Author:

Abstract:

Soil organic carbon (SOC) is well known to maintain several functions. On the one hand, being the major component of soil organic matter (SOM),it is a determinant of soil physical and chemical properties, an important proxy for soil biological activity and a measure of soil productivity. Land use management that will enhance soil carbon (C) levels is therefore important for farmers and land use planners, particularly in semiarid and sub-humid Africa where severe soil degradation and desertification are related to perpetual food crises and overall impoverishment. On the other hand, C sequestration in the soil is an important climate changemitigation measure. The global SOC stock is about 1500 Pg, which is about three times the amount of organic C in the vegetation and twice the amount of C in the atmosphere. Since tropical soils store 26% of the SOC stocks of the world they could serve as C storage reservoirs. Due to an incomplete knowledge of specific soil properties at a global scale and, high spatial variability of SOC and the different effects of the factorscontrolling SOC, regional studies of SOC are necessary to refine globalestimates. Both soil and land use data should be used in determining regional SOC. While soil factors are important, as are climatic factors, in explaining C storage or pools over long periods of time, changes in vegetation or land use determine the changes in C sequestration over shorter periods. Southeastern Tanzania serves as a typical example of soil degradation and SOC losses through land use change, deforestation and increased land pressure on the African continent. Therefore any policies geared to the significant reduction of current and future emissions as well as the sequestration of carbon from the atmosphere should target the forest and land-use sectors. Local institutions have shown their interest in tacklingthese problems and in implementing soil conservation management measures such as reforestation of degraded soils, which could be funded by the Clean Development Mechanism of the Kyoto Protocol. In this way C sequestration opportunities could go hand in hand with local conservation of natural resources as well as poverty alleviation. However, the Kyoto Protocol limits reporting of C sequestration activities to those that are both “measurable and verifiable”. Even if allowed as part of an accounting system, credit for C sequestration activities could be withdrawn if highstandards of measurement cannot be met. Tools for monitoring, verification or certification will be needed in developing countries in order to measure the changes in C pools in relation to soil type, climatic conditions, land occupation and different land management practices. This research will therefore in the first place focus on SOC budgeting in the 415 ha large study area in southeastern Tanzania, based on soil and land use properties. Secondly, SOC measuring techniques that are bothcost efficient and reliable will be developed based on the spatial variation of SOC in the study area. Mapping and upscaling of SOC in southeasternern Tanzania will be based on soil and land use maps. However, no reliable soil map was available that could be used as a base for this study. Therefore, a soil map was created, based on 280 full profile descriptions available from the literature, supplemented with 354 new profile descriptions that were made by the author in the scope of this work. Soils were classified according to the World Reference Base. Spatial delineation of the boundaries was basedon both a geology as a topography layer in a GIS-environment. The geology consists of old basement sediments of the Archean period. As a resultof relative fluctuations of land and sea levels and recurrent marine transgression and regression a series of younger deposits has been laid down parallel to the shore. From east to west, southeasternern Tanzania can be divided into a coastal area, an area of plateaus and floodplains, acentral area of plains and a western area of plateaus and uplands. The oldest soils of the region are those of the large plains situated on Basement rocks. All these soils are deep and homogeneous with clayey subsoil. The most remarkable examples are the deep red Eutric Nitisols in the low areas of the central plains. The largest part of the plains consistsof Arenic or Ferralic Phaeozems which were formed on coarse-grained gneisses. Where clay illuviation resulted in an argic horizon, mainly on slopes around the higher plains, Lixisols were formed. On the inclined terrain with gneisses and quartzites of both the northern and the southern plains, erosion lead to a finer texture and clay illuviation effected inargic horizons. This resulted in Lixisols or, when fertile A-horizons were found, Luvic Phaeozems. All soils derived from the strongly weathered sandstone beds of Lower Cretaceous origin are deep and sandy: Arenic Ferralsols, Arenosols and (Ferralic) Arenic Phaeozems are the most common soils on the high plateaus while Arenic Alisols and Acrisols wereobserved on the slopes. On marine deposits of grey clays and marls and limestones from the Upper Cretaceous and Paleogene mainly Gleysols and Vertisols are formed. A complex pattern of sandy soils was formed on the marine and terrestrial layers of the Neogene: Luvic Arenosols, Alisols and Ferralsols. Quaternary deposits are present along the main rivers andat the coastline. Soils are variable: along the rivers (Calcaric) Vertisols, Gleysols, Fluvisols and Phaeozems are seen. At the coastline Arenosols are present. Besides a soil map, a land use map is necessary for SOC upscaling across the region. Land use maps from 1997 and 2005 were used to estimate land use changes that have occurred in the study area. Several authors stated that the original vegetation of southeast Tanzania consists of forestor forest-woodlands. More specific, coastal forests are the natural vegetation type of the coastal area, plateaus and southern part of the plains, miombo woodlands are naturally occurring in the rest of the plains and mangrove forests are naturally growing at the coastline and in sheltered bays. Despite its high ecological value, the coastal forest has suffered most from human intervention: only 0.4% of the coastal forest remains. The reason is the greater population in these areas, which is a historical development. In 1997 the coastal forest remained mainly on the less populated Rondo plateau but was replaced by secondary bushland in thecoastal area, cashew on the Makonde plateau and cropland with tree crops in the southern plains. Towards 2005 cashew expanded even more on the Makonde plateau, complemented by more than 20% of cropland. The large expanse of cashew groves is typical of southeastern Tanzania, where most of the nation’s cashew is produced. The central plains and northern plateaus were formerly covered in woodland, since the extreme seasonal fluctuations in soil moisture were unsuitable for forest growth. Although the total area of woodland in the study area has not much declined, large parts of it have been disturbed and been converted to secondary vegetation. In the large plains, areas with tree crops have expanded mainly on thefertile Nitisols. The rest of the area still remains largely under woodland cover. Attributing SOC stocks to the soils and land uses that were just described can be done after calculating the SOC of 210 profiles from the reference dataset and 354 profiles sampled by the author. SOC is measured in the laboratory as the mass of C per soil mass (%C), but for upscaling toland units it is expressed as a C density (Mg C.ha-1 to a certain soil depth). This can be calculated by multiplying %C with the bulk density. However, not all these variables are given for every dataset. They can however be estimated based on other soil properties using following methods. C percentages can be measured using different techniques, for example wet oxidation, dry oxidation or reflectance spectroscopy. A cheap and fast way of measuring soil organic matter is loss-on-ignition. It involves the removal of organic matter from an oven dry sample (dried at 105°C) by combustion of the sample at medium temperature (375 to 600°C) in a temperature-regulated muffle furnace. Regression equations can be used to derive SOC from SOM by loss-on-ignition. A regional regression equation was developed based on an intercept model for a subset of 100 samples, which were selected from most types of soil and land-use in the study area. This resulted in two different regression equations, one for the topsoil and one for the subsoil. The intercept and the low value of the slopeparameter for the subsoil can be explained by weight loss of other, inorganic, soil constituents. In the soils of the study area, this additional weight loss could be attributed to the desorption of clay water from kaolinite, removal of hygroscopic water of vermiculite and illite, loss of hydration water around gibbsite and goethite and loss of crystal-lattice water from gibbsite and goethite. These minerals are very common in the old soils found in the area. With these formulae all data tested forSOM with loss-on-ignition can now be converted to SOC.The soil bulk density is the mass of an oven-dry sample of undisturbed soil per unit volume. Bulk density is an important physical property which is needed for mass to volume or area conversions of soil properties; therefore it is indispensable for the assessment of soil C stocks and nutrient pools and it is the mandatory measuring unit for soil C assessment and accounting. Unfortunately, bulk density is not always determined in routine soil surveys, as the procedures are tedious, labour-intensive,time-consuming and expensive. To overcome this problem, pedotransfer functions are frequently used, which translate basic soil data into another soil property, in this case bulk density. It has been advised to calibrate and validate specific pedotransfer functions on a regional basis toenhance the accuracy and precision of bulk density prediction. Five model types of pedotransfer functions described in literature were tested against 153 horizon descriptions. By performing multiple regressions on five models based on three variables (organic carbon, texture and depth) and performing calibration and validation tests, the best function was selected. The best model for estimating bulk densities was a logarithmic function based on clay content and depth, but not on the organic carbon content. In clay-rich soils well developed micro-structures caused by the aggregation of negatively charged clay minerals such as kaolinite withsesquioxides could lower the bulk density more than the small effect ofthe low organic carbon content. Lower bulk densities higher in the profile are probably the result of the combined effect of higher organic matter contents and more decompaction by biotic activity of soil fauna and roots in the topsoil versus the subsoil. If texture and sampling depth are available from soil profile descriptions found in literature, they can now be used to derive the bulk density, which is often lacking from these data.SOC is known to decrease with soil depth. The vertical SOC distributioncan be estimated by fitting different equations found in literature. The result showed that the vertical distribution depends on the vegetation. In woodlands and bushlands a high proportion of SOC is stored in the upper layers and SOC distribution is shallow. The relatively low decomposability in wood- and bushland could increase SOC storage in surface soils compared to grasslands and croplands. The deeper distribution of C in croplands could also be the result of soil disturbance during clearing and preparation of the field, although soil preparation and ploughing is uncommon in the study area. Cashew fields have intermediate vertical SOCdistributions. The vegetation structure shows elements of the two othersystems with a tree layer of cashew trees and sometimes an undergrowth of grass or crops. The developed equations can be used to estimate SOC at depths which are not measured in certain datasets. With the previously described methodologies it is now possible to calculate SOC stocks for the 210 profile description from the literature and the 354 profiles sampled by the author. Both soil type and land use determine SOC stocks. Soil texture influences C dynamics through the formation of organo-mineral complexes that protect C from microbial oxidation which is enhanced by the high surface area of the fine silt and clay fractions. The equilibrium soil C stock, which is the result of abalance between the rate of SOC inputs and rate of mineralization, is disturbed by land use change until a new equilibrium is eventually reached in the new ecosystem. Conversion of natural vegetation to various landuses results in very rapid declines in soil organic matter and a release of C to the atmosphere. Globally, the long-term flux of C from changesin land use (1850-2000) released 165 Pg C to the atmosphere, about 60% of it from the tropics and 8% from tropical Africa.The study of 415 Kha area was divided into landscape units by intersecting the soil map and the land use maps of 1997 and 2005, thus generatinglandscape units. SOC stocks were attributed to these landscape units by assessing the relationship between SOC, soil type and land use. This information was used to create SOC maps of the study area and to calculate the total size of the soil carbon reservoir. The average SOCstock to 1m depth in the study area is 73 Mg.ha-1, which is comparable to other studies in Eastern Africa. The soils that can store the highestamounts of SOC in the study area are Vertisols, Phaeozems, Umbrisols and Nitisols (average SOC stock of 106 Mg.ha-1). Although they cover less than 40% of the total area, they contribute to almost 60% of the total SOC stock in 2005. However, they also contributed to 50% of the total SOCloss due to land use changes between 1997 and 2005. This can be mainly attributed to the loamy Phaeozems and Umbrisols, which have a high SOC content in their A-horizon but which offer no protection in organo-mineral complexes such as in clayey soils. Alisols, Acrisols, Lixisols, Ferralsols, Arenosols and Ferralic/Arenic Phaeozems have significantly lower SOC stocks (average of 87 Mg.ha-1). SOC stocks of Gleysols are intermediate at 87 Mg.ha-1, which is mainly the result of the impeded growth of bushlands on these soils.Remnants of closed forest, woodlands and bushlands of more than 20 years have SOC stocks between 66 and 70 Mg.ha-1. The relatively low SOC stocks can be attributed to the retaining of natural vegetation on the poorest soils, to the open structure of the vegetation with often little ground cover and to the disturbances by humans and fire. However, if woodlands are preserved on fertile soils they can have high SOC stocks, for example 120 Mg.ha-1 on Nitisols.The traditional cropping system, which consists of several years of cultivation followed by several years of bush fallow, is able to maintain high SOC stocks (120 Mg.ha-1 in young fallow bushland 98 Mg.ha-1 in cropped fields without cassava). On less fertile soils and when cropping becomes continuous, cassava becomes most important, since it is the most resistant to drought and low fertility. SOC stocks under cassava are less than half (39 Mg.ha-1) of those under other crops. Shortening of the fallow period and conversion permanent agriculture will have serious implications on the SOC stocks. Another threat for the regional SOC stocks is the expansion of cashew plantations, which in 2005 already covered 33% ofthe study area. Tree crops have the second lowest average SOC stock (50Mg.ha-1) which can be attributed to the removal of grass, undergrowth and litter by farmers to facilitate harvesting of cashew seeds. Planting other crops between the young trees (<20 yr) enhances the mean SOC stocks (to 79 Mg.ha-1) but after 20 years this effect completely disappears, since crops can only be grown in the small spaces that remain between the old trees. Grasslands, rice fields and homestead gardens are less common land uses, which occur mostly on the fertile soils and which have among the highest SOC stocks (mean total values between 114 and 128 Mg.ha-1). Total SOC stocks of the study area were estimated at 34.50 Tg in 1997. Based on the changes in land use during 8 years time it was estimated that this stock reduced with 14%. It can be concluded that intensification of the agricultural system by reducing the fallow period and plantingof mainly cassava on the depleted soils as well as the large expansion of the cashew monocultures have caused a serious loss of SOC from the study area. Restoration of the traditional fallow system, intercropping ofcashew trees with crops on a substantial area of the field, also after 20 years, and reforestation could be used to restore SOC stocks. Although sequestration of SOC through aforestation or reforestation proved favourable, these measures are restricted by the ability to produce rapid, cost-effective and precise sampling schemes. Reliable measurements of SOC content are hard to get and labor-intensive due to the spatial variability. A good knowledge of the soil C stock and its spatial variation leads to a sampling strategy that is balanced between reliability and cost-effectiveness, in agreement with the error that is allowed for the purpose of the study. Factors to take into account are sample size andsample density in relation to the spatial scale of the variation. Sampling intervals can be optimized for soil mapping and for calculating average values by application of classical and geostatistical theories. In tropical forest soils geostatistic analyses were not yet used to develop sampling strategies. Therefore a SOC variability study was conducted in five common forest/woodland/plantation types (coastal dry forest, miombowoodland, teak plantation, pine plantation and cashew plantation) usingconventional statistical methods, as well as geostatistics. In the 5 forest types of this study, SOC stocks in the upper 5 cm ranges between 5 (in the cashew plantation) and 13 (in the coastal forest) Mg ha-1.The optimal sampling distance for measuring mean SOC stocks varies between 36 m (in the patchy miombo woodland) and 422 m (in the homogenized cashew plantation). Sample sizes fluctuate between 6 and 72 (1 Mg ha-1 precision) for respectively cashew plantation and coastal forest. A rectangular grid with a sample interval of 25m can be used for SOC mapping with a point kriging estimation error of 3.0 Mg ha-1 in the coastal forest,2.6 Mg ha-1 in miombo woodland, 2.2 Mg ha-1 in the teak plantation and 1.1 Mg ha-1 in the cashew plantation. Since the pine plantation has no spatial structure; samples can be arranged randomly and its best soil maphas an average C content attributed over the whole field. Refining the sampling strategy with a new spatial variability study in other forest types can be based on a regular grid with sampling distances of half the range identified in this study. It was proven that sampling schemes varystrongly as a result of the different spatial behaviour of SOC in forests and depend on the required precision and research question. Only whenthe right strategy is followed, high standards of precision can be met without economic loss or risk of statistical misinterpretation. Besides spatial structures in SOC resulting from vegetation induced patterns, other factors could locally alter SOC spatial heterogeneity. One such factor in the Tanzanian landscape is the presence of termites and their mounds, which can cover 10% of the soil surface. Termites can influence the soil properties in different ways: they can alter the soil profile development, soil physical properties, soil chemical properties and soil microbiology and have an impact on plant growth. Both increases anddecreases in mound OC levels compared to the surrounding soil have been observed, probably depending on termite species and activity status, soil type and land use. Through the impact of termites on different soil properties, termite constructions and especially mounds can be considered as ‘islands’ in ecosystems where the diversity and quantity of litter transformers, micro-predators and micro-organisms are often different from those in the surrounding soil and where nutrient release can be either enhanced or blocked depending on the age and composition of the structure. These sources of small-scale heterogeneity are increasingly seen to be vital both in the maintenance of biodiversity as in providing farmers with a ‘least risk strategy’ for crop production. No information was yet available on the termite genera which occur in southeastern Tanzania, and how they specifically influence SOC on the different typical soils of the region. Therefore a characterization of the spatial patternsof SOC in and around termite mounds in a plot was made in order to adapt sampling strategies to these heterogeneities in the soil. Three different locations were selected which included the typical variation of soiltype, altitude and landscape unit of the region. In each location threeplots of 1 ha were selected with 3 common land uses: natural vegetation, crops and a cashew grove. On each plot termite diversity was studied based on the transect method, and SOC was measured in and around different types of mounds and in reference profile pits. Termite diversity decreased from natural forests to cashew plantations and agricultural fields,reflecting the change in canopy cover and resulting effects on soil humidity. Soil feeding genera are most affected by conversion from forest to crops. On the other hand cropping favours many genera known as pests. On average 43 (small) Termitinae mounds and 5 (large) Macrotermitinae mounds are reported per ha. Organic carbon levels of mounds are comparable to those found in literature. OC in the aboveground part of Macrotermitinae mounds (large mounds) are on average 1.5 times higher than the reference micro-pits (0-40 cm). This generally small increase results from the use of saliva in building the mounds as described in literature. Mound carbon contents reflect the amount of SOC in the surrounding subsoil, which is the main building material for Macrotermitinae. OC levelsof Termitinae mounds (small mounds) are on average 3.5 times higher than the reference micro-pits, which is the result of incorporation of faeces in the mound material. The increased organic matter is stabilised in the mound through adsorption to the clay fraction. Calculating carbon densities of mounds reveals that on average a Macrotermitinae mound (aboveground) contains 751.8 kg C, while only 0.5 kg C is stored in an averageTermitinae mound. Over 1 ha these add up to 21kg OC stored in Termitinae mounds and 3759 kg OC in Macrotermitinae mounds. Thus, although Termitinae raise OC levels much more, due to the small volume of their mounds the effect on field level carbon stocks is much smaller. OC is not only enriched in the mound structures but also under and around mounds. On average the presence of termite mounds and termite modified soil increasesthe C density with 12 Mg ha-1 or 6%. In Arenosols and Ferralsols this could be of the magnitude of major land use changes, such as conversion from forest to cashew groves. Thus by overlooking the local hotspots created by termite mounds, SOC stocks can be strongly underestimated, and byimprecise sampling of these local heterogeneities, land use change induced increases or decreases of SOC could be obscured. However by taking into account termite induced heterogeneity and by sampling the different landscape components separately, coefficients of variation can be reduced. By using the average ground cover distribution across the landscape, a weighted estimate of SOC can be made. Stratified sampling in and around these mounds is absolutely necessary to correctly estimate SOC stocks of these fields, and to be able to distinguish management or land use induced changes in SOC.