DataPDF Available
New Phytologist Supporting Information
1
Article title: Soil geochemistry - and not topography - as a major driver of carbon
2
allocation, stocks and dynamics in forests and soils of African tropical montane ecosystems
3
4
Authors: Benjamin Bukombe, Marijn Bauters, Pascal Boeckx, Landry Ntaboba Cizungu, Matthew
5
Cooper, Peter Fiener, Laurent Kidinda Kidinda, Isaac Makelele, Daniel Iragi Muhindo, Boris
6
Rewald, Kris Verheyen, Sebastian Doetterl
7
8
Article acceptance date: 14 August 2022
9
10
The following Supporting Information is available for this article:
11
Fig. S1 Contribution of each diameter class to the total number of trees per unit area
12
Fig. S2 Species composition and similarities across geochemical regions
13
Fig. S3 NPP and NPP allocation for three components across geochemical regions and along
14
topographic positions
15
Fig. S4 Correlations between NPP and NPP allocation
16
Fig. S5 C:N ratio of living leaves and litter layer for the three geochemical regions
17
Fig. S6 Relative root biomass along soil depth for the three geochemical regions
18
Table S1 Mineral soil properties of the three geochemical regions
19
Table S2 General plot information
20
Table S3 Forest stands characteristics across the three investigated geochemical regions
21
Table S4 Rotated principal components and their mechanistic interpretation
22
Methods S1 Detailed descriptions to study sites, existing data, protocols and assessment
23
methods
24
Notes S1 Results of forest structure and species composition
25
26
27
28
Fig. S1. Contribution of each diameter class to the total number of trees per unit area
29
The contribution to the average number of trees per hectare by diameter at breast height for the
30
felsic, mafic and sedimentary geochemical regions. Differences between regions (Mean ± SD; n=
31
12 per DBH class per region) were assessed separately for each DBH class with letters above bars
32
indicating statistically significant differences, following Kruskal-Wallis tests and pair-wise
33
comparison using Dunn’s test (p-value < 0.05). Data taken from Doetterl et al. (2021c).
34
35
36
37
38
39
40
41
42
43
Fig. S2. Species composition and similarities across geochemical regions
44
Non-metric multidimensional scaling (NMDS) on the inventory data to assess (dis)similarities and
45
the separation of species composition and abundance for the felsic, mafic and sedimentary
46
geochemical regions n = 12; 12; 12 for mafic, felsic and sedimentary, respectively. Forest tree
47
species abundance of each plot are defined in a two-dimensional space with NMDS1 and 2
48
representing variables after dimension reduction of data derived from the first forest inventory.
49
Values on X and Y axes are rank-based scores indicating ordination distance (similarity) between
50
points and regions. Points represent plotID(s) and shaded ellipses represent regions. Points that are
51
more similar to one another are ordinated closer together.
52
53
54
55
56
Fig. S3. NPP and NPP allocation for three components across geochemical regions and
57
along topographic positions
58
(a) NPP of litterfall, wood, and fine roots across the mafic, felsic and mixed sediment geochemical
59
regions (mafic, felsic and sedimentary) and along topographic positions (PL: plateau, UP: upper
60
slope, MS: middle slope and V: valley), (b) Relative C allocation for NPP of litterfall, wood and
61
fine roots across the felsic, mafic and sediment geochemical regions and along four topographic
62
positions (mean ± SD). Different letters on top of the stacked bars indicate significant differences
63
in NPP and C allocation between topographic positions following Kruskal-Wallis tests and pair-
64
wise comparison using Dunn’s test (p-value 0.05). Tests were conducted for each component
65
and geochemical region separately.
66
67
68
69
Fig. S4. Correlations between NPP and NPP allocation
70
Pearson correlations between NPP components (fine roots, litterfall, wood, and total), and the
71
corresponding relative NPP C allocation used in our analyses as response variables. White cells
72
indicate non-significant correlations, p-value ≤ 0.05.
73
74
75
76
77
78
79
80
81
82
Fig. S5. C:N ratio of living leaves and litter layer for the three geochemical regions
83
Plot based community weighted C:N ratio of living canopy leaves and mean CN ratio of the L
84
horizon litter layer across the three geochemical regions (n=12 per region). Different letters on top
85
of boxplots indicate significant differences in C:N ratio between regions following Kruskal-Wallis
86
tests and pair-wise comparison using Dunn’s test (p-value ≤ 0.05). The lower and upper T-shape
87
whiskers indicate the minimum and maximum C:N ratio respectively. The distance between the
88
lower and upper end of the box represents the interquartile range. Horizontal line in the middle of
89
the box represents the median C:N ratio and the white dot represents the mean C:N ratio. The tests
90
were conducted separately for living canopy leaves and the litter layer.
91
92
Fig. S6. Relative root biomass along soil depth for the three geochemical regions
93
Means of relative root biomass along soil depth for O-horizon and the top 50 cm of mineral soil
94
for the mafic, felsic and sedimentary regions expressed as proportion (%) of total root biomass.
95
The horizontal gray dashed line indicates the cumulative 90 % root biomass in the assessed soil
96
volume as basis for the selected geochemical soil properties used in the analyses. Number of
97
observations aggregated for each point equal n = 8; 12; 12 for mafic, felsic and sedimentary
98
respectively. Data taken from Doetterl et al. (2021c). Note that the thickness indicates the average
99
thickness of O-horizon at each site.
100
101
102
103
104
105
Table S1. Mineral soil properties of the three geochemical regions
106
Chemical composition of the top 30 cm of mineral soils layers representing the three geochemical
107
regions. Values presented are mean ± standard deviation (n=129). Mafic (Kahuzi-Bièga forest),
108
felsic (Kibale forest), and sedimentary (Nyungwe forest). Source: Project TropSOC Database
109
Version 1.0 (Doetterl et al., 2021a,c).
110
Variables
Units
Mafic
Felsic
sedimentary
Clay
42.7±11.1
33.0±5.5
32.5±13.9
Sand
42.4±16.2
55.6±5.2
48.4±17.4
Silt
13.2±3.2
11.4±2.7
20.2±9.6
pHKCL
3.93±0.6
5.3±0.6
3.4±0.4
Total Nitrogen
0.4±0.2
0.2±0.1
0.2±0.2
C:N
10.3±5.9
12.5±6.3
56.4±70.3
Bio-P
17.9±28.5
31.2±25.7
5.5±35.9
Bases in CEC
16.0±20.5
73.0±18.1
8.8±14.5
CEC
38.4±5.5
17.6±4.5
15.3±11.8
Bases in ECEC
48.2±23.2
84.3±17.8
8.2±15.2
ECEC
12.1±4.1
14.4±4.7
7.1±3.1
Exchangeable Mg2+
1.9±0.9
2.3±0.8
0.2±0.6
Exchangeable Ca2+
5.1±4.2
11.2±4.4
0.4±2.3
Exchangeable K+
0.3±0.2
0.5±0.1
0.1±0.1
Total Ca
1600±1200
3000±900
100±1000
Total K
2800±400
2600±300
400±400
Total Mg
900±900
1400±700
500±1200
Total P
1800±700
700±400
500±700
111
112
113
114
115
116
117
118
119
Table S2. General plot information
120
General plot information, unique identifier of each plot where data were collected, international
121
country code (Democratic Republic of the Congo = DRC; Uganda = UG; Rwanda = RW), latitude
122
and longitude in decimal degree (WGS 1984; Projection EPSG 4326), topographic positions,
123
dominant soil type based on the World Reference Base for Soil Resources, geochemistry of the
124
parent material, and sampling date.
125
PlotID
Country
Latitude
Longitude
Position
Soil type
Geochemistry
Sampling date
KBPL14
DRC
-2.31249
28.7526
plateau
Mollic Nitisols (Ochric)
mafic magmatic rock
2018-04-15
KBPL13
DRC
-2.31379
28.75295
upper slope
Mollic Nitisols (Ochric)
mafic magmatic rock
2018-04-15
KBPL6
DRC
-2.31398
28.7524
upper slope
Alic Nitisols (Ochric)
mafic magmatic rock
2018-04-15
KBPL15
DRC
-2.31159
28.75269
Middle slope
Mollic Nitisols (Ochric)
mafic magmatic rock
2018-04-15
KBPL16
DRC
-2.31153
28.75294
Middle slope
Mollic Nitisols (Ochric)
mafic magmatic rock
2018-04-15
KBPL10
DRC
-2.31439
28.75246
valley
Mollic Nitisols (Ochric)
mafic magmatic rock
2018-04-15
KBPL11
DRC
-2.32866
28.72988
valley
Mollic Nitisols (Vetic)
mafic magmatic rock
2018-04-15
KBPL12
DRC
-2.32768
28.72956
valley
Mollic Nitisols (Vetic)
mafic magmatic rock
2018-04-15
UPL1
UG
0.46225
30.37403
plateau
Haplic Lixisols (Nitic)
felsic metamorphic rock
2018-06-03
UPL2
UG
0.46245
30.37347
plateau
Haplic Lixisols (Nitic)
felsic metamorphic rock
2018-06-03
UPL3
UG
0.46271
30.37291
plateau
Haplic Lixisols (Nitic)
felsic metamorphic rock
2018-06-03
UPL4
UG
0.46078
30.37271
upper slope
Haplic Lixisols (Nitic)
felsic metamorphic rock
2018-06-03
UPL5
UG
0.46083
30.37356
upper slope
Sederalic Nitisols (Ochric)
felsic metamorphic rock
2018-06-03
UPL6
UG
0.46021
30.37396
upper slope
Sederalic Nitisols (Ochric)
felsic metamorphic rock
2018-06-03
UPL7
UG
0.48398
30.35252
middle slope
Haplic Lixisols (Nitic)
felsic metamorphic rock
2018-06-03
UPL8
UG
0.4838
30.35238
middle slope
Haplic Lixisols (Nitic)
felsic metamorphic rock
2018-06-03
UPL9
UG
0.48337
30.35179
middle slope
Haplic Lixisols (Nitic)
felsic metamorphic rock
2018-06-03
UPL10
UG
0.45994
30.37355
valley
Luvic Nitisols (Endogleyic)
felsic metamorphic rock
2018-06-03
UPL11
UG
0.46054
30.37317
valley
Luvic Nitisols (Endogleyic)
felsic metamorphic rock
2018-06-03
UPL12
UG
0.46028
30.37242
valley
Luvic Nitisols (Endogleyic)
felsic metamorphic rock
2018-06-03
NPL1
RW
-2.4645
29.10346
plateau
Alic Nitisols (Ochric)
sedimentary rock
2018-05-19
NPL2
RW
-2.46337
29.09542
plateau
Acric Ferralsols (Vetic)
sedimentary rock
2018-05-19
NPL3
RW
-2.46328
29.09489
plateau
Acric Ferralsols (Vetic)
sedimentary rock
2018-05-19
NPL4
RW
-2.4623
29.09644
upper slope
Acric Ferralsols (Vetic)
sedimentary rock
2018-05-19
NPL5
RW
-2.46254
29.09666
upper slope
Acric Ferralsols (Vetic)
sedimentary rock
2018-05-19
NPL6
RW
-2.46823
29.10455
upper slope
Acric Ferralsols (Vetic)
sedimentary rock
2018-05-19
NPL7
RW
-2.46401
29.10335
middle slope
Haplic Alisols (Nitic)
sedimentary rock
2018-05-19
NPL8
RW
-2.46319
29.10213
middle slope
Haplic Alisols (Nitic)
sedimentary rock
2018-05-19
NPL9
RW
-2.46381
29.09542
middle slope
Haplic Alisols (Nitic)
sedimentary rock
2018-05-19
NPL10
RW
-2.46391
29.10289
valley
Acric Ferralsols (Gleyic)
sedimentary rock
2018-05-19
NPL11
RW
-2.46366
29.1031
valley
Acric Ferralsols (Gleyic)
sedimentary rock
2018-05-19
NPL12
RW
-2.46321
29.10369
valley
Acric Ferralsols (Gleyic)
sedimentary rock
2018-05-19
126
127
128
Table S3. Forest stands characteristics across the three investigated geochemical regions
129
Forest stands characteristics across the three investigated geochemical regions. Species richness
130
and Shannon indices, BA: species weighted tree basal area, DBH: tree diameter at breast height,
131
average tree height, average wood density, the average number of trees per hectare, RGR: relative
132
tree growth rate, 󰄵wood: wood C turnover rate, 󰄵fineroot is the fine root C turnover rate MAP: mean
133
annual precipitation, MAT: mean annual temperature, slope inclination, slope length and altitude
134
in terms of elevation above sea level. Values presented are (Mean ± SD) and the range along
135
topographic positions. Letters in brackets for each stand characteristic indicate significant
136
differences following Kruskal-Wallis tests and pair-wise comparison using Dunn’s test (p-value <
137
0.05) performed to assess differences between geochemical regions. Kruskal-Wallis tests were
138
performed on each stand separately. Source: Project TropSOC Database Version 1.0 (Doetterl et
139
al., 2021a,c).
140
141
Mafic
Felsic
Sedimentary
Parameter
Unit
Mean±SD
Range
Mean±SD
Range
Mean±SD
Range
Species richness
-
10.1±1.9 (a)
6.6-13.3
12.7±2.0 (b)
9.5-15
10.2±1.2 (a)
8.3-12.1
Shannon
-
2.2±0.3 (ab)
1.6-2.6
2.3±0.3 (b)
1.8-2.6
2.0±0.2 (a)
1.6-2.3
BA
m2.ha-1
37.0±10.3 (a)
25-67
35.4±9.71 (a)
16-51
51.3±10.0 (b)
38-69
DBH
cm
25.0±2.6 (a)
10-232
27.8±3.7 (ab)
10-157
30.1±2.9 (b)
10-138
Tree height
m
13.1±2.3 (a)
10-31
16.4±2.1 (b)
5-30
16.7±2.2 (b)
5-32
Wood density
g.cm-3
0.5 ± 0.1(a)
0.2-0.8
0.6±0.1(a)
0.2-0.8
0.6±0.1(a)
0.2-0.8
Tree density
Number.ha-1
546.9±102.9 (b)
275-706
427.1±140.0 (a)
200-713
513.0±103.4 (ab)
331-700
RGR
%
7.0±2.0 (c)
2-9
5.0±1.0 (b)
3-6
2.0±0.5 (a)
1-2
󰄵wood
year -1
0.06±0.01
0.02-0.08
0.04±0.009
0.03-0.06
0.02±0.005
0.006-0.02
󰄵fineroot
year- 1
0.3 ± 0.2
0.1-0.7
0.3 ± 0.1
0.1-0.4
0.5 ± 0.1
0.3-0.6
MAP
mm
1924
1702
1697 
MAT
°C
15.3
14-17
19.2
19-22
16.7
16 -17
Slope
%
21±20
3-60
21±20
3-55
31±30
3-60
Slope length
m
70 ± 56
50-170 
149 ± 125
55-374
101 ± 103
55- 339
Altitude
(m) a.s.l
2220 ± 38
1919-2224
1324 ± 60 
12711424
1909 ± 22 
1891-2395
142
143
144
145
Table S4. Rotated principal components and their mechanistic interpretation
146
Rotated principal component analysis for four principal components (RC) that were retained with
147
Eigenvalues >1 and proportion variance >10 %. Upper part of the table shows eigenvalues,
148
proportional, cumulative variance and mechanistic interpretation of specific RCs. Bottom part
149
represents loadings with bold marked values showing highly correlated loadings (> 0.5 or < -0.5)
150
of each RC that was used as part of the interpretation of each variable. Blank cells indicate that
151
variables are not represented by the corresponding RCs and the loadings of those variables onto
152
the RC are near zero.
153
Rotated component
RC1
RC2
RC3
RC4
Eigenvalue
6.9
4.4
3.7
2.6
Proportion Variance (%)
33.0
21.0
17.7
12.4
Cumulative Variance (%)
33.0
54.0
71.0
84.1
Mechanistic interpretation
Soil exchangeable
cations
Soil base cation
stocks
Soil CNP stocks
& NP availability
Soil
texture
Independent variables
Units
Potential cations exchange capacity
0.01 me g-1
0.5
-0.3
0.7
-0.2
Effective cations exchange capacity
0.01 me g-1
0.9
0.2
0.2
-0.1
Exhangeable potassium
0.01 me g-1
0.6
0.5
0.3
0.1
Exhangeable calcium
0.01 me g-1
0.9
0.4
Exchangeable magnesium
0.01 me g-1
0.8
0.2
0.3
-0.3
Base saturation in CEC
%
0.9
0.4
0.1
-0.1
Available phosphorus
mg kg-1
0.3
0.4
0.5
0.4
Total potassium
%
0.4
0.9
Total calcium
%
0.5
0.8
0.1
Total magnesium
%
0.5
0.7
0.4
0.2
TRB
%
0.5
0.8
0.2
0.2
Total phosphorus
%
-0.1
0.2
0.8
-0.2
pH
0.8
0.5
-0.1
C:N
-0.8
-0.4
-0.2
SOC
Mg ha-1
0.2
0.1
0.9
Ammonium
mg kg-1
0.7
0.2
0.5
Nitrate
mg kg-1
-0.5
0.2
Total nitrogen
%
0.1
0.9
0.3
Clay content
%
0.4
-0.1
-0.8
Silt content
%
-0.6
-0.6
Sand content
%
1.0
154
155
Methods S1. Detailed descriptions to study sites, existing data, protocols and assessment
156
methods
157
Climate
158
The mean annual temperature (MAT) at the study sites varies between 15.3 and 19.2 °C, and mean
159
annual precipitation (MAP) varies between 1697 and 1924 mm (Fick & Hijmans, 2017). Note that
160
climate information is derived from large scale data products at coarse spatial resolution, since no
161
long-term observations are available at the local level. However, weather stations installed and
162
started at the beginning of the monitoring period in 2018 near the study sites showed a general
163
good relation with climatic indicators used in this study (data not shown). Throughout the year,
164
the study sites are characterized by dry and rainy seasons subdivided into four seasons (weak dry,
165
December-February; strong rain, March-May; strong dry, June-August; weak rain, September-
166
November) each covering three months (Doetterl et al., 2021b,a). Vegetation cover in the study
167
area is classified as closed-canopy evergreen montane forests and show typical dominant
168
vegetation of old growth forests and primary forests (Nyirambangutse et al., 2017; Imani et al.,
169
2017). All study sites belong to National Parks established between 20 and 50 years ago, free from
170
permanent settlements and larger road systems, protected from logging and larger anthropogenic
171
disturbance for more than five decades.
172
173
Soil parent material
174
Study sites in the DRC are located in the Kahuzi-Biéga National Park (-2.31439 ° S; 28.75246 °
175
E) where soils (alic Nitisols(ochric), alic Nitisols (vetic), and mollic Nitisols (ochric)) have
176
developed from nutrient rich, mafic magmatic rocks, a result of volcanism in the East African Rift
177
System (Schlüter, 2006), further referred to as mafic region. In Uganda, study sites are located in
178
the Kibale National Park (0.46225 ° N; 30.37403 ° E) where soils (sederalic Nitisols (ochric),
179
haplic Lixisols (nitic), and luvic Nitisols (endogleyic)) have developed from felsic magmatic and
180
metamorphic rocks of intermediate nutrient content, further called the felsic region. Study sites in
181
Rwanda are located in the Nyungwe National Park (-2.463088 ° S; 29.103834 ° E) where soils
182
(haplic Acrisols (nitic), acric Ferralsols (vetic), and acric Ferralsols (gleyic)) have developed from
183
a mixture of nutrient poor sedimentary rocks, further called the mixed sediment region. These
184
sediments are mostly dominated by quartz-rich sandstone and schist layers spanning along the
185
Congo-Nile divide (Schlüter, 2006). Soils in the three study regions were deeply weathered (> 2
186
meters) with homogeneous properties in subsoil horizons and no bedrock could be reached at any
187
slope position. Soils were described and analyzed extensively across the three geochemical
188
regions, separately for each topographic position and depth-explicit (Doetterl et al., 2021a).
189
190
Topography and plot installation
191
In each slope catena, four topographic positions were covered: flat plateaus (3 - 5 %), upper slope
192
(9 - 15 %), middle slope (45 - 60 %), and valley/foot slope (1 - 3 %) of slope steepness with an
193
average slope length (plateau to valley plots) of (70 ± 56 m) in mafic, (149 ± 125 m) in felsic and
194
(101 ± 103 m) in sedimentary region (Table S3). At each topographic position, three replicate plots
195
of 40 m x 40 m in size were established, resulting in a total of 36 plots. Each experimental plot
196
was further subdivided into four 20 m x 20 m subplots in order to structure and distribute the
197
replicate sampling of soil and root cores as well as the activities performed during forest
198
inventorization evenly across the plot. Note that in January 2019, due to security concerns at the
199
beginning of our monitoring activities in the mafic region (details see below), four plots (two
200
plateaus, one upper slope, and one middle slope) were abandoned and four additional plots re-
201
established in nearby area of the remaining original plots with similar vegetation, topographic, and
202
geochemical features. This resulted in a reduced set of eight plots in the mafic region and total of
203
32 plots, allowing for root and litter monitoring at all topographic positions. Furthermore, to assess
204
relationships between potential topographic features and NPP components and their relative C
205
allocation, we derived topographic indices relevant for distinguishing differences in the local
206
landscape dependent variation in water and nutrient availability from a 30m x 30m SRTM derived
207
void filled DEM (NASA SRTM, 2013). The calculated topographic indices included the
208
topographic position index (TPI), topographic wetness index (TWI), slope length, steepness factor
209
(SL-factor), slope inclination (slope), stream power index (SPI), terrain aspect (aspect) and
210
curvature.
211
212
Forest inventory and aboveground living biomass C stock
213
In 2018, full inventories of the forest tree species and aboveground standing coarse woody (further
214
simply called “wood”) biomass were conducted on all plots following Matthews et al. (2012).
215
First, we identified all living trees with a diameter at breast height (DBH; measured at 1.3 m
216
aboveground) of ≥10 cm in each plot. Second, we calculate two alpha diversity indices (species
217
richness and H-Shannon) (Morris et al., 2014) to get insight into aspects of species diversity across
218
the investigated geochemical regions. Third, to estimate the wood biomass, we constructed stand-
219
specific height-diameter (H:D) allometric relationships using a representative subset of plot-
220
specific trees using modelHD included in R package ‘BIOMASS’ (Réjou-Méchain et al., 2017).
221
For this, 20 % of all trees distributed across all DBH classes were selected for height measurement
222
per plot. Depending on DBH class abundance, the heights of three to five individual trees per class
223
were measured using a hypsometer (Forestry Pro II, Nikon, Japan). Wood biomass, i.e. stems and
224
large branches, for each individual tree was then estimated using the allometric equation for moist
225
tropical forests as described in Chave et al.(2014). For model parameterization, species-level
226
averages of wood density (WD) were taken from the DRYAD database (Zanne et al., 2009). Where
227
species-specific WD data were not available, we consecutively used genus- or family-level mean
228
WD values for the analysis. To calculate wood productivity, we carried out a re-census in 2020.
229
Note that inventorized dead trees in the first and second census were excluded as they do not
230
contribute to the annual NPP. We then calculated wood NPP, the relative stem growth rate, and
231
wood C turnover rate per plot using the following equations:
232
233
NPPwood=󰇡Σstem2 - Σstem1
Δt 󰇢* a-1 (eq. 1)
234
235
RGR= 󰇡ln󰇛stem2󰇜-ln󰇛stem1󰇜
Δt 󰇢*100 (eq. 2)
236
237
τwood=NPPwood
Wood C stock (eq. 3)
238
239
Where NPPwood is the wood net primary productivity of a plot (Mg C ha-1 year-1), RGR the relative
240
growth rate in (% year-1),  and  is the biomass of individual stems for the first and
241
second census, respectively, is the plot area in ha, is the time between the two censuses (2.0
242
- 2.4 years), and 󰄵wood is the wood C turnover rate in (% C year-1). To enable direct comparison
243
with other studies on biomass assessment (see (Saatchi et al., 2011; Kearsley et al., 2013)), we
244
assumed that all standing wood biomass holds a C content of 50 % of dry biomass but acknowledge
245
the uncertainty related to this since e.g. wood biomass C can vary between ~ 41 - 51 % of dry
246
biomass (Martin & Thomas, 2011). To calculate standing wood C stocks, we used data collected
247
during the final, second census in 2020 only. Wood C turnover rate was then calculated as a ratio
248
of NPPwood to wood C stock. In addition, we assessed CN content and ratio of living, healthy-
249
looking (without signs of herbivory), canopy leaves (sun-exposed shoots at outer canopy), sampled
250
during the weak dry season of December 2018 February 2019 following Pérez-Harguindeguy et
251
al. (2013). Sampled leaves originate from at least 3 individual trees per species representing ≥80 %
252
of the standing basal area per plot. Community-weighed means of CN of canopy leaves were
253
calculated using dried and homogenized samples at the plot level.
254
255
Litterfall productivity
256
For litterfall measurements, ten litter traps per plot were installed and distributed evenly across
257
each plot and the subplots therein following Matthews et al. (2012); for details of litter trap
258
placement see Doetterl et al. (2021a). Traps, made of locally available charcoal sacks, had a
259
diameter (d) of 60 cm and were installed at a height of 1.0 m. Litter samples were collected every
260
two weeks for the period August 2018 to February 2020. In consequence, the strong rain and dry
261
seasons were sampled once while the weak rain and the weak dry seasons were sampled twice,
262
requiring to calculate weighted averages for each season (see below). The collected litter included
263
all organic residues collected by the traps; woody debris (d >2 cm) and dead animals were
264
discarded. After each sampling, collected litter material was broken into small pieces, mixed and
265
homogenised per plot. Material from all 10 traps per plot was pooled to obtain a composite sample
266
per plot and taken to the laboratory the day of sampling, oven-dried (70 °C, 72 h), and subsequently
267
weighed. Where mixing with hands was not possible due to the large surface area of leaves, an
268
electric blender (1000 W; TYB-315) was used to homogenize the litter material. This resulted in
269
a total of, on average, 45 pooled data points of litterfall per plot distributed over the monitoring
270
period. We then subsampled 10 g of well mixed litter material per data point and milled it using a
271
PM 400 Planetary Ball Mill (Retsch, Germany) at 400 rounds per minute for CN analysis. Litter
272
C content was then measured on a 5 mg powdered litter subsample using dry total combustion
273
(Variomax CN, Elementar GmbH, Germany). Note that three replicates were measured on 20 %
274
of the samples to assess the laboratory analytical errorshowing a standard deviation of 5 %
275
(Doetterl et al., 2021c). To represent total litter production per plot, we first calculated the average
276
daily litter productivity in each plot for season (i) using equation 6 below. Litter productivity values
277
for the seasons covered twice during sampling were first averaged before an annual average of
278
litter productivity was calculated.
279
Litdw(i)=w
a*n (eq. 4)
280
Where Litdw(i) is the average daily litterfall per season (i) (Mg ha-1 day-1), w is the total dry weight
281
(Mg) of the sample per plot, is the area of the litter traps (ha), n is the number of days per
282
sampling interval. The annual litter productivity NPPlitterfall (Mg C ha-1 year-1) was calculated from
283
the averaged litter productivity of the considered four (equal length) seasons, as the sum of 365
284
days. Litter biomass C for each plot was calculated by multiplying the measured litter C content
285
with the corresponding biomass productivity.
286
287
Root biomass and fine root production
288
Fine (d ≤ 2 mm) and coarse (d > 2 mm) root biomass and fine root production were assessed from
289
September 2018 to December 2019 on all plots, following depth-explicit sampling and
290
standardized protocols (recently summarized by Freschet et al., 2021). Prior to deciding on
291
maximum sampling depth and depth intervals, root depth distribution was assessed to 1 m depth
292
using soil profiles established at the center of a plot. This assessment revealed that fine root counts
293
on the profile wall were most frequent in organic horizons and the upper 50 cm (data not shown),
294
with approximately 90-97% of the root biomass evenly placed within the O horizon and the top
295
30 cm of mineral soil (see Fig. S6 for fine root biomass distribution to 50 cm depth). Belowground
296
standing root biomass was thus sampled to a depth of 50 cm using a soil core sampler with 6.8 cm
297
inner diameter (Vienna Scientific Instruments, Austria). Sequential coring took place once per
298
season (every three months) where one soil core per subplot was sampled, resulting in a total of
299
four cores per plot per season. More frequent sampling campaigns similar to litterfall monitoring
300
were not feasible for logistic reasons. No cases of soil compaction as a result of the incremental
301
sampling were observed. Cores were subsequently divided into five distinct depth layers: the
302
organic O horizon, and four mineral soil layers from 0 10 cm, 10 20 cm, 20 30 cm, 30 50
303
cm. After transport to the laboratory, roots were separated into fine and coarse roots. To do this,
304
each sample was rinsed within a sieve of 2 mm-mesh size positioned on top of a 1 mm sieve. The
305
two sieves together were placed on top of a bucket to collect also the smallest roots fragments.
306
Note that Coarse roots were not considered for further analyses as i) their spatial distribution is
307
considered insufficiently covered by the sequential coring approach (Ostonen et al., 2005; Yuan
308
& Chen, 2013), and ii) the contribution of relatively slow coarse root growth and turnover rates to
309
seasonal biomass productivity (using the DM method, see below) is considered minimal
310
(McCormack et al., 2015; Huang et al., 2020). This can lead to an underestimation of the total
311
NPP. However, fine roots' contribution to the terrestrial NPP range is 22%-40% (Cordeiro et al.,
312
2020) with turnover being much faster than for coarse roots, making them an important but often
313
unmeasured component of the carbon budget of forest ecosystems (Jackson et al., 1997; Ostonen
314
et al., 2005).Therefore, as a dynamic carbon pool (fast growth and high turnover rate), fine roots
315
are often seen as a major C input and contributor to SOC stocks (Lukac, 2012). Hence, for roots,
316
we reported the fine root biomass only.
317
Thus, fine roots were separated into living and dead fractions based on criteria such as color, root
318
elasticity, and the degree of cohesion of cortex, and steleroots were i.a. considered living when
319
root steles were bright and resilient (Ostonen et al., 2005; Freschet et al., 2021). The dry mass of
320
fine roots per core, horizon/layer and living/dead fraction 0.01 g dwt) was determined after
321
drying (70 °C, 72 h); standing fine root biomass C was calculated (Mg C ha-1), assuming a C
322
content of 50 % of dry biomass (as above). Fine root productivity was calculated using the
323
improved Decision Matrix (DM) method (Yuan and Chen, 2013) as a reliable method to determine
324
fine root production when rapid turnover can be assumed (Fairley & Alexander, 1985; Assefa et
325
al., 2017; Freschet et al., 2021). Briefly, to determine the fine root NPP, this method includes both
326
(significant) changes in living and dead standing root mass between two sampling dates throughout
327
the monitoring period. We aggregated the collected fine root data per season for each layer at the
328
plot level. In brief, fine root biomass production (Proot) in (g) for 90 days for each sampling point
329
and horizon/layer was then calculated taking into account variation in living () and dead 󰇛)
330
fine root mass between two consecutive sampling dates where Proot is equal to if > 0 and
331
. Proot is equal to zero if both and < 0; see Yuan & Chen (2013) for details. In order
332
to give all seasons the same weight for the year 2018 and 2019, we first calculated daily averages
333
of root biomass per unit area for each season using the following equation:
334
rootseason(i)=Proot
a*n (eq. 5)
335
336
Where rootseason(i) is the daily dry weight fine root biomass productivity per season (i) (Mg ha-1day-
337
1), Proot is the root productivity (Mg) between two sampling dates per layer and plot, is the area
338
of a soil core for each layer (ha), is the number of days each sampling interval represented. Note
339
that the DM method may underestimate fine root NPP owing to fine root losses through herbivory,
340
secondary growth of fine roots, or decomposition of dead roots faster than the sampling interval
341
(Lowatschek, 2021).
342
Similar to litter productivity, before calculating the annual root productivity, the four seasons were
343
weighed equally. That is root productivity values for the seasons covered twice were first averaged
344
to daily means before an annual average for root productivity was calculated. Similar to NPPLitterfall,
345
annual dry weight fine root productivity NPProots (Mg C ha-1 year-1) was calculated as the sum of
346
the (averaged) seasonal root productivity. Fine root biomass C for each plot was calculated by
347
assuming C content to be 50% of dry biomass (Lewis et al., 2009; Zhu et al., 2017). Fine root
348
turnover rate was calculated following Gill and Jackson (2000) and Brunner et al. (2013) ( eq. 6)
349
350
τfineroot=NPProot
fine C stock (eq. 6)
351
352
Assessing biomass C allocation
353
NPPsum in our study was calculated as the sum of components (wood growth, litterfall and fine
354
root production) for each forest plot using the following equation:
355
356
NPPsum = NPPwood + NPPlitterfall + NPProots (eq. 7)
357
358
where NPPsum is the total NPP (Mg C ha-1 year-1). Note that the NPPsum estimated in this study may
359
be biased towards underestimation because it omits several NPP terms such as volatile organic
360
emissions, and C allocation to root exudates and mycorrhizal symbionts for methodological
361
reasons (Malhi et al., 2017). We fully acknowledge that in particular the amount of C allocation
362
to tree root exudates and mycorrhizal symbionts can be substantial and strongly related to nutrient
363
availability (Treseder, 2004; Hobbie, 2006; Aoki et al., 2012; Doughty et al., 2018)--creating some
364
uncertainty particular in stands with medium- or low-nutrient availability (Buendia et al., 2014).
365
However, our calculation of NPP is based on three independent measurements (wood growth,
366
litterfall, fine root production) covering major components of NPP (Vieira et al., 2011; Doughty
367
et al., 2018) allowing for an essential characterization of plant investment into root:shoot and
368
root:leaf allocation.
369
370
C stocks of organic and mineral soil layers
371
As part of an extensive sampling campaign (Doetterl et al., 2021a), organic soil litter layers (L
372
horizon and O horizon) were sampled at eight points along the border distributed across all
373
subplots and in the center of each forest plot at the time when soil sampling took place. At each
374
sampling point, the thickness of the L and O horizons were measured with a ruler and then sampled
375
within a 5 cm x 5 cm square. When the litter layers were thin (<0.5 cm), the sampling square was
376
expanded to 10 cm x 10 cm to retrieve sufficient sample material. The nine samples of each layer
377
were combined into one composite sample per plot. In the laboratory, samples were oven-dried
378
(70 °C, 48-96 h) and weighed. To sample mineral soils, four one-meter soil cores were sampled
379
per plot (one core per subplot) using a cylindrical soil core sampler for undisturbed sampling. For
380
the purpose of this study, relating to the depth of sequential root coring (see below), we used soil
381
C data of the top 50 cm. Cores were separated into 10 cm increments and combined into depth-
382
explicit composite samples per plot. For the organic litter layers (L and O horizons), C stocks
383
were determined as the product of the litter mass per area and the litter C content. For mineral soil,
384
total C was measured to 50 cm soil depth on 1 g of grinded soil samples using dry total combustion
385
(Variomax CN, Elementar GmbH, Germany) with the laboratory analytical error assessed in the
386
same way as for litterfall (see above and Doetterl et al.(2021c)). Since no inorganic C was found
387
in any of the investigated soils (Doetterl et al., 2021b,a), total C was interpreted as SOC. C stock
388
of the bulk soil of each layer was then calculated by multiplying C content with soil bulk density
389
and the depth increment of the horizon/layer. Aligned with root sampling, the top five mineral
390
layers were summed to have SOC stocks to 50 cm mineral soil depth (Mg C ha-1).
391
To assess relationships of potential soil controls on NPP components and their relative C
392
allocation, we extracted a wide range of plot-specific geochemical properties for the top 30 cm
393
from an existing database assembled in parallel to this study (TropSOC database v1.0, Doetterl et
394
al., 2021c). We used soil properties for the top 30 cm of mineral soil only because the assessment
395
of root biomass revealed that the large majority of fine roots (approximately 90 %, relative to 50
396
cm) were found in organic litter layers and in those top mineral layers for all three investigated
397
geochemical regions (Fig. S6). Introducing properties of deeper subsoil (> 30cm) would hence
398
introduce unnecessary bias towards deeper layers with little to no roots. Included soil variables
399
covered a wide range of predictors such as soil fertility, SOC properties (SOC stocks, C:N ratio)
400
and soil texture (clay, silt and sand content).
401
402
Statistical analysis
403
To assess species composition as well as similarities and dissimilarity of species between the three
404
geochemically regions, we performed a nonmetric multidimensional scaling (NMDS) on the
405
inventory data of the plots, using the ‘vegan’ package (Oksanen et al., 2013) and following a rank-
406
based interpretation of the data. We performed NMDS analysis using the “Bray-Curtis” measure
407
of distance, and number of axes (K=3) and reported the stress value which is a measure of goodness
408
of fit. We also computed the coefficient of determination between the ordination distance and the
409
observed dissimilarity in the original data. Since NMDS does not use the absolute abundance of
410
species but rank orders. In our case rather than mafic being X units distant from the felsic region
411
and Y units distant from sedimentary, the NMDS ranks plot/region species similarities as follows:
412
The sedimentary region is the "first" most distant from mafic while the felsic region is the "second"
413
most distant. As such, it is a robust method for multidimensional analysis of tree diversity. Another
414
advantage of NMDS is that it does not rely on normally distributed data. This is important in our
415
case as there are only few abundant species across sites, but many species with site specific
416
appearance, which could result in skewed data (Legendre et al., 2005).
417
To assess differences in the distribution of C stocks in soil and aboveground biomass, as well as
418
the differences between NPP components (litterfall, wood, fine roots) and the relative C allocation,
419
collected data were analyzed for differences across topographic positions and geochemical regions
420
with data presented as means per plot standard deviation, SD). Before the analysis, we
421
conducted a residual analysis to test for the assumptions of ANOVA. For this, we first used
422
Shapiro-Wilk’s test of normality distribution and Levene's test for homogeneity of variances. The
423
tests showed that requirements of normality distribution and homogeneity were not met. Hence,
424
we continued using the non-parametric Kruskal-Wallis test as an alternative to one-way ANOVA
425
using the R package ‘pgirmess’ (Giraudoux, 2021). For a post-hoc pairwise comparison of
426
significant differences in C stock, NPP and C allocation between distinct groups, in our case
427
specific topographic positions or geochemical regions, we used Dunn’s test to identify which
428
group levels are different using the ‘rstatix’ package (Kassambara, 2021).
429
As multicollinearity and autocorrelation between independent variables was to be expected due to
430
the large number of variables and a relatively small number of aggregated observations, we
431
conducted rotated principal component analysis (rPCA) for dimension reduction (Jolliffe, 1995),
432
before regression analysis. Before conducting the rPCA, due to differences in scales and units
433
among the soil geochemical properties, we applied a transformation method on the input data using
434
the standardization technique to have a mean value of zero and standard deviation of 1 for each
435
variable. All retained rotated components (RCs) were then interpreted based on the loadings of the
436
original variables and using expert knowledge for the likely underlying mechanisms that can affect
437
C dynamics (Table S4). We used a threshold of r > 0.5 & < -0.5 to decide whether an independent
438
variable that is loaded into an RC is used for the mechanistic interpretation of the RC or not. We
439
used an eigenvalue >1 and explained the proportion of variance >10% for each RC as criteria to
440
include or exclude an RC into our models (Jolliffe, 1995; James et al., 2013).
441
To assess the effects size and direction (positive or negative) of rotated principal components on
442
different NPP compartments and their relative C allocations, we applied Bayesian multilevel linear
443
mixed effect models with intercepts set to zero to allow comparison of the effects size of rotated
444
principal components between models for the different NPP components. In the models, we set
445
the retained RCs as fixed effects and plotIDs as random effects using the ‘brms’ R package
446
(Bürkner, 2017). Compared to traditional statistical models, Bayesian approaches have the
447
advantage of taking all sources of variance into account simultaneously while still allowing for the
448
implementation and assessment of random and fixed factors (random and fixed) into a single
449
model. To assess model uncertainty, we used Monte Carlo Markov Chain algorithms (MCMC) set
450
up for a total of 3000 iterations, using 4 separate chains with 1000 warmup iterations. Confidence
451
intervals were subsequently extracted from the posterior parameter distributions, along with the
452
mean effect size, marginal R squared (indicating the proportion of total variance explained by fixed
453
effects) and conditional R squared (indicating the proportion of total variance explained by both
454
fixed and random effects) (Nakagawa et al., 2017), the root mean square error (RMSE) and the
455
ratio of performance to deviation (RPD). We interpreted a model as performing strong if RPD >
456
1.5 along the guidelines given by Chang et al.(2001). For assessing the relationship between
457
response variables and independent predictors, we used (pairwise) Pearson correlation coefficients
458
and least square regression analysis. Due to the relatively small number of replicates and therefore
459
limited statistical power, for all the statistical tests we designated p-values <0.05 as significant and
460
p-values < 0.1 as marginally significant / tendent. All statistical analyses were carried out with R
461
software (R Core Team, 2022).
462
463
Notes S1.
464
Forest structure and species composition
465
The three regions differ significantly in terms of dominant tree species composition (Fig. S2). In
466
the mafic region, the dominant species are Dombeya mukole and Alangium chinense. In the felsic
467
region, Uvariopsis congensis and Chrysophyllum gorungosanum and in the sedimentary region
468
Cleistanthus polystachyus and Syzygium guineense. A complete species list can be found in
469
Doetterl et al. (2021c). NMDS analyses yielded a stress value of 0.04, which indicates an only
470
small error between the actual dissimilarity distances and the ordination distances calculated by
471
NMDS and a good representation of dissimilarity of species composition between plots and
472
regions. Furthermore, NMDS results suggest that all forests are characterized by varying structure
473
and species composition, which is specific for each geochemical region, but similar within each
474
region (Fig. S2). Note that an analysis of the Shannon and species richness indices (Table S3) did
475
not indicate strong differences between geochemical regions. Forest stands in mafic and
476
sedimentary regions are characterized by a high number of trees per unit area compared to forests
477
in the felsic region (Table S3). Forests in the mafic region are characterized by a higher density of
478
small trees of (10 - 20 & 20 - 30 cm diameter at breast height (DBH)) while plots in the sedimentary
479
regions hold a higher density of larger diameter trees (DBH >40 cm; Fig. S1). Average tree height
480
and distribution among tree DBH classes are similar in the felsic region compared to the
481
sedimentary region but lower in the mafic region (Table S3, Fig. S1). Collected living canopy
482
leaves differ between regions with significantly wider C:N ratios in canopy leaves of the
483
sedimentary region compared to those in mafic and felsic regions (Fig. S5). Lastly, based on
484
observations during the field campaign, we noted that understory vegetation in the mafic and
485
sedimentary regions was usually higher/more dense than in the felsic region, which was almost
486
free of any understory vegetation (data not shown).
487
488
489
490
491
492
493
494
References
495
Aoki M, Fujii K, Kitayama K. 2012. Environmental Control of Root Exudation of Low-
496
Molecular Weight Organic Acids in Tropical Rainforests. Ecosystems 15: 11941203.
497
Assefa D, Rewald B, Sandén H, Rosinger C, Abiyu A, Yitaferu B, Godbold DL. 2017.
498
Deforestation and land use strongly effect soil organic carbon and nitrogen stock in Northwest
499
Ethiopia. CATENA 153: 8999.
500
Besnard S, Koirala S, Santoro M, Weber U, Nelson J, Gütter J, Herault B, Kassi J,
501
N’Guessan A, Neigh C, et al. 2021. Mapping global forest age from forest inventories, biomass
502
and climate data. Earth System Science Data 13: 48814896.
503
Brunner I, Bakker MR, Björk RG, Hirano Y, Lukac M, Aranda X, Børja I, Eldhuset TD,
504
Helmisaari HS, Jourdan C, et al. 2013. Fine-root turnover rates of European forests revisited: an
505
analysis of data from sequential coring and ingrowth cores. Plant and Soil 362: 357372.
506
Buendia C, Arens S, Hickler T, Higgins SI, Porada P, Kleidon A. 2014. On the potential
507
vegetation feedbacks that enhance phosphorus availability &ndash; insights from a process-based
508
model linking geological and ecological timescales. Biogeosciences 11: 36613683.
509
Bürkner P-C. 2017. brms: An R Package for Bayesian Multilevel Models Using Stan. Journal of
510
Statistical Software 80: 128.
511
Chang C-W, Laird DA, Mausbach MJ, Hurburgh CR. 2001. Near-Infrared Reflectance
512
SpectroscopyPrincipal Components Regression Analyses of Soil Properties. Soil Science Society
513
of America Journal 65: 480490.
514
Chave J, Réjou-Méchain M, Búrquez A, Chidumayo E, Colgan MS, Delitti WBC, Duque A,
515
Eid T, Fearnside PM, Goodman RC, et al. 2014. Improved allometric models to estimate the
516
aboveground biomass of tropical trees. Global Change Biology 20: 31773190.
517
Cordeiro AL, Norby RJ, Andersen KM, Valverde-barrantes O, Fuchslueger L, Oblitas E,
518
Hartley IP, Iversen CM, Gonçalves NB, Takeshi B, et al. 2020. Fine-root dynamics vary with
519
soil depth and precipitation in a low-nutrient tropical forest in the Central Amazonia. Plant-
520
Environment Interactions 1: 316.
521
Doetterl S, Asifiwe R, Baert G, Bamba F, Bauters M, Boeckx P, Bukombe B, Cadisch G,
522
Cooper M, Cizungu L, et al. 2021a. Organic matter cycling along geochemical, geomorphic and
523
disturbance gradients in forests and cropland of the African Tropics - Project TropSOC Database
524
Version 1.0. Earth Syst. Sci. Data Discuss. 2021: 146.
525
Doetterl S, Bauters M, Berhe AA, Chivenge P, Finke P. 2021b. Preface : Tropical
526
biogeochemistry of soils in the Congo Basin and the African Great Lakes region-Special issue in
527
Doetterl S, Bauters M, Berhe A, Chivenge P, Finke P (eds). SOIL 7:1-7.
528
Doetterl S, Bukombe B, Cooper M, Kidinda L, Muhindo D, Reichenbach M, Stegmann A,
529
Summerauer L, Wilken F, Fiener P. 2021c. TropSOC Database. V. 1.0. GFZ Data Services
530
Doughty CE, Goldsmith GR, Raab N, Girardin CAJ, Farfan-Amezquita F, Huaraca-Huasco
531
W, Silva-Espejo JE, Araujo-Murakami A, da Costa ACL, Rocha W, et al. 2018. What controls
532
variation in carbon use efficiency among Amazonian tropical forests? Biotropica 50: 1625.
533
Fairley RI, Alexander IJ. 1985. Methods of calculating fine root production in forests. In: Fitter
534
AH, (eds). Ecological interactions in soil. Special Publication Number 4 of the British Ecological
535
Society. 3747. York, England.
536
Fick SE, Hijmans RJ. 2017. WorldClim 2: new 1-km spatial resolution climate surfaces for global
537
land areas. International Journal of Climatology 37: 43024315.
538
Freschet GT, Pagès L, Iversen CM, Comas LH, Rewald B, Roumet C, Klimešová J,
539
Zadworny M, Poorter H, Postma JA, et al. 2021. A starting guide to root ecology: strengthening
540
ecological concepts and standardising root classification, sampling, processing and trait
541
measurements. New Phytologist 232: 9731122.
542
GILL RA, JACKSON RB. 2000. Global patterns of root turnover for terrestrial ecosystems. New
543
Phytologist 147: 1331.
544
Hobbie EA. 2006. Carbon allocation to ectomycorrhizal fungi correlates with belowground
545
allocation in culture studies. ecology 87: 563569.
546
Huang C, Wu C, Gong H, You G, Sha L, Lu H. 2020. Decomposition of roots of different
547
diameters in response to different drought periods in a subtropical evergreen broad-leaf forest in
548
Ailao Mountain. Global Ecology and Conservation 24: 01236.
549
Imani G, Boyemba F, Lewis S, Nabahungu NL, Calders K, Zapfack L, Riera B, Balegamire
550
C, Cuni-Sanchez A. 2017. Height-diameter allometry and above ground biomass in tropical
551
montane forests: Insights from the Albertine Rift in Africa. PLOS ONE 12: 0179653.
552
Jackson, R B, Mooney H A, Schulze, E D. 1997. A global budget for fine root biomass, surface
553
area, and nutrient contents. Proceedings of the National Academy of Sciences 94: 73627366.
554
James G, Witten D, Hastie T, Tibishirani R. 2013. An Introduction to Statistical Learning with
555
Applications in R. New York, USA: Springer New York.
556
Jolliffe IT. 1995. Rotation of principal components : choice of normalization constraints Rotation
557
of principal components : choice of normalization constraints. Journal of Applied statistics 22: 29
558
35.
559
Kearsley E, De Haulleville T, Hufkens K, Kidimbu A, Toirambe B, Baert G, Huygens D,
560
Kebede Y, Defourny P, Bogaert J, et al. 2013. Conventional tree height-diameter relationships
561
significantly overestimate aboveground carbon stocks in the Central Congo Basin. Nature
562
Communications 4: 8.
563
Legendre P, Borcard D, Peres-Neto PR. 2005. Analyzing beta diversity: partitioning the spatial
564
variation of community composition data. Ecological Monographs 75: 435450.
565
Lewis SL, Lopez-Gonzalez G, Sonké B, Affum-Baffoe K, Baker TR, Ojo LO, Phillips OL,
566
Reitsma JM, White L, Comiskey JA, et al. 2009. Increasing carbon storage in intact African
567
tropical forests. Nature 457: 10031006.
568
Lowatschek, P. 2021. Decay rates of fine roots in tropical forest soils and temperature dependency
569
of the tea bag decomposition method. Master thesis, BOKU-Universität für Bodenkultur, Vienna,
570
Austria.
571
Lukac M. 2012. Fine Root Turnover BT - Measuring Roots: An Updated Approach. In: Mancuso
572
S. (eds). Measuring Roots: Springer, Berlin, Germany, 363373.
573
Malhi Y, Girardin CAJ, Goldsmith GR, Doughty CE, Salinas N, Metcalfe DB, Huaraca
574
Huasco W, Silva-Espejo JE, del Aguilla-Pasquell J, Farfán Amézquita F, et al. 2017. The
575
variation of productivity and its allocation along a tropical elevation gradient: a whole carbon
576
budget perspective. New Phytologist 214: 10191032.
577
Martin AR, Thomas SC. 2011. A Reassessment of Carbon Content in Tropical Trees. PLOS ONE
578
6: 23533.
579
Matthews TR, Metcalfe D, Malhi Y, Phillips O, Huasco HW, Riutta T, Ruiz Jaén M,
580
Girardin C, Urrutia R, Butt N, Cain R, Menor O, and colleagues from the RAINFOR and
581
GEM networks 2012: Measuring tropical forest carbon allocation and cycling: a RAINFOR-GEM
582
field manual for intensive census plots (v2.2), Manual, Global Ecosystems Monitoring network,
583
104 pp., available at: http://gem.tropicalforests.ox.ac.uk/ (last access: 1 October 2020), Oxford,
584
England.
585
586
587
McCormack ML, Dickie IA, Eissenstat DM, Fahey TJ, Fernandez CW, Guo D, Helmisaari
588
H-S, Hobbie EA, Iversen CM, Jackson RB, et al. 2015. Redefining fine roots improves
589
understanding of below-ground contributions to terrestrial biosphere processes. New Phytologist
590
207: 505518.
591
Morris EK, Caruso T, Buscot F, Fischer M, Hancock C, Maier TS, Meiners T, Müller C,
592
Obermaier E, Prati D, et al. 2014. Choosing and using diversity indices: insights for ecological
593
applications from the German Biodiversity Exploratories. Ecology and Evolution 4: 35143524.
594
Nakagawa S, Johnson PCD, Schielzeth H. 2017. The coefficient of determination R2 and intra-
595
class correlation coefficient from generalized linear mixed-effects models revisited and expanded.
596
Journal of The Royal Society Interface 14: 20170213.
597
NASA Shuttle Radar Topography Mission (2013). Shuttle Radar Topography Mission (SRTM)
598
Global. Distributed by OpenTopography. California, USA. https://doi.org/10.5069/G9445JDF.
599
Accessed: 2021-02-18
600
Nyirambangutse B, Zibera E, Uwizeye FK, Nsabimana D, Bizuru E, Pleijel H, Uddling J,
601
Wallin G. 2017. Carbon stocks and dynamics at different successional stages in an Afromontane
602
tropical forest. Biogeosciences 14: 12851303.
603
Oksanen J, Blanchet FG, Kindt R, Legendre P, Minchin PR, O’Hara RP, et al. 2013. Package
604
“vegan.” Community Ecology Package, Version 2. Available online at: https://CRAN.R-
605
project.org/package=vegan (Accessed 2021-10-01).
606
Ostonen I, Lõhmus K, Pajuste K. 2005. Fine root biomass, production and its proportion of NPP
607
in a fertile middle-aged Norway spruce forest: Comparison of soil core and ingrowth core methods.
608
Forest Ecology and Management 212: 264277.
609
Pérez-Harguindeguy N, Díaz S, Garnier E, Lavorel S, Poorter H, Jaureguiberry P, Bret-
610
Harte MS, Cornwell WK, Craine JM, Gurvich DE, et al. 2013. New handbook for standardised
611
measurement of plant functional traits worldwide. Australian Journal of Botany 61: 167234.
612
R Core Team 2022. R 4.2.0: A language and environment for statistical computing. R Foundation
613
for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.
614
jou-Méchain M, Tanguy A, Piponiot C, Chave J, Hérault B. 2017. Biomass: an R Package
615
for Estimating Above-Ground Biomass and Its Uncertainty in Tropical Forests. Methods in
616
Ecology and Evolution 8: 11631167.
617
Saatchi SS, Harris NL, Brown S, Lefsky M, Mitchard ETA, Salas W, Zutta BR, Buermann
618
W, Lewis SL, Hagen S, et al. 2011. Benchmark map of forest carbon stocks in tropical regions
619
across three continents. Proceedings of the National Academy of Sciences of the United States of
620
America 108: 98999904.
621
Schlüter T. 2006. Geological Atlas of Africa: with Notes on Stratigraphy, Tectonics, Economic
622
Geology, Geohazard and Geosites of Each Country. Berlin,: Springer.
623
Treseder KK. 2004. A meta-analysis of mycorrhizal responses to nitrogen, phosphorus, and
624
atmospheric CO2 in field studies. New Phytologist 164: 347355.
625
Vieira SA, Alves LF, Duarte-Neto PJ, Martins SC, Veiga LG, Scaranello MA, Picollo MC,
626
Camargo PB, do Carmo JB, Neto ES, et al. 2011. Stocks of carbon and nitrogen and partitioning
627
between above- and belowground pools in the Brazilian coastal Atlantic Forest elevation range.
628
Ecology and evolution 1: 421434.
629
Yuan ZY, Chen HYH. 2013. Simplifying the decision matrix for estimating fine root production
630
by the sequential soil coring approach. Acta Oecologica 48: 5461.
631
Zanne AE, Coomes D, Jansen S, Lewis SL, Swenson NG, Chave J. 2009. Towards a worldwide
632
wood economics spectrum. Ecology Letters 12: 351366.
633
Zhu J, Hu H, Tao S, Chi X, Li P, Jiang L, Ji C, Zhu J, Tang Z, Pan Y, et al. 2017. Carbon
634
stocks and changes of dead organic matter in China’s forests. Nature Communications 8: 151.
635
636
637
638

File (1)

Content uploaded by Boris Rewald
Author content
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.