Introduction

Soil and vegetation are complexly interrelated so the growth and reproduction of terrestrial ecosystems such as forests cannot be understood without knowledge about the soil. The selective absorption of nutrient elements by different tree species and their capacity to return them to the soil brings about changes in soil properties (Singh et al. 1986). Therefore, the conversion of forests to other forms of land use types has profound effects on the distribution and supply of soil nutrients by directly altering soil properties and influencing biological transformations in the rooting zone (Murty et al. 2002). These modifications are more rapid for chemical than for physical properties of soils (Schipper and Sparling 2000) and can be more significant in the humid tropical areas due to high rainfall and temperature variation (Shepherd et al. 2000; Takoutsing et al. 2014).

In the humid forest zone of Cameroon, slash-and-burn is the principal form of natural forest conversion for agricultural purposes (Nolte et al. 2001). In practice, part of the forest is under-brushed; some of the remnant trees are felled, burned and used as ash fertilizer. But several other trees are maintained in the field and crops are cultivated under them. A mixture of annual food crops [maize (Zea mays), melon (Cucumeropsis melo) and groundnuts (Arachis hypogeae)] and perennial crops [cocoa (Theobroma cacao)] are established, usually with minimum soil surface disturbance for one to three cropping cycles. These crops are usually combined with a variety of forest products of economic and nutritional importance to the households. Towards the end of the cropping interval, competitive perennials such as banana are planted and these become a feature in pioneer vegetation and later successional events.

Thereafter, cultivated lands used for food crops are allowed to fallow for a period of time while another part of the forest is converted (Agoumé and Birang 2009). After cultivated lands are abandoned, forest succession results in large statured secondary forests within two to three decades. The species composition of the secondary forest shifts progressively toward that of the original forest over the next several decades until the two forest types become nearly indistinguishable.

Previous studies have demonstrated significant effects of land use types on selected soil physical, chemical and biological properties, and consequently on soil quality indicators (Kotto-Same et al. 1997; van Noordwijk et al. 1997; Shepherd et al. 2000; Tchienkoua and Zech 2004; Agoumé and Birang 2009). However, the knowledge of chemical properties of soils of different land use types of the humid forest zone of Cameroon is meager. Such information is useful for the detection of changes in soil quality indicators (Ishaq et al. 2002; Bewket and Stroosnijder 2003), as well as understanding how these changes influence global climatic systems and ecological processes (Abbasi et al. 2010). Vegetation in general, and tree species in particular have a greater influence on soil conditions than herbaceous plant species, due to input of litter with high lignin content, high total net primary production, and high water and nutrient demand (Binkley and Giardina 1998). Moreover, different tree species can differ significantly in their influence on soil properties as well as soil fertility (Augusto et al. 2002).

The objectives of this study were to (a) assess the chemical properties of soils in different land use types, and (b) determine if the correlations between the soil chemical properties were similar in the land use types. We hypothesized that the values for all soil chemical properties would be highest in land uses with higher tree density and lowest in the agricultural systems.

Materials and methods

Description of the study area

The study was conducted in the humid forest zone of Cameroon located between 2°48′ and 4°7′N, and between 10°55′ and 12°30′E (Fig. 1). The average elevation is 685 meters above sea level and the climate is characterized by two rainy seasons (March-June and September–November) and two dry seasons. Mean annual rainfall ranges from 1500 to 2000 mm while the average annual temperature lies between 22° and 25 °C (Gockowski et al. 2004). Most of the soils in the area are Ferralsols and Acrisols according to the World Reference Base (WRB) for Soil Resources classification system (IUSS 2006). Though generally acidic and infertile, these soils have been found to be suitable for selected food and perennial crops such as cocoa (Gockowski et al. 2004). Concentrated in the southern part of the country, the dense forests represent more than 36 % of the national area and were experiencing annual rates of deforestation of only 0.14 % between 1990 and 2000 (Duveiller et al. 2008; de Wasseige et al. 2009). However, the risk for future deforestation is high. Current deforestation is mainly caused by the expansion of agriculture in the form of slash-and-burn. The impact of the forestry sector remains limited as logging is sparse and selective, but is a cause of a degradation of ecosystem products and services supplied by forests (Bellassen and Gitz 2008; de Wasseige et al. 2009).

Fig. 1
figure 1

Map of Cameroon showing the study site (Adapted from Cameroon ASB/REALU benchmark country report, 2013)

Full size image

Selection of land use types

The general biophysical environment of the area can be characterized as a tropical humid jungle forest. Due to anthropogenic activities, the resulting spatial pattern of the area is a landscape mosaic. Transects were identified which contained the predominant land use types namely primary forest, large-statured secondary forest that was established at least 30 years ago adjacent to the primary forest, land cleared from the secondary forest 3 years ago and used for agricultural practices, land abandoned to fallow for 8–14, and mature cocoa plantations of at least 14 years old (Nolte et al. 2001). Each of these transects are referred to as ‘land use chronosequence’ where the effects across space substitute for those observed over time (Kotto-Same et al. 1997). Each chronosequence was located on near-level topography and within the same soil characteristics. Local farmers were consulted to obtain information about the age and cropping history of the different land uses types.

Description of vegetation under various land use types

The original primary forests consisted of well-spaced, large-statured trees with little understory and thick layers of surface litter. The dominant tree species within the original forests included Gossweilerodendron joveri, Erythrophloeum suaveolens, Detarium macrocarpum and Canarium schweinfurthii (Zapfack et al. 1996). The croplands (less than 3 years old) contained remnant trees from the original forest and crop mixtures such as maize (Zea mays), groundnuts (Arachis hypogaea), cassava (Manihot esculenta), cocoyams (Colocasia esculenta) and banana (Musa spp.). Annual crops were seldom planted after 2 years, but perennials such as cassava and banana remain in the croplands to compete with the encroaching weedy fallow. As a result, young fallows often contained fewer trees than recently cleared croplands. Bush fallows were well-established 5–8 years after land abandonment and were dominated by Chromolaena odorata. These were followed by tree fallows from 8 to 14 years after land abondonment, which consisted of a dense under-story and nearly closed canopy of pioneer tree species. The most common tree species in the tree fallows were Albizia zygia, Ficus exasperata and Myrianthus arboreus with understory of Aframomum longiscarpum and Haumania danckelmaniana. Chromolaena odorata was not observed in the tree fallows. The secondary forests, which were at least 30 years old, were closed-canopy and 15–20 m in height, and contained less herbaceous under-story than the tree fallows but large numbers of immature trees. Common tree species included Triplochiton scleroxylon, Terminalia supurba, Pycnanthus angolense and Lophira alata. The cocoa plantations were at least 14 years old, and contained tree species belonging to both original and secondary forests, which provided shade to the cocoa understory. Also present in the under-story were many ferns (Pteris spp.) and wild ginger (Costus afer).

Experimental design and soil sampling

Within the study area, six sites (Akonolinga, Ayos, Lekié1, Lekie2, Ebolowa and Efoulan) were chosen. In each site, four villages were selected based on settlement density, agricultural practices around the homestead, land use diversity and tree density. Field survey and soil sampling were then carried out in the villages using the quadrat approach. Nested sampling plots of variable sizes were adjusted to the four land-use types to be sampled (secondary forest, fallow, mixed cropping, and cocoa plantation). Subplots were set up in each selected main plot with considerations that attempt to capture within-plot heterogeneity. For land use types with high tree density (secondary forest and fallow), rectangular subplots (5 m × 40 m = 200 m2) were established within a main plot of one ha, avoiding the boundary of the plot. Rectangular plots have the advantage of including more of the within-plot heterogeneity, and thus are more representative in tree-based system. For land use types with low tree density (cocoa plantation and mixed cropping), square plots of (20 m × 25 m = 500 m2) were established.

In each of the 24 villages, soils were sampled in two randomly selected plots (0.5 × 0.5 m) with similar slope, gradient, and altitude in each land use type. The samples were considered true replicates given that the distance between them exceeded the spatial dependence (<13.5 m) of most soil chemical and microbial variables (Mariotte et al. 1997; Wang et al. 2011). Soil sampling was conducted using a cylinder of 100 m3. Soil samples were collected separately at depths of 0–10 and 10–30 cm in each of the two randomly selected plots. The two samples from the same depth were mixed to obtain composite samples of 500 g. In total, eight composite samples (two per land-use type) were obtained from each of the 24 villages that made up the study area, giving a total of 192 samples.

Soil samples laboratory and spectral analyses

The samples collected were air-dried and passed through a 2 mm sieve in preparation for mid-infrared (MIR) spectral analysis. A subset of 54 (30 %) soil samples were analysed using conventional laborartory methods for total carbon (TC), soil organic carbon (SOC), total nitrogen (TN), cation exchange capacity (CEC), boron (B), calcium (Ca), potassium (K) magnesium (Mg), and pH, while MIR spectroscopy was then used to estimate these properties in the remaining 134 samples. The analyses methods were as follows: soil pH in water was determined in a 1:2.5 (w/v) soil:water suspension; SOC and TC (%) were determined by chromic acid digestion and spectrophotometric analysis; TN (%) was determined from a wet acid digest and analyzed by colorimeter; exchangeable Ca, K and Mg were extracted using the Mehlich-3 procedure; CEC was determined by ammonium acetate at pH 7. Boron was extracted using CaCl2 and the resulting extract analyzed calorimetrically using a spectrophotometer. Then the 192 samples were scanned using a Bruker Alpha Drift FT MIR Spectrometer and were analyzed by mid infrared (MIR) diffuse reflectance spectroscopy using the OPUS Laboratory software (version 6.5; Bruker Optics GmbH). Measured values of the soil samples (30 %) selected for conventional laboratory analysis were calibrated to the first derivative of the reflectance spectra using partial least squares regression (PLSR) models developed in R software version 2.15.0 (R Development Core Team 2008). The regression models were then used to predict soil values for the rest of the samples under investigation. Table 1 shows the correlation coefficient and the root mean standard errors of calibration (RMSEC) between the wet chemistry and the MIR results. These results especially for the large correlation are to a large extent similar to those reported by (Shepherd and Walsh 2002; Terhoeven-Urselmans et al. 2010; Vågen et al. 2006).

Table 1 Coefficients of determination of partial least squares regression models for predicting soil chemical properties from spectral reflectance (MIR)
Full size table

Data analysis

The SAS® statistical package (SAS Institute Inc. 2004) was used for all analyses, and the significance level was α ≤ 0.05 for all tests. The distributions of all soil properties in their measurement scales were positively skewed, but the distributions were normal on the log10 scale (UNIVARIATE procedure). Values for all soil properties were therefore analyzed on the log10 scale.

Analysis of variance (MIXED procedure, restricted maximum likelihood estimation method) was used to determine if soil chemical properties differed significantly among land-use systems, soil depth, sites and villages. The ANOVA model included four main effects (land-use, soil depth, site, village nested in site) treated as fixed factors, and five two-way interactions (land-use by site, land-use by village nested in site, soil depth by site, soil depth by village nested in site, land-use by soil depth). The Tukey HSD (honestly significant difference) test was used to compare least-squares means for land-use systems, soil depths and sites.

Pearson correlation coefficients (CORR procedure) were used to investigate linear relationships between the soil chemical properties in order to better understand the possible factors that affected their buildup in the soil. Correlation coefficients were computed in each land-use system.

Results

Calibration and prediction of soil properties

Partial least-squares regression was used to determine if there was a significant relationship between spectral reflectance from MIR (independent variable) and soil properties of the 192 samples (dependent variables: 58 measured, 134 estimated using calibration model). A coefficient of determination (R2) close to one indicates that MIR is almost as good as the laboratory technique against which it is calibrated. All chemical properties were successfully estimated with R2 > 0.72 and acceptable root mean square errors (Table 1). This indicates that spectra are capable of picking up changes in soil properties not only across soil types but also across different land use types, and is consistent with research conducted by Forouzangohar et al. (2014).

Analysis of variance of soil chemical properties

The one-way ANOVA indicated that land use types, soil depth and sites exhibited significant effects on chemical concentration (Table 2). The effect of land use was significant for four properties (B, Ca, K, pH); whereas the effects of soil depth, site, and village nested in site were significant for all properties (Table 2). There was a significant interaction between land use types and site for all properties, revealing that the effect of land use types was not the same in all sites. Two other significant interactions were observed: the effect of land use was not the same in all villages for five properties (Ca, SOC, TC, TN, pH), and the effect of soil depth was not the same in all sites for five properties (Ca, Mg, SOC, TC, pH).

Table 2 Analysis of variance of soil chemical properties in different land-use systems, soil depths, sites and villages in the humid forest zone of southern Cameroon
Full size table

Among the four soil properties (B, Ca, K, and pH) that differed significantly among land use types, mean values tended to be higher in cocoa plantations followed by mixed cropping (Table 3). No significant differences were detected for pH values between cocoa plantation and mixed cropping or between secondary forest and fallow. TC, SOC, TN and CEC did not differ significantly among land use types (Table 2), but they were highest in secondary forest followed by fallow (Table 3).

Table 3 Means and standard deviations of soil chemical properties in different land-use systems, soil depths and sites in the humid forest zone of southern Cameroon
Full size table

There were no consistent trends in the mean of soil chemical properties among sites (Table 3). For example, total carbon and total nitrogen were highest in Ayos and Lekie II. Though the soils under the different land use types were all acidic, pH values were highest in Lekié I, Lekié II and Ebolowa sites, whereas CEC values were highest in Ayos, Lekié II, Ebolowa and Efoulan sites.

There was a consistent trend in the mean of properties between the two soil depths (Table 3). All the properties were higher in the topsoil (0–10 cm) than in the subsoil (10–30 cm).

Correlations between soil chemical properties

Correlations between soil chemical properties were positive and generally significant in the four land-use systems (Table 4). The non-significant correlations involved B and pH. Boron was not significantly correlated with Mg, SOC and TC in secondary forests, with SOC and TC in fallows, and with SOC, TC and TN in mixed cropping. In contrast, all correlations with B were significant in cocoa plantations. Regarding pH, the correlations with Mg, SOC, TC, TN and CEC were not significant in all four land-use systems, and the correlation with K was not significant in the secondary forest.

Table 4 Pearson correlation coefficients among soil chemical properties in four land-use systems in the humid forest zone of southern Cameroon
Full size table

Discussion

Effects of land use on soil chemical properties

Soil pH

A fertile soil generally has a pH range between 5.5 and 7.2, which makes the essential elements and nutrients available to the flora. Though all soils were acidic (pH < 6.4) in all land use types, soils in secondary forests and fallows had significantly lower pH than those in mixed cropping and cocoa plantation systems (Table 3).

The processes involved in creating the observed variation in pH among land use with diverse tree species are likely to be threefold: (1) interspecific differences in the production of organic acids from decomposing litter that change the relative quantities of exchangeable base (Ca, Mg) and acid (Al, Fe) cations in soils, (2) differences in cation uptake and allocation to biomass pools with differing turnover times, and (3) variation in the mineral content of the parent material (Gairola et al. 2012).

Higher organic acid production and lower soil pH have been noted in sites dominated by species whose litter is relatively recalcitrant to the decomposition process (Konova 1966). The high values of pH obtained in the cultivated lands can be attributed to the fact that most farms are established through the practice of slash and burn techniques. During the process, wood ashes produced have fertilizer value and the amount of ash varies with the species of wood used (Sotelo Montes et al. 2012). Wood ash has a very fine particle size, so it reacts rapidly and completely in the soil with the main effect as a liming agent (Lerner 2000). As the largest component of wood ash is calcium carbonate, it is possible that the wood ash is responsible for the higher values of pH in cultivated lands. In addition, it is known that soils formed under low tree density tend to be less acidic than soils formed under forests. This is probably because residues from evergreen trees (forest) are more acidic than those from deciduous of less dense ones.

We did not expect soil pH to be higher in the upper layer because we assumed that the upper layer would be more strongly leached by rainfall, resulting in the removal of metal cations (e.g., Ca, K and Mg) by dissolved carbonic acid and organic acids and their replacement with hydrogen ions. We suspect that the higher pH in the upper soil layer may be due to differences in organic matter content, but this was not measured in the study. However, it is commonly observed that the upper soil layer contains more organic matter than the lower soil layer (Höfle et al. 2013; Lai et al. 2013). The difference in pH between soil layers has been ascribed in previous studies to the differences in mineralization of organic matter, nitrification of the nitrogen fertilizer applied on the soil surface, and root exudation (Limousin and Tessier 2007).

The general acidity of the soils in the study area might be due in part to the high mean annual rainfall. Rainfall has been demonstrated to remove basic cations from the surface horizons through leaching, thereby lowering the pH and increasing the soil acidity (Foth 2006; Abua et al. 2010). However, this should be taken with caution, since it may take hundreds of years for new parent materials to become acidic under high rainfall. The general acidity of the soils may also be due to the chemical composition of parent materials. Soils in the study area are Ferralsols and Acrisols, and are composed of leuco-mesocratic gneisses with intrusions of pyroxenic diorites and doloritic gabbros. These are strongly-weathered soils in which acidity, high exchangeable aluminum and a low ratio of basic to total cations are the main limiting factors to permanent cropping systems.

Normally at low soil pH values (<5.5 units), strong soil acidity constrains the availability of most nutrient elements. The current intensity of agricultural activities in the study site however, suggests that soil pH on average is unlikely to have a negative impact on plant growth and agricultural productivity. Most crops produced in the area such as maize, seem to be acid-tolerant.

Cation exchange capacity (CEC)

Generally, the CEC of soils is influenced by the clay content, the soil pH and the amount of organic matter. Values for CEC in all land use types were found to be below the acceptable range of 15–25 me/100 g (Hazelton and Murphy 2007), indicating that the soils cannot hold a lot of cations (Fenton and Helyar 2007) and consequently are more likely to develop deficiencies in potassium, magnesium and other cations. Soils with CEC values within the acceptable range are expected to hold more nutrients, have greater water holding capacity, and more effectively buffer rapid changes in soil solution levels of these nutrients (Fenton and Helyar 2007). Since acceptable values of CEC normally indicate high organic matter content in the soil (Moore et al. 1998), we deduce that the soil in the study area had low organic matter content, but organic matter was not measured in the study.

No significant difference in CEC was detected among the four land-use types (Table 3) probably because CEC is an inherent soil characteristic and is difficult to alter significantly. The non-significance could also be attributed to the fact that most farmers in the study area practiced zero tillage; therefore surface soil disturbance is so minimal that its effect on CEC is negligible. In addition, trees were present in all land use types, and provided some organic matter content through litter.

Our results are inconsistent with those described by Mesfin (1980) and Gao and Chang (1996) who reported higher values of CEC in abandoned lands as compared to croplands, and attributed the differences to the organic matter content. Our results do not agree with those of Bewket and Stroosnijder (2003) and Nega (2006) who reported lower CEC in the surface than in the subsurface layers in soils. The authors attributed the differences to the intensity of agricultural production and the types of land preparation that alter the structure of the upper layer and promote leaching of nutrients.

Like the other soil properties considered in this study, CEC (meq/100 g soil) decreased from the upper to the lower soil depth in all land use types. We assume that the decrease is mainly attributed to the high organic matter content found in the upper soil layer as also reported by Sahu and Mishra (1994). In addition, the soils of the study area have been confirmed to be rich in clay content, and soil compaction and sealing of soil pores are highly expected. Consequently, nutrients responsible for high CEC are unable to move down to the lower layer through leaching.

Calcium (Ca), potassium (K) and magnesium (Mg)

The concentration of Ca and K in the soil significantly varied with land use types (Table 2). Higher values for both elements were recorded in soil under cocoa plantation followed by mixed cropping (Table 3). This is attributed mainly to the practice of slash and burn agriculture. When farmers burn the wood and other organic materials, basic cations in the ash are deposited on the soil, thereby increasing the base saturation (Voundi Nkana et al. 1998). It was also observed that some farmers use chemical fertilizers and herbicides in the farms, which could have also influenced the level of Ca and K in the soil. These two elements in the soil largely depend on the composition of parent rock materials, but the parent rock materials were similar in the study area. With the exception of the practices mentioned above, we cannot explain the differences in Ca and K among the land use types.

Magnesium did not show any significant difference among land use types, despite showing a higher value in secondary forest. This suggests that the conversion of secondary forest into other land use types may reduce the amount of Mg in the soil. The absence of significant difference in Mg among land use types could be simply due to low available Mg as it is highly sensitive with rainfall which is high in the study area. The relatively low Mg in cultivated systems could be attributed to their continuous removal with crop harvest. Assuming that soil organic matter is lower in cultivated systems, soil erosion could also account for the lower Mg content in cultivated systems.

The mean concentrations of Ca, K and Mg were significantly higher in the upper soil layer. This result probably reflects three factors that affect the level of these elements in the soils: ash deposits on the soil surface following the practice of slash-and-burn (Voundi Nkana et al. 1998); basic cations (K, Ca and Mg) pumped by vegetation from the lower to the upper soil layers; and application of waste and other forms of manure to the soil surface by farmers, although this was not a common practice. In addition, the three cations are known to behave very much the same in the soil. Their mobility in the soil is relatively low and consequently, loss of these cations through leaching should also be relatively low (Yimer et al. 2007).

Soil organic carbon (SOC) and total nitrogen (TN)

We expected higher carbon and nitrogen in secondary forests and fallows. Our expectation was based on the premises that croplands have lower inputs of organic matter and a higher rate of oxidation of soil organic matter due to agricultural activities, while soils in forest and fallow have higher inputs of carbon and nitrogen from roots and above-ground biomass (Reicosky and Forcella 1998; Yimer et al. 2007; Takoutsing et al. 2013). The non-significant differences observed in this study may be attributed to several factors: firstly, owing to the high clay content of the soils, the adsorption of organic molecules onto clay mineral surfaces provides a mechanism of organic carbon stabilization against microbial attack (Ladd et al. 1996); secondly, the concentrations of carbon and nitrogen accumulate slowly in forests as plant cover increases and produces increasingly larger amounts of litter and root biomass that decompose in the soil (García-Oliva et al. 1999); thirdly, mixed cropping systems are maintained for a maximum of 4 years and then put under fallow, and farmers practice shallow tillage which does not alter soil structure (Voundi Nkana et al. 1998). Therefore, we could assume that the organic matter in the soils of the study area is well protected against mineralization, and carbon in soils is physically protected from massive loss during felling and burning and may increase due to entry of incompletely combusted particulates depending on the intensity of the burn. Furthermore, decaying roots from felled and burned forest vegetation may contribute to soil organic matter following conversion to agriculture and subsequent fallows. This view is supported by that of García-Oliva et al. (1999) and Shrestha et al. (2006) who found that only the labile organic carbon associated in macro aggregates was destroyed by forest burning, whereas the other organic carbon fractions were not affected. Further studies are needed to show whether qualitative organic carbon changes in our study area took place after conversion from forest to croplands.

Our results are however inconsistent with those of McGrath et al. (2001) and Powers and Veldkamp (2005) who reported significant soil carbon reduction in tropical soils after the conversion of forests to croplands, and with those of Dominy et al. (2002) and Rasiah et al. (2004) who reported large decrease of soil organic matter after conversion of undisturbed forest and grassland into sugarcane plantations. Therefore, it is difficult to generalize about changes in SOC following land conversion.

SOC and TN were higher in upper soil layer, compared with the lower soil layer. Nitrogen is mostly present in the soil in the form of nitrates, and consequently, it is expected to be very mobile and move freely with moisture (Gupta and Sharma 2008). The higher nitrogen content in the upper soil layer may be attributed to higher water holding capacity due to the presence of litter and humus in the upper soil layer of the land use types.

Boron (B)

Of the mineral elements essential for plant growth, boron is frequently considered to have a minor role since it is used in relatively small quantities in plant metabolism. This, however, does not diminish the importance of maintaining adequate boron in soils. Too often, there is a tendency to concentrate on the big three mineral elements–nitrogen, phosphorus and potassium—while assuming that remaining minor element levels are adequate (Muntean 2005). We found boron level in all land use types to be within the acceptable range of 0.15–0.50 ppm (Muntean 2005), indicating that the risk for toxicity is minimal and there is no need for adding supplemental boron in the study area. Boron levels were significantly greater in the cocoa plantations and mixed cropping (Table 3). Though we may not have a clear explanation for these results, it is believed that much of the available soil boron is held tightly by soil organic material. As organic matter decomposition occurs boron is released with a portion being absorbed by plants or leached below the root zone area (especially in high rainfall/acid soil areas).

The high concentration of boron in cultivated lands could be attributed to the wood ash produced from slash and burn agriculture. Farmers use the ash as a means of improving the fertility of the land, and the ash content has trace amounts of micro-nutrients including boron. We could also attribute the higher concentration in the cocoa plantations to the use of chemical fertilizers and herbicides that are widely used in cocoa plantations in the study area. Such chemical products are known to contain trace amounts of boron. Further studies are needed to assess the influence of the applications of these products on the level of boron in the soil and its interaction with other nutrients.

We expected B concentration to increase with soil depth. This was based on the fact that B is known to be very vulnerable to leaching, and with the high rainfall in the study area, one would expect a higher concentration in the lower layer of the soil (Zia et al. 2006). However, B concentration was higher in the upper soil layer. This could be attributed to the higher content of organic matter in the upper layer. Previous studies have also reported higher concentrations of B near the soil surface where organic matter concentrations are higher (Marschner 1995; Miwa and Fujiwara 2010), confirming that soil organic matter could be the main storehouse for most of the B in the soil (Goldberg and Suarez 2012; Steiner and Lana 2013). In addition, soils of the study area are rich in clay content and consequently have fine texture, which is less susceptible to leaching.

Correlations between soil chemical properties

Pearson correlations between soil chemical properties were positive and generally significant in the four land use types (Table 4). SOC showed a positive correlation with TN in all land use types. The correlations between SOC and TN were expected because these properties are related to the amount of organic matter in the soil (Kahle et al. 2002; Takoutsing et al. 2013). Previous studies have demonstrated that soil nitrogen influences carbon stock in the soil through plant growth and litter quantity, and stabilizes soil carbon decomposition particularly in tropical areas (Nave et al. 2009). The dynamics of soil nitrogen are closely related to that of carbon: thus, any change in the level of carbon may cause a change in the level of nitrogen in the soil (Qi et al. 2007; Moges and Holden 2008).

Cation exchange capacity was positively correlated with all soil properties expect pH in all land use types (Table 4). Positive relationships were expected because most of the soil chemical properties depend on the CEC values. These consistent positive relationships and similarities among land use types also imply that soil properties in the study area are influenced by similar climatic, pedogenic, anthropogenic and biotic factors that are likely to affect their levels and their interactions in the soil.

The correlation between CEC and pH was not significant in any of the land use types (Table 4). This was unexpected because there tends to be a positive relationship between these two properties (Havlin 2009). The result suggests that CEC was not affected by pH alone, but also by other properties such as soil texture and organic matter content which were not measured in this study.

Boron was positively correlated with all soil chemical properties in the cocoa plantations, but was not significantly correlated with Mg, SOC and TC in secondary forests; with SOC and TC in fallows; and with SOC, TC and TN in mixed-cropping systems (Table 4). The level of boron in the soil is increased through the use chemical fertilizers and herbicides. The differences in correlations in the four land use systems indicate that boron might be influenced by the concentration of other properties in the soil. Previous studies have demonstrated that significant positive correlations exist between boron, carbon and CEC (Moafpouryan and Shukla 2004).

Due to some identified limitations related to the sampling design and soil properties analyses used by this study, the results may be interpreted with some cautions. Firstly, only one composite sample was collected to represent each land use type in each study village. From agroforestry point of view, soil properties can vary at micro-scales, that is, within small distances due to interactions between trees and crops through different bio-physical processes. For this reason, one composite sample alone may not be able to fully represent a land use system. Fortunately, collection zones were widened and more soil samples (n = 192) were collected therefore, help in minimizing the shortcomings of the sampling design. Intensive sampling within each land use is highly recommended for further studies so as to account for sources of soil properties variation. Secondly, the good and satisfactory predictions reported above indicate a marked usefulness of MIR/PLS in the assessment of chemical characteristics of soils; however the results of the analysis should be taken with cautions because of the uncertainty associated with the method. Only 30 % of the samples were analysed using the conventional laboratory methods and the rest were predicted using the partial least square regression (PLSR) models. Steps undertaken in the procedure of IR calibration, if not controlled can be potential sources of errors in IR predictions (Batten, 1998; Singh et al. 2011; Olale et al. 2013), and plot to plot variation can contribute to considerably high variation within the dataset. The goal for future research should be to better identify robust, stable, and improved soil spectra correlations that can lead to reliable predictions.

Conclusion and recommendations

We investigated differences in soil chemical properties among land-use types and between soil depths in the humid forest zone of Cameroon. Land use types significantly affected B, Ca, K and pH: in general, these properties were higher in cocoa plantation and mixed cropping than in secondary forest and fallow. All soil properties were significantly greater in the upper than in the lower soil depth. With the exception of boron, correlations between soil properties were positive and significant in all land use types. There was considerable variation among sites, villages within sites and samples within land use types in the villages: this probably reflects differences in prior land-use practices, age of the land use, mode of management and vegetation.

The four land-use types did not differ significantly in some of the fundamental soil chemical properties that characterize soil quality and productivity such as SOC, TN, Mg, pH and CEC. These results are attributed to the fact that (1) the soil is replenished with nutrients from the plant biomass when secondary forests are slashed and burned; (2) the fallow systems are relatively long in the study area enabling the replenishment of soil nutrient levels, (3) farmers practice minimal or shallow land preparation techniques which do not affect soil structure; (4) some farmers add household wastes and other forms of manure to their farms.

The present study did not assess how soil chemical properties change when land is converted from one use to another. Consequently, further studies with more intensive soil sampling are needed to determine the effects of land-use conversion on soil properties (chemical and physical) over time, and estimate how long it takes for significant changes to occur.