehabilitation or restoration intervention is unlikely to succeed in the long-term if the underlying factors causing degradation or vegetation composition change are not adequately addressed. In the case of Samburu, Isiolo, and Laikipia rangelands, the primary degradation cause has been pinpointed as unsustainable grazing practices leading to overgrazing. Unless grazing management across the larger Ewaso Ecosystem is restructured to avoid overuse of a constantly reducing and seasonally variable amount of available forage, then any rehabilitation or restoration programs will only have temporary benefits to the communities implementing them, and will most likely only attract the attentions of surrounding pastoral nomadic communities during drought periods. And though our vegetation data and anecdotal information from the community supports the idea that both wildlife and livestock have benefitted greatly from the increased grazing resources, further research on impact of A. reficiens presence and subsequent clearing on browse availability to wildlife and domestic livestock is also recommended before “success” can clearly be claimed.

Finally, our work here shows that it is imperative that large scale restoration projects set up treatment and control plots to monitor and evaluate relative restoration success. Easy to use mobile phone applications exist that could facilitate rapid restoration assessment, including helping design properly matched reference and control sites, as well as facilitating vegetation data collection.

1) Kimiti, D. W., Hodge, A. M. C., Herrick, J. E., Beh, A. W., & Abbott, L. E. (2017). Rehabilitation of community-owned, mixed-use rangelands: lessons from the Ewaso ecosystem in Kenya. Plant ecology, 218(1), 23-37.
2) Ali, A. H., Ford, A. T., Evans, J. S., Mallon, D. P., Hayes, M. M., King, J., … & Goheen, J. R. (2017). Resource selection and landscape change reveal mechanisms suppressing population recovery for the world’s most endangered antelope. Journal of applied ecology, 54(6), 1720-1729.
3) Kimiti, D. W., Ganguli, A. C., Herrick, J. E., & Bailey, D. W. (2020). Evaluation of Restoration Success to Inform Future Restoration Efforts in Acacia reficiens Invaded Rangelands in Northern Kenya. Ecological Restoration, 38(2), 105-113.
4) Monitoring Rangeland Health Manual. Riginos and Herrick 2010.
5) Kimiti, D. W., Riginos, C., & Belnap, J. (2017). Low‐cost grass restoration using erosion barriers in a degraded African rangeland. Restoration ecology, 25(3), 376-384.

1 Comment

  1. Great read as always.

    I like classroom, chair, carpet illustration – drives the point home and especially with plants like Acacia Reficiens.
    As you said we need to solve the underlying issue before we get to clearing the A.reficiens.

Leave a Reply

Your email address will not be published. Required fields are marked *