Teatime on Mount Kilimanjaro: Assessing climate and land-use effects on litter decomposition and stabilization using the Tea Bag Index

2.1 Study site

The study sites are located at the southwestern slope of Mt. Kilimanjaro (3°4′33″S, 37°21′12″E) and cover an elevation gradient from 951 to 4,190 m a.s.l (Table 1). Ten plots (0.25 to 1.00 ha) were selected, representing typical natural and agricultural ecosystems of the region as characterized by Hemp (2006a). The colline area, below 1,200 m, is naturally covered with savanna (SAV) woodland dominated by Acacia species (Becker et al., 2016). This natural vegetation is increasingly transformed into arable land for intensive maize (MAI) and sorghum production (Lambrechts, Woodley, Hemp, Hemp, & Nnyiti, 2002). The densely populated area between 1,200 m and 1,800 m is mainly covered by Chagga homegardens (HOM) and coffee plantations (COF). HOM are multilayered agroforestry systems with Musa ssp. and Coffea ssp. as dominant crops under fruit and remnant forest trees (e.g., Albizia schimperiana, Grevillea robusta; Hemp, 2006b). They are traditionally managed with sporadic addition of organic fertilizers and household waste (Fernandes, Oktingati, & Maghembe, 1986). Shade-coffee plantations are an intensively managed land-use type, with regular application of mineral fertilizers and pesticides. We categorized land-use intensity of these sites according to the indices proposed and calculated by Classen et al. (2015) and Schellenberger Costa et al. (2017; Table S1). These indices consider factors such as annual biomass removal, input of fertilizers and pesticides, and vegetation structure as well as surrounding land-use types.

Five natural sites were located inside the Kilimanjaro National Park along the Machame and Umbwe ridges. The lower montane forest (FLM) at 1,920 m is dominated by Macaranga kilimandscharica, Agauria salicifolia, and occasional Ocotea usambarensis. Ocotea forest (FOC) at 2,120 m is defined by the lone dominance of O. usambarensis and tree fern, such as Cyathea manniana. The forest at 2,850 m was classified as Podocarpus forest (FPO) and is dominated by Podocarpus latifolius together with Prunus africana and Hagenia abyssinica. In the subalpine zone around 4,000 m Erica forest (FER), Erica trimera is dominating and can reach up to 10-m growth height. Between 4,000 and 4,500 m, the alpine forest is displaced by Helichrysum cussion vegetation with tussock grasses (HEL) (Ensslin et al., 2015). An additional HEL plot (HEL1) was added to represent the zone of ongoing vegetation shift between Erica and HEL.

Climate at Mt. Kilimanjaro follows a bimodal rainfall regime with a short rainy season between October and December and a longer rainy season from March to May (Hemp, 2006a). Interpolated, mean annual and monthly (2011–2014) meteorological data from the study sites are available from Appelhans et al. (2014). Seasonality and site-specific data are summarized in Table S3. Mean annual precipitation varies between 663 mm and about 2,500 mm per year (Table 1). Mean annual temperature ranges from 4.5 °C to 23.7 °C.

Soils at Mt. Kilimanjaro are of similar age and developed from similar parent material, which makes comparisons between these ecosystems especially valuable. In the colline zone, soils developed on erosion deposits from Mt. Kilimanjaro and were classified as Vertisols. Soils in the forest zone were classified as Andosols with folic, histic, or umbric topsoil horizons. Their C contents are accordingly high in the upper horizons, and these soils are often underlain by C rich paleosol sequences (Zech et al., 2014). In the alpine zone, dominating soil types are mainly Leptosols and Vitric Andosols. These soils developed from volcanic rocks, such as basalt, trachyte, and olivine basalts (Dawson, 1992).

2.2 Sampling and analyses

We used the TBI, as introduced by Keuskamp et al. (2013), to assess seasonal effects on decomposition of a standardized substrate. At each of the 10 plots, nine pairs of litterbags (green tea & rooibos tea) were buried in 8-cm depth along a 100-m transect parallel to the line of the slope. The litterbags were exposed for ~90 days before collection. This was repeated during the short-wet (October 2014–December 2014), warm-dry (December 2014–March 2015), long-wet (March–June 2015), and cold-dry season (June–September 2015; Figure 1). Due to logistic restrictions, we slightly adapted the TBI protocol and dried the recovered teabags at 60 °C for 48 hr before weighing.

TBI is based on the decomposition rate constant (k) and stabilization factor (S). Both were calculated according to Keuskamp et al. (2013). In short, the k value is calculated from mass loss W after incubation time t, assuming a double-exponential decomposition due to faster decomposition of hydrolysable fractions (a) and relative increase of the more recalcitrant fraction (1-a) over time (Equation 1).

(1)

Environmental conditions can alter the stability of less recalcitrant compounds, reducing the mass loss of the originally hydrolysable (i.e., chemically labile) fraction. This inhibiting effect is therefore referred to as S (Equation 2), with a_g being the decomposed fraction and H_g the hydrolysable fraction of green tea.

(2)

2.3 Statistical analyses

The effect of elevation was assessed by linear regression at p level ≤ .05. The polynomial degree of the model fit was determined using Akaike's Information Criterion on linear, second-order, and third-order models. We identified seasonal variations by comparing slopes and intercepts of the final regression models using analysis of covariance (p ≤ .05). Effects of land use were compared separately for each elevation class (colline and montane). Significant effects were determined by using linear mixed effect model analysis of variance for nested designs with season as random factor (p ≤ .05). Seasonality of both TBI parameters (k and S) was related to seasonal amount of precipitation and mean temperature in each ecosystem using partial correlation to correct for T and P, respectively (Table S1). Continuous measurements of climatic variables were available only from SAV, FLM, FPO, and FER (Table S2), thus we limited our analysis to these sites. All statistical analyses were conducted in R 3.3.1 (R Core Team, 2016).

3.1 Effect of elevation

Mean annual decomposition rate constant k decreased logarithmically with increasing stabilization factor S (Figure 2). Average S values were highest in alpine (HEL) and sub-alpine FER ecosystems as well as in SAV. FOC exhibited the maximal k values.

Annual means of k and S were strongly affected by elevation (Figure 3). These relationships were best explained by left skewed third-order (or higher) polynomial functions (Table S2), indicating stronger effects within the colline and lower-montane zones compared with the montane and alpine zones.

Mean decomposition rate increased from k = 0.007 in SAV up to a maximum of k = 0.022 in FOC. The increase of k was followed by its decrease to around k = 0.010 in the (sub-)alpine ecosystems. Stabilization factor decreased from SAV (S = 0.33) to COF or FOC (S = 0.11) and strongly increased again to a maximum of S = 0.41 in the alpine HEL ecosystem.

3.2 Effect of seasonality

During all seasons, we found the highest decomposition rates in the midelevation forest belt (Figure 4). However, during both warm seasons, the peak is shifted upslope.

Regression slopes between k and elevation differed significantly between seasons (p ≤ .05). Maximum k values in cold-wet and cold-dry season were found at 2,220 m in FOC. During the warm-dry season, k peaks at 2,850 m (FPO). At most sites below 2,220 m, seasonal maxima were found during the longer cold-wet season with the highest precipitation. While at higher elevation, maxima occurred solely during the warm-wet season.

Seasonal variations strongly affected stabilization factor in SAV (Figure 4). In all ecosystems, the S-factor values were highest during the cold dry season. This seasonality was less influential at midelevation. Both, highest and lowest S values were measured for wet and dry season in SAV, respectively. The mean S values in SAV during cold and dry season (S = 0.54) were about 23 times higher compared with the warm-wet season (S = 0.02). The lowest seasonal fluctuation was measured for FOC, where S varied between 0.13 during the cold-dry season and 0.09 during the warm-wet season. Partial correlation between k in natural ecosystems and precipitation was significant (p ≤ .05) except at midelevation (FPO; Figure 5). At mid and high elevation (FPO & FER), k was significantly affected by temperature. The correlation between stabilization factor and seasonal precipitation linearly decreased with elevation. Contrary, the stabilization factor was significantly affected by temperature, already at FLM and above.

3.3 Effects of land use

Land-use intensification slightly increased decomposition rates and significantly decreased S values (Figure 6). In both elevation zones, mean annual k values increased by about 30% with land-use intensity, but these effects were not significant when considering seasonal variations (SAV–MAI: p = .14 & HOM–COF: p = .16). Mean annual stabilization factor in the colline zone decreased from 0.33 in SAV to 0.22 in MAI. Likewise, the stabilization factor in COF was around 20% lower compared with HOM.

4.1 Evaluation of TBI indices

All measured values of k and S and their variances were in a similar range as global reference data derived from Keuskamp et al. (2013), but mean annual k values were mainly on the lower half (Figure 2).

The mean k values in the Kilimanjaro forest belt (i.e., FOC, FLM, and FPO) were comparable with temperate forest sites from Keuskamp et al. (2013) but were not as high as in tropical moist or lowland forests. Lower slopes of Kilimanjaro region are under stronger water limitation than lowland forests in Central and South American tropics (Legates & Willmott, 1990), while lower mean annual temperature at high elevation restricts decomposition. Annual S means covered the whole range of global reference values. In cold alpine and semiarid SAV ecosystems, S was higher compared with most reference sites (except desert). This supports the underlying assumption of the TBI, that S is depending on environmental and climatic factors (Berg & Meentemeyer, 2002) and can reflect climatic limitation.

The TBI appears to be a valid and reproducible method for estimating decomposition rates and C stabilization potential at Mt. Kilimanjaro, and our results are consistent within this context. However, further improvements of the TBI method might be recommended (Didion et al., 2016). Measurements are limited to 3 months of incubation but are highly sensitive to seasonal fluctuations. If the TBI data should contribute to a global annual modelling, this should be considered in method standardization.

4.2 Effects of elevation

Elevation (i.e., climatic conditions) had a strong effect on decomposition rate and stabilization factor (Figure 7). Both parameters have their critical values at midelevation: the decomposition rate k—its maximum and the stabilization factor S—its minimum (Figure 3).

Unimodal and U-shaped patterns are typical for various ecosystem properties along montane elevation gradients (Campos, Etchevers, Oleschko, & Hidalgo, 2014; Kluge, Kessler, & Dunn, 2006). Peaks at midelevation were recently found for photosynthesis (NDVI), soil C content, litter quality, and species abundance at Mt. Kilimanjaro (Becker, Pabst, Mnyonga, & Kuzyakov, 2015; Hemp, 2006a; Pabst, Kühnel, & Kuzyakov, 2013; Röder et al., 2016). Especially the distribution of aboveground biomass is distinctly hump shaped at Mt. Kilimanjaro (Ensslin et al., 2015). The maximum occurs in FLM and FOC, between 2,000 and 2,500-m elevation. This midelevation peak of ecosystem productivity is highly correlated with precipitation, that is, water availability, (Röder et al., 2016), and it can be directly linked to decomposition patterns (Figure 3).

Seasonal temperature variations start to affect C stabilization at FLM (1,920 m) and become increasingly important at higher elevation (Figure 5). Precipitation can be seasonally limiting below FPO (<2,850 m). However, FLM and FOC represent the interception zone between mostly sufficient moisture availability and temperature. This indicates that litter-C stabilization in these ecosystems is mainly driven by amounts of litter input and productivity. At lower and higher elevation, decomposition is restrained by climatic factors.

Ecosystems at lower elevation are highly subjected to seasonal moisture limitation (Appelhans et al., 2016). Especially in semiarid environments, low water availability negatively affects litter decay rates (Incerti et al., 2011). During the rainy season, soil microbial activity in SAV strongly increases (Otieno, K'Otuto, Maina, Kuzyakov, & Onyango, 2010), and the turnover is less selective regarding organic matter quality (Davidson & Janssens, 2006). This effect is only present in semiarid elevation zones (i.e., colline and submontane). At midelevation, S values were low and unaffected by seasonality, thus, the preference of easily available substrate was rather constant throughout the year.

In upper montane and alpine environments (≥2,850 m), decomposition was strongly limited by temperature (Figure 5) and increased during the warm seasons (Figure 4). This is commonly expected because temperature sensitivity of decomposition is generally higher at low temperatures (Davidson & Janssens, 2006) and at higher elevation (Blagodatskaya, Blagodatsky, Khomyakov, Myachina, & Kuzyakov, 2016; Schindlbacher et al., 2010). Another factor that might reduce decomposition specifically in FPO (2,850 m) is the regular water logging of soil due to clouds inhibiting evaporation (Bruijnzeel & Veneklaas, 1998). However, neither negative nor positive effects of precipitation were found during the seasons (Figure 5). Strong seasonality in Erica and HEL ecosystems implies strong dependency on climate variables and low potential to adapt to fast climate changes compared with lower elevation forests (Hemp & Beck, 2001). The projected increase of surface temperature (Bradley, Vuille, Diaz, & Vergara, 2006) will reduce the stabilization of fresh C and accelerate organic matter decomposition. Therefore, future soil C losses into the atmosphere might be considerably large and fast in East African mountain ecosystems.

4.3 Effects of land use

Land-use intensification from seminatural SAV to MAI monocultures and from traditional HOM to large-scale COF decreased C stabilization and showed the tendency to increase decomposition rates (Figure 6 and Figure 7). The total content of soil organic matter (SOM) and microbial biomass commonly decrease with land-use intensification (Don, Schumacher, & Freibauer, 2011; Junior et al., 2016). This effect was also found at Mt. Kilimanjaro (Pabst et al., 2013). However, at the same time decomposition rates at Mt. Kilimanjaro tended to increase while C stabilization decreased. This is in contrast to previous findings that connected land-use intensification to decreasing decomposition rates (Attignon et al., 2004; Violita, Triadiati, Anas, & Miftahudin, 2016). Under similar environmental conditions as compared with the lower slopes of Mt. Kilimanjaro (i.e., western Kenya, 1,500 m), Kagezi et al. (2016) found decreased decomposition rates on agricultural compared with natural sites. This decrease of organic matter decomposition can be connected to the application of N fertilizers and reduced microbial biomass (Zang, Wang, & Kuzyakov, 2016). Decomposition studies tend to exhibit strong site and method specific variation (Makkonen et al., 2012), and land-use intensification was likewise found to increase decomposition of litter and SOM (Guillaume, Damris, & Kuzyakov, 2015; Lisanework & Michelsen, 1994). Decreasing decomposition with higher land-use intensity are often related to changes in decomposer communities (Kagezi et al., 2016). Recent studies from Mt. Kilimanjaro found only minor effects of land-use change on overall arthropod abundance and composition (Röder et al., 2016) but indicated accelerated organic matter turnover on agricultural sites (Becker et al., 2015). Also, glucose decomposition increases with land-use intensification from SAV to MAI fields and HOM to COF (Mganga & Kuzyakov, 2014). This is because soil microbes in these ecosystems are less efficient in SOM decomposition but at the same time more demanding for new C sources (Pabst, Gerschlauer, Kiese, & Kuzyakov, 2016), reducing S values on agricultural sites (Figure 6). This concept relates decomposition patterns primarily to the microbial decomposers nutritional status (Manzoni, Jackson, Trofymow, & Porporato, 2008). Considering the features of the TBI method (i.e., standardized litter, exclosure of exogeic, and >0.25-mm fauna), this points out the importance of pre-existing soil nutrient conditions on litter decomposition and C stabilization.