1
Supplementary Materials for:
2
Calibrating the impact of root orientation on root quantification
3
using ground-penetrating radar
4
Li Guo, Yuan Wu, Jin Chen, Yasuhiro Hirano, Toko Tanikawa, Wentao Li, Xihong Cui
5
6
L. Guo, J. Chen ()
7
College of Global Change and Earth System Science, Beijing Normal University, Beijing, 100875, China;
8
Joint Center for Global Change Studies, Beijing, 100875, China
9
E-mail: chenjin@bnu.edu.cn
10
11
Y. Wu, W. Li, X. Cui
12
State Key Laboratory of Earth Surface Processes and Resource Ecology, Beijing Normal University,
13
Beijing, 100875, China
14
Y. Hirano
15
Graduate School of Environment Studies, Nagoya University, Nagoya 464-8601, Japan
16
T. Tanikawa
17
Kansai Research Center, Forestry and Forest Products Research Institute, Nagai-kyutaro, Momoyama,
18
Fushimi, Kyoto 612-0855, Japan
19
Text S1 Forward simulation of root GPR signal
20
Until now, root GPR signals could only be simulated in two dimension (Guo et al. 2013a). The cross
21
angle of root to survey line could not be parameterized in the simulation. In this study, GPR signals of
22
46 roots were simulated as though roots and survey lines were orthogonal. Diameters of the simulated
23
roots were set from 5 to 50 mm (in increments of 1 mm, which covers the majority of root sizes in root
24
GPR studies). To compare with the field experiment, water contents of root and soil were assigned based
25
on field measurements. According to our extensive field investigation, the gravimetric water content of
26
C. microphylla roots was normally distributed, with an average of 109% and a standard deviation of 27%
27
(Guo et al. 2013b). Hence, the water content of the simulated root was randomly assigned following such
28
a normal distribution. Field soil water content measurements at four depth intervals (0-0.2 m, 5.39%;
29
0.2-0.4 m, 5.06%; 0.4-0.6 m, 4.72%; and 0.6-0.8 m, 4.39%) were input into the simulation accordingly.
30
Simulations were computed under the center frequency of 900 MHz, which is the same as the GPR
31
systems used in both the field and sand box experiments. The geometric domain of each simulation
32
included a soil background (1.2 m long and 0.8 m deep) and a 5 mm thick air layer above the soil. The
33
simulated root was located at 0.3 m depth and 0.6 m to the right boundary of the simulation domain (see
34
an example of the simulation configuration in Fig. 1a in Guo et al. 2013b). Specific rationales of root
35
GPR signal simulation were stated in Guo et al. (2013a).
36
37
Text S2 Generating amplitude areas for simulated roots at varied cross angles
38
For the simulation experiment, amplitude areas were first extracted on the simulated radargrams for
39
orthogonal roots and survey lines, i.e., A(90°). Then, the amplitude areas corresponding to the other cross
40
angles were calculated based on the sinusoidally changing pattern of A against x following the method
41
below: 1) For each type of amplitude area, the minimum A(90°) extracted from the 46 simulated roots
42
was determined. In our simulation design, the minimum A(90°) was measured corresponding to the third
43
smallest simulated root (with a diameter of 7 mm and a low gravimetric root water content of 87%). This
44
minimum A(90°) was assumed to be A(0°) for all of the simulated data (i.e., c-a). Together with A(90°)
45
(i.e., c+a) directly extracted from each simulated radargram, the values of (c-a) and (c+a) were calculated.
46
In this case, the two coefficients (a and c) could be calculated for each simulated root, and the numerical
47
relationship between A and x was fixed. Then, A(x) was computed using Eq. (2) with the calculated
48
coefficients (a and c) at 7 cross angles from 45° to 135° (in 15° steps). Finally, to approximate the truth
49
as closely as possible, a relative error (i.e., the blue lines in Fig. 3) was added to each calculated A(x) and
50
the directly extracted A(90°). The scale of the relative error was estimated using data from the sand box
51
experiment. Hence, 644 amplitude areas (46 roots × 7 cross angles × 2 amplitude area types) were
52
obtained for the simulation experiment.
53
54
References
55
Guo L, Lin H, Fan B, Cui X, Chen J (2013a) Forward simulation of root's ground penetrating radar signal:
56
simulator development and validation. Plant Soil 372:487-505. doi:10.1007/s11104-013-1710-4
57
Guo L, Lin H, Fan B, Cui X and Chen J (2013b) Impact of root water content on root biomass estimation
58
using ground penetrating radar: evidence from forward simulations and field controlled experiments.
59
Plant Soil 371:503-520. doi:10.1007/s11104-013-1710-4