We selected two study regions along the Albertine rift (Fig. ). Within each region, sites followed gradients of topography and land cover change, under similar climate but on contrasting soil parent material. The mafic region was located in the eastern Democratic Republic of Congo (DRC; further called “mafic region”). Late Miocene and Pliocene basalt deposits of volcanoes along the Mitumba mountain chain (10–2.6 Ma, ) and in the Virunga volcanic area north of the city Goma formed soil parent material with mafic geochemical features (high aluminum, iron, rock-derived base cations, and phosphorus (P) content; ). The sampled sites were located west of the city of Bukavu with latitude and longitude ranges of -2.1 to -2.6 and 28.6 to 28.9, respectively, and elevations ranging from 1360 to 2250 m a.s.l. Cropland dominates the lower altitudes and old-growth forest the higher altitudes. The felsic region was located in western Uganda (further called “felsic region”). In this region, soil parent material is primarily composed of felsic magmatic and metamorphic rocks (silica rich, gneissic granites; ) around the city of Fort Portal, in Rwenzori mountain foothills, Itwara and Matiri forest reserves, and in Kibale National Park . Mid-Holocene carbonate volcanism had occurred about 4700–5000 b.p. . This volcanic activity – specifically the deposition of ash and mudflow – has substantially influenced the topsoil characteristics. While soil types and parent material are otherwise comparable across the region, the local distribution of this volcanic deposition is highly variable (e.g. due to topography, wind patterns, and precipitation) and was unknown and unquantifiable prior to sampling. The study sites in the felsic region are located at latitudes of 0.5–0.8, longitudes of 30.1–30.9, and altitudes of 1230–1760 m a.s.l. (Fig. ).

Mean annual precipitation (MAP) in the mafic region is 1700 mm, with a rainy season from September to May and a dry season from June to August; mean annual temperature (MAT) is 17.5 °C. In the felsic region, MAP is 1300 mm and MAT is 20.4 °C . In the mafic region, soils are classified as Ferralsols and Acrisols. Nitisols, Phaeozems, Acrisols, and Luvisols dominate the felsic region . Numerous studies have analyzed volcanic materials from the Fort Portal region and found significant concentrations of primary minerals e.g.,. More detailed information on soil coverage and soil properties in the region can be found in . The topography of both regions is dominated by hillslopes with gentle to steep inclinations (mafic: 10–35 °, felsic: 9–24 °; Table ). This steep topography causes soil erosion and landslides . In both regions, hillslopes which were cleared >60 years ago showed yellowish to red colors of the iron oxides (e.g. goethite and hematite) and the dark topsoils were completely absent. Recently cleared sites were often characterized by remnant burnt tree trunks and charcoal particles. Note that also forest seedlings were still growing between the crops.

The main crops in both regions were cassava and maize, with smaller quantities of beans, sorghum, potatoes, groundnuts, coffee, onions, and amaranth. In the felsic region, all visited cropland sites were actively cropped. Here, crops still provide adequate yields (maize yields: 2–2.5 tha-1yr-1 in the Rwenzori foothills in western Uganda ). In contrast, in the mafic region, crops perform poorly (maize yields: 0.8–2.1 tha-1yr-1 in Walungu, Kabare and Kalehe territories of South Kivu ). Consequently, numerous hillslopes in the mafic region were abandoned from cropping due to infertility (information from farmers). A common practice in the mafic region is to plant Eucalyptus on heavily degraded hillslopes. In the felsic region Eucalyptus plantations are planted regardless of soil status in order to provide fuel for tea plantations. Eucalyptus in both regions were grown in monocultures, and sometimes grass, sedges, and rushes grew between the trees. The most common Eucalyptus species (family Myrtaceae, genus Eucalyptus) which are grown in the region are Eucalyptus globulus, Eucalyptus cinerea, and Eucalyptus grandis. For simplicity, we will refer to the genus name Eucalyptus only. Abandoned sites were either bare soil or covered with grasses and shrubland vegetation.

To assess years since deforestation of old-growth or pristine forest (on all cleared hillslope sites, i.e. cropland, abandoned, and Eucalyptus), we used a combination of historical aerial imagery and Landsat products. For the mafic region we used the historical, panchromatic aerial photographs from 1955–1958 from the Royal Museum for Central Africa in Belgium. For the felsic region, aerial images from the years 1955–1961 were purchased at the National Collection of Aerial Photography (NCAP). For the years after 1960, satellite-derived Landsat imagery, which is available for Eastern Africa starting from the 80s (Landsat 4–8), historical imagery from Google Earth Pro (Version 7.3.5, last access: 12 October 2023) and the deforestation maps by were combined to visually determine years since deforestation for the period 1955–2020 for all sampling locations. This approach allowed us to generate five classes: (1) undisturbed forest (old-growth or pristine forest; also labeled as 0 years since deforestation), as well as altered sites (cropland, abandoned sites, Eucalyptus monocultures) at (2) 2–7 years, (3) 10–20 years, (4) 40–60 years, and (5) >60 years since deforestation. Due to the low temporal coverage and raster resolution of the available satellite imagery, time of planting, harvesting, and replanting of Eucalyptus could not be assessed. In summary, in the mafic region we sampled a total of 60 sites with 9 forest sites, 27 cropland sites, 6 abandoned and 18 Eucalyptus sites (Table ). In the felsic region, we sampled a total of 38 sites with 9 forest sites, 26 cropland sites, and 3 Eucalyptus sites (Table ). The different number of sampling sites between both regions stems from difficulties in finding abandoned sites in the felsic region.

To quantify the SOC loss relative to the old-growth forest reference profile, we employed a SOC content-matching approach. These estimations were only possible in the mafic region due to the lack of an observed effect of time since deforestation on SOC content in the felsic region. The SOC matching approach assumes that depth profiles of SOC contents and stabilization mechanisms in the original, pre-erosion hillslope profiles resembled the depth profiles of undisturbed forest profiles. Although forest reference profiles (hillslopes and plateaus) developed on the same parent material and did not show any indication of erosion (results not shown; see ), calculations based solely on SOC matching may overestimate erosion rates and thus need to be interpreted with care. In this study, we interpret SOC losses compared to forest reference profiles as the combined effect of land conversion and higher erosion rates under cropland. The procedure was as follows: For each hillslope sample, the equivalent depth in the forest profile was determined by identifying the depth in the reference profile where the SOC content matched the hillslope topsoil content. We used linear interpolation to estimate the equivalent depth in the forest profile between the measured reference 10 cm-increments. SOC stock loss was quantified by calculating the cumulative forest SOC stock above that specific depth. To account for the inherent spatial heterogeneity of the old-growth forest, we performed this matching for each hillslope across the mean and the ±1 standard deviation (SD) of the forest soil profiles. This generated a range of estimates for both the equivalent depth in the forest profile and the SOC stock loss. Hillslope-specific estimates were aggregated by land-use type and time since deforestation to calculate group means and standard errors for SOC content, equivalent depth in forest profile, and total SOC stock loss. The total standard error (SEtotal=SEspatial2+SEref2) for equivalent depth in forest profile and SOC stock loss were calculated by propagating the spatial variance of the hillslope replicates (SEspatial=σsamples/n) and the uncertainty inherited from the forest reference range SEref=1n∑i=1n((Refhigh,i-Reflow,i)/2n), where Ref_high and Ref_low represent the upper and lower estimates derived from the ±1 SD forest limits. Soil loss rates (cmyr-1) were calculated using the equivalent depth in forest profile of cropland sites with a known history of 40–60 years since deforestation (50±10 years). This rate was then applied to sites cleared >60 years ago to estimate their time since deforestation. All reported uncertainties for rates and ages represent propagated standard errors, incorporating the combined variance from the forest reference profiles, the sampled hillslopes, and the 20-year uncertainty window.

All statistical analyses and graphical visualizations were performed using the R language and RStudio . For importing the BRUKER binary files, we used the R package opusreader2 version 0.6.2.9000 , the package prospectr version 0.2.7 for infrared data resampling and pre-processing, and the package resemble version 2.2.3 for chemometrics and calibration modeling (memory-based learning; ). For mapping, we used the R packages ggplot2 version 3.5.1 , terra version 1.7-74 and geodata version 0.5-9 , the latter providing spatial data such as country borders, elevation and land cover raster. For testing the statistical significance of SOC, TN, P_resin, and ECEC differences between the land use and years since deforestation, Type III ANOVA was performed (unbalanced data) on the log-transformed SOC, TN, ECEC and P_resin data (to fulfill the model assumptions). A pairwise comparison was done using the Tukey-HSD test. The relative changes of SOC, TN, ECEC, and P_resin on the hillslopes were calculated compared to the forest means using (hillslope SOC content - mean forest SOC content) / mean forest SOC content × 100). Uncertainties were calculated using standard error of the means (standard error of the means / mean forest SOC content × 100).

In the mafic region, a significant decrease of topsoil SOC content with time after deforestation was observed (Fig. ). Average SOC content on mafic parent material ranged from 64.4 gCkg-1 under forest decreasing to 23.3 gCkg-1 under cropland cleared >60 years ago. Abandoned fields had an average SOC content of 19.7 gCkg-1, while soil under Eucalyptus, cleared 40–60 years and >60 years ago, had an average of 40.5 and 38.1 gCkg-1, respectively. Mean relative change (± standard error of the means) in topsoil SOC content compared to the forest mean were -64 % (±9 %) on cropland cleared >60 years ago, -69 % (±15 %) on abandoned sites, and -41 % (±12 %) in soil under Eucalyptus monoculture cleared >60 years ago (Fig. ).

In topsoils of the felsic region, SOC did not exhibit a decrease along the deforestation chronosequence as in the mafic region. Nevertheless, SOC content in degraded cropland was similar to its equivalent in the mafic region. Forest topsoils in the felsic region (mean of 36.5 gCkg-1) did not show significantly higher SOC content than nearby croplands (mean of 30.9 gCkg-1), even though a slight, but not significant trend towards lower SOC content with time since deforestation was visible. Soils under Eucalyptus had the highest mean SOC content (59.6 gCkg-1) when sites were cleared 2–7 years ago and the lowest content (20.6 gCkg-1) when cleared >60 years ago (Fig. ).

In the mafic region, Fm ranged between 1.03–1.05 in the topsoils of the depth-explicit forest reference profiles and decreased to 0.72–0.78 at 60–70 cm soil depth. In forest reference profiles within the felsic region, Fm ranged between 1.04–1.07 in topsoil (0–10 cm) and decreased to 0.46–0.77 at 60–70 cm soil depth. Cropland topsoils in the mafic region cleared >60 years ago had lower Fm values (mean of 0.81 in 0–10 cm) than the corresponding forest reference topsoils (mean of 1.05; Fig. a). These measured topsoil cropland Fm values corresponded to the depths of 30–40 and 60–70 cm in the reference forest profiles (Fm = 0.68–0.85; Fig. a). In contrast, topsoils of cropland in the felsic region cleared >60 years ago had comparable Fm values to sites in the forest reference profiles (average of 1.01 in 0–10 cm).

The current SOC content in cropland, abandoned, and Eucalyptus monocultures topsoils cleared >60 years ago was equivalent to the SOC content of the reference profile at depths of 47±7, 59±13, and 24±5 cm soil depths, respectively (Fig. ). The corresponding losses of SOC stocks (cumulative SOC stocks above the calculated depths) are estimated to be 157±23, 192±44, and 85±23 tCha-1, respectively (Fig. ). We estimated that the soil above the depth of 21±5 cm had been lost from the hillslope over the last 40–60 years indicating a loss rate of 0.41±0.13 cmyr-1. Based on this loss rate, we calculated conversion ages of 114±40 years for cropland sites and 145±56 years for abandoned sites (for the sites with >60 years since deforestation).

In the mafic region, numerous hillslopes were found to be abandoned due to infertility. This was not the case, however, in the felsic region, where only one highly degraded hillslope under Eucalyptus (cleared >60 years ago) could be sampled. Mean pH_CaCl2 values (mafic: 4.17, felsic: 5.24; Fig. b, f), and the sum of base cations (mafic: 5.21 cmolckg-1, felsic: 12.90 cmolckg-1; Fig. c, g) were lower in the mafic compared to the felsic region. Consequently, mean Al³⁺ values were generally higher in topsoils of the mafic than of the felsic region (mafic: 2.4 cmolckg-1, felsic: 0.4 cmolckg-1; Fig. a, e). In the mafic region, decreasing pH_CaCl2 significantly co-varied with decreasing sum of base cations (Fig. d) and increasing Al³⁺ (Fig. b). As the ECEC decreased, Al³⁺ increased (Fig. a). Sites with low pH_CaCl2 values, high Al³⁺, and low base cations were dominated by Eucalyptus monocultures, abandoned cropland sites, and forests (Fig. a, b). For soils in the felsic region, pH_CaCl2 values significantly and positively correlated with the sum of base cations (Fig. g). In contrast, there was no clear relationship between exchangeable Al³⁺ and either pH_CaCl2 or ECEC (Fig. e, f). Topsoils under Eucalyptus monoculture, where conversion after deforestation happened just 2–7 years ago, had the highest Al³⁺ values in the felsic region (which were still much lower than for most soils in the mafic region; Fig. e, f).

ECEC was positively correlated with SOC in soils of the felsic region (R2=0.28, p<0.01) but not in the mafic region (no significant correlation; Fig. ), where ECEC values were extremely low (<10 cmolckg-1). Instead, there was a strong correlation of SOC with reactive Al and Fe (Al_PyOx and Fe_PyOx) in soils of the mafic region (Al_PyOx: R2=0.54 and p<0.01, Fe_PyOx: R2=0.36 and p<0.01; Fig. ) but not of the felsic region. Differences between the two geochemical regions were also observed for the content of these two reactive metal phases, with Al_PyOx ranging between 0.0 %–0.6 % and Fe_PyOx ranging between 0.0 %–2.1 % in soils of the felsic region and Al_PyOx ranging between 0.0 %–1.2 % and Fe_PyOx ranging between 0.0 %–2.7 % for soils of the mafic region. Similarly, we observed a decrease of reactive Al and Fe phases with time after deforestation for soils in the mafic region, while these differences were not observed for soils of the felsic region. Although clay content ranged from 28 %–76 % and 29 %–59 % in soils of the mafic and felsic region, respectively, SOC did not correlate positively with clay content in either region. In fact, a slight but significant negative correlation of SOC and clay was observed in the mafic region (R2=0.18, p<0.01; Fig. ).

A strong decrease of SOC content was observed along the deforestation chronosequence in the mafic region (Fig. ), which we attribute to post-deforestation erosion. It is widely accepted that deforestation causes severe losses of tropical SOC e.g.,. Particularly fast loss of SOC following forest conversion to cropland has been shown for a wide variety of tropical agroecosystems . However, these studies generally investigated geomorphologically more stable landforms (i.e. flat topography). Interestingly, our study in geomorphologically active and quickly eroding landscapes also shows that most SOC loss occurs within the first years after conversion (Figs. , , ), contradicting our first hypothesis where we expected more continuous losses of SOC over time after deforestation. Due to high erosion rates in the sloping landscapes of the study region, the relative change of SOC content is high compared to stable landforms, particularly for the mafic region (up to -69 %, Fig. ).

In the felsic region, soil fertility and volcanic activity exerted a stronger influence on SOC content than land cover. No clear effect of deforestation on SOC was observed: forest soils were found to have lower SOC content than the cropland sites converted 2–7 years ago (Figs. , ). This pattern likely reflects inherent pedogenic differences between land-use types rather than land-use change itself . Based on local observations, we hypothesize that croplands were historically established on more fertile soils, while forests were selectively left intact on less fertile soils. Although modern deforestation is less selective, this land-use bias remains difficult to decouple in such regional studies. Furthermore, the fact that soils in the felsic regions have developed on similar silica rich parent material, but experienced an uneven mineral replenishment from volcanic material has decoupled soil properties from the underlying lithology developed solely from the features of the parent material. This heterogeneous volcanic influence likely drives the observed geochemical variation among the sites. These findings suggest that soils with high fertility due to volcanic inputs can sustain longer and more productive farming and thus impede land abandonment. Lastly, there is no indication that the differences that we see consistently along the degradation gradient are overprinted by the climatic variability within each region (likely due to a small climate variability within each local gradient).

Our results highlight the distinct C stabilization mechanisms on weathered soils of the humid tropics, which are an understudied region . In contrast to many regions of the world clay and SOC content were not significantly correlated in either of the two geochemical regions (Fig. ). The lack of this relationship may be caused by the fact that old and deeply weathered soils of the humid tropics are dominated by clay-sized minerals with low reactive surface areas. Such clays form fewer organo-mineral associations and aggregates , which could provide protection against microbial predation of SOC . Negative or missing relationships between SOC contents and clay have also been observed in other parts of the tropics . Despite these results, clay is still often used as a universal predictor for SOC in many large-scale studies. This assumption undoubtedly leads to erroneous assumptions regarding the SOC stabilization potential of tropical soils e.g.,.

Particularly in the mafic region, our study emphasizes the importance of reactive metal phases in understanding SOC turnover and stability, as also highlighted by and (Fig. ). Sesquioxides can form resistant e.g., and old mineral-associated organic matter. The limited C sorption suggests that fresh C is minimally stabilized via new surfacing minerals (concept of the geomorphic pump; ). Several meters of weathered subsoil need to erode before less weathered minerals appear closer to the surface. However, our observations indicate that the thick subsoil layers have not fully been eroded (e.g. depths of 5–8 m, Figs. , ), and thus the weathering front has not yet reached the more nutrient-rich parent material. On one hand, fresh plant input to soil is scarce (mainly roots and root exudates) since most of the above-ground crop material in tropical African subsistence farms is harvested . On the other hand, fresh plant input that does end up being incorporated into the soil might be rather labile and decompose quickly. However, this is important as the apparent C saturation due to comparatively unreactive sesquioxides, combined with the strong (metal-related) stabilization of old SOC, likely explains the low SOC content of cropland and abandoned sites in the mafic region, which were cleared >60 years ago (∼ 19 gkg-1, Fig. ).

In the felsic region, the lack of reactive Al and Fe seemed to limit SOC stabilization, as we did not observe correlations between reactive metals and SOC (Fig. ). In this geochemical region reactive Al and Fe content that can support organo-mineral complexation over long timescales were substantially lower than in the mafic region. Comparatively young SOC ages in croplands of the felsic region (Fig. ) and the connection to key fertility indicators such as ECEC (Fig. ) highlight the importance of fresh C inputs to avoid degradation and the subsequent decrease in crop yields. This suggests that with a higher soil fertility and crop productivity than in the mafic region, more fresh C enters the SOC pool, despite a comparatively lower SOC stabilization potential. This interpretation is supported by findings in other tropical regions that show the residence time of organic C in soils is short when reactive minerals to stabilize C are lacking . More research is needed to better understand SOC stabilization mechanisms in the felsic region, for example by exploring how the region’s internal presence or absence of volcanic ashes (determined via x-ray fluorescence) or soil pH might impact SOC . Our second hypothesis (greater SOC stabilization in the mafic region due to higher contents of clay and reactive metals) can be rejected. Instead, the higher fertility levels in the felsic region likely lead to increased fresh organic matter input, resulting in higher SOC content. In contrast, the reactive surfaces in the mafic region are limited.

Our estimated soil loss rates, combined with their alignment with other studies, indicate minimal replacement of eroded SOC on cropland, Eucalyptus, and abandoned hillslope topsoils. Our FMC data confirm the low organic matter input into the SOC pool and highlight that cleared sites generally represent a deeper section of the same initial forest soil profile (see Fig. ). Interestingly, our calculated soil loss rates of 0.41±0.13 cmyr-1, which include the combined effect of land conversion and erosion, corresponds well with previous measurements of erosion rates of 0.3–0.7 cm per season (sum of reported rill and gully erosion; approx.: 0.6–1.1 cmyr-1) on croplands in South Kivu by . Using ^239+240Pu fallout radionuclides, estimated similar erosion rates of 0.4 cmyr-1 in two catchments in South Kivu (51.4 tha-1yr-1; assuming a bulk density of 1.25 gcm-3). Modeled erosion rates and estimates using remote sensing for the region were within the same range but more variable (30 tha-1yr-1 , 138 tha-1yr-1 ). Moreover, we found the presence of substantial colluvial deposits in the valleys of this region, with depths reaching up to 3 meters (work unpublished). These deposits further corroborate the extreme soil loss observed on the sampled hillslopes.

The soil loss rates and conversion ages reveal rapid degradation and abandonment of croplands on hillslopes, thereby underscoring the urgent need for sustainable farming. Interestingly, estimated that soil from an average depth of 0.5 m was eroded and mobilized into the rivers in fully deforested catchments in the same region, which would indicate a conversion age of approximately 122 years (given our average erosion rate of 0.41 cmyr-1). Overall, these estimated conversion ages correspond well with the reported history of the region, which was characterized by scattered settlements and kingdoms until the end of the 19th century and intensified deforestation with colonization in the beginning of 1900 . Ultimately, if hillslopes are abandoned after approximately 145±56 years, as our data indicate, then the region will face extreme challenges in the near future. A large proportion of the hillslopes around Bukavu were cleared in the early 20th century. These hillslopes thus already exceed the 100 years lifespan, leaving only a few decades left for farming if no measures are taken to preserve soil fertility and introduce more sustainable and conservation farming approaches. Unfortunately, these timespans are probably conservative, because population densities in the region rapidly increase . Furthermore, in the three territories of our study area (Walungu, Kabare and Kalehe), about 70 % of the agricultural land has inclinations of >10 ° (own analyses, Fig. ; sources: SRTM 30 m and ESA-CCI S2 prototype Land Cover 20 m map of Africa 2016).

Unfortunately, assessing SOC and soil loss in the felsic region proved challenging due to high and inherent system complexity and heterogeneity. Nonetheless, regional modeling suggests erosion rates comparable to those observed in the mafic region . This is supported by field trials reporting seasonal soil losses of up to 25 tha-1, matching the magnitudes measured at our mafic sites . Sustained erosion at these levels poses a critical threat to soil fertility and long-term agricultural productivity.

Low fertility, high soil acidity, and Al toxicity likely cause land abandonment of croplands in the mafic region. The low values of key fertility indicators in abandoned hillslope topsoils of the mafic region (ECEC, TN, and bioavailable-P/P_resin, Fig. ) indicate that sustainable management needs to address a multitude of elements to avoid land abandonment. Also, imbalanced nutrient stoichiometries such as low Presin:TN ratios (Fig. ) might negatively influence plant growth and microbial functions causing further unfavorable conditions for biological processes . Moreover, low Ca:Al ratios are unfavorable for plants , which is likely the case in soils of the mafic region in our study. In parallel, our data suggests that land abandonment may be driven by increasing soil acidity and Al toxicity, even if other fertility indicators such as SOC and TN remain favorable. As pH decreases towards values <4, a large proportion of Al is mobilized into soil solution as Al³⁺ , where it can be taken up and cause toxic conditions for crops . From informal conversations with farmers and our later soil data analyses, we learned that slopes with such low pH and high Al³⁺ were no longer used for cropping due to the low performance of crops, and instead abandoned or used for Eucalyptus monocultures.

Geochemical soil properties in both regions (Figs. , ) indicate that volcanic inputs slowed soil degradation in the felsic region, highlighting the role of geochemical differences in maintaining soil fertility. Mid-Holocene carbonate volcanism appears to prevent land abandonment due to the maintenance of acceptable pH values and soil fertility levels . This supports our third hypothesis that differences in the timescale of land abandonment are to be expected due to geochemically contrasting fertility conditions in soils derived from different parent material and varying levels of volcanic ashes. However, high erosion rates in the felsic region under current agricultural practices will likely still cause land abandonment in the near future on steep slopes due to a complete loss of the fertile, volcanic-ash enriched layers.

In the mafic region, our data suggest that the planting of Eucalyptus can restore SOC content to a level of croplands cleared 40–60 years ago (Fig. ), therefore recuperating only a small portion of SOC losses in topsoil caused by the initial deforestation and erosion. A similar increase of SOC under Eucalyptus has also been shown by , , and , however, others have observed no effect or even a decrease of SOC in Eucalyptus plantations . Our results support this variation in effect since recovered SOC content under Eucalyptus monocultures was highly variable. Thus, our fourth hypothesis (Eucalyptus monocultures improving SOC stocks) was partially confirmed, however not to the degree of full SOC recovery. For the felsic region, we could not investigate the effect of Eucalyptus due to the small sample size.

While Eucalyptus plantations can thrive on degraded soils, our results suggest that they may also reduce soil fertility by lowering pH and increasing Al³⁺ toxicity. The tolerance of Eucalyptus to heavily degraded soils has been reported for many tropical systems e.g.,. Subsistence farmers in South Kivu often plant Eucalyptus on hillslopes to generate viable income via charcoal or timber production once crops no longer provide reasonable yields on degraded land . However, the monoculture plantations of these non-native trees can have strong negative effects on soil fertility proxies. First, in the mafic region, soil pH_CaCl2 drastically decreased under Eucalyptus and thus reduced ECEC to very low levels (Fig. ). Due to the fact that Eucalyptus topsoil pH_CaCl2 and ECEC were lower than even those of abandoned hillslopes (Figs. , ), we argue that Eucalyptus plantations caused the additional acidification. Indeed, such a lowering of pH and ECEC by Eucalyptus has been shown before . Second, Al³⁺ saturation of the soil solution, which was on average 53 % across the Eucalyptus plantations (exchangeable Al³⁺ relatively to the ECEC; results not shown), was extremely high and toxic for most crops .

In the felsic region, negative effects of Eucalyptus on soil fertility have not been found , but our data suggest that soils under such plantations may eventually exhibit similar detrimental impacts as seen in the mafic region. However, our sample size for Eucalyptus sites in the felsic region is small for clear interpretations. Moreover, the majority of these sites were only recently cleared and likely installed on relatively fertile soil to support the local tea industry rather than as an alternative post-cropping land use. It therefore might take more time for soils of the felsic region under Eucalyptus to exhibit the same negative effects on pH and ECEC as in the mafic region. There is no evidence that suggests that plantations in the felsic region will show different fertility endpoints than described for the mafic region and elsewhere . The problem of soil degradation through Eucalyptus will further be exacerbated with planned expansions of Eucalyptus plantations . The lack of knowledge about precise plantation installation and harvest frequency adds limitations to our understanding of short and long-term effects of Eucalyptus plantations and urgently calls for more research. We conclude that the negative effects of Eucalyptus on soil fertility relativize partial improvements of SOC stocks that we observed which falsifies our fourth hypothesis (Eucalyptus monocultures improving soil fertility due to erosion control).