|Livestock Research for Rural Development 26 (9) 2014||Guide for preparation of papers||LRRD Newsletter||
Citation of this paper
Climate Change affect various sectors in Kenya, with the most vulnerable being agriculture, livestock, water, health, fisheries and tourism. Accurate estimates of soil organic carbon stocks (SOCS) in the rangelands are critical in developing strategies to help mitigate impacts of climate change. The study therefore, sought to establish the relationship between vegetation cover types and SOCS in northern rangelands of Kenya as an indirect method of estimating SOCS in the field. Landsat 5 Thematic Mapper satellite image was used to differentiate vegetation cover types and soil samples taken along the transect line laid at intervals of 50 m across each vegetation cover type. Colourimetric and core sampling methods were used to determine SOC concentrations and soil bulk densities, respectively. Analysis of variance and simple linear regression were used in the statistical analysis.
Four vegetation cover types indentified were: Acacia bush land (ABL), bare land (BRL), sparsely distributed acacia with bare ground (SAB) and sparsely distributed acacia with forbs (SAF) and. The means of SOC for each vegetation cover were different. However, soil bulk densities under BRL and SAB were similar but different from that of ABL and SAF that were alike. Further, overall mean of SOCS was 6.76±2.85 t C ha-1 for all the vegetation cover types. A positive relationship was established between the average mean values of both Normalized Difference Vegetation Index (NDVI) and Soil Adjusted Vegetation Index (SAVI) when regressed with the average mean values of SOCS. The findings suggest that vegetation indices measured with GIS are good predictors of SOCS for the study region, with the potential for extrapolation to the arid and semi-arid areas to which this ecosystem belongs.
Keywords: Climate Change, Normalized Difference Vegetation Index (NDVI)
Atmospheric concentrations of greenhouse gases (GHGs) have increased globally since the pre-industrial era due to anthropogenic activities (Houghton 2007; IPPC 2007). This has lead to an increase in average global surface temperatures (Lal 2004) with an approximate mean annual temperature increase in the range of 1-2°C recorded in Kenya (ILRI 2008). These changes have had negative effect in ecosystem structure and functions with the most vulnerable sectors being crop production, livestock water, health, fisheries and tourism. Various strategies has been developed with an aim of mitigation these threats; CO 2 sequestration in soil is one of the strategies which helps in reducing CO2 enrichment in the atmosphere (Salahuddin 2006; Roncoli et al 2007). Accurate and reliable estimates of soil organic carbon stocks (SOCS) in the rangelands are therefore critical in developing strategies to help mitigate impacts of climate change as well as improving rangeland productivity. A number of studies have been done to estimate SOCS at global scales ((Post et al 1982; Eswaran et al 1993 Sombroek et al 1993; Batje 1996), These estimates were based on information derived from different global soil maps and soil organic carbon(SOC) concentration and other attributes obtained from representative soil profiles (GEFSOC 2003). However, few studies have been conducted to determine SOCS levels and its relationships with different vegetation types in arid and semi-arid lands (ASAL) ecosystems yet ASALs are of global importance; they occupy lagers areas and have the potential to sequester significant amount of CO2 from the atmosphere (Lal 2001). This is probably because of high spatial variability of SOCS in ASALs which necessitates very high sampling densities to get accurate and reliable estimates (Bird et al 2002; Martin et al 2011). The objective of the study was therefore to examine the relationships between vegetation cover types and SOCS in rangelands of northern Kenya as a means of developing a methodology to indirectly estimate SOCS in ASALs. To achieve this, it was hypothesized that vegetation indices derived from satellite image can be used indirectly to estimate SOCS in ASALs.
The study was conducted in the grazing unit (‘rage’) of Gabra rangeland of Kalacha location in Chalbi district (Marsabit County) in northern Kenya as shown in Figure 1. The grazing unit is located between latitude 3°14'12.10'' N and 3°11'08.37'' N, and, between longitude 37°17'21.92 E and 37°22'16.36''. The unit is also classified as ecological zone VI, with a climate that is characterized by high input of solar radiation, high radiative heat losses at night, low precipitation, high moisture losses and prolonged water deficits. It receives an annual rainfall of 157 mm which exhibited both temporal and spatial variability and bimodal distribution. Drought is a common phenomenon which puts water stress on the already fragile ecosystem.
|Figure 1. Map of the study area: BRL: bare land. SAB: sparsely distributed acacia with bare ground cover,
SAF: sparsely distributed acacia with forb undergrowth and ABL: acacia bushland cover
Acacia species, shrubs and forbs were the major vegetation cover types with Acacia tortilis being the dominant tree species. The other species identified were; Acacia seyal, Salvadora persica, Acacia nubica, Balanite spp and colonies of Hyphaena coriacea (doum palms). At the time of the study, larger areas were completely bare with only desert pavements on the surface. Soils are sandy loam in texture, saline, sodic and calcareous, shallow to moderate deep and pale brown, with the parent material being sand mixed with some volcanic ashes.
Landsat 5 Thematic Mapper (TM) satellite image with a resolution of 30 x 30 m was processed with ERDAS IMAGINE 9.1 software and used to classify vegetation into four different cover types namely; acacia bushland cover (ABL), sparsely distributed acacia with forb undergrowth (SAF) and sparsely distributed acacia with bare ground cover (SAB). The four identified vegetation cover types formed the basis of stratifying the ‘rage’ grazing unit into four strata. The geographical positions centers of the four vegetation cover types were identified, marked, recorded and loaded into Trimble Goex GPS system. Subsequently, ground-truthing to locate each vegetation cover type in the field was done with the help of a GPS, Marsabit county physical map and experienced herders selected from the site.
A transect line of 750 m in length was laid across each vegetation type and sampling of soils done at interval of 50 m along the transect line, giving a total of 15 replicates per each cover type and a total of 60 samples. Few samples (15) were collected due to poor accessibility and heterogeneous patchy vegetation which is a characteristic of ASALs in Kenya. Additionally, there was a prolonged drought which reduced the vegetation cover significantly at the time of the study limiting the length of the transect line. Soil samples were collected at a depth of 0-15 cm in an area of 1 m2 in a Z pattern using a soil auger along the transect line. Sampling to a depth of 0-15 cm was possible only beyond which an impermeable rock layer made it impossible particularly on Bare Land cover and (BRL) and sparsely distributed acacia with bare ground cover (SAB). Four sub-samples of soil were collected at every corner of the 1 m2 mixed in a larger plastic bucket, and 500 g sample was pooled out and taken for laboratory analysis. Soil samples for determination of BD was sampled along the laid transect line of 750 m at an interval of 50 m on each vegetation cover type. Coring rings of known volume (100 cm3) were used to collect the samples. All the soil samples were labeled and Geographical position of every sampling point taken using soil using GPS (Trimble GeoXT) and recorded.
Analysis of the total SOC concentration was done at the National Agricultural Research Laboratories (NARL) in Nairobi using the colourimetric method (Anderson and Ingram 1993). This method is a wet-oxidation procedure that uses potassium dichromate with external heat as shown below in equation 1:
2Cr2O72-+ 3C + 16H+ → 4Cr3++ 3CO2 + 8H2O (Equation 1)
Soil samples were first dried at room temperature before passing them through a 2 mm sieve. The coarse fraction (particles > 2 mm) was weighed to determine the percentage of the coarse fraction. One gram of the soil sample of fine fraction (from <2 mm) was scooped, grounded and passed through a 0.05 mm screen into a labeled digestion tube. The standard samples and reagent blanks were also included in each step of analysis. 2 ml of deionized water was then added to each soil sample using a pipette (deionized water was not added to the standards since it was already added during preparation stage), 10 ml of 5% potassium dichromate solution was then added into both the standards and sample tubes, and potassium dichromate allowed to completely wet the sample. Slowly, 5 ml of concentrated H2SO4 technical grade was added from a bottle-top dispenser in drops of 1 ml of the acid at a time while swirling on a vortex mixture to avoid violent reaction and then the digest was heated at 150° C for 30 minutes. The samples were then removed from the heater and allowed to cool; 50 ml of 0.4% Barium Chloride solution was then finally added and allowed to stand overnight to ensure complete mixing. Carbon concentration was then read on the spectrophotometer at 600 nm. No calibration curves were drawn since the spectrophotometer was computerized and gave the results automatically after it was calibrated with two standards of 0 and 12.5 mg C/ml as the lowest and the highest values respectively.
Analysis of soil bulk density was done following the methodology described by Cresswell and Hamilton (2002). Oven-proof container was first weighed before carefully pushing out the trimmed soil cores into it. The oven-proof container and the soil were again weighted and the weight recorded. The same procedure was repeated for all the sixty soil core samples before oven drying it in a well ventilated oven at 105°C for 48 hours until the weights of the soil were constant. The container with the soils was removed from the oven and cooled in the desiccator before weighing the container and the oven dried soil. Soil bulk density (g cm-3) was then calculated as depicted in equation 2 below:
BD Sample = ODW Sample /CV Sample (Equation 2)
where; BD Sample is the bulk density (g cm-3) of the soil sample,ODWSample the mass (g) of oven dried soil core and CVSample the core volume (cm 3) of the soil sample.
Calculation of SOCS
The SOCS (t C ha-1) were calculated from the SOC concentration (%) obtained from laboratory analyses as indicated in equation 3 below:
C (t ha -1) = C (%)* ρ * D (Equation 3)
where C (t ha -1) is the SOCS (t ha-1), C (%) the concentration of SOC (%), ρ the soil bulk density (g cm -3) and D the depth of sampling in cm (0-15 cm).
To estimate SOCS indirectly from vegetation indices, Normalized Difference Vegetation Index (NDVI) and Soil Adjusted Vegetation Index (SAVI) were computed from Landsat TM 5 reflectance image using ERDAS IMAGINE 9.1 software. NDVI for each pixel was derived according to the relationship described by Rouse et al (1973) as shown in equation 4:
NDVI = NIR –R / NIR + R (Equation 4)
where; NIR and R are reflectance in Near Infra – Red and Red bands, respectively. The SAVI was computed using the formula of Huete (1988) as indicated below in equation 5 below:
SAVI = (1 + n) (NIR – R) / NIR + R + n (Equation 5)
where n = 0.5. The SAVI adjusted factor n is used to compensate for the influence of varying soil backgrounds on the measured plant index and is typically assigned a value of n = 0.5 (Huete 1988).
Statistical analyses of each of the measured variable were performed with general procedures model (GLM) procedure of SAS version 9.0 software (SAS Institute Inc., 2010). Homogeneity of variance and normality of distribution of all dependent data were verified graphically prior to analyses. Statistically significant interactions were subjected further to one way ANOVA with a SLICE command of PROC GLM. Multiple comparisons of means of SOC concentration, soil bulk densities and SOCS for the four vegetation cover types were done by Tukey’s HSD (P<0.05). The relationship between vegetation indices (NDVI and SAVI) and SOCS was derived from simple linear regression. Graphic presentation was done with Sigma plot (Version 12.0).
The means of SOC concentrations (%) for each vegetation cover type observed at a depth of 0-15cm are presented in Figure 2 (ABL, 0.59±0.14; BRL, 0.18±0.05; SAF, 0.47±0.04; SAB, 0.28±0.07). The overall mean of SOC concentration for the four vegetation cover types was 0.38±0.18%, with a coefficient of variation (CV) of 21.26 (r2 = 0.86, P<0.05). The SOC concentrations were different under the four vegetation cover types (P<0.05).
|Figure 2. Mean soil organic carbon concentration (%) at a depth of 0-15 cm for each vegetation cover type; the bars represent
standard errors of the mean. a, b, c, d are different (P<0.05). BRL: bare land. SAB: sparsely distributed acacia with
bare ground cover, SAF: sparsely distributed acacia with forb undergrowth and ABL: acacia bushland cover
The separate means of soil bulk densities (in g cm -3) for each of the four different vegetation cover types at a depth of 0-15 cm are given in Figure 3 (ABL, 1.15±0.10; BRL, 1.32±0.08; SAF, 1.14±0.10; SAB, 1.23±0.14). The overall mean was 1.23±0.14 g cm-3, with a CV of 9.04 (r 2 =0.52, P<0.05). Soil bulk densities under BRL and SAB were similar but different from that of ABL and SAF that were alike (P<0.05), with BLR and SAB having higher mean values compared to those of SAF and ABL, and ABL being the least.
|Figure 3. Mean soil bulk densities in g cm -3 at a depth of 0-15 cm for each vegetation cover type, the bars represent standard errors of the mean.
a, b, are different (P<0.05). BRL: bare land SAB: sparsely distributed acacia with bare ground cover,
SAF: sparsely distributed acacia with forb undergrowth cover and ABL: acacia bushland cover.
Figure 4 shows least squares means (t C ha-1) of SOCS under the four different vegetation cover types at a depth of 0-15 cm (ABL, 10.05±2.02; BRL, 3.53±0.79; SAF, 8.23±0.97; SAB, 5.42±1.51). The overall mean of SOCS was 6.76±2.85 (r2 = 0.85, P < 0.05) with a CV of 19.44. The SOCS differed among the four vegetation cover types (P<0.05), with a lower and higher mean value for BRL and ABL, respectively.
|Figure 4. Mean soil organic carbon stocks in t C ha -1 at a depth of 0-15 cm for each vegetation cover type; the bars represent standard
errors of the mean. a, b, c, d are different (P<0.05). BRL: bare land. SAB: sparsely distributed acacia with bare ground cover,
SAF: sparsely distributed acacia with forb undergrowth and ABL: acacia bushland cover
Figure 5 shows the relationship between the average means of SOCS (t C ha-1) and the average NDVI of four vegetation cover types. The model was evaluated using the goodness of fit creteria determined by R2, where the highest R2 values inidicate best fit of the model to the data. The SOCS had a high positive correlation with NDVI (R2=0.89), indicating that as NDVI increases, SOCS will also increase. A similar trend was observed on the relationship between the average means of SOCS and the average means of SAVI (R2 = 0.89) as shown in Figure 6.
|Figure 5. Relationship between the evarage means of SOCS (t C ha-1) and the average means of Normalized Difference Vegetation
Index (NDVI value) of four vegetation cover types. BRL: bare land, SAB: sparsely distributed acacia with bare ground
cover SAF: sparsely distributed acacia with forb undergrowth and ABL: acacia bushland cover
Figure 6. Relationship between the evarage means of SOCS
(t C ha-1) and the average means of no soil adjusted vegetation
The differences in SOCS and SOC concentration as also observed by Yimer et al (2006) and Li et al (2010) are highly associated with vegetation cover type; suggesting that there is a fundamental difference in net carbon assimilation under different vegetation cover types. Higher SOCS and SOC concentration recorded under ABL can be linked to higher litter production which is incorporated into the soil from A. tortilis (dominant plant species) compared to other vegetation cover types. It may also be ascribed to reduced solar radiation input into the soil due to protection from the canopy cover, leading to relatively higher soil moisture concentration which promotes higher root decomposition under ABL. The least mean values observed for BRL could be due to; loss of root carbon, leading to a reduction in carbon inputs from the roots and leaf litter, reduced microbial activity due to increased soil temperature relative to ABL and SAF as it was also noted by Mills and Fey (2004) and loss of top soil through erosion and increased soil compaction, causing an increase in soil bulk density. Soil compaction reduces water infiltration and increases runoff during rainy the seasons under BRL, causing a decrease in the water available for plant growth. The overall mean of SOCS recorded is comparable to those of 0-18 t C ha-1 in a depth of 1m reported earlier by GEFSOC (2003) for hot arid land (Zone VII) of northern Kenya, and 7.5- 9.9 t C ha-1 for degraded rangelands of west Africa (Batjes 1999. Lower SOCS observed can be attributed to soil structural degradation that has taken place through pulverization, compactions, soil particle dispersion, low organic matter inputs, high pH, high salinity and extremely high exchangeable sodium percentage. Overgrazing in the rangeland under low but highly variable precipitation, both in space and time, and high solar radiation inputs may also account for the lower SOCS levels. Overgrazing destroys the most palatable and useful species in the plant mixture and reduces the density of plant cover, leading to an increase in erosion hazards which deplete SOC pools.
The higher values separate means of soil BD densities under BRL and SAB relative to those of ABL and SAF were comparable to other observations in Kenya (Verdoodt et al 2009; Muya et al 2011; Kahi et al 2009). This is due to low incorporation of organic carbon in the soil as indicated by Pande and Yamamoto (2006). Loss of vegetative and litter cover coupled with rangeland degradation allows direct impact of rain drops on bare soils resulting to enhanced splash impacts, mechanical crust formation, surface sealing that reduce water infiltration in to soil. The lower soil bulk density values under ABL and SAF may also be explained by improved soil micro porosity due to improved microclimate, higher SOC concentration input and improved soil aggregate relative to BRL and SAB.
The positive corellation established between SOCS and NDVI and; SAVI indicates that as NDVI and SAVI increases SOCS aslo incresses, Nisha Wani and Dadhwal (2010) observed similar relationships between NDVI values derived from remote sensing data and soil carbon densities. This is expalined by the fact that NDVI have a direct relationship with photoysnthetic activities in green plants, this implies that more carbon is assimilated into the plants and by extension to the soil. This positive correlation is important in assessing carbon inventory, expecially in vast areas charaterized by heterogenous vegetation and rough terrain. However, it should also be noted that vegetation cover types alone as assessed in the study cannot solely and fully explain the behaviour of SOCS. There is need to consider the effect of grazing intensity, animal contribution to spatial SOCS redistribution in the grazing unit through defecation (manure deposits), site specific soil type information and profile attributes in order to quantify specific amounts of carbon sequestered into the soil by plants only. It is worth noting that the drought which was experienced in the region during data collection was a major limiting factor as it impeded the existence of different vegetation types in the grazing unit. The depth of soil sampling was only possible to a depth of 0-15 cm due to the hard rocky material in BRL and SAB. It is therefore, recommended that a satellite image with high spatial resolution, e.g., of <5 m, like the IKONOS, Quickbird and SPOT HRV be used in future so as to obtain clear information on vegetation types, particularly in areas characterized by patchy vegetation growth, for the purposes of accurate classification of vegetation cover types.
We are grateful to the GrassNet Project funded by DAAD (Deutscher Akademischer Austausch Dienst) for financial support, University of Hohenheim (Stuttgart, Germany) Kassel University (Germany), KARI (Marsabit) and Gabbra pastoral community for their moral and spirit support during the study.
Anderson J M and Ingram J S 1993 Tropical Soil Biology and Fertility: A Handbook of Methods: CAB International, Wallingford, Oxon, UK, pp. 171-221.
Batjes N H 1996 Total carbon and nitrogen in the soils of the world: Journal of European Soil Science 47: 151-163.
Batjes N H 1999 Management Options for Reducing CO2 Concentration in the Atmosphere by Increasing Carbon Sequestration in the Soils. Report 410-200-031, Dutch National Research Programme on Global Air Pollution and Climate Change, Technical Paper 30, International Soil Reference and Information Center, Wageningen, The Netherlands, pp 14-141.
Bird S B, Herrick J E, Wander M and Wright S F 2002 Spatial heterogeneity of aggregate stability and soil carbon in semi – arid rangelands: Environmental pollution, 116: 445-455.
Cresswell H P and Hamilton G J 2002 Bulk density and pore space relations: In: McKenzie, N., Coughlan, K., Cresswell, H. (Eds), Soil Physical Measurement and Interpretation for Land Evaluation. CISRO Publishing, Melbourne, Australia, pp. 35-58.
Eswaran H, Van den Berg E and Reich P 1993 Organic-carbon in soils of the world: Journal of Soil Science Society of America 57: 192-194.
GEFSOC 2003 Assessment of Soil Organic Carbon (SOC) Stocks and Change at National scale. Global Environmental Facility (GEF) Co-financed Project, Coordinated by the University of Reading (UK): Last retrieved on 25th June, 2011.
Houghton R A 2007 Balancing the global budget: Annual Review Earth Planet 35: 313-47.
Huete A R 1988 A Soil adjusted vegetation index (SAVI): Remote Sensing of Environment 25: 295-309.
ILRI 2008 Implications of Climate Change on Rangeland Productivity in Kenya. Clip Policy Brief. International Livestock Research Institute, Nairobi, Kenya, pp 1-2.
IPCC 2007 Summary for Policymakers: In: Climate Change; Impacts, Adaptation and Vulnerability: Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change: Cambridge University Press, Cambridge, UK, pp. 7‐22.
Kahi C H, Ngugi R K, Mureithi S M and Ng’ethe J C 2009 The canopy effects of prosopis juliflora and Acacia tortilis (hayne) trees on herbaceous plant species and soil physico-chemical properties in Njemps flats, Kenya. Tropical and Subtropical Agro Ecosystems 10: 441-449.
Lal R 2001 “Myths and Facts about Soils and the Greenhouse Effect,” in Soil Carbon Sequestration and the Greenhouse Effect. Soil Science Society of America, Inc., USA, pp. 9-26.
Lal R 2004 Soil carbon sequestration to mitigate climate change: Goederma 123: 1-22
Li P, Wang Q, Endo T Zhao X and Kakubari Y 2010 Soil organic carbon stock is closely related to above-ground vegetation properties in cold-temperate mountainous forests. Geoderma 154: 407–415.
Martin M P, Wattenbach M, Smith P, Meersman J, Jolivet C, Boulonne L and Arrouay D 2011 Spatial distribution of soil organic carbon stocks in France. Journal of Biogeosciences 8: 1053–1065.
Mills A J and Fey M V 2004 Transformation of thicket to savanna reduces soil quality in the Eastern Cape, South Africa. Plant and Soil 265: 153–163.
Muya E M, Obanyi S, Ngutu M, Sijali I V, Okoti M P, Maingi M and Bulle H 2011 The physical and chemical characteristics of soils of Northern Kenya arid-lands: opportunity for sustainable agricultural production. Journal of Soil Science and Environmental Management 2: 1-8.
Nisha Wani A V and Dadhwal V K 2010 Assessment of agricultural crop and soil carbon pools in Madhya Pradesh, India. Tropical ecology 51: 11 – 19.
Pande T N and Yamamoto H 2006 Cattle treading effects on plant growth and soil stability in the mountain grasslands of Japan. Land Degradation Development 17: 419-428.
Post W M, Emanuel W R, Zinke P J and Stangenberger A G 1982 Soil carbon pools and world life zones. Nature 298: 156-159.
Roncoli C, Jost C, Perez C, Moore K and Ballo A 2007 Carbon sequestration from common property resources: lessons from community-based sustainable pasture management in north-central Mali. Agricultural Systems 94: 97–109.
Rouse J M Haas R H, Schell J A and Deering D W 1973 Monitoring vegetation systems in the great plains with ERST. In: Proceedings of the 3rd ERTS Symposium NASA SP-351, pp 301-317.
Salahuddin M J 2006 Monitoring Spatial Variation in Soil Organic Carbon Using Remote Sensing and Geographic Information Systems: Ph.D. Thesis, Environmental Resources and Policy Graduate School, Southern Illinois University Carbondale, USA, UMI. Number 3229815.
SAS Institute Inc. 2010 Base SAS® 9.2 Procedures Guide: Statistical Procedures. Third Edition, Cary, NC: SAS Institute Inc. USA.
Sombroek W G, Nachtergaele F O and Hebel A 1993 Amounts dynamics and sequestering of carbon in tropical and subtropical soils. Ambio 22: 417–426.
Verdoodt A, Mureithi S M, Ye L and van Rans E 2009 Chronosequence analysis of two enclosure management strategies in degraded rangeland of semi-arid Kenya. Agriculture, Ecosystems and Environment 129: 332–339.
Yimer F, Stig L and Abdu A 2006 Soil organic carbon and total nitrogen stocks as affected by topographic aspect and vegetation community in the Bale Mountains, Ethiopia. Geoderma 135: 335–344.
Received 30 April 2014; Accepted 24 August 2014; Published 5 September 2014
Go to top